Superalloys 2020: Proceedings of the 14th International Symposium on Superalloys [1st ed.] 9783030518332, 9783030518349

Citation preview

The Minerals, Metals & Materials Series

Sammy Tin
Mark Hardy
Justin Clews
Jonathan Cormier
Qiang Feng
John Marcin
Chris O’Brien
Akane Suzuki Editors

Superalloys 2020 Proceedings of the 14th International Symposium on Superalloys

123

Editors Sammy Tin Illinois Institute of Technology Chicago, IL, USA

Mark Hardy Rolls-Royce Derby, UK

Justin Clews PCC Structurals Portland, OR, USA

Jonathan Cormier ISAE-ENSMA Chasseneuil-du-Poitou, France

Qiang Feng University of Science and Technology Beijing, China

John Marcin Collins Aerospace Windsor Locks, CT, USA

Chris O’Brien ATI Specialty Materials Monroe, NC, USA

Akane Suzuki GE Global Research Niskayuna, NY, USA

ISSN 2367-1181 ISSN 2367-1696 (electronic) The Minerals, Metals & Materials Series ISBN 978-3-030-51833-2 ISBN 978-3-030-51834-9 (eBook) https://doi.org/10.1007/978-3-030-51834-9 © The Minerals, Metals & Materials Society 2020 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

This volume is a collection of papers from the 14th International Symposium on Superalloys, to be held on September 12–16, 2021, at the Seven Springs Mountain Resort, Seven Springs, Pennsylvania, USA. The proceedings is sponsored by the International Seven Springs Symposium Subcommittee of the TMS International Affairs Committee, the TMS Structural Materials Division (SMD), and the TMS High Temperature Alloys Committee, and co-sponsored by the Institute of Materials, Minerals and Mining (IOM3).

Preface

The purpose of the International Symposium on Superalloys, which takes place every four years, is to provide a forum for researchers, producers, and users to exchange recent technical information regarding the high-temperature, high-performance materials that are used in gas turbine engines and related products. The principal goal of the symposium is to highlight new initiatives and future growth opportunities for superalloys, recent advances in the understanding of their behavior, and progress in integrating them into new systems. The first symposium, held in 1968, emphasized phase instabilities in superalloys. Since then, the scope of the symposium has expanded considerably to cover all aspects of research, development, manufacture, and application of these materials. Over the years, the symposium has developed rich traditions, encompassing a high-quality peer-reviewed publication, which is presented before the conference, single-session presentations, and lively discussions during and after formal sessions, which are facilitated by the Seven Springs Mountain Resort. Participation from the international superalloy community in the technical program has always been key to the success and reputation of this symposia, so due to the unprecedented COVID-19 pandemic, the Superalloys 2020 Organizing Committee and TMS have rescheduled the 14th International Symposium on Superalloys to September 12–16, 2021. The collected proceedings is being published and released to the community as scheduled in September 2020 to ensure that that the reported findings are timely and up to date. This, the Fourteenth Symposium, will be taking place at a time when advances in the superalloy community have been largely driven by the development of property models, computational tools, processing methods, and innovative characterization techniques. For example, 3D mesoscale through atomic-scale characterization, machine learning algorithms, integrated computational materials engineering (ICME), and physics-based property models have all contributed to improve the processing and performance of existing materials, while accelerating the development of new alloys. As highlighted in the collection of proceedings, the development and application of innovative technologies in academia, industry, and government laboratories have been critical for improving the overall life cycle of superalloys. For the first time, the keynote address of this symposium will be a joint presentation from representatives of an engine OEM and a superalloy supplier. Christian Dumont, Chief of the Materials and Processing Modeling Department at Aubert & Duval, and Arnaud Longuet, an expert in the mechanics of high-temperature materials at Safran Aircraft Engines, will provide a unique overview of how data and information generated from process modeling tools used by the supply chain have been integrated into lifing methodologies used by the engine’s original equipment manufacturer (OEM). Starting with the Second Symposium in 1972, each symposium and its corresponding published proceedings have been dedicated to an individual as a means of honoring his or her contributions to the superalloy industry. The Fourteenth International Symposium is dedicated to Pierre Caron, a true pioneer and innovator in our field. Further details of Pierre’s career and contributions can be found on the following pages. Finally, it should be noted that this symposium would not have been possible without the efforts of the current and past members of the committees that serve the International Symposium on Superalloys. The Program Committee for the Fourteenth Symposium, listed below, vii

viii

Preface

was responsible for preparation of the technical program, including critical review of abstracts and manuscripts for originality, technical content, and pertinence to industry. The TMS staff, particularly Trudi Dunlap, Jennifer Booth, Matt Baker, and Doug Shymoniak, devoted considerable effort to organizing all other aspects of the symposium. Sammy Tin, Chair Mark Hardy Justin Clews Jonathan Cormier Qiang Feng John Marcin Chris O’Brien Akane Suzuki

Contents

Part I

Keynote

Advanced Modeling Tools for Processing and Lifing of Aeroengine Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Arnaud Longuet, Christian Dumont, and Eric Georges Part II

3

Alloy Development

Developing Alloy Compositions for Future High Temperature Disk Rotors . . . . M. C. Hardy, C. Argyrakis, H. S. Kitaguchi, A. S. Wilson, R. C. Buckingham, K. Severs, S. Yu, C. Jackson, E. J. Pickering, S. C. H. Llewelyn, C. Papadaki, K. A. Christofidou, P. M. Mignanelli, A. Evans, D. J. Child, H. Y. Li, N. G. Jones, C. M. F. Rae, P. Bowen, and H. J. Stone Development of AGAT, a Third-Generation Nickel-Based Superalloy for Single Crystal Turbine Blade Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Rame, P. Caron, D. Locq, O. Lavigne, L. Mataveli Suave, V. Jaquet, M. Perrut, J. Delautre, A. Saboundji, and J.-Y. Guedou Segregation of Solutes at Dislocations: A New Alloy Design Parameter for Advanced Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lola Lilensten, Philipp Kürnsteiner, Jaber Rezaei Mianroodi, Alice Cervellon, Johan Moverare, Mikael Segersäll, Stoichko Antonov, and Paraskevas Kontis Ni–Co-Based Wrought Superalloys Containing High W—Microstructure Design for a Balance of Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Akane Suzuki, Steve J. Buresh, Richard DiDomizio, Scott M. Oppenheimer, Soumya Nag, Ian M. Spinelli, P. R. Subramanian, Stephen G. Pope, and Jon C. Schaeffer On the Influence of Alloy Composition on Creep Behavior of Ni-Based Single-Crystal Superalloys (SXs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . O. M. Horst, S. Ibrahimkhel, J. Streitberger, N. Wochmjakow, P. Git, F. Scholz, P. Thome, R. F. Singer, C. Körner, J. Frenzel, and G. Eggeler Platinum-Containing New Generation Nickel-Based Superalloy for Single Crystalline Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jérémy Rame, Satoshi Utada, Luciana Maria Bortoluci Ormastroni, Lorena Mataveli-Suave, Edern Menou, Lucille Després, Paraskevas Kontis, and Jonathan Cormier

19

31

41

52

60

71

ix

x

Contents

Development and Application of New Cast and Wrought Ni-Base Superalloy M647 for Turbine Disk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Naoya Kanno, Masaya Higashi, Ryosuke Takai, Shigehiro Ishikawa, Kota Sasaki, Kenji Sugiyama, and Yoshinori Sumi Alloy Design and Microstructural Evolution During Heat Treatment of Newly Developed Cast and Wrought Ni-Base Superalloy M647 for Turbine Disk Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kenji Sugiyama, Yoshinori Sumi, Naoya Kanno, Masaya Higashi, Ryosuke Takai, Shigehiro Ishikawa, and Kota Sasaki c′ Thermodynamic Simulation and Experimental Validation of Phase Stability in Ni-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kyle Ventura, David Beaudry, Alex Aviles, Anna Kapustina, Phillip Draa, Kirtan Patel, Raymond Snider, and Gerhard Fuchs Composition and Temperature Stability of η and d Phases for Future Nickel-Base Superalloys for Turbine Disks Application . . . . . . . . . . . . . . . . . . . . Laurane Finet, Vladimir A. Esin, Vincent Maurel, and Loïc Nazé Advanced Alloy Design Program and Improvement of Sixth-Generation Ni-Base Single Crystal Superalloy TMS-238 . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tadaharu Yokokawa, Hiroshi Harada, Kyoko Kawagishi, Toshiharu Kobayashi, Michinari Yuyama, and Yuji Takata Phase Equilibria Among A1/TCP/GCP Phases and Microstructure Formation in Ni–Cr–Mo System at Elevated Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . Ryota Nagashima, Ryosuke Yamagata, Hirotoyo Nakashima, and Masao Takeyama A New Co-free Ni-Based Alloy for Gas Turbine and Exhaust Valve Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Karl A. Heck, Ning Zhou, Samuel J. Kernion, Danielle Rickert, and Filip Van Weereld On the Influence of Alloy Chemistry and Processing Conditions on Additive Manufacturability of Ni-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Joseph N. Ghoussoub, Yuanbo T. Tang, Chinnapat Panwisawas, André Németh, and Roger C. Reed

82

91

103

112

122

131

142

153

Alloying Effects on the Competition Between Discontinuous Precipitation Versus Continuous Precipitation of d/η Phases in Model Ni-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Satoru Kobayashi, Tomoki Otsuka, Rikuryo Watanabe, Kyosuke Sagitani, Masaki Okamoto, and Kako Tokutomi

163

Precipitate Phase Stability and Mechanical Properties of Alloy 263 and Variants in Wrought or Cast Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Martin Detrois, Paul D. Jablonski, and Jeffrey A. Hawk

171

Part III

Blade Alloy Behavior

Creep, Fatigue, and Oxidation Interactions During High and Very High Cycle Fatigue at Elevated Temperature of Nickel-Based Single Crystal Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Cervellon, J. Z. Yi, F. Corpace, Z. Hervier, J. Rigney, P. K. Wright, C. J. Torbet, J. Cormier, J. W. Jones, and T. M. Pollock

185

Contents

xi

Fretting Fatigue Life Extension for Single Crystal Ni-Based Superalloy by Applying Optimized Surface Texturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Okazaki, R. Balavenkatesh, S. Yamagishi, and M. Sakaguchi

196

Initiation of Fatigue Cracks in a Single-Crystal Nickel-Based Superalloy at Intermediate Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yaqi Huang, Dong Wang, Jian Shen, Yuzhang Lu, Langhong Lou, and Jian Zhang

208

Effect of Re on Long-Term Creep Behavior of Nickel-Based Single-Crystal Superalloys for Industrial Gas Turbine Applications . . . . . . . . . . . . . . . . . . . . . . Fan Lu, Longfei Li, Stoichko Antonov, Yufeng Zheng, Hamish L. Fraser, Dong Wang, Jian Zhang, and Qiang Feng Evaluation and Comparison of Damage Accumulation Mechanisms During Non-isothermal Creep of Cast Ni-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . Stoichko Antonov, Wenrui An, Satoshi Utada, Xiaotong Guo, Caspar Schwalbe, Weiwei Zheng, Catherine M. F. Rae, Jonathan Cormier, and Qiang Feng High-Temperature Pre-deformation and Rejuvenation Treatment on the Microstructure and Creep Properties of Ni-Based Single-Crystal Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Satoshi Utada, Jérémy Rame, Sarah Hamadi, Joël Delautre, Lorena Mataveli Suave, Patrick Villechaise, and Jonathan Cormier Evidence of Short-Range Order (SRO) by Dislocation Analysis in SingleCrystal Ni-Based Matrix Alloys with Varying Re Content After Creep . . . . . . . . Florence Pettinari-Sturmel, Joël Douin, Fabian Krieg, Ernst Fleischmann, and Uwe Glatzel

218

228

240

253

Rationalisation of the Micromechanisms Behind the High-Temperature Strength Limit in Single-Crystal Nickel-Based Superalloys . . . . . . . . . . . . . . . . . Daniel Barba, Ashton J. Egan, Yilun Gong, Michael J. Mills, and Roger C. Reed

260

Local Mechanical Properties at the Dendrite Scale of Ni-Based Superalloys Studied by Advanced High Temperature Indentation Creep and Micropillar Compression Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lukas Haußmann, Steffen Neumeier, Markus Kolb, Johannes Ast, Gaurav Mohanty, Johann Michler, and Mathias Göken

273

Phenomenological Modeling of the Effect of Oxidation on the Creep Response of Ni-Based Single-Crystal Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jean-Briac le Graverend and Seungjun Lee

282

Prediction of Rafting Kinetics of Practical Ni-Based Single-Crystal Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Yusuke Matsuoka, Yuhki Tsukada, and Toshiyuki Koyama

292

Creep Anisotropy in Single-Crystal Superalloy DD6 near the [001] Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jian Yu, J. R. Li, S. Z. Liu, and Mei Han

303

Equations to Predict the Elastic Modulus of the Individual Gamma and Gamma-Prime Phases in Multi-component Ni-Base Superalloys . . . . . . . . . . Takuma Saito, Makoto Osawa, Tadaharu Yokokawa, Hiroshi Harada, Toshiharu Kobayashi, Kyoko Kawagishi, and Shinsuke Suzuki

312

xii

Contents

Crystal Plasticity Mechanism of Temperature-Dependent Crack Propagation in a Single Crystal Nickel-Based Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . Xiaosheng Chen, Motoki Sakaguchi, Shiyu Suzuki, Hirotsugu Inoue, and Masakazu Okazaki Micro-mechanisms of Cyclic Plasticity at Stress Concentrations in a Ni-Based Single-Crystal Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alessandro Piglione, Jian Yu, Jinqian Zhao, Chengbo Xiao, Fionn Dunne, and Minh-Son Pham Tensile, Low Cycle Fatigue, and Very High Cycle Fatigue Characterizations of Advanced Single Crystal Nickel-Based Superalloys . . . . . . . . . . . . . . . . . . . . . Luciana Maria Bortoluci Ormastroni, Satoshi Utada, Jérémy Rame, Lorena Mataveli Suave, Kyoko Kawagishi, Hiroshi Harada, Patrick Villechaise, and Jonathan Cormier Competing Mechanism of Creep Damage and Stress Relaxation in Creep-Fatigue Crack Propagation in Ni-Base Superalloys . . . . . . . . . . . . . . . . Shiyu Suzuki, Motoki Sakaguchi, Ryota Okamoto, Hideaki Kaneko, Takanori Karato, Kenta Suzuki, and Masakazu Okazaki Part IV

324

333

341

352

Component Manufacture and Repair

Microstructure and Material Properties of Alloy 718/713LC Joints Using Orbital Friction Welding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Björn Hinze

365

Enhancing the Efficiency and Surface Integrity of Chemical Cleaning During Repair of Ni-Base Superalloy Rotating Disks . . . . . . . . . . . . . . . . . . . . . E. Huron, N. Tibbetts, Z. Bolukoglu, and T. Webster

374

HIP + ITF of SS-PREP® Superalloy Powder . . . . . . . . . . . . . . . . . . . . . . . . . . . . Zonghong Qu, Shujin Liang, Yunjin Lai, Pingxiang Zhang, Jiaming Song, Ruimin Bai, and Cheng Luo

384

An Integrated HIP Heat-Treatment of a Single Crystal Ni-Base Superalloy . . . . Benjamin Ruttert, Inmaculada Lopez-Galilea, and Werner Theisen

391

3 Ton Melting with CaO Desulfurization of Ni-Base Single Crystal Superalloy TMS-1700, Simulating a Recycling of Used Turbine Blades . . . . . . . . . . . . . . . . Tadaharu Yokokawa, Hiroshi Harada, Kyoko Kawagishi, Masao Sakamoto, Makoto Osawa, Yuji Takata, Michinari Yuyama, Toshiharu Kobayashi, Takuya Sugiyama, and Shinsuke Suzuki On Optimising Ring-Rolling Manufacturability of C&W Nickel Superalloys for Aero-engine Turbine Disc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fauzan Adziman, Ryosuke Takai, Yuanbo T. Tang, Shigehiro Ishikawa, Daniel Barba, Enrique Alabort, Andre Nemeth, Naoya Kanno, and Roger Reed High-Resolution Diffraction Imaging of Misorientation in Ni-Based Single Crystal Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Robert Albrecht, Maciej Zubko, Kamil Gancarczyk, and Dariusz Szeliga Assessment of Mechanical and Metallurgical Features of Inconel 680 Weld Metal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rafaella S. Silva, Émerson M. Miná, Giovani Dalpiaz, Ricardo M. Reppold, Marcelo T. P. Paes, Marcelo F. Motta, Cleiton C. Silva, and Hélio C. de Miranda

400

408

421

431

Contents

xiii

Part V

Disk Alloy Manufacture

Gamma Prime Precipitate Evolution During Hot Forging of a c–c′ Ni-Based Superalloy at Subsolvus Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marcos Pérez, Christian Dumont, and Sebastien Nouveau Dynamic and Post-dynamic Recrystallization During Supersolvus Forging of the New Nickel-Based Superalloy—VDM Alloy 780 . . . . . . . . . . . . . . . . . . . . Juhi Sharma, Masood Hafez Haghighat, Bodo Gehrmann, Charbel Moussa, and Nathalie Bozzolo Aspects of High Strain Rate Industrial Forging of Inconel 718 . . . . . . . . . . . . . . A. Reshetov, N. Stefani, O. Bylya, B. Krishnamurthy, and P. Blackwell Tuning Strain Localization in Polycrystalline Nickel-Based Superalloys by Thermomechanical Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. A. Charpagne, J. C. Stinville, A. T. Polonsky, M. P. Echlin, S. P. Murray, Z. Chen, N. Bozzolo, J. Cormier, V. Valle, and T. M. Pollock Impact of Coarse c′ Phase on Recrystallization Modeling in New Ni-Based Superalloy M647 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Takashi Nishimoto, Takuma Okajima, Kenta Yamashita, Qiaofu Zhang, Jiadong Gong, and Greg Olson Development of a Prediction Model and Process–Microstructure–Property Database on Forging and Heat Treatment of Superalloy 720Li . . . . . . . . . . . . . . Nobufumi Ueshima, Chuya Aoki, Toshio Osada, Satoko Horikoshi, Akira Yanagida, Hideyuki Murakami, Toshiki Ishida, Yoko Yamabe-Mitarai, Katsunari Oikawa, Nobuki Yukawa, and Jun Yanagimoto

441

450

461

471

482

491

Phase-Field Modeling of c′ and c″ Precipitate Size Evolution During Heat Treatment of Ni-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Felix Schleifer, Michael Fleck, Markus Holzinger, Yueh-Yu Lin, and Uwe Glatzel

500

Characteristic Flow Behaviour of c + cʹ Duplex and Its Significant Applications in Hot Working Process of Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Beijiang Zhang, Wenyun Zhang, Jiantao Liu, Shuo Huang, and Shifu Chen

509

Abnormal Grain Growth in the Presence of Grain Boundary Pinning Precipitates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. G. Fahrmann and D. A. Metzler

519

Part VI

Disk Alloy Behavior

The Effect of Shot Peening on the Ductility and Tensile Strength of Nickel-Based Superalloy Alloy 720Li . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. J. Jackson, J. Rolph, R. C. Buckingham, and M. C. Hardy Metallurgical Mechanisms upon Stress Relaxation Annealing of the AD730TM Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Malik Durand, Jonathan Cormier, Patrick Villechaise, Jean-Michel Franchet, Christian Dumont, and Nathalie Bozzolo Metallurgical Analysis of Direct Aging Effect on Tensile and Creep Properties in Inconel 718 Forgings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alexis Nicolaÿ, Jean-Michel Franchet, Nathalie Bozzolo, and Jonathan Cormier

535

546

559

xiv

Contents

Experimental and Simulation Study of the Effect of Precipitation Distribution and Grain Size on the AD730TM Ni-Based Polycrystalline Superalloy Tensile Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marco Panella, Loïc Signor, Jonathan Cormier, Marc Bernacki, and Patrick Villechaise

570

Understanding the Effects of Alloy Chemistry and Microstructure on the Stress Relaxation Behavior of Ni-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . Linhan Li, Joshua McCarley, Eugene Sun, and Sammy Tin

579

Is the Carbon Content Really an Issue for the LCF Durability of Forged c/c′ Ni-Based Disk Alloys? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adèle Govaere, Anne-Laure Rouffié, Jean-Michel Franchet, Daniel Galy, Christian Dumont, Alexandre Devaux, Coraline Crozet, Paraskevas Kontis, Patrick Villechaise, and Jonathan Cormier High Temperature Dwell Fatigue Crack Growth in Cold-Worked and Direct-Aged 718PlusTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrew Merrison, Hangyue Li, Paul Bowen, and Wei Li Contribution of Primary c′ Precipitates in the Deformation Creep Mechanisms in the Ni-Based Polycrystalline AD730TM Superalloy . . . . . . . . . . . . . . . . . . . . . Florence Pettinari-Sturmel, Muriel Hantcherli, Winnie Vultos, Cécile Marcelot, Maud Tisseyre, Robin Cours, Joël Douin, Patrick Villechaise, and Jonathan Cormier Strengthening Mechanisms of Ni–Co–Cr Alloys via Nanotwins and Nanophases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Gan, J. Gu, M. A. Monclús, X. Dong, Y. S. Zhao, H. Y. Yu, J. H. Du, M. Song, and Z. N. Bi Role of Non-metallic Inclusions and Twins on the Variability in Fatigue Life in Alloy 718 Nickel Base Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Damien Texier, Jean-Charles Stinville, Marie-Agathe Charpagne, Zhe Chen, Valery Valle, Patrick Villechaise, Tresa M. Pollock, and Jonathan Cormier Effect of Nb Alloying Addition on Local Phase Transformation at Microtwin Boundaries in Nickel-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. J. Egan, Y. Rao, G. B. Viswanathan, T. M. Smith, M. Ghazisaeidi, S. Tin, and M. J. Mills

591

603

613

619

629

640

A New Approach to Strength Prediction of Ni-Base Disk Superalloys with Dual-Phase c/c′ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . L. Wu, T. Osada, I. Watanabe, T. Yokokawa, T. Kobayashi, and K. Kawagishi

651

Crystal Plasticity Model for Nickel-Based Superalloy René 88DT at Elevated Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Monica A. Soare, Shenyan Huang, and Mallikarjun Karadge

659

Deformation Mechanisms of c′ and c″ Co-precipitates in IN718 . . . . . . . . . . . . . Longsheng Feng, Donald McAllister, Christopher H. Zenk, Michael J. Mills, and Yunzhi Wang High-Throughput Approaches to Establish Quantitative Process–Structure–Property Correlations in Ni-Base Superalloy . . . . . . . . . . . . . Nishan M. Senanayake, Semanti Mukhopadhyay, and Jennifer L. W. Carter

669

680

Contents

xv

An Approach Toward Understanding Unstable Gamma Prime Precipitate Evolution and Its Effect on Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nicholas Krutz, Chen Shen, Mallik Karadge, Ashton J. Egan, Justin R. Bennett, Timothy Hanlon, and Michael J. Mills Modeling Creep of Ni-Base Superalloys for Applications in Advanced Ultra-supercritical Power Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Monica Soare, Chen Shen, and Vito Cedro III Stress-Induced Variant Selection of c″ Phase in Inconel 718 During Service: Mechanism and Effects on Mechanical Behavior . . . . . . . . . . . . . . . . . . . . . . . . . Hailong Qin, Zhongnan Bi, Ruiyao Zhang, Tung Lik Lee, Hongyao Yu, Hai Chi, Dongfeng Li, Hongbiao Dong, Jinhui Du, and Ji Zhang Enhancing the Creep Strength of Next-Generation Disk Superalloys via Local Phase Transformation Strengthening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . T. M. Smith, T. P. Gabb, K. N. Wertz, J. Stuckner, L. J. Evans, A. J. Egan, and M. J. Mills Part VII

691

702

713

726

Environmental Behavior

Characterization of the Benefit of APS Flash Coatings in Improving TBC Lifetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bruce A. Pint, Michael J. Lance, Ercan Cakmak, Kenneth A. Kane, J. Allen Haynes, Edward J. Gildersleeve, and Sanjay Sampath Hot Corrosion and Creep Properties of Ni-Base Single-Crystal Superalloys . . . . Yutaka Koizumi, Kyoko Kawagishi, Tadaharu Yokokawa, Michinari Yuyama, Yuji Takata, and Hiroshi Harada Investigation into the Effects of Salt Chemistry and SO2 on the Crack Initiation of CMSX-4 in Static Loading Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Duarte Martinez, N. I. Morar, M. Kothari, G. Gibson, J. Leggett, J. C. Mason-Flucke, J. R. Nicholls, G. M. Castelluccio, and S. Gray Using Rapid Thermal Annealing for Studying Early Stages of High-Temperature Oxidation of Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . Dorota Kubacka, Yolita M. Eggeler, Nicklas Volz, Steffen Neumeier, and Erdmann Spiecker Measurement and Evaluation of Co-existing Crack Propagation in Single-Crystal Superalloys in Hot Corrosion Fatigue Environments . . . . . . . . L. Brooking, C. Ferguson, J. Mason-Flucke, G. Gibson, J. Leggett, I. Palmer, J. R. Nicholls, and S. Gray

739

747

753

763

771

Critical Hafnium Content for Extended Lifetime of AM1 Single Crystal Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . F. Pedraza, R. Troncy, A. Pasquet, J. Delautre, and S. Hamadi

781

Laboratory-Scale Replication of Deposit-Induced Degradation of High-Temperature Turbine Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Matthew Kovalchuk and Brian Gleeson

789

Suppression of Sulfur Segregation at Scale/Substrate Interface for Sixth-Generation Single-Crystal Ni-Base Superalloy . . . . . . . . . . . . . . . . . . . Kyoko Kawagishi, Chihiro Tabata, Takuya Sugiyama, Tadaharu Yokokawa, Jun Uzuhashi, Tadakatsu Ohkubo, Yuji Takata, Michinari Yuyama, Shinsuke Suzuki, and Hiroshi Harada

798

xvi

Contents

Development of a New Coating Compatible with Third-Generation Nickel-Based Superalloys and Thermal Barrier Coatings . . . . . . . . . . . . . . . . . . A. Saboundji, V. Jaquet, L. Mataveli Suave, and J. Rame Recent Progress in Local Characterization of Damage Evolution in Thermal Barrier Coating Under Thermal Cycling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vincent Maurel, Lara Mahfouz, Vincent Guipont, Basile Marchand, Fabrice Gaslain, Alain Koster, Anne Dennstedt, Marion Bartsch, and Florent Coudon

804

813

Computational Methods to Accelerate Development of Corrosion Resistant Coatings for Industrial Gas Turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R. Pillai, K. Kane, M. Lance, and B. A. Pint

824

The Influence of Hot Corrosion Damage on the Low Cycle Fatigue Fracture Modes of a Disk Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J. Hart, M. Task, and M. Bochiechio

834

Part VIII

Alternative Materials

Recent Developments in the Design of Next Generation c′-Strengthened Cobalt–Nickel Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stephane A. J. Forsik, Ning Zhou, Tao Wang, Alberto O. Polar Rosas, Austin D. Dicus, Gian A. Colombo, Andrea Ricci, and Mario E. Epler Supersolvus Hot Workability and Dynamic Recrystallization in Wrought Co–Al–W-Base Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Katelun Wertz, Donald Weaver, Dongsheng Wen, Michael S. Titus, Rajiv Shivpuri, Stephen R. Niezgoda, Michael J. Mills, and S. Lee Semiatin Effects of Al, Cr, and Ti on the Oxidation Behaviors of Multi-component c/cʹ CoNi-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Xiaoli Zhuang, Longfei Li, and Qiang Feng Microstructure and Tensile Properties of a CoNi-Based Superalloy Fabricated by Selective Electron Beam Melting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sean P. Murray, Kira M. Pusch, Andrew T. Polonsky, Chris J. Torbet, Gareth G. E. Seward, Peeyush Nandwana, Michael M. Kirka, Ryan R. Dehoff, Ning Zhou, Stéphane A. J. Forsik, William Slye, and Tresa M. Pollock

847

857

870

880

Microstructural Effects on Creep Properties in a Co-Base Single Crystal Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . H. J. Zhou, L. F. Li, S. Antonov, and Q. Feng

891

Castability and Recrystallization Behavior of c′-Strengthened Co-Base Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. Volz, C. H. Zenk, T. Halvaci, K. Matuszewska, S. Neumeier, and M. Göken

901

The Effect of Alloying on the Thermophysical and Mechanical Properties of Co–Ti–Cr-Based Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Christopher H. Zenk, Nicklas Volz, Andreas Bezold, Laura-Kristin Huber, Yolita M. Eggeler, Erdmann Spiecker, Mathias Göken, and Steffen Neumeier Atomic Structure and Chemical Composition of Planar Fault Structures in Co-Base Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Malte Lenz, Mingjian Wu, Junyang He, Surendra K. Makineni, Baptiste Gault, Dierk Raabe, Steffen Neumeier, and Erdmann Spiecker

909

920

Contents

xvii

Effects of Ti and Cr Additions in a Co–Ni–Al–Mo–Nb-Based Superalloy . . . . . . Nithin Baler, Prafull Pandey, Mahendra Pratap Singh, Surendra Kumar Makineni, and Kamanio Chattopadhyay Machine Learning Assisted Design Approach for Developing c′-Strengthened Co-Ni-Base Superalloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Min Zou, Wendao Li, Longfei Li, Ji-Cheng Zhao, and Qiang Feng The Yield Strength Anomaly in Co–Ni Design Space . . . . . . . . . . . . . . . . . . . . . K. V. Vamsi, Sean P. Murray, and Tresa M. Pollock Part IX

929

937 948

Additive

Microstructure and Mechanical Properties of Additively Manufactured Rene 65 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrew Wessman, Jonathan Cormier, Florence Hamon, Kelsey Rainey, Sammy Tin, Dhruv Tiparti, and Laura Dial Novel Approach for Suppressing of Hot Cracking Via Magneto-fluid Dynamic Modification of the Laser-Induced Marangoni Convection . . . . . . . . . . . . . . . . . A. Seidel, L. Degener, J. Schneider, F. Brueckner, E. Beyer, and C. Leyens Effect of Carbide Inoculants Additions in IN718 Fabricated by Selective Laser Melting Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tzu-Hou Hsu, Kai-Chun Chang, Yao-Jen Chang, I-Ting Ho, Sammy Tin, Chen-Wei Li, Koji Kakehi, Chih-Peng Chen, Kuo-Kuang Jen, Ho-Yen Hsieh, and An-Chou Yeh 3D Characterization of the Columnar-to-Equiaxed Transition in Additively Manufactured Inconel 718 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrew T. Polonsky, Narendran Raghavan, McLean P. Echlin, Michael M. Kirka, Ryan R. Dehoff, and Tresa M. Pollock

961

972

982

990

Anisotropic Deformation and Fracture Mechanisms of an Additively Manufactured Ni-Based Superalloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1003 Cheng-Han Yu, Ru Lin Peng, Mattias Calmunger, Vladimir Luzin, Håkan Brodin, and Johan Moverare Microstructural Control and Optimization of Haynes 282 Manufactured Through Laser Powder Bed Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1014 K. A. Christofidou, H. T. Pang, W. Li, Y. Pardhi, C. N. Jones, N. G. Jones, and H. J. Stone Additive Manufacturability of Nickel-Based Superalloys: Composition-Process Induced Vapourization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1024 Chinnapat Panwisawas, Yuanbo Tony Tang, Joseph Ghoussoub, and Roger C. Reed Strain Monitoring During Laser Metal Deposition of Inconel 718 by Neutron Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1033 S. Cabeza, B. Özcan, J. Cormier, T. Pirling, S. Polenz, F. Marquardt, T. C. Hansen, E. López, A. Vilalta-Clemente, and C. Leyens Development of a New Alumina-Forming Crack-Resistant High-c′ Fraction Ni-Base Superalloy for Additive Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . 1046 Ning Zhou, Austin D. Dicus, Stephane A. J. Forsik, Tao Wang, Gian A. Colombo, and Mario E. Epler

xviii

Contents

The Effect of Heat Treatment on Tensile Yielding Response of the New Superalloy ABD-900AM for Additive Manufacturing . . . . . . . . . . . . 1055 Yuanbo T. Tang, Joseph N. Ghoussoub, Chinnapat Panwisawas, David M. Collins, Sajjad Amirkhanlou, John W. G. Clark, André A. N. Németh, D. Graham McCartney, and Roger C. Reed Computational Design of Additively Printable Nickel Superalloys . . . . . . . . . . . . 1066 Adarsh Shukla, Sanket Sarkar, A. Durga, Raghav Adharapurapu, Laura Dial, and Sanjay K. Sondhi Mechanical Performance of a Non-weldable Ni-Base Superalloy: Inconel 738 Fabricated by Electron Beam Melting . . . . . . . . . . . . . . . . . . . . . . . 1075 Michael M. Kirka, Patxi Fernandez-Zelaia, Yousub Lee, Peeyush Nandwana, Sean Yoder, Obed Acevedo, and Daniel Ryan Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1085

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1091

Dedication

The 14th International Symposium on Superalloys is dedicated to Pierre Caron for his substantial contributions to our field. He has been a pioneer and innovator in the development of Ni-based single-crystal superalloys, their processing, property optimization, and compatibility with coatings. Dr. Pierre Caron was born in Paris, France, in September 1954. He attended Paris XI University from 1971 to 1979, earning a postgraduate degree in special metallurgy in 1976 and a Ph.D. degree in metallurgy in 1979. After a one-year military internship at the French Atomic Energy Commission, he joined ONERA, the French Aerospace Research Center, in November 1980 as a research engineer. As his career progressed, Pierre was successively appointed Head of the Cast Superalloys Group, Head of the High Temperature Materials Research Unity, and finally Special Advisor in the field of superalloys. In 2005, he successfully defended a habilitation thesis to supervise research and advise Ph.D. students and then he received the title of senior researcher at ONERA. Since his retirement in December 2016, he has kept in touch with the world of superalloys as a scientific consultant. Pierre began at ONERA in the Department of Materials initially under the supervision of Dr. Tasadduq Khan who led the activities on nickel-based superalloys for single-crystalline blades. During his entire career, this topic has been the central theme of his research activities, although he also made important contributions to studies on intermetallic high-temperature alloys and

xx

in the field of wrought powder metallurgy superalloys (alloys NR3, NR6, and N19). He contributed to the design and the development of new single-crystal superalloy chemistries (AM1, AM3, MC2, MC-NG alloys) for applications in aircraft and helicopter engines to fulfill the requirements of the French gas turbine engine manufacturers of the Safran Group. AM1 is currently used as high-pressure turbine blade and vane material in the M88-2 engine powering the Rafale fighter, in the TP400-D6 European engine for the Airbus A400M military transport airplane, and in the SaM146 engine produced by PowerJet, a joint venture between Safran and NPO Saturn, for powering of the Sukhoi Superjet 100 regional aircraft. AM3 and MC2 alloys were introduced in various versions of the Arriel 2 and Arrius 2 Safran helicopter engines. Pierre also participated in the development of the SC16 superalloy that was optimized for large single-crystal blades in industrial gas turbines (IGTs). His interest in this area extended to the coordination of European research BRITE-EURAM program SC2, which led to the development of the SCA425 and SCB444 single-crystal superalloys for IGT. Pierre contributed also to the development of the THYMONEL 2 and THYMONEL 8 superalloys specifically designed for single-crystalline blades of a hydrogen turbo-pump in rocket engines. Very recently, he made a significant contribution to the design of the AGAT singlecrystal superalloy that was developed to meet the demanding requirements of advanced Safran Aircraft Engines. The other research activities performed by Pierre covered various topics including: the application of transmission electron microscopy to analyze deformation mechanisms, effects of solidification parameters, heat treatments, anisotropy, microstructure and microstructural instabilities on the mechanical behavior, effects of minor alloying elements on oxidation resistance, interactions with protective coatings, optimization of the single-crystal superalloy/thermal barrier system, strengthening c′-Ni3Al-based and b-NiAl-based intermetallic alloys, and contributions to the development of Co-based and Cr-based alloys. Several of these studies were performed in collaboration with a number of French academic research laboratories, giving Pierre the opportunity to explore fundamental aspects of the physics and the chemistry of various high-temperature alloys. Pierre is Author/Co-author of about 160 research papers in journals and conference proceedings and, as Co-inventor, made applications for 15 patents. From 1988 to 2014, he was Teacher and then Educational Leader at the National Conservatory of Arts and Crafts, Paris, France, in presenting the training course “Properties and Applications of Superalloys.” He was Co-organizer of Superalloy Symposiums at the EUROMAT 1999 and THERMEC 2003 Conferences, Member of the Program Committee of the 11th International Symposium on Superalloys, and Member of the Organizing Committee of the Euro Superalloys 2010 and Euro Superalloys 2014 Conferences. He received in 1988 in Hanover, Germany, the Dr. Ernst Zimmerman Memorial Award for its contribution to the promotion of propulsion technology in the field of aeronautics. As Co-author, he received the Best Paper Award of the International Symposium on Superalloys in 1988 and 2012. In 2017, he was named Knight in the Order of Academic Palms, delivered by the French Ministry of National Education, Higher Education and Research.

Dedication

Dedication

xxi

Last but not least, Pierre has always been a source of knowledge and inspiration for younger generations from both academia and industry, sharing in a very kind way all the experience he has gained from his professional activities. Past Dedicatees 1972—Clarence Bieber 1976—Falih Damara 1980—Rudy Thielman 1984—Gunther Mohling 1988—Herb Eiselstein 1992—Carl Lund 1996—John Radavich 2000—Wilfred “Red” Couts 2004—Fred Pettit 2008—Raymond Decker 2012—Anthony Giamei 2016—Louis W. Lherbier

Best Paper Award

The following paper was selected by the Awards Committee of the International Symposium on Superalloys as the winner of the Best Paper Award for the 14th symposium: “Enhancing the Creep Strength of Next Generation Disk Superalloys via Local Phase Transformation Strengthening” by T. M. Smith1, T. P. Gabb1, K. N. Wertz2, J. Stuckner1, L. J. Evans1, A. J. Egan3, M. J. Mills3; 1

NASA Glenn Research Center, 21000 Brookpark Rd., Cleveland, OH 44135, USA; 2 Air Force Research Laboratory, Wright-Patterson Air Force Base, OH 45433, USA; 3 Center for Electron Microscopy and Analysis and the Department of Materials Science and Engineering, The Ohio State University, Columbus, OH, USA

The selection was based on the following criteria: originality, technical content, pertinence to the superalloys and gas turbine industries and academic community, and clarity and style.

xxiii

Committee Members (14th International Symposium on Superalloys)

Program Chair Sammy Tin Program Committee Justin Clews Jonathan Cormier Qiang Feng Mark Hardy John Marcin Chris O’Brien Akane Suzuki Awards Committee Eric Huron Gern Maurer Tresa Pollock Roger Reed

xxv

Part I Keynote

Advanced Modeling Tools for Processing and Lifing of Aeroengine Components Arnaud Longuet, Christian Dumont, and Eric Georges

Abstract

Introduction

Lifing is one of the main challenges for aeroengine manufacturers. For fatigue prediction, attention has been focused on the crack initiation mode depending on stress level and initial microstructure. Microstructure prediction during the component manufacturing, especially for final heat treatments and final forging operations, is required if it is to be included in fatigue analysis. Reliable tools are now available for basic nickel-based alloys such as Inconel 718. For other alloys, notably c/c′ alloys, research is still being performed in close partnership with academia. Globally, two main trends are emerging; first, one of our main interests is to develop the modeling capability for the entire manufacturing process, including ingot conversion and billet forging. Second, new approaches are still under development by introducing more physical considerations through full-field models, which are very useful for a better understanding of specific issues such as heterogeneous grain growth. From a component lifing point of view, the initial state of stress is also a key parameter to be considered. One method for the control of residual stresses is application of a pre-spinning process. Finally, a standard lifing methodology is explained and improvements are proposed; in particular, size effect is used to model notch specimen life considering surface or internal initiation. Keywords












Inconel 718 c/c′ alloys Modeling Recrystallization Lifing Residual stresses Manufacturing process

A. Longuet Safran Aircraft Engines VILLAROCHE Materials and Processes Division, Rond-Point René Ravaud, 77550 Moissy-Cramayel, France C. Dumont (&)
E. Georges R&D Department, Aubert & Duval, BP1, 63770 Les Ancizes, France e-mail: [email protected]




Aeroengine turbine disks are among the most challenging components to design. Material choice and microstructure evolution are crucial to the performance and durability of a part design, which is dependent on the application, required fatigue/creep life and temperatures reached. The design process for such a component is iterative. To select a material for a particular application, specifications need to be given by the pre-design and the design department. Highly representative material properties are necessary for models to accurately predict the life and integrity of a part. Disk qualification packages are submitted for certification to airworthiness authorities (e.g., FAA and EASA). The two most critical risks to avoid are disk burst in case of over-speed and disk failure due to fatigue. Material properties and their response to processing are of the utmost importance to meet certification requirements. Often times, the material performance for the selected application is too low with respect to the design requirements and a new superalloy is desired. However, multiple approaches can be used in an aim to meet the design requirements: Improve the material by a change in the chemical composition, improve the manufacturing process (i.e., conversion, forging, heat treatment, machining) to have better material properties, or improve the material’s constitutive modeling approaches to better understand any conservatism in the predictive methodology. For decades, many research programs have been devoted to microstructure prediction, from ingot casting to final closed die forging of parts. For the forging process, models are based on post-processing of thermomechanical histories during part manufacturing, deduced from finite element modeling, Inconel 718 being the material of choice [1–3]. This presentation will be oriented toward improving the modeling tools that can help with improving forge process and lifing methodologies of forged rotating components. For

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_1

3

4

microstructure prediction, we propose to address more specifically the three following points: – The reliability of the thermomechanical description of processes as a requirement for realistic microstructure prediction – The relevance of new full-field models for microstructure prediction – The specificity of c/c′ alloys versus Inconel 718. Several examples presented in this article are from Aubert & Duval and Safran Aircraft Engines in house research activities, as well as from French Industry–University research programs, through joint PhD students or large-scale programs like the OPALE, DIGIµ, and TOPAZE chairs [4– 6] supported by the ANR (French National Agency for Scientific Research), Safran, and/or Aubert & Duval. It should be noted that research activities on hightemperature materials can be performed on very applied (industrial) topics in French academic laboratories due to a well-established intellectual property (IP) system. The IP is shared in most cases between the university and the industrial partner to existing agreements. Governmental initiatives encouraging joint industry–university research are further strengthened by providing funding, e.g., half of Ph.D. student grants if hired by industry or half of the financial support for large-scale research programs (like the above-mentioned ones) for more fundamental research activities.

Effect of Microstructure on Inconel 718 Life Inconel 718 (IN718) is the most widely used superalloy for disk applications. It is low cost compared to c/c′ superalloys and relatively easy to process and has good mechanical properties up to 650 °C. The forging process produces different microstructures, mainly in terms of grain size. The understanding of crack initiation mode on the durability of

A. Longuet et al.

IN718 is critical to be able to estimate the scatter linked to each crack initiation mechanism [7]. Figure 1a shows that the surface crack initiation (example in Fig. 1b) mode has a lower scatter than the internal crack initiation mode. It is of the utmost importance at the stress level where the two crack initiation modes are competing to be able to draw the minimal master curve used for the component design, and to understand the possible origins of scatter. Turbine disks always show variation in grain size due to the forging process. Among all the microstructural features critical for component lifing, grain size is the most important for IN718, while this is not always true for other c/c′ alloys [8–12]. Figure 2 shows the effect of grain size on IN718 fatigue life. In the same way, the fatigue life variability is dependent on the crack initiation mechanism and hence on the grain size. For a surface crack initiation mode, carbides/nitrides for 10 ASTM (*10 µm) material or twin boundaries for 7 ASTM (*30 µm) material can be crack starters [8, 10–13]. But the fatigue life remains identical between these two modes. For the internal crack initiation mode, as 10 ASTM IN718 material already initiates on grains/at twin boundaries, the initiation mode does not change. However, the fatigue life is reduced with larger grain size. A more detailed review of the data is presented in Figs. 3 and 4. LCF strain-controlled fatigue tests were performed on two different IN718 materials (different forging routes) with nearly the same grain size at an intermediate strain level where crack initiation could occur at the surface with a low life or internally with higher lives [8]. According to Fig. 3, it is clear that material 1 has a lower life compared to material 2. In an unexpected way, material 1 initiates mainly on grains/twin boundaries near the surface and material 2 internally on nonmetallic inclusions, nitrides especially. Better fatigue life is expected with internal crack initiation compared to (sub)surface initiation, but usually grain initiation gives better lives than inclusion initiation. The explanation for this result can be found in Fig. 4. The two materials have a very similar average grain size, but the

Fig. 1 Nature of crack initiation mode of IN718 in a S–N diagram at low and intermediate temperature (i.e., T < 500 °C) (a) and typical LCF surface crack initiation from a nonmetallic inclusion (a nitride in this case) at intermediate temperature—adapted from [10] (b)

Advanced Modeling Tools for Processing and Lifing …

Fig. 2 Effect of grain size on fatigue life of Inconel 718 (G—ASTM grain size)

5

grain size distribution of material 1 has a longer tail, with sizes even larger than maximum nitride size. Crack initiation occurring on the largest microstructural feature between nonmetallic inclusions and twins, the results of Fig. 3 are confirmed. A similar trend is also observed in AD730TM c/c′ alloy in LCF at 450 °C [9], despite a reduced variability in LCF life compared to IN718. Moreover, it has been clearly shown by Texier et al. that at fixed grain sizes and by considering only nonmetallic inclusion crack initiation, the precipitation state is critical in terms of fatigue life variability. In addition, an alloy with a higher content of c″ (i.e., lower d content) leads to greater fatigue life variability for IN718 [10]. All these considerations show that both the microstructure and initial state of stress/strain should be controlled carefully throughout the component manufacturing process. Modeling tools are now used regularly in order to improve both the forging and heat treatment routes. The main goal is to predict the microstructures and residual stress levels. Thus, a review is proposed in the next subsections on the developments and capability of “metallurgical and mechanical post-processing modules” implemented in commercial process simulation software packages.

Microstructure Modeling Closed Die Forging Fig. 3 Comparison between LCF lives of different IN718 forgings at an intermediate strain level and low temperature (i.e., T < 500 °C)— adapted from [8]

Commercial software packages such as FORGE or DEFORM based on Johnson–Mehl–Avrami–Kolmogorov (JMAK) formulations [14] are used to model closed die forging of IN718. They are very useful for testing different forging routes for each new application (geometry of the blank, lubrication, die temperature, etc.). Moreover, correlation between modeling and experimental results is a good way to validate process control and the prediction of thermomechanical history at different locations in the forged part. However, this basic approach may become unsatisfactory for more complex situations such as: – Multistep forming processes (rolling or ring rolling, open die forging, etc.) – Adiabatic heating over the d-solvus, leading to a faster dissolution of d phase and consequently to unexpected grain growth.

Fig. 4 Grain size histograms and maximum carbide and nitride sizes for Fig. 2 materials—adapted from [8]

In such cases, modeling equations must be improved, leading to specific experiments to assess new parameters. By considering again the same two examples:

6

– Recovery effects between each deformation step must be introduced for unrecrystallized grains in the description of the residual strain coming from the previous deformation pass [15]. – In contrast to a simple thermal treatment, deformation significantly increases the precipitate dissolution rate [16]. All these improvements lead to more reliable and complex models, which come progressively close to physical and mean-field approaches. Residual strain accumulated between each pass during an incremental process can thus be considered as describing dislocation density. However, JMAK models are still largely used since the sensitivity analysis and parameter identification are easier, with their main parameters being directly connected to thermomechanical data (strain, temperature, etc.). Moreover, physical models are more difficult to implement in computation codes, because we have to manage larger numbers of parameters, increasing with the level of detail for microstructure description: kinetics of misorientation of sub-grains, twins [17]. Ultimately, it becomes impossible to get a complete map of microstructure parameter distribution (grain size, recrystallized fraction etc.) through metallurgical post-processing. Each thermomechanical operation, extracted from the initial finite element model, needs to be treated separately. In this sense, full-field models are much more suitable to account for microstructure evolution. A large research program is currently in progress in France which aims at developing the DIGIµ™ software package, based on the state-of-the-art numerical methods and metallurgical models, but optimized to enable their industrial use [18–21]. The collaboration between industrial and academic partners runs in a specific framework with interactive governance and shared IP principles that have been set up. Similar to what has been done in the past for other software developments, such as FORGE, the interaction between industry and academics proceeds through the following steps: – New numerical developments are proposed by the academia with agreed-upon needs expressed by the industrial partners, in terms of metallurgical phenomena, materials, and processes. – The new software implementation and their tutorials are delivered by the research laboratory (CEMEF—MINES ParisTech, Sophia Antipolis) to the industrial partners through a software editor company (Transvalor). – Industrial partners provide a feedback on the software usage and parameter identification for their materials of interest. The latter issue is also subjected to common work performed between several industrial companies of the funding consortium.

A. Longuet et al.

– Important feedback is provided for new proposals and requests for further software development that lead to new academic research programs. Founded and co-funded by industrial partners and the French government (through the French National Agency for Scientific Research, ANR), this process leads to a strong partnership that is highly efficient, flexible, and agile. Even if they are not IP free at the beginning, these numerical tools are ultimately intended to be disseminated on the market by the software company. However, this process offers numerous benefits for the initial funders like exclusive use period, discount for maintenance, and definition of future development plans. Based on a level-set description of polycrystals, the DIGIµ framework has been developed in order to model the evolution of an actual grain size distribution during static, dynamic, and post-dynamic recrystallization. In this case, the main goal is not to simulate an actual forging process, but it can be very useful for optimization of model parameters. One of the main interests of this approach consists also in carrying out numerical experiments in order to study specific phenomena occurring during subsequent solution heat treatments. One typical example deals with heterogeneous grain growth [22], which can be uncontrolled and lead to large grain sizes, with a deleterious impact on mechanical properties [23]. For this specific concern, three main parameters governing grain boundary motion have to be taken into account: – Capillarity driving force directly connected to grain size distribution – Distribution of second-phase particles, such as d phase in IN718 (volume fraction and size) – Difference of stored energy between neighboring grains. An example is shown in Fig. 5 adapted from [22]. Pattern (a) represents an initial microstructure on IN718 (100  100 µm) with a volume fraction of 4% of round d particles 0.8 µm in diameter. The blue grain is free of stored energy, while the surrounding grains are characterized by a stored energy equal to 200 kJ/m3. Subsequent heat treatment at 985 °C is then calculated with DIGIµTM. This difference in stored energy leads to the faster growth of the blue grain after 600 s (b), 1800 s (c), and 3600 s (d). Similar phenomena have also been reported in c/c′ alloys and attributed to static recrystallization under critical stored energy conditions [24]. Obviously, this approach does not pretend to be representative of a real microstructure and process. However, it can give useful guidelines for understanding unexpected microstructure evolution occurring during heat treatments,

Advanced Modeling Tools for Processing and Lifing …

7

From a purely computational point of view, we have to address three main challenges:

Fig. 5 Modeling of grain growth with DIGIµTM on a theoretical microstructure on IN718 (extracted from [22])

on a rather wide range of materials (IN718, c/c′ alloys, etc.) and different processes (ring rolling, bar cogging, etc.). For a given initial microstructure in terms of grain size and secondary phase particles, it is then possible to assess the distribution of stored energy which can trigger heterogeneous grain growth. This phenomenon is more sensitive when grain size decreases due to the higher capillarity effect (typically from grain size finer than 10 µm). However, the link between thermomechanical history and a specific distribution of stored energy between neighboring grains leading to heterogeneous growth is not always so easy to establish.

Upstream Process: Ingot Conversion Another challenge consists of microstructure prediction during ingot conversion and billet manufacturing [25]. In order to meet the final part requirements, microstructure needs to be controlled at the earlier stages of the product and typically on billets. For c/c′ alloys such as Rene 65 or AD730TM, it is well known that fine and homogeneous grain size must be obtained directly at this step since recrystallization of large and elongated unrecrystallized grains cannot be completed during closed die forging [26, 27]. Thus, the trend is toward modeling of the full process, from ingot to the final part, as shown in Fig. 6.

– As open die forging consists of a large number of various operations (upsetting, bar forging, furnace reheating, etc.), we need to gather a lot of process parameters for accurate modeling of hundreds of press strokes. Improvement of data recording and more continuous communication between the press and computational codes is an important way to move forward [28]. – Regarding microstructure prediction models, specific adaptations have to be implemented in order to take into account all the phenomena occurring simultaneously at different locations of the bars (dynamic, metadynamic, or static recrystallization). Moreover, as these models are usually built from equiaxed initial microstructure, specific attention has to be paid to the evolution of as-cast microstructure during the early stages of the process conversion. – Lastly, ingot conversion involves several reheating steps in furnaces. In several cases, we do not necessarily look after a full recovery of a homogeneous temperature distribution within the bars. During some processes, we can have some beneficial effect of a progressive temperature decrease at the core of the bar during the whole forging process, leading to microstructure refinement. As shown in Fig. 7 in the case of IN718, heating rate should be in some areas fast enough in order to avoid massive and detrimental precipitation. For all these reasons, an accurate prediction of thermal history in furnaces is also a key point to have an excellent microstructure control. Special attention has to be paid to this part of the process through specific modeling tools, dedicated to furnace behavior. Taking into account all these considerations, a rather accurate prediction of microstructure evolution during ingot conversion can be obtained. For example, forging route optimization on IN718 for larger diameter bars up to 356 mm has been proposed.

Modeling Capability Versus Alloys Finally, for processes, from ingot to forged parts, modeling including microstructure prediction can be achieved on IN718. However, additional development needs to be carried out for ring rolling for two main reasons. First, the reliability of thermomechanical prediction must be improved for this complex process: pronounced spatial strain gradient, true displacement rate of the press ram, etc. An exemplary result obtained is shown in Fig. 8. When the press reached its

8

A. Longuet et al.

Fig. 6 Flowchart for microstructure modeling from ingot to closed die forging on IN718

Fig. 7 Temperature evolution close to a bar surface during cogging and reheating—consequences on microstructures for IN718 alloy

power limit, this last parameter can significantly decrease due to the pressure loss in the hydraulic circuits. Thus, if the actual behavior of the press is not correctly taken into account, forging time will be in that case largely underestimated, leading to a cascade of incorrect predictions (temperature distribution, load, and microstructures). Moreover, some specific conditions make identification of parameters for microstructure prediction more challenging, including high strain rates (up to 30 s−1) and low strain for each pass (lower than 0.1). Of course, these models must take into account precipitation which is involved in controlling grain growth by Smith–Zener pinning effects after primary recrystallization: d phase for IN718, primary c′ particles for Rene 65, Udimet 720Li, or AD730TM alloys. However, rather slight interactions have been observed between d phase and recrystallization for IN718:

Fig. 8 Comparison between real (dashed line) and predicted (solid line) load (red) and ram speed (blue) during closed die forging of a turbine gas disk in Inconel 706

– Accelerated dissolution during hot deformation above the solvus temperature as mentioned above – Local segregation of niobium, leading to local variation of d solvus, and consequently local variation in grain size – Increase of nucleation rate during dynamic recrystallization – Large amounts of d phase can slow down metadynamic recrystallization [17]. Even if these interactions can lead to some changes in microstructure, both mechanisms (primary recrystallization and d precipitation) can be studied separately in a first approach for IN718. For c/c′ alloys, the situation is more challenging, especially during cogging, but more specifically during the first deformation steps below the c′ solvus. For example, a fine and coherent primary precipitation can be

Advanced Modeling Tools for Processing and Lifing …

Fig. 9 Back-scattered electron (BSE) image (a) and orientation color-coded electron back-scattered diffraction (EBSD) map (b) with grain boundaries (>15°) plotted black of a hetero-epitaxially recrystallized grain in the Rene 65 alloy. BSE image at a recrystallization front in the AD730TM alloy (c). BSE image in the longitudinal section of an AD730TM alloy billet (d). Orientation color-coded EBSD map (e) with grain boundaries (>15°) plotted black and twin boundaries plotted white of an overgrown grain with twin-related c′ precipitates (arrowed in the zoomed insert) in the AD730TM alloy

observed in coarse, elongated, and unrecrystallized grains [27]. They can be considered as hard grains during hot deformation, surrounded by soft material, corresponding to fine recrystallized grains (an example of such a microstructure can be seen in Fig. 9d). Therefore, the large grains cannot accumulate enough strain hardening and stored energy in order to advance the recrystallization process. Examinations at different steps of billet conversion show that these large

9

grains are inherited from the early stages without any significant evolution. Therefore, significant increase of deformation is less efficient in order to improve recrystallization fraction [26]. The best solution is then to act on precipitation. Figure 9 shows other examples of recently reported mechanisms which are specific to c/c′ alloys. Hetero-epitaxial recrystallization (HEREX—Fig. 9a and b) is a mechanism by which a recrystallized grain arises from inverse precipitation of c phase at the rim of a c′ precipitate and subsequent growth driven by stored energy consumption [29–31]. The striking feature of HEREX grains is that they have the same crystallographic orientation as the precipitate they originate from, which led to the terminology of hetero-epitaxial recrystallization. Figure 9c shows the complex mechanism at play when a recrystallization front migrates at sub-solvus temperatures, thus in a microstructure with c′ precipitates [32]. The proposed mechanism proceeds by dissolution of the deformed grain precipitates at the recrystallization front, followed by re-precipitation on the other side in the recrystallized grain, the whole process keeping the precipitates coherent with the matrix grains on both sides. The classical Smith–Zener pinning mechanism and model usually considered in recrystallization and grain growth simulations are far from being sufficient in such a mechanism. The last example concerns overgrown grains (Fig. 9e) which develop specifically in elongated recovered grains which can be found in forged billets when the billet conversion route was not optimized enough to fully recrystallize the microstructure [26, 27]. Such an elongated recovered grain can be seen on the left side of Fig. 9d, and the c′ precipitate size is much finer in those areas than it is in the recrystallized equiaxed grains. The mechanism by which those overgrown grains can develop is also driven by stored energy consumption like that of Fig. 5, but it goes along with a dissolution and re-precipitation mechanism at the recrystallization front which leads to c′ precipitates with particular shape and orientation (notably twin relationship with the overgrown grain, as highlighted in the zoomed insert of Fig. 9e). This only occurs if the recrystallizing grain satisfies the condition of being misoriented about a 〈111〉axis with the recovered grain it grows into [33, 34]. Reliable microstructure prediction for c/c′ alloy forgings will definitely require at least the implementation of the coupling between recrystallization and phase transformation phenomena into the metallurgical models, and likely also the consideration of the local crystallographic texture and orientation relationships. Due to the strong interaction between precipitation and recrystallization, modeling microstructure evolution during forging is by far more complex for these alloys than for IN718 and no suitable tool is available up to now. Knowledge improvement in c′ precipitation

10

morphology, size distribution, etc., is still necessary, with dependence on many parameters (cooling and heating rate, strain hardening, etc.).

Residual Stress and Mechanical Properties Beyond microstructure prediction during forging, the next step for modeling deals with final heat treatments and mechanical properties. For IN718, three main topics are involved in these investigations: – Cooling rate modeling after annealing – Constitutive laws for mechanical behavior – Validation by residual stress measurements. Such simulations have already been presented previously [35, 36]. For this alloy, a rather high level of maturity and reliability has been achieved, with accurate characterization of the relaxation behavior [37], except for the direct aged version of the alloy for which work remains to be done. For this specific case, strain hardening inherited from forging must be taken into account for residual stress predictions with specific development, such as taking into account how the thermomechanical (forging) history affects subsequent c′/ c″ precipitation during the precipitation heat treatment. By comparing results obtained on mechanical properties and residual stress, we can specify what could be the best compromise between both kinds of properties, depending on customer requirements [35], as shown in Fig. 10. Fig. 10 Heat treatment design and optimization through the compromise between mechanical properties and residual stress for Inconel 718

A. Longuet et al.

For each specific case, master curves giving the evolution of mechanical properties versus the magnitude of residual stresses can be established in order to specify what are the best heat treatment conditions and the more appropriate quenching media. Finally, these tools are very helpful for component design, one trend in place now in concurrent development between the supply chain and the end-user. As for microstructure modeling (see previous subsection), our knowledge and capability are less advanced for c/c′ alloys. Equivalent methodologies can be used. However, as shown in Fig. 11, two main differences can be raised: – Very high level of residual stress after quenching due to the specific thermomechanical behavior of this class of alloys, most of the hardening precipitation occurring during cooling from the solution heat treatment – Rather poor residual stress relief during aging. For this reason, slower quenching conditions must be considered while mechanisms of stress relief during aging are not still clearly understood, depending on metallurgical state (c′ size distribution) and mechanical state (level of stress) just after quenching. Such characterizations are presently in progress for c/c′ alloys [38, 39]. A typical example for AD730TM alloy is shown in Fig. 12. To better understand the intrinsic relaxation behavior of c/c′ alloys and to feed accurate constitutive equations for simulations as the ones presented in Fig. 11, relaxation tests at different temperatures and initial maximum stresses and microstructure are performed. As a striking result from Fig. 12, one can

Advanced Modeling Tools for Processing and Lifing …

11

Fig. 11 Residual stress distribution measured on laboratory samples in AD730 by contour method: a on as quenched conditions (annealed at 1080 °C/4 h–15 s transfer time and polymer quenching) and b after aging (760 °C/8 h)

observe a weak stress relaxation at 730 °C for durations well beyond classical ones used for aging heat treatments (i.e., typically from 4 to 20 h for these alloys). Moreover, by using single-crystalline specimens of the polycrystalline version of the alloy, it unambiguously shows a contribution of grain boundaries to the relaxation (creep) mechanisms [39], which are in good agreement with previous creep results from Thébaud et al. [40]. Additional investigations must be performed in order to obtain a fully predictive model. Beyond the control and validation of manufacturing routes, all these data are necessary for original engine manufacturers to predict the mechanical behavior during component service, including distortion during final machining and lifing. An adapted aging heat treatment may have to be found to ensure machinability of the components and further improve in-service properties.

Use of Pre-spinning to Address Residual Stress Concerns It has been shown in the previous subsection that it is not always possible to lower residual stresses with thermal treatments, especially for c/c′ alloys due to the almost immediate precipitation upon cooling after solution heat treatment. The main effect is seen when the disk is machined

to its final shape. For large components like high-pressure turbine disks, it is not an issue, because the residual strains remain low with respect to the size of the part. However, for low-pressure turbine disks, which are very thin and larger in diameter, it is difficult to achieve the prescribed dimensions during further machining, due to residual stress redistribution during material removal. One way to avoid this effect is to pre-spin the disks up to a certain rotation speed or to a diameter growth. Two useful strategies are available to define the criterion: trial and error and process modeling, as shown in Fig. 13. First an elastoplastic model is computed to predict the growth/swelling of the component as a function of the rotation speed. Then, a few tests are performed to validate the model. Unfortunately, the first results are not always sufficiently precise and the model would need adjustments. When the prescribed machining dimensions are reached, while limiting the plastic strain during pre-spinning, the parameter set is fixed. Another advantage of using this process is to avoid the initial increase of the disk diameter during the first few engine cycles due to material hardening. On the other hand, if the plastic strain generated by the pre-spinning is too high, material properties could be reduced.

Advanced Modeling Tools for Lifing of Turbine Disks Usual lifing methods for rotating components use the “hotspot method” applied to finite element simulations. This means that a life criterion is computed from stress tensor and

Fig. 12 Stress relaxation evolutions in AD730TM alloy at 730 °C comparing a sub-solvus fine grain (FG) material and its equivalent 〈001〉 oriented single crystal (SX 〈001〉). Results from [38, 39]

Fig. 13 Effect of pre-spinning on residual stresses of low-pressure turbine disk

12

A. Longuet et al.

then compared in each point of the structure to the S–N diagram in order to compute the fatigue life of the structure. It is well known that the life of a notched specimen is higher than the life of a plain specimen (see Fig. 14) at same iso-local von Mises stress. Multiple mechanical approaches have been designed to take this phenomenon into account. Among these, one can cite critical distance [41], stress averaging [42], and modification of the life criterion with the introduction of the stress gradient [43]. But usually these methods are only valid in a narrow set of conditions. Another way is to consider the effect of the stressed volume. Fatigue life is usually controlled by a weakest link mechanism. This means that when a first constituent of the material breaks then the whole part is broken (see Eq. 1). Introducing this principle, a volume effect is shown in Fig. 15. In fact, the larger the specimen is, the higher is the probability to contain a critical crack initiation site within the stressed volume (i.e., gauge part of a fatigue specimen). Y   1  Pstruct ðNa; rÞ ¼ 1  Pea lem ðNa; rÞ ð1Þ a e lemts

The size effect model has been widely used to take into account high-cycle fatigue life of notched specimen [44]. Fatigue life to failure can be divided into two main parts: life to crack initiation and crack growth propagation life. The crack initiation part is the only one to be subjected to a size effect since when one (or many) crack(s) are present, the crack propagation is deterministic. Even if there is an effect of stress gradient and size on crack propagation life, it is not linked to a scale/size effect. The stress gradient and the size of the crack propagation path change the stress intensity factor and the end of life criterion. In order to define the limit between initiation and propagation, a crack size is needed. Figure 16 shows how to compute crack propagation life on low-cycle fatigue specimens. For each fatigue test performed, only life to failure is stated. Specific fatigue tests

Fig. 14 Notch effect of IN718 and effect of initiation (“at bulk”)

Fig. 15 Example of weakest link application on a structure with multiple elements

with thermal marking have been performed to validate crack propagation modeling up to a0 crack size [45, 46]. As it is difficult to model crack propagation up to 30 µm (average initiation size), a0 is taken as a parameter of the model. This analysis allows us to determine life to initiation of a crack of a0 size. A Weibull distribution is used to model life distribution for each pseudo-stress level (see Eq. 2):  mðr;TÞ ! V N e lem Pa ðNa; r; VÞ ¼ 1  exp  lnð2Þ Vref N0 ðr; TÞ ð2Þ with Vref the volume of the plain specimen, V the current volume, N0 the mean life to initiation, and m the scatter parameter. N0 and m are functions of pseudo-stress and temperature to allow the complete description of domain. In this study, only surface initiation is considered, so the reference volume has to be changed to the one attached to the surface in order to be consistent. The same equation could have been written with areas instead of volumes.

Fig. 16 Crack propagation life analysis

Advanced Modeling Tools for Processing and Lifing …

In order to validate the model, a life prediction from notched specimens has to be performed. The implementation of the model is direct in finite element codes: For each element, a survival probability is computed with the local temperature, stress, and volume of the element. Then, a sum is done all over the structure. This can be done for different levels of probability: 50% for average and 0.0135% for an equivalent −3r log-normal distribution probability, with r being the standard deviation. To be able to compare notch specimen life to rupture in the model, the crack growth propagation life must be computed on notched specimens. Figure 17 shows very good agreement between the tests and the curve considering the few specimens tested. The part application is as simple as it is on the specimen except that the evolution of stress regarding time is more complex, given real service conditions. In Eq. (2), r is taken as an equivalent stress. This means that the same method is used: same rainflow-counting algorithm (i.e., a method to reduce/simplify the spectrum of varying stress), same R-ratio effect and same multiaxiality effect. Finally, the multiplicity of the area must be taken into account to determine the life of the part. This method shows that a substantial life increase can be computed with this method if the solicited volume (volume with a local equivalent stress larger than, e.g., 90% of the maximum local stress magnitude) of the part is low enough. It also shows that the direct comparison between a notched specimen and a part can be difficult if the involved volumes are very different. To be able to consider shot peening applied to components, the same modeling has to be done for cases involving internal crack initiation. Moreover, the way residual stresses inherited from shot peening evolve during service has also to be taken into account.

Fig. 17 Comparison between notched specimen experimental life to rupture and prediction

13

Conclusions and Prospects This article shows that in order to model the life of components, one has to take into account metallurgical considerations regarding crack initiation. The example of grain size effect on fatigue life of IN718 is very interesting because it shows that if two specimens are tested at different strain levels, then two different conclusions may arise. One will show a life debit, whereas the other does not. Therefore, it is critical to understand on which metallurgical feature crack initiation occurs as a function of temperature, stress level, etc., to be able to model the effect of grain size on life and to design the right test plan. To achieve this goal, the mechanical engineer and metallurgist must work together. The actual challenge is to be able to transpose the results obtained on specimens to the parts. It is shown that the use of the “hotspot” method for life determination is highly conservative. A way to reduce it is to understand the different phenomena at stake. Life can be divided into crack initiation and crack propagation phases. Each one must be modeled in the case of stress gradient to improve life computation of components. The modeling of residual stresses is also very important to predict life and final geometry of forged components like disks/rings. Therefore, such an effort has been made on process modeling. Historically, the first steps for advanced modeling in component manufacturing started with microstructure prediction during closed die forging of disks in Inconel 718 or Waspaloy, by means of purely phenomenological models. Progressively, this approach has been extended toward three different directions: – Introduction of more complex tools (mean-field models or full-field software) in order to have a better understanding and prediction of specific phenomena occurring during closed die forging, such as abnormal grain growth, recovery effects between several processing steps with complex thermal histories (cooling and reheating in furnaces) – Modeling development for manufacturing processes beyond closed die forging modeling for upstream processes (ingot conversion and microstructure prediction) and downstream processes (heat treatment and quest for the best compromise between highest mechanical properties and lowest residual stress distribution) – Adding new materials for these approaches, especially c/c′ alloys such as Rene 65, Udimet 720Li, or AD730TM. In this case, more complex phenomena have to be taken into account and more clearly understand the interaction between c′ precipitates and recrystallization mechanisms,

14

A. Longuet et al.

kinetics of stress relaxation during aging depending on initial state of stress or on initial c′ size distribution. For these new investigations, metallurgical questions are not the only challenge we have to face. A good prediction of the thermomechanical history is obviously required: cooling rates during quenching for c/c′ alloys with specific equipment (air blown, gas pressure furnace, etc.) or strain and temperature distribution during multi-steps and complex processes such as open die forging. Improved accuracy can then be achieved with specific tools: CFD modeling for heat treatments, fully instrumented press and manipulator for forging, etc. As mentioned above, heat treatment and residual stress control are still a challenging question leading to several research programs in academic laboratories and industry. In the same way, many teams are involved in the development of mean-field or full-field models for microstructure prediction. As we can predict a grain size distribution instead of an average value, these models are of great interest, as damage can initiate usually on the larger grains (fatigue properties) or on the finer ones (creep properties at high temperatures). These tools open new fields of investigation for specific applications such as dual microstructure disks [47–51] or for a better understanding of abnormal grain growth [22, 52]. For this last point and thanks to these new approaches, some basic explanations have been raised now for a better understanding of the root cause for heterogeneous or multimodal grain size distribution. However, the capability for predicting and for preventing this on real processes for disk manufacturing remains challenging and open to a large field of new investigations. Finally, we have identified two main topics with too few investigations up to now. The first one deals with the ring rolling process, whether it is the reliability of the thermomechanical history or the prediction and the consequences on microstructure evolution. Secondly, the knowledge on the interaction between the microstructure and its response to the ultrasonic inspection should be improved in order to minimize the background noise and to optimize the capability to detect the smallest potential defect. Acknowledgements The authors wish to thank Pr. Nathalie Bozzolo and Pr. Marc Bernacki both from CEMEF (Ecole des Mines de Paris, Sophia Antipolis, France) and Dr. Jonathan Cormier from ISAE-ENSMA (Institut Pprime, Poitiers, France) for their contributions, review, valuable advice, and fruitful discussions.

References 1. Tin S, Lee P, Kermanpur A et al (2005) Integrated modeling for the manufacture of Ni-based superalloy discs from solidification to final heat treatment. Met. Mater. Trans. A 36(9):2493–2504.

2. Kermanpur A, Lee P, Tin S et al (2005) Integrated model for tracking defects through full manufacturing route of aerospace discs. Mater. Sci. Tech. 21(4):437–444. 3. Zouari M, Bozzolo N, Loge RE (2016) Mean field modelling of dynamic and post-dynamic recrystallization during hot deformation of Inconel 718 in the absence of d phase particles. Mater. Sci. Eng. A 655:408–424. 4. Chaire DIGIMU, https://chaire-digimu.cemef.mines-paristech.fr/. 5. Chaire OPALE, https://chaire-opale.cemef.mines-paristech.fr/. 6. Chaire TOPAZE, https://chaire-topaze.cemef.mines-paristech.fr/. 7. Pineau A, Antolovich SD (2009) High temperature fatigue of nickel-base superalloys–a review with special emphasis on deformation modes and oxidation. Eng. Fail. Anal. 16(8): 2668–2697. 8. Abikchi M (2012) Effets des microstructures induites par le procédé de forgeage sur la durée de vie de pièces en Inconel 718 DA, Ph.D Thesis, Ecole Nationale Supérieure des Mines de Paris: Paris, France. 9. Govaère A, Rouffié AL, Franchet JM et al (2020) Is the Carbon Content Really an Issue for the LCF Durability of Forged c/c′ Ni-based Disk Alloys? 14th International Symposium on Superalloys 2020, Accepted for publication. 10. Texier D, Casanova-Gomez A, Pierret S et al (2016) Microstructural features controlling the variability in low cycle fatigue properties of alloy Inconel 718DA at intermediate temperature. Met. Mat. Trans. A 47:1096–1109. 11. Texier D, Cormier J, Villechaise P et al (2016) Crack initiation sensitivity of wrought direct aged alloy 718 in the very high cycle fatigue regime: the role of non-metallic inclusions. Mat. Sci. Eng. A 678:122–136. 12. Texier D, Stinville JC, Echlin MP et al (2019) Short crack propagation from cracked non-metallic inclusions in a Ni-based polycrystalline superalloy. Acta Mater. 165:241–258. 13. Texier D, Stinville JC, Charpagne MA et al (2020) Role of Non-Metallic Inclusions and Twins on the Variability in Fatigue Life in Alloy 718 Nickel Base Superalloy. 14th International Symposium on Superalloys 2020, Accepted for publication. 14. Shen G, Semiatin S, Shivpuri R (1995) Modeling microstructural development during the forging of Waspaloy. Met. Mater. Trans. A 26:1795–1803. 15. Dumont C, Georges E (2011) Microstructure prediction during incremental processes for hot forming of 718 alloy. Adv. Mater. Res. 278:186–191. 16. Giraud R, Hervier Z, Cormier J et al (2013) Strain effect on the c′ dissolution at very high temperatures of a nickel-based single crystal superalloy. Met. Mater. Trans. A 44:131–146. 17. Thomas JP, Bauchet E, Dumont C et al (2004) EBSD investigation and modeling of the microstructural evolutions of superalloy 718 during hot deformation. Paper presented at Superalloys 2004, TMS: Warrendale, PA, USA, 959–968. 18. Maire L, Scholtes B, Moussa C et al (2017) Modeling of dynamic and post-dynamic recrystallization by coupling a full field approach to phenomenological laws. Mat. & Des. 133: 498–519. 19. Bernacki M, Bozzolo N, De Micheli P et al (2019) Numerical Modeling of Recrystallization in a Level Set Finite Element Framework for Application to Industrial Processes, In Recrystallization: Types, Techniques and Applications, 1st edition, Nova Science Publishers, Inc. 20. Transvalor, 2D & 3D microstructure prediction with DIGIMU® https://www.youtube.com/watch?v=BFo9hz8Pwj4&list=PLM66D 2Ym7FYziu57vJiKV5Wf-PXW-rgDK&index=2. 2019. 21. Transvalor, https://www.transvalor.com/en/cmspages/digimu.86. html. 22. Agnoli A (2013) Origin of Inhomogeneous Grain Growth in Inconel 718 Forging, Ph.D Thesis, Ecole Nationale Supérieure des Mines de Paris: Sophia Antipolis, France.

Advanced Modeling Tools for Processing and Lifing … 23. Flageolet B, Yousfi O, Dahan Y et al (2010) Characterization of microstructures containing abnormal grain growth zones in 718 alloy. Paper presented at the 7th International Symposium on Superalloy 718 and Derivatives, TMS: Warrendale, PA, USA, 595–606. 24. Charpagne MA, Franchet JM, Bozzolo N (2018), Overgrown grains appearing during sub-solvus heat treatment in a polycrystalline c-c′ Nickel-based superalloy. Mat. & Des. 144:353–360. 25. Antolovich B, Evans M (2000) Predicting grain size evolution of UDIMET alloy 718 during the” cogging” process through the use of numerical analysis. Paper presented at Superalloys 2000, TMS: Warrendale, PA, USA, 39–47. 26. Perez M, Dumont C, Nodin O et al (2018) Impact of forging direction on the recrystallization behaviour of nickel base superalloy AD730 billet material at subsolvus temperatures. Mat. Charact. 146:169–181. 27. Vernier S (2019) Microstructure Evolution of the Nickel Based Superalloy AD730TM during Industrial Forgings, Ph.D Thesis, Ecole Nationale Supérieure des Mines de Paris: Sophia Antipolis, France. 28. Brochure on FERROTRON® and LACAM® system, Minerals Technologies International: Easton PA 18042, USA. 29. Charpagne MA, Vennegues P, Billot T et al (2016) Evidence of multimicrometric coherent c′ precipitates in a hot-forged c–c′ nickel-based superalloy. J. Microscopy 263:106–112. 30. Charpagne MA, Billot T, Franchet JM et al (2016) Heteroepitaxial recrystallization: A new mechanism discovered in a polycrystalline c-c′ nickel based superalloy. J. All. Comp. 688:685–694. 31. Charpagne MA, Billot T, Franchet JM et al (2016). Heteroepitaxial recrystallization observed in Ren 65TM and Udimet 720TM: a new recrystallization mechanism possibly occuring in all low lattice mismatch c-c′ superalloys? in Superalloys 2016. Paper presented at Superalloys 2016, TMS: Warrendale, PA, USA, 417–426. 32. Seret A, Moussa C, Bernacki M et al (2018) On the coupling between recrystallization and precipitation following hot deformation in a c-c′ nickel-based superalloy. Met. Mater. Trans. A 49:4199–4213. 33. Vernier S, Franchet JM, Dumont C et al (2018) A Mechanism Leading to c′ Precipitates with {111} Facets and Unusual Orientation Relationships to the Matrix in c–c′ Nickel-Based Superalloys. Met. Mater. Trans. A 49:4308–4323. 34. Vernier S, Franchet JM, Dumont C et al (2018) c′ precipitates with a twin orientation relationship to their hosting grain in a c-c′ nickel-based superalloy. Scr. Mater. 153:10–13. 35. Dahan Y, Nouveau S, Georges A et al (2014) Residual Stresses in Inconel 718 Disks. Paper presented at Eurosuperalloys 2014, Matec Web of Conferences: Presqu’île de Giens, France, 10003. 36. Gostic W (2012) Application of Materials and Process Modeling to the Design, Development and Sustainment of Advanced Turbine Engines. Paper presented at Superalloys 2012, TMS: Warrendale, PA, USA, 1–12. 37. Semiatin S, Fagin P, Goetz R et al (2019) Effect of Test Method on Stress-Relaxation Behavior of Alloy 718. Met. Mater. Trans. A 50:1397–1408.

15 38. Durand M (2020) Stress relaxation mechanisms in Ni-based polycrystalline disk superalloys during aging heat treatments, Ph. D. Thesis, Ecole Nationale Supérieure des Mines de Paris: Sophia Antipolis, France. 39. Durand M, Cormier J, Villechaise P et al (2020) Metallurgical mechanisms upon stress relaxation annealing of the AD730TM superalloy. 14th International Symposium on Superalloys 2020, Accepted for publication. 40. Thébaud L, Villechaise P, Crozet C et al (2018) Is there an optimal grain size for creep resistance in Ni-based disk superalloys? Mater. Sci. Eng. A 716:274–283. 41. Taylor D (2007) The theory of critical distances: a new perspective in fracture mechanics. Elsevier, Oxford, UK. 42. Pijaudier-Cabot G, Bažant ZP (1987) Nonlocal damage theory. J. Eng. Mech., 113:1512–1533. 43. Nadot Y, Billaudeau T (2006) Multiaxial fatigue limit criterion for defective materials. Eng. Fract. Mech. 73:112–133. 44. Vayssette B, Saintier N, Brugger C et al (2019) Numerical modelling of surface roughness effect on the fatigue behavior of Ti-6Al-4 V obtained by additive manufacturing. Int. J. Fat. 123:180–195. 45. Dorémus L, Nadot Y, Hénaff G et al (2015) Calibration of the potential drop method for monitoring small crack growth from surface anomalies–Crack front marking technique and finite element simulations. Int. J. Fat. 70:178–185. 46. Dorémus L, Cormier J, Villechaise P et al (2015) Influence of residual stresses on the fatigue crack growth from surface anomalies in a nickel-based superalloy. Mat. Sci. Eng. A 644:234–246. 47. Taboada Michel H, Sasaki Reda L, Effgen Santos G et al (2016) Mechanical properties of cast & wrought hybrid disks. Paper presented at Superalloys 2016, TMS: Warrendale, PA, USA, 539– 548. 48. Panella M, Signor L, Cormier J et al (2020) Experimental and Simulation Study of the Effect of Precipitation Distribution and Grain Size on the AD730TM Ni-Based Polycrystalline Superalloy Tensile Behavior. 14th International Symposium on Superalloys 2020, Accepted for publication. 49. Mitchell R, Lemsky J, Ramanathan R et al (2008) Process development & microstructure & mechanical property evaluation of a dual microstructure heat treated advanced nickel disc alloy. Paper presented at Superalloys 2008, TMS: Warrendale, PA, USA, 347–356. 50. Gabb T, Kantzos P, Telesman J et al (2011) Fatigue resistance of the grain size transition zone in a dual microstructure superalloy disk. Int. J. Fat. 33:414–426. 51. Gayda J, Gabb T, Kantzos P The effect of dual microstructure heat treatment on an advanced nickel-base disk alloy. Paper presented at Superalloys 2004, TMS: Warrendale, PA, USA, 323–329. 52. Agnoli A, Bernacki M, Logé R et al (2015), Selective growth of low stored energy grains during d sub-solvus annealing in the Inconel 718 nickel-based superalloy. Met. Mater. Trans A 46:4405–4421.

Part II Alloy Development

Developing Alloy Compositions for Future High Temperature Disk Rotors M. C. Hardy, C. Argyrakis, H. S. Kitaguchi, A. S. Wilson, R. C. Buckingham, K. Severs, S. Yu, C. Jackson, E. J. Pickering, S. C. H. Llewelyn, C. Papadaki, K. A. Christofidou, P. M. Mignanelli, A. Evans, D. J. Child, H. Y. Li, N. G. Jones, C. M. F. Rae, P. Bowen, and H. J. Stone

Abstract

Two new alloy compositions for possible disk rotor applications have been examined. Both were intended to have higher c0 content than the existing alloy, RR1000, and be produced using powder metallurgy and isothermal forging to enable forgings to show a consistent coarse grain microstructure. Small pancake forgings of the new alloys and RR1000 were made and from these, blanks were cut, solution heat treated, cooled at measured rates and aged. Results of screening tests to understand the tensile, creep and dwell crack growth behavior, oxidation resistance and phase stability of these new alloys and coarse grain RR1000 are reported. The development alloys were similar in composition but exhibited different

M. C. Hardy (&)  C. Argyrakis  R. C. Buckingham  P. M. Mignanelli  D. J. Child Rolls-Royce Plc, Derby, UK e-mail: [email protected] A. S. Wilson  K. A. Christofidou  N. G. Jones  C. M. F. Rae  H. J. Stone Department of Materials Science and Metallurgy, University of Cambridge, Cambridge, UK H. S. Kitaguchi  S. Yu  C. Jackson  S. C. H. Llewelyn  H. Y. Li  P. Bowen School of Materials and Metallurgy, University of Birmingham, Birmingham, UK K. Severs Allegheny Technologies Incorporated (ATI) Forged Products Cudahy, Cudahy, WI, USA A. Evans Bundesanstalt Für Materialforschung Und –Prüfung (BAM), Berlin, Germany E. J. Pickering Department of Materials, University of Manchester, Manchester, UK C. Papadaki Department of Engineering Science, University of Oxford, Oxford, UK

tensile and creep properties, phase stability and resistance to oxidation damage. Despite attempts to minimize variation in microstructure from heat treatment, differences in c0 size distribution were found to influence tensile and creep behavior. One of the new alloys (Alloy 2) showed improved yield and tensile strength compared to RR1000. Alloy 2 displayed similar initial creep strain behavior to RR1000 but superior resistance to subsequent creep damage, producing longer creep rupture lives. All of the alloys showed crack retardation at low stress intensity factor ranges (DK) from 3600 s dwell cycles at 700 °C in air. This occurred whilst crack growth was intergranular. Alloy 1 was found to precipitate C14 Laves phase from long term exposure at 800 °C. Like RR1000, r phase was not detected in the new alloys after 750 h at 800 °C. Keywords

Powder metallurgy Material properties



Phase stability



CALPHAD



Introduction High bypass ratio turbofan aircraft engines and operating cycles are continuously evolving to provide improved efficiencies for reduced fuel consumption and emissions [1, 2]. However, whilst propulsive and aerodynamic optimizations of aircraft engines are possible, the increased demands upon superalloys, which are used in the hot section parts, limit the thermal efficiency improvements that can be achieved. The requirements for reduced engine core sizes and increased temperatures and stresses pose a complex set of seemingly conflicting property requirements for the materials considered for safety-critical disk rotor applications. Specifically, materials with higher strength levels are needed to reduce the size and weight of components. Whilst this necessitates the

© Copyright Rolls-Royce plc 2020. All rights reserved. S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_2

19

20

M. C. Hardy et al.

development of compositions with increased amounts of the gamma prime ðc0 Þ phase, further optimization is possible by using a fine grain size. Yet such grain structures produce less appealing time dependent crack growth behavior [4], which may limit the design life of the component or the interval between inspections. This is more relevant in today’s engines as high climb rates are increasingly required by commercial airlines to move aircraft more quickly to altitude to reduce fuel burn [5]. Therefore acceptable strength is required from coarse grain microstructures, which demands effective precipitation strengthening from alloy design and control of grain size in near net shaped forgings. Inevitably, this is only possible using powder metallurgy to minimize elemental segregation to length scales of a micron (lm) or less for these complex, multi-component alloys with high levels of reactive elements (Al, Ti, Ta etc.) [6–8]. Subsequent hot deformation of consolidated powder compacts produces billet material with extremely fine grains, which enables superplastic flow of the work piece during isothermal forging at high temperatures and low strain rates, to make the desired near net disc shapes [6–8]. A uniform average grain size of 20–40 lm can then be created by super-solvus solution heat treatment. This microstructure produces an ideal balance in material properties between tensile strength and resistance to time dependent crack growth. Otherwise, optimization of strength in the hub of the disc and resistance to time dependent crack growth and creep in the rim and diaphragm can be achieved by producing dual microstructure forgings, albeit with greater design and manufacturing complexity as well as increased cost [4]. To achieve the desired material properties, careful consideration is required of the volume fraction, number density, size and morphology of c0 precipitates, as well as the heat treatments and rates of cooling that produce these. Alloy design and development is further complicated by minor grain boundary phases, surface degradation from oxidation and type II hot corrosion, coarsening and dissolution of c0 particles and possible precipitation of detrimental topologically close packed (TCP) phases as a result of long duration exposures at temperatures above 700 °C. This paper discusses the design of 2 development alloys and evaluates their microstructure and material properties, comparing them against an existing Rolls-Royce alloy, RR1000 [9].

Alloy Design The first priority in designing the new alloys was to achieve the required yield strength in a coarse grain microstructure. Of the strengthening mechanisms that give rise to the performance of nickel base superalloys, precipitation hardening is the most significant. Strengthening occurs as a result of creating fault energies from anti-phase boundaries and stacking faults when paired dislocations penetrate c0 particles. It is understood then that a new alloy should contain a larger number density of (preferably smaller) c0 particles, compared to existing alloys to provide a higher level of precipitation hardening. This understanding is formalized in models that are available in the literature [10–14]. These correlate the critical resolved shear stress or flow stress of the alloy with the volume fraction and size of c0 particles, and the anti-phase boundary (APB) energy. In practice, the new alloys in this study (Table 1) were designed by specifying a volume fraction of c0 of about 50–53%. It was considered that this range should produce sufficient levels of improvement over the existing alloy, RR1000, which has 45% c0 , but without incurring significant detriments to time dependent crack growth behavior or difficulties in raw material and component manufacture. As c0 is described by Ni3X, where X is predominantly Al with progressively smaller proportions of Ti, Ta and Nb, the specified volume fraction of c0 was defined according to: 13:15 atomic% [ Al þ Ti þ Ta þ Nb [ 12:65 atomic% ð1Þ and elemental ranges of 6.55 to 7.15 at. % for Al, 3.3 to 3.7 at. % for Ti, 1.2 to 1.7 at. % for Ta and 0.8 to 1.0 at. % for Nb. These ranges were determined [15–17] by considering the effect of these elements on (i) strengthening potential (or APB energy), (ii) the c0 dissolution or solvus temperature (Tsolvus), (iii) the propensity for eta (η) formation, (iv) the stability of primary MC carbides, (v) oxidation resistance (notably Ti), and (vi) dwell crack growth behavior (notably Nb). The first 4 of these factors (i–iv) were assessed using phase diagram modelling [18–20] and Thermo-Calc databases TCNi6 and 7. These were the current databases when the development alloys were designed. The work of Crudden et al. [13] was also influential in optimizing strengthening

Table 1 Nominal compositions (in atomic percent) of development alloys [15–17] and RR1000 [9] at.%

Ni

Co

Cr

Mo

W

Fe

Mn

Al

Ti

Ta

Nb

Si

Hf

B

C

Zr

Alloy 1

Bal.

15.0

14.0

1.3

1.0

1.0

0.6

7.0

3.5

1.6

0.9

0.90

0

0.14

0.15

0.05

Alloy 2

Bal.

15.6

14.0

1.4

1.1

1.0

0

6.8

3.5

1.6

0.9

0.32

0

0.14

0.15

0.06

RR1000

Bal.

17.9

16.5

3.0

0

0

0

6.4

4.3

0.6

0

0

0.16

0.08

0.13

0.03

Developing Alloy Compositions for Future High Temperature Disk …

potential. They showed that the composition of the c0 particles, i.e. the concentration of Ti, Ta and Nb atoms that replace Al atoms, has a profound effect on the APB energy and therefore yield stress. Tsolvus was important in the alloy design to ensure that forgings, which will receive a super-solvus solution heat treatment to produce a coarse grain microstructure, do not suffer from incipient melting at grain boundaries and quench cracking. This is particularly problematic in alloys that contain high levels of c0 and B. The risk was minimized by limiting Tsolvus to values of about 1165 °C. Whilst Co and Cr also have significant influence on Tsolvus, this target value principally restricted the Al content. The Ti ranges were then selected to improve oxidation resistance (which was correlated to the Cr/Ti ratio in at. %), to avoid η formation, to minimize Tsolvus and to maintain a stable MC carbide in alloys with Ta and Nb. Specifically, Antonov et al. [21] have shown that η and delta (d) phase can be avoided by ensuring that the Al, Ti, Ta and Nb values in atomic % are such that: Al [ 0:85 Ti þ Ta þ Nb

ð2Þ

Tantalum and Nb were added to the development compositions at appropriate levels to satisfy Eqs. (1) and (2) with the caveat that Nb be limited to 1 at.% (about 1.6 weight %) as there were concerns that excess Nb would be detrimental to time dependent crack growth. This limit was considered to be safe, albeit conservative as evidence in the literature [22] indicates that the effect of Nb content (up to about 1.7 wt%) on dwell crack growth behavior is less significant than grain size and size of c0 precipitates. The presence of Ta and Nb in c0 was understood to be beneficial as these elements show slower rates of diffusion in Ni compared to Al and Ti, which reduces coarsening during material manufacture and component operation. It was also considered that sufficient quantities of Ta and Nb should be added to the new alloys to develop stable MC carbides (where M can be Ti, Ta or Nb), which would resist decomposition at lower temperatures to M23C6 carbides. These latter grain boundary carbides were regarded as detrimental as they remove Cr from the c matrix adjacent to grain boundaries [23], reduce oxidation resistance and elevated temperature fatigue crack nucleation life [24]. Unlike Ti and Nb (due to large volume changes from forming Nb2O5 [25]), Ta may not be detrimental to oxidation resistance and has been shown to improve time dependent crack growth resistance [26]. The negative impact of adding Ta is the increase in raw material cost and with Nb, the increase in density. Yield stress models are more useful if they include terms for grain size (d) and the size of secondary c0 particles, i.e. those produced from quenching after solution heat treatment. The model shown below, proposed by Parthasarathy et al.

21

[11] was used in comparing the yield stress of new compositions with those of established alloys.


ry ¼ 1  fc0



*

8 9 + > 0 > < = k kðc þ c0 Þ c MðCRSSÞ þ pffiffiffiffiffiffiffiffiffiffiffiffiffi þ fc0 M  s0c þ qffiffiffiffiffi > dðc þ c0 Þ : ; d0c >

ð3Þ where fc0 is the volume fraction of c0 particles, CRSS is the critical resolved shear stress, M is the Taylor factor for polycrystals, d is grain size, k is the Hall-Petch coefficient for c and c0 phases and s0c is the friction stress that opposes dislocation motion from grain boundary precipitated c0 particles. Further details of the model can be found in [27]. More attention was paid to strength of the c phase for producing resistance to creep deformation. Initially, work was conducted to determine a minimum Co content for achieving the required creep strain behavior and a Tsolvus value below 1165 °C. A minimum Co content was sought to reduce the propensity for r formation [28], to promote improved resistance to type II hot corrosion damage since the melting temperature of Na2SO4-CoSO4 eutectic is 565 °C [29], compared to Na2SO4–NiSO4, which melts at 671 °C [30], and to minimize raw material costs. The benefits of Co in lowering the stacking fault energy [28, 31] and in producing more annealing twins are well documented [32]. The latter reduces effective grain size, which is important for fatigue crack nucleation life for coarse grain microstructures at temperatures below 650 °C. A further, less established benefit of Co is its ability to influence the size of secondary c0 particles, particularly those in intergranular locations. For a given cooling rate from super-solvus solution heat treatment, increasing Co content reduces the size of secondary c0 precipitates [33]. The other contribution to c strength was provided by Mo and W. Whilst established alloys have shown that high concentrations of Mo and W are necessary for good creep resistance, their values in the development compositions were limited by concerns regarding phase stability and density. Two approaches were used to provide an indication of phase stability was understood in 2 ways, both of which used the results of phase diagram modelling (Table 2). Firstly, to predict the solvus temperature for detrimental TCP phases, notably r (which is rich in Cr, Mo and W), to ensure that these were minimized and at least below the value for RR1000. Secondly, predictions of the atomic fraction of elements in the c phase were used in the second approach to calculate the average energy of d orbitals of alloying transition metals (Mdc), after Morinaga et al. [34]. The usefulness of this approach relies on defining a critical average Mdc value, below which a r free microstructure is likely. Guedou et al. [35] proposed that alloys can be designed using an average Mdc value of 0.915. In latter work, Reed

22 Table 2 Results of phase diagram modelling for the alloys in Table 1, calculated using the Thermo-Calc software with the TCNi7 database

M. C. Hardy et al. Alloy

c0 Tsolvus(°C)

Tsolidus(°C)

r Tsolvus(°C)

Ave Mdc

(%)

Alloy 1

1102

1214

863

0.917

0.04

Alloy 2

1092

1203

847

0.911

0.13

RR1000

1121

1232

909

0.905

0.08

c0 Tsolvus and r Tsolvus are the solvus temperatures for the c0 and r phases, respectively, and Tsolidus is the incipient melting temperature. Ave Mdc is the average energy of d orbitals of alloying transition metals, after Morinaga et al. [36], and d is the lattice misfit. These were calculated at 600 °C

et al. [3] calculated that alloys for which the average Mdc number was less than 0.88 would be free of r phase. At the concept stage, a value similar to that for RR1000 was considered acceptable for the development alloys in this current work. The aim in designing the alloys was to develop low or no coherency strain to minimize coarsening of c0 precipitates during time at temperatures above 700 °C. One measure of coherency strain is lattice misfit (d), given by Eq. (4).   2 ac 0  ac d¼ ð4Þ ac0 þ ac The values of d in Table 2 were estimated for 600 °C
 using the respective lattice parameters of c (a) and c0 a0c , which were calculated from molar volume values of the phases from phase diagram modelling and Avogadro’s constant. Resistance to environmental damage was sought by reducing the amount of Ti, thereby increasing the Cr/Ti ratio. It is understood that Ti dopes the chromia scale [36], where it segregates primarily along grain boundaries and forms large rutile nodules above the chromia scale [37]. It was found that for RR1000, the initial rates for the thickening kinetics of the chromia scale are considerably higher than those for relatively Ti-free chromia in alloys such as Inconel 718 and ATI 718PlusTM [36]. For Alloys 1 and 2 in Table 1, it was proposed that further improvements in resistance to environmental damage could be achieved by adding Si and Mn. Pedrazzini et al. [38] reported that after 100 h at 800 °C, an oxide scale in a Ni-base alloy with 1 at. % Mn consisted of an outer layer of NiCr2Mn2O4 and a subsequent inhomogeneous mix of chromia, spinel MnCr2O4 and rutile (Ti,Cr)O2. A 3 fold reduction in oxide thickness was observed compared to RR1000. Whilst this is encouraging, it is unclear whether the reduction was due to the presence of Mn or a consequence of the low Ti content (1 at. % Ti) and the high Cr/Ti ratio of 16. The beneficial effects of Mn can also be attributed to its ability to scavenge S and form high melting point sulfides. This reduces the available S in the alloy that can form low melting point Ni3S2, which gives rise to high temperature grain boundary embrittlement of Ni-Cr alloys [39].

Zirconium performs a similar role, scavenging O and S, and is known to provide improved high temperature tensile ductility and strength, creep life and rupture strength, and dwell crack growth resistance [40–42]. The S content in the small scale heats for producing powder was 10 ppm for the development alloys and RR1000. There was no cause to significantly alter the level of C in the development alloys from that in RR1000. However, B content was increased compared to that in RR1000, which was defined for ingot as well as powder metallurgy. The aim of the higher B value was to produce beneficial grain boundary cohesion and toughness from elemental B or isolated M3B2 boride particles [40–42] but without reducing the precipitation of intergranular secondary c0 . Iron was intentionally added to the development alloys at a level of about 1 at. % to facilitate the use of solid scrap from powder billet and machining chips in alloy manufacture. It was considered that such levels of iron would not be detrimental to alloy stability, and would behave like Co in reducing c0 Tsolvus.

Material Manufacture Powder of the development compositions and RR1000 were produced at Allegheny Technologies Incorporated (ATI) Specialty Materials (Robinson) in Pittsburgh, PA, USA. These were sieved to a final screen size of -270 mesh (53 lm), filled into 3 inch diameter mild steel containers and hot isostatic pressed (HIP). End slices from the HIP compacts were used for experimental work to verify c0 Tsolvus values, and to evaluate phase stability and oxidation damage from stress-free thermal exposures at temperatures up to 800 °C. Bars of 76 mm in height and about 67 mm in diameter were machined from the compacts for isothermal forging at ATI Forged Products in Cudahy, WI, USA. The bars were forged down to a height of about 18 mm. From these, circular section test piece cylinders were extracted at mid-height locations for heat treatment. Solution heat treatment was conducted at 20 °C above the c0 Tsolvus followed by cooling at a nominal rate of 1.1 °C/s ± 0.3 °C/s between Tsolvus and Tsolvus-90 °C. RR1000 cylinders were then given a post-solution heat treatment (P-SHT) of 16 h at 760 °C

Developing Alloy Compositions for Future High Temperature Disk …

and static air cooled. Cylinders of Alloy 1 and 2 received a P-SHT of 2 h at 850 °C, followed by 4 h at 800 °C, then static air cooled.

Experimental Work Prior to forging and heat treatment work, differential scanning calorimetry (DSC) and heat treatment trials were undertaken to determine the c0 Tsolvus and the incipient melting temperatures. For DSC, a 5 mm diameter  1 mm thick disk was cut from as-HIP material for each alloy and tested with a heating/cooling rate of 10 °C/min under flowing Ar. Samples for isothermal oxidation and thermal exposure tests were also prepared from as-HIP material, which was subsequently heat treated to conditions stated in the material manufacture section. These samples were 20  10  2 mm in size with chamfered edges on the faces of interest. At least one of these surfaces was polished to a 6 lm finish using diamond solution. The samples were held in open alumina boats and exposed in bench top furnaces at 800 °C for times up to 1000 h. Samples for assessment of phase stability were encapsulated in glass tubes under Ar and exposed in the same or similar furnaces. Following thermal exposure, oxide phases in Alloy 1 were identified using X-ray diffraction with a Cu Ka source. To protect the oxide scale during sectioning and metallographic preparation, samples were sputter coated with gold and electroplated with either Ni or Cu. Sections from oxidation and phase stability samples were prepared for scanning electron microscopy using SiC paper and diamond solution or colloidal silica (for Alloy 2). An electrolytic extraction was undertaken on the phase stability sample for Alloy 2 according to ASTM E963-95 [43]. X-ray diffraction was conducted on the extracted residue using a Cu Ka source and a Ni filter. After heat treatment, samples from forged material were prepared using standard metallographic techniques for characterization of microstructure. After polishing to a 0.10 lm finish using alumina solution, samples were electrolytically etched using 10% phosphoric acid solution. An optical microscope was used to acquire micrographs for grain size analyses. From these, grain boundaries were traced manually to enable the determination of grain areas and subsequent calculation of equivalent diameter values. Gamma prime precipitates were characterized using ImageJ Software from at least 5 backscattered electron images that were taken during examination of polished samples. As reported previously [44], the size and area fraction of tertiary c0 precipitates that were determined from this method were found to be comparable to those from high resolution scanning transmission electron microscopy (STEM).

23

Round bar laboratory test pieces, with a gage diameter of 4 mm, were machined from the near net sized cylinders for tensile and creep testing. They had 2 ridges along the parallel section, 20 mm apart, which enabled an extensometer to be attached. Tensile testing was performed at room temperature, 600, 700 and 800 °C at a constant strain rate of 0.01 per s. Creep testing was conducted at 650 °C with a stress of 1000 MPa, at 700 °C with a stress of 800 MPa, at 750°C with a stress of 600 MPa and at 800 °C with a stress of 300 MPa. Due to the limited material, only a single test piece could be evaluated for each test condition. Crack growth behavior was evaluated at 700 °C in square section 5  5 mm corner crack test pieces using a direct current potential difference technique, with electrodes welded either side of a 0.1 mm wide starter slit. Test pieces were initially pre-cracked at room temperature before they were heated to 700 °C and subject to about 2000 baseline (0.25 Hz) fatigue cycles at a stress ratio of 0.1 and a peak load of 12 kN. Dwell cycles were then applied from a stress intensity factor range (DK) of about 15 to 30 MPa√m. These were trapezoidal waveforms (1-X-1-1) in which X was a hold period of 3600 s at peak load, also at 12 kN. Where possible, tests were completed after a further period of baseline fatigue cycles.

Results from the Experimental Work The findings from DSC and heat treatment trials showed that the temperature for complete dissolution of c0 was found to be about 1160–1165 °C for both development alloys and about 1145–1150 °C for RR1000. Similarly, the onset of incipient melting was detected at temperatures over 1200 °C for all of the alloys, typically about 1210 °C. These latter values are very close to the predictions in Table 2. Average grain size values of forged material after heat treatment are summarized in Table 3, with average size and volume fraction data for secondary and tertiary c0 precipitates. Values for one standard deviation are also provided in Table 3. The data confirm that a consistent grain size of about 20 lm was produced. The shape and size of c0 precipitates in the 3 alloys can be understood from the backscattered electron images in Fig. 1. They show that secondary c0 precipitates in the development alloys are more irregular in shape than those in RR1000, are larger in size and have a greater number density. RR1000 has a higher volume fraction of tertiary c0 precipitates, which are about a half or one-third of the size of those in Alloy 1 and 2. Data from tensile tests at room temperature and 800 °C are presented in Fig. 2 in the form of 0.2% proof stress (Rp0.2%) and tensile strength (Rm) values for the 2 development alloy compositions and RR1000. Whilst the tensile

24

M. C. Hardy et al.

Table 3 Results from characterization of microstructure. SGP and TGP are secondary and tertiary c0 respectively

Alloy 1

c0 volume fraction (%)

Measured dT/ dt (°C/s)

Average size (diameter) Gamma grain (lm)

SGP (nm)

TGP (nm)

SGP

TGP

Total

0.9 ± 0.1

20.6 ± 0.5

273 ± 8

23.7 ± 0.7

52.4 ± 1.0

1.2 ± 0.1

53.6 ± 1.0

Alloy 2

1.4 ± 0.1

21.6 ± 0.2

238 ± 7

30.8 ± 1.4

50.7 ± 0.7

1.9 ± 0.4

52.5 ± 0.5

RR1000

1.1 ± 0.2

18.2 ± 0.6

149 ± 4

10.3 ±0.1

39.3 ± 0.6

4.5 ± 1.1

43.8 ± 1.8

Values for one standard deviation follow average values. Over 600 and 3000 SGP and TGP particles, respectively, were measured for each alloy

Fig. 1 Backscattered electron images of c0 precipitates after heat treatment for (i) Alloy 1, (ii) Alloy 2 and (iii) RR1000 at a magnification of 20,000. Image (iv) is a 100, 000 magnification image of the red box in (iii)

strength properties of Alloy 1 are similar to those for RR1000, the data show that Alloy 2 provides useful improvements over RR1000. Creep rupture life data for the development alloys and RR1000 are compared using the Larson-Miller Parameter (LMP) in Fig. 3, where LMP ¼

ð273 þ hÞð28 þ log tÞ 1000

ð5Þ

in which h is temperature and t is the creep rupture life. It is evident that Alloy 2 shows improved creep rupture life compared to RR1000. Alloy 1, however, produced very

similar creep rupture lives to those from RR1000. Creep strain data for one of the test conditions (700 °C, 800 MPa) are plotted in Fig. 4. There is a significant difference in the shape of the creep curves for these alloys. Alloy 1 accumulates creep strain much faster than RR1000 and Alloy 2, which show similar creep strain behavior up until about 0.4-0.5% creep strain. Further creep deformation appears to result in much faster rates of damage accumulation in RR1000 than in Alloy 2. The results of crack growth testing are plotted in Fig. 5 in terms of crack growth rate per cycle (da/dN) versus stress intensity factor range (DK). The data show that a significant

Developing Alloy Compositions for Future High Temperature Disk …

strength (MPa)

1500

 

Alloy 1 % over RR1000 T (°C) Rp0.2% Rm 20 1.5 2.9 800 3.4 0.5

1300   

1100

                     

           

900

       

700

          

   

Alloy 2 % over RR1000 T (°C) Rp0.2% Rm 20 6.6 6.4 800 7.2 9.5

    

0.1

20°C, Rp0.2% 800°C, Rp0.2% 20°C, Rm 800°C, Rm

Predicted yield stress (MPa) RR1000 Alloy 1 Alloy 2 20 1218 1335 1397 800 902 967 1013

T (°C)

RR1000 0.08

creep strain

1700

25

0.06

0.04

0.005 Alloy 2

0.004

Alloy 1

0.003



0.02

 

RR1000

0.002



0.001

RR1000 Alloy 1 Alloy 2

0

Fig. 2 0.2% proof stress (Rp0.2%) and tensile strength (Rm) data for Alloy 1, Alloy 2 and RR1000 at 20 and 800 °C. Predicted values of yield stress from equation [3] are shown in the lower right table

0 0

0

100

200

10

300

20

400

30

500

time (hours) Fig. 4 Creep strain data measured from tests at 700 °C with a nominal stress of 800 MPa

1100

650°C

Alloy 1 Alloy 2

900

RR1000

700°C stress (MPa)

Alloy 2

Alloy 1

700

750°C 500

300

t = time to rupture (hours) θ = temperature in °C

800°C

100 27

29

31

33

LMP (273 + θ )(28 + logt)/1000 Fig. 3 Time to creep rupture data for Alloy 1, Alloy 2 and RR1000 shown in terms of the Larson-Miller Parameter (LMP) from creep tests at 650, 700, 750 and 800 °C

increase in growth rate is produced in all of the alloys on the application of 3600 s dwell cycles at DK values of about 18 MPa√m. Subsequently, however, cracks were found to retard rapidly, i.e. reduced rates of crack growth were observed with increasing DK. To ensure that tests continued, baseline cycles were re-applied to DK values of about 23-25 MPa√m. At which point, dwell cycles were continued, with high rates of crack growth that were approximately 35

times faster than those from baseline cycles. It is interesting to note that the crack in the Alloy 1 test piece showed another period of crack retardation to a DK of about 27 MPa√m and then started to grow in the expected manner, showing increases in growth rate with increasing DK. In contrast, cracks in the Alloy 2 and RR1000 test pieces continued to accelerate, with the highest rates of crack growth in the Alloy 2 test pieces. Examination of fracture surfaces in a SEM confirmed that intergranular crack growth occurred from all 3600 s dwell cycles, even during crack retardation. Scanning electron microscope images of oxidation damage on polished surfaces of Alloy 1 and 2 after 1000 h at 800 °C are shown in Fig. 6. Previous work [36] on RR1000 has identified that the outer scale consists of rutile (TiO2) then chromia (Cr2O3). Beneath these, there is internal oxidation damage in the form of alumina (Al2O3) intrusions. Both of the development alloys also displayed these oxidation products but their depths were smaller than those for RR1000, which are indicated in Fig. 6 by the red arrows that are labelled 1 and 2 for the scale and alumina intrusions respectively. The data for RR1000 were taken from experimental work reported in [36, 45]. The comparison shows that whilst both development alloys offer improved oxidation resistance, the improvement is most significant for Alloy 2. The resolution of the Alloy 1 image in Fig. 6 does not allow further insights into oxidation products or indications of elemental migration. However, the results from X-ray diffraction in Fig. 6 show the presence of Cr2MnO4 spinel in the scale in addition to rutile and chromia.

26

1 RR1000 (3600 s)

700°C in air

RR1000 (1 s) Alloy 1 (3600 s)

0.1 3600s dwell cycles

da/dN (mm/cycle)

Fig. 5 Crack growth rate (da/ dN) versus stress intensity factor range (DK) data for Alloy 1, Alloy 2 and RR1000 at 700 °C from testing in air using baseline (1-1-1-1 s) and dwell (1-3600-1-1 s) cycles

M. C. Hardy et al.

Alloy 1 (1 s) Alloy 2 (3600 s)(1)

0.01

Alloy 2 (3600 s)(2)

crack retardation

Alloy 2 (1 s)(1)

x 35

0.0001

0.25 Hz fatigue cycles

0.00001 10

60

ΔK(MPa√ m) Fig. 6 Secondary electron image of oxidation products in Alloy 1 (top left) on a polished surface after 1000 h at 800 °C. Backscattered electron image of oxidation products in Alloy 2 (top right) on a polished surface after 1000 h at 800 °C. The red arrows are the depths of scale (1) and alumina intrusion (2) respectively that were measured after an identical exposure on a polished coarse RR1000 surface [36, 45]. Lower chart shows X-ray diffraction data from a polished Alloy 1 sample after 1000 h at 800 °C

Alloy 2 (1 s)(2)

0.001

Developing Alloy Compositions for Future High Temperature Disk …

It has recently been reported that RR1000 is free of undesirable TCP phases after a thermal exposure of 1000 h at 800 °C [46]. In contrast, the backscattered electron image in Fig. 9 shows that extensive grain boundary decoration was present in Alloy 1 after a 750 h exposure at 800 °C. Evidence in terms of a selected area electron diffraction pattern from TEM has confirmed that this phase, rich in Mo, Nb, Co, Cr, Fe, W and Ta, was C14 Laves phase. There was little or no decoration of grain or prior particle boundaries in Alloy 2. Bright particles were present in backscattered electron images. From EDS, these were found to be either rich in Ti, Ta, Nb and C or rich in Mo and B but depleted in Ni, Co and Cr. X-ray diffraction confirmed that the former were MC carbides but peaks, presumably for Mo rich borides, were not identified.

Discussion Whilst a consistent grain size has been produced in the experimental material, measured cooling rates (Table 3) were unfortunately found to be at the ends and middle of the expected variation. This has been at the detriment of tensile and creep properties for Alloy 1 but to the benefit of Alloy 2. As such the average size of secondary c0 precipitates for Alloy 1 was larger than that for Alloy 2 and RR1000. Given the low predicted misfit value (Table 2), which has been found to agree well with experimental measurements [47], it is possible that Alloy 1 is particularly susceptible to the anomalous coarsening and splitting phenomenon that was identified by Mitchell et al. [48, 49] during aging of RR1000 at 800 °C. Such behavior may have grown the secondary c0 precipitates at the expense of tertiary c0 particles, both in terms of size and volume fraction. The shape of secondary c0 precipitates in Fig. 1 (ii) also suggest that this phenomenon occurs in Alloy 2. The differences in cooling rate and consequently secondary c0 size, are likely to be major contributors to the observed differences in tensile and creep properties between Alloy 1 and Alloy 2. The trends are similar to those reported by other workers, notably Groh [50] for Waspaloy, in which improved properties were achieved by a higher cooling rate from solution heat treatment. Predicted values of yield stress are included in Fig. 2. These were calculated from equation [3], using the method described in [13, 27] for determination of APB energy and the data in Table 3. The model correctly predicts the trends in yield stress behavior that is observed in these alloys after heat treatment. Furthermore, if the grain size and the size distribution of secondary c0 particles were the same in Alloys 1 and 2, the model indicates that the yield stress value for Alloy 1 at 20 °C is within 32 MPa of the predicted value for Alloy 2 in Fig. 2.

27

Whilst the size and number density of tertiary c0 precipitates are not included in equation [3] for predicting yield stress, these aspects of microstructure are critical for minimizing the accumulation of creep strain during sustained load tests or conversely, for beneficial stress relaxation behavior during sustained strain tests. This is evident from Fig. 4. RR1000 has a lower volume fraction of c0 than the development alloys, by 8–10%, but offers a competitive resistance to creep strain accumulative at the selected test conditions (700 °C, 800 MPa) as a result of a relatively high volume fraction of fine tertiary c0 precipitates. Alloy 2 shows very similar creep strain versus time data to RR1000 although the spaces between the secondary c0 precipitates are smaller in Alloy 2 but the size of tertiary c0 precipitates is a factor of 3 larger and there are fewer of them. However, Alloy 2 shows improved resistance to creep above strains of 0.4–0.5%. It appears that significant creep damage occurs at lower creep strains in RR1000 than in Alloy 2. It is postulated that such damage nucleates from grain boundary particles such as carbides, borides or oxides. RR1000 has fewer borides that may pin grain boundaries and hinder grain boundary displacement but it shows a significantly higher concentration of small oxide particles, due to the Hf content in the alloy. Alloy 1 accumulates creep strain much faster than the other 2 alloys. This is likely to result from having larger secondary c0 precipitates and the smallest volume fraction of tertiary c0 precipitates. The observed effects of c0 size and volume fraction can be predicted using the Orowan model adapted by Galindo-Nava and Rae [51]. Evidence is not presented here but has been reported by Christofidou et al. [27] for other development compositions. The inferior creep performance of Alloy 1 is not considered to result from phase instability as Laves phase was not detected after long term exposures at 700 °C. The inelastic deformation behavior shown by Alloy 1 is beneficial during dwell crack growth testing. This is because nominally elastic material constrains the inelastic zone around the crack tip and imposes an essentially strain controlled loading configuration at the crack tip. Under these conditions, the crack tip stresses in Alloy 1 are able to decay much faster during the sustained load than those in Alloy 2. The behavior of the development alloys shown in Fig. 5 is consistent with the findings of previous work that have investigated the effects of microstructure on dwell crack growth rates [44, 52–54]. In coarse grain microstructures, in particular, it has been found that cracks at low values of DK retard during dwell cycles. It is understood that oxide intrusions form ahead of a stationary or a slow growing crack, with an increase in volume that induces compressive stresses in the matrix in the vicinity of the crack tip [55]. This and crack blunting from inelastic deformation are likely causes of the observed retardation behavior. Significantly

28

higher growth rates are produced at DK values greater than 25 MPa√m from continuous intergranular crack growth. Further evidence is required to understand the mechanisms that give rise to this behavior. It is encouraging, however, to find that development alloys, which contain a higher number density of c0 precipitates and have higher yield stress values than RR1000 show similar time dependent crack growth behavior at 700 °C. The images in Figs. 6 and 7 indicate that the development alloys have improved resistance to oxidation damage compared to RR1000. Although the reduced Ti content and the slightly higher Cr/Ti ratio in the new alloys may contribute significantly to this improvement, more detailed investigations are necessary to identify the presence of Ta oxides [56] and SiO2 particles below the chromia scale. Similarly, the importance of Cr2MnO4 spinel in the scale of Alloy 1 needs to be understood but is beyond the scope of the current study. Alloy 1 was shown to precipitate C14 Laves phase after long term exposure at 800 °C while Alloy 2 and RR1000 were found to not to form undesired TCP phases. Interestingly, none of the alloys precipitated r phase, which for Fig. 7 Backscattered electron image of Alloy 1 after 750 h at 800 °C

M. C. Hardy et al.

these alloys supports the use of phase diagram modelling to predict r Tsolvus and c composition for average Mdc values. It has been established [57] that additions of 0.5 and 1 wt% Si. to cast B-1900, 713C and MAR-M200 produce hexagonal Mo(Ni,Si)2 Laves phase during solidification. However, the reduced amount of Si in Alloy 2 did not lead to the precipitation of Laves phase. It should also be noted that a similar development alloy, D8 in reference [58] also precipitated Laves phase but it contained no Si. It did contain 0.9 wt% Fe and high levels of Nb, W and Cr, with moderate additions of Mo.

Summary This paper examined 2 new alloy compositions, which were designed for possible disk rotor applications. Both were intended to have higher c0 content than the existing alloy, RR1000, and be produced using powder metallurgy and isothermal forging to enable forgings to be made that show a consistent coarse grain microstructure. Small pancake forgings of the new alloys and RR1000 were made and from

Developing Alloy Compositions for Future High Temperature Disk …

these, blanks were solution heat treated, cooled at measured rates and aged. The development alloys were similar in composition but they exhibited different tensile and creep properties, phase stability and resistance to oxidation damage. Despite attempts to minimize variation in microstructure from heat treatment, differences in c0 size distribution were found to influence tensile and creep behavior. One of the new alloys (Alloy 2) showed improved yield and tensile strength compared to coarse grain RR1000. Alloy 2 displayed similar initial creep strain behavior to RR1000 but superior resistance to subsequent creep damage, producing longer creep rupture lives. All of the alloys showed crack retardation at low stress intensity factor ranges (DK) from 3600 s dwell cycles at 700 °C in air. This occurred whilst crack growth was intergranular. In agreement with work in the literature, this behavior was attributed to relaxation of crack tip stresses and the development of compressive crack tip stresses from the formation of crack tip oxide intrusions. At higher DK values, the rates of continuous dwell crack growth in the development alloys were similar to those in coarse grain RR1000 although the new alloys showed a higher number density of c0 precipitates. Alloy 1 was found to precipitate C14 Laves phase from long term exposure at 800 °C. This was considered to be due to excessive Si content in an alloy that contained high levels of Mo, W, Cr and Nb. Like RR1000, r phase was not detected in the new alloys after 750 h at 800 °C. Acknowledgements This work was supported by Rolls-Royce plc and the Rolls-Royce/EPSRC Strategic Partnership under EP/H022309/1, EP/H500375/1 and EP/M005607/1. Dr Hardy would like to thank Rolls-Royce colleague Dr Han Tai for his support and encouragement in this work, Professor Roger Reed and Dr David Crudden from the University of Oxford for fruitful discussions and Mr Joe Muha, formerly of ATI Specialty Materials, Robinson, for his assistance in making the powder compacts.

References 1. European Commission, (2001), European Aeronautics: A Vision for 2020. 2. European Commission, (2011), Flightpath 2050 Europe’s vision for aviation. https://doi.org/10.2777/50266. 3. Reed RC, Mottura A, Crudden DJ, (2016), Alloys-by-design: towards optimization of compositions of nickel-based superalloys. In: Superalloys 2016, (Ed. by Hardy MC et al.), TMS, Warrendale, p 15–23. 4. Mitchell RJ, Lemsky JA, Ramanathan R, Li HY, Perkins KM and Connor LD, (2008), Process development & microstructure & mechanical property evaluation of a dual microstructure heat treated advanced nickel disc alloy. In: Superalloys 2008, (Ed. by Reed RC et al.), TMS, Warrendale, p 347–356. 5. https://www.icao.int/Meetings/EnvironmentalWorkshops/Documents/ ICAO-TransportCanada-2006/Anderson_ops.pdf.

29 6. Dreshfield RL, Gray HR, (1984), P/M Superalloys – a Troubled Adolescent, NASA Technical Memorandum, TM-83623, Lewis Research Center, Cleveland. 7. Anderson R, (1990), Powder Metallurgy at Pratt and Whitney, The International Journal of Powder Metallurgy, 26, (2):171–178. 8. Smythe S, (2008), Superalloy powders: an amazing history, Advanced Materials and Processes, November: 52–55. https:// www.asminternational.org/documents/10192/1887146/amp16611 p052.pdf/6664af73-fa87-4c2f-8b94-d420d5c438e3/AMP16611P052. 9. Hardy MC, Zirbel B, Shen G, Shankar R, (2004), Developing damage tolerance and creep resistance in a high strength nickel alloy for disc applications. In: Superalloys 2004, (Ed. by Green KA et al.), TMS, Warrendale, p 83–90. 10. Reppich B, (1982), Some new aspects concerning particle hardening mechanisms in c′ precipitating Ni-based alloys – I. theoretical concept, Acta Metall., 30:87–94. 11. Parthasarathy TA, Rao SI, Dimiduk DM, (2004), A fast spreadsheet model for the yield strength of superalloys. In: Superalloys 2004, (Ed. by Green KA et al.), TMS, Warrendale, p 887–896. 12. Kozar RW, Suzuki S, Milligan WW, Schira JJ, Savage MF, Pollock TM, (2009), Strengthening mechanisms in polycrystalline multimodal nickel-base superalloys, Met. Trans. A, 40, (7):1588– 1603. 13. Crudden DJ, Mottura A, Warnken N, Raeisinia B, Reed RC, (2014), Modelling of the influence of alloy composition on flow stress of high-strength nickel-based superalloys, Acta Mater., 75:356–370. 14. Galindo-Nava EI, Connor LD, Rae CMF, (2015), On the prediction of the yield stress of unimodal and multimodal c′ nickel-base superalloys, Acta Mater., 98:377–390. 15. Hardy M, Reed R, Crudden D, (2018), European Patent Specification EP 3 112 485 B1, 15 August 2018. 16. Hardy MC, Reed RC, Crudden D, (2019), United States Patent US 10,266,919 B2, 23 April 2019. 17. Hardy MC, Reed RC, Crudden D, (2019), United States Patent US 10,422,024 B2, 24 September 2019. 18. Saunders N, (1996), Phase diagram calculating for Ni-Based Superalloys. In: Superalloys 1996, (Ed. by Kissinger RD et al.), TMS, Warrendale, PA, USA, p 101–110. 19. Small CJ, Saunders N, (1999), The application of CALPHAD techniques in the development of a new gas-turbine disk alloy, MRS Bulletin, 24, (4): 22–26. 20. Saunders N, Fahrmann M, Small CJ, (2000), The application of CALPHAD calculations to Ni-Based Superalloys. In: Superalloys 2000, (Ed. by Pollock TM et al.), Warrendale, p 803–811. 21. Antonov S, Detrois M, Helmink RC, Tin S, (2015), Precipitate phase stability and compositional dependence on alloying additions in c-c′-d-η Ni-based superalloys, Journal of Alloys and Compounds, 626:76–86. 22. Telesman J, Kantzos P, Gayda J, Bonacuse PJ, Prescenzi A, (2004), Microstructural variables controlling time-dependent crack growth in a P/M superalloy. In: Superalloys 2004, (Ed. by Green KA et al.), TMS, Warrendale, p 215–224. 23. Evans HE, (1988), Cavity formation and metallurgical changes induced by growth of oxide scale, Materials Science and Technology, 4, (12):1089–1098. 24. Sudbrack CK, Evans LJ, Garg A, Perea DE and Schreiber DK, (2016), Characterization of grain boundaries and associated minor phases in disk alloy ME3 exposed at 815 °C. In: Superalloys 2016, (Ed. by Hardy MC et al.), TMS, Warrendale, p 927–936. 25. Connolley T, Starink MJ, Reed PAS, (2000), Effect of oxidation on high temperature fatigue crack initiation and short crack growth in Inconel 718. In: Superalloys 2000, (Ed. by Pollock TM et al.), TMS, Warrendale, p 435–444.

30 26. Huron ES, Bain KR, Mourer DP, Gabb T, (2008), Development of high temperature capability P/M disk superalloys. In: Superalloys 2008, (Ed. by Reed RC et al.), TMS, Warrendale, p 181–189. 27. Christofidou KA, Hardy MC, Li HY, Argyrakis C, Kitaguchi H, Jones NG, Mignanelli PM, Wilson AS, Messé OMDM, Pickering EJ, Gilbert RJ, Rae CMF, Yu S, Evans A, Child D, Bowen P and Stone HJ, Met. Trans., 49A (9):3896–3907. 28. Tien JK, Jarrett RN, (1983), Effects of Cobalt in Nickel-Base Superalloys, NASA Contractor Report CR168308, NASA Lewis Research Center, Cleveland. 29. Luthra KL, (1982), Low temperature hot corrosion of cobalt-base alloys: part I. morphology of the reaction product, Met. Trans. A, 13, (10):1843–1852. 30. Lillerud KP, Kofstad P, (1984), Sulfate-Induced hot corrosion of nickel, Oxidation of Metals, 21, (5/6):233–270. 31. Reed RC, (2009), The Superalloys, Fundamentals and Applications, Chapter 2, Cambridge University Press, ISBN 9780511541285, https://doi.org/10.1017/CBO9780511541285. 32. Randle V, (1997), Mechanism of twinning-induced grain boundary engineering in low stacking-fault energy materials, Acta Mater., 47, (15):4187–4196. 33. Christofidou KA, Jones NG, Pickering EJ, Flacau R, Hardy MC, Stone HJ, (2016), The microstructure and hardness of Ni-Co-Al-Ti-Cr quinary alloys, Journal of Alloys and Compounds, 688:542–552. 34. Morinaga M, Yukawa N, Adachi H, Ezakil H, (1984), New PHACOMP and its application to alloy design. In: Superalloys 1984, (Ed. by Gell M et al.), TMS, Warrendale, p 523–532. 35. Guédou J-Y, Augustins-Lecallier I, Nazé L, Caron P, Locq D, (2008), Development of a new Fatigue and creep resistant PM nickel-base superalloy for disk applications. In: Superalloys 2008, (Ed. by Reed RC et al.), TMS, Warrendale, p 21–30. 36. Cruchley C, Evans HE, Taylor MP, Hardy MC, Stekovic S, (2013), Chromia layer growth on a Ni-based superalloy: sub-parabolic kinetics and the role of titanium, Corrosion Science, 75:58–66. 37. Pedrazzini S, Rowlands BS, Turk A, Parr IMD, Hardy MC, Bagot PAJ, Moody MP, Galindo-Nava E, Stone HJ, (2019), Partitioning of Ti and kinetic growth predictions on the thermally grown chromia scale of a polycrystalline nickel-based superalloy, Met. Trans A., 50, (7):3024–3029. 38. Pedrazzini SD, Child DJ, West G, Doak SS, Hardy MC, Moody MP, Bagot PAJ, (2016), Oxidation behaviour of a next generation polycrystalline Mn containing Ni-based superalloy, Scripta Materialia, 113:51–54. 39. White CL, Schneibel JH, Padgett RA, (1983), High temperature embrittlement of Ni and Ni-Cr alloys by trace elements, Met. Trans. A, 14, (4):595–610. 40. Gayda J, Gabb TP, Miner RV, (1985), Fatigue Crack Propagation of Nickel-Base Superalloys at 650 °C, NASA Technical Memorandum 87150, NASA Lewis Research Center, Cleveland. 41. Garosshen TJ, Tillman TD, McCarthy GP, (1987), Effects of B, C, Zr on the structure and properties of a P/M nickel base superalloy, Met. Trans. A, 18, (1):69–77. 42. Jain SK, Ewing BA, Yin CA, (2000), The development of improved performance PM Udimet® 720 turbine disks. In: Superalloys 2000, (Ed. by Pollock TM et al.), TMS, Warrendale, p 785–794.

M. C. Hardy et al. 43. ASTM E963-95, Standard Practice for Electrolytic Extraction of Phases from Ni and Ni-Fe Base Superalloys Using a Hydrochloric-Methanol Electrolyte, ASTM International, West Conshohocken. 44. Shulz F, Li HY, Kitaguchi H, Child D, Williams S, Bowen P, (2018), Influence of tertiary gamma prime (c′) size evolution on the dwell crack growth behaviour in CG RR1000, Met. Trans. A, 49, (9):3874–3884. 45. Cruchley S, Taylor MP, Evans HE, Hardy MC, Child DJ, (2014), Characterisation of subsurface oxidation damage in Ni based superalloy, RR1000, Materials Science and Technology, 30, (15):1884–1889. 46. Wilson AS, Christofidou KA, Evans A, Hardy MC, Stone HJ, (2019), Comparison of methods for quantification of topologically closed packed phases in Ni-based superalloys, Met. Trans. A, 50, (12):5925–5934. 47. Papadaki C, (2019), Optimising Creep Resistance of Next Generation Nickel-Based Superalloys via Tailoring of c′ Precipitates Using Different Heat Treatments, Confirmation of Status Report, Trinity College, University of Oxford, Oxford. 48. Mitchell RJ, Preuss M, (2007), Inter-relationships between the composition, c′ morphology, hardness, and c-c′ mismatch in advanced polycrystalline nickel-base superalloys during aging at 800 °C, Met. Trans. A, 38, (3):615–627. 49. Chen Y, Prasath Babu R, Slater TJA, Bai M, Mitchell R, Ciuca O, Preuss M, Haigh SJ, (2016), Acta Materialia, 110:295–305. 50. Groh JR, (1996), Effect of cooling rate from solution heat treatment on Waspaloy microstructure and properties. In: Superalloys 1996, (Ed. by Kissinger RD et al.), TMS, Warrendale, p 621–626. 51. Galindo-Nava EI, Rae CMF, (2016), Microstructure-senstive modelling of dislocation creep in polycrystalline FCC alloys: Orowan theory revisited, MSEA, 651:116–126. 52. Gabb TP, Gayda J, Telesman J, Garg A, (2008), The effects of heat treatment and microstructure variations on disk superalloy properties at high temperature. In: Superalloys 2008, (Ed. by Reed RC et al.), TMS, Warrendale, p 121–130. 53. Li HY, Sun JF, Hardy MC, Evans HE, Williams SJ, Doel TJA, Bowen P, (2015), Effects of microstructure on high temperature dwell fatigue crack growth in a coarse grain PM nickel based superalloy, Acta Mater., 90:355–369. 54. Telesman J, Gabb TP, Ghosn LS, (2016), Separating the influence of environment from stress relaxation effects on dwell fatigue crack growth in a nickel-base disk alloy. In: Superalloys 2016, (Ed. by Hardy MC et al.), TMS, Warrendale, p 551–560. 55. Evans HE, Li HY, Bowen P, (2013), A mechanism for stress-aided grain boundary oxidation ahead of cracks, Scripta Materialia, 69:79–182. 56. Sudbrack CK, Draper SL, Gorman TT, Telesman J, Gabb TP, Hull DR, (2012), Oxidation and the effects of high temperature exposures on notched fatigue life of an advanced powder metallurgy disk superalloy. In: Superalloys 2012, (Ed. by Huron ES et al.), TMS, Warrendale, PA, p 863–872. 57. Miner RV Jr. (1977), Effects of silicon on the oxidation, hot-corrosion and mechanical behavior on two cast nickel-base superalloys, Met. Trans. A, 8, (12):1949–1954. 58. Reed R, Crudden D, Raeisinia B, Hardy M, (2016), European Patent Specification EP2 894 234 B1, 19 October 2016.

Development of AGAT, a Third-Generation Nickel-Based Superalloy for Single Crystal Turbine Blade Applications J. Rame, P. Caron, D. Locq, O. Lavigne, L. Mataveli Suave, V. Jaquet, M. Perrut, J. Delautre, A. Saboundji, and J.-Y. Guedou

Abstract

Keywords

The new third-generation single crystal superalloy AGAT has been developed for aircraft engine turbine blade applications. Alloy design procedure is described and AGAT alloy properties are presented and compared with those of, respectively, first-, second-, and third-generation AM1, CMSX-4 and CMSX-10 alloys. AGAT alloy exhibits high creep resistance at very high temperature (1200 °C) compared with first- and second-generation superalloys while maintaining moderate density (8870 kg m−3) and stable microstructure unlike the third-generation superalloy. High cycle fatigue (HCF) and low cycle fatigue (LCF) properties of AGAT alloy are similar to second-generation CMSX-4 alloy. AGAT solution heat treatment allows suppressing the c/c′ interdendritic eutectic pools at a temperature 30 °C lower than for CMSX-10 with a shorter duration. Oxidation resistance of AGAT alloy at 1150 °C is lower than that of second but higher than that of third-generation reference superalloys. AGAT shows low sensitivity to secondary reaction zone (SRZ) formation under b-NiPtAl bond coat (BC) and great spallation resistance of YPSZ EB-PVD thermal barrier coating (TBC) compared with reference alloys. Finally, single crystal turbine blades were successfully manufactured through industrial processes to be tested in engine conditions.

Nickel based Alloy design

J. Rame (&)
J. Delautre
J.-Y. Guedou Safran Aircraft Engines, 171 Boulevard de Valmy, 92702 Colombes, France e-mail: [email protected] P. Caron
D. Locq
O. Lavigne
M. Perrut ONERA, DMAS, Université Paris-Saclay, 92322 Châtillon, France L. Mataveli Suave
V. Jaquet
A. Saboundji Safran Tech, PFX, 171 Boulevard de Valmy, 92700 Colombes, France










Single crystal Superalloy Creep Oxidation TBC




Context and Background Nickel-based superalloys are currently key materials for high temperature components of aircraft turboengines, as first stage single crystal (SC) turbine blades and vanes. During the past four decades, nickel-based SC superalloys have undergone significant chemistry changes to increase their creep resistance at high temperature. Particularly, main creep gains were successively achieved by (i) introducing rhenium up to 3 wt% in the second-generation SC superalloys, (ii) increasing the level of rhenium up to about 6 wt% in the third-generation SC superalloys, and (iii) adding ruthenium in high containing rhenium superalloys in order to stabilize their microstructure (fourth-generation SC superalloys) [1–3]. However, these compositional changes possibly led to the occurrence of microstructure instability phenomena such as cellular colonies, observed along low-angle grain boundaries and within dendrite cores, and/or secondary reaction zones (SRZ) in coated superalloys [4]. SRZ occur beneath the primary diffusion zone between the bond coat and the alloy. Their depth can grow up to hundreds of microns at service temperature due to interdiffusion phenomena. Since the mechanical strength of SRZ is significantly lower than that of the c/c′ superalloy, this drastically reduces mechanical properties of the component, particularly for cooled components with thin walls. It has been shown that these instabilities were mostly due to excess rhenium additions [5]. Thus, development of new superalloys for coated components such as high pressure (HP) turbine blades must take into account the compatibility with bond coats. ONERA and Safran have conducted several alloy development programs since the 1980s [6] resulting in the

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_3

31

32

J. Rame et al.

introduction of first-generation SC superalloys in aircraft and helicopter gas turbine engines. Indeed, AM1 superalloy [7] is used for HP blades and vanes of M88, TP400, and SaM146 engines, respectively, powering the Dassault Rafale fighter, the Airbus A400M military airplane, and the Sukhoi Superjet 100 regional airplane. Moreover, AM3 [8] and MC2 [7, 9] alloys were developed to manufacture the HP SC blades of helicopter Safran engines. However, future aircraft require increasing engine service temperatures demanding the introduction of new superalloys able to withstand high stress levels in extreme environments. For this purpose, the development of a new SC superalloy was undertaken for a specific turbine blade application, taking into account the final product as the three-part component: superalloy—BC—TBC. This paper focuses on the design and characterization of the AGAT alloy [10].

Alloy Design Procedure The AGAT alloy chemistry was designed to reach targeted properties for a specific HP turbine blade. Key parameters considered were: (i) density (targeted 950 °C) is mainly influenced by diffusion controlled creep mechanisms, such as climb and cross-slip of matrix dislocations, which depend on superalloy chemistry and microstructure [1]. Increasing the content of refractory elements such as rhenium, tungsten, tantalum, and molybdenum, characterized by low diffusion coefficients, has therefore a beneficial effect on the high temperature creep life of these alloys. Moreover a high volume fraction of c′ phase is needed in the very high temperature regime to maintain creep properties to be as high as possible. The content of the various alloying elements in AGAT alloy has therefore been carefully balanced to obtain maximum strengthening of the c matrix and a c′ solvus temperature as high as possible. In particular, the concentration of alloying elements such as rhenium, chromium, and molybdenum partitioning preferentially to the c matrix was tuned to avoid the precipitation of deleterious topologically close-packed (TCP) phase particles. Thermodynamic data such as c′ solvus temperature, solidus temperature, and volume fraction of c′ phase at high temperature were estimated using the CALPHAD method (Thermo-Calc software, TCS Ni-based superalloys Database v8) and empirical models to get information for the choice of the optimized alloy chemistry. The proneness to TCP phase formation was evaluated by combining Thermo-Calc calculations and the “New PHACOMP” method devised on the basis of molecular orbital calculations of electronic structures [12]. An average Md parameter calculated from the chemical composition of the alloy must have a value less than 0.985 in order to preclude formation of TCP phase particles. No TCP phase precipitation at high temperature was targeted.

 39:31 C Co  38:75 C Fe  51:48 C Mo  34:6 C W  37:09 CV  32:94 C Nb  31:28 CTa þ 11:07 C Ti þ 107:40 C C þ 0:5176 ðC Mo Þ2  1:4006 C Co
CTi  26:57 C Re  31:306 C Ru

Tensile Strength In the temperature regimes where the c′ precipitates are sheared by the matrix dislocations, the intrinsic resistance of

Development of AGAT, a Third-Generation Nickel-Based Superalloy …

these particles is of prime importance to ensure a high level of alloy strength. It concerns the tensile strength at low temperatures, typically less than 800 °C. A GPR (gamma prime resistance) criterion was thus calculated to quantify the c′ strength level. This criterion is defined as followed, taking into account the atomic concentration of c′-former elements in the alloy (the higher the GPR value, the higher the c′ strength):

33

Walston et al. [18]. The following formula was determined by analyzing experimental data on SC superalloys with chemistry close to that of the René N6 third-generation SC superalloy. It allows for calculation of the parameter SRZ (%) which is the linear percentage of SRZ in the superalloy below a b-NiPtAl coating after a hold for 400 h at 1093 °C:

GPR ¼ ½C Ti þ CTa þ ðC W =2Þ=CAl

½SRZð%Þ1=2 ¼ 13:88 C Re þ 4:10 C W  7:07 CCr  2:94 C Mo  0:33 C Co þ 12:13

where CX is the atomic percent content of the element X in the alloy.

where CX is the atomic percent content of the element X in the alloy.

Alloy Oxidation Resistance

Castability

Hot corrosion and high temperature oxidation resistances are key parameters for component durability particularly for uncoated surfaces or hot spots where coating can be exposed to extreme environment. For HP turbine blade superalloys, the main issue concerns oxidation resistance which is linked to their aluminum and chromium contents. These elements promote the formation of a protective thermally grown oxide layer (mainly Al2O3). The addition of chromium enhances the formation and stability of this alumina protective layer. However, chromium addition is limited by rhenium and tungsten additions, as all these elements partition preferentially to c phase and promote TCP phase formation. Thus, a balance has to be made between those elements to retain good oxidation and creep properties while avoiding precipitation of TCP phases. Otherwise, minor additions of silicon and hafnium have shown to improve high temperature oxidation resistance [13, 14], as well as reactive elements such as yttrium or lanthanum [15, 16]. As the sulfur content in nickel-based superalloys is a critical parameter for oxidation resistance [17], it was considered in the AGAT alloy design. Aluminum and chromium activities (aAl and aCr) in the alloy at service temperature (from 900 to 1150 °C) were evaluated from Thermo-Calc calculations. Knowing the experimental oxidation resistance of reference commercial alloys, activities of chromium and aluminum in these alloys were used to evaluate oxidation resistance of AGAT. A higher value of chromium activity has been considered to be in favor of better oxidation resistance.

Castability of the new superalloy was estimated by calculating its propensity to freckle formation through the use of the no-freckles parameter (NFP) defined by Konter et al. for SC nickel-based superalloys with a rhenium content close to 3 wt% [19]. The NFP value must be  0.7 and  1 to reduce the risk of freckling during casting of such superalloys:

Compatibility with Coating Compatibility with an alumina-former metallic coating was a key objective in the development of AGAT alloy. Proneness to SRZ formation under an usual b-NiPtAl metallic coating was estimated using an empirical formula developed by

NFP ¼ ðCTa þ 1:5 C Hf þ 0:5 C Mo  0:5 C Ti Þ=ðC W þ 1:2 C Re Þ where CX is the weight percent content of the element X in the alloy.

Results and Discussion Alloy Design Taking into account the different criteria described above and a detailed analysis of the literature data, seven experimental alloys were designed and cast by vacuum induction melting (*10 kg heat for each alloy). Single crystal bars with a 〈001〉 growth orientation were cast at the laboratory scale at ONERA using conventional Bridgman process in order to perform screening tests and to select the alloy best satisfying the targeted specifications. The properties of these experimental alloys cannot be detailed here. This article thus focuses on the presentation of the selected AGAT alloy. The rhenium content was set intermediate between the values of 3 and 6 wt%, characterizing, respectively, the second- and third-generation SC superalloys. The objective was to reduce the drawbacks associated with high levels of rhenium, such as excessive cost and density, and proneness to TCP phase and SRZ formation, while maintaining a high level of mechanical strength at high temperatures. This

34

J. Rame et al.

allowed introducing a higher chromium content than in CMSX-10 for example that is favorable to environmental resistance. The cobalt content was set at a quite high level to promote microstructural stability (TCP phase and SRZ) even if that is contrary to a high c′ solvus temperature. The level of tantalum was kept high to promote both low and high temperature mechanical strengths. Silicon and hafnium were added to ensure a high oxidation resistance and a very low content of sulfur was imposed for a strong TBC adherence. The chemical composition (Table 1) and calculated properties (Table 2) of the selected AGAT alloy are compared with those of the commercial reference superalloys AM1 [7], CMSX-4 [20] and CMSX-10 (K version) [21]. These alloys are respectively first-, second-, and third-generation superalloys currently used for SC turbine blade applications in several civil or military flying engines. The computed AGAT density is lower than that of CMSX-10 and higher than the values obtained for CMSX-4 and AM1. The calculated c′ volume fraction Fc′ at 1200 °C in AGAT alloy is higher than in CMSX-4 and AM1 and lower than in CMSX-10. The same ranking is logically observed for the calculated values of the c′ solvus temperature. The creep strength at 1200 °C of the AGAT alloy was then expected to be less than for CMSX-10 but higher than that of CMSX-4. The GPR criterion indicates a stronger c′ resistance in AGAT alloy compared with CMSX-4 or CMSX-10 alloys that should be beneficial to the tensile strength at low temperature. According to the results of Thermo-Calc and Md calculations, microstructure should be stable in AGAT alloy with a low tendency to TCP phase precipitation. Based on the aCr values at 1150 °C, AGAT alloy was expected to exhibit slightly lower oxidation resistance than CMSX-4 and higher than CMSX-10. The negative value of [SRZ(%)]1/2 for CMSX-4 is typical of alloys not prone to SRZ formation under a b-NiPtAl coating, while the high positive value calculated for CMSX-10 is indicative of a strong tendency to form SRZ that was actually reported for this alloy [21]. The weak positive value calculated for AGAT suggests a low proneness to SRZ formation. Finally, based on NFP criterion calculation, proneness of AGAT alloy to freckling was estimated to be lower than for CMSX-4 and CMSX-10. The values of [SRZ (%)]1/2 and NFP were not calculated for AM1 as this alloy does not contain rhenium.

Alloy Casting Several 150 kg heats of the AGAT alloy were produced using a low sulfur vacuum induction melting process for the alloy development program. AM1, CMSX-4, and CMSX-10 commercial alloys were provided by an industrial manufacturer. Single crystal parts and bars for AGAT and reference alloys (〈001〉 growth orientation) were cast at the Safran Aircraft Engines foundry (Gennevilliers, France) using conventional Bridgman process. Single crystal bars were used for heat treatment optimization and material characterization. Single crystal turbine blades were produced to assess AGAT alloy castability in comparison with commercial superalloys. The components were manufactured using conventional lost-wax casting process used for industrial production of turbine blades. Dozens of parts were cast and then inspected by non-destructive testing methods. No foundry defects such as freckles were found after chemical etching of the parts or test bars. Good castability of AGAT alloy was thus confirmed in accordance with the NFP criterion.

Heat Treatments Heat treatments were evaluated to obtain an optimal microstructure maximizing the mechanical strength. Such a microstructure has to be free of interdendritic c/c′ eutectic pools and of incipient melting events. A regular distribution of c′ precipitates with a mean size within the range 400– 500 nm was aimed to optimize the creep strength [22]. Differential thermal analyses indicate AGAT c′ solvus and solidus temperatures of respectively 1310 and 1340 °C. Thus, the solution heat treatment (SHT) window is 30 °C (incipient melting temperature minus c′ solvus temperature) making AGAT alloy heat treatment compatible with conventional industrial furnaces. A 10 h SHT at 1335 °C with a 300 °C min−1 cooling rate is able to suppress the c/c′ interdendritic eutectic pools and to obtain a fine distribution of c′ precipitates. Due to the heating ramp to 1335 °C (to avoid incipient melting), total duration of the SHT was about 25 h. A large SHT window with reasonable dwell temperature and duration compared with reference alloys is beneficial for heat treatment of industrial components with a

Table 1 Chemical composition of AGAT, AM1, CMSX-4, and CMSX-10 alloys (in wt%) Alloy

Ni

Cr

Mo

Co

W

Re

Al

Ti

Ta

Hf

Others

S

AM1

Bal.

7.5

2

6.5

5.5

–

5.3

1.2

8

0.05

–

CMSX-4 and CMSX-10), – Fatigue properties similar to CMSX-4, – Stable microstructure (no TCP phases), – Solution heat treatment at 1335 °C with a large SHT window (30 °C), – Good castability (no freckles), – Low sensitivity to SRZ formation under a b-NiPtAl coating (0.5) at temperatures ranging from 1050 to 1200 °C [8] in contrast to conventional Ni-based SX superalloys. Thus, the target was to enhance mechanical properties at near solvus temperature and to increase oxidation resistance by enhancing the formation of protective oxide layer [10]. Compared to rhenium (2867 $ kg−1) or ruthenium (9107 $ kg−1), the main drawback of platinum (25,571 $ kg−1) is a higher cost [11]. However, rhenium and ruthenium are strategic materials which can be subjected to shortages while platinum is a more convenient material to supply due to higher worldwide production [12]. Moreover, in the process development of a turbine blade or vane, platinum is commonly used in bond coats (e.g., NiPtAl) to enhance their oxidation resistance [13]. Adding platinum directly in the alloy composition could reduce or even suppress the need for further addition of platinum during the bond coating process. The following sections present and discuss alloy design criteria, material processing, solution heat treatment (SHT) optimization, and mechanical and oxidation properties.

Alloy Design Procedure The alloy design procedure was oriented to target the following specifications: (i) density < 8.9 g cm−3; (ii) high oxidation and corrosion resistance; (iii) compatibility with usual bond coats; (iv) heat treatment window compatible with industrial standards; (v) good castability; (vi) high mechanical strength close to alloy solvus temperature; (vii) reasonable alloy cost. Alloy properties were simulated using empirical criteria or thermodynamic modeling to guide the design toward the targeted specifications. CALPHAD («CAlculation of PHAse Diagrams») models using Thermo-Calc Software [14] (TCS Ni-based superalloys database v9) were employed for thermodynamic property predictions at equilibrium such as transformation temperatures (c′ solvus, solidus, liquidus), volume fraction of each phase, and activities of elements. Alloy density is a key parameter for rotating components such as turbine blades. Indeed, considering the rotational

J. Rame et al.

speed of high-pressure (HP) turbine blade (from *10,000 rpm for large civil turbofans up to *50,000 rpm for turboshaft engines for helicopters during take-off conditions), such components experience tremendous levels of centrifugal force during service. The Hull formula allows for density estimation of simulated compositions based on a mixing law whose parameters were fitted using experimental data [15]. Here, alloy density was predicted from a derivative of Hull formula [16] whose coefficients were tailored to take into account the addition of platinum element. Hot corrosion and oxidation resistance are key parameters to consider when designing SX superalloys for HP turbine blades or vanes. Indeed, such components are exposed to severe environments considering the temperature and composition of gas at the exit of the combustion chamber [4]. Oxidation and corrosion resistance of the Ni-based superalloys in the temperature range of 900–1200 °C are strongly linked to their aluminum and chromium contents, which promotes the formation of protective Al2O3 and Cr2O3 layers [17]. Addition of chromium assists in forming a protective and stable layer of a–Al2O3 for a given aluminum content. Thus, as a first approximation, chromium and aluminum contents in the alloy in the 900–1200 °C temperature range were used to assess oxidation and hot corrosion resistance. These quantities were predicted using Thermo-Calc (mass content of each element in the c phase). HP turbine components are commonly used in combination with bond coats to improve their environmental resistance and to enhance bonding of thermal barrier coating. However, interactions between the coating and the alloy substrate can lead to instabilities such as “secondary reaction zone” (SRZ) [18] depending on the alloy and bond coat chemistries. SRZ drastically reduces mechanical properties of the component by disrupting the SX superalloy microstructure. Thus, compatibility with coating was considered in the alloy development. Proneness to SRZ formation beneath usual NiPtAl metal coating was checked using the Walston et al. formula [19]. The SRZ% criterion indicates the percentage of SRZ in the superalloy below coating; a high SRZ % value increases the propensity to SRZ formation. Solution heat treatment window is an important parameter that should be considered when designing a SX superalloy. Indeed, the temperature and duration of this treatment have strong impact on final cost of a turbine blade or vane. Moreover, the SHT window must be compatible with industrial furnace specifications in order to fully solution c′precipitates while avoiding incipient melting. The SHT window is then taken as the incipient melting temperature minus the c′ solvus temperature; both temperatures were simulated with Thermo-Calc Software. Alloy castability is also a significant parameter for industrial applications, having a direct impact on scrap rates

Platinum-Containing New Generation Nickel-Based Superalloy …

73

of produced parts. Freckles are one of the main defects found in blades or vanes produced by investment casting [20]. They form due to thermosolutal convection occurring in the interdendritic liquid region during solidification. Density inversion in the mushy zone which is related to interdendritic segregations is the source of this specific flow. Interdendritic segregation is closely linked to alloy chemistry. Thus, several works report studying and modeling freckle development [20–22]. In TROPEA design, the assessment of freckle formation (no-freckle parameter; NFP) was estimated using Konter’s formula [23]. An NFP  0.7 reduces the risk of freckle formation during casting. Mechanical strength was assessed by considering different parameters: c solid solution strengthening, c′ precipitate volume fraction, c/c′ lattice mismatch, c′ solvus and alloy solidus temperature, c′ coarsening rate, and microstructure stability. Matrix strengthening was tuned via rhenium, molybdenum, cobalt, and tungsten additions. Thermodynamic data such as c′ solvus, solidus, and volume fraction of c′ precipitates at high temperature were simulated with Thermo-Calc Software. It has been shown that a 60–70% volume fraction of c′ phase was optimal for creep resistance [24]. Moreover, high solvus temperature alloys promote high creep strength by increasing residual c′ volume fraction at high temperature [25]. Long-term microstructural stability was evaluated through the tendency to form topologically closed-packed (TCP) phases (by Thermo-Calc), given the fact that these deleterious phases have a negative impact on an alloy mechanical strength. The lattice mismatch between c and c′ phases is an important parameter to consider, having a direct impact on c′ particle morphology. On the one hand, high misfit that disrupts coherency of c/c′ interface promotes cuboidal shape precipitates. On the other hand, close to zero misfit leads to spherical precipitates [26]. Moreover, the absolute value and sign of misfit strongly determine the directional coarsening under stress at high temperature [27]. Thus, c/c′ mismatch (both at 25 °C and at 1100 °C) was considered and calculated from JMatPro® software [28].

Coarsening kinetics of precipitates strongly affects the creep properties and is linked to alloy chemistry. Indeed, it is well known that refractory elements such as rhenium slow down coarsening kinetics at high temperature [29–31]. Moreover, it has been shown that platinum decreases interdiffusion rates above 1150 °C, hence limiting precipitate coarsening [7]. Alloy cost was evaluated according to mass percentage and market price of alloying elements. Obviously, this is a significant parameter to consider when designing a new alloy for industrial applications.

Results Design of TROPEA Alloy Nominal chemistries of TROPEA and reference alloys are reported in Table 1 as well as the actual chemistry of TROPEA cast SX bars measured by inductively coupled plasma atomic emission spectroscopy (ICP-AES) [32]. The composition of several alloys used for reference is also given. CMSX®-4 [33] and René N5 [34] alloys are second-generation SX superalloys largely employed in the hot sections of aero-engines and industrial gas turbines. René N6 [35] and CMSX®-4 Plus [36] are third-generation superalloys. PX-5 is an experimental alloy containing rhenium and platinum developed by Van Sluytman et al. [7]. The TROPEA chemistry was tailored by the parameters and criteria previously described. The targeted density was lower than that of third-generation superalloys. Chromium content was maintained to the levels typical of second-generation SX superalloys, to maintain good oxidation and corrosion resistance compared to third-generation superalloys. Hafnium was added to promote oxidation resistance. Titanium and tantalum were added to strengthen c′ precipitates at high temperature. Platinum was added to decrease coarsening kinetics of precipitates at high temperature, to stabilize c′ volume fraction near c′ solvus

Table 1 Chemical composition of TROPEA, CMSX®-4, CMSX®-4 Plus, René N5, René N6, and PX-5 Ni-based superalloys (in wt%) Alloys

Density

Ni

Cr

Mo

TROPEA

8.83

Bal.

6.50

0.60

6.41

0.59

CMSX®-4

8.70

Bal.

6.4

0.6

René N5*

8.63

Bal.

6.9

1.7

7.6

CMSX®-4 Plus

8.93

Bal.

3.5

0.6

10

René N6*

8.97

Bal.

4.4

1.1

PX-5**

–

Bal.

6.2

1.5

TROPEA ICP

*0.05% C, 0.004% B **0.02% C, 0.015% B; 0.02% Zr; 0.2% Si

Co

W

Re

Al

Ti

Ta

Pt

Hf

S (ppm)

6.00

1.00

5.60

1.00

9.00

2.00

0.10

–

8.94

6.06

0.99

5.41

1.00

9.10

1.95

0.09

6.9

9.7

6.4

2.9

5.6

1.1

6.5

0

5.1

3.4

6.3

0

6.7

0

0.17

*2

6.0

4.8

5.7

0.85

8.0

/

0.1

70 at.%

Cr

Phases present

Bal.

Bal.

Bal.

Bal.

Bal.

5 0 20

20

11

6

12

25 30 15

18

22

27

32

c + oP6*

–

–

–

–

–

–

71.4

10.1

18.5

70.9

10.6

18.5

P

–

–

–

48.4

9.1

42.5

46.1

9.1

44.8

–

–

–

c

73.6

5.2

21.1

74.1

4.9

21.0

74.6

5.4

21.0

NiMo

50.5

3.9

45.7

48.2

6.9

44.9

50.2

4.0

45.8

c

75.2

0.0

24.8

–

–

–

–

–

–

–

–

–

NiMo

52.8

0.0

47.2

–

–

–

–

–

–

–

–

–

c

65.0

20.7

14.3

65.9

20.7

13.4

67.5

20.3

12.2

69.7

21.3

9.1

c + oP6*

–

–

–

–

–

–

–

–

–

66.3

20.8

12.9

P

–

–

–

42.2

19.5

38.3

43.7

19.0

37.4

42.9

19.9

37.2

c

64.1

20.7

15.1

65.7

20.9

13.4

–

–

–

68.4

20.9

10.6

c + oP6*

–

–

–

–

–

–

–

–

–

63.7

20.6

15.7

P

41.9

19.6

38.5

41.4

19.6

39.0

–

–

–

41.9

19.6

38.5

69.4

10.9

19.8

45.3

10.4

44.3

71.3

6.4

22.3

c

68.8

11.4

19.8

69.8

13.5

16.6

–

–

–

oP6*

–

–

–

–

–

–

–

–

–

P

45.5

10.6

43.9

43.8

12.5

43.7

c

71.5

6.6

22.0

73.4

6.5

20.1

–

–

–

c + oP6*

–

–

–

–

–

–

–

–

–

P

–

–

–

47.8

5.5

46.7

–

–

–

NiMo

50.4

4.9

44.7

49.6

4.7

45.7

–

–

–

50.1

5.0

44.8

c

67.9

12.9

19.2

69.8

13.0

17.2

–

–

–

–

–

–

oP6*

–

–

-

–

–

–

–

–

–

67.7

12.8

19.5

P

44.6

11.8

43.6

43.9

12.0

44.1

–

–

–

44.8

11.9

43.3

*Average composition of matrix –Not analyzed

lower temperatures. Let us consider the case I (■) in Fig. 12 where the high-temperature c phase decomposes into three phases of c, oP6 and TCP. The terminal composition of each phase in the three-phase tie-triangle corresponds to the point of c, d, and a. As shown in Fig. 3b, TCP phase preferentially precipitates at grain boundaries with precipitate free zone (PFZ) of the oP6 phase,

indicating a local equilibrium between c and TCP phases there. This gives a tie-line ac by compositional analysis of each phase. Then, by measuring the volume fraction of the TCP phase precipitated, point b can be determined by drawing a line (dotted) from the point a crossing through the point I(■) using a lever rule. Then, we can determine the point d, the terminal composition of the oP6 phase, by

Phase Equilibria Among A1/TCP/GCP Phases and Microstructure …

139

Fig. 12 A schematic illustration showing how to determine the composition of oP6 single-phase region

drawing a line from the point c crossing through the point b, by measuring the volume fraction of the oP6 phase in the area away from the grain boundaries. The same method is applied to determine the terminal compositions of the points e and f for II(■) and those of g, i and j for III(■). Thus, the phase boundary of oP6 phase in equilibrium with c phase can be determined, but the other side of the phase boundary cannot be determined at this moment because the c phase of 11Cr–22Mo and 12Cr–32Mo at 1473 K completely transforms to oP6 phase at 973 K (Table 1). However, it is unlikely that a large atom of Mo (rMo = 1.40 Å) can occupy the small sublattice site of Ni (rNi = 1.25 Å), suggesting almost no expansion toward the Ni-poor side beyond the equi-nickel line of 66.6 at.%. From these results, it is apparent that the congruent temperature of the binary Ni2Cr phase (873 K) increases by more than 200 K by replacing Cr with 18 at.% Mo, and it transforms to c + P phases through a peritectoid reaction (oP6 ! c + P), thereby leading to the formation of two three-phase regions of c + oP6 + P and c + oP6 + NiMo. When Mo solute atoms are in solutioned in Ni2Cr, this tremendously increases the phase stability of the oP6 structure. It should also be noted that none of the commercially available thermodynamic databases can reproduce the experimentally determined isothermal sections where the oP6 phase exists by calculation, and the database has to be modified. Such efforts are currently in progress in our group.

Fig. 13 A vertical section at 15 at.% Cr in Ni–Cr–Mo ternary system

Microstructure Formation In the previous sections, it was found that the c phase is in equilibrium with TCP phases at higher temperatures and the c phase decomposes to oP6 phase by aging at lower temperatures. Figure 13 shows a vertical section at 15 at.% Cr, nearly parallel to the tie-line between the c and P phases, in this system, drawn based on the isothermal sections at different temperatures. This vertical section opens many possibilities to design various novel microstructures by multi-step heat treatments within a certain composition range of Mo, as schematically shown in Fig. 14, where after homogenization in the c single-phase region (a), the grain boundary can be decorated with P phase through a heat treatment in c + P two-phase region [step I (b)], followed by aging at lower temperatures in the region of either c + oP6 two-phase or oP6 single phase to form the oP6 phase within the grain interiors [step II (c)]. In step I, the grain boundary area decorated by TCP phases can be precisely tuned from 0% to 100% by changing the degree of Mo supersaturation in the c matrix along with the temperature and time of aging in the two-phase region. It is interesting to note that the oP6 phase exists as a Kurnakov-type compound, indicating that spinodal decomposition from c phase to oP6 phase would take place. This could further widen the variety of

140

R. Nagashima et al.

Fig. 14 Schematic illustrations of novel microstructures to be designed by multi-step heat treatment: a solution treated, b decoration of grain boundaries with P phase (step I), c formation of oP6 phase within grain interiors (step II)

microstructures that can be engineered during step II; the oP6 phase precipitates by nucleation and growth mechanism in the two-phase region outside the spinodal line, whereas it could form a modulated microstructure at compositions that reside inside the region. The microstructure shown in Fig. 7a would be formed by the former mechanism, whereas the others shown in Figs. 3 and 7b might be formed by the latter mechanism. Ongoing investigations are being conducted to detail the decomposition process and will be reported in future studies.

Summary Phase equilibria among the A1 (c-fcc), Ni2Cr (oP6) and TCP phases in Ni–Cr–Mo system at temperatures ranging from 973 K to 1473 K have been investigated. A thermodynamically stable Ni2Cr single-phase region exists as an island up to about 1073 K at a composition of around Ni–13Cr– 19Mo, which is approximately 200 K higher than that found in the binary Ni–Cr system. The oP6 phase decomposes to c and P (oP56) phase through a peritectoid transformation (oP6 ! c + P), thereby leading to two distinct three-phase regions of c + oP6 + P and c + oP6 + NiMo. The decomposition product of the oP6 phase coherently forms in c matrix, having an orientation relationship of {110}c// (100)oP6, c//[010]oP6, its morphology changes from Widmansttaten-type to tweed-type with an increase in the supersaturation of Mo in the prior c phase. In contrast, the P phase preferentially precipitates along the c grain boundaries. From these results, an innovative approach for designing the composition and engineering the microstructure of novel Ni-base superalloys has been proposed such that TCP phases can be employed to decorate the grain boundaries and fine oP6 phase precipitates are present within the grain interiors. Combining grain boundary precipitation strengthening along with conventional precipitate

strengthening within the grain interiors may enable advances in the development of new classes of Ni-base superalloys.

References 1. Watanabe R, Chiba Y, Kunou T (1975) Alloy design of solid solution strengthened nickel-base superalloy. Tetsu-to-Hagane 61 (10):2405–2414. 2. Yukawa N (1983) New PHACOMP for superalloys. Denki-Seiko 54 (4):275–287. 3. Seiser B, Drautz R, Pettifor DG (2011) TCP phase predictions in Ni-based superalloys: structure maps revisited. Acta Mater. 50 (2):749–763. 4. Takeyama M (1986) Ph.D. theses, Tokyo Institute of Technology. 5. Takeyama M et al. (1986) Effect of grain boundary precipitates on high temperature creep properties of Ni-20Cr-Nb-W alloys. Tetsu-to-Hagané 72 (10):1605–1612. 6. Matsuo T, Kikuchi M, and Takeyama M (1991) Strengthening mechanisms of Ni-Cr-W based superalloys for very high temperature gas cooled reactors, Proc. 1st Inter. Conf. Heat-Resistant Materials: 601–614. 7. Takeyama M (2008) Fundamentals of physical metallurgy for high-temperature materials -Microstructure control and design-. Paper presented at the ISIJ 194th and 195th Nishiyama Memorial Lecture, Tokyo and Osaka, Jun.:1–23. 8. Tarigan I et al. (2011) Novel concept of creep strengthening mechanism using grain boundary Fe2Nb Laves phase in austenitic heat resistant steel. Materials Reseach Society Symp., Proc., 1295:317–322. 9. Hasebe Y, Hashimoto K, Takeyama M (2011) Phase equilibria among c-Fe/Fe2Nb(TCP)/Ni3Nb(GCP) Phases in Fe-Ni-Nb ternary system at elevated temperatures. J. Jpn. Inst. Met. 75 (4): 265–273. 10. Takeyama M (2012) Recent trend on materials development for A-USC power plants. Denki-Seiko 83 (1):27–33. 11. Takeyama M et al. (2014) Creep of the novel austenitic heat resistant steels strengthened by Fe2Nb Laves phase under steam condition. Proc. 3rd International ECCC Conf., Rome, Italy.: 703– 710. 12. Takeyama M (2017) Materia Jpn. 56 (3): 145–150. 13. Miller MK, et al. (2002) Microstructural characterization of Haynes® 242™ alloy. Mater. Sci. Eng. A 327:89–93. 14. Ul-Hamid A et al. (2004) Cyclic oxidation behaviour of a Ni-Mo-Cr alloy at 800 °C. Anti- Corrosion Methods and Materials 51:339–347.

Phase Equilibria Among A1/TCP/GCP Phases and Microstructure …

141

15. Dymek S et al. (2006) Influence of plastic deformation and prolonged ageing time on microstructure of a Haynes 242 alloy. J. Microsc. 224: 24–26. 16. Ul-Hamid A et al. (2007) Evolution of oxide scale on a Ni-Mo-Cr alloy at 900 °C. Mater. Charact. 58:13–23. 17. Dymek S et al. (2010) Microstructure stability and mechanical properties of an age-hardenable Ni-Mo-Cr alloy subjected to long-term exposure to elevated temperature. Mater. Charact. 61:769–777. 18. Fahrmann M, Srivastava SK, Pike LM (2012) Development of a new 760 °C (1400 °F) capable low thermal expansion alloy. Superalloys 2012, ed. E.S. Huron et al. (Warrendale, PA: TMS, 2012) 769–777. 19. Fahrmann MG, Srivastava SK, Pike LM (2014) Haynes 244 alloy a new 760 °C capable low thermal expansion alloy. MATEC Web of Conferences 14, 17004. https://doi.org/10.1051/matecconf/ 20141417004. 20. Nash P (ed) (1991) Phase diagram of binary nickel alloys. ASM International, Materials Park, Ohio. 21. Turchi PEA, Kaufman L, Liu ZK (2006) Modeling of Ni-Cr-Mo based alloys: Part I-phase stability. CALPHAD 30:70–87.

22. Rideout S et al. (1951) Intermediate phases in ternary alloy systems of transition elements. JOM 3:872–876. 23. Bloom DS, Grant NJ (1954) An investigation of the systems formed by chromium, molybdenum, and nickel. JOM 6:261–268. 24. Raghavan M et al. (1984) Determination of isothermal sections of nickel rich portion of Ni-Cr-Mo system by analytical electron microscopy. Metall. Mater. Trans. A 15A:783–792. 25. Zhou SH et al. (2005) First-principles calculations and thermodynamic modeling of the Ni-Mo system. Mater. Sci. Eng. A 397(1– 2):288–296. 26. Brink C, Shoemaker DP (1955) A variation on the a-phase structure: the crystal structure of the P phase, Mo-Ni-Cr. Acta Crystallogr. 8:734–735. 27. Sinha AK, Buckley RA, Hume-Rothery W (1967) Equlibrium diagram of the iron-molybdenum system. J. Iron Steel Inst., London 205:191–195. 28. Virkar AV, Raman A (1969) Alloy chemistry of r (b-U)-related phases. II. Z. Metallkd. 60:594–600.

A New Co-free Ni-Based Alloy for Gas Turbine and Exhaust Valve Applications Karl A. Heck, Ning Zhou, Samuel J. Kernion, Danielle Rickert, and Filip Van Weereld

Abstract

Introduction

A new Ni-based cast-and-wrought alloy designed for high temperature strength, stability, notch ductility, and minimal elevated temperature dwell fatigue crack growth rate has been laboratory developed and scaled up in the mill. Based on studies of a relatively wide compositional space, a Co-free composition was selected. Other properties, including tensile strength, stress rupture, creep, low cycle fatigue, oxidation, and sulfidation resistance, as well as hot workability were also studied. Several heat treatments were developed to achieve property balances as appropriate for various end-user applications such as those limited by damage tolerance, creep, fatigue, or tensile strength. Microstructural stability after extended exposure at temperatures ranging from 704 to 871 °C was studied. Multiple 12-ton heats have been successfully processed as VIM/VAR and VIM/ESR ingots converted into forged billets, rolled bar, and strip products. Potential applications for this alloy include turbine disks, gas turbine engine casings, high temperature fasteners, heavy duty diesel engine exhaust valves, and high temperature gaskets. This paper discusses alloy and process development as well as microstructure–property relationships of the alloy. Keywords













Dwell fatigue crack growth Stress rupture Microstructure Stability Oxidation resistance Gamma prime Grain size

K. A. Heck
N. Zhou (&)
S. J. Kernion
D. Rickert
F. Van Weereld Carpenter Technology Corporation, Reading, PA, USA e-mail: [email protected] K. A. Heck e-mail: [email protected]

Structural Ni or Ni–Fe-based superalloys that are designed to operate at elevated temperatures (i.e., >600 °C) typically require high strength and creep resistance. However, as strength and creep resistance are increased in such alloys, they can become more susceptible to environmental effects such as those aggravated by oxygen in the atmosphere [1]. This susceptibility can manifest itself as notch brittleness and/or an increase in crack growth rate [2]. With regard to crack growth rate, nickel-base superalloys may be relatively resistant to crack propagation when fatigue cycled at a relatively fast rate, but exhibit increased growth rate when the alloy is stressed at low frequencies with a dwell hold during each load cycle [3]. One theory for such sensitivity is a change in the crack propagation mechanism with increased load dwell times which provides time for oxygen to diffuse down grain boundaries to form an oxide layer within the crack. That oxide layer then may act as a wedge when the load is released, advancing the crack tip movement at a faster overall rate [4]. In nickel-base superalloys, in addition to oxidation behavior, compositional and structural factors influence strength and creep resistance properties and thus crack growth rate. Such factors include the effects of solid solution strengthening, precipitation strengthening (such as c′ precipitation); anti-phase boundary energy of the precipitates; volume, size, and coherency of the precipitates in the matrix; grain size; grain boundary structure (precipitation, composition, morphology, and mismatch) [1, 5]; as well as low levels of certain potent elements in the grain boundaries. Localized strain at the crack tip relieved via creep relaxation also promotes crack tip blunting [6, 7]. Thus, it is desirable to have a nickel-base superalloy that not only provides high temperature strength and creep resistance, but also resistance to crack growth during stress cycling in oxidizing environments. One of the key objectives for this study was to achieve dwell fatigue crack growth

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_14

142

A New Co-free Ni-Based Alloy for Gas Turbine and Exhaust …

143

resistance equivalent to or greater than that of Waspaloy at 704 °C with a 90–100 s hold time at maximum stress intensity factor and stable strength.

Alloy Design Methodology A compositional space predominantly outside that of existing commercialized alloys was studied with the range shown in Table 1. Thermodynamic modeling with JMatpro Ni-Fe database (Version 6.2.1) was used to screen compositions to avoid detrimental phases and target c′ as the stable strengthening phase. Multiple 18 kg heats were vacuum induction melted and cast into ingots. The ingots were homogenized at 1177 °C for 24 h then forged from 1177 °C into bars (2 cm by 4.4 cm cross-section) with a 500-ton hydraulic press. Forging temperature was lowered somewhat in alloys with high refractory content when cracking was experienced. Annealing and aging studies were conducted to determine c′ solvus temperatures and time-temperature-hardness behavior to predict strength as a function of composition and heat treatment. This also served to establish annealing temperatures to produce similar microstructures in the various compositions for equivalent comparison of grain size dependent properties. An example is shown in Fig. 1 to demonstrate how base Ni and Co content of a group of eight alloys impacts the choice of annealing temperature as measured by the onset of grain growth and drop in hardness at the c′ solvus temperature. Dwell fatigue crack growth experiments were performed with compact tension specimens according to ASTM E647. Samples were annealed at 1010 °C for 1 h followed by a two-step age: 843 °C/4 h and 732 °C/16 h. This aging treatment shortens the total time required to achieve peak hardness compared to a single isothermal aging step at the peak hardening temperature 732 °C. Tests were conducted at 704 °C in laboratory air with R = 0.1 and a 100 s hold at peak load. During testing, it was found that the series of heats with 10% Cr had relatively good 704 °C cracking resistance at 10 Hz, but behavior began to differentiate from 15% Cr heats with 10 s holds. The 10% Cr heats cracked very rapidly when using a 100 s dwell time. Heats containing 15% Cr showed much better hold time cracking resistance, which indicates the impact of oxidation resistance on dwell fatigue.

Fig. 1 Anneal temperature for onset of grain growth and hardness drop as a function of Ni, Co content (all contain about 15% Cr, 1.8% Ti, 2% Al, 3% Nb, 1% W, Fe as balance)

Tensile tests at 21, 538, and 704 °C as well as 704 °C/551.6 MPa stress rupture tests were also performed to determine the balance of tensile and creep performance vs. dwell time crack growth resistance. Results from 15% Cr heats demonstrate that dwell crack growth rates are inversely related to creep resistance: alloys high in W and/or Co have greater creep resistance, but exhibit faster crack growth rates; while alloys without Co but relatively high in Mo had shorter stress rupture lives, they exhibited much slower crack growth rates. A heat of the chosen base composition was also produced with 18% Cr and was found to embrittle after long time exposure at 704 °C; likely due to sigma phase formation as predicted by JMatPro. Thus, 15% Cr and 4% Mo were selected to maintain alloy phase stability. About 2% Al and 1.8% Ti were selected for c′ formation. Al was kept at a relatively high level to interact with Cr for oxidation resistance. Ti, while a potent strengthener, also causes rapid age hardening and tends to make thermo-mechanical processing and welding more difficult. While Ti dramatically increases tensile strength, it also promotes subsurface internal oxidation and nitridation at higher temperature in air [8]. In addition, 3.7% Nb was added for solid solution strengthening and to compliment Al and Ti for c′ strengthening. The nominal composition of the selected alloy EXP-G27 is shown in Table 2. This alloy is cobalt-free, although cobalt and additional refractory elements can be added if higher creep strength is desired at the expense of crack growth resistance, stress rupture notch ductility and greater volatility in raw material costs.

Table 1 Alloy design composition range Element

Ni

Co

Cr

Al

Nb

Ti

Mo

W

Fe

B

C

Zr

Mg

wt%

Balance

0–10

10–15

1–3

0–4

0–3

0–8

0–8

0–15.5

0.005

0.03

0.03

0.001

144 Table 2 Nominal composition of EXP-G27 (in wt%)

K. A. Heck et al. Element

Ni

Fe

Cr

Al

Nb

Ti

Mo

Zr

B

C

Mg

wt%

Balance

15.3

15.0

2.0

3.7

1.8

4.0

0.03

0.005

0.03

0.001

Fig. 2 Dwell fatigue crack growth results at 704 °C, R = 0.1, 100 s dwell

The effects of grain size were also studied by altering the annealing temperature. c′ supersolvus annealing at 1135 °C for 1 h results in coarse grain, while c′ subsolvus annealing at 982 °C for 1 h retains the fine grain in the forged sample. The annealing temperature not only affects grain size but also alters c′ morphology, which will be described later. It was discovered that coarse grain not only increases rupture life but also significantly improves the crack growth resistance compared to fine grain structures. The dwell fatigue crack growth test results for both fine grain and coarse grain EXP-G27 are shown in Fig. 2. The phenomenon that coarser grain size can slow down dwell fatigue crack growth rate at 704 °C was observed for other candidate alloys as well as EXP-G27. It is hypothesized that crack tip relaxation during 704 °C dwell is related to lowered yield strength (reduced Hall–Petch strengthening).

Properties After a comprehensive study of grain size and hardness response from heat treatment temperature and time, alloy EXP-G27’s microstructure stability, mechanical properties (tensile, creep, fatigue), oxidation, and sulfidation resistance were evaluated. A fine grain heat treatment with an annealing at 1010 °C (1 h with oil quench) then aging (843 °C/4 h, air cool + 732 °C/16 h, air cool) provides higher strength. This heat treatment develops a uniform grain structure (Fig. 3a) and fine, uniform c′ (Fig. 3b). A coarse grain heat treatment

utilizing a supersolvus anneal of 1133 °C/1 h oil quench and aging at 982 °C/4 h oil quench + 843 °C/4 h, air cool + 732 °C/16 h, air cool maximizes creep and dwell fatigue crack growth resistance. This treatment develops a coarse grain structure (Fig. 3c) and a bimodal c′ particle size distribution (Fig. 3d). The bimodal c′ size distribution offers higher high temperature tensile ductility compared to the structure with a uniform fine c′ size distribution. Variations of the fine grain heat treatment, such as 1010 °C/1 h oil quench + 857 °C/4 h, air cool + 732 °C/4 h air cool, were developed to reduce the total heat treatment time with minimum impact on the properties. The equilibrium c′ volume percent for both fine grain (Fig. 3b) and coarse grain heat treatment (Fig. 3d) is about 25% and the c′ solvus temperature is about 1000–1010 °C. Microstructural stability was also characterized. Coarse grain heat treatment samples were exposed to temperature from 704 to 871 °C for 500 to 10,000 h. A transition of the c′ size distribution from bimodal to unimodal between 760 and 815 °C after 2500 h can be seen in Fig. 4. While strength is lowered by high temperature exposure due to the coarsening of c′, the alloy still retains useful strength for some applications. The presence of delta, eta, or sigma phase has not been observed. However, MC and M6C carbides, where M = (Mo, Nb) slowly precipitate at grain boundaries when exposed for extended periods of time as shown in Figs. 5c, d. This results in lowered room temperature ductility after exposure. This behavior has been demonstrated in other established alloys such as alloy 625. The rate and extent to which this occurs is affected by C and Nb content, as well as Si when C is at lower levels [9]. Room temperature and elevated temperature tensile strengths of EXP-G27 in two heat treated conditions (fine grain and coarse grain) are shown in Fig. 6 together with Waspaloy [10] for comparison. Similar to EXP-G27, Waspaloy can be heat treated with a supersolvus solution treatment at 1079 °C to increase grain size for longer stress rupture life (Waspaloy A), or by a subsolvus anneal at 1010 °C for finer grain size and higher tensile strength (Waspaloy B). The strength of fine grain EXP-G27 is higher than fine grain Waspaloy B at all tested temperatures. Coarse grain EXP-G27 has higher strength than Waspaloy A when tested at temperature above 704 °C. Stress rupture results are shown in Fig. 7. Waspaloy B has higher creep strength compared to fine grain EXP-G27. The coarse grain Larson–Miller (LM) curve is slightly lower

A New Co-free Ni-Based Alloy for Gas Turbine and Exhaust …

145

Fig. 3 Grain size and c′ morphology after heat treatments. a and b show fine grain heat treatment (grain size ASTM 6-8 for hot rolled bars), c and d show coarse grain heat treatment (grain size is typically ASTM 0-1)

Fig. 4 Long-term exposure of coarse grain EXP-G27 at various temperatures

146

Fig. 5 Grain boundary morphology after exposure. a 871 °C/2500 h; b 815 °C/2500 h; c 760 °C/5000 h; d 704 °C/10000 h Fig. 6 Tensile strength of EXP-G27 compared with Waspaloy (Waspaloy data from [10])

K. A. Heck et al.

A New Co-free Ni-Based Alloy for Gas Turbine and Exhaust …

Fig. 7 Stress rupture of EXP-G27 compared with Waspaloy (Waspaloy data from [10])

than that for Waspaloy A [10]. It is possible to further improve the stress rupture of EXP-G27 by introducing elements such as Co and W at the expense of dwell fatigue resistance and higher raw material costs. Low cycle fatigue was tested at 650 and 704 °C according to ASTM E606 (axial fatigue, strain controlled sinusoidal waveform) with test frequency of 0.5 Hz, R = 0. A comparison was made with Waspaloy tested at 704 °C [11] and 650 °C [12]. As shown in Fig. 8, the LCF life of Waspaloy is in between that of fine grain and coarse grain EXP-G27. The surface oxidation of an EXP-G27 sample exposed at 872 °C for 502 h is shown in Figs. 9 and 10. The surface

Fig. 8 Low cycle fatigue of EXP-G27 compared with Waspaloy (axial, strain controlled sinusoidal waveform with test frequency of 0.5 Hz, R = 0. Waspaloy data from [11, 12])

147

structure is a mixture of Fe and Ti oxides followed by a continuous Cr oxide subsurface layer. Alumina fingers and Ti nitrides can be seen with the presence of some Mo, Nb rich phases. Similar to Waspaloy [13], the aluminum level in EXP-G27 does not promote the formation of a continuous alumina layer. However, the chromia layer is continuous and dense which hinders the formation of less desirable internal oxidation along grain boundaries. For high temperature applications such as exhaust valve in petroleum/diesel engines and turbine components in aero-engines, sulfidation resistance is important because of the sulfur content in the fuel. Thus, sulfidation testing was performed at 871 °C in laboratory air for 80 h with alloy EXP-G27 exposed to a mixture of sulfates of Ca, Ba, and Na as well as a carbon addition, and compared to Waspaloy. Cross-sections of the exposed samples are shown in Fig. 11. Both alloys exhibited comparable sulfidation resistance by comparing the depth of the oxidation/sulfidation affected region. EDS maps of the EXP-G27 sample are shown in Fig. 12. EXP-G27 expresses Nb and Ti rich sulfides.

Hot Workability Gleeble™ tension tests with strain rate +5.0 s−1 were performed at various temperatures from 900 to 1200 °C to identify the proper hot working window. Figure 13a, b shows the ultimate tensile strength and the reduction of area percent for both EXP-G27 and Waspaloy. The dependence

148

Fig. 9 BSE image of surface oxidation after 872 °C/502 h exposure

Fig. 10 EDS map of the surface layer after 872 °C/502 h exposure

K. A. Heck et al.

A New Co-free Ni-Based Alloy for Gas Turbine and Exhaust … Fig. 11 Sulfidation attack after 80 h exposure at 871 °C in a reducing sulfur-rich environment. a EXP-G27; b Waspaloy

Fig. 12 EDS map of EXP-G27 after 871 °C/80 h sulfidation exposure

149

150

K. A. Heck et al.

Fig. 13 Comparisons of alloy EXP-G27 and Waspaloy’s hot working behavior: a Gleeble tensile UTS at strain rate +5.0 s−1; b Gleeble tensile Ra % at strain rate +5.0 s−1; c Gleeble compressive flow stress at strain rate: −5.0 s−1; d Gleeble compressive flow stress at strain rate—0.1 s−1

of UTS on temperature is similar for both alloys. However, the alloy EXP-G27 has a wider hot working window than Waspaloy (Using RA > 60% as the criteria, EXP-G27 has a window size about 180 °C while Waspaloy’s window size is about 115 °C). Gleeble™ compression tests were also performed at 982 to 1093 °C at different strain rates to obtain the flow stress. Figure 13c shows the comparison between the alloy EXP-G27 and Waspaloy with a compressive strain rate of −5.0 s−1, and Fig. 13d demonstrates that with a compressive strain rate of −0.1 s−1. Both alloys show comparable flow stress dependencies on temperature and strain rate. With assistance from Gleeble testing data and process modeling results, various product forms of EXP-G27 have been successfully manufactured including rings produced from a 305 mm diameter forged billet using an upset, punch and ring roll process (Fig. 14), hot rolled bar (Fig. 15), and cold rolled, annealed strip (Fig. 16). Fig. 14 Upset, punched, and rolled ring made from 305 mm diameter forging billet

A New Co-free Ni-Based Alloy for Gas Turbine and Exhaust …

151

Fig. 15 Hot rolled 25.4 mm diameter bar

Fig. 16 Cold rolled 1 mm thick strip

Summary Alloy EXP-G27 development was largely based on screening by 704 °C dwell fatigue crack growth performance. The property balances and workability of EXP-G27 have been characterized. Strengthened by c′ phase, this alloy has mechanical properties and microstructural stability suitable for high temperature applications. Properties can be further tailored by altering heat treatments for different applications, i.e., coarse grain for optimum creep and dwell crack growth resistance and fine grain for optimum tensile strength and low cycle fatigue resistance. A continuous chromia layer forms during elevated temperature exposure which provides alloy EXP-G27 with good oxidation and sulfidation resistance. Without Co, the raw material cost variability of this alloy is significantly lower than that for Waspaloy. Co can be added to improve stress rupture at the cost of dwell fatigue crack growth resistance. The thermal mechanical processability of EXP-G27 is similar to, if not better than, that of Waspaloy. To date, multiple 12-ton heats were successfully produced through VIM/VAR and VIM/ESR. The final product form of alloy EXP-G27 includes forged billet, bar, coil, and strip.

Acknowledgements The authors want to acknowledge Ms. Emily Holcombe for assisting the chemistry screening experiments and Frisa Forjados, S. A. de C. V. for ring rolling trials.

References 1. Heck K, Smith D, Wells D, Holderby H (1988), “The physical metallurgy of a silicon-containing low expansion superalloy,” Superalloys 1988: 151–160, Edited by S. Reichman, D.N. Duhl, G. Maurer, S. Antolovich and C. Lund, Metallurgical Society of AIME. 2. Bain K, Pelloux R (1984), “Effect of oxygen on creep crack growth in PM/HIP nickel-base superalloys,” Superalloys 1984: 387–396, Edited by M. Gell, C.S. Kortovich and R. Briknell, Metallurgical Society of AIME. 3. Gayda J, Gabb T, Miner R (1985), “Fatigue crack propagation of nickel-base superalloys at 650 °C,” NASA Technical Memorandum 87150, Cleveland, Ohio. 4. He M, Evans A (2010), “A model for oxidation-assisted low cycle fatigue of superalloys,” Acta Materialia 58(2): 583–591. 5. Danflou H, Macia M, Sanders T, Khan T (1996), “Mechanisms of formation of serrated grain boundaries in nickel base superalloys,” Superalloys 1996: 119–127, Edited by R. D. Kissinger, D. J. Deye, D. L. Anton, A. D. Cetel, M. V. Nathal, T. M. Pollock, and D. A. Woodford, The Minerals, Metals & Materials Society.

152 6. Viggo T (2004), “On fatigue crack growth in ductile materials by crack tip blunting,” Journal of the Mechanics and Physics of Solids 52(9):2149–2166. 7. Yu S, Li H, Hardy M, McDonald S, Bowen P (2014), “Mechanisms of dwell fatigue crack growth in an advanced nickel disc alloy RR1000,” MATEC Web of Conferences, 03002(14): 1–6. 8. Chen J.H, Rogers P.M, Little J.A (1997), “Oxidation behavior of several chromia-forming commercial nickel-base superalloys,” Oxidation of Metals 47:381–410. 9. Floreen S, Fuchs G. Yang W (1994), “The metallurgy of alloy 625,” Superalloys 718,625,706 and Various Derivatives: 13–37, Edited by E.A. Loria, The Minerals, Metals& Materials Society. 10. ”Alloy Digest,” (2002), “Unitemp Waspaloy, filing code: Ni-46”, Publisher: ASM International, Materials Park, Ohio, www. asminternational.org.

K. A. Heck et al. 11. Ott E, Groh J, Sizek H (2005), “Metals affordability initiative: application of Allvac alloy 718Plus for aircraft engine static structural components,” Superalloys 718,625,706 and Various Derivatives: 35–45, Edited by E.A. Loria, The Minerals, Metals & Materials Society. 12. Yeom J, Williams S, Park N (2001), “Low-cycle fatigue life prediction for Waspaloy,” Materials at High Temperatures 19 (3):153–161. 13. Forsik A, Polar Rosas A, Wang T, Colombo G, Zhou N, Kernion S, Epler M (2018), “High-temperature oxidation behavior of a novel Co-based superalloy,” Metallurgical and Materials Transactions A 49A: 4058–4069.

On the Influence of Alloy Chemistry and Processing Conditions on Additive Manufacturability of Ni-Based Superalloys Joseph N. Ghoussoub,Yuanbo T.Tang, Chinnapat Panwisawas, André Németh, and Roger C. Reed

Abstract

Additive manufacturing trials are carried out on two new nickel-based superalloys designed specifically for this processing method. Their performance—with emphasis on their capability to resist cracking—is assessed by comparing with the two legacy alloys IN939 and CM247LC. The two new alloys are found to have demonstrably superior printability. Thermophysical testing and quantitative characterization, particularly via stereology, are used to help rationalize the physical basis of the improved manufacturability displayed. Keywords

Additive manufacturing • Ni-based superalloys • Alloy design

melting, CMSX-4 for single-crystal processing, and Rene 95 for powder-based routes [1]. Probably, new alloys are needed if the AM process is to be exploited commercially. In this paper, the suitability of the nickel-based superalloys for the AM process is studied. As is well known, the differing heat transfer characteristics of the process mean that defects such as cracks and pores can occur. While the precise mechanisms by which these defects arise are not entirely agreed upon at present, many studies have shed light on their nature [2–4]. Characterization at high spatial and chemical resolution is needed to better elucidate the factors influencing defect formation. Specifically, systematic processing trials are needed to compare the behaviour of different alloys: both the heritage ones but also the new grades being designed specifically for this process. The work reported here was motivated with these factors in mind.

Introduction

Alloy Design Approach

Existing grades of superalloy—so-called heritage alloys— were not designed to be used for additive manufacturing (AM). Instead, investment casting or else machining of cast/ wrought or powder metallurgy stock has been the processes on which the industry has been built. Of course, it may well be true that the AM process can be made to work for the heritage alloys. But this may not be the case. After all, the history of the field confirms that a new process usually ushers in new grades of superalloy. Examples include IN718 for vacuum

There has been limited focus on the development of new Ni-based superalloys designed to overcome one of the limitations of additive manufacturing, namely the propensity to cause cracking during processing. The specific mechanisms by which cracking may occur referred to in welding and additive manufacturing literature are the hot cracking mechanisms, solidification and liquation cracking, and those occurring in the solid state such as strain-age cracking; see Fig. 1. An alloy’s susceptibility to each crack mechanism is dependent on the alloy composition. For example, the freezing range as well as the terminal solidification products determines the susceptibility of a given alloy to the formation of solidification cracks [5,6]. In addition, low melting point solidification products such as carbides, borides, and γ  -phase eutectic increase the vulnerability to liquation cracking in the heat-affected zone [7]. Strain-age cracking can occur during cooling to the base-plate temperature and/or during the subsequent heat treatment. The key challenge is the hardening caused by γ  precipitation in conjunction

J. N. Ghoussoub (B) · Y. T. Tang · C. Panwisawas · R. C. Reed Department of Materials, University of Oxford, Parks Road, Oxford OX1 3PH, UK e-mail: [email protected] A. Németh OxMet Technologies Ltd., Yarnton,Kidlington, Oxford OX5 1QU, UK R. C. Reed Department of Engineering Science, University of Oxford, Parks Road, Oxford OX1 3PH, UK

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_15

153

154

J. N. Ghoussoub et al.

Fig. 1 a Microstructure property relationships in different temperature regimes and associated cracking modes, b conceptual optimization of alloy properties

with the presence of process-induced stresses [8]. The link to composition (Al+Ti) is often captured in the so-called weldability diagram [9]. In this paper, two novel Ni-based superalloys designed for crack-free additive manufacture are investigated, ABD850AM and -900AM, whose compositions are given in Table 1 [10]. The equilibrium γ  volume fraction φγ  of each alloy at 500 ◦ C as determined through ThermoCalc TCNI8 database is found in Table 1. For more on the heat treatment and mechanical behaviour of ABD-900AM, see companion paper in this volume by Tang et al. [11]. Both alloys have been isolated via a large-scale modelling-based approach termed Alloys-by-Design (ABD) [12,13]. The method combines high-throughput computational thermodynamics with physical leading-edge scientific models to quantify the effect of alloy chemistry on the risk to one of the outlined crack mechanisms during additive manufacture. In addition, engineering properties such as yield strength and creep resistance, and thermophysical properties such as oxidation resistance and thermal expansion/contraction are included. For a more detailed description of the ABD approach, see [14]. The use of a computational tool for alloy design is motivated by the large number of alloying elements in Ni-base superalloys which results in a considerable composition space. Empirical design is not feasible for a

multi-objective alloy design optimizing the processability and high-temperature properties trade-off. Moreover, the Alloys-by-Design approach takes into consideration the advantages of powder metallurgy and the rapid cooling rates of this process. Segregation of alloying elements during additive manufacture is significantly reduced compared to casting processes for which conventional alloys have been developed. As a result, a new class of Ni-based superalloy for additive manufacture can be isolated.

Experimental Methods Feedstock Material Four different feedstock materials have been studied: the legacy alloys IN939 and CM247LC and the above-mentioned new superalloys ABD-900AM and ABD-850AM. While the alloys investigated vary in γ  volume fraction, IN939 and CM247LC serve as benchmarks with highly desirable corrosion resistance and mechanical properties, respectively. The feedstock powders were produced via argon gas atomization. Their particle size distributions were measured with a MasterSizer 2000 particle size analyser in accordance to ASTM B822. Not strictly identical, the D50’s were found to be similar at 32.3, 33.5, 33.1, and 36.5 µm, respectively; they flow

Table 1 Measured alloy powder compositions (wt.%) Alloy

Ni

Al

Co

Cr

Mo

Nb

Ta

Ti

W

C

B

Hf

φγ 

ABD-850AM ABD-900AM CM247LC IN939

Bal Bal Bal Bal

1.5 2.1 5.5 1.8

18.3 20.3 9.3 18.8

19.3 17.1 8.3 22.5

2.00 2.09 0.55 0.00

0.50 1.85 0.00 1.00

0.37 1.21 3.20 1.30

2.64 2.39 0.73 3.60

4.95 3.06 9.66 1.60

0.040 0.047 0.070 0.160

0.010 0.003 0.020 0.011

– – 1.35 –

0.27 0.34 0.75 0.33

155

On the Influence of Alloy Chemistry and Processing …

adequately and have particle size distributions such as to result in samples largely free of lack of fusion defects [15]. Scanning electron microscopy (SEM) images of the powders are shown in Fig. 2.

time spent travelling between laser pulses. An experimental window has been selected based on empirical evidence in order to investigate the influence of changing printing parameters within a small window where legacy alloys are printed free of excessive porosity and lack of fusion defects, allowing for an isolated investigation of cracking. The printing parameters in the bulk of the samples were varied with respect to velocity and hatch distance, h distance , the two process variables which define the spatial distribution of energy input as illustrated in Fig. 3. The processing parameters used for the experimental trials are listed in Table 2. After each 30 µm thick layer is processed, the hatching pattern frame of reference is rotated by 67◦ relative to the scanning that was previously performed [17].

Quantification of Cracking

Fig. 2 Scanning electron microscopy (SEM) images of powder employed to produce specimens

Additive Manufacturing Trials In order to assess the printability of various alloys, the same geometry was used for all alloys processed. This consisted of building cubes of dimensions 10 × 10 × 10 mm3 ; see Fig. 3. These were manufactured with 16 inverted pyramid legs in order to allow for easy sample removal from the base plate. Samples were produced via selective laser melting (SLM) with a Renishaw AM 400 pulsed fibre laser system of wavelength 1075 nm, laser spot size of diameter 70 µm, and build plate size of 250 × 250 mm2 . Additive manufacturing was carried out under an argon atmosphere using powder layer thickness of 30 µm and laser power of 200 W. The laser scan path was kept constant for all samples printed; a meander hatching pattern was performed on each layer followed by a border finish consistent with Renishaw’s recommended practice as shown in Fig. 3. On the borders, a laser velocity of 50.0 cm/s was employed for all samples manufactured. As is consistent with industrial practice, the laser velocity v is defined as follows for the pulse laser system: v=

dpoint texposure + tdistance

(1)

where dpoint is the distance between the laser pulses, texposure is the time spent depositing the laser pulse, and tdistance is the

Quantitative stereology was performed on the as-printed material, in order to assess printability on the basis of cracking propensity. This was accomplished using a Zeiss Axio A1 optical microscope and MIPAR software package. For each sample, ten optical images of the XZ plane were analysed at 100× magnification. Cracking severity is expressed here in terms of the number of cracks per unit area. Individual crack length is defined by the maximum calliper diameter, the largest line length fitting across a crack. Features of maximum calliper diameter smaller than 10 µm are rejected from the analysis due to unreliable resolution at such magnification. Many studies have investigated the variance in microstructure with increasing build height, showing that the initial hundreds of layers incur significant microstructure heterogeneity [18]. In the light of these findings, the samples are analysed in the XY plane at the sample height midpoint of 5 mm and the XZ plane from the sample height midpoint to the top surface. Images used for crack quantification were taken from regions of greater than 500 µm inwards of the sample edge, as shown in Fig. 4. Additively manufactured materials are highly prone to spatial variance in cracking, with surface-connected cracks posing a unique problem due to their inability to be removed via hot isostatic pressing (HIP).

Microstructure and Phase Characterization Microstructures were analysed by SEM and electron backscatter diffraction (EBSD) analysis performed with a Zeiss Merlin Gemini 2 field-emission gun scanning electron microscope (FEG-SEM) equipped with a Bruker EBSD detector. Differential scanning calorimetry (DSC) was performed with a NETZSCH 404 F1 Pegasus DSC. Heating was carried out from 700 to 1450 ◦ C at 10 ◦ C/min followed by cooling back down to 700 ◦ C at the same rate. The specific

156

J. N. Ghoussoub et al.

Fig. 3 a Sample geometry, build plate, b laser path, and c process variables—adapted from [16]—shown across multiple length scales

heat capacity Cp was determined in accordance to ASTM E1269, and the solidus and liquidus temperatures were determined following Chapman and Quested [19,20].

Results On the Propensity for Alloys to Crack at a Single Processing Condition

Fig. 4 Optical micrograph of CM247LC material at the sample corner, as observed in the XY plane with a black-and-white threshold applied

Images of the XZ plane of four alloys printed given conditions E are shown in Fig. 5. At this condition, many cracks are apparent in legacy alloys CM247LC and IN939. However, with the resolution employed, no cracks were resolved in the ABD alloys. Such findings were confirmed through observation of both the XZ and XY planes.

Table 2 Sample processing conditions in the bulk Sample

dhatch (µm)

dpoint (µm)

v (cm/s)

A B C D E F G H I J K

50 70 90 50 60 60 70 70 80 80 90

60 70 80 80 70 80 60 80 60 70 60

75.0 87.5 100 100 87.5 100 75.0 100 75.0 87.5 75.0

157

On the Influence of Alloy Chemistry and Processing …

displayed as contour plots in Fig. 7. These illustrate that for both legacy alloys, the cracking is most severe when processed with conditions A and C, where the hatch distance and velocity are either both smallest and both largest, respectively. In summary, this quantitative comparison of alloy printability reveals two significant findings: the degree to which cracking occurs is highly dependent on both alloy composition and processing conditions.

Fig. 5 Optical micrographs taken of the XZ plane at the 5 mm height of the as-fabricated material for all four alloys under investigation. Micrographs have had a black-and-white threshold applied. The regions investigated display cracks and minor porosity in CM247LC and IN939, and some µm length-scale porosity in ABD-850AM

On the Influence of Processing Conditions The analysis of the four alloys processed at a single condition is now expanded to a range of conditions. The number of cracks per unit area, the index of cracking severity, is plotted in Fig. 6 against hatch distance and laser velocity, and fitted with an interpolating function. Figure 6 shows two ways by which cracking can be reduced: through varying processing conditions and changing alloy. In this case, the latter appears to have been more successful. In order to better understand the influence of processing conditions, surfaces in Fig. 6 are

Fig. 7 Experimentally determined cracking severity as a function of hatch distance and laser velocity fitted with interpolating contours for a CM247LC, b IN939, the solid points represent the samples as shown in Table 2

On the Presence of Multiple Crack Mechanisms Fig.6 Superimposed surfaces representing the crack severity for IN939, CM247LC, and the ABD alloys as a function of processing conditions

Imaging of individual cracks by means of SEM allows for an assessment of the mechanism by which they occurred. Solidi-

158

J. N. Ghoussoub et al.

fication cracks, recognizable by the presence of intact dendrite arms [21], are found in the sub-optimal regions of the processing diagram. Samples A and C, having cracked to the largest extent, both exhibit solidification-type cracks; an example is shown in Fig. 8. Such is the case for both CM247LC and IN939 alloys. No γ  precipitates could be resolved through SEM of samples prepared by 3 V, 10 s, 10% phosphoric acid electropolish or polished to 0.04 µm colloidal silica finish, and this is due to the ultra-fast cooling rates present during the SLM process. However, the presence of γ  precipitates of size too small to resolve through SEM is possible. As the γ  precipitates are either suppressed or very small, it is concluded that the contribution of the strain-age cracking mechanism is minimal. Solid-state cracking may occur regardless due to insufficient material ductility in the presence of excessive thermal gradient-induced stresses [22]. Consequently, all nonsolidification-type cracking is classified here as solid-state cracking. An example of one such crack is shown in Fig. 8. For IN939 and CM247LC, cracks of both mechanisms are found in samples in the regimes of most pronounced cracking.

Discussion While the above results shed light on the influence of processing and alloy composition on the extent to which cracking occurs, several pertinent questions arise that demand further rationalization. Why do the ABD alloys demonstrate improved additive manufacturability? What is the extent of the contribution from each respective mechanism? Is there a link between reduction in cracking severity and the material microstructure? These questions are addressed in what follows.

Fig. 8 Scanning electron micrographs taken of CM247LC sample A, processed at 50 µm hatch distance and 75 cm/s laser velocity on XZ plane, showing examples of solidification cracking and solid-state cracking

Considerations of the Role of Solidification Range and Path The reduction in solidification cracking achieved through alloy design demonstrated in Figs. 5 and 6 may be rationalized through the effect of chemistry on solidification characteristics. During the final stage of the solidification process, liquid remains in isolated interdendritic regions. If this liquid is retained over too wide a temperature range, or is otherwise unable to fill any gaps between dendrites that open up due to the contraction associated with solidification and cooling, then solidification cracking may occur [5,23]. Clyne proposed a solidification cracking susceptibility coefficient (CSC) as the ratio of the time at which the material is vulnerable to local solidification cracking to the time where stress relief is feasible [24]. Values of CSC are determined from experimental DSC results, shown in Fig. 9, given the assumptions that (a) only one phase transformation is occurring, (b) the

Fig. 9 Specific heat capacity of the four alloys as a function of temperature as determined through DSC, data shown is of the first heating cycle

159

On the Influence of Alloy Chemistry and Processing …

fraction solidified is proportional to the enthalpy divided by the total enthalpy of the phase transformation as determined by integration of Cp − T , and (c) the ratio of the vulnerable period to stress relief period is proportional to the corresponding temperature ranges [25] as follows CSC =

T99% − T90% T90% − T40%

(2)

where the Tx% are the temperatures associated with a remaining solid fraction of x%. The experimentally determined liquidus, solidus, freezing range, and CSC values are shown in Table 3. In addition, the onset of γ  dissolution is determined by the inflexion in Cp −T , where the presence of γ  is detected due to the slow heating rate of 10 ◦ C/min. Table 3 summarizes each alloys respective CSC and cracks/ mm2 ; these do not correlate. Clyne notes that dynamic processes with more complex mechanics may have their cracking susceptibility better estimated by magnitude of their susceptible period [24]. It is indeed found that the magnitude of the solidification range correlates with the extent to which

cracking occurs in respective alloys, but findings suggest the need for a more involved cracking susceptibility criteria. Several studies have investigated the influence of alloy chemistry on cracking during additive manufacturing. While the effects of major alloying elements are generally understood due to extensive welding research in the past [22,26], the influence of minor additions such as carbon and boron remains a topic of discussion [27,28], as it remains to some extent in welding [29–31].

On the Quantification of Cracking Mechanisms

Fig. 11 Quantification of each crack mechanism via eccentricity for both alloys processed using conditions A, B, and C

Having identified the presence of solidification as well as solid-state cracks, the question arises—can the extent of the contribution from each respective mechanism be determined? Observation of cracks in the XZ plane allows for the distinction to be made as solid-state cracks are characteristically long and straight, while solidification cracks, possessing jagged gaps between dendrites, are more complex features; see Fig. 8. Binning of cracks on the basis of their eccentricity e: the extent to which they are circular or elongated, as computed by the fit of an ellipse, serves to bin respective cracking mechanisms,  e=

Fig. 10 Crack eccentricity distribution for a single image of CM247LC sample A illustrating the threshold values of e at which SC, U, and SS cracking modes are distinguished

1−

b2 a2

(3)

where b and a are the minor and major axis lengths, respectively, for b < a. A circle and a straight line have e of 0 and 1, respectively. As shown in Fig. 8, solidification cracks (SC) are less eccentric, by contrast, solid-state cracks (SS) are highly eccentric. Analysis of an image with attention to specific cracks allows for the development of a crack eccentricity

160

J. N. Ghoussoub et al.

Fig. 12 Contour plot of a average grain size (diameter), b texture component vs processing conditions for CM247LC, c texture of CM247LC at conditions A, B, and C as shown through pole figures, and the in-plane inverse pole figure (IPF-Z)

161

On the Influence of Alloy Chemistry and Processing … Table 3 DSC results and measures of solidification cracking susceptibility Alloy

γ  dissolution (◦ C) TLiquidus (◦ C)

TSolidus (◦ C)

TL - TS (◦ C)

Cracks/mm2

CSC

IN939 CM247LC ABD-900AM ABD-850AM

1086 1251 1084 1061

1203 1260 1264 1275

125 116 98 77

>25.6 >19.8 70 µm). At R = 0.2 − 0.3, the nature of the initiation site or mode failure is referenced in Fig. 5 by the shape of the points. When increasing the stress ratio from 0.05 to 0.3, a shift from a surface to an internal initiation from a pore is observed in CMSX-4 specimens. Interestingly, contrary to CMSX-4, the fatal crack in coated René N5 specimens initiates dominantly from the surface of the coating layer for all frequencies tested. Longitudinal sections of specimens also reveal that creep damage is less influential at intermediate stress ratios when decreasing the number of cycles and then the time to rupture. Figure 7a presents the microstructure of a René N5 specimen tested at R = 0.2, 900 Hz that failed at 1.1  107 cycles, equivalent to 3.4 h. No c′- rafting is observed, whereas for other conditions where more fatigue damage is expected but where the time to rupture is longer (tr * 27 h), c′-rafting is well developed, see Fig. 7b, c. At 20,000 Hz and R = 0.3, highly localized slip bands and the precipitation of bright particles are observed at the vicinity of the crack initiation site, see Fig. 7c. For this condition, it is unclear if the fatal crack propagates first in mode I or along octahedral planes.

Discussion Fatigue-Oxidation Interactions Observations performed on specimens tested under fully reversed conditions (R = −1) or at a low stress ratio (R = 0.05) reveal competing surface and internal crack initiation. At low frequency or intermediate frequency and

190

A. Cervellon et al.

Fig. 6 Fracture surface of CMSX-4 specimens tested at 1000 °C, R = 0.3, ra = 156 MPa: a f = 0.5 Hz, Nf = 7.52  104 cycles; b f = 70 Hz, Nf = 4.40  106 cycles; c f = 20,000 Hz, Nf = 1.94  109 cycles. The red arrows point out the main crack initiation site with the exception of (a), which does not present a single but a multitude of initiation sites

Fig. 7 Longitudinal sections of specimens tested at: a René N5, 983 ° C, 900 Hz, R = 0.2, rmax = 430 MPa, Nf = 1.10  107 cycles. No rafting of the precipitates is observed; b CMSX-4, 1000 °C, 70 Hz, R = 0.05, rmax = 430 MPa, Nf = 6.85  106 cycles. c′-rafting is observed; c CMSX-4, 1000 °C, 20,000 Hz, R = 0.3, rmax = 463 MPa, Nf = 1.94  109 cycles. c′-rafting is observed in the volume. Slip bands, bright particles, and a layer of highly deformed microstructure are observed at the vicinity of the crack initiation pore. The last micrograph is in back-scattered electron (BSE) imaging. All micrographs have the same scale bar

long-term tests (> 10 h), fatal cracks initiate from the oxide layer. By increasing the frequency, the potency of the environment is reduced, and this favors internal initiation. In this case, crack initiation takes place from the largest stress concentrators found in the volume, namely casting pores. For each frequency, a shift from internal to surface initiation should be observed, impacted by both the oxidation and the fatigue properties of the alloy. The oxidation resistance, through the ability to form a stable, non-porous, and adherent oxide layer will tend to increase the time range for internal initiations and shift to longer durations the initiations from the surface. AM1 [16], CMSX-4 [17], and René N5 [18] alloys form an adherent Al2O3 oxide layer at

1000 °C, known to provide good oxidation resistance. The shift from internal to surface initiations should be then similar for these three alloys. Likewise, the SX casting process, through its impact on the pore size, will affect the number of cycles at which initiation transfers from the volume to the surface, as already noted in the LCF regime [19]. For the same loading conditions, surface initiation will be observed sooner for specimens with small defects. This is supported by recent observations of surface initiations on hot isostatic pressed specimens tested at 20,000 Hz [9]. The presence of other defects such as carbides, known to decrease fatigue properties [20, 21], will also affect this shift. This competition is less evident on coated René N5, as failure essentially always originates from the coating layer under the testing conditions investigated here. The presence of a coating layer has an influence on the crack initiation stage and the early crack growth. No tests were performed under fully reversed conditions with this material, which do not enable us to discuss the potential impact of the coating layer during pure fatigue.

Creep Damage and Equivalent Stress Method The application of a high mean stress at stress ratios superior to 0.6 activates time-dependent mechanisms at high temperature. In these conditions, HCF and VHCF life is dependent on the time to rupture rather than the number of cycles, which was already noted in the HCF [7] and VHCF [9] regimes. The fracture surfaces and longitudinal sections also indicate that failure mechanisms are more associated with creep rather than cyclic fatigue damage. If the mean stress rm applied during the fatigue test is plotted as a function of the time to rupture tr, all frequencies are then merged on one single S-N curve for both materials, as shown in Fig. 8a. This representation is compared to predictions made with creep laws identified with isothermal creep tests at the same temperature as the fatigue tests. For CMSX-4, a Rabotnov–Kachanov law identified at 1000 °C is used:

Creep, Fatigue, and Oxidation Interactions During High …

191

Fig. 8 a Mean stress rm of the fatigue test plotted as a function of the time to rupture tr at R = 0.6 − 0.8 for CMSX-4 tested at 1000 °C and René N5 tested at 983 °C; equivalent stress plotted as a function of the time to rupture for b CMSX-4 at 1000 °C and c René N5 at 983 °C. René N5 and CMSX-4 creep laws were identified previously by pure isothermal creep tests at the same temperature. René N5 creep law is an identified Larson– Miller Parameter from [22] and CMSX-4 law is a Rabotnov–Kachanov law


r 1 r tr ¼ rþ1 A

ð1Þ

where r is the applied stress in MPa and is defined as the mean stress rm of the fatigue test. A and r are material constants. For René N5, a Larson–Miller Parameter (LMP) from Walston et al. [22] is used. Creep predictions are plotted in Fig. 8a as curves. In the conditions and time range studied, René N5 and CMSX-4 have very similar creep properties [23]. Both alloys are from the second generation of Ni-based SX superalloys, with a rhenium content of around 3 wt%. In Fig. 8a, data points for R = 0.6 − 0.8 follow the associated time to rupture predicted by creep laws for each material. Predicted lives are lower than the test data at 900 Hz and R = 0.6, indicating that the reduction of stress ratio from 0.8 to 0.6 results in more interaction with fatigue and/or oxidation, particularly with coated specimens. One should note that the 20,000 Hz tests in René N5 are closer to the times to rupture estimated by a creep law than the ones at 900 Hz, despite the increase in frequency. These tests were performed with pulses/pauses of 0.5/2 s. During the pause, the vibrations are stopped, but the mean load is still applied, meaning that creep damage under rm also happens when no fatigue cycles are performed. For the same number of cycles, the time to rupture tr is 5 times longer with the pulse/pause parameters used than a continuous test. Then, for a given number of cycles, the tests performed on René N5 at 20,000 Hz are longer than those performed at 900 Hz and explain why creep damage is more prominent for René N5 at these high frequencies. The comparison with pure creep law predictions has also been applied to intermediate stress ratios 0.05
R
0.3. However, all fatigue lives are clearly overestimated when rm is used in Eq. 1. Likewise, all fatigue lives are underestimated if the maximum stress rmax is used in both equations. Pure creep laws are not sufficient to predict the HCF and

VHCF lives if the alternating stress component is not considered for such ratios. To study this effect, the contribution of the alternating stress on the creep damage has been estimated by using an equivalent stress req in Eq. 1. This method considers that the alternating stress contributes to the creep damage when decomposing the fatigue cycle into small increments of time Dt [24]. On each increment, the stress ri is assumed constant as follows: ri ¼ rm þ ra  sinð2pf  Dti Þ

ð2Þ

The portion of creep damage induced by the increment of time Dti over the time to rupture tr,i at ri is calculated with a creep law, in our case a Rabotnov–Kachanov one (Eq. 1). Creep damage over one fatigue cycle is computed by summing these ratios. The number of cycles to failure is determined by considering that failure will happen when creep damage is equal to 1: Nf 

X Dti tr;i

¼1

ð3Þ

Based on the testing frequency, the time to rupture tr can be computed. The equivalent stress req is then determined by inversing Eq. 1 and can be defined as follows: req ¼ rm þ xra

ð4Þ

where x is a variable dependent on the stress ratio and on the material constants used in the creep law. The equivalent stress is plotted as a function of the time to rupture for CMSX-4 and René N5 alloys in Fig. 8b, c, respectively. From Fig. 8b, fatigue tests with durations greater than 5 h are well predicted for CMSX-4 for all positive stress ratios. Even the tests performed at R = 0.05 and 70 Hz, which present crack initiation sites and fracture surfaces highly affected by oxidation, are well predicted. The 20,000 Hz pulsed tests on René N5 are following the tendency even at a

192

A. Cervellon et al.

low stress ratio of R = 0.2 compared to 900 Hz. According to a linear summation, the creep damage developed during the pauses represents 50 and 75% of the total damage at R = 0.2 and R = 0.6, respectively. Tests at 900 and 60 Hz at R = 0.2 are the less well predicted. In these conditions, fracture surfaces are also highly affected by oxidation, but a coating is present. Compared to CMSX-4 results, it reveals that the coating affects/accelerates the crack initiation and the first stages of crack propagation and then reduces the fatigue life. The detrimental impact of coatings is mostly observed for conditions where the fatigue component is more important, i.e., low stress ratios.

Linear Summation of Fatigue and Creep Damage Highlighted in orange in Fig. 8b, c, some fatigue lives do not follow creep predictions, particularly for conditions of high frequency and/or short times to rupture where the contribution of pure fatigue cannot be ignored. In these conditions, fatigue life is then affected by a creep-fatigue interaction, without a relevant impact of oxidation according to the fracture surfaces. The contributions to the total damage D were then considered as a summation of the independent damage contributions from creep and fatigue, as proposed by Wright et al. [7]: D ¼ DF þ DC ¼

N t þ Nf tr

ð5Þ

where DF and DC are the pure fatigue and pure creep damage contributions, respectively. The creep laws presented in the prior section and the equivalent stress method were used to compute the pure creep damage. It should be noted that from this approach, a “creep damage” is computed from the equivalent stress method for fully reversed conditions. If one assumes a similar creep resistance in tension and compression, the creep mechanism is not active under fully reversed conditions. However, several studies have shown an asymmetry in creep strength [25, 26]. Only the positive part of the fatigue cycle has been then considered as damaging for the R = −1 condition. For the pure fatigue law, a fatigue indicator parameter (FIP) developed by Steuer et al. [10] was used. The approach considers that without an environmental influence, pure fatigue life is mainly controlled by crack initiation and the casting pore size. This dependence has been several times validated by studies in the LCF [10], HCF [27, 28], and VHCF [9, 29, 30] domain and is also confirmed by the fracture surfaces described here. The FIP is described as follows:   lDr DK FIP ¼ 1þk ð6Þ E DKthreshold

where µ is the Schmid factor of octahedral planes, Dr the applied tensile stress, E the Young’s modulus, k a constant taken equal to 1, DKthreshold a long crack propagation threshold equal to 10 MPa m1/2 [31], and DK the mode I stress intensity factor calculated according to Murakami’s equation [32]: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffi DK ¼ YDr p Apore ð7Þ where Y is a constant equal to 0.5 and 0.65 for, respectively, internal and surface initiations. Apore is the equivalent area of the casting pore that acts as the crack initiation site. The reader is referred to [10] for more details on this approach. The FIP has been calculated for all pure fatigue specimens that presented a single and internal crack initiation from a casting pore. A good fit with a Basquin-type law computed in [9] between the FIP and Nf is found for the fully reversed tests performed in this study (not shown to keep a reasonable number of figures). The coefficients calculated in [9] were then used in the following for the pure fatigue law. If Eqs. 6 and 7 are directly used in the linear summation, fatigue damage is clearly overestimated for high stress ratios, which are mostly subjected to creep damage. In this work, the applied tensile stress Dr is replaced by the alternating stress ra, which has been shown to give good estimations. The predictions obtained from the linear damage summation are presented in Fig. 9a. A good overall prediction is achieved when considering the number of tests and different loading conditions performed. Tests at 60 Hz and R = −1 are also well predicted despite failures mainly caused by an environmental effect. The loading conditions applied induced failures in a regime where they can either originate from the surface or from the volume without a high impact on the fatigue life. For longer fatigue lives (>108 cycles) though, predictions may overestimate the actual lives. These over-estimations are for instance spotted at 900 and 20,000 Hz in Fig. 9a for René N5 specimens as soon as 106 cycles, where the presence of the coating shifts the competition between internal to surface initiations to lower number of cycles. In Fig. 9b and c are reported the values of the creep damage contribution DC computed from the linear summations for CMSX-4 and AM1, and René N5, respectively. Each curve represents DC computed for a specific condition of stress ratio and frequency as a function of the number of cycles to failure. As expected, the value DC increases with the number of cycles and the decrease of the frequency, clearly highlighted in Fig. 9b. Figure 9c shows that most of the loading conditions tested on CMSX-4 and AM1 lead to a dominant damage contribution (DC > 0.8 or Slit 90 + porous under a given rapp. It is worthy to note that this order corresponds to that in the stick/slip length ratio, c/a (Table 2), and that in the FF lives (Fig. 3). Another important aspect is that the magnitude of Q via the porous pad was lower than that via the Slit 90, whereas the former should have higher stiffness to the FF loading direction in comparison with the latter, by mounding the porous material into the slit pad. These experiments indicate that the introduction of slit and porous structures seems to be working effectively to release the contact stress field at the FF crack nucleating site. These backgrounds will be discussed again in the next chapter.

FF Life Prediction Stress State at Contacting Interface Between Two Elastic Bodies For the purpose of discussing the FF failure behavior, it is interesting to consider the stress state by employing a model as presented in Fig. 8a. In the model, two elastic bodies are contacted by normal compression load, P, and at the same time, an external remote stress, rapp, is applied to the bottom substrate. When the friction is significant at the interface, the adhesive force is supposed to work. This implies that the substrate surface is subjected to a tangential force, Q [12]. In the actual FF test, the direction of Q is reversed and repeated, since the applied load direction is reversed.

Fretting Fatigue Life Extension for Single Crystal … Fig. 6 Fretting surface roughness measured by white beam microscope (CMSX-4/CMSX-4, Slit 0, R.T.)

201

FF loading axis

Outer edge of pad

Outer edge of pad A

A’ Fretting fatigue crack nucleation site

A B B’

B 0.2 mm

Stick/slip boundary

10 µm

Roughness [μm]

0.5 mm

Roughness profile along the line B-B’ along the line A-A’

B

B’

A’

A Stick region

Micro slip region

Table 2 Brief summary of tangential force and stick/slip ratio under the respective test conditions Nomalized stress amplitude in FF test, ra/rB (MPa)

Contact pad surface geometry

Normal Force (in net section stress) (MPa)

Length of stic area, c (in average) (mm)

Ratio of Stic/slip length, c/a (in average)

Variation of Tangential force during FF test (N)

Tangential force at Nf/ 2 (N)

Fretting fatigue life (cycles)

0.24

Flat

100

1.67

0.94

110–135

125

7.1  105

0.24

Flat

70

1.60

0.92

118

118

5.0  105

0.24

Slit 0°

100

1.59

0.69

90–122

112

5.5  105

0.24

Slit 90°

100

1.22

0.82

85–97

95

3.1  106

0.24

Dimple

100

1.51

0.80

80–85

90

1.3  106

When a normal load high enough is to apply on the surface of elastic body, e.g., by a high stiffness indenter, a local plastic yielding starts from an inner site of the structure [12]. When both the P and Q are applied at the same time, a plastically yielding site approaches to the contacting surface [12]. This must be also the case in the present FF test. Hence, it is reasonable to pay a great attention to the stress state of the contacting interface. Such a state has been analyzed by Mindlin [12]. One important finding is that the

adhesive and micro-slip (or partial slip) regions are mixed depending on the relative magnitude of P to Q, as observed in this work, as well. By combining the contact mechanics theories [12, 13], the interface stress state can be known derived for the case in which an external uniform remote stress, rapp, is applied to the bottom elastic body, those are derived in this work and given by;

202

M. Okazaki et al.

Here, the size 2a and po denote the size of contact area, and the maximum contact stress derived by Hertz theory, respectively. Here, the geometry of indenter is assumed to be a long cylinder that has elastic modulus, Ep, thickness t, and the radius, Rp. In the above equations, the regions, −a < x< 2c − a, and 2c – a < x < a, correspond to the adhesive and micro-slip areas, respectively. Note, the size of c/a is changed by relative magnitude of Q/P and the frictional coefficient, l [12]. These equations derived here, of course, agree with those in [12] which have been derived for the case when the rapp is zero. The values of principle stress, r1, and the principle shear stress, s1, and the Mises equivalent stress, req, are directly calculated by neglecting sxy. These stresses are evaluated by, 
0:5 rxx þ rzz rxx  rzz 2 þ þ s2zx 2 2 
0:5 rxx  rzz 2 s1 ¼ þ s2zx 2 o0:5 1 n req ¼ pffiffiffi ðrxx  ryy Þ2 þ ðryy  rzz Þ2 þ ðrzz  rxx Þ2 þ 6ðszx g2 2 r1 ¼ Fig. 7 Hysteresis loops between applied load and resultant tangential force under different surface texturing (CMSX-4/CMSX-4, R.T.) "


 1=2 # x x 2 rxx ðx   aÞ ¼ 2lpo þ 1 a a 2 3 ( 2 )1=2 x þ d x þ d 5 þ rapp þ 1 þ 2lpo ðc=aÞ4 c c 
x2 1=2 x rxx ða  x  c  dÞ ¼ po 1  2lpo a a  xþd þ rapp þ 2lpo ðc=aÞ c 
x2 1=2 x rxx ðc  d  x  aÞ ¼ po 1  2lpo a a 2 ( )1=2 3  xþd xþd 2 4 5 þ rapp þ 2lpo ðc=aÞ þ 1 c c " 
 1=2 # x x 2 rxx ða  xÞ ¼ 2lpo  1 a a 2 3 ( 2 )1=2 x þ d x þ d 5 þ rapp þ 2lpo ðc=aÞ4  1 c c  
x2 1=2
x2 1=2 szx ða  x  c  dÞ ¼ lpo 1  lpo ðc=aÞ 1  a c 
x2 1=2 szx ðc  d  x  aÞ ¼ lpo 1  a 
x2 1=2 rzz ða  x  aÞ ¼ po 1  a

ð1Þ with sffiffiffiffiffiffiffiffiffiffiffi 8RP 2 1  m21 1  m22 p0 ¼ 2P=ðpatÞ; a ¼ ; ¼ þ ; pEeq t Eeq E1 E2 ð2Þ sffiffiffiffiffiffiffiffiffiffiffiffiffiffi d c Q ¼1 ¼1 1 : a a lP

ð3Þ Hereinafter, a similar materials combination, e.g., CMSX-4/CMSX-4 system, is considered for quantitative discussion, otherwise stated. The numerical calculation of each stress components along the contacting interface on the substrate side is shown in Fig. 8, where the stress is normalized by po, and the “leading” and “trailing” edge depict the sites at which the tangential force, Q, starts and ends to work toward the right hand side relatively to the bottom (see Fig. 8a). Once the ratio Q/P is given as a parameter, the stick/slip boundary can be calculated by Eqs. (1), (2). There the shear stress shows the maximum. Of particular importance is that the maximum of req defined by Eq. (3) is always achieved at the stick/slip boundary, irrespective the values of l and Q/P. The increase in l makes this trend more pronounced (compare Figs. 8b with d). The second peak of req appears at the leading edge; however, the difference between the two points gets vague, with increasing the applied stress, rapp/po (compare Figs. 8b with c). This implies the stress state at the stick/slip boundary is more sensitive to the contact conditions, and that the plastic yielding must predominantly originate from there. The more quantitative and conventional expressions to denote the stress concentration at the lading and at the slick/slip boundary will be given in Sect. 4.3 (Eq. 8). These characteristics provide reasonable explanation why the FF cracks were predominantly nucleated from the stick/slip boundary, especially under low stress amplitude FF test condition (Fig. 3). It is also interesting to note from Fig. 8 that the sign of principal stress, r1, at the stick/slip boundary is changed from negative (in compression) to positive (in tension),

Fretting Fatigue Life Extension for Single Crystal …

203

(a)

(b)

(c)

(d)

Fig. 8 Stress analysis along the contact interface. a A mechanical model and b–d calculated results; b Q/P = 0.2, l = 0.3, rapp/po = 0.5, c Q/P = 0.2, l = 0.2, rapp/po = 2, d, Q/P = 0.2, l = 0.5, rapp/po = 0.5

when the frictional coefficient increases, depending on the level of rapp/po. This suggests that once the FF cracks nucleated at the stick/slip boundary, they may propagate rapidly, under higher l structural systems. This assumption also agreed with many experimental works [2, 3, 8]. The effect of test temperature on the FF lives can be considered in a macroscopic manner by incorporating its influence into the value of l in the present calculation. The reason why the FF reduction was a little relaxed at 600 °C than at R.T (Fig. 3) can be explained reasonably because the l was higher at R.T. than at 600 °C [3].

Life Prediction of FF Lives As shown in Fig. 7, the tangential force, Q, is almost proportional to the applied load, P to the substrate specimen in the present FF testing system, whereas the magnitude of Q has been treated as an independent variable from P. From a macroscopic viewpoint, the driving force of Q is the elastic deformation of substrate to which a remote stress is applied, and the resultant resisting force, Q’ should work so that the contact pad follows the substrate deformation, as illustrated in Fig. 9, where Q and the frictional force Q′ should be balanced: i.e., Q + Q′ = 0. The measurement of Q given in Fig. 7 supports this aspect: hence,

204

M. Okazaki et al.

Qs:Trac on force so that it follows the substrate deforma on Qs Balanced state by deforming elastically Contact pad Load P

Substrate

σapp Qp(=Qs) Elas cally deformed line (Schema cs)

Ap

Contact pad Contacting interface

As

Substrate

Load transferring thickness: Ap and As. Fig. 9 Illustration of macroscopic load transfer model via contact interface in the FF test Fig. 10 Example of estimated contact stress field during the FF loading under Rapp = 0

Q  P  sapp

ð4Þ

(the symbol, “  ”, expresses “proportion to”). Here, the Q and Q′ are associated with shear mode dominantly at a narrow interface thickness areas (denoted by Ap and As), which is approximated.   rapp 1 1 1 Q¼ þ E s As E s Ap E p

ð5Þ

Whereas it is too hard to show explicitly the depth, Ap and As, it must be reasonable to assume that they are in order of normal force affected area, or they can be related to the po in Eqs. (1), (2); in a similar material combination system Ap = 2P/(ppo) = As. Hence, by combining Eqs. (4) and (5), the ratio Q/P can be expressed by,     Q Ep rapp Gp rapp ¼g ffi g0 ; ð6Þ P pp0 Es pp0 Gs where the η and η′ are parameters to represent a load transfer efficiency between the contact pad and the substrate, under the influence of micro-slip. The introduction of shear moduli, Gs and Gp, into Eq. (6) is due to the background that the load transfer between two materials must be mainly realized by shear mode stress. Equation (6) can successfully explain such an experimental result that the magnitude of Q decreased when the relative elastic modulus ratio, Ep/Es, decreases (see Table 2 and Fig. 7). By substituting Eq. (6) into Eqs. (1), (2), the stress field at the stick/slip boundary can be calculated for a given FF loading under remote stress ratio of Rapp. Figure 10 presents the calculated stress ratio in principal shear stress, Rs, and

the range Ds1, at the stick/slip boundary as a function of relative intensity of maximum applied stress, rapp,max to po. Here, the calculation was carried out for the case of Rapp = 0, employing tentative values of η′Ep/Es = 0.95 and the frictional coefficient, l = 0.2. It should be noted that the ratio, Rs, and the range Ds1 are different from those given by the remote FF loading. Generally speaking, the values of Ds1 are reduced, and the Rs increases, respectively, compared with those in the plain fatigue. These features get more pronounced when the ratio of contact stress is stronger (or decreasing rapp,max/po in Fig. 10). Based on these numerical calculations, the FF lives for the similar materials combination can be predicted, on the basis of the plain fatigue live curve. The process is illustrated in Fig. 11 upon comparing between the plain fatigue and FF lives. One of the key points is how to convert the remote stress field to the contact stress field especially at the stick/slip boundary; Step 1. This can be done via the calculation; such as Fig. 10. The other key points are how to estimate the effect of stress ratio on fatigue lives; Step 2, and how to convert back from the contact stress field into the remote axial shear a stress field; Step 3. For the former, the characteristic slip system in the SC alloy, or the fracture on crystallographic planes (Fig. 4), is affected by the Schmid factor. For the latter, on the other hand, the semi-empirical approach by Soderberg was used in this work [14]. According to this empirical approach, the fatigue strength under a given R ratio, (ra(R)), is approximated on the basis of the fatigue strength under the stress ratio of 0 (ra(R = 0)), which is presented by [14]

Fretting Fatigue Life Extension for Single Crystal …

205

Fig. 11 New method to predict the FF lives

ra ðRÞ 1  ¼
ra ðR ¼ 0Þ 1  rB a þ a rB ð1 þ RÞ rY

with

rY ð1RÞ

rðR ¼ 0Þ a¼ rB

ð7Þ

where a is the normalized stress amplitude by tensile strength which gives the number of cycles to failure under stress ratio of R = 0. In this work, the value of a is obtained by 0.56 from Fig. 3. Summarizing these processes, the FF live curves are estimated and compared with the experimental lives in Fig. 12. Here, the magnitude of po is approximated by rY, yield strength of substrate material, since the local yielding onsets when po reaches to rY [12], and the gross yielding under the indenter does when po reaches to 3rY, respectively. Note again that this condition corresponds to the state when plastic yield begins at the stick/slip boundary in the model shown in Fig. 8a. While the predicted FF life is sensitive to the value of l, the FF lives seem to be fairly estimated agreeing with the experimental data.

Effect of Surface Texturing on Contact Pad As shown in Fig. 3, the 718/CMSX-4 combination seems to be more critical than the similar material combination from the viewpoint of FF lives. Here, the elastic modulus of alloy 718 is higher than that of CMSX-4. On the other hand, as

Fig. 12 Predicted FF lives and the comparison with experimental results for the CMSX-4/CMSX4 combination

shown in the Sect. 3, the introduction of slit into the contact pad was beneficial to improve the FF lives, associating with the reduction in the tangential force Q (Figs. 3 and 7). Figure 13 explains a mechanical background why the introduction of slit made Q lower. Since no normal stress is working at the vacant area, szx should be also zero there by the Hertz theory. Noting that the integration product of szx

206

Fig. 13 Schematic illustration showing the change in contact stress field by introducing porous structures

along the contact area produces Q, it can reasonably be interpreted why the magnitude of Q in the slitting pad is lowered, compared with that in the bulk surface pad. However, it is necessary to discuss more quantitatively whether the reduction in Q is always beneficial. The role of Q is not simple. As an example, as suggested by Eqs. (1), (2) and Fig. 8, the reduction in Q leads to the same direction as the increase in l which contributes to move the stick/slip boundary near to the trailing edge, whereas the stress state is influenced by the change of Q as indicated by Fig. 8. By the numerical calculations via Eqs. (1), (2), and (3), the following semi-empirical expressions are derived in this work to express the peak stress at the leading edge and that at the stick/slip boundary:   req ðleadingÞ po Kleading  ffi 1þl ð1  gÞ0:5 rapp rapp   req ðstick/slip boundÞ po 1:2 ffi 1 þ 0:2l Kstick=slipbound  req ðleadingÞ rapp Q g¼ lP ð8Þ where Kleading and Kstick/slip bound are the stress concentration factor in equivalent stress (Eq. 3) relatively to the remote applied stress, rapp. In the above approximate equations, the magnitudes of Kleading and Kstick/slip bound are proportional to l, (/lP) and po/rapp. However, it should be noted that there is a difference in power, or sensitivity, for each variable. Hereinafter, a case in which the simple slitting is introduced into the contact pad made by the same material as the substrate by the area fraction of Vf, will be considered, for a brief but more semi-quantitative discussion. Since in the present experiment, the contact force, P, is given so that the net section nominal stress was kept constant (Sect. 2), the P is proportional to (1 − Vf). Eeq appearing in Eq. (2) also varies almost proportionally to (1 − Vf)/(2 − Vf), employing

M. Okazaki et al.

a simple rule for elastic stiffness, or a rule of mixture. Combining these with Eq. (6), the ratio, Q/P, changes proportionally to (2 − Vf)0.5, compared with that of non-slitting pad. Applying the similar calculation, the value of po in Eqs. (2) and (8), changes by (1 − Vf)/(2 − Vf)0.5 relatively to the non-slitting pad. By introducing these semiquantitative calculation into Eq. (8), it is estimated that the stress concentration factors, Kleading and Kstick/slip bound, are reduced by about 20–30%, when Vf = 0.5 and l = 0.5. In these calculations, the most influencing factors are the values of l and po/rapp. The above discussion is more or less too macroscopic. Considering the case when the slits are introduced near the original site at which the stick/slip boundary should be produced in the bulk contact pad, a stick/slip boundary is supposed to be shifted from the original site to the new site. There is a possibility that the multiple stick slip boundaries appear. These phenomena may significantly change the contact stress field. However, these types of microscopic approaches are necessary for future works further, employing finite element calculations. Recently, there has been a research from another aspect, informing that the surface texturing has a role to stabilize the tribological response [15]. Methods to fabricate the textured surface are also requested in future, where the laser pinning and the additive manufacturing methods are expected as potential tools.

Conclusions Fretting fatigue (FF) lives of the CMSX-4 SC alloy have been experimentally measured, associating with different surface texturing counterpart contact pads, in similar and dissimilar material combinations at room and elevated temperatures. A stress field analysis has also been performed along the contact interface. Through these processes, a new life estimation method was proposed. The findings can be summarized as follows. (1) The fatigue strength of the single crystal Ni-based superalloys was significantly influenced by the contact stress field: the FF strength was significantly reduced, compared with the plain fatigue. The measurement of the FF damaged surface topography, as well as the fracture surface observation, indicated that the FF crack was predominantly initiated from the so-called the stick/slip boundary and then propagated associating with crystallographic features. The EBSD analysis showed that the fretting damage progressed at thin layer associated with significant crystallographic rotation by the influence of contact stress field. (2) The experimental works indicated that the surface texturing on the contact pad and the introduction of porous

Fretting Fatigue Life Extension for Single Crystal …

structures were effective to improve the FF lives, without changing the geometry of major substrate. (3) On the basis of contact mechanics theories, some important equations to express the stress filed along the contacting interface between two elastic bodies have been derived as functions of frictional coefficient and the ratio of tangential force to normal load. The numerical calculation of the stress state at the stick/slip boundary provided reasonable mechanical explanation of the FF failure features that agreed with the experimental observations. (4) A new method to predict the FF lives was proposed, which showed reasonable agreement with the experimental data. The method can also provide the rational knowledge on the influencing factors on FF failures.

Acknowledgements The authors express their gratitude to a financial support by the Grain-in-Aid for Scientific Research by JSPS (No. 16H02304, Category A). Sumitomo Denko Co, Japan, supplied the porous material under courtesy.

References 1. Gostic W (2012) Application of materilas and process modeling to the design, development and sustainment of advanced turibine engines. In: Hardy M, Huron ES, Reed RC, Hardy MC, Mills MJ, Montero RE, Portella PD, Telesman J (ed), Superalloys 2012, John Wiley& Sons, Inc., Hoboken, NJ, p 3–12. 2. Huang X, Gibson TE, Zhang M, Neu RW (2009) Fretting on the cubic face of a single-crystal Ni-base superalloy at room temperature. Tribology International, 42(6):875–885.

207 3. Balavenkatesh R, Baba S, Okazaki M (2016) Influence of crystal orientation on cyclic sliding friction and fretting fatigue behavior of single crystal Ni-Base superalloys. In: Hardy M, Huron E, Glatzel U, Griffin B, Lewis B, Rae K, Seethraman V, Tin S (ed) Superalloys 2016, John Wiley& Sons, Inc., Hoboken, NJ, USA: p 395–404. 4. Waterhouse, RB (1981) Fretting fatigue. Applied Science Publishers, London. 5. Hoeppner, DW (ed) (2000) Fretting fatigue; Current technology and practice. ASTM, West Conshocken, PA. 6. JSME standard method of fretting fatigue testing, JSME S 015-2002, The Japan Society of Mechanical Engineers, (2002) Tokyo. 7. Nishioka K, Hirakawa K(1968) Fretting fatigue failure. Trans. JSME 34 (268): 2068–2073. 8. Mutoh Y, Jayaprakash M (2011) Tangential stress range–compressive stress range diagram for fretting fatigue design curve. Tribology International, 44:1394–1399. 9. Prasad, S, Michael J, and Christenson, T (2003). EBSD studies on wear-induced subsurface regions in LIGA nickel. Scripta Materialia, 48(3):255–260. 10. Hattori H, Nakamura M, Watanabe T, A New approach to the prediction of the Fretting fatigue life that considers the shifting of the contact edge by wear:19–30 in Ref. [5]. 11. Okazaki M, Hiura T, and Suzuki T (2000) Effect of local cellular transformation on fatigue small crack growth in CMSX-4 and CMSX-2 at high temperature. In: Pollock TM, Kissinger RD, Bowman RR, Green KA, McLean M, Olson SL, Schrra JJ (ed) (2000) Superalloys 2000, TMS: p 505–514. 12. Johnson KL (1985), Contact mechanics, Cambridge University Press, U.K. 13. Mindlin RD (1949), Compliance of elastic bodies in contact, Transactions of the ASME 16: 259–268. 14. Soderberg CR (1936), Factors of safety and working stress. Trans. ASME 52: 13–28. 15. Shimada K, Hirai T, Mizutani, M, Kuriyagawara T (2018) Fabrication of functional surface by ultrasonic assisted cutting. J. Japan Soc. Abrasive Tech., 62: 39–44.

Initiation of Fatigue Cracks in a Single-Crystal Nickel-Based Superalloy at Intermediate Temperature Yaqi Huang, Dong Wang, Jian Shen, Yuzhang Lu, Langhong Lou, and Jian Zhang

Abstract

Fatigue tests under a series of stress amplitudes at 760 °C of a single-crystal nickel-based superalloy were conducted to investigate the relationships between loading conditions and initiation sites of fatigue cracks. It was found that initiation sites of the dominant fatigue crack gradually transferred from pores to MC carbides with an increase of the stress amplitude. At low stress amplitude, the fatigue crack that causes the final fracture initiated at pores and the fatigue life depended on size and shape of pores. Although the carbides on surface cracked as a result of oxidation and cyclic loading, the cracks did not penetrate the carbide/matrix interface. However, at high stress amplitude, the cracks in the carbide can penetrate the carbide/matrix interface rapidly under applied stress, and following crack propagation resulted in decisive fatigue fracture. Compared to blocky carbides on the surface, script carbides at subsurface were more likely to induce fatigue crack initiation. Keywords




Single-crystal Ni-based superalloy initiation Stress amplitude




Fatigue crack

Introduction Nickel-based single-crystal (SX) superalloys are widely used in advanced turbine engines due to their exceptional high temperature strength [1–3]. They endure cyclic stresses Y. Huang
D. Wang
J. Shen
Y. Lu
L. Lou
J. Zhang (&) Superalloys Division, Institute of Metal Research, Chinese Academy of Sciences, Shenyang, 110016, China e-mail: [email protected] Y. Huang School of Materials Science and Engineering, University of Science and Technology of China, Hefei, 230026, China

during service, and fatigue is one of the most common modes of failure in SX components [4–6]. The initiation of fatigue cracks and its early stages of propagation is very important as they contribute to most of the fatigue life [6–10]. A large number of studies have shown that the discontinuous or inhomogeneous microstructures (such as pores, carbides, or eutectics) on the surface or subsurface in single-crystal superalloy are easy to induce irreversible cyclic dislocation slip during cyclic loading, which leads to local plastic strain accumulation and crack initiation [11– 15]. As for the specific crack initiation place, it mainly depends on the applied stress, temperature, frequency, and the size, shape, and location of defects [5, 15–22]. Generally speaking, the crack initiation occurs at the largest pore in the alloy [5, 19, 20, 22, 23]. In addition, as the distance of pore from the surface decreases or the irregularity of pore increases, the effect of the pores on fatigue life becomes more obvious [20, 21, 24, 25]. When hot isostatic pressing eliminates porosity in alloy, fatigue crack may also initiate from carbides or eutectics [12, 13, 19]. For longer tests (low stress or frequency), the oxidized layer may become the main crack initiation sites [6, 19]. Carbides are brittle and usually possess poor oxidation properties [25–31]. During fatigue deformation, cracks of carbides and formation of cracks at the carbides/matrix interface have been observed [10, 26, 32–35]. Up to now, there is limited information on where fatigue cracks exactly originated from except for some recent reports on very high cycle fatigue (VHCF) behaviour of SX alloys [11, 19]. The dominant fatigue crack initiation sites may be related to the loading conditions [15, 16]. However, the details are still missing. In this study, the fatigue tests under a series of stress amplitudes at 760 °C were conducted, and a special focus is paid to the relationships between loading conditions and the initiation sites of the dominant fatigue cracks. The effect of the size, shape, and distribution of the pores and carbides on the fatigue crack initiation is also explored.

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_20

208

Initiation of Fatigue Cracks in a Single-Crystal Nickel …

Experiments Material The first-generation single-crystal Ni-based superalloy PWA 1483 was used in this study. The nominal composition (in wt %) of this alloy was 0.05 C, 12.0 Cr, 9.0 Co, 4.0 W, 2.0 Mo, 3.4 Al, 4.0 Ti, and Ni bal. Cast bars of U16  255 mm were produced by two different directional solidification (DS) processes (high rate solidification, HRS and liquid metal cooling, LMC). A spiral grain selector was used in all castings. Orientation measurements were employed using electron backscattered diffraction (EBSD). The cast bars within 10° deviation from were cut into *50-mm-long bars and subjected to a standard heat treatment (1230 °C/2 h + 1250 °C/4 h, AC + 1080 °C/4 h, AC). Then, these heat-treated bars were machined into fatigue specimens (gauge section, 1 mm  1 mm  1 mm), as shown in Fig. 1. The gauge section in specimens was ground and polished with 2.5-lm diamond polishing paste in the longitudinal direction to achieve a surface finish (Ra) of *0.4 lm. Edges and both sides of the gauge section were polished with 3000 grit SiC abrasive paper.

Fatigue Test and Microstructure Characterization The fatigue tests were performed at 760 °C in air using a MTS 810 servo hydraulic testing machine under a constant stress control mode, and a series of stress amplitudes were applied during testing. A triangle waveform with a loading frequency of 5 Hz or 1 Hz and stress ratio of R = 0.1 (minimum stress rmin/maximum stress rmax) were used. Metallographic samples were mechanically ground, polished, and etched in a solution of 10 g CuSO4, 50 ml HCl, and 50 ml H2O. The optical metallographic microscope (OM-IMc5) and scanning electron microscope (FE-SEM) equipped with electron diffraction spectrum (EDS) were employed to characterize the microstructure of the alloy before and after fatigue tests. The fracture surfaces of fatigue Fig. 1 a Schematic and b photograph of the fatigue specimen

209

specimens were also examined by SEM. The X-ray computed tomography (XCT) with laboratory-based Xradia Versa XRM-500 system was used for 3D characterization of pores in the gauge Sect. (1 mm  1 mm  1 mm). The projected images with about 1.0 lm pixel size were recorded. Then, these projections were processed by beam hardening correction and smoothing, and visualized using the software of Avizo fire 7.1. The pores with equivalent diameter over 4 lm were counted in the quantitative analysis to ensure the accuracy of results. Some of the specimens were interrupted during fatigue tests for microstructure and XCT examination.

Results Microstructure Before Fatigue Test Figure 2 shows the microstructure of specimens before fatigue testing. The typical dendritic structures (Fig. 2a, d) reveal that the decrease in the primary dendrite arm spacing (PDAS) of LMC sample compared to that of HRS sample was remarkable. Pores and carbides in the interdendritic region were observed. It can be seen from Fig. 2b and e that all carbides with blocky or script morphology were MC-type enriched in Ta and Ti (see the EDS inset in Fig. 2b). The size of carbides and pores in LMC sample is much smaller than those in HRS sample. The statistical results of pores from Fig. 2c and f are shown in Fig. 3a, b. Most pores were measured to possess an equivalent diameter (diameter of sphere with the same volume) around 5 * 10 lm. The average pore size was measured as 14.3 lm in HRS and 9.6 lm in LMC samples, respectively. Large pores with equivalent diameter above 50 lm were observed in HRS specimen, whilst the largest pore in LMC specimen was below 40 lm. Volume fraction of pore in LMC sample (0.09%) is also significantly lower than that in HRS sample (0.15%). Results obtained from XCT in present work are in good agreement with previous results obtained from traditional quantitatively metallography method [36, 37].

210

Y. Huang et al.

Fig. 2 a and d OM, b and e backscattered electron (BSE) micrographs, and (c and f) XCT results of specimens before fatigue test. a–c Alloys solidified by HRS, d–f alloys solidified by LMC

Fig. 3 Size distribution of the pores in a HRS and b LMC samples

Fatigue Tests The results of fatigue tests are listed in Tables 1 (HRS samples) and 2 (LMC samples). Size refers to the equivalent diameter of a pore or the maximum length of a carbide, while aspect ratio is defined as the ratio between minimum width and maximum length of a pore or carbide. Location is the distance to the nearest surface of specimen. In both HRS and LMC specimens, the initiation sites of dominant fatigue crack gradually transferred from pores to

MC carbides with increase of stress amplitude ra. The critical stress amplitude was observed as 495 MPa in HRS specimens, and 517.5 MPa in LMC specimens, respectively. In the specimens tested at low ra, pores were the crack initiation sites. The sizes of the pore-induced fatigue failure were larger than average value, whilst the aspect ratio was lower than the average. It is interesting to note that the location of these pores shown in Tables 1 and 2 indicates that the pore-induced final failure distribute rather randomly in the specimens. Script carbides were the predominant crack

Initiation of Fatigue Cracks in a Single-Crystal Nickel …

211

Table 1 Initiation sites of fatigue crack in alloys solidified by HRS Specimens

Frequency/Hz

ra/MPa

Crack sources

Size/lm

Aspect ratio

Location/lm

Nf (105)

S1

5

405

Pore

42

0.27

90

12.46874

S2

5

450

Pore

54

0.32

227

3.74674

S3

5

450

Pore

40

0.43

14

2.94212

S4

5

450

Pore

55

0.32

490

2.50274

S5

5

495

Pore

50

0.39

37

1.51222

Averagea

–

–

–

14.3 ± 0.01

0.63 ± 0.002

–

–

S6

5

495

Script carbide

99

0.04

5

0.32770

S7

1

517.5

Script carbide

93

0.14

26

0.16072

S8

1

517.5

Blocky carbide

28

0.38

0

0.15001

S9

1

540

Script carbide

121

0.04

11

0.13349

S10

1

540

Script carbide

94

0.06

25

0.08859

Averageb

–

–

–

18.4 ± 0.02

0.39 ± 0.003

–

–

Data obtained from pores in XCT results (1 mm  1 mm  1 mm) b Data obtained from carbides in metallographic samples (averaged from 60 to 70 carbides) a

initiation sites at high ra. Only one crack initiated on blocky carbide located very near the corner of the surface in specimen S8. It is quite clear that large subsurface carbides with complex shape were vulnerable sites to induce the fatal fatigue crack.

Initiation of Fatigue Cracks Under Low Stress Amplitude All fatigue crack-induced final fracture initiated from pores at low stress amplitude ra (  495 MPa in HRS specimens and  517.5 MPa in LMC specimens). The typical fracture surfaces and pore morphology are shown in Fig. 4. It can be viewed that the pore-induced fatigue crack were with irregular shape and large size.

Figure 5 shows the surface morphologies of different specimens during fatigue under low amplitude. It is interesting to see that all MC carbides on the surface experienced cracking. However, the cracks cannot propagate through the carbide/matrix interface during fatigue (from early stage to final failure, Fig. 5a–c). Figure 6 shows the detailed observation of a cracked carbide on surface. Figure 6a, b are the morphology of cracked carbides at different stages of fatigue. Figure 6c reveals the EDS linear scanning result, and the dashed lines are corresponding to that in Fig. 6b. It can be seen from Fig. 6 that the carbides on the surface were oxidized in early stage of fatigue. The oxidized carbide provided a rapid diffusion passage of oxygen and promoted further oxidation of carbide and carbide/matrix interface (Fig. 6a). The oxidized carbides cracked easily under the cyclic loading during

Table 2 Initiation sites of fatigue crack in alloys solidified by LMC

a

Specimens

Frequency/Hz

ra/MPa

Crack sources

Size/lm

Aspect ratio

Location/lm

Nf (105)

S11

5

450

Pore

40

0.20

198

5.62666

S12

5

495

Pore

26

0.31

62

5.32255

S13

5

495

Pore

33

0.26

243

2.96954

S14

5

517.5

Pore

36

0.27

125

1.33718

Averagea

–

–

–

9.6 ± 0.01

0.57 ± 0.004

–

–

S15

1

517.5

Script carbide

29

0.11

4

0.09408

S16

1

517.5

Script carbide

114

0.14

3

0.07737

S17

1

540

Script carbide

110

0.04

2

0.02394

Averageb

–

–

–

10.6 ± 0.01

0.37 ± 0.003

–

–

Data obtained from pores in XCT results (1 mm  1 mm  1 mm) Data obtained from carbides in metallographic samples (averaged from 60 to 70 carbides)

b

212

Y. Huang et al.

Fig. 4 SEM micrographs showing the fracture surfaces in a and c S5 and b and d S11. c–d Magnifications of the regions highlighted by rectangles in (a–b)

Fig. 5 BSE micrographs showing the cracked carbides. a 16141 cycles (*0.11Nf), b 70,000 cycles (*0.28Nf), and c after failure

fatigue (Fig. 6b). Figure 6d shows oxidized carbides on a polished surface after heated at 760 °C for 30 min as reference (all samples were hold at 760 °C for 20–30 min in fatigue tests before loading). All carbides were oxidized, but most of the carbides did not crack.

Initiation of Fatigue Crack Under High Stress Amplitude The carbides were predominant fatigue crack initiation sites in the specimens tested at high stress amplitude ra

Initiation of Fatigue Cracks in a Single-Crystal Nickel …

213

Fig. 6 BSE micrographs showing the cracked carbide a at 1160 cycles and b after failure, c EDS linear scanning result of the cracked carbide as marked by the black arrow in Fig. 6b, and d surface carbides after heated at 760 °C for 30 min without loading

(  495 MPa in HRS specimens and  517.5 MPa in LMC specimens). The typical fracture surfaces and surface morphology are shown in Fig. 7. The cracked script carbides at subsurface were observed on the fracture surfaces of most specimens (as marked by arrows in Fig. 7d, e). Only one fatigue crack-induced final fracture initiated on blocky carbide located very near the corner on the surface in S8 (as marked by arrow in Fig. 7f). As shown in Fig. 8, the cracking of MC carbides on the surface of fatigue specimens at high stress amplitude was also observed. However, it was very different from the cracked carbides shown in Fig. 5. The cracks passed through the carbide/matrix interface and continued to propagate. A longer crack could form easily by connecting of such

cracks (Fig. 8b), leading to the acceleration of crack propagation. Figure 9 exhibits the longitudinal section in a fatigue specimen showing a cracked carbide at subsurface. It can be seen clearly that the micro-crack first formed at the surface and encountered a carbide on its propagation path. The carbide was oxidized when exposed to air. This probably accelerated the propagation of the crack. However, it is worthy to note that the carbide located more closely to surface did not experience oxidation and fracture, as marked by circle in Fig. 9. Figure 10 shows a pore with irregular shape in specimen S6 after its failure. No micro-crack initiating from the sharp corner of the pore was observed.

Fig. 7 SEM micrographs showing the fracture surfaces and surface morphology in different specimens. a and d S6, b and e S17, and c and f S8. d–f Magnifications of the regions highlighted by rectangles in (a–c)

214

Y. Huang et al.

Fig. 8 SEM micrographs showing the cracked carbides a at 3000 cycles (0.2Nf) and b after failure

Discussion Effect of Loading Condition on Initiation Site of Fatigue Crack It was found from the present experimental results that the initiation sites of the dominant fatigue crack gradually transferred from pores to MC carbides with increase of stress amplitude. It is well known that fatigue occurs from crack initiation (from a pore or carbide in this work), early propagation, and formation of a short/long crack, and follows by quick propagation leading to final fracture. Where the dominant fatigue crack originates from depends on not only the micro-crack initiation but also the possibility of micro-crack propagation. Fig. 9 BSE micrograph showing a cracked carbide at subsurface

• Formation of a micro-crack

Fig. 10 SEM micrograph showing the pore in S6 after failure

The micro-crack initiating from pores was associated with the higher stress concentration and localized plastic deformation around the pores [11, 20, 24, 38]. Under the cyclic loading during fatigue, the severe plastic deformation was observed around the pores due to the repeated shearing of the surrounding c and c′ phases, causing the crack initiation [6, 11, 38]. Therefore, a certain loading cycles are necessary to induce the crack initiation around a pore [20]. If the fatigue life is not long enough, as the case of specimen S6, micro-crack did not form around pores (Fig. 10). All carbides on surface cracked under low/high stress amplitude as shown in Figs. 5 and 8. The detailed characterization reveals that the cracking of carbides was attributed to a combined effect of oxidation and cyclic loading (Fig. 6). Porous Ta and Ti oxide formed and easily cracked under the cyclic loading. It is interesting to note that the cracks

Initiation of Fatigue Cracks in a Single-Crystal Nickel …

215

generally observed within carbides, i.e. not at the interface of carbide and matrix, which indicates that oxidation plays an important role. • Propagation of the micro-crack All carbides cracked at the early stage of fatigue. However, the present observations indicate that whether the crack can enter the matrix (early propagation) is critical. This is closely related to the stress intensity threshold of MC carbide ΔKth-MC. The micro-crack is able to propagate when the stress intensity factor (SIF) range ΔK at the front of a crack is higher than the threshold ΔKth. The ΔK is typically calculated considering a mode-I opening and estimated by [7] DK ¼ YDra1=2

ð1Þ

where Y is the shape factor, Dr is stress range, and a is the crack length. In this work, the stress intensity factor range ΔK around a MC carbide can be calculated with a single-edged straight crack assumption [39], as follows: pffiffiffiffiffiffi DK ¼ Dr paF ða=W Þ ð2Þ where Dr (=2ra) is the applied stress range used in tension– tension fatigue tests, a is the crack length, and W is the width of specimen. F (a/W) is a function of a/W to account for the geometry effect. The initial stress intensity factor range ΔK around a pore was estimated using the following equation [40]: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi DK ¼ YDr p Adefect ð3Þ where Adefect is the area of the pore measured using ImageJ software on the fracture surface, and Y the shape factor which is equal to 0.5 for internal and to 0.65 for subsurface initiations [41]. Pore at a distance from the surface greater than or equal to half of its mean diameter is considered as internal pore, whereas pore located closer to the surface or at the surface is considered as subsurface pore [42]. The results illustrated in Fig. 11 show the calculated ΔK around a pore or a cracked MC carbide as a function of the number of cycles to failure Nf. Since under low stress amplitude ra, the cracks in carbides did not penetrate the carbide/matrix interface (Fig. 5), and it can be speculated from this work that ΔKth-MC is greater than 6.2 MPa m1/2, obviously higher than the stress intensity threshold of pore ΔKth-pore reported as 1.1 * 2 MPa m1/2 for superalloys [20, 43]. More fatigue tests are underway to obtain the precise value of ΔKth-MC. Under low stress amplitude ra (  495 MPa in HRS specimens and  517.5 MPa in LMC specimens), crack initiated in carbides, but cannot enter the matrix due to ΔK < ΔKth-MC. Consequently, fatigue crack initiated at

Fig. 11 Stress intensity factor range ΔK at the front of a micro-crack as a function of the number of cycles to failure Nf. Dashed lines represent the data of ΔKth-pore from Refs. [20, 43], respectively

pores and was able to propagate until failure. However, at high stress amplitude ra (  495 MPa in HRS specimens and  517.5 MPa in LMC specimens), it can be seen clearly in Fig. 8 that the cracks can easily pass through the carbide/matrix interface as ΔK > ΔKth-MC. Thus, a micro-crack in the matrix formed quickly, resulting in decisive fatigue fracture.

Effect of Size, Shape, and Distribution of Pores and MC Carbides It is obvious that large and irregular pores are vulnerable sites for crack initiation. It seems that the pore-induced final failure distributed rather randomly in specimens. This may be attributed to the small specimens in the present experiments. On the other hand, the thickness of our sample is comparable to the SX blades with thin-walled structure; it may suggest that the pores at any sites in the SX blades are potential threats to the fatigue performance. Most fatigue crack-induced final fracture initiated from script carbides on subsurface at high ra. The script carbides have large size and small aspect ratio. It is believed that the complex stress distribution exacerbates the stress concentration around them. Moreover, effect of oxidation during fatigue tests is also important. The cracking of the script carbides arising from the combination of oxidation and cyclic loading provided a fast propagation path for fatigue crack. Only one crack initiated on blocky carbide located very near the corner on the surface. Since the surface of the specimen suffered complex environment and stress condition, the cracked blocky carbide with smaller size and regular shape may also induce crack initiation.

216

Comparison of Two Different Solidification Processes The LMC technique with a higher temperature gradient compared to HRS process achieves microstructural improvements including reduced fraction and size of casting pores, refined carbides and eutectics, and smaller PDAS. This generally leads to a much better fatigue property in LMC specimens [20, 24, 44–48]. The present results are in good agreement with previous observations. LMC specimens (S11 * S14) had a longer fatigue life compared to that of HRS specimens (see Tables 1 and 2). The critical loading condition was 517.5 MPa in LMC alloys, whilst the critical stress amplitude of 495 MPa was found in HRS alloys. However, three LMC specimens (S15 * S17) exhibited a shorter life compared to that of the HRS specimens. This is probably due to the limited number of tests and small gauge length of the specimens in the present experiments. The carbides led to final failure in these specimens (S15 * S17) all had very large size and located very close to the surface.

Summary (1) The initiation site of dominant fatigue crack gradually transferred from pores to MC carbides with increase of stress amplitude in a single-crystal nickel-based superalloy at intermediate temperature (760 °C). Large pores and carbides with irregular shape were critical crack initiation sites for dominant fatigue cracks. (2) Under the condition of low stress amplitude ra (  495 MPa in HRS specimens and  517.5 MPa in LMC specimens), the fatigue crack initiated at pores and propagated until failure. The fatigue life depended on size and shape of pores. The carbides on surface cracked as a result of oxidation and cyclic loading, but the cracks did not penetrate the carbide/matrix interface. (3) At high stress amplitude ra (  495 MPa in HRS specimens and  517.5 MPa in LMC specimens), the cracks in the carbide can penetrate the carbide/matrix interface rapidly under applied stress, and following crack propagation resulted in decisive fatigue fracture.

Acknowledgements This work was financially supported by the National Natural Science Foundation of China (Grant Nos: 51631008 and 91860201), National Science and Technology Major Project (2017-VII-0008-0101 and 2017-VI-0003-0073), and Key Deployment Projects of the Chinese Academy of Sciences (ZDRW-CN-2019-01). The authors are grateful for those supports.

Y. Huang et al.

References 1. Reed RC, Tao T, Warnken N (2009) Alloys-By-Design: Application to nickel-based single crystal superalloys. Acta Mater. 57:5898–5913. 2. Reed, RC (2006) The superalloys: fundamentals and applications. Cambridge University Press, Cambridge. 3. Pollock TM, Tin S (2006) Nickel-based superalloys for advanced turbine engines: chemistry, microstructure and properties. J. Propuls. Power 22:361–374. 4. Zhang YL, Wang XG, Li JG, Chen Y, Yang YH, Meng J, Liu JL, Liu JD, Zhou YZ, Sun XF (2019) The low-cycle fatigue deformation mechanisms of two single crystal superalloys at room temperature and 600 °C. Scr. Mater. 171:122–125. 5. Bortoluci Ormastroni LM, Mataveli Suave L, Cervellon A, Villechaise P, Cormier J (2020) LCF, HCF and VHCF life sensitivity to solution heat treatment of a third-generation Ni-based single crystal superalloy. Int. J. Fatigue 130:105247. 6. Rémy L, Geuffrard M, Alam A, Köster A, Fleury E (2013) Effects of microstructure in high temperature fatigue: lifetime to crack initiation of a single crystal superalloy in high temperature low cycle fatigue. Int. J. Fatigue 57:37–49. 7. Suresh, S (1998) Fatigue of materials. Cambridge University Press, Cambridge (UK). 8. Stephens RR, Grabowski L, Hoeppner DW (1993) The effect of temperature on the behaviour of short fatigue cracks in Waspaloy using an in situ SEM fatigue apparatus. Int. J. Fatigue 15:273–282. 9. Zhang XP, Wang CH, Chen W, Ye L, Mai YW (2001) Investigation of short fatigue cracks in nickel-based single crystal superalloy SC16 by in-situ SEM fatigue testing. Scr. Mater. 44:2443–2448. 10. Ma XF, Shi HJ (2014) In situ SEM studies of the low cycle fatigue behavior of DZ4 superalloy at elevated temperature: effect of partial recrystallization. Int. J. Fatigue 61:255–263. 11. Cervellon A, Hémery S, Kürnsteiner P, Gault B, Kontis P, Cormier J (2020) Crack initiation mechanisms during very high cycle fatigue of Ni-based single crystal superalloys at high temperature. Acta Mater. 188:131–144. 12. Yi JZ, Torbet CJ, Feng Q, Pollock TM, Jones JW (2007) Ultrasonic fatigue of a single crystal Ni-base superalloy at 1000 ° C. Mater. Sci. Eng. A 443:142–149. 13. Ruttert B, Meid C, Roncery LM, Lopez-Galilea I, Bartsch M, Theisen W (2018) Effect of porosity and eutectics on the high-temperature low-cycle fatigue performance of a nickel-base single-crystal superalloy. Scr. Mater. 155:139–143. 14. Han GM, Zhang ZX, Li JG, Jin T, Sun XF, Hu ZQ (2012) High cycle fatigue behavior of a Nickle-based single crystal superalloy DD98M at 900 °C. Acta Metall. Sin. 48:170–175. 15. Morrissey RJ, John R, Porter WJ (2009) Fatigue variability of a single crystal superalloy at elevated temperature. Int. J. Fatigue 31:1758–1763. 16. Prasad K, Sarkar R, Gopinath K (2016) Role of shrinkage pores, carbides on cyclic deformation behaviour of conventionally cast nickel base superalloy CM247LC® at 870 °C. Mater. Sci. Eng. A 654:381–389. 17. Chieragatti R, Rémy L (1991) Influence of orientation on the low-cycle fatigue behaviour of MAR-M200 single crystals. Part I: Fatigue life behaviour. Mater. Sci. Eng. A 141:l–9. 18. Fleury E, Rémy L (1993) Low cycle fatigue damage in nickel-base superalloy single crystals at elevated temperature. Mater. Sci. Eng. A 167:23–30.

Initiation of Fatigue Cracks in a Single-Crystal Nickel … 19. Cervellon A, Cormier J, Mauget F, Hervier Z, Nadot Y (2018) Very high cycle fatigue of Ni-based single crystal superalloys at high temperature. Metall. Mater. Trans. A 49:3938–3950. 20. Lamm M, Singer RF (2007) The effect of casting conditions on the high-cycle fatigue properties of the single-crystal nickel-base superalloy PWA 1483. Metall. Mater. Trans. A 38:1177–1183. 21. Jiang R, Bull DJ, Evangelou A, Harte A, Pierron F, Sinclair I, Preuss M, Hu XT, Reed PAS (2018) Strain accumulation and fatigue crack initiation at pores and carbides in a SX superalloy at room temperature. Int. J. Fatigue 114:22–33. 22. Cervellon A, Cormier J, Mauget F, Hervier Z (2017) VHCF life evolution after microstructure degradation of a Ni-based single crystal superalloy. Int. J. Fatigue 104:251–262. 23. Hong HU, Choi BG, Kim IS, Yoo YS, Jo CY (2011) Characterization of deformation mechanisms during low cycle fatigue of a single crystal nickel-based superalloy. J Mater. Sci. 46:5245–5251. 24. Steuer S, Villechaise P, Pollock TM, Cormier J (2015) Benefits of high gradient solidification for creep and low cycle fatigue of AM1 single crystal superalloy. Mater. Sci. Eng. A 645:109–115. 25. Liu YH, Kang MD, Wu Y, Wang MM, Gao HY, Wang J (2017) Effects of microporosity and precipitates on the cracking behavior in polycrystalline superalloy Inconel 718. Mater. Charact. 132:175–186. 26. Reuchet J, Rémy L (1983) Fatigue oxidation interaction in a superalloy—application to life prediction in high temperature low cycle fatigue. Metall. Trans. A 14A:141–149. 27. Connolley T, Reed PAS, Starink MJ (2003) Short crack initiation and growth at 600 °C in notched specimens of Inconel 718. Mater. Sci. Eng. A 340:139–154. 28. Kontis P, Alabort E, Barba D, Collins DM, Wilkinson AJ, Reed RC (2017) On the role of boron on improving ductility in a new polycrystalline superalloy. Acta Mater. 124:489–500. 29. Kontis P, Collins DM, Wilkinson AJ, Reed RC, Raabe D, Gault B (2018) Microstructural degradation of polycrystalline superalloys from oxidized carbides and implications on crack initiation. Scr. Mater. 147:59–63. 30. Mataveli Suave L, Serrano Muñoz A, Gaubert A, Benoit G, Marcin L, Kontis P, Villechaise P, Cormier J (2018) Thin-wall debit in creep of DS200 + Hf alloy. Metall. Mater. Trans. A 49A: 4012–4028. 31. Kontis P, Mohd Yusof HA, Pedrazzini S, Danaie M, Moore KL, Bagot PAJ, Moody MP, Grovenor CRM, Reed RC (2016) On the effect of boron on grain boundary character in a new polycrystalline superalloy. Acta Mater. 103:688–699. 32. Reuchet J, Rémy L (1983) High temperature low cycle fatigue of MAR-M 509 superalloy II: the influence of oxidation at high temperatures. Mater. Sci. Eng. 58:19–32. 33. Chu ZK, Yu JJ, Sun XF, Guan HR (2008) High temperature low cycle fatigue behavior of a directionally solidified Ni-base superalloy DZ951. Mater. Sci. Eng. A 488:389–397.

217 34. Ma XF, Shi HJ, Gu JL (2010) In-situ scanning electron microscopy studies of small fatigue crack growth in recrystallized layer of a directionally solidified superalloy. Mater. Let. 64:2080– 2083. 35. Zhang L, Zhao LG, Roy A, Silberschmidt VV, McColvin G (2019) In-situ SEM study of slip-controlled short-crack growth in single-crystal nickel superalloy. Mater. Sci. Eng. A 742:564–572. 36. Brundidge CL, Vandrasek D, Wang B, Pollock TM (2012) Structure refinement by a liquid metal cooling solidification process for single-crystal Nickel-base superalloys. Metall. Mater. Trans. A 43A:965–976. 37. Roskosz S, Adamiec J (2009) Methodology of quantitative evaluation of porosity, dendrite arm spacing and grain size in directionally solidified blades made of CMSX-6 nickel alloy. Mater. Charact. 60:1120–1126. 38. Lukas P, Kunz L (2001) Cyclic slip localisation and fatigue crack initiation in fcc single crystals. Mater. Sci. Eng. A 314:75–80. 39. Tada H, Paris PC, Irwin GR (1973) The stress analysis of cracks handbook. 3rd ed, ASME Press. 40. Murakami Y, Endo M (1994) Effects of defects, inclusion and inhomogeneities on fatigue strength. Fatigue 16:163–182. 41. Beretta S, Blarasin A, Endo M, Giunti T, Murakami Y (1997) Defect tolerant design of automotive components. Int. J. Fatigue 19:319–333. 42. Mayer H, Papakyriacou M, Zettl B, Stanzl-Tschegg SE (2003) Influence of porosity on the fatigue limit of die cast magnesium and aluminium alloys. Int. J. Fatigue 25:245–256. 43. Nie BH, Zhao ZH, Liu S, Chen DC, Ouyang YZ, Hu ZD, Fan TW, Sun HB (2018) Very high cycle fatigue behavior of a directionally solidified Ni-base superalloy DZ4. Materials 11:98–109. 44. Lecomte-Beckers J (1988) Study of microporosity formation in Nickel-base superalloys. Metall. Trans. A 19:2341–2348. 45. Elliott AJ, Tin S, King WT, Huang SC, Gigliotti MFX, Pollock TM (2004) Directional solidification of large superalloy castings with radiation and liquid-metal cooling: A comparative assessment. Metall. Mater. Trans. A 10:3221–3231. 46. Liu CB, Shen J, Zhang J, Lou LH (2010) Effect of withdrawal rates on microstructure and creep strength of a single crystal superalloy processed by LMC. J. Mater. Sci. Technol. 26: 306– 310. 47. Liu L, Huang TW, Qu M, Liu G, Zhang J, Fu HZ (2010) High thermal gradient directional solidification and its application in the processing of nickel-based superalloys. J. Mater. Process. Technol. 210:159–165. 48. Brundidge CL, Pollock TM (2012) Processing to fatigue properties: benefits of high gradient casting for single crystal airfoils. In: Huron ES, Reed RC, Hardy MC, Mills MJ, Montero RE, Portella PD, Telesman J (ed) Superalloys 2012. TMS, Warrendale, p 379–385.

Effect of Re on Long-Term Creep Behavior of Nickel-Based Single-Crystal Superalloys for Industrial Gas Turbine Applications Fan Lu, Longfei Li, Stoichko Antonov, Yufeng Zheng, Hamish L. Fraser, Dong Wang, Jian Zhang, and Qiang Feng

Abstract

Understanding the long-term creep behavior (creep lives longer than 5,000 h) of nickel-based single-crystal (SX) superalloys is of great interest for the service safety of industrial gas turbine (IGT) blades. However, understanding the influence of various factors on long-term creep behavior of nickel-based SX superalloys has always been a challenge. In this study, the creep behavior of nickel-based SX superalloys with or without 2 wt.% Re addition were investigated at 900 °C and 200 MPa. Microstructural characterization was systematically conducted to elucidate the microstructural evolution of the experimental alloys after creep rupture tests. The results indicated that the creep lifetime was increased significantly by 2 wt.% Re addition, which dramatically reduced the creep strain rate and caused the lattice misfit of c/c′ phases to inverse from positive to negative, leading to N-type rafting exhibiting a stronger barrier to dislocations during creep. The evolution of lattice misfit of the c/c′ phases was closely associated with the elemental partitioning behavior between the c/c′ phases, which was determined as a consequence of “time accumulation effect” and corresponding elemental diffusion special for prolonged time. This study provides a new insight in the effect of Re on the long-term creep life and microstructure F. Lu
L. Li
S. Antonov
Q. Feng (&) State Key Laboratory for Advanced Metals and Materials, University of Science and Technology Beijing, Beijing, 100083, China e-mail: [email protected] Y. Zheng Department of Chemical and Materials Engineering, University of Nevada Reno, Reno, NV 89557, USA H. L. Fraser Center for the Accelerated Maturation of Materials, The Ohio State University, Columbus, OH 43212, USA D. Wang
J. Zhang Superalloys Division, Institute of Metal Research, Chinese Academy of Sciences, Shenyang, 110016, China

evolution of nickel-based SX superalloys. Such knowledge will be helpful to provide the guideline of superalloys design for large-scale IGT blades. Keywords







Long-term creep Single crystal misfit Elemental partitioning




Re addition




Lattice

Introduction Nickel-based single-crystal (SX) superalloys are widely utilized for hot section components, such as turbine blades, where excellent high-temperature properties and environmental resistance are required [1]. Considering the significance of the creep performance for industrial applications, so far research activities in this field have been mainly aimed at short-term performance under high temperature and low stress, known as typical service conditions for aeroengine turbine blades [2, 3]. With regard to the particularity of service condition for aeroengine turbine blades, the creep mechanism by accelerated thermal–mechanical coupling experiments (temperature: 1050–1100 °C, creep lives shorter than 1,000 h) is basically applicable to the whole service life of blades. In contrast, the service temperature for industrial gas turbine (IGT) blades is usually no more than 950 °C; nevertheless, the stable working time is usually as long as thousands of hours [4]. Undoubtedly, the creep mechanism developed previously cannot provide a suitable guidance for the composition design and microstructure control of SX superalloys for long-term application of IGT blades, although some research has already focused on the long-term creep behavior and microstructural evolution using typical commercial nickel-based SX superalloys [5, 6]. Therefore, it is necessary to explore the long-term creep mechanism of SX superalloys under the service conditions of IGTs (where creep lives are longer than 5,000 h).

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_21

218

Effect of Re on Long-Term Creep Behavior of Nickel-Based …

219

For nickel-based SX superalloys, it is widely acknowledged that Re is necessary for achieving the required mechanical properties, and the addition of a small amount of Re can significantly improve the short-term high-temperature creep strength [1, 7–9]. To date, although many studies associated with long-term creep were conducted using alloys with Re addition, recent publications mainly focus on the mechanical properties and microstructural evolution [10]. Limited studies focus on the effect of Re on long-term creep behavior and deformation mechanisms of nickel-based SX superalloys for prolonged time. Early studies suggest that the creep behavior of nickel-based SX superalloys mainly depend on the microstructural evolution, dislocation movements, and lattice misfit [3, 11], while their interactive effect on the long-term creep behavior still remains unclear. In comparison with the short-term creep, the “time accumulation effect” on the long-term creep becomes more apparent, and the diffusion of alloying elements becomes more sufficient, which might lead to the prominent effect of elemental partitioning on the creep behavior of nickel-based SX superalloys. In addition, early studies also reported the beneficial effect of Re on improving hot corrosion resistance of nickel-based SX superalloys designed for IGT blades, which promotes further investigations on the mechanical properties of Re-containing alloys [12]. To design new nickel-based SX superalloys for IGTs with regard to both temperature capability and long-term mechanical performance, a deeper understanding of interactive effects of parameters on the creep behavior of nickel-based SX superalloys and the effect of Re addition are required. In this study, the effect of Re on the long-term creep behavior of nickel-based SX superalloys for IGTs was studied by means of multi-scale characterization, and the synergistic relationships among the elemental partitioning behavior, lattice misfit, micro/submicronstructure, and creep life was established.

Experimental Two experimental alloys were designed based on the first-generation nickel-based SX superalloys used in IGT blades without or with 2wt.% Re addition, named as alloy 0Re5Ta and alloy 2Re5Ta, respectively. Both alloys were directionally solidified (DS) as single-crystal bars (15 mm in

diameter and 120 mm in length) using the conventional Bridgman method at the Institute of Metal Research, Chinese Academy of Science. The primary dendrite arm spacing (PDAS) was measured as 383.7 and 342.5 lm for alloy 0Re5Ta and alloy 2Re5Ta, respectively. SX bars were prepared with the longitudinal axes within 5 degree from [001] orientation and subject to a full heat treatment process, including a two-step solution treatment and a two-step aging treatment, defined as the “Initial” state. The alloy compositions and heat treatment procedure are listed in Table 1. All of the following creep tests and characterizations were based on the initial state. NETZSCH STA 449C differential scanning calorimetry (DSC) was used to measure the c′ solvus temperatures of the experimental alloys with a heating rate of 10 ° C/min. The c′ solvus temperatures were determined at 1238 and 1227 °C for alloy 0Re5Ta and alloy 2Re5Ta, respectively. After the full heat treatment, the bars were machined, along the longitudinal orientation, into tensile creep specimens with gauge length of 25 mm and a gauge diameter of 5 mm. Creep rupture tests were conducted at 900 °C and 200 MPa for both alloys until rupture followed by air cooling. Metallographic specimens for microstructural investigation were prepared on cross sections for the initial state samples and longitudinal sections for creep rupture samples, 5 mm away from the fracture surfaces to avoid the necking regions. The specimens were etched in a solution of 1% HF, 33% CH3COOH, 33% HNO3, and 33% H2O (by volume). The microstructures were examined using a ZEISS SUPRA 55 field-emission scanning electron microscope (FE- SEM) equipped with an energy-dispersive X-ray spectroscope (EDS). In order to quantitatively characterize the microstructural evolution, image analysis software (Image-J and Image-Pro Plus) were used to measure the size of c′ precipitates and the channel width of c phase in dendrite cores. The volume fraction of c′ precipitates was determined by photoshop software using the standard point count method according to the Chinese Standard GB/T 15749-2008, which is similar to ASTM E562-2008. The sub-microstructures were observed by transmission electron microscopy (TEM). Thin TEM disks were cut perpendicular to the stress axis of the creep samples and were manually ground to 60 lm thickness. Then, the TEM foils were prepared by twin-jet thinning technique in a solution of 8% HClO4, 4% CH3COOH, 88% C2H5OH at −25 °C and 38 V.

Table 1 Compositions of the experimental alloys (in wt%) Alloy

Ni

Co

Cr

Al

Mo

W

Ta

Ti

Re

0Re5Ta

Bal.

10

10

4

0.5

4.5

5

3

0

2Re5Ta

Bal.

10

10

4

0.5

4.5

5

3

2

Full Heat Treatment: 1230 °C/2 h + 1255 °C/2 h; 1130 °C/4 h + 870 °C/20 h for alloy 0Re5Ta 1240 °C/2 h + 1265 °C/2 h; 1130 °C/4 h + 870 °C/20 h for alloy 2Re5Ta

220

The sub-microstructural observation of the thin foils was conducted using a FEI TECHNAI F30 TEM operated at 300 kV. The lattice parameters of c and c′ phases were determined by BRUCKER D8 Advanced high-resolution X-ray diffraction (XRD) at 900 °C. Single-crystal bulk plates parallel to (001) for the initial state samples or (010) plane for creep samples (away from the necking regions) were cut into 1 mm in thickness for testing. To minimize the effect of specimen oxidation, the chamber was evacuated at room temperature prior to heating. During the measurements, the sample holder and heating plate assembly were placed in a vacuum chamber at 1 indicates elements partitioning preferentially to the c phase, represented by the bars in the right side of the figure. By contrary, ki0 1000 °C) accelerates microstructure evolution to decrease internal elastic strain [6, 7]. In the previous studies, a room-temperature (RT) PD was introduced to the AM1 Ni-based SX superalloy in between solution heat treatment and aging heat treatments to mimic accidental PD during manufacturing. During subsequent aging treatment at 1100 °C, precipitate coarsening and deformation-pore nucleation were enhanced in the vicinity of pre-deformation slip bands [8, 9]. This pre-deformation between the solution and the aging drastically decreased creep life and ductility of AM1 at temperatures above 950 °C and rupture specimen showed planar fracture surface parallel to a {111} pre-deformation slip plane [8]. Under

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_23

240

High-Temperature Pre-deformation and Rejuvenation Treatment …

241

creep loading, microstructure coarsened band in the material (formerly {111} slip band) accumulated creep damage locally and intensively by evolving into inclined c′-raft structure [9]. Faster microstructure coarsening in the inclined c′-raft enhanced nucleation, coarsening, and coalescence of creep voids in the band. Recrystallization started from enlarged voids and spread along the band which ultimately became a fracture plane [8, 9]. Previous studies had focused on the effect of pre-deformation at RT that reproduces a microstructure of a nonconforming component from Safran Aircraft Engines’ foundry [8, 9]. Pre-deformation at higher temperatures is another interest considering possible PD source such as thermal contraction during cooling sequence of the solution treatment [1]. Furthermore, concept of rejuvenation treatment that dissolves c′ precipitates and restores serviceable microstructure can be effective to erase detrimental effect of pre-deformation [10–15]. Scientific study of these production-related situations is essential for controlling yield rate of a foundry. To understand the effect of PD during production, microstructure and creep properties of Ni-based SX superalloy AM1 after different pre-deformations and heat treatment procedures were examined. The additional supersolvus rejuvenation heat treatment was applied to the SX superalloy with different tensile plastic damage before aging treatments to verify rejuvenation capability. In addition, effect of RT PD and rejuvenation capability was similarly evaluated in third-generation Ni-based SX superalloy CMSX-4 Plus, which is a possible candidate for non-cooled LPT blades.

1050 °C/140 MPa and at 850 °C/500 MPa, and for CMSX-4 Plus at 1150 °C/110 MPa and at 1050 °C/200 MPa using radiant creep furnaces used in the previous studies [8, 9]. Creep test was performed once for each material and creep testing condition. Both before pre-deformation and creep test, specimen’s cylindrical surface was mechanically polished up to P4000 SiC abrasive paper to remove residual stresses. Cylindrical specimens were cut in either (010) or (110) plane after each preparation stages or creep tests, mechanically polished down to 1-µm diamond powder spray, and then chemically etched using aqua regia. Their microstructures were characterized using optical microscope and field emission gun scanning electron microscopy.

Experimental Procedures Cast SX bars (14 mm diameter, longitudinal crystal orientation within 10° from direction) of Ni-based SX superalloys AM1 [16] (first generation) and CMSX-4 Plus [17] (third generation) were prepared by the Bridgman directionally solidifying method. They were solution heat-treated and machined into tensile testing specimens with same geometry as previous studies (14 mm gauge length and *4.0 mm diameter) [8, 9, 18]. A plastic strain was introduced to the AM1 specimens by tensile PD at RT, 750, and 950 °C with a strain rate of 5.0  10−4/s. CMSX-4 Plus was pre-deformed only at RT. Two stages of aging heat treatment were applied after PD. Additional rejuvenation treatment was added after PD before aging treatments to remove detrimental consequences of the pre-deformation. Chemical composition of alloys and heat treatment conditions used in this study is listed in Tables 1 and 2, respectively. Rejuvenation treatment and aging treatments were performed using resistive furnace in air with accuracy of ±2 °C. Tensile creep tests were performed for AM1 at

Results Tensile Plastic Deformations and Aging Treatments Typical tensile curves of pre-deformation (plastic strain *0.8%) are shown in Fig. 1 with their yield strength labeled. The highest tensile strength at 750 °C and decreased tensile strength at 950 °C compared to RT are common characteristics that can be explained by c/c′ misfit and deformation processes in a Ni-based superalloy consisting of the FCC disordered matrix and ordered L12 structure [20, 21]. Two separated curves shown as Fig. 1b are PD at 950 ° C paused at yield point for about 1 min and resumed until the target plastic strain. Dotted line shows stress relaxation and creep deformation occurred during the pause. In the as-solutioned state, AM1 has higher yield strength compared to CMSX-4 Plus at RT which is consistent with the tensile properties after full heat treatment [22]. Specimens with higher or lower plastic strain were obtained by simply interrupting tensile deformation later or earlier. Microstructure after PD (plastic strain *0.8%) at RT and at 750 °C showed similar precipitation coarsening in both parallel and perpendicular directions to the tensile direction during subsequent aging (Fig. 2a, b). Tensile and compressive residual stresses in the vicinity of {111} slip planes activated plasticity within the c phase during aging treatment at 1100 °C to form precipitates with tendency of N-type or P-type directional coarsening [7–9]. After PD at 950 °C and following aging treatments, precipitates are no longer cuboidal for all area and they show no sign of a band like other two pre-deformation temperatures (Fig. 3). c′ precipitates after PD at 950 °C and aging treatments tend to coarsen in the transverse direction to the deformation direction (Fig. 3a, b), which can be called “pre-rafted” structure. Tendency of pre-rafts became shorter and disturbed as plastic strain increases from 0.36 to 2.17%. Comparing Fig. 3b, d, continuous tensile deformation results in larger

242

S. Utada et al.

Table 1 Nominal chemical composition of alloys in weight percentage (Ni bal.) Alloy

Cr

Co

W

Mo

Ta

Al

Ti

Hf

Re

Solvus (°C)

AM1 [16]

7.6

6.7

5.6

2.0

8.0

5.2

1.2

0.05

–

1268 ± 5 [19]

CMSX-4 Plus [17]

3.5

10.0

6.0

0.6

8.0

5.7

0.85

0.1

4.8

1334 ± 5 [19]

Table 2 Heat treatment conditions of alloys Alloy

Solution

Rejuvenation

Aging 1

Aging 2

AM1 [16]

1300 °C, 3 h, AQ

1270–1290 °C, 20 min, AQ

1100 °C, 5 h, AQ

870 °C, 16 h, AQ

CMSX-4 Plus [17]

1313 °C, 2 h ! 1318 °C, 2 h ! 1324 °C, 6 h ! 1335 °C, 6 h, AQ

1310–1335 °C, 20 min-1 h, AQ

1165 °C, 6 h, AQ

870 °C, 20 h, AQ

AQ air quench, GFC gas fan cooling

1400

PD at 750°C (AM1)

1200

1138MPa

Stress (MPa)

PD at RT (AM1)

990MPa

1000

820MPa

800

PD at RT (CMSX-4Plus)

695MPa (a) (b) PD at 950°C (AM1)

600 400 200 0

0

0.5

1

1.5

2

Strain (%)

Fig. 1 Tensile curves of pre-deformation at RT, 750 °C, and 950 °C for AM1 and at RT for CMSX-4 Plus. Two types of PD at 950 °C: continuous deformation (a) and paused at yield point (b)

precipitates whereas the paused tensile deformation results in smaller precipitates with round edges and homogeneous size distribution. The resulting microstructure after various PD procedures suggests that the transition temperature from the shearing regime to the c/c′ interface by-passing regime is in between 750 and 950 °C [21, 23]. Microstructures of CMSX-4 Plus specimens with PD at RT followed by first and second aging treatment are shown in Fig. 2c, d. Precipitates were no longer cuboidal and coarsened in vicinity of former {111} slip planes, due to internal elastic strain around the slip plane. Difference from AM1 is the dislocation network structure in the c/c′ interface appeared in the band after full heat treatment (Fig. 2d).

Rejuvenation Treatment After Plastic Deformation Microstructure is successfully restored for AM1 specimen with RT PD by adding rejuvenation treatment at 1290 °C/

20 min (Fig. 4c, similar result for PD at 750 °C). Because rejuvenation treatments under 1275 °C resulted in coarsening of precipitates (Fig. 4a, b), appropriate rejuvenation temperature seemed to be in the 1280–1290 °C temperature range. Dendritic structure is clearer in the low-mag image shown in Fig. 5a compared to Fig. 5b, and the brighter area corresponds to interdendritic area with non-cubic precipitates. This kind of contrast difference remains in wider area when plastic strain applied at RT is higher or when the specimen is pre-deformed at 950 °C (Fig. 5d), meaning that microstructure of those areas has not been restored completely. Figure 6 is an example of the contrast border after rejuvenation and subsequent aging on the specimen with PD at 950 °C. Slip plane traces appear as coarsened precipitates after rejuvenation at 1280 °C when the specimen has been submitted to higher plastic strain (parallel to dotted lines in Fig. 5c). Such traces disappeared when rejuvenation was at 1290 °C; however, the contrast boarders were remaining. Considering all the results, rejuvenation treatment should be applied to AM1 at 1290 °C for 20 min to restore optimal microstructure. Recrystallization was observed in the specimens with higher plastic strain (=2.17%) applied at 950 °C after rejuvenation at lower temperature of 1280 °C (Fig. 7a). Recrystallization is confirmed to occur in the primary dendrite arms because grains in that region have cuboidal c′ precipitates that facing different orientation to the (010) cut surface of the specimen. Recrystallized grains are connected to the specimen surface which was similarly seen in another study about the impact of PD on the recrystallization sensitivity during solution treatment [2]. Specimen that was paused after reaching the yield stress (Fig. 1b) had its microstructure fully recrystallized during the rejuvenation at 1290 °C, including the interdendritic area (Fig. 7c). Results of rejuvenation tests on CMSX-4 Plus pre-deformed at RT are shown in Fig. 8. Similarly to AM1, rejuvenation temperature below c′ solvus at 1310 °C did not help but degraded the microstructure with severe precipitate

High-Temperature Pre-deformation and Rejuvenation Treatment …

(b)

(a)

Tensile direction

Fig. 2 Microstructure of AM1 (a, b) and CMSX-4 Plus (c, d), after PD at RT and subsequent two aging treatments. Dotted lines in (a) and (c) are indicating areas of microstructure coarsened band and ovals in (d) are showing dislocation network structures

243

1µm

5µm

AM1 PD at RT / Strain = 1.00%

(c)

(d)

(d) 1µm

5µm

CMSX-4 Plus PD at RT/ Strain = 0.79% Fig. 3 Microstructure of AM1 after PD at 950 °C with different plastic strain followed by two aging treatments. Deformation of (a–c) are by Fig. 1a-type continuous deformation and (d) is by Fig. 1b-type deformation paused after yielding

(b) Strain = 0.87% / type (a) Tensile direction

(a) Strain = 0.36% / type (a)

1µm (c) Strain = 2.17% / type (a)

(d) Strain = 1.01% / type (b)

244

S. Utada et al.

Fig. 4 Microstructure of AM1 with PD at RT (plastic strain = 0.87%) followed by rejuvenation and two aging treatments

5µm

5µm (b) Rejuv at 1275°C Tensile direction

(a) Rejuv at 1270°C

5µm

(c) Rejuv at 1290°C

Interdendritic

Fig. 5 Dendritic scale images of AM1 with pre-deformed at RT (a–c) and at 950 °C (d) followed by rejuvenation and two aging treatments. Dotted lines in (c) are showing direction of slip band traces

(a) RT/Strain = 0.87% Rejuv at 1280°C

(b) RT/Strain = 0.87% Rejuv at 1290°C

Tensile direction

Dendrite core

5µm

1mm (c) RT/Strain = 1.86% Rejuv at 1280°C

coarsening (Fig. 8a). Slip plane traces were not visible after rejuvenation at 1325 °C. However, precipitates in interdendritic areas are still coarse which indicates that the rejuvenation temperature is not yet optimal (Fig. 8b).

(d) 950°C/Strain = 0.87% Rejuv at 1280°C

Microstructure is restored both in primary dendrite arms and in the interdendritic areas after rejuvenation at 1330 °C (Figs. 8c, d). Therefore, suitable rejuvenation condition for CMSX-4 Plus with PD at RT is at 1330 °C for 1 h.

High-Temperature Pre-deformation and Rejuvenation Treatment … Fig. 6 Microstructure of AM1 after PD at 950 °C with plastic strain of 0.87% followed by rejuvenation at 1290 °C for 20 min and two aging treatments. Dotted line indicates border between full-resorted dendrite arm and interdendritic area

245

Eutectic

Tensile direction

Dendrite arm

Interdendritic area 10µm

Dendrite arm

Interdendritic area

5µm

Creep Behavior with Various Preparation Conditions Result of creep rupture tests on AM1 at 1050 °C/140 MPa and at 850 °C/500 MPa are presented in Fig. 9. Rejuvenation condition for creep specimens was at 1290 °C/20 min. Creep results for specimens without and with PD at RT are adopted from the previous studies [8, 9]. Additional rejuvenation treatment is shown to be very effective to restore original creep properties of AM1 for both creep testing conditions. This is clearly owing to restored microstructure by rejuvenation after PD (Fig. 4c). Specimen with PD at 950 °C also performed well at 1050 °C with smaller primary creep strain. However, it showed very poor creep durability at 850 °C, even shorter than the specimen pre-deformed at RT. Microstructures of AM1 specimens without PD and with PD at 950 °C after creep tests at 1050 °C/140 MPa and at 850 °C/500 MPa are shown in Fig. 10. They showed similar ductility and appearance after creep rupture test at 1050 ° C/140 MPa. However, c′-rafts in a dendrite core far from fracture surface are coarser in the specimen pre-deformed at 950 °C (Fig. 10b) compared to the other one without PD (Fig. 10a). On the other hand, rupture characteristics at 850 °C/500 MPa were clearly different. Directional coarsening with former-matrix c-particles structure was observed in the area far from fracture surface (Fig. 10c) which is similar to the AM3 after creep test at the same condition [24]. In comparison, the specimen with PD at 950 °C has

less and smaller creep voids with wavy c′-rafts (Fig. 10d). Meanwhile, AM1 specimen with PD at RT after the creep rupture at 850 °C/500 MPa has two different microstructures, outside and inside the microstructure coarsened band as presented in Fig. 11. Examples of a crack propagating from a coarsened void lying on the intersection of two different bands are indicated by arrows in Fig. 11a. Microstructure outside the band is similar to that of a specimen without PD consisting of c-channel and c-particles, but less developed compared to Fig. 10c because of shorter creep time. c′ precipitates inside the band have further coarsened both vertical and horizontal direction from microstructure before creep, and a shearing trace is also observed (Fig. 11b). To understand the effect of alloy composition on the creep durability and rejuvenation capability of pre-deformed specimen, similar tests were conducted on high-Re containing Ni-based SX superalloy CMSX-4 Plus. Figure 12 presents creep curves of CMSX-4 Plus tested at 1150 °C/110 MPa and at 1050 °C/200 MPa. Creep life of CMSX-4 Plus drastically decreased at 1150 °C/110 MPa if the material was pre-deformed between solution and aging treatments. Inclination of c′-rafts, creep void coarsening, and crack propagation along the microstructure coarsened band on the {111} planes are processes lead to very short creep life and a planar fracture surface shown in Fig. 13a, c. This is the same creep mechanism occurred in AM1 with PD at RT at 1050 °C/140 MPa [9]. As determined previously,

246

S. Utada et al.

Fig. 7 Recrystallized microstructures of AM1 with PD at 950 °C after rejuvenation treatment. (a, b) are from Fig. 1a, and (c) are deformed like Fig. 1b. Arrows are indicating original grains which are not recrystallized

100µm

100µm

(b) Strain = 2.17 % (a) Rejuv. 1290°C/20min Tensile direction

(a) Strain = 2.17 % (a) Rejuv. 1280°C/20min

1mm (c) Strain = 1.01 % (b), Rejuv. 1290°C/20min

Tensile direction

Fig. 8 Microstructure of CMSX-4 Plus with PD at RT (plastic strain = 0.79%) after rejuvenation and two aging treatments

5µm (b) Rejuv. at 1325°C/1h Tensile direction

(a) Rejuv. at 1310°C/20min

5µm (c) Dendrite core

(d) Interdendritic area Rejuv. at 1330°C/1h

High-Temperature Pre-deformation and Rejuvenation Treatment … 20

(a) 1050 C/140MPa

Without PD PD at 950 C Strain=0.79%

Creep strain (%)

15

10

PD at RT + Rejuv. Strain=0.71% PD at RT Strain=0.74%

5

0 0

100

200

300

400

Time (h) 25

(b) 850 C/500MPa

Creep strain (%)

Without PD

PD at 950 C Strain=0.71%

20

PD at RT + Rejuv. Strain=0.79%

15

PD at RT Strain=0.79%

10

5

0 0

50

100

150

200

250

300

Time (h)

Fig. 9 Creep curves of AM1 with different preparation processes tested at 1050 °C/140 MPa (a) and at 850 °C/500 MPa (b)

CMSX-4 Plus was rejuvenated at 1330 °C for 1 h after PD at RT and the creep properties have been restored to a comparable level to the original material. Contrary to creep at 1150 °C/110 MPa, the creep resistance of CMSX-4 Plus at 1050 °C/200 MPa is hardly affected by a pre-deformation. Failure mechanisms at this creep condition were similar regardless of PD; however, crack propagation in predeformed specimen seems to be starting from bigger creep voids on the band (Fig. 13b, d). Although the microstructure coarsened band can be identified as distorted c′-rafts (Fig. 13d), creep damage accumulation was not strongly pronounced compared to the creep at 1150 °C/110 MPa or in case of AM1 [9].

Discussion Microstructure Behavior of AM1 After Pre-deformation Different precipitate coarsening behaviors at temperatures below 750 °C (Fig. 2) and at 950 °C (Fig. 3) are due to the transition of tensile deformation mechanisms [20, 21]. As

247

discussed in other studies, accelerated precipitation coarsening after lower temperature PD (
750 °C) is due to the relaxation of the elastic stress field surrounding slip plane during aging above 1000 °C [7–9]. Horizontal directional coarsening observed in the specimen with PD at 950 °C (strain < 0.9%) and subsequent agings is attributed to the deformation mechanism that proceeds by dislocations introduced in c matrix side of c/c′ interface and its inhomogeneous distribution between horizontal and vertical channels. Relaxation of c/c′ misfit strain by rearrangement of introduced dislocations during aging treatments is the main driving force of prior mild directional coarsening, or “pre-rafting.” Such coarsening mechanism is similar to the c′-rafting [25, 26], and similar subsequent aging effect after high-temperature PD has been discussed previously [7, 27]. Two deformation types at 950 °C, continuous and paused at yield point, have been presented as stress–strain curve (Fig. 1) and microstructures (Fig. 3). Trend of increasing plastic strain with more disturbed microstructure is indirectly telling that dislocation migration is anisotropic at the beginning and then it migrates in both horizontal and vertical matrix channels as plastic strain increases. Clear difference between Fig. 3b, d gives assumption that creep and stress relaxation during temporary interruption (Fig. 1b) changed dislocation distribution and affected resumed deformation mechanism. Round and smaller precipitate in Fig. 3d is an indication that Fig. 1b-type deformation introduced higher dislocation density than Fig. 1a-type. However, the detail of the coarsening mechanism is not clear from the observed results.

Creep Properties of AM1 If a target creep life for application is determined to be ±10% of an original material, pre-rafted microstructure (Fig. 3b) of the specimen pre-deformed at 950 °C is acceptable level under creep condition of 1050 °C/140 MPa. This mild c′-rafting is assumed to continue after applying creep load at 1050 °C to have the microstructure shown in Fig. 10b. During creep deformation, this specimen showed very small primary creep strain compared to other three specimens. Decreased primary creep strain by pre-straining was also reported by Drew et al. [28]. This suggests that the plasticity involved in directional coarsening is not so active, resulting in degraded c′-raft (Fig. 10b) and decreased creep life compared to the original material [28]. Although it was not well developed like Fig. 10a, this homogeneous microstructure helped avoiding very intense localized damage accumulation in the microstructure coarsened band that was observed in AM1 with PD at RT crept at 1050 °C/140 MPa [9].

(a)

(b)

Tensile direction

Fig. 10 Microstructure of AM1 specimens after creep test at 1050 °C/140 MPa (a, b) and at 850 °C/500 MPa (c, d). Specimen without PD (a, c) and with PD at 950 °C (b, d), all observed at 5 mm from rupture surface. Plastic strain = 0.71% for both (b) and (d)

S. Utada et al.

1050°C/140MPa

248

5µm

(d) Tensile direction

850°C/500MPa

(c)

1µm Without PD

(a) Tensile direction

Fig. 11 Microstructure of AM1 specimen with PD at RT (plastic strain = 0.79%) after creep test at 850 °C/500 MPa. Dotted lines in (a) are showing microstructure coarsened band that came from pre-deformation. Precipitate scale image inside the band is shown in white rectangle (b) and an area surrounded by dotted line is showing microstructure shearing. Black arrows are indicating cracks initiating from the void

With PD at 950°C

(b)

(b) 1µm

Creep properties of pre-deformed AM1 at 850 °C/500 MPa have not been deeply discussed [8]. Firstly, the original material performed similarly to other first generation superalloys [24, 29, 30]. Dislocations fill up matrix during primary and secondary stage, forms dislocation network that surrounds c′ precipitates, and then precipitates are sheared by dislocations during the tertiary creep stage [29, 30]. Although temperature is not so high, microstructure

5µm

after creep rupture shows that both precipitation coarsening and void formation took place during deformation. On the contrary, creep loading at 850 °C/650 MPa on a third-generation alloy with lower diffusivity exhibits clear slip band and no directional coarsening within the ruptured specimen [31]. Since coarsened precipitates on a specimen with PD at RT (Fig. 2b) have round edge that suggests dislocations migrating in the matrix side of c/c′ interface, it is

High-Temperature Pre-deformation and Rejuvenation Treatment … 30 Without PD 1050 /200MPa

Creep strain (%)

CMSX-4 Plus

PD at RT Strain=0.64% 1050 /200MPa

20

PD at RT Strain=0.50% 1150 /110MPa

10

Without PD 1150 /110MPa

PD at RT + Rejuv. Strain=0.57% 1150 /110MPa

0 0

20

40

60

80

100

120

140

Time (h)

Fig. 12 Creep curves of CMSX-4 Plus with different preparation processes tested at 1150 °C/110 MPa and 1050 °C/200 MPa

likely that already decent amount of dislocations existing in the microstructure coarsened band that enhance dislocations penetration/shearing into the c′ precipitates. In the meantime, crystallographic creep void growth and cracks initiating from this void (Fig. 11a) suggest that faster internal diffusion is also involved in the creep damage localization to the band like the case for creep at higher temperatures [9]. Similarly, very short creep life of AM1 with PD at 950 °C in this higher stress creep can be explained from microstructures both before and after the creep test. This is like the pre-rafted CMSX-2 with very high density of particle shearing dislocations [29], or like AM3 with very slow cooling from c′ solvus that exhibits irregular and larger precipitates [24]. As discussed previously, existence of dislocations in c matrix after full aging treatments is predicted. Creep curve of AM1 with PD at 950 °C in Fig. 9b does not have primary creep stage, and creep rate is accelerating from the beginning. Therefore, dislocation shearing assumed to start at very early stage of creep deformation and easily transitioned into the failure. Despite decreased creep life, specimen with PD at 950 °C maintained its ductility at similar level to the original AM1 meaning that this deformation occurred homogenously. It can be concluded that high-temperature pre-straining of Ni-based SX superalloy component must be avoided because of its poor creep properties at higher stress condition.

Rejuvenation Capability After Pre-deformation Microstructure and creep properties were restored by rejuvenation treatment (without hot-isostatic pressing). Rejuvenation treatment is known to effectively restore creep damaged Ni-based superalloys [10–14]; however, this study is the first one to show the possibility of rejuvenating SX superalloys with tensile PD. Although recrystallization is

249

always a risk for rejuvenation, research on the condition which may cause recrystallization is not available. To discuss rejuvenation capability, some rejuvenation results on crept Ni-based SX superalloys are taken from other studies [11–13] and they are plotted in Fig. 14. According to the result from this study, PD at higher temperature can cause recrystallization during rejuvenation treatment. Similar relationship of recrystallization sensitivity after high-temperature PD (especially at PD temperature between 900 and 1000 °C) with subsequent annealing above 1280 °C on a SX superalloy has been reported in the study for understanding recrystallization during solution treatment [4, 5]. Since deformation at 950 °C is driven by dislocation by-passing at c/c′ interface [21], plastic strain is macroscopically homogeneous. However, higher dislocation density than PD at RT should be accumulated at the c side of the c/c′ interface that enhances microstructure evolution in all areas during aging (Fig. 3). Dislocation filling up in the interface is probably the source of recrystallization occurred during rejuvenation treatment. In other study about rejuvenation after creep deformation at 950 °C, microstructure was restored without recrystallization even with much higher plastic strain (up to 5%, Fig. 14) [13, 15]. In these studies, typical directionally coarsened microstructures were observed before rejuvenation treatment and the precipitate size is obviously larger compared to the ones observed in the present study. Meanwhile, compressive plastic deformation at 980 °C is reported to introduce stacking faults into c′ precipitates during deformation and it may facilitate recrystallization during solution treatment [4, 5]. However, such substructure is unlikely to appear neither during creep deformation with lower applied stress nor during tensile deformation with lower strain rate. In René N5 after creep at 950 °C, recrystallization was reported in the vicinity of carbide which may have different morphology compared to carbon-free superalloys [11]. These are the indications that dislocation types, distribution, and quantity determined by deformation mechanisms have different reordering and annihilating behavior during dissolution of c′ phase, and the degree of plastic strain cannot be a simple indicator for boundary condition. Further experiments on the rejuvenation treatment and microstructure analyses are necessary to clearly identify the precise condition that activates recrystallization. From a practical point of view, rejuvenation treatment is a very useful process to put damaged component back into production line. However, some remarks can be made from the experimental results. As presented in Fig. 4a and 8a, pronounced microstructure coarsening can occur if the rejuvenation temperature is too low and dissolution of c′ phase is incomplete. This works as very high-temperature annealing where diffusion and plasticity are extremely active and enhances very fast microstructure evolution. Because creep

250

S. Utada et al.

mm

(a)

(b)

Tensile direction

1mm

(c) Tensile direction

(d)

10µm Fig. 13 Microstructure of CMSX-4 Plus with PD at RT after creep tests at 1150 °C/110 MPa (a, c) and at 1050 °C/200 MPa (b, d). Plastic strain = 0.50% for (a, c) and 0.64% for (b, d)

specimen surface since it acts as free surface during PD and locally higher stored energy can activate recrystallization. Although it did not give huge debit in the creep tests of cylindrical specimen, it may have an impact for actual turbine component that has thin aerodynamic profiles.

Plastic strain before rejuvenation (%)

6 AM1, AM1, strain strain rate = 5.0×10-4 5.0 10-4 /s René 2012 [ref]) René N5, N5, 206MPa 206MPa (Rettberg (Rettberg,et et al., al. [11]) ERBO/1C, 2017 [ref]) ERBO/1C, 160MPa 160MPa (Ruttert (Ruttert,et et al., al. [12]) ERBO/1, 2019 [ref]) ERBO/1, 350MPa 350MPa (Horst (Horst,et etal., al. [13])

4

Recrystallized

Effect of Alloy Compositions

2

0 0

200

400 600 800 Deformation temperature ( )

1000

1200

Fig. 14 Relationship between deformation temperature and plastic strain before rejuvenation treatment for different alloys. Filled plots are showing successful rejuvenation. Plastic strain values adopted from other studies [11–13] are by creep with a constant tensile loading

life of SX superalloys can be significantly reduced [24, 32], this situation is much worse compared to the one without rejuvenation. Another risk is a surface recrystallization which has been observed in rejuvenated specimens of AM1 tested at 850 °C/500 MPa and CMSX-4 Plus tested at 1150 °C/110 MPa. Plastic strain is locally higher at the

Creep behavior of pre-deformed CMSX-4 Plus at 1150 °C/110 MPa was similar to AM1 at 1050 °C/140 MPa. Contrarily, pre-deformation gave less impact on the material’s creep properties at 1050 °C/200 MPa. These differences can be explained by the intrinsic creep behavior of CMSX-4 Plus at 1050 °C and the microstructure before creep tests. Dislocation networks and precipitate edges observed in the microstructure of CMSX-4 Plus (Fig. 2c, d) are the main difference when it is compared with that of AM1 (Fig. 2a, b). This suggests that the c/c′ interfacial coherency is still present in CMSX-4 Plus, at least partly, whereas dislocations have filled up in the matrix of AM1 that decreases interfacial misfit resulting in round edges. Two reasons can be predicted: simply, c/c′ lattice misfit of CMSX-4 Plus is much higher (in absolute value), and

High-Temperature Pre-deformation and Rejuvenation Treatment …

251

deformation was not enough to distinguish misfit strain during aging treatment. Or, plastic strain by a single slip was smaller in CMSX-4 Plus due to a lower yield stress (lower resistance to the precipitate shearing) and stored elastic energy near the slip plane was smaller resulting less dislocation density after aging treatments. Creep curve of CMSX-4 Plus at 1050 °C/200 MPa shows almost no primary stage and gradually increases strain rate. Although c′-rafts are observed in creep ruptured microstructure (Fig. 13d), rafting may not start immediately in this creep condition. Instead, relatively good creep properties of this alloy among other SX superalloys are possibly owing to slower internal diffusion that retards c′-rafting even at 1050 °C. Creep failure seemed to start from former slip plane in pre-deformed specimen (Fig. 13b), but localized damage accumulation was not drastic because of low diffusivity and precipitates in the band were still capable of exhibiting their function by remaining interfacial misfit. And also, with high volume fraction of c′ phase at 1050 °C, creep deformation is more in the domain of topological inversion [33] which the case that damage is spreading over longer distances [9]. Still, pre-deformation has a huge impact on its creep properties at 1150 °C, a creep deformation with pronounced primary, secondary, and tertiary stages. Elongated and inclined c′-rafts formed after primary stage, and the band acted as diffusion paths that enhance creep void nucleation/growth, leading to recrystallization along the band and finally the early failure [9]. Remaining elastic strain field in the band is a possible source of the raft inclination. As a summary, introducing RT pre-deformation severely affects creep durability of Ni-based SX superalloys at a temperature with relatively high internal diffusion rate that activates early formation of c′-rafts. Dislocation scale analyses are necessary for clear understanding of the creep behavior of pre-strained CMSX-4 Plus and its comparison with AM1.

1. Microstructure evolved differently during aging treatments after a plastic deformation in precipitate shearing regime (
750 °C) and in the c/c′ interface dislocation by-passing regime (950 °C). Mild directional coarsening was observed after pre-deformation at 950 °C with subsequent aging treatments. Creep properties at 1050 °C/140 MPa was serviceable; however, this pre-rafted microstructure showed poor creep life at 850 °C/500 MPa. 2. Rejuvenation treatment successfully restored microstructure of AM1 after room-temperature plastic deformation with plastic strain below 1.86% or plastic deformation at 950 °C with plastic strain below 0.87%. It was also successful for CMSX-4 Plus with plastic strain of 0.79%. Suitable rejuvenation conditions are at 1290 °C for 20 min for AM1 and at 1330 °C for 1 h for CMSX-4 Plus. Rejuvenated specimens showed creep properties equivalent to the original materials, and they can be considered as serviceable materials. 3. Room-temperature pre-deformation decreased creep life of CMSX-4 Plus at 1150 °C/110 MPa similarly to AM1 at 1050 °C/140 MPa. On the contrary, impact of a plastic deformation was not significant at 1050 °C/200 MPa for CMSX-4 Plus. Slower microstructure evolutions in pre-deformed CMSX-4 Plus during aging treatments and remaining c/c′ interfacial misfit in the band are possible reasons of minor effect of pre-deformation to the creep at lower temperature.

Conclusions A plastic deformation at room-temperature, at 750 °C and at 950 °C was introduced after solution treatment and before aging treatments. Its effect on microstructure and creep properties of AM1 single-crystal superalloy has been investigated. Rejuvenation treatment has also been introduced after plastic deformation for microstructure restoration. CMSX-4 Plus was similarly tested to understand the effect of chemical composition. Pre-deformation introduced at any temperature can cause drastic creep life reduction of a Ni-based single-crystal superalloy. Rejuvenation treatment after pre-deformation can restore serviceable microstructure for turbine blade applications. Detailed conclusions can be drawn as follows:

Acknowledgements Safran Aircraft Engines is acknowledged for its financial support, for providing material, and for their continuous interest in the subject. Mr. Christophe Audic and Mr. Sebastian Blas (both formerly at Safran Aircraft Engines), Dr. Sandrine Charles, Dr. Elodie Drouelle, and Dr. Nicolas Leriche (all at Safran Aircraft Engines) are gratefully acknowledged for their suggestions and discussions. Mrs. Florence Hamon (at Institut Pprime) is acknowledged for her technical support.

References 1. Pierret S, Etter T, Evans A, Van Swygenhoven H (2013) Origin of localized rafting in Ni-based single crystal turbine blades before service and its influence on the mechanical properties. Acta Mater 61:1478–1488. https://doi.org/10.1016/j.actamat.2012.11.024. 2. Bürgel R, Portella PD, Preuhs J (2000) Recrystallization in single crystals of nickel base superalloys. In: Superalloys 2000. The Minerals, Metals & Materials Society, Warrendale, Pennsylvania, p 229–238. https://doi.org/10.7449/2000/Superalloys_2000_229_ 238. 3. Cox DC, Roebuck B, Rae CMF, Reed RC (2003) Recrystallisation of single crystal superalloy CMSX–4. Mater Sci Technol 19:440– 446. https://doi.org/10.1179/026708303225010731. 4. Li Z, Fan X, Xu Q, Liu B (2015) Influence of deformation temperature on recrystallization in a Ni-based single crystal

252

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16. 17.

18.

S. Utada et al. superalloy. Mater Lett 160:318–322. https://doi.org/10.1016/j. matlet.2015.07.120. Li Z, Xu Q, Liu B (2016) Experimental investigation on recrystallization mechanism of a Ni-base single crystal superalloy. J Alloys Compd 672:457–469. https://doi.org/10.1016/j.jallcom. 2016.02.149. Choi BG, Jo CY, Hong HU, et al (2011) Effect of pre-strain on microstructural evolution during thermal exposure of single crystal superalloy CMSX-4. Trans Nonferrous Met Soc China 21:1291– 1296. https://doi.org/10.1016/s1003-6326(11)60855-8. Sakaguchi M, Ike M, Okazaki M (2012) Microstructural changes in a single crystal Ni-base superalloy induced by plastic straining. Mater Sci Eng A 534:253–259. https://doi.org/10.1016/j.msea. 2011.11.066. Hamadi S, Hamon F, Delautre J, et al (2018) Consequences of a room-temperature plastic deformation during processing on creep durability of a Ni-based SX superalloy. Metall Mater Trans A 49:4246–4261. https://doi.org/10.1007/s11661-018-4748-3. Utada S, Rame J, Hamadi S, et al (2020) Kinetics of creep damage accumulation induced by a room-temperature plastic deformation introduced during processing of AM1 Ni-based SX superalloy. Mater Sci Eng A 789:139571. https://doi.org/10.1016/j.msea.2020. 139571. VanDrunen G, Liburdi J (1977) Rejuvenation of used turbine blades by hot isostatic processing. In: Proceedings of the Sixth Turbomachinery Symposium. p 55–60. https://doi.org/10.21423/ R1MQ1G. Rettberg LH, Tsunekane M, Pollock TM (2012) Rejuvenation of nickel-based superalloys GTD444(DS) and René N5(SX). In: Superalloys 2012. John Wiley & Sons, Inc., Hoboken, New Jersey, p 341–349. https://doi.org/10.1002/9781118516430.ch37. Ruttert B, Bürger D, Roncery LM, et al (2017) Rejuvenation of creep resistance of a Ni-base single-crystal superalloy by hot isostatic pressing. Mater Des 134:418–425. https://doi.org/10. 1016/j.matdes.2017.08.059. Horst OM, Ruttert B, Bürger D, et al (2019) On the rejuvenation of crept Ni-Base single crystal superalloys (SX) by hot isostatic pressing (HIP). Mater Sci Eng A 758:202–214. https://doi.org/10. 1016/j.msea.2019.04.078. Pal J, Srinivasan D, Cheng E (2016) Effect of rejuvenation heat treatment and aging on the microstructural evolution in Rene N5 single crystal Ni base superalloy blades. In: Superalloys 2016. John Wiley & Sons, Inc., Hoboken, New Jersey, p 285–291. https://doi.org/10.1002/9781119075646.ch31. Ruttert B, Horst O, Lopez-galilea I, et al (2018) Rejuvenation of single-crystal Ni-base superalloy turbine blades: Unlimited service life? Metall Mater Trans A 49:4262–4273. https://doi.org/10.1007/ s11661-018-4745-6. Davidson JH, Fredholm A, Khan T, Theret J-MCF (1986) Single-crystal alloy has a nickel base matrix. Patent FR8320986. Wahl JB, Harris K (2016) CMSX-4 Plus single crystal alloy development, characterization and application development. In: Superalloys 2016. John Wiley & Sons, Inc., Hoboken, New Jersey, p 25–33. https://doi.org/10.1002/9781119075646.ch3. Cormier J, Jouiad M, Hamon F, et al (2010) Very high temperature creep behavior of a single crystal Ni-based superalloy under

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

complex thermal cycling conditions. Philos Mag Lett 90:611–620. https://doi.org/10.1080/09500839.2010.489887. Cormier J (2016) Thermal cycling creep resistance of Ni-based single crystal superalloys. In: Superalloys 2016. John Wiley & Sons, Inc., Hoboken, New Jersey, p 383–394. https://doi.org/10. 1002/9781119075646.ch42. Thornton PH, Davies RG, Johnston TL (1970) The temperature dependence of the flow stress of the c′ phase based upon Ni3Al. Metall Trans 1:207–218. https://doi.org/10.1007/bf02819263. Milligan WW, Antolovich SD (1987) Yielding and deformation behavior of the single crystal superalloy PWA 1480. Metall Trans A 18:85–95. https://doi.org/10.1007/bf02646225. Bortoluci Ormastroni LM, Utada S, Rame J, et al (2020) Tensile, low cycle fatigue and very high cycle fatigue characterizations of advanced single crystal nickel-based superalloys. In: Superalloys 2020. Matan N, Cox DC, Carter P, et al (1999) Creep of CMSX-4 superalloy single crystals: effects of misorientation and temperature. Acta Mater 47:1549–1563. https://doi.org/10.1016/s13596454(99)00029-4. Steuer S, Hervier Z, Thabart S, et al (2014) Creep behavior under isothermal and non-isothermal conditions of AM3 single crystal superalloy for different solutioning cooling rates. Mater Sci Eng A 601:145–152. https://doi.org/10.1016/j.msea.2014.02.046. Pollock TM, Argon AS (1994) Directional coarsening in nickel-base single crystals with high volume fractions of coherent precipitates. Acta Metall Mater 42:1859–1874. https://doi.org/10. 1016/0956-7151(94)90011-6. Ichitsubo T, Koumoto D, Hirao M, et al (2003) Rafting mechanism for Ni-base superalloy under external stress: elastic or elastic– plastic phenomena? Acta Mater 51:4033–4044. https://doi.org/10. 1016/s1359-6454(03)00224-6. Sakaguchi M, Okazaki M (2018) Distinctive role of plastic and creep strain in directional coarsening of a Ni-base single crystal superalloy. Mater Sci Eng A 710:121–128. https://doi.org/10. 1016/j.msea.2017.10.085. Drew GL, Reed RC, Kakehi K, Rae CMF (2004) Single crystal superalloys: The transition from primary to secondary creep. In: Superalloys 2004. p 127–136. https://doi.org/10.7449/2004/ Superalloys_2004_127_136. Caron P, Henderson PJ, Khan T, McLean M (1986) On the effects of heat treatments on the creep behaviour of a single crystal superalloy. Scr Metall 20:875–880. https://doi.org/10.1016/00369748(86)90458-8. Pollock TM, Argon AS (1992) Creep resistance of CMSX-3 nickel base superalloy single crystals. Acta Metall Mater 40:1–30. https:// doi.org/10.1016/0956-7151(92)90195-k. Li Y, Wang L, Zhang G, et al (2019) Creep anisotropy of a 3rd generation nickel-base single crystal superalloy at 850 °C. Mater Sci Eng A 760:26–36. https://doi.org/10.1016/j.msea.2019.05.075. Caron P, Khan T (1983) Improvement of creep strength in a nickel-base single-crystal superalloy by heat treatment. Mater Sci Eng 61:173–184. https://doi.org/10.1016/0025-5416(83)90199-4. Epishin A, Link T, Brückner U, Portella PD (2001) Kinetics of the topological inversion of the c/c′-microstructure during creep of a nickel-based superalloy. Acta Mater 49:4017–4023. https://doi. org/10.1016/s1359-6454(01)00290-7.

Evidence of Short-Range Order (SRO) by Dislocation Analysis in Single-Crystal Ni-Based Matrix Alloys with Varying Re Content After Creep Florence Pettinari-Sturmel, Joël Douin, Fabian Krieg, Ernst Fleischmann, and Uwe Glatzel

Abstract

Introduction

TEM characterization of the deformation micromechanisms after interrupted creep test at 980 °C is proposed in the case of two single-crystal Ni-based matrix alloys without rhenium (Re) and with 9 wt% Re. This paper is aimed to propose an identification of the physical parameter at the origin of the strengthening effect of Re. The TEM observations carried out in the alloy with 9 wt % Re indicate that the deformation proceeds through the propagation of long dislocation pile-ups. A leading pair is observed within these pile-ups. As this dislocation configuration is a signature of short-range order (SRO), a quantitative approach is used to evaluate the SRO degree and thus the resistance due to SRO. The experimental positions of the dislocations are used to evaluate the total elastic force experienced by the dislocations. The strength associated with the SRO is evaluated to: c0 = 50 − 60 mJ/m2. The observation of the matrix alloy without Re indicates the presence of both individual dislocations and some short dislocation pile-ups. It is also associated with a lower SRO degree. It can be concluded that the Re effect leads to an increase of the SRO degree. SRO is still present at high temperature (930 °C), whereas it has been shown to disappear in similar matrix alloy with a lower amount of Re. Rhenium appears then to promote SRO. It can be at the origin of its well-known strengthening effect. Keywords

Rhenium




Short-range order




Dislocation pile-ups

F. Pettinari-Sturmel (&)
J. Douin CEMES-CNRS, Université de Toulouse, 29 Rue Jeanne Marvig, BP 94347 31055 Toulouse, France e-mail: [email protected] F. Krieg
E. Fleischmann
U. Glatzel University Bayreuth, Metals and Alloys, Bayreuth, Germany

For several decades, the temperature resistance of Ni-based superalloys has been increased by new processing routes, i.e., changing from wrought to single-crystal alloys and by modification the alloy composition [1]. The huge research activity on Ni-based superalloy optimization has led to identify the ideal microstructure, which consists in a well-defined initial c′ particle size, c′ volume fraction, and a lattice misfit ranging around 0.2–0.5%
. The precipitation strengthening is thus exploited to its maximum. Additionally, solid solution strengthening of the soft c-matrix has been of high interest to further increase the mechanical properties. Since at low creep strains, deformation is mainly constrained to the matrix [2] whose creep resistance can be directly increased by solid solution hardening. Therefore, the improvement of creep resistance by solid solution hardening is another route for increasing the high-temperature strength of Ni-based superalloys. Besides the development of the Ni-based superalloys, several investigations have concerned the solid solution hardening in faced-centered cubic nickel-based alloys [3–15]. These studies have shown that rhenium (Re), tungsten (W), and molybdenum (Mo) are the strongest solid solution-hardening elements. As Re has the strongest effect, it has been introduced in successive superalloy generations since the 1980s. The Re addition has been identified as essential for the high-temperature resistance of single-crystal superalloys as Re has been shown to dramatically improve the high-temperature creep properties, especially slowing down coarsening of c′ phase and rafting. Thereby, the effect of Re has been extensively investigated. A possible explanation for the large Re effect on the creep rate was the formation of rhenium clusters in the matrix. These clusters could act as obstacles against dislocation movement. The existence of these clusters has remained controversial for a long period: rhenium clusters were claimed by several researchers [16, 17] and were identified using 1-D atom probe field ion microscopy, but

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_24

253

254

F. Pettinari-Sturmel et al.

this was later refuted by Mottura et al. using the extended X-ray absorption fine structure technique and more sophisticated statistical analysis of modern 3-D atom probe tomography [18, 19]. The second-generation superalloy DD6 (2 wt% Re) has been investigated by Ge et al. [20] using scanning transmission electron microscopy, high angle annular dark field imaging, and energy dispersive X-ray spectroscopy. They made two different observations: In the initial state (after heat treatment, before creep), Re was randomly distributed in the matrix phase. In a recent paper, Mottura and Reed [21] have started by reviewing the hypothesis of Re-clusters, and using density functional theory calculations, they have rejected definitively the presence of clusters in the c-phase. They have concluded with two important factors: (i) the preferred partitioning of Re to the matrix (c) phase, where dislocation activity preferentially occurs during the tertiary creep regime and (ii) a retardation effect on dislocation segments at c/c′ interfaces. More recently, evidence of Re and Mo segregation (up to 2.6 and 1 at.%) along with Cr and Co to the dislocations inside of c′ precipitates has been reported in a second generation Ni-based single-crystal superalloy, after creep deformation at 750 °C under an applied stress of 800 MPa by Wu et al. [22]. The mechanism of solid solution hardening especially at high temperatures is very complex and still not fully understood. One possible explanation is a change of materials properties such as the stacking fault energy cSF, the diffusivity, and/or the elastic modulus as reported in the literature [7, 23–26]. In previous studies on the effect of Re on tensile properties in c-phases of Ni-based superalloys, the group of Pettinari-Sturmel has concentrated its effort on the evaluation of the stacking fault energy [27]. It is shown that the influence of Re and W on the stacking fault energy in c-phases of Ni-based superalloy was similar in the temperature range of 25–1050 °C and does slightly evolve from 750 °C. The influence of alloying elements on diffusivity, stacking fault energy has been also discussed as possible reasons for solid solution hardening, but they were not able to explain all the experimental effects of Re [14, 15]. At least, the presence of short-range order (SRO) has been evidenced by Pettinari et al. in c-phases containing Re

Table 1 Nominal alloy compositions of single-phase (matrix) and two-phase (matrix/c′) alloys

Alloy

or W and has been described as nano-obstacles acting against the propagation of dislocation [12, 28]. A strong correlation is established between the distribution of the dislocations after tensile tests and SRO: The presence of dislocation pile-ups with a pairing of the leading pair is a clear signature of SRO. This SRO has been shown to control the deformation micromechanisms during tensile tests up to 750 °C and has been attributed to the most effective strengthening effect in these alloys. Actually, no fundamental grounded model is able to fully explain the effect of Re on creep deformation. This paper focuses on the SRO due to Re and its possible effect on the creep deformation micromechanisms at high temperature in specially produced single-phase, single-crystal alloys with the composition of the matrix of two different superalloys with varying Re content. This paper is focused on TEM experiments performed after interrupted creep test only at 980 °C and 50 MPa.

Experimental To quantify the contribution of the refractory elements like rhenium, tungsten, and molybdenum to the solid solution strengthening, different alloys with the calculated matrix composition of the conventional and model nickel-based superalloys which served as reference alloys were casted [14, 15, 29]. For the first time, single-phase single-crystalline material with systematically varied content of heavy elements was tested in creep [14, 15]. The results are very clean data for the solid solution strengthening potential, which are not influenced by any grain boundary or particle strengthening effects. This allows for quantification of the effectiveness of the refractory elements on solid solution hardening of the matrix phase. Thereby, single crystals containing 0, 3, 6, and 9 wt% Re have been casted. The present study is focused on the alloys with 0 and 9 wt%. The investigated alloy compositions are listed in Tables 1 and 2. The produced single crystals have been solution-treated in a vacuum cold wall furnace to homogenize the sample and remove casting segregation of strong segregating elements like rhenium. Creep samples have been prepared using wire erosion.

Concentration in wt% Al

Co

Mo

Re

Ta

Ti

W

Hf

Ni

7.9

0.5

–

6.0

1.0

8.0

0.1

66.3

9.6

6.6

0.6

3.0

6.5

1.0

6.4

0.1

60.6

19.8

18.4

1.4

–

0.2

0.1

9.1

–

49.6

16.7

1.3

9.0

0.2

0.1

8.3

–

45.0

CMSX-3

5.6

4.6

CMSX-4

5.6

MSX Re0 (matrix CMSX-3)

1.4

MSX Re9 (matrix CMSX-4)

1.4

18.0

Cr

Evidence of Short-Range Order (SRO) by Dislocation Analysis … Table 2 Nominal alloy compositions of single-phase (matrix) alloys

Times

255

Concentration in at.% Al

Co

Cr

Mo

Re

Ta

Ti

W

Hf

Ni

MSX Re0 (matrix CMSX-3)

3.1

20.3

21.4

0.9

–

0.1

0.1

3.0

–

51.1

MSX Re9 (matrix CMSX-4)

3.3

19.6

20.0

0.9

3.1

0.1

0.1

2.9

–

50.0

The creep tests have been performed at 980 °C with a stress of 50 MPa. The specimen′s macroscopic loading axis has been oriented along the [001] crystallographic orientation. The tests have been carried out in vacuum or protective gas to avoid the building of an oxide layer, which distort the creep measurements. The temperature has been measured and controlled by a thermocouple, which is spot welded in-between two opposite measuring spikes to the creep sample. Strain measurement has been done visually and contactless using a CCD-camera by light emission of the sample or against backlight. Due to the resistive sample heating, the samples are cooled very quickly after the creep test is interrupted. This fast cooling prevents dislocation movement, which is a big advantage for subsequent TEM investigation. The second benefit of this heating method is that creep tests can be interrupted at different strains. The creep tests have been interrupted at 1% strain in order to freeze in the dislocation structure for TEM observations. The experimental results are reported in details by Fleischmann et al. in [15]. The creep rates at 980 °C and 50 MPa are 7.5  10−4 s−1 and 10−6 s−1 in the alloys with 0 and 9 wt%, respectively. The creep resistance increases strongly with increasing Re content. The TEM foils were cut parallel to the [001] crystallographic direction, polished mechanically and then prepared by twinjet polisher using a solution made by 5% perchloric acid, 35% glycerol in methanol. The polishing conditions were −20 °C and 30 V. Conventional TEM observations were carried out using a JEM2010 operating at 200 kV. The characterization of the dislocations has been undertaken using TEM under two beam conditions (and sometimes using weak beam conditions). The Burgers vector b of the dislocations has been determined using the familiar criterions for dislocation invisibility, g
b ¼ 0 for perfect dislocations.

Results and Discussion A typical example of TEM observations performed in the sample with 9 wt% Re is illustrated in Fig. 1. Different extensive dislocation pile-ups are visible. The characterization of the dislocations reveals that the glide plane is the
 same for all the observed dislocations; it is the 111 plane.

The dislocations within the pile-ups noted P1, P3, and P4 are visible with the diffraction vector g = 020 and invisible with g ¼ 111. This leads to the determination of their Burgers vector b ¼ a2 ½011, whereas the dislocations within the pile-up noted P2, which are not visible with g = 020 and   visible with g ¼ 111 have a Burgers vector b ¼ a2 101 . The activation of so many dislocations within parallel glide planes is an evidence of a strong localized deformation. These dislocations are classical a/2 dislocations, gliding on {111} planes, as expected in faced cubic-centered phase, as it is the case for the c-phase. The presence of such long pile-ups is the signature of short-range order (SRO) as already mentioned in the literature in c-phases containing either only W (4 at.%) or a combination of Re and W (4 at.% Re and 2 at.% W) [12]. The origin of planar slip resulting in heterogeneous deformation has been discussed in the literature. Both the stacking fault energy and SRO may promote heterogeneous deformation. The stacking fault energy has been found to be not the physical parameter controlling the formation of dislocation pile-ups [30, 31]: SRO constraints the dislocations to planar glide as SRO is at the origin of a local friction stress higher than the applied stress, so that the dislocations move as pile-up to overcome this high friction stress The first dislocations destroy the local order, and the following ones benefits from the softening by propagating in the same glide plane. The slip of successive dislocations leads to reduced SRO resulting in a diffuse antiphase boundary, the main part of the energy being provided by the first dislocation slip step [30]. As a consequence, the two first dislocations are strongly linked and appear to be paired. Some examples of dislocation pile-ups observed in the present samples are given in Fig. 2. These TEM micrographs correspond to four different pile-ups with different lengths (two pile-ups noted P1 and P2 are imaged using two diffraction conditions). A pairing of the leading dislocations is always observed. The distance between these successive dislocations have been determined, using geometrical corrections, taking into account different angles (angle between the dislocation line and the trace of the glide plane on the observation plane, the character of the dislocation, the inclination of the glide plane with respect to the foil plane), so that the real distances between the dislocations in their glide plane have been evaluated with a precision of about 10%.

256

F. Pettinari-Sturmel et al.

Fig. 1 Observations of dislocation pile-ups in samples interrupted after 1% creep strain with 9 wt% Re. The diffraction vector is g = 020, the normal to the foil is closed to [002] direction, the glide plane
of allthe observed dislocations is 111

Fig. 2 Observations of dislocation pile-ups in samples interrupted after 1% creep strain with 9 wt% Re. The diffraction vector is g ¼ 020 for Fig. 2a, b, c. The normal to the foil is closed to [002] direction, and the glide plane of all the
observed  dislocations is 111 . The same dislocation pile-up is imaged in Fig. 2c, d using two different diffraction vectors

For all the investigated pile-ups, the distance between the two leading dislocations ranges between 20 and 25 nm, whereas the other distances range from 40 to 80 nm approximately. In order to obtain a quantification of the effect of SRO in the present samples, the methodology developed by Pettinari et al. and Saada et al. [28, 32, 33] has been used. It is worth mentioning that the distribution of the first dislocations of the pile-up is slightly affected by the presence of the free surfaces, so that this method can be applied although it is based on

measurements performed in a thin foil. The method is based on the determination of the elastic interaction forces experienced by the dislocations within the pile-up and due to the other dislocations, using the experimental position of the dislocations. Considering a dislocation pile-up submitted to an homogeneous applied force (−bsa) in a disordered solid solution with a friction force bsf (resulting from “classical” solid solution effect due to size effect, modulus effect of the solute contents), each dislocation p experiences a total force Fp as follows:

Evidence of Short-Range Order (SRO) by Dislocation Analysis …

257

Fig. 3 Schematic representation of a dislocation pile-up within a short-range-ordered solid solution


 Fp ¼  b ta  tf þ Rp

ð1Þ

where Rp is the total elastic force exerted on the dislocation P p by the other dislocations: Rp ¼ i6¼p Fip , and Fip is the elastic force exerted by the dislocation i on the dislocation p. In the case of a short-range-ordered solid solution, the force cp resulting from SRO acting on the dislocation p has to be taken into account (as illustrated in Fig. 3). At equilibrium, the total force Fp equals to zero so that force cp is given by:
 gp ¼ b t a  t f  Rp ð2Þ Saada et al. have shown that the term (sa − sf) is close to zero and that the SRO effect is essentially governed by the value for the first dislocation, and the value of c0 is then chosen to quantify the SRO in our samples using the following equation: g0 ¼ R0

ð3Þ

It can be easily understood that the controlling parameter is the distance between the two leading dislocations, as the elastic interaction force is inversely proportional to the distance between two dislocations: the lowest this distance is, the highest the elastic interaction stress is, and the highest is the stress due to SRO. With a distance between the leading dislocations between 20 and 25 nm, the obtained c0 values range from 54 to 61 mJ/m2 with an error estimated to be around 10%. For example, the determination of the total elastic force exerted on the first dislocation leads to: – For the pile-up P1 observed in Fig. 1, c0  55 mJ/m2 – For the pile-up P3 observed in Fig. 1, c0  61 mJ/m2 – For the pile-up observed in Fig. 2a, c0  56 mJ/m2. These values have to be compared with the energies obtained in other c-phases containing 4 at.% Re and 2 at.% taken from [12, 28, 32]. In these alloys, after tensile tests at 25 °C, the friction force c0 due to SRO is close to 30 mJ/m2 and after tensile tests at 750 °C, the force c0 is around 10 mJ/m2. In our case, a value higher than 50 mJ/m2 is a clear evidence of a strengthening effect due to SRO, which may be promoted by the higher content in Re. It is also worth mentioning that there are some rare areas where the dislocation pile-ups are different from those observed

Fig. 4 Observations of dislocation pile-ups in samples interrupted after 1% creep strain with 9 wt% Re using the diffraction vector g = 220

previously as illustrated in Fig. 4. They are not so well organized, and some individual dislocations may be observed. This can be due to a non-homogeneous distribution of SRO within the sample. Nevertheless, the prevalent deformation mechanism is the propagation of long dislocation pile-ups. TEM observations after creep in the sample without Re are illustrated in Fig. 5. Differences can be pointed out: pile-ups are less frequent, individual dislocations are observed, and thereby, the deformation is more homogenous. The dislocations seem to be distributed randomly. In some areas, short pile-ups can be identified and paired leading dislocations are observed as illustrated in Fig. 6. In this sample, the number of dislocations within pile-ups is less than in the sample with 9 wt% Re. In the case of the pile-up illustrated in Fig. 6, the total elastic interaction has been evaluated and leads to c0  45 mJ/m2. The presence of both individual dislocations and some short dislocation pile-ups indicates that SRO could exist even in the absence of Re (as identified in the previous study of Pettinari et al. [28]). If the value of c0 is used to quantify the SRO “degree” and thus its strengthening effect, it can be supposed that, in these areas where individual dislocations are observed, the value of c0 is low. The important point is that long dislocation pile-ups is not the prevalent

258

F. Pettinari-Sturmel et al.

Fig. 5 General TEM overview of crept sample with 0 wt% Re. Some dislocation pile-ups and individual dislocations are observed simultaneously. The diffraction vector is g = 020

around 900 °C in other similar c-fcc phases [15]. And has been associated with a value for c0 being lower than 10 mJ/m2. In the present investigation, the value of c0 is five times higher. The presence of SRO may be also due to Re and could be attributed to the “Re effect.”

Conclusions

Fig. 6 TEM observation of a dislocation pile-up in the sample interrupted after 1% creep strain with 0 wt% Re. Two leading paired dislocations are observed

deformation mechanism. Thus, SRO seems to be restricted to some areas and/or is less efficient in the sample without Re. It can be concluded that the SRO degree is much lower in the sample without Re. These experimental results with and without Re indicate the existence of a higher SRO with a high content of Re, even at elevated temperature of 980 °C. This refractory element Re promotes local chemical order, as suggested by atomistic calculations [21]. This results in a SRO present at high temperature, which contributes to the hardening of the matrix phase because it creates high friction stress acting against the dislocation propagation. Another important result is the observation at this temperature, of dislocation pile-ups. It is surprising to observe such a localized deformation resulting from creep tests at 980 °C. It is well documented that SRO vanishes when the temperature increases. SRO has been observed to vanish

TEM observations on deformed specimens indicate the presence of long dislocation pile-ups after creep at 980 °C/50 MPa in c-phases of Ni-based superalloys containing 9 wt% Re. In the sample without Re, the deformation is more homogeneous and the observed pile-ups are much shorter, indicating that the deformation is less localized. The more localized deformation and pairing of leading dislocations are a direct evidence of short-range order (SRO). This local order appears then to be the creep controlling deformation mechanism in the alloy with Re, as these dislocation pile-ups are the observed prevalent micromechanisms. The friction force due to this SRO has been evaluated as it could explain the strength of the investigated alloy. It ranges between 50 and 60 mJ/m2, which is a high value considering the usual energies associated with SRO in other c-phases and especially at this temperature. This SRO seems to be promoted by Re, even at a fairly high temperature of 980 °C. This local atomic arrangement in the presence of Re may explain the exceptional creep behavior at high temperatures of Ni-based superalloys containing Re. Acknowledgments F. PS gratefully acknowledges N. Clément, A. Coujou, and P. Caron who have been pioneers on this topic (relationship between SRO and dislocations in Re-containing c-phases) and gave her the opportunity to develop the quantitative approach of SRO using dislocation pile-ups. F. PS and J. D want to dedicate this paper to Professor Armand Coujou and Professor Georges Saada, who sadly passed away on July, 1, 2016, and on August, 30, 2019, respectively.

Evidence of Short-Range Order (SRO) by Dislocation Analysis …

259

References 17. 1. Reed RC The Superalloys: Fundamentals and Applications. Cambridge: Cambridge University Press. doi:110.1017/ CBO9780511541285. 2. Pollock TM, Argon AS (1992) Creep resistance of CMSX-3 nickel base superalloy single crystals. Acta Metall Mater 40 (1): 1–30. https://doi.org/10.1016/0956-7151(92)90195-k. 3. Pelloux RM, N.J. G (1960) Solid solution and second phase strengthening of nickel alloys at high and low temperatures Title. Trans Metall Soc AIME 218 232–237. 4. Monma K, Suto H, Oikawa H (1964) High-Temperature Creep of Nickel-Chromium Alloys (On the Relation Between High-Temperature Creep and Diffusion in Nickel Base Solid Solutions, IV). J Japan Inst Met 28 (5): 253–258. https://doi.org/ 10.2320/jinstmet1952.28.5_253. 5. Monma K, Suto H, Oikawa H (1964) High-Temperature Creep of Nickel-Tungsten Alloys (On the Relation between High-Temperature Creep and Diffusion in Nickel-Base Solid Solutions. VI). J Japan Inst Met 28 (6): 304–307. https://doi.org/ 10.2320/jinstmet1952.28.6_304. 6. Davies CKL, Davies PW, Wilshire B (1965) The effect of variations in stacking-fault energy on the creep of nickel-cobalt alloys. Philos Mag 12 (118): 827–839. https://doi.org/10.1080/ 14786436508218920. 7. Johnson WR, Barrett CR, Nix WD (1972) The high-temperature creep behavior of nickel-rich Ni-W solid solutions. Metall Mater Trans B 3 (4): 963–969. https://doi.org/10.1007/bf02647673. 8. Matsuo T, Ueki M, Takeyama M, Tanaka R (1987) Strengthening of nickel -base superalloys for nuclear heat exchanger application. J Mater Sci 22 (6): 1901–1907. https://doi.org/10.1007/bf01132914. 9. Nathal M V., Diaz JO, Miner R V. (1988) High Temperature Creep Behavior of Single Crystal Gamma prime and Gamm Laloys. MRS Proc 133 269. https://doi.org/10.1557/proc-133-269. 10. Shinoda T, Masuda-Jindo K, Suzuki T (1990) Extra Contribution of Tm Solutes to the Solute-Dislocation/Surface Interactions in Nickel. MRS Proc 186 (I): 283. https://doi.org/10.1557/proc-186-283. 11. Clement N, Coujou A, Jouiad M, Caron P, Kirchner H.O.K., Khan T (1996) Local Order and Mechanical Properties of the gamma Matrix of Nickel-Base Superalloys. Superalloys 1996 (Eighth Int Symp 239–248. https://doi.org/10.7449/1996/ superalloys_1996_239_248. 12. Pettinari F, Prem M, Krexner G, Coujou A, Kirchner H.O.K, Clement N (2001) Local order in industrial and model c phases of superalloys. Acta Mater 49 (13): 2549–2556. https://doi.org/10. 1016/s1359-6454(01)00103-3. 13. Gan B, Tin S (2010) Assessment of the effectiveness of transition metal solutes in hardening of Ni solid solutions. Mater Sci Eng A 527 (26): 6809–6815. https://doi.org/10.1016/j.msea.2010.06.071. 14. Fleischmann E, Konrad C, Preußner J et al. (2015) Influence of Solid Solution Hardening on Creep Properties of Single-Crystal Nickel-Based Superalloys. Metall Mater Trans A 46 (3): 1125– 1130. https://doi.org/10.1007/s11661-014-2727-x. 15. Fleischmann E, Miller MK, Affeldt E, Glatzel U (2015) Quantitative experimental determination of the solid solution hardening potential of rhenium, tungsten and molybdenum in single-crystal nickel-based superalloys. Acta Mater 87 350–356. https://doi.org/ 10.1016/j.actamat.2014.12.011. 16. Blavette D, Caron P, Khan T (1986) An atom probe investigation of the role of rhenium additions in improving creep resistance of

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29. 30.

31.

32.

33.

Ni-base superalloys. Scr Metall 20 (10): 1395–1400. https://doi. org/10.1016/0036-9748(86)90103-1. Wanderka N, Glatzel U (1995) Chemical composition measurements of a nickel-base superalloy by atom probe field ion microscopy. Mater Sci Eng A 203 (1–2): 69–74. https://doi.org/ 10.1016/0921-5093(95)09825-9. Mottura A, Wu RT, Finnis MW, Reed RC (2008) A critique of rhenium clustering in Ni-Re alloys using extended X-ray absorption spectroscopy. Acta Mater 56 (11): 2669–2675. https://doi.org/ 10.1016/j.actamat.2008.01.032. Mottura A, Warnken N, Miller MK, Finnis MW, Reed RC (2010) Atom probe tomography analysis of the distribution of rhenium in nickel alloys. Acta Mater 58 (3): 931–942. https://doi.org/10.1016/ j.actamat.2009.10.008. Ge BH, Luo YS, Li JR, Zhu J (2010) Distribution of rhenium in a single crystal nickel-based superalloy. Scr Mater 63 (10): 969–972. https://doi.org/10.1016/j.scriptamat.2010.07.019. Mottura A, Reed RC (2014) What is the role of rhenium in single crystal superalloys? MATEC Web Conf 14 01001. https://doi.org/ 10.1051/matecconf/20141401001. Wu X, Makineni SK, Kontis P, Dehm G, Raabe D, Gault B (2018) On the segregation of Re at dislocations in the c′ phase of Ni-based single crystal superalloys. 4 109–114. https://doi.org/10.1016/j. mtla.2018.09.018. Barrett CR, Sherby OD (1965) Influence of stacking-fault energy on high-temperature creep of pure metals. Trans Metall Soc AIME 233 116–1119. Bonesteel R., Sherby O. (1966) Influence of diffusivity, elastic modulus, and stacking fault energy on the high temperature creep behavior of alpha brasses. Acta Metall 14 (3): 385–391. https://doi. org/10.1016/0001-6160(66)90096-4. Karunaratne MSA, Carter P, Reed RC (2000) Interdiffusion in the face-centred cubic phase of the Ni–Re, Ni–Ta and Ni–W systems between 900 and 1300 °C. Mater Sci Eng A 281 (1–2): 229–233. doi:http://dx.doi.org/10.1016/S0921-5093(99)00705-4. Sherby OD, Burke PM (1968) Mechanical behavior of crystalline solids at elevated temperature. Prog Mater Sci 13 323–390. https:// doi.org/10.1016/0079-6425(68)90024-8. Pettinari F, Douin J, Saada G, Caron P, Coujou A, Clement N (2002) Stacking fault energy in short-range ordered c-phases of Ni-based superalloys. Mater Sci Eng A 325 511–519. doi:doi: https://doi.org/10.1016/s0921-5093(01)01765-8. Pettinari-Sturmel F, Douin J, Coujou A, Clement N (2006) Characterisation of short-range order using dislocations. Int J Mater Res 97 (3): 200–204. https://doi.org/10.3139/146.101243. Fleischmann E (2013) PhD Thesis. University Bayreuth. Schwander P, Schönfeld B, Kostorz G (1992) Configurational Energy Change Caused by Slip in Short-Range OrderedNi–Mo. Phys status solidi 172 (1): 73–85. https://doi.org/10.1002/pssb. 2221720109. Plessing J, Achmus C, Neuhäuser H et al. (1997) Short-Range Order and the Mode of Slip in Concentrated Cu-based Alloys. Zeitschrift für Met Int J Mater Res Adv Tech 88 (8): 630–635. Pettinari-Sturmel F, Saada G, Douin J, Coujou A, Clement N (2004) Quantitative analysis of dislocation pile-ups in thin foils compared to bulk. Mater Sci Eng A 387–389 (1-2 SPEC. ISS.): 109–114. https://doi.org/10.1016/j.msea.2004.05.048. Saada G, Douin J, Pettinari-Sturmel F, Coujou A, Clement N (2004) Pile-ups in thin foils: application to transmission electron microscopy analysis of short-range-order. Philos Mag 84 (8): 807– 824. https://doi.org/10.1080/14786430310001646718.

Rationalisation of the Micromechanisms Behind the High-Temperature Strength Limit in Single-Crystal Nickel-Based Superalloys Daniel Barba, Ashton J. Egan,Yilun Gong, Michael J. Mills, and Roger C. Reed

Abstract

The peculiar atomic structure of γ  precipitates [Ni3 (Al/Ti)-L12 ] in Ni-based superalloys produces high-energy faults when dislocations glide them, giving their significant strength at high temperatures. The mechanisms behind the strength failure of these alloys above 700–800 ◦ C are still controversial. Recent advances in atomic resolution microscopy have allowed to study these mechanisms with unprecedented detail. In our study, we have characterised in a careful systematic study a SX-[001] superalloy from RT to 1000 ◦ C. Multiscale microscopy (TEM and SEM) has been combined with physical metallurgy and atomistic modelling to fully understand the correlation between the strength drop and the observed changes in the γ  shearing mechanism. Our results show that, far from previous beliefs, the initial failing of alloy strength is not a consequence of the activation of dislocation climbing. Instead, there is a transition between three different mechanisms: (T < 750 ◦ C) continuous planar stacking faults below, (T = 750 ◦ C) APB shearing at the strength peak anomaly and (T > 800 ◦ C) extensive twin deformation after D. Barba (B) Universidad Politecnica de Madrid, Madrid, Spain e-mail: [email protected] A. J. Egan · M. J. Mills Department of Materials Science and Engineering/Center for Electron Microscopy, The Ohio State University, Columbus, OH, USA e-mail: [email protected] M. J. Mills e-mail: [email protected] Y. Gong Department of Materials, University of Oxford, Oxford, UK e-mail: [email protected] R. C. Reed Department of Engineering Science, University of Oxford, Oxford, UK e-mail: [email protected]

the yield drop. Local chemical changes around the γ  shearing dislocations boost these changes, thus producing the sudden drop of strength. Keywords

TEM • Deformation mechanisms • Mechanical testing • DFT

Introduction Resistance to high-temperature plasticity is a significant advantage of the nickel-based superalloys, particularly when in single-crystal form. This important characteristic comes from a careful arrangement of shear resistant γ  precipitates. The high-energy faults formed when dislocations glide through these precipitates are known to provide this unique strength at high temperatures. This important mechanism becomes ineffective when the shearing dislocations have enough mobility to overcome the γ  precipitates at high temperature [6,32]. Historically, this was the consensus reason for the sudden drop of strength observed above 700 ◦ C in these alloys. But is this the case? In the recent years, advances in atomic resolution chemical microscopy (atom probe and TEM-EDX) have allowed the study of the population of mechanisms in this temperature regime with unprecedented detail [2,5,7,11,26,29] complementing previous studies on plasticity in Ni3 Al system [9,30,31]. These studies have shown the presence of other complex diffusion mechanisms different from the traditional dislocation climbing [2,7]: traditional dislocation shearing is aided by segregation processes leading to changes in the plastic strength above 700 ◦ C. However, there is a key-point missing: these studies are constrained to limited deformation regimes, thus not providing a complete picture of the mechanism. What is clearly needed to extract a convincing theory is a systematic study between the temperature, appearance of segregation

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_25

260

261

Rationalisation of the Micromechanisms Behind …

micromechanisms and loss of strength. This is a critical step to design improved grades of superalloys. In this work, we present a high-resolution characterisation study of the evolution of the micromechanisms and segregation processes as a function of the temperature, and we will associate the transition between the different deformation mechanisms with the loss of strength in the alloy. The study is structured as follows: first, a commercial SX-superalloy is mechanically tested at seven different temperature conditions (from RT to 900 ◦ C). Second, critical samples before, at, and after the yield drop are selected for advanced TEM-EDX analysis. Third, the microsegregation levels and their chemical species associated with the dislocation shearing are quantified for each temperature. Finally, this observed behaviour is rationalised and modelled using density functional theory (DFT), and the observed yielding behaviour is extracted from the different micromechanism transitions.

Methods Material and Experimental Methods The single-crystal superalloy MD2 of composition Ni11.2Al-9.3Co-5.3Cr-2.6W-2Ta-1.65Ti-1.33Mo-0.2Si0.03Hf (at.%) is used in this study. The material was solution treated at 1275 ◦ C for 8 h, followed by ageing for 6 h at 1080 ◦ C, and finally at 870 ◦ C for 16 h. The orientation of the bulk crystal was checked using backscattered electron diffraction (EBSD) analysis prior to extracting creep samples. A deviation from the ideal nominal orientation is 001 with less than 5◦ error. Compression experiments under a constant displacement rate of the cross-head producing an initial strain rate of ε˙ = 10−5 s−1 were performed at different temperatures covering the strength drop regime of the alloy (testing temperatures = 20, 400, 650, 750, 800, 850, and 900 ◦ C). Cylinders of 5 mm diameter and 5 mm height were employed for the compression tests in an Instron servo-electric machine. In addition to the cross-head displacement, digital image correlation (DIC) was used to track the sample deformation for all tests. The tests were stopped for all specimens when 5% of total DIC strain was reached. Once the test was stopped, the furnace was open and displaced on rails away from the loading frame leaving the sample cooling down in the air. Loading on the sample was kept until a temperature of 200 ◦ C was reached. Post-mortem examination prior to scanning transmission electron microscopy (STEM) analysis was carried out in order to identify the deformation mechanisms after plastic deformation. One of the axial facets of the samples were grinder and polished finished with colloidal silica in order to identify the 001 crystal poles of the sample perpendicular to the compression orientation. Once identified, the cylinders

were cut along one of the 001 planes perpendicular to the compression direction in order to extract STEM foils normal to 011 orientation using an FEI Helios NanoLab DualBeam 600 focused ion beam (FIB). This assures that planar faults are viewed edge-on using high-angle annular dark field (HAADF) zone axis imaging. Samples were thinned at 5 kV and then further cleaned using a Fischione NanoMill. Energydispersive X-ray analysis (EDX) of the foils was performed on an image-corrected Titan3TM 60–300 kV with a Super-X detector utilising the Bruker ESPRIT software. Integrated line scans were conducted and quantified through Cliff–Lorimer analysis [10] using experimental Kα energies for Ni, Co, Al, Cr, and Ti. Lα was used for the case of Mo. The Cu specimen holder signal was avoided by using the Mα lines for Ta and W since the Lα Ta and W peaks corresponded too closely to a Cu peak to be accurately considered. Deconvolution for the W and Ta Mα peaks, as well as background subtraction, was used to reduce the influence of bremsstrahlung. For EDS line scan analysis, to reduce the noise and smooth the data, small amounts of averaging were used. Using this setup, the noise can be reduced and no large effects on the variation of composition along the line scans were observed. Detrimental artefacts from foil thickness to the EDS spectrum like beam spreading or low EDS counts were mitigated by controlling the specimens thickness during the thinning process on the FIB. The FIB foil was thinned to around an estimated 20– 50 nm where both foil thickness effects are observed to be avoided. Higher atomic resolution STEM analysis was performed using a probe-corrected Titan3TM 60–300 kV.

DFT Calculations Density functional theory (DFT) calculations were performed to rationalise the experimental results through the calculation of antiphase boundary (APB) fault energies γ . The change in fault energy produced by segregation to the fault is expressed as: segregated

γ =

(E fault

segregated

− E perfect

Ni3 Al Ni3 Al ) − (E fault − E perfect )

A

(1) segregated segregated Ni3 Al Ni3 Al where E fault , E perfect , E fault , and E perfect are the energies of a segregated APB cell, a perfect segregated cell, a Ni– Al APB cell, and a Ni–Al perfect cell, respectively. These energy terms are extracted from Ni3 Al supercell calculations consisting on 12-{111} layers with 16 atoms on each layer, with a perfect stacking sequence or introducing and APB defect in the cell. To quantify the segregation effect on the stacking fault energy, Co, Cr, and Mo atoms were introduced in the Ni1 sites at the APB plane [21]. All calculations were performed using the projector augmented-wave (PAW) method [3] as implemented in the Vienna ab initio simulation

262

package (VASP) [13,14] with the provided PAW potentials [15]. The generalised gradient approximation (GGA) with Perdew–Burke–Ernzerhof (PBE) [20] parametrisation was utilised for the exchange-correlation functional. A 3×3×1 Monkhorst–Pack k-point mesh [19] was used to sample the Brillouin zone in all calculations with a cut-off energy of 350 eV. Ions were fully relaxed with a force criterion of 5 × 10−3 eV/Å. Spin polarisation was included in all calculations with initial magnetic moments set as 1.5 μ B for Co, 0.6 μ B for Cr, 0.8 μ B for Ni, and zero for Mo & Al.

Results In this section, the mechanical and experimental results extracted from MD2 alloy will be presented. First, the mechanical response of the alloy is analysed, and the selection of temperature conditions for TEM samples is motivated. Then, the mechanisms operating at each of the mechanical regimes are elucidated using STEM atomic resolution microscopy. Although a large number of STEM images of each temperature conditions have been studied, a reduced set of representative ones is presented in this work. Finally, the local chemical changes associated with each of the faults are studied using TEM-EDX technique.

Mechanical Results The mechanical behaviour for the different testing conditions is presented in Fig. 1a. Due to the high nonlinearity of the elastic regime in the curves, the yield stress Fig. 1b is calculated by translating by 0.2% deformation a straight line with an

D. Barba et al.

approximate slope equal to the linear elastic regime of each curve and identifying the intersection with the stress–strain curve. It can be observed a reduction of the hardening levels as the temperature increases and a sudden drop of properties after 750 ◦ C. The strength of the alloy present a flat region up to 650 ◦ C with a slight increase at 750 ◦ C and finally a sudden drop for T > 750◦ . This behaviour has been previously observed in other SX-superalloys [22,23]. In order to rationalise micro-mechanically the anomalous yielding behaviour between 650–750 ◦ C and the subsequent yield drop specific samples between 650 and 850 ◦ C were selected for atomic resolution TEM-EDX microstructural and chemical analysis of their deformation structures. The selected conditions are: (1) 650 ◦ C before the yield drop, (2) 750 ◦ C just as the yield drop, and (3) 800 ◦ C and (4) 850 ◦ C after the yield drop.

Deformation Mechanisms Analysis The deformation structures of the selected samples are presented next. STEM images of the 650 ◦ C sample are presented in Fig. 2 corresponding to the temperature region below yield anomaly. The image shows a high density of continuous faults preferentially along one slip direction, with a reduced number of them along a second slip system (Fig. 2b). These deformation structures are in agreement with other studies performed on the same alloy under creep conditions [1,2]. These faults extend both in γ  and γ phases, with dislocation pile-ups formed in the latter phase. More detailed analysis presented later confirms a population of SESFs between these faults, with a minority evolving to a microtwin stage. An example of STEM images of the 750 ◦ C sample is presented in Fig. 3, corresponding to the peak of anomalous yield.

Fig. 1 a Mechanical stress–strain curve of MD2 superalloy for the different testing conditions experimented in this study. b Yield stress of MD2 alloy under compression at ε˙ = 10−5 s−1 as a function of temperature and detail of the testing conditions at which TEM samples are extracted

Rationalisation of the Micromechanisms Behind …

263

Fig. 2 STEM micrograph of the deformation structures observed at 650 ◦ C, overview (a) and detail (b). A high density of continuous planar faults is observed along the whole sample

Fig. 3 STEM micrograph of the deformation structures observed at 750 ◦ C, overview (a) and detail (b). Dissociated dislocation faults (later confirmed as APBs) populate the whole TEM sample

The deformation structures are completely different from the ones observed at 650 ◦ C. Most of the faults are now formed by dissociated dislocations, presenting short extensions. Only a small number of extensive continuous faults have been spotted in the foil. This indicates a change of deformation mechanism from single leading dislocations on each plane to a sys-

tem of leading and trailing dislocations with a fault between them (partials for the case of extrinsic and intrinsic faults and full dislocations for antiphase boundaries-APBs). This will be explained with further detail in the discussion section. Atomic-level resolution analysis on some of these faults has confirmed their APB structure with enhance contrast due to

264

D. Barba et al.

Fig. 4 STEM micrograph of the deformation structures observed at 800 ◦ C. High density of extensive microtwinning along γ and γ  phases is observed

a) 850°C - Overview

b) 850°C - Detail

Microtwins

500 nm

Microtwins

250 nm

Fig. 5 STEM micrograph of the deformation structures observed at 850 ◦ C, overview (a) and detail (b). As in for 800 ◦ C, a high density of microtwins is observed extending along both γ and γ  phases

elemental segregation (more details in the chemical analysis section). Finally, STEM images of the deformation structures corresponding to the yield drop region at 800 ◦ C and 850 ◦ C are presented in Figs. 4 and 5. The predominant deformation mechanism here is microtwinning. Extensive continuous bands run all along the TEM studied regions following

the preferential slip system. These bands have been identified as microtwins using atomic resolution TEM as microtwins (see the following section). The thickness of these microtwin bands varies from tens of atomic planes to nanometres, usually organised in bundles. This clearly indicates a change of deformation mechanism from APBs to extensive microtwins along the whole sample, presumably producing the sudden

Rationalisation of the Micromechanisms Behind …

drop of strength observed mechanically in Fig. 1 as indicated in the discussion section. As a summary, there is a transition between (1) a predominance of continuous faults below the yielding anomalous temperature (T < 750 ◦ C), to APB shearing at the anomalous peak temperature (T = 750 ◦ C), and finally the predominance of extensive microtwinning at the yield drop (T > 750 ◦ C). Chemical analysis of the fault types observed at different temperatures is presented next.

Detailed Analysis of Fault Structures In this section, representative examples of the faults observed for each temperature will be used for TEM-EDX chemical analysis. For the case of 650 ◦ C, a double-fault structure is analysed as presented in Fig. 6. Atomic order analysis of the faults confirms their extrinsic nature (Fig. 6a). The faults can be either complex faults (CESFs) or regular faults (SESFs) as the present analysis cannot distinguish between them. The faults present higher contrast in HAADF analysis indicating local changes in composition. EDX analysis confirms this. EDX maps are presented in Fig. 6b, c showing a strong segregation of Cr to the faults and less pronounced for the case of Co. These maps have been integrated parallel to the fault to show the concentration profiles perpendicularly to the faults as indicated in Fig. 6d, e. The concentration profiles confirm the segregation of Cr (≈ +1 at.%) and Co (≈ +0.5 at.%). These increases are compensated by the depletion of Al (−1 at.%) at the fault. No significant fluctuations were found in the other elements analysed (Ni, Mo, Nb, Ti, Ta, W). The segregation/depletion of these elements is in accordance with previous works on the same alloy [2] and other alloy systems [11,16,25,28], but all of them under creep loading conditions. Furthermore, recent modelling work of Mianroodi et al. [17] rationalises the segregation of Co and Cr to complex faults in γ  in Ni–Al–Co ternary system. The chemical analysis for a representative fault observed at 750 ◦ C is presented in Fig. 7. A fault terminating inside a γ  precipitate is selected as indicated in see Fig. 7a. Atomic analysis of the STEM image in Fig. 7a shows no planar sequence disorder along the fault, vanishing the possibility of being an extrinsic or intrinsic fault. Therefore, this fault is presumably an APB. The enhanced contrast at the fault indicates again changes in the local composition. EDX maps of the faults confirm this, with Cr and Mo segregating slightly to the fault. A slight depletion of Ni is also observed. No significant changes are observed for the other elements. This elemental segregation is less pronounced and differs from the observed in the case of the SESFs at 650 ◦ C (Co is not segregated here, while Mo is segregated and instead of Al, Ni is depleted). On top of that, the segregation levels of Cr are considerably lower (+≈ 0.3 at.% Cr) than for the 650 ◦ C case.

265

Finally, the extensive microtwins observed at 800 ◦ C and 850 ◦ C are studied in detail in Figs. 8 and 9, respectively. The microtwin boundaries appears to be segregated of Cr in both cases (≈ +0.4 at.% Cr) and of Co for the case of the microtwin at 800 ◦ C (≈ +0.2 at.% Co). Depletion of Al at the twin boundaries is also observed (≈ −0.3 at.% Al). The concentration of all elements inside the twin recovers the γ  composition. This is in accordance with other microtwin observation in the literature [2,7,12,27,28]. As a summary, all the different types of faults observed for the four temperatures present a similar segregation phenomenon. Variations in their quantity and type of elements are identified between continuous faults/microtwins and APB. The connection between these segregation events, the different deformation mechanisms, and the observed mechanical performance are discussed in more detail in the following section.

Discussion and Rationalisation of the Observed Deformation Mechanisms In this section, each of the three different deformation mechanisms observed under mechanical compression of SX-MD2 alloy is discussed in detail. Then, the relationship between the different mechanisms, segregation phenomenon, and observed mechanical behaviour is rationalised. The transition between the three different governing mechanisms along with the connection to mechanical strength of the alloy is summarised in Fig. 10. The three different deformation mechanisms present different dislocation structures and physical processes: • T750 ◦ C - Microtwinning: Finally, at higher temperatures, APB dissociated dislocations reduce their presence and a high density of microtwin bands appears. These types of faults are formed by continuous pass of 112 partial dislocation on adjacent {111} planes leaving a population of blocked trailing dislocations at the γ -γ  interface [1,28]. The segregation profiles are again similar to the continuous faults. However, the segregation is just confined at the twin boundaries, leaving the core of the twin unsegregated. This has great implications on the mechanical performance as, once the first twin layer is created, no further segregation elements need to be diffused to the twin from the surrounding material. From the observed deformation mechanism, chemical analysis, and temperature dependencies, it is clear that segregation and other time-dependent processes control the transition between different mechanisms and thus, the observed mechanical behaviour at high temperatures of superalloys. These are discussed next. The transitions between different deformation mechanisms define the observed mechanical dependence on temperature. For the first transition from continuous faults to APB shearing, there is a change of segregated elements present at the fault (from Co to Mo segregation). In order to arise light on this change of segregated elements, DFT analysis on segregated APBs has been performed as shown in Fig. 11. All the atoms of Ni in Ni1 sites at the APB fault plane has been replaced by Co, Cr, and Mo as indicated in Fig. 11a. The calculated fault energies for the different segregated APBs are presented in Fig. 11b. These results show the powerful effect of Mo and Cr in reducing the

stacking fault energy and thus the stress required for the creation of the APB fault, especially Mo. Cr effect is in accordance with previous results from Rao et al. [21]. Thus, the transition to APBs might be triggered by the increased mobilities by the higher temperatures of these slow diffusion elements such as Mo [4,18] observed in the APB and not in the continuous stacking faults. However, once an APB segregated fault is created at the γ -γ  interface, the gliding of this fault is limited to the diffusion of the segregation cloud coupled with it. This increases the stress required during constant displacement rate tests. In the case of the second mechanism transition, the factor triggering the formation of thick microtwins at higher temperatures is the necessity of piling up dislocations on adjacent {111} gliding planes. This phenomenon is statistically unlikely to happen without climbing of the shearing dislocation. This should include necessarily climbing short distances from the interception of the dislocation at the γ -γ  interface. This short climbing process requires a high mobility of the dislocation segments only possible at higher temperatures [6,8] which might explain the appearance of thick microtwins only above 750 ◦ C for this alloy. On the other hand, once the first layer of a microtwin formed, no additional segregation elements are needed as the initial segregated twin boundary just needs to advance to the adjacent {111} plane as the gliding partial dislocation advances [1,24]. This has strong implications on the rate limiting processes governing the dislocation motion, as diffusion distances reduce considerably, to just one atomic spacing. At constant strain rate tests, this produce a drop of the stress required for extending these twins and thus deforming the material. Based on these premises and experimental observations, we can infer that the yield drop observed in this alloy experimentally is produced by the transition from APB faults with high gliding stress to deformation microtwins with low stress required for their extension.

271

Rationalisation of the Micromechanisms Behind …

a)

b)

APB plane unsegregated

Ni1 sites segregated

APB energy 191 mJ/mm2

Co -69 mJ/mm 2

Cr -281 mJ/mm 2

[211]

Mo -433 mJ/mm 2

[011]

Ni

Al

Co,Cr,Mo (Ni1 sites)

Fig. 11 a Atomic arrangements at the {111} APB plane of the unsegregated and segregated DFT supercells used in this study. All the Ni1 sites at the APB fault are substituted by segregate elements; b change is APB energy calculated from DFT when all Ni1 sites at the APB are substituted by the indicated elements

Conclusions

cal response of this alloy. The transition between different faults might be guided by the different levels and types of segregation types along with the necessity of partial dislocation climbing for microtwinning which is enhanced at higher temperatures.

The relationship between deformation mechanisms and strength evolution with temperature of the SX-superalloy MD2 has been studied experimentally. The following conclusions arose from this study: 1. The yield strength response of the SX-MD2 alloy as a function of temperature can be split in three different regions: (1) flat strength region up to ≈ 750 ◦ C, (2) anomalous yield peak at about 750 ◦ C, and finally (3) a sudden strength drop about ≈ 750 ◦ C. 2. Each of these three regions presents different deformation mechanisms elucidated by TEM analysis: (1) high density of 1–2 layer continuous faults extending to both γ and γ , (2) APB shearing within the γ  precipitates formed by coupled dislocations, (3) extensive microtwinning bands at the nm level extending to both γ and γ  . 3. All the faults studied show local segregation at their core (or twin boundaries for the case of the microtwinning). Region (1) and (3) faults (SESFs and microtwinning) are enriched with Cr and Co and depleted with Al. Faults at the yield strength peak (3) are enriched with Cr and Mo, while in this case there is a depletion of Ni at the fault. 4. Atomistic modelling of the segregation phenomenon to APB faults shows that the introduction of Cr and Mo atoms in the Ni1 sites of the fault reduces the energy abruptly, thus producing a positive driving force for the segregation of these elements and more importantly promotes the creation of APBs within the γ  precipitates. However, the gliding of these APBs through the γ  precipitates is restricted to the diffusion velocity of these elements. 5. The dislocation structures for the different observed mechanisms has been postulated. These have been used to rationalise the observed transition between deformation mechanism and the connected changes in the mechani-

Acknowledgements The authors are grateful to J. Moverare and M. Segersäll for the provision of the studied material. The authors also thank Steve Kench and Enrique Alabort for their assistance and advice. Funding from the USAF Air Force is acknowledged under grant FA955018-1-7000. AJG and MJM acknowledge the support of the National Science Foundation and the DMREF program under grant #1922239.

References 1.

2.

3. 4.

5.

6.

7.

Barba, D., Pedrazzini, D., Collins, A., Wilkinson, A., Moody, M., Bagot, P., Jérusalem, A., Reed, R.: On the microtwinning mechanism in a single crystal superalloy. Acta Materialia 127, 37– 40 (2017). https://doi.org/10.1016/j.scriptamat.2016.08.029. http:// www.sciencedirect.com/science/article/pii/S1359646216304055 Barba, D., Smith, T., Miao, J., Mills, M., Reed, R.: Segregationassisted plasticity in ni-based superalloys. Metallurgical and Materials Transactions A 49(9), 4173–4185 (2018) Blöchl, P.E.: Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994). https://doi.org/10.1103/PhysRevB.50.17953 Chen, Y., Francis, E., Robson, J., Preuss, M., Haigh, S.: Compositional variations for small-scale gamma prime (γ  ) precipitates formed at different cooling rates in an advanced ni-based superalloy. Acta Materialia 85, 199 – 206 (2015) Eggeler, Y., Müller, J., Titus, M., Suzuki, A., Pollock, T., Spiecker, E.: Planar defect formation in the γ  phase during high temperature creep in single crystal coni-base superalloys. Acta Materialia 113, 335–349 (2016). http://www.sciencedirect.com/science/article/pii/ S1359645416302531 Epishin, A., Link, T.: Mechanisms of high-temperature creep of nickel-based superalloys under low applied stresses. Philosophical Magazine 84(19), 1979–2000 (2004) Freund, L., Messé, O.M., Barnard, J., Göken, M., Neumeier, S., Rae, C.: Segregation assisted microtwinning during creep of a polycrystalline L12-hardened Co-base superalloy. Acta Materialia 123, 295–

272

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

D. Barba et al. 304 (2017). https://doi.org/10.1016/j.actamat.2016.10.048. http:// linkinghub.elsevier.com/retrieve/pii/S1359645416308151 Haghighat, S.H., Eggeler, G., Raabe, D.: Effect of climb on dislocation mechanisms and creep rates in γ  -strengthened ni base superalloy single crystals: A discrete dislocation dynamics study. Acta Materialia 61(10), 3709 – 3723 (2013) Hirsch, P.B.: A new theory of the anomalous yield stress in l12 alloys. Philosophical Magazine A 65(3), 569–612 (1992). https:// doi.org/10.1080/01418619208201539 Hoeft, H., Schwaab, P.: Investigations towards optimizing EDS analysis by the cliff-lorimer method in scanning transmission electron microscopy. X-Ray Spectrometry 17(5), 201–208 (1988). https:// doi.org/10.1002/xrs.1300170509 Kontis, P., Li, Z., Collins, D.M., Cormier, J., Raabe, D., Gault, B.: The effect of chromium and cobalt segregation at dislocations on nickel-based superalloys. Scripta Materialia 145(Supplement C), 76 – 80 (2018). https://doi.org/ 10.1016/j.scriptamat.2017.10.005. http://www.sciencedirect.com/ science/article/pii/S1359646217305912 Kovarik, L., Unocic, R., Li, J., Sarosi, P., Shen, C., Wang, Y., Mills, M.J.: Microtwinning and other shearing mechanisms at intermediate temperatures in Ni-based superalloys. Progress in Materials Science 54, 839–873 (2009) Kresse, G., Furthmüller, J.: Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 (1996). https://doi.org/10.1103/PhysRevB.54. 11169 Kresse, G., Furthmüller, J.: Efficiency of ab-initio total energy calculations for metals and semiconductors using a planewave basis set. Computational Materials Science 6(1), 15 – 50 (1996). https://doi.org/10.1016/0927-0256(96)00008-0. http:// www.sciencedirect.com/science/article/pii/0927025696000080 Kresse, G., Joubert, D.: From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 59, 1758–1775 (1999). https://doi.org/10.1103/PhysRevB.59.1758 Makineni, S.K., Kumar, A., Lenz, M., Kontis, P., Meiners, T., Zenk, C., Zaefferer, S., Eggeler, G., Neumeier, S., Spiecker, E., Raabe, D., Gault, B.: On the diffusive phase transformation mechanism assisted by extended dislocations during creep of a single crystal conibased superalloy. Acta Materialia 155, 362 – 371 (2018). https:// doi.org/10.1016/j.actamat.2018.05.074. http://www.sciencedirect. com/science/article/pii/S1359645418304713 Mianroodi, J.R., Shanthraj, P., Kontis, P., Cormier, J., Gault, B., Svendsen, B., Raabe, D.: Atomistic phase field chemomechanical modeling of dislocation-solute-precipitate interaction in ni–al–co. Acta Materialia 175, 250 – 261 (2019). https:// doi.org/10.1016/j.actamat.2019.06.008. http://www.sciencedirect. com/science/article/pii/S1359645419303672 Minamino, Y., Yoshida, H., Jung, S.B., Hirao, K., Yamane, T.: Diffusion of platinum and molybdenum in Ni and Ni3 Al. In: Diffusion in Materials DIMAT 1996, Defect and Diffusion Forum, vol. 143, pp. 257–262. Trans Tech Publications Ltd (1997)

19. Monkhorst, H.J., Pack, J.D.: Special points for brillouin-zone integrations. Phys. Rev. B 13, 5188–5192 (1976) 20. Perdew, J.P., Burke, K., Ernzerhof, M.: Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865–3868 (1996). https://doi.org/10.1103/PhysRevLett.77.3865 21. Rao, Y., Smith, T., Mills, M., Ghazisaeidi, M.: Segregation of alloying elements to planar faults in γ  -Ni3 Al. Acta Materialia 148, 173– 184 (2018) 22. Reed, R.: The Superalloys: Fundamentals and Applications. Cambridge (2006) 23. Reed, R., Rae, C.: 22 - Physical metallurgy of the nickel-based superalloys. In: D. Laughlin, K. Hono (eds.) Physical Metallurgy (Fifth Edition), pp. 2215–2290. Elsevier, Oxford (2014) 24. Smith, T.: Orientation and alloying effects on creep strength in nibased superalloys. Ph.D. thesis, The Ohio State University (2016) 25. Smith, T., Esser, B., Antolin, N., Carlsson, A., Williams, R., Wessman, A., Hanlon, T., Fraser, H., Windl, W., McComb, D., Mills, M.: Phase transformation strengthening of hightemperature superalloys. Nature Communications 7, 13434 (2016). https://doi.org/10.1038/ncomms13434 http://www.nature. com/articles/ncomms13434#supplementary-information 26. Smith, T., Esser, B., Antolin, N., Viswanathan, G., Hanlon, T., Wessman, A., Mourer, D., Windl, W., McComb, D., Mills, M.: Segregation and η phase formation along stacking faults during creep at intermediate temperatures in a Ni-based superalloy. Acta Materialia 100, 19–31 (2015). https://doi.org/10.1016/j.actamat.2015.08.053. http://linkinghub.elsevier.com/retrieve/pii/S1359645415006308 27. Smith, T., Good, B., Gabb, T., Esser, B., Egan, A., Evans, L., McComb, D., Mills, M.: Effect of stacking fault segregation and local phase transformations on creep strength in nibase superalloys. Acta Materialia 172, 55 – 65 (2019). https:// doi.org/10.1016/j.actamat.2019.04.038. http://www.sciencedirect. com/science/article/pii/S135964541930240X 28. Smith, T., Rao, Y., Wang, Y., Ghazisaeidi, M., Mills, M.: Diffusion Processes During Creep at Intermediate Temperatures in a Ni-based Superalloy. Acta Mater. 141, 261–272 (2017). https:// doi.org/10.1016/j.actamat.2017.09.027. http://www.sciencedirect. com/science/article/pii/S1359645417307607 29. Smith, T., Unocic, R., Deutchman, H., Mills, M.: Creep deformation mechanism mapping in nickel base disk superalloys. Materials at High Temperatures 33(33), 1–12 (2016). https://doi.org/10.1080/ 09603409.2016.1180858 30. Suzuki, H.: Segregation of solute atoms to stacking faults. Journal of the Physical Society of Japan 17(2), 322–325 (1962). https://doi. org/10.1143/JPSJ.17.322 31. Veyssiere, P., Douin, J., Beauchamp, P.: On the presence of super lattice intrinsic stacking faults in plastically deformed ni3al. Philosophical Magazine A 51(3), 469–483 (1985). https://doi.org/10. 1080/01418618508237567 32. Zhu, Z., Basoalto, H., Warnken, N., Reed, R.: A model for the creep deformation behaviour of nickel-based single crystal superalloys. Acta Materialia 60(12), 4888–4900 (2012)

Local Mechanical Properties at the Dendrite Scale of Ni-Based Superalloys Studied by Advanced High Temperature Indentation Creep and Micropillar Compression Tests Lukas Haußmann, Steffen Neumeier, Markus Kolb, Johannes Ast, Gaurav Mohanty, Johann Michler, and Mathias Göken

Abstract

Chemical inhomogenities due to dendritic solidification of Ni-based superalloys result in different local microstructures with varying mechanical properties. New indentation creep test methods allow probing of the local creep properties at the dendrite scale at high temperatures. The as-cast single crystalline Ni-based superalloy ERBO1A (a derivative alloy of CMSX–4) was investigated and electron-probe microanalysis (EPMA) measurements revealed strong segregation of, e.g., Re and W in the dendritic region and, e.g., Ta in the interdendritic region. Indentation creep experiments at 750 °C and micropillar compression tests at 785 °C were conducted in both regions, and a higher creep strength was found in the dendritic region compared to the interdendritic region. Theoretical models for solid solution hardening as well as c′ precipitation hardening confirm these results, since they predict a higher strength in the dendritic region than in the interdendritic region. Compared with the fully heat treated state, a smaller difference in the local mechanical

L. Haußmann (&)
S. Neumeier
M. Kolb
M. Göken Friedrich-Alexander-Universität Erlangen-Nürnberg, Department of Materials Science & Engineering, Institute I: General Materials Properties, Martensstr.5, 91058 Erlangen, Germany e-mail: [email protected] S. Neumeier e-mail: [email protected] J. Ast
G. Mohanty
J. Michler Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Mechanics of Materials and Nanostructures, Feuerwerkerstraße 39, 3602 Thun, Switzerland J. Ast Fraunhofer-Institut für Keramische Technologien und Systeme IKTS, Korrelative Mikroskopie und Materialdaten, Äußere Nürnberger Strasse 62, 91301 Forchheim, Germany G. Mohanty Tampere University, Materials Science and Environmental Engineering, 33014 Tampere, Finland

properties or even a reverse strength ratio of the dendritic and interdendritic region can be expected. Keywords





Ni-based superalloy Indentation creep compression Dendritic segregations




Micropillar

Introduction Microstructural inhomogeneities and concentration gradients are present in nearly all metallic alloys. In Ni- and Co-based superalloys, the dendritic microstructure, diffusion zones between substrate and coatings, c′ precipitates and other intermetallic phases determine the macroscopic mechanical properties. Nanomechanical testing methods are very well suited to determine local differences of the mechanical properties [1–3]. Especially, indentation creep testing is a well-suited technique for such investigations, since it allows an examination of the local creep properties at elevated temperatures. Chu and Li [4] introduced already in 1977 an indentation creep method with a cylindrical flat punch indenter. Commonly used pyramidal indenter tips degrade during indentation creep experiments, with a significant influence on the results [5, 6]. However, until today such indentation creep experiments using cylindrical indenters were usually carried out with relatively large indenter tips of diameters  1 mm [7–9]. Cylindrical flat punch indenters were also used for special applications like characterization of thin films with very small indenter tips of diameters  1 µm [10, 11] or at low temperatures [4, 12]. Pyramidal indenters have been used quite intensively to study the individual properties of c, c′, and TCP phases [1–3, 13]. However, these pyramidal sharp indenter tips suffer from many limitations when used to study high temperature properties, particularly during long-term indentation creep tests. A crucial factor for nanomechanical high temperature

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_26

273

274

L. Haußmann et al.

creep experiments is the thermal drift rate. Most nanomechanical methods show relatively high thermal drift rates at high temperatures in the range of nm/s which makes them unsuitable for long-term high temperature experiments. A suitable new testing method was recently introduced by Matschkal–Amberger et al. [14] using a thermomechanical analyzer with a focused ion beam (FIB) milled cylindrical flat punch indenter with a diameter of 20 µm for indentation creep experiments which only leads to thermal drift rates in the range of nm/h. In this study, this new indentation creep method was used together with micropillar compression tests to investigate the influence of segregations at the dendritic and interdendritic regions on the local mechanical properties of the single crystalline Ni-based superalloy ERBO1, which is a derivative alloy of CMSX–4.

Experimental Procedures The nominal composition of Ni-based superalloy ERBO1 is given in Table 1. The material was cast by Doncasters Castings GmbH using the Bridgman process as single crystalline plates in [001] orientation. All investigations were performed on as-cast samples. In the following, the as-cast condition is referred to as ERBO1A, whereas the fully heat treated condition is referred to as ERBO1C. Further information on the material can be found in [15]. After standard metallographic sample preparation procedure, the microstructure was characterized using a scanning electron microscope (SEM) (Crossbeam 1540 EsB, Zeiss), and the chemical composition was probed by energy dispersive X-ray spectroscopy (EDS) (Inca Energy 350, Oxford Instruments). The chemical analysis on the dendrite scale was determined by electron-probe microanalysis (EPMA) Table 1 Nominal composition of ERBO1 in at %

Concentration/at.% ERBO1

Fig. 1 a Freestanding micropillar of Ni-based superalloy ERBO1 with diameter d and force F prior to mechanical testing. b FE simulation of indentation with a cylindrical flat punch indenter and (inset) FIB milled sapphire indenter tip for indentation creep experiments

(JXA-8100, Jeol) operating at a voltage of 20 kV, spot size of 5 lm, and a dwell time of 5 ms. The analyzing crystals used for detecting the elements were LiF (Co,W,Ta), TAP (Al,Ni,Hf), PETJ (Ti,Cr), and LIFH (Mo,Re). Freestanding micropillars were prepared after the widely used top-down methodology [16] using focused ion beam milling (FIB) (Crossbeam 540, Zeiss) in dendritic (DR) and interdendritic (IR) regions of the ERBO1A alloy. All micropillars were milled to a diameter of 4 µm and a height of 12 µm using concentric annular milling patterns with decreasing current and diameter. The starting current for rough milling was 30 °kv/15 nA which was gradually decreased to a fine milling current of 30 kV/ 100 pA. Micropillar compression testing was carried out at 785 °C using a high temperature nanoindenter (Alemnis AG), equipped with a tungsten carbide flat punch indenter (6 µm diameter). Prior to the experiments, the contact drift was minimized by adjusting the sample and tip temperature independently using thermal drift minimization protocols outlined in [17, 18]. The tests were performed in constant displacement control mode with an imposed displacement rate of 10 nm/s. To calculate the flow stress, the diameter measured at the top region of the pillar was used, as indicated in Fig. 1a. A modified thermomechanical analyzer (TMA) (402 F3 Hyperion, Netzsch) was used to perform indentation creep experiments under Argon atmosphere at 750 °C and at stress levels of 800 MPa and 1350 MPa. A cylindrical, constant cross-section flat punch indenter (20 µm diameter), FIB milled (Helios Nanolab 600I, FEI) from a conical sapphire indenter (SYNTON-MPD AG) with an individual manufactured shaft was used to ensure constant stress with increasing indentation depth (see inset Fig. 1b). This indenter geometry allows a conversion of the indentation creep data into equivalent uniaxial data for r and e_ with the method reported by Matschkal–Amberger et al. [14] as indicated by the FE

Ni

Co

Al

W

Ti

Ta

Cr

Re

Mo

Hf

Bal.

9.8

12.4

2.1

1.3

2.2

7.5

1.0

0.4

0.03

Local Mechanical Properties at the Dendrite Scale …

simulation in Fig. 1b. For the calculations, conversion factors C1 = 0.5 and C2 = 0.564 [14] were used, which were calculated by crystal plasticity finite element modeling (FEM) simulations for pure Ni at 650 °C. To achieve a thermal equilibrium and to reduce the thermal drift, the temperature of the sample and the low force were held constant for 2.5 h prior testing. To avoid oxidation of the samples, all tests were performed under Argon atmosphere. In addition, a Ti foil as oxygen getter was placed close to the sample. The validation of the method is given in [14]. In addition to the creep tests, the microstructure was analyzed close to the indentation sites by FIB cross sections and EDS analysis to determine the c′ volume fraction and precipitate size.

275

analyze the EPMA data. By sorting the concentration values of each element of the EPMA map, it can be estimated when a particular position solidified. This method was already proposed in the literature and used to investigate segregations in Ni- and Co-based superalloys [19–21]. The data points for Re and Ta concentrations from Fig. 2b, d are assigned to the concentration difference of Ta and Re (cTa– cRe), as this value increases constantly until the end of the solidification. A solid fraction fs between 0 and 1 is then assigned to the sorted concentrations, and the indents placed in the IR and DC (see next section) can be linked with the prevailing chemical concentrations, see Fig. 2e.

Local Mechanical Properties

Results and Discussion Microstructure and Segregation Behavior In the BSE images of the microstructure of ERBO1A in Fig. 2a, a brighter contrast reveals an accumulation of heavy elements, such as W or Re, in the DR. The elemental distribution was analyzed in detail by the EPMA-maps. They demonstrate that Re and W segregate in the dendrite cores (DC) and arms, whereas Ta segregates in the IR, see Fig. 2b–d. SEM investigations revealed that the different element concentrations lead to a varying c′ volume fraction, from 54.8% in the dendrite core to 66.0% in the IR close the c/c′ eutectic. Parsa et al. have reported a c′ volume fraction of 72% in the dendrite core and 77% in the IR on the same alloy in the fully heat treated condition [15]. In order to assign particular concentrations to the dendritic microstructure, a special sorting algorithm is used to Fig. 2 Microstructure of ERBO1A, a SEM micrograph of a dendrite in the BSE contrast, and concentration on the dendrite scale of b Re, c W, and d Ta determined by EPMA. Re and W segregate to the dendrite core, whereas Ta segregates strongly to the interdendritic region. e Shows the concentrations of Re and Ta in dependence of the solid fraction fs. The regions of the indents in the DR and IR shown in Fig. 3 are also marked in (c)

In Fig. 3, some of the residual indents after the indentation creep experiments performed at 800 MPa and 750 °C are shown. Figure 3a, c illustrates an indent in the DR, whereas Fig. 3b, d displays an indent positioned in the IR at different magnifications. The resulting data of the indentation creep experiments, conducted at 800 and 1350 MPa at 750 °C, is plotted in Fig. 4. The indentation depth and indentation strain rate point out that the creep strength in the DR is slightly higher than in the IR. To compare the determined values with macroscopic tensile creep experiments, the minimum strain rates are plotted in a Norton plot together with results from Wollgramm et al. [22], as shown in Fig. 4b. These results are also from alloy ERBO1, but in the fully heat-treated condition (ERBO1C). Therefore, the c′ volume fraction is higher, which explains the lower creep rates. It should also to be noted that ERBO1C exhibits a double creep minimum

276

L. Haußmann et al.

Fig. 3 Residual indents after loading with 800 MPa at 750 °C (a) in the dendritic region (DR) and (b) in the interdendritic region (IR) of the alloy ERBO1A. Higher magnifications of the indents are shown in (c) and (d)

Fig. 4 a Results from indentation creep tests in the DR and IR tested at 750 °C under 800 MPa and 1350 MPa. The creep data at 800 MPa corresponds to the indents shown in Fig. 2. b Norton plot of the indentation creep experiments in comparison with macroscopic tensile creep experiments of the fully heat treated alloy ERBO1C [22]

[22] similar to other Ni- and Co-based superalloys [23, 24]. The duration of the indentation creep experiments is rather short with 3 h and comparable to the duration until the first creep minimum in the macroscopic creep experiments, which is reached after about 10 h. Contrarily, the second creep minimum is reached after 100 h or more. Therefore, the first creep minimum was selected for comparison with Table 2 Concentration at the indentation sites in the DR and IR (see Fig. 3) measured via EDS

Region

Concentration/at % Co

DR IR

the macroscopic tensile creep experiments. The determined stress exponent of 4.7 from the indentation creep experiments is in very good agreement with the stress exponent of 5 from the macroscopic tensile creep experiments. The chemical composition measured via EDS at the locations of the indents sites (see Fig. 3) is listed in Table 2. The solid fraction at the particular position can then be

Ni

Al

9.9

61.1

15.6

W 3.1

Ti 0.9

Ta 1.6

Cr 6.0

Re 1.3

Mo 0.4

±0.7

±2.4

±2.5

±0.2

±0.1

±0.5

±1.0

±0.3

±0.1

9.3

60.9

14.8

2.6

1.2

2.3

7.3

1.0

0.6

±0.3

±1.3

±0.8

±0.1

±0.0

±0.1

±1.1

±0.2

±0.0

Local Mechanical Properties at the Dendrite Scale …

277

Fig. 5 a Micropillar in the DR and IR of the alloy ERBO1A and b higher magnification of one of the freestanding micropillars

Fig. 6 Results from micropillar compression testing in the DR and IR of the alloy ERBO1A at 785 °C, a stress–strain curves and b yield strength Rp0.2

calculated from the measured concentrations of Re and Ta together with Fig. 2e. The results emphasize the difference in microstructure at varying positions. While the microstructure at the indentation site in the DR solidifies in the beginning, with a maximum solid fraction fs of 21%, the indentation site in the IR solidifies between 40 and 80% solid fraction. Besides the indentation creep experiments, micropillar compression tests at selected positions were conducted. As shown in Fig. 5, the micropillars were prepared in different regions of the microstructure and were subsequently tested at 785 °C. As of now, this is the highest temperature at which micropillar compression tests have been performed on superalloys. No drift correction was conducted, since the thermal drift is very low with around 2 nm/min. Especially for the determination of the yield strength Rp0.2, which is reached within short experiment duration, the thermal drift should have a rather small influence on the measured data. The resulting stress–strain diagrams are shown in Fig. 6a and the determined yield strengths in Fig. 6b. The different regions —DR and IR—are separated by different colors. An evaluation of the yield strength Rp0.2 of pillars from the DR and IR results in an about 10% higher strength in the DR. This is in

good agreement with the slightly improved creep strength in the DR. It has to be noted that one micropillar in the IR shows a significantly lower strength compared to the others. The SEM image of this pillar shows an interface like contrast change at the bottom, which could indicate a sub-grain boundary. The strength level is in general comparable to literature data from CMSX-4 [6] and René N5 [25], albeit no direct comparison is possible due to different micropillar diameters, orientations, and testing temperatures. Shade et al. [25] have also reported higher strengths in the DR measured on pillar diameters of 5 µm and orientation at room temperature. A higher hardness in the DR is also reported in the literature [26, 27].

Correlation of Microstructure and Local Mechanical Properties For a detailed analysis of the increased strength in the DR, a local microstructure analysis was carried out at the indentation creep sites, and the strength contribution from solid solution hardening and precipitation hardening was estimated.

278

L. Haußmann et al.

The strength rsum of superalloys results mainly from the solid solution hardening in the c phase rMK and the c′ precipitation hardening rP: rsum ¼ rMK þ rP

0 cci

ð1Þ

c0 =c

and the partitioning coefficients ki;DR=IR were calculated for

the DR and IR from the composition of the c and c′ phase determined by Parsa et al. [15] in the fully heat treated state. The c′ volume fractions fV;DR=IR were directly determined at the indentation sites of the experiments at 750 °C under 800 MPa to a volume fraction of fV;DR = 58.0 ± 3.9% in the DR and fV;IR = 62.3 ± 2.0% in the IR. Since the partitioning

i

The solid solution hardening coefficient Ki describes the element specific contribution to the strength increase in dependence of the concentration cci . The Labusch model [29] suggests for the strengthening exponent n to be 2/3, which is used here. For the estimation of the strength contribution from solid solution hardening in the c phase in the DR and IR, the elements W, Re, Ta, and Mo will be considered, since they contribute strongly to the solid solution hardening [30]. The solid solution hardening coefficients Ki of W, Re, Ta, and Mo in Ni were calculated from data of [31, 32] to 1.5 GPa/at.%2/3, 1.5 GPa/at.%2/3, 1.8 GPa/at.%2/3 and 1.6 GPa/at.%2/3, respectively. The used data for the calculation was generated at 77 K, and therefore, the solid solution hardening coefficients at room temperature could be significantly lower. However, the calculated strengthening contributions are only estimations, and the qualitative results should be equal at both temperatures. The concentration of the alloying element in the c phase cci was estimated with Eq. (3): cci

c0 =c

coefficients ki are only slightly different in the DR and IR [15], it is assumed that they are equal to those of the as-cast state. Nevertheless, the calculated concentrations of the c and c′ phase in Table 3 are only an estimation. According to the model of Gypen and Deruyttere [28], together with the estimated concentrations for W, Re, Ta, and Mo in the gamma phase (see Table 3), solid solution hardening in the c phase rMK is 276 MPa in the DR and 258 MPa in the IR. This increased strength by 18 MPa in the DR compared to the IR is mainly due to the higher content of the strong solid solution hardening elements W and Re in the c phase of the DR. The strength increase from the interaction of dislocations with c′ precipitates rP is defined as followed: rP ¼ M
sP

ð3Þ

c0 =c

0

sP ¼

Whereas cci is the concentration of an element in the c′ c0 =c ki

phase, and is the partitioning coefficient between the c and c′ phase. The concentration in the c′ phase cci was estimated by Eq. (4): Table 3 Concentration of the c and c′ phase estimated by Eqs. (1) and (2) in at.%

ð5Þ

where sP is the critical resolved shear stress for precipitation hardening and M is the Taylor factor and has a value of 3 [33]. In order to estimate the c′ precipitation hardening, the unified approach of Galindo–Nava [34] for c′ precipitate cutting is used for the critical resolved shear stress sP:

0

ki

ð4Þ

The concentration ci was taken from the determined chemical composition at the indentation sites (see Table 2),

The increase of strength by solid solution hardening in Ni can be estimated after Gypen and Deruyttere [28] from the sum of the contributions of individual alloying elements: !n X 1=n c DrMK ¼ K i
ci ð2Þ

cci ¼

c0 =c

ci
ki;DR=IR i ¼ h  c0 =c 1  fV;DR=IR þ fV;DR=IR
ki;DR=IR

cAPB
l 2b
ðK1 þ 2r Þ

ð6Þ

While cAPB is the antiphase boundary energy, l is the segment length of the leading partial dislocation, b is the

Region

Phase

Co

Ni

DR

c

16.2

51.8

3.5

4.0

0.1

0.1

12.7

3.0

0.7

±1.2

±2.0

±0.6

±0.3

±0.0

±0.0

±2.0

±0.6

±0.2

5.4

67.9

24.4

2.4

1.5

2.6

1.3

0.1

0.2

±0.4

±2.6

±3.9

±0.2

±0.2

±0.8

±0.2

±0.0

±0.1

16.0

50.8

3.7

3.5

0.1

0.1

16.3

2.4

1.0

±0.5

±1.1

±0.2

±0.1

±0.0

±0.0

±2.4

±0.4

±0.1

5.3

67.1

21.6

2.0

1.8

3.7

1.8

0.1

0.3

±0.2

±1.4

±1.2

±0.0

±0.0

±0.2

±0.3

±0.0

±0.0

c′ IR

c c′

Al

W

Ti

Ta

Cr

Re

Mo

Local Mechanical Properties at the Dendrite Scale …

279

burgers vector, K1 is the average precipitate distance along the dislocation line, and r is the precipitate radius. Since the precipitate radius in the DR and IR is larger than the radius with the highest counteracting force rm (see Eq. 7), strong pair-coupling can be assumed. G
b2 2cAPB

rm ¼

ð7Þ

In this case, l is defined by (8) and K1 by (9) with L as the effective precipitate distance (10).  1=2 l ¼ 2 r 2  ðr  rm Þ2

ð8Þ

K1 ¼ L  l

ð9Þ

 L¼

2p 3fV

1=2

ð10Þ

The calculation refers to the cutting of edge dislocations within c′ precipitates. Here, it is worth to be mentioned that the model of Galindo–Nava et al. [34] is valid for spherical precipitates, while the precipitates in ERBO1A are more cubic. This could have a significant influence on the calculated strength contribution from precipitation hardening. The c′ precipitate sizes in the DR and IR were also directly determined at the indentation sites of the experiments at 750 °C under 800 MPa to 480 nm in the DR and 600 nm in the IR. An shear modulus G of 90.2 GPa at 1000 K was calculated after [35] with the elastic constants of ERBO1C [36] and a Burgers vector b of 0.254 nm that was taken from [37] for the c′ hardened Ni-based superalloy Nimonic PE16. Considering the different concentrations of elements in the c′ phase in the DR and IR, the antiphase boundary (APB) energy was estimated after the model of Crudden et al. [38]: cAPB ¼ c0APB þ

n  X i¼1

ui
cci

0



ð11Þ

Whereas c0APB describes the APB energy of Ni3Al and is 195 ± 13 mJ/m2. The model assumes a linear relationship between the contributions of the individual alloying elements i. The coefficient ui describes the change of the APB energy 0

by an element i, which has the concentration cci in the c′ phase. The coefficients to describe the APB energy change ui of Cr, Mo, W, Ta, and Ti were calculated to −1.7 mJ/m2, −1.7 mJ/m2, 4.6 mJ/m2, 27.1 mJ/m2 and −15 mJ/m2,

Therefore, the calculated strength contribution by precipitation hardening rP in the DR is 297 MPa, while it is 279 MPa in the IR. This reveals that the strength in the DR is 18 MPa higher than in the IR, despite the higher c′ volume fraction and the higher APB energy in the IR due to the higher concentration of c′ forming element Ta. The decisive factor is the smaller c′ precipitate size in the DR (see Eq. 6) which compensates the lower c′ volume fraction and APB energy under the assumption of c′ precipitate cutting by strong pair-coupling of the dislocations. The theoretically calculated strength contributions result in a total strength of 573 MPa in the DR and 537 MPa in the IR. The 36 MPa higher strength of the DR supports the results of the indentation creep and micropillar compression tests, in which also slightly better mechanical properties in the DR could be determined. Laplanche et al. [39] also investigated the DR and IR of the alloy ERBO1 but by an in situ SEM micromechanical test technique at room temperature and in the fully heat treated condition ERBO1C. Contrary to the as-cast condition in this study, they observed a higher strength in the IR than in the DR. There the elemental distribution was more homogeneous in the heat treated condition, the c′ volume fraction in the IR of 77% and in the DR of 72% was much higher, but the precipitate size with nearly equal values of 600 nm in the IR and 590 nm in the DR was much closer to each other. This might lead to a higher strength of the IR due to additional strength contributions of 5% from the slightly higher c′ volume fraction and 6% from the narrower c′ channels, as stated in [39]. Thus, for the fully heat-treated state, a smaller difference in the local mechanical properties can be expected, despite the still existing but lower segregations, especially of Re. Even a reversal of the strength ratio of DR and IR is possible, as shown by Laplanche et al. [39], depending on the local c′ morphology and volume fraction.

Conclusions The local microstructural and mechanical differences of the as-cast single crystalline Ni-based alloy ERBO1A were investigated by SEM, EDS, and EPMA measurements and by indentation creep and micropillar compression tests at elevated temperatures. Together with theoretical calculations on the local strength, the following conclusions can be drawn:

0

respectively [38]. The concentrations cci can be found in Table 3. This results in an APB energy cAPB of 297 mJ/m2 in the DR and 328 mJ/m2 in the IR. It is unlikely that the alloying elements Co, Ni, Al, and Re lead to a further increase of the APB energy in the IR, as Co, Ni, and Re have a comparable concentration in the c′ phase in the DR and IR.

• ERBO1 shows a distinct dendritic microstructure in the as-cast state, and the solid solution hardening elements Re and W segregate in the dendritic region (DR), whereas the c′ forming element Ta segregates in the interdendritic region (IR).

280

• The indentation creep experiments at 750 °C reveal a two times lower creep rate in the DR than the IR, and the micropillar compression tests at 785 °C show an about 10% higher yield strength in the DR than in the IR. • The calculated strength increase by solid solution hardening is 276 MPa for the DR and 258 MPa for the IR due to higher concentrations of W and Re. The calculated strength increase by c′ precipitation hardening is 297 MPa for the DR and 279 MPa for the IR. This is caused by a smaller c′ precipitate size in the DR, despite a higher c′ volume fraction and a higher APB energy in the IR due to an enrichment of Ta. • The calculations are in good agreement with the nanomechanical measurements of a higher strength in the DR than in the IR of the as-cast condition. However, the differences might vanish or even invert in the fully heat treated condition, due to less segregation, higher c′ volume fractions, and a more similar c′ size between dendrite core and interdendritic regions.

Acknowledgements The authors gratefully acknowledge funding by the Deutsche Forschungsgemeinschaft (DFG) through projects A6 of the Collaborative Research Center SFB/TR 103 “From Atoms to Turbine Blades—a Scientific Approach for Developing the Next Generation of Single Crystal Superalloys” and thank S. Giese and C. Schunk from the Institute I: General Materials Properties at FAU for their support for milling the micropillars. Furthermore, the authors thank Siwen Gao from the Interdisciplinary Center for Advanced Materials Simulation of the Ruhr University Bochum for the FE simulation image.

L. Haußmann et al.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

References 1. Neumeier S, Pyczak F, Göken M (2011) Influence of rhenium and ruthenium on the local mechanical properties of the c and c′ phases in nickel-base superalloys. Philos Mag 91:4187–4199. https://doi. org/10.1080/14786435.2011.607139. 2. Rehman H ur, Durst K, Neumeier S, Parsa AB, Kostka A, Eggeler G, Göken M (2015) Nanoindentation studies of the mechanical properties of the l phase in a creep deformed Re containing nickel-based superalloy. Mater Sci Eng A 634:202– 208. https://doi.org/10.1016/j.msea.2015.03.045. 3. Göken M, Kempf M (1999) Microstructural properties of superalloys investigated by nanoindentations in an atomic force microscope. Acta Mater 47:1043–1052. https://doi.org/10.1016/ s1359-6454(98)00377-2. 4. Chu SNG, Li JCM (1977) Impression creep; a new creep test. J Mater Sci 12:2200–2208. https://doi.org/10.1007/bf00552241. 5. Trenkle JC, Packard CE, Schuh CA (2010) Hot nanoindentation in inert environments. Rev Sci Instrum 81:073901. https://doi.org/10. 1063/1.3436633. 6. Korte S, Stearn RJ, Wheeler JM, Clegg WJ (2012) High temperature microcompression and nanoindentation in vacuum. J Mater Res 27:167–176. https://doi.org/10.1557/jmr.2011.268. 7. Dorner D, Röller K, Skrotzki B, Stöckhert B, Eggeler G (2003) Creep of a TiAl alloy: a comparison of indentation and tensile

18.

19.

20.

21.

22.

testing. Mater Sci Eng A 357:346–354. https://doi.org/10.1016/ s0921-5093(03)00205-3. Mathew MD, Naveena, Vijayanand D (2013) Impression Creep Behavior of 316LN Stainless Steel. J Mater Eng Perform 22:492– 497. https://doi.org/10.1007/s11665-012-0290-4. Sundar RS, Kutty TRG, Sastry DH (2000) Hot hardness and creep of Fe3Al-based alloys. Intermetallics 8:427–437. https://doi.org/10. 1016/s0966-9795(99)00118-1. Cross GLW, O′Connell BS, Pethica JB, Rowland H, King WP (2008) Variable temperature thin film indentation with a flat punch. Rev Sci Instrum 79:013904. https://doi.org/10.1063/1.2830028. Rowland HD, King WP, Cross GLW, Pethica JB (2008) Measuring Glassy and Viscoelastic Polymer Flow in MolecularScale Gaps Using a Flat Punch Mechanical Probe. ACS Nano 2:419–428. https://doi.org/10.1021/nn700211g. Chu SNG, Li JCM (1979) Impression creep of b-tin single crystals. Mater Sci Eng 39:1–10. https://doi.org/10.1016/00255416(79)90164-2. Durst K, Göken M (2004) Micromechanical characterisation of the influence of rhenium on the mechanical properties in nickel-base superalloys. Mater Sci Eng A 387–389:312–316. https://doi.org/ 10.1016/j.msea.2004.03.079. Matschkal-Amberger D, Kolb M, Neumeier S, Gao S, Hartmaier A, Durst K, Göken M (2019) New flat-punch indentation creep testing approach for characterizing the local creep properties at high temperatures. Mater Des 183:108090. https://doi.org/10.1016/ j.matdes.2019.108090. Parsa AB, Wollgramm P, Buck H, Somsen C, Kostka A, Povstugar I, Choi P-P, Raabe D, Dlouhy A, Müller J, Spiecker E, Demtroder K, Schreuer J, Neuking K, Eggeler G (2015) Advanced Scale Bridging Microstructure Analysis of Single Crystal Ni-Base Superalloys: Advanced Scale Bridging Microstructure Analysis. Adv Eng Mater 17:216–230. https://doi.org/10.1002/adem.201400136. Greer JR, De Hosson JThM (2011) Plasticity in small-sized metallic systems: Intrinsic versus extrinsic size effect. Prog Mater Sci 56:654–724. https://doi.org/10.1016/j.pmatsci.2011.01.005. Conte M, Mohanty G, Schwiedrzik JJ, Wheeler JM, Bellaton B, Michler J, Randall NX (2019) Novel high temperature vacuum nanoindentation system with active surface referencing and non-contact heating for measurements up to 800 °C. Rev Sci Instrum 90:045105. https://doi.org/10.1063/1.5029873. Wheeler JM, Brodard P, Michler J (2012) Elevated temperature, in situ indentation with calibrated contact temperatures. Philos Mag 92:3128–3141. https://doi.org/10.1080/14786435.2012.674647. Koßmann J, Zenk CH, Lopez-Galilea I, Neumeier S, Kostka A, Huth S, Theisen W, Göken M, Drautz R, Hammerschmidt T (2015) Microsegregation and precipitates of an as-cast Co-based superalloy – microstructural characterization and phase stability modelling. J Mater Sci 50:6329–6338. https://doi.org/10.1007/ s10853-015-9177-8. Karunaratne MSA, Cox DC, Carter P, Reed RC (2000) Modelling of the Microsegregation in CMSX-4 Superalloy and its Homogenisation During Heat Treatment. In: Green K, Kissinger R, Pollock TM (eds) Superalloys 2000. The Minerals, Metals & Materials Society, Warrendale PA, pp 263–272. Ganesan M, Dye D, Lee PD (2005) A technique for characterizing microsegregation in multicomponent alloys and its application to single-crystal superalloy castings. Metall Mater Trans A 36:2191– 2204. https://doi.org/10.1007/s11661-005-0338-2. Wollgramm P, Bürger D, Parsa AB, Neuking K, Eggeler G (2016) The effect of stress, temperature and loading direction on the creep behaviour of Ni-base single crystal superalloy miniature tensile specimens. Mater High Temp 33:346–360. https://doi.org/10. 1080/09603409.2016.1186414.

Local Mechanical Properties at the Dendrite Scale … 23. Xue F, Zenk CH, Freund LP, Hoelzel M, Neumeier S, Göken M (2018) Double minimum creep in the rafting regime of a single-crystal Co-base superalloy. Scr Mater 142:129–132. https://doi.org/10.1016/j.scriptamat.2017.08.039. 24. Wu X, Wollgramm P, Somsen C, Dlouhy A, Kostka A, Eggeler G (2016) Double minimum creep of single crystal Ni-base superalloys. Acta Mater 112:242–260. https://doi.org/10.1016/j.actamat. 2016.04.012. 25. Shade PA, Uchic MD, Dimiduk DM, Viswanathan GB, Wheeler R, Fraser HL (2012) Size-affected single-slip behavior of René N5 microcrystals. Mater Sci Eng A 535:53–61. https://doi. org/10.1016/j.msea.2011.12.041. 26. Rehman H ur (2016) Solid Solution Strengthening and Difusion in Nickel- and Cobalt-based Superalloys. Ph.D. thesis, Friedrich-Alexander Universität Erlangen-Nürnberg. 27. Tromas C, Arnoux M, Milhet X (2012) Hardness cartography to increase the nanoindentation resolution in heterogeneous materials: Application to a Ni-based single-crystal superalloy. Scr Mater 66:77–80. https://doi.org/10.1016/j.scriptamat.2011.09.042. 28. Gypen LA, Deruyttere A (1977) Multi-component solid solution hardening: Part 1 Proposed model. J Mater Sci 12:1028–1033. https://doi.org/10.1007/bf00540987. 29. Labusch R (1970) A Statistical Theory of Solid Solution Hardening. Phys Status Solidi B 41:659–669. https://doi.org/10. 1002/pssb.19700410221. 30. Roth HA, Davis CL, Thomson RC (1997) Modeling solid solution strengthening in nickel alloys. Metall Mater Trans A 28:1329– 1335. https://doi.org/10.1007/s11661-997-0268-2. 31. Mishima Y, Ochiai S, Hamao N, Yodogawa M, Suzuki T (1986) Solid Solution Hardening of Nickel – Role of Transition Metal and B-subgroup Solutes. Trans Jpn Inst Met 27:656–664. https://doi. org/10.2320/matertrans1960.27.656.

281 32. Diologent F, Caron P (2004) On the creep behavior at 1033 K of new generation single-crystal superalloys. Mater Sci Eng A 385:245–257. https://doi.org/10.1016/s0921-5093(04)00925-6. 33. Cottrell AH (1965) Dislocation and plastic flow in crystals. Clarendon Press, Oxford. 34. Galindo-Nava EI, Connor LD, Rae CMF (2015) On the prediction of the yield stress of unimodal and multimodal c′ Nickel-base superalloys. Acta Mater 98:377–390. https://doi.org/10.1016/j. actamat.2015.07.048. 35. Pottebohm H, Neite G, Nembach E (1983) Elastic properties (the stiffness constants, the shear modulus and the dislocation line energy and tension) of Ni-Al solid solutions and of the Nimonic alloy PE16. Mater Sci Eng 60:189–194. https://doi.org/10.1016/ 0025-5416(83)90001-0. 36. Demtröder K, Eggeler G, Schreuer J (2015) Influence of microstructure on macroscopic elastic properties and thermal expansion of nickel-base superalloys ERBO/1 and LEK94. Mater Sci Eng Technol 46:563–576. https://doi.org/10.1002/mawe.201500406. 37. Nembach E, Neite G (1985) Precipitation hardening of superalloys by ordered c′-particles. Prog Mater Sci 29:177–319. https://doi. org/10.1016/0079-6425(85)90001-5. 38. Crudden DJ, Mottura A, Warnken N, Raeisinia B, Reed RC (2014) Modelling of the influence of alloy composition on flow stress in high-strength nickel-based superalloys. Acta Mater 75:356–370. https://doi.org/10.1016/j.actamat.2014.04.075. 39. Laplanche G, Wieczorek N, Fox F, Berglund S, Pfetzing-Micklich J, Kishida K, Inui H, Eggeler G (2018) On the influence of crystallography and dendritic microstructure on micro shear behavior of single crystal Ni-based superalloys. Acta Mater 160:173–184. https://doi.org/10.1016/j.actamat.2018.08.052.

Phenomenological Modeling of the Effect of Oxidation on the Creep Response of Ni-Based Single-Crystal Superalloys Jean-Briac le Graverend and Seungjun Lee

Abstract

A new constitutive equation predicting the evolution of oxide thickness is proposed and implemented in a crystal plasticity framework with a hardening-based damage density function. The oxidation model depends on the initial surface roughness as well as the amount of accumulated plastic strain and stress triaxiality. The effect of oxidation on the mechanical behavior and damage is considered at the flow stress level by modifying the amplitude of the kinematic hardening. The oxidationaltered kinematic hardening is able to predict the effect of oxidizing environments on the mechanical behavior, specifically the plastic strain rate in the secondary creep stage, and lifetime. In addition, the oxidation model has also been tested in 3D by performing a finite-element simulation on a notched specimen subjected to a creep load. It revealed that the model is able to predict surface roughness and oxide thickness distributions in 3D and for multiaxial stresses. Keywords















Ni-based superalloy Creep Oxidation Damage Crystal plasticity Surface roughness FE simulation

Introduction High-temperature components in jet and turbo engines are exposed to harsh corrosive environments that are detrimental to the mechanical behavior and lifetime. Indeed, oxidation J.-B. le Graverend (&)
S. Lee Department of Aerospace Engineering, Texas A&M University, College Station, TX 77843, USA e-mail: [email protected] J.-B. le Graverend Department of Materials Science Engineering, Texas A&M University, College Station, TX 77843, USA

comes with a depleted zone [1–6], vacancies injection, and acceleration of diffusion processes while forming an oxide scale, i.e., the succession of NiO, spinels Ni(Cr, Al)2O4, and Al2O3 layers [7–10], by outward cationic diffusion toward the external surface of the substrate [1, 11–13]. The growth of fast-growing cationic-type oxides or the Kirkendall effect during the interdiffusion enables a continuous injection of vacancies due to interfacial reactions. The vacancies in excess may be reinjected into the material and, therefore, alter the mechanical behavior either by producing deformation pores or by coarsening with already existing pores, namely solidification and homogenization pores [14–16]. In fact, Hancock [17] and Hales et al. [18–20] consider the appearance of these cavities as proof of the injection of vacancies. Furthermore, diffusion is also involved in two main creep mechanisms: dislocation climb to bypass c′ particles [21–23] and morphological changes of c′ precipitates by migration of c/c′ interfaces [24–27]. Therefore, oxidation has a “static” effect on creep rate by enhancing microstructure degradation, and a “dynamic” effect by accelerating dislocation motion. These two effects are coupled and lead to higher creep rates and shorter time to failure. Furthermore, oxidation also has an effect on fatigue life since cracks initiate more quickly when the surface integrity, which is related to the surface geometry and residual stress, is poor [28–32]. Indeed, a higher surface roughness fosters strain localization that is the premise of crack initiation [33–36]. It is, therefore, essential to have a more transverse consideration of the phenomena occurring at high temperature in superalloys, which will improve the thermomechanical response and lifetime prediction of critical structures. However, despite the numerous effects of oxidation on the mechanical behavior and lifetime, much of the work focused on oxidation-assisted crack initiations and propagations [37–41]. This work focuses on the effect of oxidation on the mechanical behavior and, consequently, on the damage evolution of Ni-based single-crystal superalloys. Thus, the effect of oxidation on crack initiations and

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_27

282

Phenomenological Modeling of the Effect of Oxidation …

propagations, which mainly happen during the tertiary creep stage or late in a creep-fatigue loading, is out of the scope. When superalloys are exposed to oxidizing environments, the growth of the oxide scale affects the microstructural stability, such as in the c′ depleted zone where the volume fraction of the c′ phase is almost reduced to zero [1]. It is why oxidation affects the mechanical behavior and lifetime of superalloys [1, 42]. The fundamental reason of this phenomenon is the depletion process of the c′ phase during the formation of the oxide scale. This increases the effective stress and speeds up the damage kinetics. Even if the level of temperature is the obvious parameter affecting the oxidation kinetics, viz. the higher the temperature is, the faster the oxidation kinetics [5], one should keep in mind that oxidizing-particle flow rate, surface roughness (Ra) [43–45], and plastic strain [46] also play a role. In fact, a higher oxidation kinetics was measured for Ra = 91 nm than for Ra = 19 nm on a Ni–4.0Cr–5.7Al single-crystal superalloy [11]. Even if the authors did not discuss this result, this suggests that there may be a critical value of the initial surface roughness that will favor oxidation the most. This result is consistent with what the authors found out with regard to the effect of flow rate on the kinetics of oxidation, namely the kinetics depends on the flow rate and not in a linear way [47]. These two results can be understood using an analogy with catalytic reaction for which the number of active sites is in competition in a nonlinear manner with the activation energy for the oxidation process [48]. This is materialized by the peak shape of the activation energy which is dependent on the size of the involved oxidizing molecules. Despite this fundamental mechanism originating at the nano-scale, this will not be considered in the present paper since it would require a full chemical study that is beyond the scope of what is envisioned for this paper. It is why a phenomenological approach on the effect of surface roughness is employed. In addition to surface roughness, plastic deformation will also be considered. Its effect on the oxidation kinetics is two-fold. On one hand, oxidation is a diffusion process that is enhanced by pipeline diffusion [49]. On the other hand, plasticity has implications in surface chemistry since dislocations propagating on slip systems and emerging at the surface of specimens create steps that modify the surface roughness and, therefore, the oxidation kinetics. It is why the effect of plasticity is critical to be considered and will lead to a faster oxidation kinetics when and where the amount of plasticity is higher. Finally, contrary to many already-developed approaches that specifically formulate a damage density function for oxidation [50–53], oxide growth is hereby phenomenologically modeled and affects the mechanical behavior, which, consequently, changes the kinetics of damage through a hardening-based damage density function developed in [54].

283

In the following, the crystal plasticity model as well as the oxidation model is presented and calibrated for the first-generation Ni-based single-crystal superalloy MC2. The model is then tested on the reported creep data in vacuum and oxidizing environments as well as by in 3D on a notched specimen subjected to creep.

Modeling of the Mechanical Behavior and Damage A crystal plasticity framework already described in [54, 55] will be employed. The model follows a micro–macroapproach. In addition, it is a microstructure-sensitive model, namely c′ rafting is taken into consideration through an Orowan stress implemented in an isotropic hardening r s defined at the slip system level s. c_ s is the viscoplastic shear strain rate on a given slip system s (see Eq. 1) in which the effective resolved shear stress sseff is obtained by the tensorial product between the macroscopic effective stress tensor e r ¼ r=ð1  DÞ, as described in [56], and the orientation tensor ms , calculated knowing the normal to the slip system plane ns and the slip direction in this plane ls . xs is the nonlinear kinematic hardening on the slip system s.  1n 3 20    sseff  xs   r s A 5sign ss  xs c_ s ¼ Cvisco sinh4@ eff K     with xs ¼ Cx as where a_ s ¼ sign sseff  xs  Dx as c_ s rffiffiffi Nsyst X 2 GB s s j where q_ s ¼ ð1  bqs Þ_cs and r ¼ r0 þ bQ hsj q þ 3 w j¼1 ð1Þ where Cvisco , n, K, Cx , Dx , b, and Q are temperature-dependent material parameters. r0s is the initial radius of the yield curve on the slip system s. ½h is the interaction matrix whose form has been described in [57] and is assumed to have all its components equal to 1. B is the magnitude of the Burgers′ vector. G is the shear modulus. w is the width of the c channels (nm) whose the rate-sensitive constitutive equations are presented in [55]. q j is the isotropic state variable on the slip system j that models the evolution of the dislocation density on each slip systems and is, therefore, related to dislocation hardening. The damage density function used to compute the effective stress is the one developed by le Graverend in [54] and provided in Eq. 2. This damage function is coupled to the mechanical behavior which means that any change in the plastic strain rate due to an external factor, such as oxidation, leads to a modification of the kinetic of damage and, therefore, to a shorter lifetime.

284

J.-B. le Graverend and S. Lee

Table 1 Material parameters used for the mechanical behavior and damage at 1150 °C for the MC2 alloy

D_ ¼

n

K (MPa)

3.6

550

1.1

Cx (MPa)

Dx

A (MPa)

20200

650

230

!rðFv0 Þ  X 12 i rqi  x  þ  ð1  DÞk   i¼1 A B

Cvisco

ð2Þ

where D is the damage variable and D_ is, therefore, the damage kinetics, r ðFv0 Þ is a function of the initial volume fraction of pores Fv0 , rqi is one of the components of the isotropic hardening for the slip system i related to the evolution of dislocations. A, B, and k are temperature-dependent material parameters. The parameters used for the mechanical behavior and damage of the MC2 alloy are gathered in Table 1. The material parameters for the mechanical behavior and damage were obtained by first calibrating the oxidation model against the thermogravimetric analysis (TGA) results from Pei et al. [43] (see next section) and then by fitting the creep curve at 1150 °C/80 MPa in air from Viguier et al. [58].

Modeling the Effect of Oxidation on the Mechanical Behavior and Damage This section is dedicated to depicting the choices that have been made by the authors to have a kinetics of oxidation depending on the surface roughness, the accumulated plastic strain, and the amount of stress triaxiality. Once the oxidation model is presented, the way to model its effects on the mechanical behavior and lifetime is also discussed. Pei et al. [43] considered the effect of surface roughness on the oxidation kinetics at 1000 °C in air for a Ni–4.0Cr– 5.7Al single-crystal superalloy. These authors investigated four surface roughnesses by thermogravimetry analysis (TGA): 19, 90, 182, and 509 nm. They found that decreasing the surface roughness increases the oxidation kinetics up to a value between 19 and 90 nm. Away from this surface roughness, the oxidation kinetics is slower. It is why the   function f Rap in Eq. 3 has been developed to account for the coupling between surface roughness and oxidation kinetics. A Lorentzian function was found to better depict what was experimentally obtained by Pei et al. However, one should understand that the obtained material parameters are for oxidation in air (molecular diameter of dioxygen 0.292 nm). It means that corrosion in different environments

r0 (MPa)

Q (MPa)

b

GB (MPa)

1.3

68

2866

B (MPa)

k

r

55

2

5.5

6.5

with smaller or larger molecules will lead to different distributions for the coupling between surface roughness and corrosion kinetics due to different activation energies [59]. Unfortunately, there is no result on the coupling between corrosion kinetics, surface roughness, and molecule size. This prevents the authors to refine the expression provided in Eq. 3. Similarly, the effect of the flow rate on the oxidation kinetics is not taken into consideration because of insufficient or unavailable experimental data, even if it would require to be considered to be more representative of what happens in engines, for instance.   In the function f Rap , Rap accounts for the effect of plasticity on the surface roughness, as shown by Becker [60] in the case of sheet forming. Indeed, this author found that there is a linear relationship between the surface roughness Ra and the amount of deformation. In addition, the same author found that the slope is affected by the multiaxiality: a biaxial deformation leading to a smaller slope than for a uniaxial loading. It is why the surface roughness is modeled to increase with the amount of accumulated plastic strain p and with a slope depending on a stress triaxiality factor   3r  H  ¼ I1 , as in [61], where J2 e r is the von Mises J2 e r J2 e r stress of the effective stress tensor, rH is the hydrostatic stress, and I1 is the first stress invariant. Finally, the oxidation constant Koxp will not be a constant and will also depend on the amount of accumulated plastic strain. In fact, Reuchet et al. [46] showed that the higher the strain amplitude during mechanical cycling at 900 °C is, the thicker the oxide scale, as also shown by Bucklow and Skelton on a Cr–Mo–V steel at 500 °C [62]. The oxidation behavior of most of alloys is known to be linear, parabolic, logarithmic, or often a combination of the three when the oxidation constant is assumed to be constant with time [63]. It can be pointed out that some studies account for the variation of the oxidation constant depending on the oxidation stage [64, 65]. For the simplicity of the modeling process, a logarithmic oxidation behavior is assumed, as this behavior has been observed in many thermogravimetric tests for Ni-based single-crystal superalloys [7, 43, 46], and accounts for the effect of the surface roughness as well as of plasticity on the kinetics of oxidation, as described in the previous paragraphs.

Phenomenological Modeling of the Effect of Oxidation …

285

 eox   e_ ox ¼ Koxp exp  f Rap with Kox     1 p 0:5C Koxp ¼ Kox 1 þ  and f Rap ¼  where  e p Rap  l 2 þ ð0:5CÞ2 0 1 3r H Rap ¼ Ra @1 þ 10   pA r J2 e

ð3Þ where e_ ox is the kinetics of the oxide-scale thickness, Kox is a material parameter, and p is the accumulated plastic strain qffiffiffiffiffiffiffiffiffiffiffiffiffiffi given by 23 ep : ep with ep the plastic strain tensor given by P s s   c m . f Rap is a Lorentzian distribution function. C and

intrinsic mechanical potential of dissipation. Indeed, Dox continuously decreases its magnitude, which means that X : a_ continually becomes smaller and the potential of dissipation gradually evolves to be larger. The material parameters used for the oxidation process in the MC2 alloy are gathered in Table 2. The parameters related to the oxidation model, i.e., in Eq. 3, were obtained by reproducing the work done by Pei et al. [43]. The parameters in Eq. 4 were obtained by fitting the normalized curves in air and in ArH2 given by Dryepondt et al. at 1150 °C [1] to obtain the proper magnitude of softening brought by the oxidation process. The reader is invited to look at the next section for more details.

s

l are constants that determine the shape, scale, and location of the distribution. As previously expressed, a few were done with regard to the effect of oxidation on the mechanical behavior. Dryepondt et al. [1] are almost the only ones who showed the effect of oxidation on creep at 1150 °C. They considered two environments: synthetic air and hydrogenated argon (ArH2). Even if these authors did not provide the stress level, it is usual for the same group to investigate 80 MPa at this level of temperature [58]. Dryepondt′s experiments show that oxidation has no effect on the amplitude of the primary creep stage, but has a large effect on the plastic strain rate of the secondary creep stage, namely the plastic strain rate is increased when the environment is more oxidizing. It would mean that the viscous flow is affected by oxidation but with a delay. It is why only the viscous flow was modified and a damage density function dedicated to the effect of oxidation was not added to the density function presented in Eq. 2, contrary to what was previously developed by many researchers [66–68]. Even though the viscous flow should be modified, the parameters n, K, and Cvisco (see Eq. 1) are not the ones that should be affected since a fast modification of these parameters by the oxidation process would lead to different amplitudes of the primary creep stage between tests performed in air or in a less oxidizing environment, like ArH2. It is why it was decided to modify the kinematic hardening, which is related to the state of internal micro-stress concentration, by making the oxidation phenomenon decreases its amplitude. This means that oxidation softens the mechanical behavior leading to an increase in the plastic strain rate. The equation for the kinematic hardening (Eq. 1) was, therefore, modified as followed: xs ¼

Cx s a with Dox ¼ 1 þ ðAox eox Þnox Dox

ð4Þ

where Aox and nox are material parameters. It is hereby important to point out that the modification of the kinematic hardening by Dox does not question the positivity of the

Results and Discussion The Lorentzian distribution was first tested in comparison with the results obtained by Pei et al. [43] who performed TGA analyses to study the effect of surface roughness on the kinetics of oxidation at 1000 °C (see Fig. 1). The simulations for the Pei′s surface roughnesses consist of holding the temperature at 1150 °C for 100 h without any load to mimic the TGA analyses. This comparison is not aimed to be quantitative since the considered temperatures are different —1150 °C in the present study—and Pei et al. plotted the mass gain when the oxide thickness is hereby considered. Figure 1 clearly highlights that the oxidation kinetics is affected by the surface roughness in a similar fashion that what was obtained by Pei et al. Also, having higher mass gain or oxide thickness for higher temperatures is consistent with what obtained by Hussain et al. [69]. The magnitude of eox is consistent with the magnitude found by Bensch et al. [64] for the René N5 at 980 °C. In order to compare the effects of oxidation predicted by the model with the experimental results obtained by Dryepondt et al. [1], it is necessary to calibrate the model at 1150 °C. The crystal plasticity model with the damage density function has been calibrated at 1150 °C with a creep test at 80 MPa in air on the MC2 alloy [58] with a surface roughness of 182 nm—value used from now on (see Fig. 2). The model perfectly reproduces the primary and secondary creep stages as well as the lifetime for the temperature/stress condition. It is, therefore, now possible to compare the predicting capability of the developed model with the experimental results from Dryepondt et al. [1] (see Fig. 3). The increase of the plastic strain rate in the secondary creep stage is perfectly reproduced by the model without modifying the amplitude of the primary creep stage (Fig. 3a). The oxidation in ArH2 was obtained by considering what Dryepondt et al. [1] obtained by TGA analyses under ArH2 and synthetic air at

286 Table 2 Material parameters used for oxidation at 1150 °C and for the MC2 alloy

J.-B. le Graverend and S. Lee Kox (lm)

e

C (lm)

l (lm)

Aox (lm−1)

nox

1

0.0002

0.08

0.08

0.064

10

Fig. Comparison between the evolution of the mass gain experimentally obtained by Pei et al. [43] at 1000 °C for a Ni-4.0Cr-5.7Al single-crystal superalloy for different surface roughnesses and the evolution of the oxide thickness predicted by the model presented in the previous section at 1150 °C

Fig. Creep test at 1150 °C/80 MPa on the Ni-based single-crystal superalloy MC2. The experimental result is adapted from [57]

1150 °C. They, indeed, found that the mass gain for ArH2 is 42% less than for synthetic air after 7 h. To obtain the same qualitative result, Kox is divided by 70 compared to the one for synthetic air. The model predicts a longer lifetime for ArH2 than for synthetic air. It is because the oxide thickness is smaller leading to less softening of the mechanical behavior and, therefore, to a longer lifetime. The curve for a creep test at 1150 °C/80 MPa in vacuum was also added to show the maximum reachable lifetime predicted by simulation. These results are consistent with what was expected to be obtained. Figure 3b shows how eox as well as the damage

density function with and without oxidation evolve. Oxidation acts like a higher stress which increases the nonlinearity of the damage evolution. It is in accordance with the physics of oxidation that creates a depleted zone and, therefore, decrease the cross-sectional area leading to a higher effective stress, as predicted by the Continuum Damage Mechanics. The nonlinear accumulation of damage with the level of stress was, for instance, obtained by Chaboche [70] by introducing a stress-dependent order-2 polynomial function for the parameter k. To verify that oxidation has no effect on fast mechanical loading, a strain-controlled tensile test at

Phenomenological Modeling of the Effect of Oxidation …

287

Fig. 3 a Comparison between the experimental results in air and in ArH2 and the simulations performed with the model presented in the previous section. The curves are normalized by the time to rupture in air. The experimental results are adapted from [3]. b Evolution of eox and the damage density function D with and without oxidation. tr is the time to rupture for the simulation considered

1150 °C and 10−3 s−1 was simulated up to 3% of deformation (Fig. 4). The two curves—with and without oxidation—are the same. That means that the growth of the oxide scale during the 30 s test is not enough to modify the mechanical response of the alloy, which one can expect for very short experiments, as it was shown by Wu and Baker for strain rates around 10−2 s−1 on FeAl single crystal [71]. Dryepondt et al. [1] also performed two extra tests during which the environment was either switched from air to ArH2 or from ArH2 to air (see Fig. 5). Before switching, the model nicely matches the experimental curves. However, if the model is able to predict the change in the plastic strain rate when the environment is switched from air to ArH2, it also predicts a large effect when the environment is switched from ArH2 to air, whereas it is not the case. Indeed, the plastic strain rate is experimentally decreased after the switch and was attributed to a dynamic interaction between the injection of defects due to oxidation and the creep deformation mechanisms. This phenomenon cannot be

(a)

depicted by the current model that does not consider vacancy injections and their interactions with dislocations. It would require to physically consider the oxide scale and to perform discrete dislocation dynamics simulations to refine the formulation of eox. The effect of the plastic deformation on the oxidation kinetics has also been investigated by considering four pre-plastic deformation on the mechanical behavior and lifetime during a creep test at 1150 °C/80 MPa (see Fig. 6). As expected, a pre-deformation leads to shorter lifetime during creep (see Fig. 6a), as shown by Ayrault [72] and Nathal [73] who, respectively, studied the effect of pre-compression on tension creep at 1050 °C/170 MPa and pre-tension on compression creep at 1000 °C/207 MPa. Figure 6b clearly shows that pre-deformation acts as a softening phenomenon. It was previously explained that increasing the nonlinearity of the damage evolution is similar to an increase in the applied stress and vice versa. Also, it was shown in Fig. 3 that oxidation increases the

(b)

↗

Fig. 4 Simulation of a strain-controlled monotonic tensile test at 1150 °C and 10−3 s−1 up to 3% of deformation with and without the effect of oxidation

288

J.-B. le Graverend and S. Lee

Fig. 5 Comparison between the experimental results when the environment is either switched from air to ArH2 (blue curves) or from ArH2 to air (red curves). The black arrows show when the environment is switched for the two conditions. The curves are normalized by the time to rupture in air. The experimental results are adapted from [3]

nonlinearity of damage. Nonetheless, Fig. 6c shows that the higher the pre-plastic strain is, the faster the oxidation kinetics and the thicker the oxide scale, as found by Reuchet et al. [46] in the case of oxide thickness for samples pre-deformed cyclically. Thus, the effect of plasticity on the oxidation kinetics considered in Eq. 3 is qualitatively consistent with previous experimental results.

A finite-element simulation was also carried out to display the 3D capability of the model with multiaxial state of stresses. A creep test at 1050 °C/F = 165 N, i.e., a nominal stress of 45 MPa, on an asymmetric bi-notched specimen, as in [55, 74], was simulated. Figure 7 shows the distribution of the accumulated plastic strain p, the damage density function D, the surface roughness Rap , and the oxide

(a)

(b)

(c)

Fig. 6 Simulation of creep curves for the MC2 alloy at 1150 °C/80 MPa that have been subjected to pre-deformations in tension. a Creep curves for the four pre-plastic deformation investigated, namely 0.2, 0.4, 0.6,

and 0.8%, b and c are, respectively, the evolution of damage and eox during the creep simulations presented in (a)

Phenomenological Modeling of the Effect of Oxidation …

evcum, p 0.007230

(a)

damage, D 0.3213

(b)

289

Rap ( m) 0.1870

(c)

eox ( m) 14.7

0.006628

0.2945

0.1865

14.4

0.006025

0.2677

0.1861

14.2

0.005423

0.2410

0.1857

13.9

0.004820

0.2142

0.1853

13.6

0.004218

0.1874

0.1849

13.4

0.003615

0.1606

0.1845

13.1

0.003013

0.1339

0.1841

12.9

0.002410

0.1071

0.1837

12.6

0.001808

0.0803

0.1832

12.3

0.001205

0.0536

0.1828

12.1

0.000603

0.0268

0.1824

11.8

0.0

0.0

0.1820

11.6

(d)

Fig. 7 a Accumulated plastic strain p, b damage D, c Rap, and d eox distributions at failure after 27.8 h at 1150 °C/F = 165 N

thickness eox. eox is the most important at the center as well as at the notch which is consistent with what was experimentally found in [55, 74]. Indeed, cracks always start at the notch, which is not entirely due to oxidation because of the multiaxial state of stresses, but oxidation is known to help initiating and propagating cracks at this level of temperatures [75, 76]. Even if the FE simulation revealed that the oxidation model gives coherent distribution fields for the surface roughness and the oxide thickness, it remains, however, critical to account for other possible modifications of the activation energy of the oxidation reaction, such as the type of oxidizing molecules, if one want to apply this model to components exposed to realistic environments.

Summary A model has been developed to predict oxide growth depending on the surface roughness, the amount of plastic strain, and the stress triaxiality. Oxidation has been considered to modify the kinematic hardening, which subsequently alters the mechanical behavior and damage. The developed modeling approach is able to predict that oxidation acts as a higher stress that increases the nonlinearity of damage evolution. The proposed model is not only capable of predicting large changes in the plastic strain rate during the secondary creep stage because of oxidation, it can also predict what occurs in 3D in terms of oxide thickness and surface roughness distributions for multiaxial state of stresses. It is, therefore, envisioned that the phenomenological oxide growth model and its effect on the mechanical behavior and lifetime can ultimately be utilized on components subjected to in-service environments.

Acknowledgements The simulations were performed using the computing resources from Laboratory for Molecular Simulation (LMS) and High Performance Research Computing (HPRC) at Texas A&M University. The authors are grateful to Mr. James Fillerup and the financial support from AFOSR through Award No.: FA9550-17-1-0233 to carry out this study.

References 1. Dryepondt S, Monceau D, Crabos F, Andrieu E (2005) Static and dynamic aspects of coupling between creep behavior and oxydation on MC2 single crystal superalloy at 1150 °C. Acta Mater. 53:4199–4209. 2. Krupp U, Christ H-J (2000) Selective oxidation and internal nitridation during high-temperature exposure of single-crystalline nickel-base superalloys. Metall. & Mater. Trans. A 31:47–56. 3. Ang J, Sato A, Kawagishi K, Harada H (2006) The Oxidation Characteristics of the gamma-gamma’ Tie-Line Alloys of a Nickel-Base Superalloy. Mat. Trans. 47:291–294. 4. Akhtar A, Hook MS, Reed RC (2005) On the oxidation of the third-generation single-crystal superalloy CMSX-10. Metall. & Mater. Trans. A 36:3001–3017. 5. Angerman CL (1972) Long-term oxidation of superalloys. Ox. Met. 5:149–167. 6. Chateau E, Rémy L (2010) Oxidation-assisted creep damage in a wrought nickel-based superalloy: Experiments and modelling. Mater. Sci. Eng. A 527:1655–1664. 7. Bensch M, Preußner J, Hüttner R, Obigodi G, Virtanen S, Gabel J, Glatzel U (2010) Modelling and analysis of the oxidation influence on creep behaviour of thin-walled structures of the single-crystal nickel-base superalloy René N5 at 980 °C. Acta Mater. 58:1607– 1617. 8. Akhtar A, Hegde S, Reed RC (2006) The oxidation of single-crystal nickel-based superalloys. JOM 58:37–42. 9. Sato A, Chiu YL, Reed RC (2011) Oxidation of nickel-based single-crystal superalloys for industrial gas turbine applications. Acta Mater. 59:225–240. 10. Göbel M, Rahmel A, Schütze M (1993) The isothermal-oxidation behavior of several nickel-base single-crystal superalloys with and without coatings. Ox. Met. 39:231–261.

290 11. Francis R, Lees DG (1989) Evidence for vacancy injection during the oxidation of iron. Mat. Sci. Eng. A 120–121:97–99. 12. Peraldi R, Monceau D, Pieraggi B Correlations Between Growth Kinetics and Microstructure for Scales Formed by High-Temperature Oxidation of Pure Nickel. I. Morphologies and Microstructures. Ox. Met. 58:249–273. 13. Perusin S, Monceau D, Andrieu E (2005) Investigations on the Diffusion of Oxygen in Nickel at 1000 ° C by SIMS Analysis. J. Electrochem. Soc 152:E390–E397. 14. Dunnington BW, Beck FH, Fontana MG (1952) The Mechanism of Scale Formation On Iron at High Temperature. Corrosion 8:2–12. 15. Deacon C, Loretto MH, Smallman RE (1985) Behaviour of vacancies during oxidation. Mat. Sci. Tech. 1:344–350. 16. Engell H, Wever F (1957) Some basic problems of the formation and adherence of scale on iron. Acta Metall. 5:695–702. 17. Hancock P (1977) Influence of vacancies produced by oxidation on the mechanical properties of Ni and Ni-Cr alloys. Paper presented at Vacancies, Bristol, 13–16 September 1976. 18. Hales R, Hill AC (1972) The role of metal lattice vacancies in the high temperature oxidation of nickel. Corros. Sci. 12:843–853. 19. Gibbs GB, Hales R (1977) The influence of metal lattice vacancies on the oxidation of high temperature materials. Corros. Sci. 17:487–507. 20. Hales R, Smallman RE, Dobson PS (1968) The Formation and Growth of Vacancy Loops in Magnesium. P. Roy. Soc. Lond. A Mat. 307:71–81. 21. Benyoucef M, Clément N, Coujou A (1993) Transmission electron microscope in situ deformation of MC2 superalloy at room temperature. Mat. Sci. Eng. A 164:401–406. 22. Fredholm A, Strudel J-L (1984) On the creep resistance of some nickel base single crystals. Paper presented at International Symposium on Superalloys, Seven Spring, Pennsylvania, 7-11 October, 1984. 23. Pollock TM, Argon AS (1992) Creep resistance of CMSX-3 nickel base superalloy single crystals. Acta Metall. & Mater. 40:1–30. 24. Nabarro FRN (1997) The chemical driving force for rafting in superalloys. Scripta Mater. 37:497–501. 25. Louchet F, Hazotte A (1997) A model for low stress cross-diffusional creep and directional coarsening of superalloys. Scripta Mater. 37:589–597. 26. Reed RC, Matan N, Cox DC, Rist MA, Rae CMF (1999) Creep of CMSX-4 superalloy single crystals: effects of rafting at high temperature. Acta Mater. 47:3367–3381. 27. Buffière JY, Ignat M (1995) A dislocation based criterion for the raft formation in nickel-based superalloys single crystals. Acta Metall. & Mater. 43:1791–1797. 28. Mughrabi H (2006) Specific features and mechanisms of fatigue in the ultrahigh-cycle regime. Int. J. Fat. 28:1501–1508. 29. Andrews S, Sehitoglu H (2000) A computer model for fatigue crack growth from rough surfaces. Inter. J. Fat. 22:619–630. 30. Huang Q, Ren JX (1991) Surface integrity and its effects on the fatigue life of the nickel-based superalloy GH33A. Inter. J. Fat. 13:322–326. 31. Antolovich SD, Baur R, Liu S (1980) A mechanistically based model for high temperature LCF of Ni base superalloys. Paper presented at International Symposium on Superalloys, Seven Springs, PA, USA. 32. Wright PK (1988) Oxidation-fatigue interactions in a single-crystal superalloy In: Solomon H. et al. (Ed.) Low Cycle Fatigue, ASTM International, West Conshohocken, Pennsylvania, p 558–575. 33. Valsan M, Sastry DH, Rao KBs, Mannan SL (1994) Effect of strain rate on the high-temperature low-cycle fatigue properties of a nimonic PE-16 superalloy. Metall. & Mater. Trans. A 25:159–171.

J.-B. le Graverend and S. Lee 34. McMahon CJ, Coffin LF (1970) Mechanisms of damage and fracture in high-temperature, low-cycle fatigue of a cast nickel-based superalloy. Metall. Trans. 1:3443–3450. 35. Hyzak JM, Bernstein IM (1982) The effect of defects on the fatigue crack initiation process in two P/M superalloys: Part ii. surface-subsurface transition. Metall. Trans. A 13:45–52. 36. Fleury E, Rémy L (1993) Low cycle fatigue damage in nickel-base superalloy single crystals at elevated temperature. Mat. Sci. Eng. A 167:23–30. 37. Miller MP, McDowell DL, Oehmke RLT (1992) A Creep-Fatigue-Oxidation Microcrack Propagation Model for Thermomechanical Fatigue. J. Eng. Mater. Tech. 114:282–288. 38. Okazaki M, Yamazaki Y (1999) Creep-fatigue small crack propagation in a single crystal Ni-base superalloy, CMSX-2: Microstructural influences and environmental effects. Inter. J. Fat. 21:S79–S86. 39. Suzuki A, Gao Y, Lipkin D, Singhal A, Krug M, Konitzer DG, Almer J, Pollock TM, Bewlay B (2014) Oxide-assisted crack growth in hold-time low-cycle-fatigue of single-crystal superalloys. Paper presented at EuroSuperalloys Giens, France, 12–16 May, 2014. 40. Evans AG, He MY, Suzuki A, Gigliotti M, Hazel B, Pollock TM (2009) A mechanism governing oxidation-assisted low-cycle fatigue of superalloys. Acta Mater. 57:2969–2983. 41. Lafata MA, Rettberg LH, He MY, Pollock TM (2018) Oxidation-Assisted Crack Growth in Single-Crystal Superalloys during Fatigue with Compressive Holds. Metall. & Mater. Trans. A 49:105–116. 42. Mallikarjuna HT, Caley WF, Richards NL (2017) Oxidation Kinetics and Oxide Scale Characterization of Nickel-Based Superalloy IN738LC at 900 °C. J. Mater. Eng. Perform. 26:4838–4846. 43. Pei H, Wen Z, Li Z, Zhang Y, Yue Z (2018) Influence of surface roughness on the oxidation behavior of a Ni-4.0Cr-5.7Al single crystal superalloy. App. Surface Sci. 440:790–803. 44. Evans JL (2010) Effect of Surface Roughness on the Oxidation Behavior of the Ni-Base Superalloy ME3. J. Mater. Eng. Perform. 19:1001–1004. 45. Amaro RL, Antolovich SD, Neu RW, Singh PM (2016) High temperature oxidation and c′ depletion in the single-crystal superalloy PWA 1484. Mater. at High Temp. 33:476–488. 46. Reuchet J, Remy L (1983) Fatigue oxidation interaction in a superalloy—application to life prediction in high temperature low cycle fatigue. Metall. Trans. A 14:141–149. 47. Lee S (2019) Instabilities on High Temperature Isothermal Creep Performance in Ni-Based Single Crystal Superalloys. Master’s thesis, Texas A&M University. 48. Smith MA, Zoelle A, Yang Y, Rioux RM, Hamilton NG, Amakawa K, Nielsen PK, Trunschke A (2014) Surface roughness effects in the catalytic behavior of vanadia supported on SBA-15. J. Catal. 312:170–178. 49. Embury JD, Deschamps A, Brechet Y (2003) The interaction of plasticity and diffusion controlled precipitation reactions. Scripta Mater. 49:927–932. 50. Neu RW, Sehitoglu H (1989) Thermomechanical fatigue, oxidation, and Creep: Part II. Life prediction. Metall. Trans. A 20:1769– 1783. 51. Chaboche JL, Gallerneau F (2001) An overview of the damage approach of durability modelling at elevated temperature. Fatigue Fract. Eng. Mater. Struct. 24:405–418. 52. Wang R-Z, Zhu S-P, Wang J, Zhang X-C, Tu S-T, Zhang C-C (2019) High temperature fatigue and creep-fatigue behaviors in a Ni-based superalloy: Damage mechanisms and life assessment. Int. J. Fat. 118:8–21.

Phenomenological Modeling of the Effect of Oxidation … 53. Dyson BF, Osgerby S (1987) Modelling creep-corrosion interactions in nickel-base superalloys. Mater. Sci. Tech. 3:545–553. 54. le Graverend J-B (2019) A hardening-based damage model for fast-evolving microstructures: Application to Ni-based single crystal superalloys. Int. J. Plast. 123:1–21. 55. le Graverend J-B, Cormier J, Gallerneau F, Villechaise P, Kruch S, Mendez J (2014) A microstructure-sensitive constitutive modeling of the inelastic behavior of single crystal nickel-based superalloys at very high temperature. Int. J. Plast. 59:55–83. 56. Lemaître J, Krajcinovic D (1987) Continuum damage mechanics: theory and applications. Springer-Verlag, Place. 57. Franciosi P (1985) The concepts of latent hardening and strain hardening in metallic single crystals. Acta Metall. 33:1601–1612. 58. Viguier B, Touratier F, Andrieu E (2011) High-temperature creep of single-crystal nickel-based superalloy: microstructural changes and effects of thermal cycling. Phil. Mag. 91:4427–4446. 59. Stará I, Nehasil V, Matolín V (1995) The influence of particle size on CO oxidation on Pd/alumina model catalyst. Surf. Sci. 331– 333:173–177. 60. Becker R (1998) Effects of strain localization on surface roughening during sheet forming. Acta Mater. 46:1385–1401. 61. Zamrik S, Mirdamadi M, Davis DC (1993) A Proposed Model for Biaxial Fatigue Analysis Using the Triaxiality Factor Concept In: (Ed.) Adv. Multiaxial Fatigue, ASTM International, West Conshohocken, Pennsylvania, p 85–106. 62. Skelton RP, Bucklow JI (1978) Cyclic oxidation and crack growth during high strain fatigue of low alloy steel. Metal Sci. 12:64–70. 63. Younes CM, Allen GC, Nicholson JA (2007) High temperature oxidation behaviour of single crystal superalloys RR3000 and CMSX-4. Corros. Eng. Sci. Tech. 42:80–88. 64. Bensch M, Sato A, Warnken N, Affeldt E, Reed RC, Glatzel U (2012) Modelling of High Temperature Oxidation of Alumina-Forming Single-Crystal Nickel-Base Superalloys. Acta Mater. 60:5468–5480.

291 65. Hou XM, Chou KC (2009) Quantitative interpretation of the parabolic and nonparabolic oxidation behavior of nitride ceramic. J. Eur. Ceram. Soc. 29:517–523. 66. Gallerneau F, Chaboche J-L (1999) Fatigue Life Prediction of Single Crystals for Turbine Blade Applications. Int. J. Damage Mech. 8:404–427. 67. Kaminski M, Kanouté P, Kruch S, Busso EP, Chaboche J-L (2016) A high temperature fatigue damage model for single crystal superalloys. Mater. at High Temp. 33:412–424. 68. McLean M, Dyson BF (2000) Modeling the Effects of Damage and Microstructural Evolution on the Creep Behavior of Engineering Alloys. J. Eng. Mater. Tech. 122:273–278. 69. Hussain N, Shahid KA, Khan IH, Rahman S (1995) Oxidation of high-temperature alloys (superalloys) at elevated temperatures in air. II. Ox. Met. 43:363–378. 70. Chaboche JL (1978) Description thermodynamique et phenomenologique de la viscosite cyclique avec endommagement. 71. Wu D, Baker I (2001) The effect of environment and strain rate on the room temperature tensile properties of FeAl single crystals. 9:57–65. 72. Ayrault D (1989) Fluage à haute température de superalliages base nickel monocristallins. PhD Thesis, Ecole Nationale Supérieure des Mines de Paris (France). 73. Nathal MV, Mackay RA (1987) The stability of lamellar c–c′ structures. Mater. Sci. & Eng. 85:127–138. 74. le Graverend J-B, Pettinari-Sturmel F, Cormier J, Hantcherli M, Villechaise P, Douin J (2018) Mechanical twinning in Ni-based single crystal superalloys during multiaxial creep at 1050 °C. Mater. Sci. Eng. A 722:76–87. 75. Gabb TP, Gayda J, Miner RV (1986) Orientation and temperature dependence of some mechanical properties of the single-crystal nickel-base superalloy René N4: Part II. Low cycle fatigue behavior. 17:497–505. 76. Yi JZ, Torbet CJ, Feng Q, Pollock TM, Jones JW (2007) Ultrasonic fatigue of a single crystal Ni-base superalloy at 1000 °C. 443: 142–149.

Prediction of Rafting Kinetics of Practical Ni-Based Single-Crystal Superalloys Yusuke Matsuoka, Yuhki Tsukada, and Toshiyuki Koyama

Abstract

Introduction

Directional coarsening of c′ phase (rafting) in Ni-based single-crystal superalloys during tensile creep at 1273 K is simulated by the phase-field (PF) method. A number of PF simulations are performed with various values of PF model parameters. The obtained results are used to train a neural network (NN) to enable fast and accurate prediction of the rafting time (traft ) from the values of model parameters. Material parameters of first-, second-, third-, and fourth-generation superalloys are estimated from their chemical compositions for predicting traft using the trained NN. The traft of several practical superalloys are predicted in the tensile stress range of 130–190 MPa. The NN prediction results show that traft tends to be longer along with the order of alloy generation. Furthermore, creep rupture time (trup ) of practical superalloys is estimated based on the Larson–Miller parameter method. It is found that there is a positive correlation between traft and trup , and the correlation becomes stronger with increasing the magnitude of external tensile stress. Keywords







Ni-based single-crystal superalloy Creep Phase-field method Neural network




Rafting

Y. Matsuoka (&)
Y. Tsukada
T. Koyama Department of Materials Design Innovation Engineering, Graduate School of Engineering, Nagoya University, Furo-Cho, Chikusa-Ku, Nagoya, 464-8603, Japan e-mail: [email protected] Y. Tsukada e-mail: [email protected] T. Koyama e-mail: [email protected]




Ni-based single-crystal superalloys are widely used as blade materials in aero-engines and land-based gas turbines [1]. Driven by industrial demands for the high-temperature alloys with superior properties, superalloys have been developed and repeatedly modified for many decades [2–6]. Since superalloys usually contain as many as ten alloying elements and the alloy design space is extremely large, some guidelines and strategies are necessary to enable the efficient development of a new alloy which meets industrial demands for cost, creep life, oxidation resistance, and other important operating needs [1, 7–10]. There have been efforts to predict certain properties of Ni-based superalloys from the chemical composition [7–10]. Recently, models have been proposed [11–13] for predicting creep behavior, which is one of the most important properties of superalloys. To enable a physically based quantitative prediction of creep, information about the microstructure of superalloys and its temporal evolution during high-temperature creep needs to be incorporated into the prediction model. The typical microstructure of Ni-based single-crystal superalloys is composed of c and c′ phases: cuboidal c′ precipitates coherently in the c phase with face-centered cubic lattice and aligns along the crystallographic directions of the c phase. The c′ phase coarsens toward directions perpendicular to the [001] tensile stress axis and forms plate-like microstructure during high-temperature creep. This phenomenon is called rafting, which is presumed to affect creep properties of superalloys [1, 14–21]. The rafting is caused by the atomic diffusion of constituent elements, and its driving force is closely related to the cphase plasticity introduced during creep [19]. Hence, the time to rafting (traft) depends on both various material parameters and creep conditions. This is why the experimental cost for investigating the rafting in various superalloys is extremely high and the relationship between the rafting and creep properties has not been fully revealed.

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_28

292

Prediction of Rafting Kinetics of Practical Ni-Based Single …

293

For simulating the rafting in superalloys, a phase-field (PF) model was developed with considering both c′ morphological change and c-phase plasticity during creep simultaneously [14]. Three-dimensional (3D) simulations based on the model predicted the rafting kinetics of a practical Ni-based single-crystal superalloy [15, 16]. The model is capable of simulating the rafting kinetics of a wide variety of practical alloys by changing the values of model parameters. However, the 3D simulation takes months depending on the values of model parameters. Hence, it is still difficult to overview the relationship between the PF model parameters and the rafting kinetics through exhaustive simulation study. Most recently, it has been shown that the fast and accurate prediction of traft can be achieved by combining PF simulations and neural network (NN) analysis [17]. The purpose of this study is to predict traft and explore the relationship between traft and the time to rupture (trup) of practical superalloys. We first perform PF simulations to generate a dataset of the relationship between the PF model parameters and traft. Then, the generated dataset is used for training a NN, which enables us to predict traft instantaneously when the values of PF model parameters are given. Second, using the trained NN, we consider predicting traft of several practical alloys, whose PF model parameters are estimated from the chemical composition.

hð/i Þ ¼

gð/i Þ ¼

4  X i¼1

4  X i¼1

/3i ð10  15/i þ 6/2i Þ



4 X 4 X  /2i ð1  /2i Þ þ a /2i /2j

ð5Þ

ð6Þ

i¼1 j6¼i

0

where GðcÞ and Gðc Þ are the Gibbs energy densities of the c 0 and c′ phases, respectively. The GðcÞ and Gðc Þ are given by 0 0 GðcÞ ¼ W ðcÞ f 2 and Gðc Þ ¼ W ðc Þ ð1  f Þ2 , where W ðcÞ and 0 0 W ðc Þ are Gibbs energy coefficients. The f ðcÞ and f ðc Þ are the local values of the field variable f for c and c′ phases, respectively [23]. The w is the height of double-well potential, j/ is the gradient energy coefficient, and Eel is the elastic strain energy. The Eel is given by

r

 n o  1 Cijkl ðr; tÞ eij ðr; tÞ  e0ij ðr; tÞ ekl ðr; tÞ  e0kl ðr; tÞ  rappl eij dr ij
2

ð7Þ

Phase-Field Model The PF model developed by Tsukada et al. [14–16] is employed. The local volume fraction of the c′ phase f ðr; tÞ and the structural order parameters /i ðr; tÞ with i = 1, 2, 3, 4, which distinguish four different ordered domains of the c′ phase, are selected as field variables. Here, r and t denote position coordinate and time, respectively. The relation between f ðr; tÞ and the concentration field cðr; tÞ is described as follows:

Here, Cijkl is the elastic constant, eij is the total strain, e0ij is is the external stress, and
eij is the the eigenstrain, rappl ij uniform macroscopic strain. The total strain is calculated by solving the mechanical equilibrium equation using an iterative perturbation approach [24]. The eigenstrain is defined by e0ij ðr; tÞ ¼ e0 dij hð/i Þ þ epij ðr; tÞ þ ecij ðr; tÞ 0

ð1Þ

0

where 0 cðcÞ and 0 cðc Þ are the equilibrium concentrations in the c and c′ phases, respectively. The temporal evolution of the field variables is calculated by solving the Cahn–Hilliard and Allen–Cahn equations [22]: @f ðr; tÞ dGtotal ¼ Mf r2 @t df ðr; tÞ

ð3Þ

ð4Þ

Eel ¼

Phase-Field Simulation

ði ¼ 1; 2; 3; 4Þ

Here, Gtotal is the total free energy of microstructure, Mf is the c/c′ diffusion mobility, and L is the structural relaxation coefficient. The total free energy is defined by 2 3 ðcÞ ðcÞ ðc0 Þ ðc0 Þ Z f1  hð/i ÞgG ðf Þ þ hð/i ÞG ðf Þ 6 7 4 7dr þ Eel Gtotal ¼ 6 j/ X 4 5 2 þ wgð/i Þ þ ðr/i Þ r 2 i¼1

Z 

Calculation Methods

cðr; tÞ  0 cðcÞ f ðr; tÞ ¼ 0 ðc0 Þ 0 ðcÞ c  c

@/i ðr; tÞ dGtotal ¼ L @t d/i ðr; tÞ

ð2Þ

e0 ¼ ðaðc Þ  aðcÞ Þ=aðcÞ

ð8Þ ð9Þ

0

where e0 is the c/c′ lattice misfit, aðcÞ and aðc Þ are lattice parameters of the c and c′ phases, respectively, and dij is the Kronecker delta function. Note that in Eq. (8), the inelastic strain is divided into the plastic term epij and the creep term ecij ; the epij develops time-independently at a position where the von Mises yield criterion is exceeded, while the ecij develops time-dependently with obeying the five-power-law creep equation. Both epij and ecij are assumed to be confined to

294

Y. Matsuoka et al.

the c phase. The evolution of epij is given by the following equation [25]: @epij ðrp ; tÞ dEshear ¼ Kijkl p @t dekl ðrp ; tÞ

ð10Þ

where rp represents a position where the von Mises yield criterion is exceeded (the equivalent stress (
r) is higher than c the yield stress (rys )), and Kijkl is the kinetic coefficient. The Eshear is the shear strain energy, which is given by Z  n o 1 Cijkl ðr; tÞ eij ðr; tÞ  e0ij ðr; tÞ Eshear ¼ 2 ð11Þ r i   ekl ðr; tÞ  e0kl ðr; tÞ  sappl eij dr ij
Here, eij and sij are the deviatoric strain and deviatoric stress, respectively. The temporal evolution of ecij is given by @ecij ðr; tÞ 3 4 r ðr; tÞsij ðr; tÞ ¼ C
2 @t

ð12Þ

phase (rcys ), the creep coefficient (C), and the external tensile stress (rappl). We generated a total of 10,105 simulation-based datasets, which were used to train a NN, as described in Sect. 2.2. Note that the value of C was changed while the value of the c/c′ diffusion mobility (Mf) was fixed; this is equivalent to the case where the simulations are performed with changing the ratio of Mf to C, because the units of both Mf and C contain the dimension of time.

Neural Network Analysis A feedforward NN model is used to predict traft from PF model parameters. The NN structure consists of one input layer with four nodes x ¼ ðx1 ; x2 ; x3 ; x4 ÞT ¼ ðe0 ; rcys ; C; rappl ÞT , two hidden layers, and one output layer with one node y ¼ logðtraft Þ (logarithm of traft). Both input and output parameters are normalized between 0 and 1 using minimum and maximum values of the each parameter [17]. ðlÞ

The output of a node in the l-th layer si is calculated by


is the equivalent stress. where C is the creep coefficient and r

Simulation Conditions Morphological evolution of the c′ phase was simulated at 1273 K under the external tensile stress along the [001] direction. 3D simulations were performed with 32  32  32 computational cells for the domain of 256  256  256 nm3. A single cuboidal c′ phase was placed at the center of the computational domain. Considering the periodic array of the c′ particles in practical superalloys, the periodic boundary condition was assumed in the simulations. The structural relaxation coefficient L was set large enough to ensure that the microstructure evolution was diffusion controlled. Model parameters used in this study are listed in Table 1. Simulations were performed with various values of the lattice misfit (e0), the yield stress of c

Table 1 Model parameters used in phase-field simulations [15]

ðlÞ si

¼g

ðl1Þ m X

j¼1

! ðlÞ ðl1Þ wij sj

ðlÞ þ hi

ð13Þ

where mðl1Þ is the number of nodes in the (l − 1)-th layer, ðlÞ

ðlÞ

wij is the weight, and hi is the bias. Note that l = 0 corð0Þ

responds to the input layer (si ¼ xi ), l = 1, 2 the two hidden layers, l = 3 the output layer. The gðxÞ is the activation function; the hyperbolic tangent function (gðxÞ ¼ tanhðxÞ) was used for l = 1, 2, and the linear function (gðxÞ ¼ x) was used for l = 3. The NN was trained using 7070 data randomly chosen out of the total of 10,105 PF simulation data. The remaining 3035 data were used as test data to evaluate the generalization capability of the trained NN. The error function is defined as follows:

Temperature, T/K

1273

c/c′ diffusion mobility, Mf/J–1 mol m2 s–1

2.15  10–20

Gibbs energy coefficient, W/J m–3

W(c) = 1.33  108, W(c′) = 1.56  108 –3

Height of double-well potential, w/J m

1.07  107

Gradient energy coefficient, j//J m–1

3.41  10–10

Elastic constant, Cijkl/GPa

C11 = 204.9, C12 = 150.8, C44 = 94.0

ðcÞ

ðc0 Þ C11

ðcÞ

= 251.6,

ðc0 Þ C12

ðcÞ

ðc0 Þ

= 194.5, C44 = 95.0

Lattice misfit, e0

–0.005 to –0.001

Yield stress of c phase, rcys /MPa

90–150

Creep coefficient, C/MPa–5 s–1

1.99  10–19 to 7.97  10–17

External tensile stress, rappl/MPa

130–190

Prediction of Rafting Kinetics of Practical Ni-Based Single …

EðwÞ ¼

7070 hn  i2  o 1X PFðnÞ log traft y xðnÞ ; w N 2 n¼1

w¼



ð1Þ ð2Þ ð3Þ ð1Þ ð2Þ ð3Þ wij ; wij ; wij ; hi ; hi ; hi

T

295

ð14Þ

Estimation of Material Parameters of Practical Alloys To simulate traft of practical superalloys, we consider estimating the values of material parameters ðe0 ; rcys ; C; Mf Þ from the chemical composition xm (in atomic fraction). The composition of component m in the c and c′ phases (xcm and 0 xcm ) is calculated by the CALPHAD method using Thermotech Ni-data (ver. 8) [27]. The lattice misfit (e0) is estimated from the lattice parameters of the c and c′ phases, aðcÞ 0 and aðc Þ . The lattice parameters are assumed to obey the Vegard’s law: X aðcÞ ¼ 0:3587 þ Acm xcm ðnmÞ ð16Þ m

0

X m

0

0

Acm xcm ðnmÞ

ð17Þ 0

The constants and coefficients (Acm and Acm ) are determined based on previous studies [1, 28]. The yield stress of c phase (rcys ) is formulated as rcys ¼ 30:0 þ

X m

Bm xcm ðMPaÞ

ð18Þ

The constant and coefficients (Bm ) are determined based on previous studies [4, 29, 30]. According to Sherby et al. [31], the creep rate is given as e_ c ¼ AD

nc o3:5 SF ðrappl Þ5 Gb

ð19Þ

where A is a constant, D is the diffusion coefficient, cSF is the stacking fault energy, G is the shear modulus, and b is the magnitude of burgers vector. Comparing Eqs. (12) and (19), the following relation is derived: C / Dc3:5 SF

1 D ¼ P xc

ð15Þ

where the symbol with subscript N represents the normalized value. Adam [26] was adopted to optimize the values of weights and biases; the batch size was set to 32, and the epoch size was set to 20,000.

aðc Þ ¼ 0:3616 þ

alloys can be estimated from the values of D and cSF . The value of D is estimated by the following equation [32]:

ð20Þ

As the value of the creep coefficient (C) for the CMSX-4 alloy has been estimated [15], the values of C for other

ð21Þ

m

m

Dm

where Dm is the diffusion coefficient of component m [32– 35]. According to Harada et al. [6], the stacking fault energy (cSF ) is formulated as X cSF  103 ¼ 12:914 þ Sm xcm Gb m

ð22Þ

The constant and coefficients (Sm ) are determined based on previous studies [6, 36]. The diffusion mobility (Mf ) is estimated by the following equation [15]: ðmÞ

Mf

Mmm ¼ 0 2 xcm  xcm

ð23Þ

ðmÞ

where Mf is the diffusion mobility of solute component m and Mmm is the atomic diffusion mobility of solute component m. The Mmm is calculated using multicomponent diffusion mobility database for Ni-based superalloys [37]. ðmÞ

According to Eq. (23), values of Mf are estimated for all components m in each alloy. Assuming that the microstructure evolution is controlled by the atomic diffusion of the element with the smallest diffusion mobility, the ðmÞ

smallest value of Mf is used for Mf . Values of coefficients used for estimating material parameters are listed in Table 2.

Results PF Simulations Figure 1 shows PF simulation results of the morphological change of c′ phase and the evolution of the equivalent inelastic strain field of the CMSX-4 alloy during creep at 1273 K under the 160 MPa tensile stress along the [001] direction. It is seen that the cuboidal c′ phase gradually changes into a plate-like morphology, connects with adjacent c′ phases, and finally forms a rafted structure. The inelastic strain accumulates in the c channels normal to the external tensile stress axis. This is attributed to the fact that in the horizontal c channels, the internal stress derived from the c/c′ misfit is intensified by applying the external tensile stress along the [001] direction. As discussed in previous studies [14, 18, 19], the inelastic strain introduced in c channels causes the anisotropic relaxation of the c/c′ misfit,

296 Table 2 Values of coefficients used for estimating material parameters at 1273 K

Y. Matsuoka et al. Element

Acm (nm)

0

Acm (nm)

Bm (MPa)

Dm (10−17 m2 s−1)

Sm

Al

0.0179

–

100

297

−23.8

Co

0.0023

0.0000

100

48.3

−11.3

Cr

0.0110

−0.0004

800

72.1

−22.4

Mo

0.0478

0.0208

700

33.9

−35.6

Nb

0.0670

0.0433

–

33.0

−27.1

Ni

–

–

–

42.2

–

Re

0.0441

0.0262

700

2.82

−30.0

Ru

0.0334

0.0129

950

8.33

−20.0

Ta

0.0700

0.0500

850

100

−27.1

Ti

0.0422

0.0249

800

213

−82.2

W

0.0444

0.0194

750

11.8

−72.0

Hf

0.1040

0.0762

–

–

–

Fig. 1 Simulation results of time evolution of c′ morphology and inelastic strain field of the CMSX-4 alloys during creep at 1273 K under the external tensile stress of 160 MPa along the [001] direction: a t = 0 h, b t = 6.3 h, c t = 7.6 h, and d t = 9.4 h

and the resultant diffusion potential difference between horizontal and vertical c channels leads to the morphological change of c′ phase (rafting) [38, 39]. In Fig. 2, we compare the macroscopic creep rate–time curves obtained by the PF simulation and experiment [40]. It is seen that the experimental result for the CMSX-4 alloy in the primary creep stage is well reproduced by the simulation except that a sharp and temporary increase in the creep rate is observed in the simulation at about t = 7.6 h. The temporary increase in the creep rate is derived from the drastic change in the internal stress field when the neighboring c′

phases connect with each other to form the rafted structure [14–16]. Although the increase in the macroscopic creep rate caused by the rafting has been reported in some alloys [41], it is not clearly observed in the CMSX-4 alloy as shown in Fig. 2. This is due to the fact that in the CMSX-4 alloy, all the c′ particles do not raft simultaneously because of the inhomogeneity of the size and shape of the c′ particles [16]. In this study, to determine traft for a number of simulation results in a consistent way, the traft is defined as the time when the macroscopic creep rate temporarily increases. The formation of the rafted structure was observed in all the

Prediction of Rafting Kinetics of Practical Ni-Based Single …

Fig. 2 Creep rate–time curve of the CMSX-4 alloy during creep at 1273 K under the external tensile stress of 160 MPa along the [001] direction. Solid symbols represent phase-field simulation data, and open symbols represent experimental data [40]

10,105 simulations with different values of model parameters ðe0 ; rcys ; C; rappl Þ, and the dataset of the relationship between the values of model parameters and traft was obtained.

Neural Network Prediction The dataset generated by PF simulations was used to train the NN that predicts traft from the values of model parameters ðe0 ; rcys ; C; rappl Þ. The time to rafting simulated by the PF NN PF method (traft ) and that predicted by the trained NN (traft ) PF NN are compared in Fig. 3. We see that the values of traft and traft are close to each other in the wide range of traft , which holds true not only for the training data but also for the test data. This indicates that the generalization capability of the trained NN is high. To evaluate the prediction accuracy quantitatively, the root-mean-square error (ERMS ) for the test data was calculated by vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u 3035 hn o i2 u 1 X PFðnÞ ERMS ¼ t logðtraft Þ yðxðnÞ ; w Þ ð24Þ N 3035 n¼1 where w is the optimum parameter vector. The value of ERMS is 0.0479, which is equivalent to the error of ±4.9% for traft. Therefore, the prediction error of the trained NN is low.

297

Fig. 3 Comparison between the time to rafting simulated by the PF phase-field method (traft ) and that predicted by the trained neural NN network (traft ). Diamond symbols represent training data and plus symbols represent test data

Time to Rafting and Time to Rupture of Practical Superalloys Practical superalloys investigated in this study are listed in Table 3 along with their compositions, estimated material parameters, and references to Larson–Miller parameters (LMP) used for the estimation of creep rupture life. Correlations between the estimated material parameters  e0 ; rcys ; C; Mf are shown in Fig. 4. We see from Fig. 4a that the value of rcys has a correlation to the value of e0 ; the alloy with a larger negative value of e0 tends to have a higher value of rcys . In addition, as for first-generation alloys, the magnitude of e0 is small and the value of rcys is low. On the other hand, it is found from Fig. 4b that the value of C does not differ depending on the alloys and has no clear correlation to the value of e0 . It is apparent from Fig. 4c that the values of Mf for first-generation alloys are significantly larger than those for second-, third-, and fourth-generation alloys. Figure 5a, b show the time to rafting predicted by the NN trained NN (traft ) and the time to rupture estimated by the LMP method (trup ), respectively. Note that experimental data of trup under low-stress conditions are lacking in some alloys. It is shown that traft and trup become shorter with increasing the magnitude of rappl , and they tend to be longer in accordance with the order of alloy generation. Additionally, it is found that traft is notably short under external

298

Y. Matsuoka et al.

Table 3 Chemical composition and estimated material parameters of practical superalloys at 1273 K Alloy

Composition (at.%) Al

Co

Cr

Mo

Re

Ta

Ti

W

Others

Ni

Alloy Gen.

e0 (10−3)

CMSX-2

12.4

4.7

9.2

0.4

–

2.0

1.2

2.6

–

Bal.

1st

−1.6

CMSX-4

12.5

10.3

7.6

0.4

1.0

2.0

1.3

2.0

–

Bal.

2nd

−2.3

rcys (MPa)

95.2 120

C (10−18 MPa−5 s−1)

Mf (10−20J−1 mol m2 s−1)

12.0

47.0

6.64

2.15

Ref. to LMP

[4] [42]

CMSX-8

13.0

10.5

6.4

0.4

0.5

2.7

0.9

2.7

0.1 Hf

Bal.

2nd

−2.3

105

CMSX-10

13.2

2.1

2.4

0.3

2.0

2.8

0.3

1.7

0.1 Nb

Bal.

3rd

−2.4

106

NKH71

11.1

–

14.0

0.3

–

1.9

1.5

2.5

–

Bal.

1st

−2.1

NKH72

10.9

3.9

14.1

0.3

–

1.9

1.6

2.5

–

Bal.

1st

−1.7

TMS-26

11.9

8.7

6.8

1.2

–

2.6

–

3.7

–

Bal.

1st

−3.1

TMS-75

13.7

12.6

3.6

1.3

1.7

2.0

–

2.0

–

Bal.

3rd

−4.2

109

4.40

[43]

TMS-82+

12.2

8.2

5.8

1.2

0.8

2.2

0.6

2.9

–

Bal.

2nd

−4.0

115

6.07

5.92

[42]

TMS-82 15Co

11.0

16.0

4.9

1.0

0.8

1.9

2.7

2.6

–

Bal.

2nd

−3.1

116

1.03

5.90

[45]

2.87 61.5

5.79

[2]

2.94

[43]

99.6

9.67

45.2

[44]

98.2

4.77

43.0

[44]

5.02

34.4

[4]

112

11.4

TMS-82 0Co

12.2

–

5.5

1.2

0.8

2.2

0.6

2.9

–

Bal.

2nd

−4.2

117

5.20

[45]

TMS-82 + Re

12.3

8.1

5.5

1.2

1.2

2.0

0.7

2.9

–

Bal.

3rd

−4.4

117

6.69

4.92

[46]

TMS-82 + Ru

12.1

8.1

5.4

1.2

0.8

2.0

0.7

2.9

1.2 Ru

Bal.

4th

−4.5

126

7.78

6.91

[46]

ReneN5

13.9

8.2

8.1

1.3

1.0

2.3

–

1.6

0.1 Hf

Bal.

2nd

−4.3

107

6.39

4.76

[47]

DD6

12.9

9.5

5.2

1.3

0.7

2.6

–

2.7

0.3 Nb

Bal.

2nd

−3.6

113

6.60

5.67

[5]

SRR99

12.1

5.0

9.1

–

–

1.0

2.7

3.2

–

Bal.

1st

−2.2

97.4

4.67

46.1

[3]

RR2000

11.1

13.9

10.5

1.7

–

–

4.6

–

–

Bal.

1st

−2.4

90.7

7.40

45.4

[3]

stresses higher than 160 MPa in some first-generation alloys. NN Figure 6 shows the correlation between traft and trup under the external tensile stresses of 130, 160, and 190 MPa. It is shown that there is a positive correlation between traft and trup , and the correlation becomes stronger with increasing the magnitude of rappl . The sample correlation coefficient between logðtraft Þ and logðtrup Þ is 0.41, 0.61, and 0.85 for the external tensile stress of 130, 160, and 190 MPa, respectively.

Discussion Material Parameters and Time to Rafting The correlation between e0 and rcys shown in Fig. 4a should be explained in terms of the size of solute atoms in the c phase. The coefficients of the Vegard’s law have a correlation to the size of solute atoms, and large solute atoms increase the lattice parameter [1]. On the other hand, from the viewpoint of the solution hardening [48], large solute atoms in the c phase lead to an increase in rcys . Therefore, if large solute atoms are selectively partitioned into the c

26.4

phase, a large negative value of e0 and a high value of rcys would be achieved. In Fig. 4a, it is also worth noting that first-generation alloys tend to have small negative values of e0 and low values of rcys ; these parameter conditions are assumed to shorten traft based on previous studies [17]. It is shown in Fig. 4c that first-generation alloys have large values of Mf compared to second-, third-, and fourth-generation alloys. This is due to the fact that Re has been added to post-first-generation alloys. The diffusion of Re in Ni solid solution is known to be slower than that of other alloying elements [32]. In our estimation of Mf (see ðmÞ

Sect. 2.3), Re had the smallest value of Mf in the alloying elements in the Re-containing alloys, while Co or W was the slowest diffusion species in the Re-free alloys. Since the value of Mf determines the migration rate of the c/c′ interface (see Eq. 2), large Mf values shorten traft . This is another reason why traft is short in first-generation alloys.

Time to Rafting and Alloy Generation As described in Sect. 4.1, first-generation alloys have the values of parameters (e0 , rcys , and Mf ) that reduce traft .

Prediction of Rafting Kinetics of Practical Ni-Based Single …

299

Fig. 5 Time to rafting at 1273 K predicted by the trained neural network (a) and time to rupture at 1273 K estimated by the Larson– Miller parameter method (b)

Fig. 4 Correlations between the estimated material parameters of practical superalloys at 1273 K: a yield stress of c phase (rcys ) versus lattice misfit (e0 ), b creep coefficient (C) versus lattice misfit (e0 ), and c diffusion mobility (Mf ) versus lattice misfit (e0 ). The generations of superalloys are distinguished by different symbols

Actually, the predicted traft of first-generation alloys tend to be short compared to second-, third-, and fourth-generation alloys as shown in Figs. 5a and 6. It should be noted that even though guidelines for the control of traft have not been reported, traft tends to be longer in accordance with the order of alloy generation. Another finding is that in some first-generation alloys, traft drastically reduces under stresses higher than 160 MPa (see Fig. 5a). Previous studies [17] have shown that the value of rcys has significant effect on traft when the ratio of rappl to rcys is relatively high. Since the values of rcys are low in first-generation alloys (see Fig. 4a), additional calculations were performed using the trained NN to confirm the effect of NN rcys on traft . Figure 7 shows traft predicted by the trained NN with setting the lower limit of rcys as 110 MPa; all the alloys with the values of rcys  110 MPa are assumed to have the

300

Y. Matsuoka et al.

Fig. 6 Correlation between time to rupture and time to rafting at 1273 K under the external tensile stress of a 130 MPa, b 160 MPa, and c 190 MPa. The generations of superalloys are distinguished by different symbols

value of rcys = 110 MPa in the calculations. It is seen that the drastic reduction in traft under external stresses higher than 160 MPa is significantly suppressed in first-generation alloys (compare Figs. 5a and 7). This result indicates that the characteristic stress dependence of traft in first-generation alloys is attributed to the low values of rcys .

Time to Rafting and Time to Rupture It has been reported that under high-stress conditions, the formation of rafted structure decreases the creep resistance because it increases the c-channel width and lowers the Orowan stress [13, 40]. We see from Fig. 6c that there is a strong positive correlation between traft and trup under 190 MPa. This result is consistent with the idea that the rafting decreases the creep strength under high-stress conditions. On the other hand, it is seen from Fig. 6 that the

correlation between traft and trup is reduced with decreasing rappl . This indicates that additional descriptors might be needed other than traft for predicting trup under low-stress conditions. It has been reported that the rafted structure increases the creep strength until it collapses [21]. In other words, the formation of complete and stable rafted structure is responsible for increasing the creep strength under low-stress conditions [49, 50]. It is assumed that the dislocation network formed at the c/c′ interface of the rafted structure acts as a barrier for the dislocation movement [51]. Hence, the stability of the rafted structure could be a descriptor for predicting trup . Further study on the correlations between trup , traft , and the stability of the rafted structure would be required to predict the creep strength of superalloys in high-temperature and low-stress conditions.

Conclusions A number of PF simulations of c′ rafting in Ni-based single-crystal superalloys during tensile creep at 1273 K were performed with various values of PF model parameters. The obtained simulation data were used for training a NN, which enabled fast and accurate prediction of traft from the values of PF model parameters. We estimated material parameters of several practical superalloys from their chemical compositions and predicted traft in the tensile stress range of rappl = 130−190 MPa using the trained NN. Obtained results are as follows.

Fig. 7 Time to rafting at 1273 K predicted by the trained neural network. All the alloys with values of rcys  110 MPa are assumed to have the value of rcys = 110 MPa in the calculations

1. First-generation alloys tend to have small negative values of e0 , low values of rcys , and large values of Mf . These conditions make traft of first-generation alloys shorter than those of second-, third-, and fourth-generation alloys. The predicted traft of practical superalloys tend to be longer in accordance with the order of alloy generation.

Prediction of Rafting Kinetics of Practical Ni-Based Single …

2. In first-generation alloys, traft drastically reduces under external stresses higher than 160 MPa. This is attributable to low values of rcys , which increases the amount of time-independent plastic strain introduced in the c phase during creep. 3. There is a positive correlation between traft and trup , and the correlation becomes stronger with increasing the magnitude of rappl . The result is consistent with the idea that the rafted structure decreases the creep strength under high-stress conditions. In addition, the result shows that additional descriptors other than traft might be required to predict trup in high-temperature and low-stress conditions.

301

11.

12.

13.

14. Acknowledgements This work was supported by JST PRESTO (Grant Number JPMJPR15NB). 15.

References 1. Reed, RC (2006) The superalloys - fundamentals and applications. Cambridge University Press, Cambridge. 2. Wahl JB, Harris K (2012) New single crystal superalloys, CMSX®-7 and CMSX®-8. In: Huron ES, Reed RC, Hardy MC, Mills MJ, Montero RE, Portella PD, Telesman J (eds) Superalloys 2012. The Minerals, Metals & Materials Society, Pittsburgh; John Wiley & Sons, Hoboken, p 179–188. 3. Ford DA, Arthey RP (1984) Development of single crystal alloys for specific engine applications. In: Gell M, Kortovich CS, Bricknell RH, Kent WB, Radavich JF (eds) Superalloys 1984. American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc., New York; The Metallurgical Society of AIME, Warrendale, p 115–124. 4. Ohta Y, Nakagawa YG, Shinichi O (1990) Fabrication and high temperature properties of single crystal component of advanced Ni-base superalloys. Tetsu to Hagane 76:940–947. https://doi.org/ 10.2355/tetsutohagane1955.76.6_940. 5. Li JR, Zhong ZG, Tang DZ, Liu SZ, Wei P, Wei PY, Wu ZT, Huang D, Han M (2000) A low-cost second generation single crystal superalloy DD6. In: Pollock TM, Kissinger RD, Bowman RR, Green KA, McLean M, Olson SL, Schirra JJ (eds) Superalloys 2000. The Minerals, Metals & Materials Society, Warrendale, p 777–783. 6. Harada H, Yamazaki M (1979) Alloy design for c′ precipitation hardening nickel-base superalloys containing Ti, Ta, and W. Tetsu to Hagane 65:1059–1068. https://doi.org/10.2355/tetsutohagane1955. 65.7_1059. 7. Reed RC, Tao T, Warnken N (2009) Alloys-by-design: application to nickel-based single crystal superalloys. Acta Mater 57:5898– 5913. https://doi.org/10.1016/j.actamat.2009.08.018. 8. Reed RC, Zhu Z, Sato A, Crudden DJ (2016) Isolation and testing of new single crystal superalloys using alloys-by-design method. Mater Sci Eng A 667:261–278. https://doi.org/10.1016/j.msea. 2016.04.089. 9. Watanabe R, Kuno T (1975) Alloy design of nickel-base precipitation hardened superalloys. Tetsu to Hagane 61:2274– 2294. https://doi.org/10.2355/tetsutohagane1955.61.9_2274. 10. Menou E, Rame J, Desgranges C, Ramstein G, Tancret F (2019) Computational design of a single crystal nickel-based superalloy

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

with improved specific creep endurance at high temperature. Comput Mater Sci 170:109194. https://doi.org/10.1016/j. commatsci.2019.109194. Oruganti R, Shukla A, Nalawade S, Sarkar S, Sivakumar KGV, Vishwanath T, Sondhi S, Wessman A, Wei D, Powell A, Bain K, Schaeffer J, Peck A, Arnett M, Shastry G, Mastromatteo F (2019) A microstructure-based model for creep of gamma prime strengthened nickel-based superalloys. J Eng Mater Technol 141:1–5. https://doi.org/10.1115/1.4040554. Zhu Z, Basoalto H, Warnken N, Reed RC (2012) A model for the creep deformation behaviour of nickel-based single crystal superalloys. Acta Mater 60:4888–4900. https://doi.org/10.1016/j. actamat.2012.05.023. Fedelich B, Künecke G, Epishin A, Link T, Portella P (2009) Constitutive modelling of creep degradation due to rafting in single-crystalline Ni-base superalloys. Mater Sci Eng A 510– 511:273–277. https://doi.org/10.1016/j.msea.2008.04.089. Tsukada Y, Koyama T, Kubota F, Murata Y, Kondo Y (2017) Phase-field simulation of rafting kinetics in a nickel-based single crystal superalloy. Intermetallics 85:187–196. https://doi.org/10. 1016/j.intermet.2017.02.017. Tsukada Y, Koyama T, Murata Y, Miura N, Kondo Y (2014) Estimation of c/c′ diffusion mobility and three-dimensional phase-field simulation of rafting in a commercial nickel-based superalloy. Comput Mater Sci 83:371–374. https://doi.org/10. 1016/j.commatsci.2013.11.028. Tsukada Y, Murata Y, Koyama T, Miura N, Kondo Y (2011) Creep deformation and rafting in nickel-based superalloys simulated by the phase-field method using classical flow and creep theories. Acta Mater 59:6378–6386. https://doi.org/10.1016/j. actamat.2011.06.050. Matsuoka Y, Tsukada Y, Koyama T (2019) Phase-field simulation and machine learning for predicting rafting kinetics in Ni-based superalloys. In: Shingledecker J, Takeyama M (eds) Joint EPRI– 123HiMAT international conference on advances in high temperature materials. ASM International, Materials Park, p 496–505. Matan N, Cox DC, Rae CMF, Reed RC (1999) On the kinetics of rafting in CMSX-4 superalloy single crystals. Acta Mater 47:2031–2045. https://doi.org/10.1016/s1359-6454(99)00093-2. Ichitsubo T, Koumoto D, Hirao M, Tanaka K, Osawa M, Yokokawa T, Harada H (2003) Rafting mechanism for Ni-base superalloy under external stress: elastic or elastic-plastic phenomena? Acta Mater 51:4033–4044. https://doi.org/10.1016/s1359-6454(03)00224-6. Pollock TM, Argon AS (1994) Directional coarsening in nickel-base single crystals with high volume fractions of coherent precipitates. Acta Metall Mater 42:1859–1874. https://doi.org/10. 1016/0956-7151(94)90011-6. Kubo Y, Miura N, Kondo Y, Murata Y (2011) Creep properties of non-Re single crystal nickel-based superalloy, containing 12%Cr at 1273K. CAMP-ISIJ 24:337. Chen LQ (2002) Phase-field models for microstructure evolution. Annu Rev Mater Sci 32:113–140. https://doi.org/10.1146/annurev. matsci.32.112001.132041. Kim SG, Kim WT, Suzuki T (1999) Phase-field model for binary alloys. Phys Rev E 60:7186–7197. https://doi.org/10.1103/ physreve.60.7186. Hu SY, Chen LQ (2001) A phase-field model for evolving microstructures with strong elastic inhomogeneity. Acta Mater 49:1879–1890. https://doi.org/10.1016/s1359-6454(01)00118-5. Guo XH, Shi SQ, Ma XQ (2005) Elastoplastic phase field model for microstructure evolution. Appl Phys Lett 87:1–3. https://doi. org/10.1063/1.2138358. Kingma DP, Ba JL (2014) Adam: a method for stochastic optimization. Paper presented at ICLR 2015, San Diego, California, 7–9 May 2015.

302 27. Saunders N, Fahrmann M, Small CJ (2000) The application of CALPHAD calculations to Ni-based superalloys. In: Pollock TM, Kissinger RD, Bowman RR, Green KA, McLean M, Olson SL, Schirra JJ (eds) Superalloys 2000. The Minerals, Metals & Materials Society, Warrendale, p 803–811. 28. Biermann H, Strehler M, Mughrabi H (1996) High-temperature measurements of lattice parameters and internal stresses of a creep-deformed monocrystalline nickel-base superalloy. Metall Mater Trans A 27:1003–1014. https://doi.org/10.1007/bf02649768. 29. Yeh AC, Tin S (2005) Effects of Ru and Re additions on the high temperature flow stresses of Ni-base single crystal superalloys. Scr Mater 52:519–524. https://doi.org/10.1016/j.scriptamat.2004.10. 039. 30. Jenkins WD, Digges TG (1952) Effect of temperature on the tensile properties of high-purity nickel. J Res Natl Bur Stand 48:313–321. https://doi.org/10.6028/jres.048.039. 31. Bonesteel RM, Sherby OD (1966) Influence of diffusivity, elastic modulus, and stacking fault energy on the high temperature creep behavior of alpha brasses. Acta Metall 14:385–391. https://doi.org/ 10.1016/0001-6160(66)90096-4. 32. Pröbstle M, Neumeier S, Feldner P, Rettig R, Helmer HE, Singer RF, Göken M (2016) Improved creep strength of nickel-base superalloys by optimized c/c′ partitioning behavior of solid solution strengthening elements. Mater Sci Eng A 676:411– 420. https://doi.org/10.1016/j.msea.2016.08.121. 33. Karunaratne MSA, Carter P, Reed RC (2000) Interdiffusion in the face-centred cubic phase of the Ni-Re, Ni-Ta and Ni-W systems between 900 and 1300°C. Mater Sci Eng A 281:229–233. https:// doi.org/10.1016/s0921-5093(99)00705-4. 34. Boettinger WJ, Kattner UR, Coriell SR, Chang YA, Mueller BA (1995) Development of multicomponent solidification micromodels using a thermodynamic phase diagram data base. Paper presented at Seventh Conference in a Series on Modeling, Casting, and Welding Processes, London, 10–15 September 1995. 35. Karunaratne MSA, Reed RC (2003) Interdiffusion of the platinum-group metals in nickel at elevated temperatures. Acta Mater 51:2905–2919. https://doi.org/10.1016/s1359-6454(03) 00105-8. 36. Zacherl CL, Shang SL, Kim DE, Wang Y, Liu ZK (2012) Effects of alloying elements on elastic, stacking fault, and diffusion properties of fcc Ni from first-principles: implications for tailoring the creep rate of Ni-base superalloys. In: Huron ES, Reed RC, Hardy MC, Mills MJ, Montero RE, Portella PD, Telesman J (eds) Superalloys 2012. The Minerals, Metals & Materials Sociey, Pittsburgh; John Wiley & Sons, Hoboken, p 455–461. 37. Campbell CE, Boettinger WJ, Kattner UR (2002) Development of a diffusion mobility database for Ni-base superalloys. Acta Mater 50:775–792. https://doi.org/10.1016/s1359-6454(01)00383-4. 38. Zhou N, Shen C, Mills MJ, Wang Y (2007) Phase field modeling of channel dislocation activity and c′ rafting in single crystal Ni-Al. Acta Mater 55:5369–5381. https://doi.org/10.1016/j.actamat.2007. 06.002.

Y. Matsuoka et al. 39. Zhou N, Shen C, Mills MJ, Wang Y (2008) Contributions from elastic inhomogeneity and from plasticity to c′ rafting in single-crystal Ni-Al. Acta Mater 56:6156–6173. https://doi.org/ 10.1016/j.actamat.2008.08.027. 40. Miura N, Kondo Y, Matsuo T (2003) Relation between creep rate during accelerating creep stage and gamma channel thickness in single crystal nickel-based superalloy, CMSX-4. Tetsu to Hagane 89:1240–1247. https://doi.org/10.2355/tetsutohagane1955.89.12_ 1240. 41. Izuno H, Yokokawa T, Koizumi Y, Odaka S, Harada H (2004) A Creep Constructive Equation of Ni-Base Superalloys Using Structural Parameters. J Japan Inst Metals 68:526–529. 42. National Institute for Materials Science, Toshiba corporation (2004) TMS-82+ . In: Second generation single crystal superalloy. http://sakimori.nims.go.jp/catalog/TMS-82+Catalogue-rev1.pdf. Accessed 17 January 2020. 43. National Institue for Materials Science, Kawasaki Heavy Industries, Ltd. (2006) TMS-75. In: Third generation nickel base single crystal superalloy. http://sakimori.nims.go.jp/catalog/TMS75TMD103-rev.pdf. Accessed 17 January 2020. 44. Murata Y, Hashizume R, Yoshinari A, Aoki N, Morinaga M, Fukui Y (2000) Alloying effects on surface stability and creep strength of nickel based single crystal superalloys containing 12mass% Cr. In: Pollock TM, Kissinger RD, Bowman RR, Green KA, McLean M, Olson SL, Schirra JJ (eds) Superalloys 2000. The Minerals, Metals & Materials Society, Warrendale, p 285–294. 45. Suzuki T, Yokokawa T, Kobayashi T, Koizumi Y, Harada H, Imai H (2007) Effect of cobalt on microstructural parameters and mechanical properties of Ni-base single crystal superalloys. J Japan Inst Metals 71:233–238. 46. Yokokawa T, Koizumi Y, Kobayashi T, Harada H (2006) Effect of Re and Ru additions to second generation nickel-base single crystal superalloy TMS-82 + . J Japan Inst Metals 70:670–673. 47. Koizumi Y, Yokokawa T, Harada H, Kobayashi T (2006) Database of creep property for nickel-base single crystal superalloys, RenéN4, RenéN5 and CMSX-4. J Japan Inst Metals 70:176– 179. 48. Fleischer RL (1963) Substitutional solution hardening. Acta Metall 11:203–209. https://doi.org/10.1016/0001-6160(63)90213-x. 49. Heckl A, Neumeier S, Göken M, Singer RF (2011) The effect of Re and Ru on c/c′ microstructure, c-solid solution strengthening and creep strength in nickel-base superalloys. Mater Sci Eng A 528:3435–3444. https://doi.org/10.1016/j.msea.2011.01.023. 50. Caron P, Khan T (1983) Improvement of creep strength in a nickel-base single-crystal superalloy by heat treatment. Mater Sci Eng 61:173–184. https://doi.org/10.1016/0025-5416(83)90199-4. 51. Zhang JX, Murakumo T, Koizumi Y, Kobayashi T, Harada H, Masaki S (2002) Interfacial dislocation networks strengthening a fourth-generation single-crystal TMS-138 superalloy. Metall Mater Trans A Phys Metall Mater Sci 33:3741–3746. https://doi.org/10. 1007/s11661-002-0246-7.

Creep Anisotropy in Single-Crystal Superalloy DD6 near the [001] Orientation Jian Yu, J. R. Li, S. Z. Liu, and Mei Han

Abstract

This paper has studied the anisotropy in the creep properties of DD6 alloy under different temperatures and applied stresses. The results reveal that the anisotropic creep of DD6 alloy near the [001] orientation is strongly influenced by the temperature in the range of 650–980 °C. The anisotropy in the primary creep strain and rupture lifetime at an intermediate temperature of 760 °C is dependent on the applied stress. Compared with the specimens oriented close to the [001]–[111] boundary, the specimens oriented close to the [001] direction and the [001]–[011] boundary exhibit lower primary creep strains and longer rupture lifetimes at intermediate temperatures and high applied stresses. With the increase in the testing temperature or the decrease in the applied stress, the anisotropic creep behavior of the alloy near the [001] orientation disappears. The mechanism of anisotropic creep is attributed to heterogeneous c′ precipitate deformation by {111} slip. Keywords

Single-crystal superalloy Orientation




Anisotropy




Creep



DD6

Introduction Ni-based single-crystal superalloys have been widely used in modern aeroengines and industrial gas turbines for the manufacture of turbine blades and vanes [1]. Due to the nature of single crystals, various properties including creep behavior exhibit a strong dependence on the orientation of J. Yu (&)
J. R. Li
S. Z. Liu
M. Han Science and Technology on Advanced High Temperature Structural Materials Laboratory, Beijing Institute of Aeronautical Materials, Beijing, 100095, China e-mail: [email protected]

the alloy, even if the orientation of the alloy is within 15° of the [001] orientation. In practice, the loading axis of single-crystal turbine blades deviates from the exact [001] orientation, which can be as high as 15° from [001] orientation [2]. The creep behavior of single-crystal superalloy is now known to be strongly temperature, stress, and orientation dependent. A comprehensive study on the influence of orientation on the creep behavior of the single crystal alloys MAR-M200 and MAR-M247 at intermediate temperature was conducted by MacKay and Maier [3]. They discovered long lifetimes for specimens oriented near [111] or [001], and short lifetimes for specimens oriented near [011], while in near the [001] orientation within 25°, the lifetimes for specimens close to the [001]–[011] boundary were significantly longer than those close to the [001]–[111] boundary. The creep anisotropy behavior at intermediate temperature (*750 °C) and high stresses has also been reported in other single crystal superalloys, such as CMSX-4 [4, 5] and PWA 1484 [6] alloys. In contrast, at high temperatures (980– 1050 °C), creep behavior is relatively insensitive to the orientation [7, 8]. Previous work has shown that the creep behavior of DD6 single-crystal superalloy is clearly dependent on the orientation at the intermediate temperature 760 °C and high stress 785 MPa near the [001] orientation [9]. In service, turbine blades are subjected to severe conditions of creep. The majority of loading in the core and root of cooled blade is borne at intermediate temperatures (*750 °C), while many of the outer regions of turbine blades experience high temperatures up to 1050 °C [10]. To take full advantage of the capabilities of single-crystal superalloys, it is necessary to further investigate the anisotropy in creep properties near [001] orientation under different temperatures and applied stresses. The work reported here is part of a wider study aimed at elucidating the deformation mechanisms operating in single-crystal superalloys under various conditions of temperature, stress, and orientation.

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_29

303

304

Materials and Experimental Procedure The material investigated in this work is DD6 single-crystal superalloy, which is a low-cost second generation singlecrystal superalloy with 2 wt% rhenium [11]. The singlecrystal bars were directionally solidified in a Bridgman furnace with the seeding technique. A standard heat treatment schedule was applied to all the specimen bars as follows: 1290 °C/1 h + 1300 °C/2 h + 1315 °C/4 h/AC + 1120 °C/4 h/AC + 870 °C/32 h/AC. The final microstructure consists of cuboidal c′ with a mean particle size of 0.45 lm occupying a volume fraction of approximately 65% in the c matrix. After the standard heat treatment, creep rupture specimens with a gauge length of 25 mm and a diameter of 5 mm were machined from the bars, and the surface of specimens was mechanically polished. The initial orientation of the tensile axes of the single crystals was determined by X-ray Laue diffraction using the grip ends of the specimens. The orientation of the specimens was characterized by h and q. The definitions of h and q are shown in Fig. 1. In this work, three types of initial orientation of specimens within 20° of the [001] orientation were investigated, namely close to the [001] pole of the standard stereographic triangle, close to the [001]–[011] boundary, and close to the [001]–[111] boundary. For the orientation of specimens close to the [001] pole with h less than 4.8°, q is undefined. For the orientation of specimens close to the [001]–[011] boundary with h from 16.7° to 19.7°, q is less than 16.4°. For the orientation of specimens close to the [001]–[111] boundary with h from 14.7° to 16.8°, q is greater than 33.1°. The creep tests were performed in air at temperatures ranging from 650 to 980 °C. To investigate the influence of stress on the anisotropy of creep, stresses from 450 to 785 MPa were applied to the specimens tested at 760 °C. All the specimens were creep tested using machines that applied a constant load. To study the resulting microstructural deformation in the bulk of the test pieces, the samples were cut parallel to the low-index crystallographic plans for transmission electron microscopy (TEM) analysis after the creep tests. The disks were thinned by mechanical abrasion and then electron-polished in a solution of 10% perchloric acid in ethyl alcohol. The resulting foils were examined in an FEI Tecnai G20 TEM operating at 200 kV.

Results and Discussion Anisotropic creep behavior of the alloy near the [001] orientation Table 1 is a summary of the creep data of DD6 alloy near the [001] orientation at temperature from 650 to 980 °C. The

J. Yu et al.

data listed in the table are the creep test conditions, specimen identification, initial orientation, life and elongation after the test, primary creep strain, and deviation of the rupture lifetimes. The elastic strains have been subtracted from these data in this work. To better understand the anisotropy in the creep behavior near the [001] orientation, the typical creep curves of DD6 alloy with orientations close to the [001] pole, [001]–[011] boundary, and [001]–[111] boundary under intermediate temperatures of 760 °C/785 MPa and 850 °C/650 MPa, and high temperature of 900 °C/400 MPa and 980 °C/250 MPa are shown in Fig. 2a, b, respectively. The creep behavior of the alloy at both intermediate temperature (  850 °C) and at high temperature (  900 °C) exhibits a clear difference. The creep curves of the alloy preformed under the intermediate temperature indicate a clearly incubation period followed by the usual three creep stages, and the primary creep of the alloy exhibits considerable strains, such as 2.4% primary strain of specimen A5 at 760 °C and 785 MPa. No incubation periods of creep are observed as the creep temperature increased to 900 °C, and the primary strain of the alloy creep at high temperatures is very small. After the primary stage, the secondary stage of the alloy at both the intermediate temperature and high temperature is characterized by low and stable creep rates and is followed by the tertiary creep stage. The creep characteristics of DD6 alloy near the [001] orientation are the same as those of the typical second-generation single-crystal superalloy CMSX-4 [12] crept at 700–980 °C. It is emphasized that the specimens near the [001] orientation exhibit a wide variation in primary creep at the intermediate temperature. Compared with specimens oriented close to the [001]–[111] boundary, the specimens oriented close to the [001] direction and the [001]–[011] boundary show a low primary creep strain. The primary creep strains of specimens A5 and T85 with orientations close to the [001] pole are 2.4% and 0.5%, respectively, and the primary creep strains of specimens I1 and TL85 with orientations close to the [001]–[011] boundary are 4.6% and 0.5%, respectively. While the primary creep strains of specimen G2 and TR85 with orientations close to the [001]– [111] boundary can be as high as 14.5% and 2.3%, respectively. The creep rupture lifetimes are also strongly influenced by the orientation at the given intermediate temperature test conditions, as observed in Table 1 and Fig. 2a. In general, the creep rupture lifetimes of the alloy near the [001] orientation rank in the following order: lifetime at the [001] pole> lifetime at the [001]–[011] boundary> lifetime at the [001]–[111] boundary. The anisotropic creep rupture lifetimes of DD6 alloy closely agree with the results of Mar-M247 and Mar-M200 alloys tested at 760 °C [3, 13].

Creep Anisotropy in Single-Crystal Superalloy DD6 …

305

Fig. 1 Initial orientation of specimens in the [001] corner of the standard stereographic triangle

Table 1 Compilation of creep data for DD6 alloy near the [001] orientation at different temperatures and stresses Temperature/stress

Specimen

Initial orientation

Life

Elongation

Primary creep strain

Deviation of rupture lifetimes

Region

h/°

q/°

/h

/%

/%

T65

[001]

1.3

19.3

1680a

3.7a

Not measured

>1.6

TR65

[001]–[111]

15.6

42.0

1013

23.7

A5

[001]

2.2

30

536

19.2

2.4

42.2

I1[9]

[001]–[011]

16.7

9.4

223

26.3

4.6

G2

[001]–[111]

15.8

41

12.7

30.6

14.5

T85

[001]

0.9

15

161

19.4

0.5

TL85

[001]–[011]

18.8

5.7

98.5

18.4

0.5

Creep test at different temperatures 650 °C/800 MPa 760 °C/785 MPa

850 °C/650 MPa

900 °C/400 MPa

980 °C/250 MPa

TR85

[001]–[111]

16.8

42.3

52.2

22.3

2.3

T90

[001]

0.5

12

494

37.2

Very small

TL90

[001]–[011]

19.7

5.6

462

24.0

Very small

TR90

[001]–[111]

16.1

40.0

472

25.2

Very small

T98

[001]

2.4

15

236

20.9

Very small

TL98

[001]–[011]

17.1

7.9

256

34.1

Very small

TR98

[001]–[111]

16.3

38.0

269

25.1

Very small

3.0

1.0

1.1

(continued)

306

J. Yu et al.

Table 1 (continued) Temperature/stress

Specimen

Initial orientation Region

Life

Elongation

Primary creep strain

/%

/%

h/°

q/°

/h

4.7

23.0

500a

3.9a

1.9

14.7

a

5.2a

2.9

Deviation of rupture lifetimes

Creep test at 760 °C with different applied stresses 760 °C/680 MPa

760 °C/600 MPa

P6 L6

[001]–[011]

R6

[001]–[111]

14.7

38.5

236

28.7

8.5

P60

[001]

2.8

35

500a

0.7a

0.5

L60

[001]–[011]

17.1

9

500a

1.3a

1.0

a

4.0a

R60 760 °C/550 MPa

760 °C/450 MPa

a

[001]

[001]–[111]

17.3

15.5

41

500

500

a

>2.1

2.8 a

P5

[001]

4.8

19.9

500

0.43

0.43

L5

[001]–[011]

18.8

15.6

500a

0.31a

0.31

a

a

R5

[001]–[111]

16.0

36.5

500

0.53

0.54

P4

[001]

4.6

29.1

500a

0.10a

0

a

a

L4

[001]–[011]

17.7

16.4

500

0.05

0

R4

[001]–[111]

15.6

33.1

500a

0.10a

0

Means that the tests were interrupted

Fig. 2 Creep curves of DD6 alloy with typical orientations at intermediate temperature of 760 °C/785 MPa and 850 °C/650 MPa (a) and high temperature of 900 °C/400 MPa and 980 °C/250 MPa (b)

The anisotropic creep in DD6 alloy near the [001] orientation is strongly influenced by the temperature, as shown in Fig. 2. Compared with the anisotropic creep behavior of the alloy near the [001] orientation at intermediate temperatures of 760 °C/785 MPa and 850 °C/650 MPa, the creep curves and rupture lifetimes of three specimens near the [001] orientation have no obvious difference at high temperatures of 900 °C/400 MPa and 980 °C/250 MPa. To quantify the anisotropic creep behavior, the deviations of the creep rupture lifetimes, which has been defined as the longest life divided by the shortest life of the alloy near the [001] orientation, are calculated and listed in Table 1. The deviation of creep rupture lifetimes is more than 1.6 at intermediate temperatures from 650 to 850 °C. In contrast, the deviation of creep rupture lifetimes is close to 1 at high temperatures of 900 and 980 °C. This means that the anisotropy in creep behavior near the [001] orientation disappears gradually with the increase in temperature to 900 °C.

Influence of the applied stress on the anisotropic creep behavior at 760 °C The anisotropic creep behavior of DD6 alloy at intermediate temperature 760 °C is influenced by the applied stresses, as shown in Table 1, Fig. 3a–c. The anisotropic creep behavior of DD6 alloy at the intermediate temperature vanishes as the applied stress decreased. When the applied stresses are at least 600 MPa, the incubation periods of all the tested specimens are very short. The anisotropic creep behavior can be observed in the creep test at 760 °C/ 785 MPa (as shown in Fig. 2a), in which the specimens oriented close to the [001] pole and the [001]–[011] boundary exhibit lower primary strains and longer rupture lifetimes than the specimens oriented close to the [001]– [111] boundary. When the applied stress is lower than 600 MPa, the incubation periods of the specimens are long time, and no anisotropic creep behavior occurs in the alloy near the [001] orientation.

Creep Anisotropy in Single-Crystal Superalloy DD6 …

307

Fig. 3 Creep curves of DD6 alloy near the [001] orientation at 760 °C/680 MPa (a), 760 °C/600 MPa (b), and 760 °C/550 MPa and 760 °C/450 MPa (c); the extent of primary creep strain of DD6 alloy creep at 760 °C as a function of applied stress (d)

The applied stress has a significant impact on the primary creep strain of DD6 alloy at a testing temperature of 760 °C, as shown in Table 1 and Fig. 3. The primary creep strain of the alloy near the [001] orientation decreases when the applied stress decreases, and the incubation period becomes dominant as the applied stress is lower than 550 MPa. Three types of specimens are still in the incubation period when the creep test is interrupted at 500 h under the lower applied stress of 450 MPa. Previous work [14] has reported that the extent of primary creep strain decreased linearly with the applied stress, and a threshold stress exists for primary creep to be trigger in CMSX-4 alloy at intermediate temperature of 750 and 850 °C. The primary strain of DD6 alloy for these three types orientation is plotted against the applied stress at 760 °C in Fig. 3d. The threshold stresses are 513, 528, and 536 MPa for DD6 alloy with orientations close to the [001] pole, [001]–[011] boundary, and [001]–[111] boundary creep at 760 °C, respectively. There is no obvious difference in the threshold stress of DD6 alloy within 20° of the [001] orientation.

Deformation mechanism of the alloy at intermediate temperature TEM micrographs of DD6 alloy with orientations close to the [001] pole, [001]–[011] boundary, and [001]–[111] boundary after the creep test at 760 °C under different applied stresses are shown in Fig. 4. All the microstructures of the specimens are interrupted at 500 h except specimens A5, I1, G2, and R6, which crept until rupture. The samples are cut parallel to the (001) crystallographic planes. The stacking fault results of the shearing of the c/c′ structure by {111} slip can be widely seen in the c′ precipitates after the creep testing under an applied stress higher than 550 MPa. The heterogeneous c′ precipitate deformation can be seen in specimens near the [001] orientation under high applied stress. There are two directions of stacking faults in the c′ precipitates for the specimens oriented close to the [001] pole, as marked by the arrows in Fig. 4a, d, and the [001]–[011] boundary, as marked by the arrows in Fig. 4b, e, and only one direction of stacking faults in the c′ precipitates for the specimens oriented close to the

308

J. Yu et al.

[001] (a)

(c)

(b)

SF

[001]-[111]

[001]-[011]

785 MPa

SF

SF

(d)

(f)

(e)

SF SF 680 MPa

SF

(g)

(h)

(i)

(j)

(k)

(l)

550 MPa

450 MPa

Fig. 4 TEM micrographs of DD6 alloy with different applied stresses at 760 °C, where B = [001]: a specimen A5, b specimen I1, c specimen G2, d specimen P6, e specimen L6, f specimen R6, g specimen P5, h specimen L5, i specimen R5, j specimen P4, k specimen L4, l specimen R4

Creep Anisotropy in Single-Crystal Superalloy DD6 …

309

[001]–[111] boundary as marked by the arrows in Fig. 4c, f. There is clear evidence that at least two {111} slip systems are operating throughout the creep deformation for specimens with orientations close to the [001] pole and [001]–[011] boundary. In contrast, the specimens with orientations close to the [001]–[111] boundary exhibit predominantly one {111} slip system. The TEM micrographs can be rationalized by Schmid factors of {111} slip systems near the [001] orientation. According to the Schmid factor, there are four equally stressed {111} slip systems close to the [001] pole, a duplex slip system along the [001]–[011] boundary, and a single slip system along the [001]–[111] boundary. The anisotropic primary creep strain and creep rupture lifetimes are associated with the heterogeneous c′ precipitate deformation by {111} slip at an intermediate temperature under a high applied stress (>550 MPa). Previous work has demonstrated that the nucleation and propagation of the a ribbons in creep deformation produced considerable primary creep strain [9, 15] and the principal

650

/800 MPa

hardening mechanism for the transition from primary to secondary creep is the interaction between {111} systems [9]. Thus, multiple {111} systems cause more hardening during creep than the one predominant {111} system does. Therefore, the primary creep strain and creep rupture lifetimes of the alloy near the [001] orientation are anisotropic. The heterogeneous c′ precipitate deformation by {111} slip decreases with decreasing applied stress. There are few stacking faults in the c′ precipitates of the creep at 550 MPa, as shown in Fig. 4g–i. Below the threshold stress of *513 MPa, only a few long and straight dislocations resulting from (111) slip are restricted in the narrow matrix channel of the creep at 450 MPa, as shown in Fig. 4j–l. It is apparent that the deformation of heterogeneous c′ precipitate by {111} slip cannot occur at low applied stresses, especially at an applied stress below the threshold stress. Thus, anisotropic creep behavior does not occur at an intermediate temperature and low applied stress.

850 /650 MPa

(b)

(a)

900 /400 MPa

(c)

[001]

SF SF

[001]-[111]

(d)

(f)

(e)

SF

SF

Fig. 5 TEM micrographs of DD6 alloy near [001] orientation at different temperatures, where B = [001]: a specimen T65 oriented close to the [001] pole at 650 °C/800 MPa; b specimen T85 oriented close to the [001] pole at 850 °C/650 MPa; c specimen T90 oriented close to the [001] pole at 900 °C/400 MPa; d specimen TR65 oriented close to the [001]–[111] boundary at 650°C/800 MPa; e specimen TR85 oriented close to the [001]–[111] boundary at 850 °C/650 MPa; f specimen TR90 oriented close to the [001]–[111] boundary at 900 °C/400 MPa

310

Creep mechanism at different temperatures The microstructures of the alloy near the [001] orientation under the different temperatures exhibit different creep mechanisms, as shown in Fig. 5. There are two directions of stackings faults in the c′ precipitates for the specimens oriented close to the [001] pole, as marked by the arrows in Fig. 5a, b, and there is one direction of stacking faults in c′ precipitates for the specimens oriented close to the [001]– [111] boundary, as marked by the arrows in Fig. 5d, e, for creep at 650 and 850°C. When the temperature is increased to 900 °C, the stacking fault in the c′ precipitates disappears, and the dislocation network arises from {111} slip around the c′ precipitates, as shown in Fig. 5c, f. It is obvious that heterogeneous c′ precipitate deformation by {111} slip occurs at temperatures below 850 °C, while mainly c precipitate deformation by {111} slip occurs at high temperatures 900 °C. It is emphasized that the heterogeneous c′ precipitate deformation by {111} slip is gradually reduced as the temperature increased, even at intermediate temperatures. The density of stacking faults in the c′ precipitates of the creep at 850 °C (Fig. 5c, d) is lower than those of the creep at 650 °C (Fig. 5a, b) and 760 °C (Fig. 4a, c). The {111} slip is mainly restricted to the c channel, which resulted in homogeneous deformation [16]. Therefore,

Fig. 6 Time to 5% creep strain and rupture lifetime in hours plotted as a function of orientation for CMSX-4 alloy at 800 °C under a constant stress of 767 MPa [5], PWA 1484 alloy creep at 732 °C under a constant load of 759 MPa [6], and DD6 alloy creep at 760 °C under a constant load of 785 MPa, respectively

J. Yu et al.

the anisotropy of creep under high stress is gradually reduced with increasing of testing temperature and disappears at high temperature (  900 °C). Anisotropic creep behavior of the other single crystal superalloy The anisotropic creep behavior of the second-generation single-crystal superalloy DD6 at intermediate temperature under a high stress has been reported in this article. The comparison of creep properties of DD6 alloy to the other second-generation single-crystal alloy CMSX-4 [5] and PWA 1484 [6] creep at similar intermediate temperature are shown in Fig. 6. The first numbers in parentheses represent the times to 5% creep strain, and the second numbers in parentheses represent the rupture lifetimes in hours. Times to 5% creep strain of DD6 alloy originate from Fig. 2a. The time to 5% creep strain and rupture lifetimes is plotted in Fig. 6 for CMSX-4 alloy creep at 800 °C under a constant stress 767 MPa [5], PWA 1484 alloy creep at 732 °C under a constant load of 759 MPa [6], and DD6 alloy creep at 760 °C under a constant load of 785 MPa, respectively, as a function of initial orientation. Which the times to 5% creep strain and rupture lifetimes of the single-crystal superalloy with the orientation close to the [001] pole and [001]–[011] boundary are longer than those close to the [001]–[111] boundary. The anisotropic creep properties of CMSX-4 and

Creep Anisotropy in Single-Crystal Superalloy DD6 …

PWA 1484 alloys closely agree with the results of DD6 alloy at intermediate temperature. It should be noted that the creep properties of DD6 and PWA 1484 alloys are performed in constant load tests. The actual stress is more than the initial stress after the large primary creep strain. The creep anisotropy is further amplified by the effect of primary creep strain on the effective stress levels during the secondary and tertiary creep. While the creep properties of CMSX-4 alloy are performed in constant stress tests. The effect of primary creep strain on the actual stress is much less pronounced for CMSX-4 alloy under constant stress tests. The anisotropy of the rupture lifetimes of CMSX-4 alloy can be reduced by constant stress creep tests.

Summary and Conclusion This work has presented the anisotropic creep behavior and mechanisms of DD6 alloy near the [001] orientation at different temperatures and applied stresses. The results reveal that the anisotropic creep for DD6 alloy near the [001] orientation is strongly influenced by the temperature in the range of 650–980 °C. Moreover, the anisotropy of the primary creep strain and rupture lifetime at an intermediate temperature of 760 °C is influenced by the applied stress. Compared with specimens oriented close to the [001]–[111] boundary, the specimens oriented close to the [001] pole and [001]–[011] boundary exhibit lower primary creep strains and longer rupture lifetimes at an intermediate temperature and a high applied stress. The mechanism of anisotropic creep is attributed to the heterogeneous c′ precipitate deformation by {111} slip. With the increase in the temperature or the decrease in the applied stress, the heterogeneous c′ precipitate deformation by {111} slip is gradually reduced, and the anisotropic creep behavior of the alloy near the [001] orientation disappears.

References 1. Reed RC (2006) The superalloys: fundamentals and applications. Cambridge University Press, Cambridge.

311 2. McLean M (1983) Directionally solidified materials for high temperature service. The metals society, London. 3. MacKay RA, Dreshfield RL, Meier RD (1980) Anisotropy of Nickel-base superalloy single crystals. In: Tien JK, Kent WB (eds) Superalloys 1980, Warrendale, PA, 1980. ASM, pp 385–394. 4. Sass V, Glatzel U, Feller-Kniepmeier M (1996) Creep anisotropy in the monocrystalline nickel -base superalloy CMSX-4. In: Kissinger RD, Deye DJ, Anton DL et al. (eds) Superalloys 1996, Warrendale, PA, 1996. TMS, pp 283–290. 5. Sass V, Schneider W, Mughrahbi H (1994) On the orientation dependence of the intermediate-temperature creep behaviour of a monocrystalline nickel-base superalloy. Scr Metall Mater 31 (7): 885–890. 6. Shah DM, Vega S, Woodard S, Cetel AD (2004) Primary creep in nickel-base superalloys. In: Green KA, Pollock TM, Harada H et al. (eds) Superalloys 2004, Warrendale, PA, 2004. TMS, pp 197–206. 7. Matan N, Cox DC, Carter P, Rist MA, Rae CMF, Reed RC (1999) Creep of CMSX-4 superalloy single crystals: effects of misorientation and temperature. Acta Mater 47 (5):1549–1563. 8. Caron P, Ohta Y, Nakagawa YG, Kahn T (1988) Creep deformation anisotropy in single crystal superalloy. In: Reichman S, Duhl DN, Maurer G, Antolovich SD, Lund C (eds) Superalloys 1988, Warrendale, PA, 1988. TMS, pp 215–224. 9. Yu J, Li JR, Zhao JQ, Han M, Shi ZX, Liu SZ, Yuan HL (2013) Orientation dependence of creep properties and deformation mechanism in DD6 single crystal superalloy at 760 °C and 785MPa. Mater Sci Eng A 560:47–53. 10. MacLachlan DW, Knowles D (2000) Creep-behavior modeling of the single-crystal superalloy CMSX-4. Metall Mater Trans A 31 (5):1401–1411. 11. Li JR, Zhong ZG, Tang DZ, Liu SZ, Wei P, Wei PY, Wu ZT, Huang D, Han M (2000) A Low-cost Second Geneution Single Crystal Superalloy DD6. In: Pollock TM, Kissinger RD, Bowman RR et al. (eds) Superalloys 2000, Warrendale, PA, 2000. TMS, pp 777–783. 12. Schneider W, Hammer J, Mughrahbi H (1992) Creep deformation and rupture behaviour of the monocrystalline superalloy CMSX-4 - A comparison with alloy SRR99. In: Antolovich SD, Stusrud RW, MacKay RA et al. (eds) Superalloys 1992, Warrendale, PA, 1992. TMS, pp 589–598. 13. MacKay RA, Maier RD (1982) The influence of orientation on the stress rupture properties of Nickel-Base Superalloy Single Crystals. Metall Trans A 13A:1747–1754. 14. Rae CMF, Reed RC (2007) Primary creep in single crystal superalloys: Origins, mechanisms and effects. Acta Mater 55 (3): 1067–1081. 15. Rae CMF, Zhang L (2009) Primary creep in single crystal superalloys some comments on effects of composition and microstructure. Mater Sci Technol 25 (2):228–235. 16. Murakami H, Yamagata T, Harada H, M. Yamazaki (1997) The influence of Co on creep deformation anisotropy in Ni-base single crystal superalloys at intermediate temperatures. Mater Sci Eng A 223 54–58.

Equations to Predict the Elastic Modulus of the Individual Gamma and Gamma-Prime Phases in Multi-component Ni-Base Superalloys Takuma Saito, Makoto Osawa, Tadaharu Yokokawa, Hiroshi Harada, Toshiharu Kobayashi, Kyoko Kawagishi, and Shinsuke Suzuki

Abstract

Strength of Ni-base single-crystal superalloys under high temperature and low stress creep usually is enhanced by formation of c/c′ raft structure and larger aspect ratio of c′ phase in the c/c′ raft structure. Elastic misfit between c and c′ phases is one of the most important factors to control the aspect ratio of the c′ phase in the c/c′ raft structure formed under external stress. The aspect ratio of the c′ phase is controlled by kinetics for the c/c′ raft structure formation, which is affected by a strain inhomogeneity caused by this elastic misfit between the c and c′ phases under external stress. To realize a new alloy design approach to control the aspect ratio of the c′ phase in the c/c′ raft structure, this research aimed to obtain the regression equations which can predict elastic modulus of the individual c and c′ phases for multi-component Ni-base single-crystal superalloys based on measurements of elastic modulus of Ni-base single-crystal alloys. Elastic modulus of the individual c and c′ phases of various kinds of Ni-base single-crystal alloys was measured by using rectangular parallelepiped resonance (RPR) method. Using the analyzed and referenced elastic modulus, regression equations for predicting longitudinal elastic modulus of the individual c and c′ phases and its temperature and composition dependence were obtained. Detailed analysis of the elastic modulus and its composition dependence was executed to clarify the contribution of each element on the elastic

T. Saito (&)
M. Osawa
T. Yokokawa
H. Harada
T. Kobayashi
K. Kawagishi National Institute for Materials Science, 1-2-1 Sengen, Tsukuba, Ibaraki 305-0047, Japan e-mail: [email protected] T. Saito
S. Suzuki Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo, 169-8555, Japan

modulus. At 900 °C, Re, Ta, Ti, Al, and Mo reduce the longitudinal elastic modulus in the c phase. On the other hand, Ru, Re, Ta, Ti, Al, W, and Mo enlarge the elastic modulus in the c′ phase. Keywords





Single crystal Alloy design misfit Raft structure




Elastic modulus




Elastic

Introduction Ni-base superalloys are required to have excellent high-temperature strength for high thermal efficiency of jet engines and land-based gas turbines. Under low stress and high temperature creep conditions, the c/c′ raft structure is beneficial for creep resistance, because it effectively prevents the motion of dislocations, like climb [1]. Furthermore, larger aspect ratio of the c′ phase which can indicate larger area of horizontal c/c′ interface and more continuous c′ phase horizontally in the c/c′ raft structure shows larger creep resistance [2, 3]. Controlling the aspect ratio of the c′ phase is one of the promising alloy design approaches for turbine blade materials for high pressure turbines. We focused on the aspect ratio of the c′ phase in the c/c′ raft structure to establish new alloy design approach. We aim to develop new alloys having vicinity of 60% c′ volume fraction which shows the longest creep life at a condition such as 1100 °C and 137 MPa [4, 5]. In this region of the c′ volume fraction, alloys maintain c/c′ raft structure for longer time although alloys having more than the vicinity of 60% c′ volume fraction tend to show topological inversion earlier to degrade the c/c′ raft structure resulting in shorter creep life [4]. The aspect ratio of the c′ phase could be determined by initial distribution of the c′ phase before deformation, and kinetics for the c/c′ raft structure formation during deformation at high temperature.

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_30

312

Equations to Predict the Elastic Modulus …

313

\1 [ Usually, negatively larger lattice misfit between the c and c′ phases can contribute to aligned distribution of the c′ phase as far as the c and c′ phases are coherent at their interface. The better aligned c′ phase results in formation of the c/c′ raft structure having larger aspect ratio of the c′ phase through its geometry change [2, 3]. We have already controlled the lattice misfit precisely [6] by using Alloy Design Program (ADP) [7]. Controlling the lattice misfit, we have been developing optimum Ni-base single-crystal superalloys for each application. The definition of the lattice misfit is shown in Eq. (1). dL ¼

ac 0  ac ac

ac : Lattice constant of c phase, m

ð1Þ

ac0 : Lattice constant of c0 phase, m \2 [ Faster formation of the c/c′ raft structure during primary creep tends to form the c/c′ raft structure having larger aspect ratio of the c′ phase [8]. Kinetics for the c/c′ raft structure formation depends on interdiffusion rate of constituent elements, and driving force for the c/c′ raft structure formation. \2:1 [ Composition of alloys affect the atomic diffusion [9]. Heavy elements like W, Ta, and Re could reduce the average diffusion rate, results in slowing down of the c/c′ raft structure formation. At the same time, as the diffusion rate is controlled by diffusion path and acting diffusion mechanisms, an activation energy of it controls the rate. For example, region around dislocations provides a chance so-called of pipe diffusion, which is faster than lattice diffusion. Therefore, the pipe diffusion could accelerate the c/c′ raft structure formation more rapidly than the lattice diffusion [10] \2:2 [ The c/c′ raft structure never forms without driving force based on free-energy difference. In addition to this, larger driving force for the c/c′ raft structure formation could accelerate the c/c′ raft structure formation in the kinetics. The essential material parameter causing the driving force for the c/c′ raft structure formation under external stress is not the lattice misfit but an elastic modulus misfit between c and c′ phases, hereafter elastic misfit [11, 12]. A definition of the elastic misfit in this report is shown in Eq. (2).

dE ¼

Ec 0  Ec Ec

Ec : h100ilongitudinal elastic modulus of c phase/GP Ec0 : h100ilongitudinal elastic modulus of c0 phase/GP ð2Þ In this report, we will focus on this driving force for the c/ c′ raft structure formation and try to establish a quantitative prediction of this elastic misfit on the Ni-base single-crystal superalloys. Mechanisms of the c/c′ raft structure formation and its driving force are described more precisely below. The mechanism of the c/c′ raft structure formation is understood by elastic [11], and elastic-plastic [13] regimes. The comprehension on the c/c′ raft structure formation is schematically shown in Fig. 1. Fundamental driving force for the c/c′ raft structure formation in both the regimes is based on an anisotropic c/c′ interfacial strain distribution [14]. Especially, under a tensile stress condition with negative lattice misfit alloys, strain inhomogeneity [15] emerges especially in a longitudinal c/c′ interface due to the elastic misfit as far as coherent c/c′ interface preserved. This strain inhomogeneity is called “external strain inhomogeneity” in this report. The external strain inhomogeneity in the longitudinal c/c′ interface should contribute to the larger anisotropy of the interfacial strain distribution. In this situation, the c/c′ raft structure formation is supposed to be advanced by reducing the coherent longitudinal c/c′ interface, because the longitudinal c/c′ interface is a source of higher external strain inhomogeneity, which seems to be more unstable than any other regions due to its higher elastic energy. On the other hand, in the elastic-plastic regime, reacted interfacial dislocations on a horizontal c/c′ interface can increase the driving force for the c/c′ raft structure formation [16]. Relationship between the driving force and the kinetics for the c/c′ raft structure formation is described below. We assume that the kinetically fast c/c′ raft structure formation does not give enough time for dislocations to enter the longitudinal c channel. Based on this reason, the kinetically fast c/c′ raft structure formation could prevent such a dislocation from depositing at the coherent longitudinal c/c′ interface. In this way, reacted dislocations at the coherent longitudinal c/c′ interfaces formed for releasing the coherent elastic stress could not be formed. This means the driving force for the c/c′ raft structure formation can be kept even during the formation of the c/c′ raft structure, resulting in larger aspect ratio of the c′ phase. On the other hand, the elastic-plastic regime takes longer time to initiate than the elastic regime. In addition to this, multiplied dislocations

314

T. Saito et al.

(a) Elastic regime γ′ phase γ phase

Horizontal γ channel Longitudinal γ channel

Strain inhomogeneity Strain inhomogeneity under external stress caused by caused by (External strain inhomogeneity) Isotropic Anisotropic interfacial strain interfacial strain

(b) Elastic-plastic regime

Reacted interfacial dislocations to release

Strongly anisotropic interfacial strain

Fig. 1 Mechanism for c/c′ raft structure formation under high temperature and low stress creep

begin to deposit on the c/c′ interface mainly on the horizontal c/c′ interface and reacted interfacial dislocations could be formed even on the longitudinal c/c′ interface which could dramatically decrease the driving force for the c/c′ raft structure formation more often in the elastic-plastic regime than the elastic regime. From these assumptions, the c/c′ raft structure formation should be progressed within the elastic regime as much as possible to obtain larger aspect ratio of the c′ phase. Therefore, we could obtain the c/c′ raft structure having larger aspect ratio of the c′ phase owing to the kinetically faster c/c′ raft structure formation in the elastic regime by enlarging the elastic misfit. Actually, some researches have already demonstrated that the elastic misfit under external stress could have some influence on the kinetics for the c/c′ raft structure formation by phase field simulation [17, 18]. As described above, larger elastic misfit is beneficial to realize larger aspect ratio of the c′ phase in the c/c′ raft structure. Presently, ADP can precisely predict equilibrium compositions of the c and c′ phases [7] and the lattice misfit from the predicted composition of the c′ phase by using various regression equations [6]. However, we cannot estimate the elastic misfit by using present ADP, because there is no regression equation of the elastic modulus for each phase to calculate elastic misfit in ADP. Moreover, dataset of the elastic modulus already reported is not enough to derive every coefficient of compositional elements in Ni-base superalloys by regression analysis [19–22]. In this research, to control the aspect ratio of the c′ phase in the c/c′ raft structure by changing the elastic misfit, we aimed to establish regression equations to predict

the longitudinal elastic modulus of the individual c and c′ phases, respectively. The regression equations are based on an empirical approach and can be adapted to multi-component Ni-base single-crystal superalloys. To establish regression equations, first, the elastic modulus of the individual c and c′ phases in Ni-base alloys is analyzed by the rectangular parallelepiped resonance (RPR) method. Second, the regression equations of the longitudinal elastic modulus are obtained by using analyzed and referenced elastic modulus. Finally, effects of each element on the longitudinal elastic modulus at 900 °C are evaluated by the comparison of the regression coefficients of each element.

Experimental Procedure Alloys for Elastic Modulus Measurement Alloy compositions of the individual c and c′ phases are listed in Table 1. Alloys with a single phase of c (c phase sample) and c′ (c′ phase sample) were designed so that their compositions are the same as those of c and c′ phases in two-phase samples at 900 °C, respectively. The equilibrium compositions of Ni–Al [23], Ni–Al–W [7], Ni–Al–Ta, and Ni–Al–Re [24] were estimated from published phase diagrams, while those of Ni-Al–Ta were analyzed in Ni–Al–Ta c + c′ two-phase sample (Ni–9.61Al–4.78Ta, at.%) by an electron probe microanalyzer (EPMA, Shimadzu EPMA-1610) after striking with a hammer and recrystallization heat treatment process at 1240 °C for 30 h followed by aging at 900 °C for 1500 h. In the case of pure Ni and

Equations to Predict the Elastic Modulus … Table 1 Designed compositions of single-crystal samples

Phase

315 Sample

Composition (at.%, Ni; bal.) Co

c

c′

Cr

Mo

W

Al

Ti

Ta

Hf

Re

Ru

Pure Ni

–

–

–

–

–

–

–

–

–

–

Ni–13.0Al

–

–

–

–

13.0

–

–

–

–

–

Ni–14.4Al

–

–

–

–

14.4

–

–

–

–

–

Ni–Al–Ta

–

–

–

–

6.3

–

3.4

–

–

–

Ni–Al–W

–

–

–

3.0

7.0

–

–

–

–

–

Ni–Al–Re

–

–

–

–

13.6

–

–

–

0.8

–

TMS-26

14.4

13.4

2.3

5.5

3.2

–

0.9

–

–

–

TMS-82+

13.3

12.0

2.2

3.8

3.9

0.2

0.7

0.01

1.9

–

TMS-75

19.7

7.3

2.3

2.2

5.4

–

0.7

0.01

3.9

–

TMS-196

10.1

11.1

2.3

2.0

4.9

–

0.6

0.01

5.2

5.3

TMS-238

12.5

15.3

1.4

1.6

5.0

–

0.8

0.01

6.3

5.7

Ni–23.2Al

–

–

–

–

23.2

–

–

–

–

–

Ni–Al–Ta

–

–

–

–

12.1

–

7.8

–

–

–

Ni–Al–W

–

–

–

3.0

21.1

–

–

–

–

–

Ni–Al–Re

–

–

–

–

22.0

–

–

–

1.2

–

TMS-26

5.9

2.7

0.7

3.2

16.5

–

3.9

–

–

–

TMS-82+

5.2

2.2

0.7

2.5

17.1

0.9

3.0

0.05

0.2

–

TMS-75

8.4

1.4

0.7

1.9

18.6

–

2.9

0.05

0.4

–

TMS-196

3.7

1.7

0.8

1.6

17.8

–

2.7

0.05

0.5

2.0

TMS-238

4.8

2.2

0.5

1.3

17.3

–

3.4

0.05

0.6

2.1

Ni–13 Al, they are designed not to have equilibrium composition but single c phase. At the same time, the equilibrium compositions of the c and c′ phase of Ni-base superalloys, TMS-26, TMS-82 + , TMS-75, TMS-196, and TMS-238 at 900 °C were predicted by using ADP. Single crystals having designed individual c and c′ compositions were obtained by using a directional solidification furnace. Raw materials of these elements were measured using an electronic balance so that the compositions correspond to Table 1 and the total weight of the elements is approximately 2 kg. The raw materials were melted in an

Al2O3 crucible and held at 1600 °C by using a vacuum induction melting furnace. The molten metal was poured into a mold with eight cylinders having grain selectors to obtain a single crystal. The mold was pulled down at 200 mm/h in this furnace producing single crystalline specimens. After the casting, solution heat treatment, primary aging treatment, and secondary aging treatment were applied to these single crystals in a vacuum furnace. Ni–14.4Al, Ni– 23.2Al, Ni–Al–W, Ni–Al–Ta, and Ni-Al–Re samples were heat treated at 1255 °C for 5 h, followed by primary aging at

(c)

(b)

(a)

100μm Fig. 2 Microstructure of Ni-Al-Ta samples after heat treatment, a c phase alloy after solution treatment, primary and secondary aging, b c + c′ two-phase alloy after recrystallization process at 1240 °C for

10μm

100μm

30 h followed by aging at 900 °C for 1500 h, c c′ phase alloy after solution treatment, primary and secondary aging

316

1100 °C for 4 h and secondary aging at 870 °C for 20 h. c phase samples and c′ phase samples of TMS-26 [25], TMS-82+ [8], TMS-75 [26], TMS-196 [27], and TMS-238 [28] were subjected to the heat treatment for their original c + c′ two-phase samples. Finally, single-crystal rods (diameter 11 mm and longitudinal length 140 mm) were obtained by cutting the samples. Microstructure of obtained single-crystal rods was observed by SEM (JEOL, JSM-6060) and all the rods were confirmed to exhibited expected microstructure although c′ phase samples tended to have small amount of c phase. As an example, microstructure of Ni–Al–Ta samples is shown in Fig. 2.

T. Saito et al.

(a)

Sample Preparation for RPR Method Approximately 4 mm rectangular parallelepiped samples were obtained by following steps. The crystal orientations of single-crystal rods were measured by using X-ray back Laue method with an X-ray generator (Rigaku, SA-HF3) and {100} of the single-crystal rods was determined. Rectangular parallelepiped samples with edge lengths of 4.5 mm were obtained by cutting with a precision cutting machine (HEIWA TECHNICA HS-100) or a wire EDM machine (Brother, HS-300). This step was performed so that each plane of the rectangular parallelepiped samples is to be parallel to {100}. Then, one of the {100} of the sample was attached to a sample holder with a 2-axis goniometer (SERVTEC Co. Ltd., M-127). Crystal orientation of the sample attached to the sample holder was measured with X-ray back Laue method. The sample direction was aligned to and was polished by using 30-lm diamond slurry and lapping plates. Subsequently, the measured plane of the rectangular parallelepiped samples was attached to another sample holder (SERVTEC Co. Ltd., M-6325) and the plane of the rectangular parallelepiped sample initially attached to the sample holder with a 2-axis goniometer was polished. The polished plane in this step was polished to be parallel to the opposite plane already polished. Finally, the rectangular parallelepiped samples with all planes parallel to {100} and edge lengths of approximately 4 mm were obtained by repeating these steps on each of their planes. To obtain clearly separated spectrum of resonance frequency for easy analysis, each edge length of rectangular parallelepiped sample was made to be different length to each other. The difference was approximately between 0.01 and 0.10 mm. Furthermore, the edge lengths and the weights of the rectangular parallelepiped samples were measured using a micrometer and an electronic balance, respectively. Subsequently, the average edge lengths of the rectangular parallelepiped samples in the same direction were calculated. The volumes of the rectangular parallelepiped samples were calculated by using the average edge lengths under the

(b)

Fig. 3 Rectangular parallelepiped resonance (RPR) method, a the flow chart of the RPR method, b the apparatus to measure the resonance frequency of rectangular parallelepiped samples

assumption that the samples were perfectly rectangular parallelepiped. The densities of the rectangular parallelepiped samples were calculated through division of their weight by their volume.

RPR method The elastic moduli of the rectangular parallelepiped samples were determined through the experimental and analytical steps by using the RPR method [29–31] as shown in Fig. 3a. In the experimental step, an apparatus (Japan Techno Plus Co. Ltd., CC-HT) was used to measure the resonance spectrum of the sample as shown in Fig. 3b. During this experiment, the vacuum was maintained at 1:0  104 Pa. The temperature of the rectangular parallelepiped sample was controlled by using a heater from room temperature to 1100 ° C and was measured using an R-type thermocouple set close to the sample. The rectangular parallelepiped sample was placed between the Al2O3 rods. Then, sign-wave voltage was generated in the function generator and its frequency was swept from 250 to 1250 kHz. The applied voltage was converted into vibration through a piezo element. Then, the

Equations to Predict the Elastic Modulus …

317

S¼

X X fi;jsim  fi;jmea j

i

!2

800

EY-1 EZ-1

600 400 200

2

1

600

EX-1

701 Measured resonance frequency

Simulated resonance frequency

4

3

800

Voltage V /mV

vibration was transferred to the rectangular parallelepiped sample through one of the Al2O3 rods. In this situation, the rectangular parallelepiped sample was resonated during sweeping when the conveyed frequency was coincident with the resonance frequency of the rectangular parallelepiped sample. The resonance of the sample was conducted to another piezo element through another Al2O3 rod at opposite side of the sample. The resonance was converted into voltage again through another piezo element. Then, the amplifier amplified the voltage sufficiently to analyze the resonance frequency spectrum with an oscilloscope. Finally, the resonance data were transferred to a computer. In the analytical step, the resonance frequency spectrum of the rectangular parallelepiped sample was simulated by assuming the elastic modulus, inputting density, and edge length of the sample by using a software program (Japan Techno Plus Co. Ltd., f-calcTM). This program can simulate resonance frequency of OD, EV, OX, OY, OZ, EX, EY, and EZ resonance modes from 1st to 8th degree. The analyzed value of elastic modulus of the rectangular parallelepiped samples was determined by fitting the simulated resonance frequency spectrum to the measured spectrum. From Eq. (3), the degree of coincidence between the measured and simulated spectra was evaluated [31]. The simulated and measured spectra were judged to be consistent when the value presented in Eq. (3) was smaller than 0.5%.

EX-1

EY-1 EZ-1

600

4

400

3

2

200

1

800

EY-1 EZ-1

504

EX-1

600 400 200 800 600

3

2

1

EY-1 EZ-1

402

4

EX-1

2

400 4

200

3

1

0 280

290

300

310

320

330

340

Frequeny f /kHz Fig. 4 Measured and simulated resonance frequency spectra of rectangular parallelepiped sample of the Ni-Al-Ta c phase sample at 400, 500, 600, and 700 °C, respectively

fi;jsim

fi;jsim : i - th simulated resonance frequency of j - mode; kHz ð3Þ

Results of RPR Method Figure 4 shows a temperature dependence of resonance frequency of Ni–Al–Ta c phase sample. In the measured spectrum, almost all peaks shifted to a lower-frequency side with the increase in temperature. However, some peaks showed slight temperature dependence: for example, peak 3 in Fig. 4. These peaks originated from the Al2O3 rods; the elastic modulus of ceramics generally shows subtle temperature dependence. Only the resonance frequency spectrum originating from the rectangular parallelepiped sample was determined by removing the peaks originating from ceramics. Figure 5 shows the temperature dependence of EX-mode resonance frequency of Ni–Al–Ta c phase sample. To avoid taking fault fitting of resonance frequency spectrum in the analytical step, continuous decrease of resonance frequency depending on temperature was confirmed in every resonance

1200

Resonanance frequency f /kHz

fi;jmea : i - th measured resonance frequency of j - mode; kHz

1000

EX-8 EX-7

800

Simulated value

EX-6 EX-5

600

EX-4 EX-3

400

EX-2 200 0

EX-1

Measured value

0

200

400

600

800

1000

1200

Temperature T / C

Fig. 5 Temperature dependence of EX-mode resonance frequency for the rectangular parallelepiped sample of the Ni–Al–Ta c phase sample

mode from 1st to 8th degree. Fitting of measured spectrum of resonance frequency by simulated spectrum of resonance frequency was executed again and again by assuming different initial elastic modulus until continuous temperature dependence in spectrum of resonance frequency was

318

T. Saito et al.

achieved. The elastic modulus used in successful fitting is called as “analyzed elastic modulus” in this report. Here, elastic modulus used in this manuscript will be explained. Generalized Hooke’s law is shown in Eq. (4). rij ¼ cijkl ekl rij : stress

ð4Þ

cijkl : elastic stiffness tensor ekl : strain

The Regression Equations for Predicting the Elastic Modulus

ðc11  c12 Þðc11 þ 2c12 Þ c11 þ c12 c11 ; c12 : elastic stiffness; GPa E¼

ð5Þ

The analyzed longitudinal elastic modulus of the c phase samples and their temperature dependence is plotted in Fig. 6a, and that of the c′ phase samples is plotted in Fig. 7a, respectively.

longitudinal elastic modulus E /GPa

PureNi

140 130 120 110

Ni-14.4Al

Ni-13.0Al

Ni-12.7Al[19]

Ni-Al-W

Ni-Al-Ta

Ni-Al-Re

TMS-26

TMS-82+

TMS-75

TMS-196

TMS-238

CMSX-4[21]

CMSX-4[22]

modeled

100 90 80 70

(a)

60 50

0

200

400

600

800

Temperature T /ºC

1000

1200

Regression Equations Including Temperature and Composition Dependence A regression equation of the longitudinal elastic modulus of the c phase showing temperature and composition dependence was obtained by regression analysis on the chemical compositions and analyzed elastic modulus of the c phase samples. Likewise, that of the c′ phase was derived from chemical compositions and analyzed elastic modulus of the c′ phase samples. The composition used in this regression analysis is listed in Table 1. For this regression analysis, analyzed elastic modulus and composition in this research and references on Ni–Al samples [19, 20] were used together. Composition of referenced Ni–Al c phase samples is Ni–12.7Al [19], on the other hand, those of Ni–Al c′ phase samples are Ni–23.2Al, Ni–24.0Al, and Ni–25.0Al [20]. The referenced elastic modulus of Ni–Al samples [19, 20] was also obtained by the RPR method.

Analyzed longitudinal elastic modulus E /GPa

Face centered cubic-based structure, for example, c phase and c′ phases of Ni-base superalloys, has three independent elastic moduli based on its symmetry of crystal structure. By Voigt notation, cijkl in Eq. (4) is restricted to three independent elastic moduli, c11, c12, and c44, in case of face centered cubic structure. These c11, c12, and c44 are used to simulate resonance frequency spectrum in the analytical step. Analyzed c11 and c12 were converted into the longitudinal elastic modulus using Eq. (5).

150

It is noted that the elastic modulus of pure Ni was not analyzed below its Curie temperature, 350 °C, because the elastic modulus of pure Ni can be affected by magnetostriction below its Curie temperature [32] even in high frequency region of this RPR method between 250 and 1250 kHz.

150 140

Data used for the regressoin analysis

130

Prediction(CMSX-4 modeled[21])

120

Prediction(CMSX-4[22])

110 100 90 80 70

(b)

60 50 50

60

70

80

90

100

110

120

130

140

150

Predicted longitudinal elastic modulus E /GPa

Fig. 6 The longitudinal elastic modulus of c phase samples (a) Temperature dependence, (b) Difference between analyzed and predicted values

Equations to Predict the Elastic Modulus … Ni-23.2Al Ni-24.0Al[20] Ni-Al-W Ni-Al-Re TMS-82+ TMS-196 CMSX-4[21]

140 130 120

modeled

110

Ni-23.2Al[20] Ni-25.0Al[20] Ni-Al-Ta TMS-26 TMS-75 TMS-238 CMSX-4[22]

100 90 80 70

(a)

60 50

0

200

400

600

800

1000

1200

Analyzed longitudinal elastic modulus E /GPa

longitudinal elastic modulus E /GPa

150

319 150 Data used for the regression analysis

140

Prediction(CMSX-4 modeled[21])

130

Prediction(CMSX-4[22])

120 110 100 90 80 70

(b)

60 50 50

60

70

80

90

100

110

120

130

140

150

Predicted longitudinal elastic modulus E /GPa

Temperature T/ºC

Fig. 7 The longitudinal elastic modulus of c′ phase samples (a) Temperature dependence, (b) Difference between analyzed and predicted values

The equations for predicting the elastic modulus of each phase were expressed by Eq. (6). Regression coefficients of these equations were obtained by the regression analysis. The term cAl was modified into 25-cAl in this equation for c′ phase, because the composition at the starting point of this equation for the c′ phase should be Ni3Al. Based on these ways to make regression equations, the constant term of this equation for the c phase means the longitudinal elastic modulus of pure Ni at −273.15 °C, on the other hand, the constant term of this equation for the c′ phase means that of Ni3Al at −273.15 °C, as physical meaning. Regression coefficients of Eq. (6) obtained from this regression analysis are shown in Fig. 8. Amount of Hf is much smaller than any other elements in all alloys. Therefore, in this regression analysis, effect of Hf on elastic modulus is ignored. E ¼ D0 þ DT T þ DT 2 T 2 þ DT 3 T 3 þ

X ðDci ci þ DAi Ai Þ i

Ai ¼ ci T : Product of concentration of i-th element and temperature; ðat:% ÞK ci : Concentration of i - th element,ðat:%Þ T : Temperature; K D0 : Regression constant; GPa DT n : Regression coefficients for n - th power of temperature,GPa
Kn Dci : Regression coefficients of ci ; GPa
ðat:%Þ1 DAi : Regression coefficients of Ai ; GPa
ðat:%Þ1 K1

ð6Þ Figures 6b and 7b show the difference between predicted and analyzed values of elastic modulus. The accuracy of the equations was estimated roughly by the comparison between

analyzed and predicted values of referenced CMSX-4 [21, 22]. The composition of the individual c and c′ phases of CMSX-4 used to predict the elastic modulus by using Eq. (6) is listed in Table 2. Composition of the individual c and c′ phases of modeled CMSX-4 was quoted from the reference [21]. On the other hand, equilibrium composition of the individual c and c′ phases of actual CMSX-4 [22] was predicted by using ADP at 900 °C. The predicted elastic modulus is shown in Figs. 6a and 7a as lines and shown in Figs. 6b and 7b as plots, respectively. The referenced values of modeled CMSX-4, obtained by Siebörger et al. [21] using a free-free beam resonance method, show good agreement with the predicted values in both the individual c and c′ phases. On the other hand, the referenced values obtained by Dye et al. [22] are different from the predicted values in the case of the c phase. There are two possible reasons why Eq. (6) for the c phase did not predict elastic modulus of the c phase of CMSX-4 obtained by Dye et al. [22] accurately. First possibility is that, their values are obtained assuming an isotropic elastic body. Therefore, they ignore shear modulus c44 in their analysis [22]. c11, c12, and the longitudinal elastic modulus could be changed in their analysis if they assume c44. This result might indicate that the elastic modulus of the c phase of CMSX-4 is more influenced by an anisotropy of the sample than that of the c′ phase. Second possibility is that. Regression coefficient of Ti in Eq. (6) for the c phase might have lower reliability than any other elements. CMSX-4 contains Ti; therefore, the elastic

320

T. Saito et al.

(a) (c)

γ phase

0

γ phase

100

120

140

160

180

ARu 200

ARe ATa

Regression coefficient D /GPa

(b)

ATi

CRu

A25-Al AAl

CRe

AW

CTa

AMo

CTi

ACr

C25-Al CAl CW

ACo

CMo

T3

CCr

T2

CCo

T -20

0

20

40

-0.1

60

-0.05

0

0.05

Regression coefficient D

Regression coefficient D /GPa

Fig. 8 Regression coefficients of the longitudinal elastic modulus shown in Eq. (6), a constant D0, b concentration of i-th element Dci , c temperatures DT n and product of concentration of i-th element by temperature DAi

modulus of CMSX-4 predicted by Eq. (6) for the c phase is affected by the value of the regression coefficient of Ti. As Fig. 8 shows, the regression coefficient of Ti is extremely positively larger than any other elements, indicating its lower accuracy of the value in the regression coefficient of Ti. Among c phase samples used for regression analysis, only TMS-82 + contains small amount of Ti, 0.16 at.%. This small range of Ti in regression analysis of Eq. (6) for the c phase might be the reason. To obtain more accurate regression coefficients especially for Ti, elastic modulus measurement of samples containing larger amount of Ti and

regression analysis by including these data should be performed in future.

Contribution of Composition on Elastic Modulus at 900 °C Equation (6) contains temperature terms such as product of concentration of i-th element and temperature Ai in addition to n-th power of temperature T n . Therefore, composition contribution on elastic modulus at specific temperature is not

Table 2 Compositions of the individual c and c′ phases of CMSX-4 in the references to predict their elastic modulus by Eq. (6) Phase c

c′

Sample

Co

Composition (at.%, Ni;bal.) Cr

Mo

W

Al

Ti

Ta

Hf

Re

Ru

Reference of composition

Reference of elastic modulus

CMSX-4 modeled

18.6

20.0

0.8

2.7

–

–

–

–

3.0

–

[21]

[21]

CMSX-4

16.3

16.5

0.7

2.6

4.5

0.3

0.7

0.01

2.4

–

Prediction by ADP at 900 °C

[22]

CMSX-4 modeled

–

–

–

–

23.0

2.0

1.0

–

–

–

[21]

[21]

CMSX-4

6.5

2.6

0.2

1.8

17.0

1.8

3.0

0.05

0.2

–

Prediction by ADP at 900 °C

[22]

Equations to Predict the Elastic Modulus …

321

clear for the purpose of alloy design. To make clear the role of composition in the longitudinal elastic modulus, regression analysis using the longitudinal elastic modulus obtained by Eq. (6) at 900 °C was executed. Composition shown in Table 1 and referenced Ni–Al samples [19, 20], and their longitudinal elastic modulus predicted by Eq. (6) were used together. Regression equations used in this analysis can be expressed as Eq. (7). X E ¼ E0 þ ðDci ci Þ ð7Þ i

In the same way as Eq. (6), the term cAl of Eq. (7) was modified into 25-cAl in the equation for the c′ phase. The constant term of Eq. (7) for the c phase means the longitudinal elastic modulus of pure Ni at 900 °C, on the other hand, the constant term of Eq. (7) for the c′ phase means that of Ni3Al at 900 °C as physical meaning, respectively. Regression coefficients of Eq. (7) are shown in Fig. 9. Figure 9a shows constant coefficients, on the other hand, Fig. 9b shows coefficients of i-th element. In Fig. 9b, regression coefficients of Re, Ta, Ti, Al, and Mo show large negative values in the equation of the c phase. This indicates addition of Re, Ta, Ti, Al, and Mo could reduce the longitudinal elastic modulus in the c phase. On

(a)

γ phase γ phase

0 0

(b)

50 100 Regression coefficient D /GPa

150

CRu CRe CTa CTi C25-Al CAl CW CMo CCr CCo -4

-2

0

2

4

6

8

Regression coefficient D /GPa Fig. 9 Regression coefficients of the longitudinal elastic modulus at 900 °C shown in Eq. (7), a constant D0 , b concentration of i-th element Dci

the other hand, regression coefficients of Ru, Re, Ta, Ti, Al, W, and Mo in the equation of the c′ phase tend to show large positive values. This indicates Ru, Re, Ta, Ti, Al, W, and Mo could increase the longitudinal elastic modulus in the c′ phase. These analyses indicate that by controlling to specific amounts of alloying elements, the elastic misfit could easily be changed, although we have to consider the effect of partitioning ratio of each element between c and c′ phases [33] to strictly understand the role of each element on the elastic misfit. This guideline of the elements, especially for Ti, could be evolved by performing regression analysis using Ti-rich alloys.

Conclusions In this research, we established regression equations of the longitudinal elastic modulus of the individual c and c′ phases for the multi-component Ni-base superalloys by using newly analyzed elastic modulus in this report and referenced values. Furthermore, to understand effects of composition on the longitudinal elastic modulus of the individual c and c′ phase at 900 °C, an analysis was executed by using ADP and the regression equations for the longitudinal elastic modulus established in this report. Conclusions are listed below. (1) Elastic modulus of the individual c and c′ phases of pure Ni, Ni–Al, Ni–Al–X (X=Ta, W, Re), TMS-26, TMS-82+ , TMS-75, TMS-196, and TMS-238 up to 1100 °C were obtained. (2) Regression equations of the longitudinal elastic modulus including temperature and composition dependence of the individual c and c′ phases for the multi-component Ni-base single-crystal superalloys have been derived using analyzed and referenced elastic modulus. (3) By regression analysis of the longitudinal elastic modulus at 900 °C, it was found that Re, Ta, Ti, Al, and Mo tend to decrease the longitudinal elastic modulus in the c phase. On the other hand, Ru, Re, Ta, Ti, Al, W, and Mo tend to increase the longitudinal elastic modulus in the c′ phase. Still there is a room for improving the accuracy of regression coefficients of each element on the longitudinal elastic modulus. We are planning to improve this accuracy by adding newly analyzed elastic modulus and execute regression analysis. Updated regression equations of the elastic modulus are going to be integrated into ADP [5]. This integration could realize control of the elastic misfit. Moreover, this accelerates development of new optimum

322

Ni-base single-crystal superalloys by assisting formation of the c/c′ raft structure having larger aspect ratio of the c′ phase. Besides, the proposed equations can also be used in various kinds of simulation including phase field methods, first principles calculation, and improving the prediction of mechanical properties by better simulating the elastic stress field based on actual elastic modulus of Ni-base superalloys having multi-component elements. Acknowledgements This work was supported by Japan Science and Technology (JST), under the Advanced Low Carbon Technology Research and Development Program (ALCA) project “Development of Direct and Complete Recycling Method for Superalloy Turbine Aerofoils,” JST ALCA Grant Number JPMJAL1302, Japan.

References 1. Pearson DD, Lemkey FD, Kear BH (1980) Stress coarsening of c′ and its influence on creep properties of a single crystal superalloy. Superalloys 1980, Champion, Pennsylvania. TMS, 513–520. 2. Yamagata T, Harada H, Nakazawa S, Yamazaki M (1986) Effect of Ta/W ratio in c′ phase on creep strength of Nickel-base single crystal superalloys. Trans. Iron Steel Inst. Jpn. 26:638–641. https:// doi.org/10.2355/isijinternational1966.26.638. 3. Nathal MV, Mackay RA (1987) The stability of lamellar c-c′ structures. Mater. Sci. Eng. 85:127–138. https://doi.org/10.1016/ 0025-5416(87)90474-5. 4. Murakumo T, Koizumi Y, Kobayashi T, Harada H (2004) Creep strength of Ni-base single-crystal superalloys on the c/c′ tie-line. Superalloys2004, Champion, Pennsylvania. TMS, 155–162. 5. Yokokawa T, Harada H, Kawagishi K, Kobayashi T, Yuyama M, Takata Y (2020) Advanced alloy design program and improvement of 6th generation Ni-base single crystal superalloy TMS-238. Superalloys 2020, Champion, Pennsylvania, accepted for publication in the same volume. 6. Yoshitake S, Yokokawa T, Ohno K, Harada H, Yamazaki M (1994) Temperature dependence of gamma/gamma-prime lattice parameter misfit for Ni-base superalloys. Proc. 5th Liege conf. “Materials for Advanced Power Engineering 1994” part 1, Liege, 875–881. 7. Harada H, Ohno K, Yamagata T, Yokokawa T, Yamagata M (1988) Phase calculation and its use in alloy design program for Nickel-base superalloys. Superalloys 1988, Champion, Pennsylvania. TMS, 733–742. 8. Hino T, Kobayashi T, Koizumi Y, Harada H (2000) Development of a new single crystal superalloy for industrial gas turbines. Superalloys 2000, Champion, Pennsylvania. TMS, 729–736. 9. Svetlov IL, Golovko BA, Epshin AI, Abalkin NP (1992) Diffusional mechanism of c′-phase particles coalescence in single crystals of nickel-base superalloys. Scr. Metall. 26:1353–1358. https://doi.org/10.1016/0956-716x(92)90648-x. 10. Kontis P, Collins DM, Cormier J, Raabe D, Gault B (2018) The effect of chromium and cobalt segregation at dislocations on nickel-based superalloys. Scr. Mater. 145:76–80. https://doi.org/ 10.1016/j.scriptamat.2017.10.005. 11. Pineau A (1976) Influence of uniaxial stress on the morphology of coherent precipitates during coarsening – elastic energy considerations. Acta Metall. 24:559–564. https://doi.org/10.1016/00016160(76)90101-2. 12. Nabbaro FRN, de Villiers HL (1995) Physics of creep and creep-resistant alloys. CRC press, Florida.

T. Saito et al. 13. Socrate S, Parks DM (1993) Numerical determination of the elastic driving force for directional coarsening in Ni-superalloys. Acta Metall. Mater. 41:2185–2209. https://doi.org/10.1016/0956-7151 (93)90389-a. 14. Ichitsubo T, Koumoto D, Hirao M, Tanaka K, Osawa M, Yokokawa T, Harada H (2003) Rafting mechanism for Ni-base superalloy under external stress: elastic or elastic-plastic phenomena?. Acta Mater. 51:4033–4044. https://doi.org/10.1016/s13596454(03)00224-6. 15. Eshelby JD (1957) The determination of the elastic field of an ellipsoidal inclusion and related problems. Proc. Roy. London Ser. A 241:376–396. https://doi.org/10.1098/rspa.1957.0133. 16. Osawa M, Shiraishi H, Yokokawa T, Harada H, Kobayashi T (2004) 3D-FEM calculations of rafting in Ni-base superalloys based on high temperature elastic and lattice parameters. Superalloys 2004, Champion, Pennsylvania, TMS, 977–985. 17. Zhou N, Shen C, Mills MJ, Wang Y (2008) Contributions from elastic inhomogeneity and from plasticity to c′ rafting in single-crystal Ni-Al. Acta Mater. 56:6156–6173. https://doi.org/ 10.1016/j.actamat.2008.08.027. 18. Zhou N, Shen C, Sarosi PM, Mills MJ, Pollock TM, Wang Y (2009) c′ rafting in single crystal blade alloys: a simulation study. Mater. Sci. Technol. 25:205–212. https://doi.org/10.1179/ 174328408x361472. 19. Prinkhodko SV, Carnes JD, Isaak DG, Ardell AJ (1998) Elastic constants of a Ni-12.69 at.% Al alloy from 295 to 1300 K. Scr. Mater. 38:67–72. https://doi.org/10.1016/s1359-6462(97) 00448-x. 20. Prinkhodko SV, Carnes JD, Isaak DG, Yang H, Ardell AJ (1999) Temperature and composition dependence of the elastic constants of Ni3Al. Metall. Mater. Trans. A 30:2403–2408. https://doi.org/ 10.1007/s11661-999-0248-9. 21. Siebörger D, Knake H, Glatzel U (2001) Temperature dependence of the elastic moduli of the Nickel-base superalloy CMSX-4 and its isolated phases. Mater. Sci. Eng. A 298:26–33. https://doi.org/10. 1016/s0921-5093(00)01318-6. 22. Dye D, Coakley J, Vorontsov VA, Stone HJ, Rogge RB (2009) Elastic moduli and load partitioning in a single-crystal nickel superalloy. Scr. Mater. 61:109–112. https://doi.org/10.1016/j. scriptamat.2009.03.008. 23. Okamoto H (1993) Al-Ni (Aluminum-Nickel). J. Phase. Equilibria 14:257–259. https://doi.org/10.1007/s11669-004-0163-0. 24. Miyazaki S, Murata Y, Morinaga M (1994) Site occupancy of Re atoms in Ni3Al (c′) and c-c′ phase equilibria in Ni-Al-Re ternary alloys. Tetsu-to-Hagané 80:73–77. https://doi.org/10.2355/ tetsutohagane1955.80.2_161. 25. Kobayashi T, Harada H, Yamazaki M (1987) Nickel-base single crystal superalloys developed for a national project in Japan. Proc. Tokyo Int. Gas Turbine Congress Part3, Tokyo, Gas Turbine Society of Japan, 230–243. 26. Kobayashi T, Koizumi Y, Nakazawa S, Yamagata T, Harada H (1997) Design of high rhenium containing single crystal superalloys with balanced intermediate and high temperature creep strengths. Proc. of the fourth international Charles Parsons turbine conference on “Advances in Turbine Materials, Design and Manufacturing”, Newcastle upon Tyne, Institute of Materials, 766–773. 27. Sato A, Harada H, Yeh AC, Kawagishi K (2008) A 5th generation SC superalloys with balanced high temperature properties and processability. Superalloys 2008, Champion, Pennsylvania, TMS, 131–138. 28. Kawagishi K, Yeh AC, Yokokawa T, Kobayashi T, Koizumi Y, Harada H (2012) Development of an oxidation-resistant high-strength sixth-generation single-crystal superalloy

Equations to Predict the Elastic Modulus … TMS-238. Superalloys 2012, Champion, Pennsylvania, TMS, 189–195. 29. Demarest Jr. HH (1971) Cube-resonance method to determine the elastic constants of solids. J. Acoust. Soc. Am. 49:768–775. https://doi.org/10.1121/1.1912415. 30. Ohno I (1976) Free vibration of a rectangular parallelepiped crystal and its application to determination of elastic constants of orthorhombic crystals. J. Phys. Earth 24:355–379. https://doi.org/ 10.4294/jpe1952.24.355. 31. Yasuda H, Koiwa M (1991) Determination of elastic constants of single crystals with cubic symmetry by the rectangular

323 parallelepiped resonance method. J. Phys. Chem. 52:723–729. https://doi.org/10.1016/0022-3697(91)90174-x. 32. Siegel S, Quimby SL (1936) The variation of Young’s modulus with magnetization and temperature in nickel. Phys. Rev. 49:663– 670. https://doi.org/10.1103/physrev.49.663. 33. Harada H, Yokokawa T, Ohno K, Yamagata T, Yamazaki M (1990) Calculation of gamma-prime/gamma equilibrium in multi-component nickel-base superalloys. Proc. High Temperature Materials for Power Engineering 1990, Liege, European Committion, 1387–1396.

Crystal Plasticity Mechanism of Temperature-Dependent Crack Propagation in a Single Crystal Nickel-Based Superalloy Xiaosheng Chen, Motoki Sakaguchi, Shiyu Suzuki, Hirotsugu Inoue, and Masakazu Okazaki

Abstract

Temperature-dependent fatigue crack propagation in a Ni-based single crystal superalloy was experimentally and numerically investigated in a single crystal Ni-based superalloy. Fatigue crack propagation tests at room temperature 300, 450, and 700 °C were conducted using four types of compact specimens with different combinations of crystal orientations in loading and crack propagation directions. It was revealed in the experiments that the crack propagated along slip planes in crystallographic cracking manner at room temperature, while the cracking mode transitioned from the Mode I to crystallographic cracking at 300, 450, and 700°C. Mode I stress intensity factor range DKI values at the transitions depended on the testing temperature as well as crystal orientation. To interpret these temperature-dependent crack propagation, a crystal plasticity finite element analysis was conducted by taking into account the 3D inclined crack plane and the activity of slip planes in front of the crack. Slip plane activity, proposed as a damage parameter, could rationalize the fatigue crack propagation rates both during the crystallographic and Mode I cracking. It has been found that crack propagation resistance for crystallographic cracking is more or less the same at low temperature, X. Chen
M. Sakaguchi (&)
S. Suzuki
H. Inoue Tokyo Institute of Technology, O-okayama 2-12-1, Meguro-ku, Tokyo, Japan e-mail: [email protected] X. Chen e-mail: [email protected] S. Suzuki e-mail: [email protected] H. Inoue e-mail: [email protected] M. Okazaki Nagaoka University of Technology, Kamitomioka 1603-1, Nagaoka, Niigata, Japan e-mail: [email protected]

while that for Mode I cracking decreases with the increase of the temperature. This damage parameter also provided an explanation of the critical condition that induces the transition from Mode I to crystallographic cracking. Keywords











Single crystal Ni-based superalloy Fatigue crack Crystal plasticity Crystal orientation Finite element analysis

Introduction Over the last three decades, many studies have been carried out to predict the remaining life of superalloy components based on fracture mechanics focusing on crack propagation behavior [1–12]. Two distinctive failure modes have been found in fatigue crack propagation in single crystal and directionally solidified superalloys, which are dependent on both testing frequency and temperature. At higher temperature, the Mode I cracking is dominant where fatigue crack propagates in the direction perpendicular to the maximum applied principal stress [1–3, 10, 11]. At lower temperature, fatigue crack propagation takes place in the Stage I manner along the crystallographic slip planes inclined to the loading axis [1, 4–10]. This crystallographic Stage I cracking is also apparent at even higher temperatures if the crack size is similar to or smaller than the grain size [6, 7]. Although some pioneering studies have provided phenomenological interpretations of these temperaturedependent fatigue crack propagation [3, 4], several issues have not been fully clarified yet, especially regarding the role of anisotropic plastic deformation that localizes on the specific slip planes in face-centered cubic (FCC) crystals. Based on the anisotropic plastic deformation on {111} slip planes, crack path has been predicted [13] and secondary crystallographic orientation-dependent deformation has been

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_31

324

Crystal Plasticity Mechanism of Temperature Dependent …

revealed [14]. However, the correlation between the anisotropic plastic deformation and fatigue crack propagation is still unclear, especially regarding the effect of temperature. To interpret the temperature-dependent crack propagation and intrinsic driving force in the crystallographic cracking and the typical Mode I cracking, precise examination of slip activities is required taking account of (1) the geometry of the 3D inclined crack, (2) elastic anisotropy, and (3) anisotropic plastic deformation on octahedral {111} slip planes in FCC single crystal. In this study, temperature-dependent fatigue crack propagation in a single crystal Ni-based superalloy was investigated. At first, crack propagation tests at room temperature, 300, 450, and 700 °C were conducted using four types of compact (C(T)) specimens with different combinations of primary and secondary orientations. Then, a crystal plasticity finite element analysis was carried out, taking into account of the 3D inclined crack plane and the activity of slip planes in front of the crack.

Experimental Procedure A material employed in this study was a single crystal Ni-based superalloy, which is partially modified from CMSX-4 as shown in Table 1. Heat treatments were carried out, consisting of solution treatment of 1277 °C  2 h + 1288 °C  2 h + 1296 °C  3 h + 1304 °C  3 h + 1313 °C  2 h + 1316 °C  2 h + 1277 °C  2 h, followed by aging treatment of 1140 °C  6 h + 871 °C 20 h. After the heat treatments, compact (C(T)) specimens were machined from the plates. The geometry of the C (T) specimen is 20.0 mm (width)  19.2 mm (height) 1.0 mm (thickness), which is the same with the previous study [8]. The length between notch tip and loading hole is 3.5 mm. Four types of C(T) specimens with different crystal orientations were prepared, changing the combination of primary (loading direction) and secondary orientations (crack propagation direction) as shown in Fig. 1. Here, the first and second Miller indices for the specimen are the primary and secondary orientations, respectively. The surfaces of all the specimens were polished to a mirror-like finish using emery paper and Al2O3 powder before the tests. Fatigue crack propagation tests were conducted at room temperature (RT) 300, 450, and 700 °C, using an electrohydraulic machine coupled with an induction heating system. Temperature on the specimen’s surface was monitored and controlled by type K thermocouple. Temperature

Table 1 Chemical composition of the single crystal Ni-based superalloy in this study (wt%)

325

distribution in the gauge section was arranged within ±5 °C. Loading amplitude was kept constant during the test (DKI increasing test) under a loading ratio of R = 0.4 with a loading frequency at 10 Hz. Similar to previous study [8], DKI is calculated based on the ASTM E647 standard, which is defined as: DP 2þa DKI ¼ pffiffiffiffiffi ð0:886 þ 4:64a B W ð1  aÞ3=2  13:32a2 þ 14:72a3  5:6a4 Þ;

ð1Þ

where a equals a/W, and DP, B, W, and a are loading range, specimen thickness, specimen width, and projected crack length, respectively. As projected crack length is adopted in the calculation, the crack path does not affect the value of DKI. Before the test, a 0.2 mm pre-crack at room temperature and a 0.2 mm pre-crack at tested temperature were introduced to eliminate the influence of the initial notch. After pre-crack, fatigue crack propagation tests were started at DKI = 13 MPa m1/2. Two digital microscopes (KEYENCE VHX-5000 and SHODENSHA TG200HD-Me) were used to observe the crack on the front and back surfaces and measure the corresponding projected crack length.

Experimental Results Fracture Surface Figure 2 shows the fracture surfaces and corresponding 3D schematic illustrations for the four specimens (a) , (b) , (c) , and (d) tested at (i) RT, (ii) 300 °C, (iii) 450 °C, and (iv) 700 °C. In the 3D illustrations, Mode I planes are green, while {111} planes are distinguished by blue (plane 1), purple (plane 2), red (plane 3), and yellow (plane 4) according to the illustrations in Fig. 1. Figure 2a shows the fracture surface for specimen. At RT (Fig. 2a–(i)), the crack propagates along slip planes from the very beginning and sometimes deflects from one slip plane to another. At 300 °C (Fig. 2a– (ii)), 450 °C (Fig. 2a–(iii)), and 700 °C (Fig. 2a–(iv)), the crack initially propagates on the Mode I plane in the low DKI regime and then transitions to the crystallographic cracking along the {111} slip plane at a certain DKI level. This transition begins at the surface of DKI of 14.0 MPa m1/2 at 300 °C, 17.3 MPa m1/2 at 450 °C, and 28.0 MPa m1/2 at 700 °C. At all three temperatures, the crystallographic crack

Co

Cr

W

Al

Ti

Ta

Mo

Hf

Re

Ni

9.6

6.5

6.4

5.7

1

6.6

0.6

0.1

3

Bal.

326

X. Chen et al.

Fig. 1 Schematic illustration for octahedral slip planes in four types of C(T) specimens

gradually expands and the transition process completes when DKI reaches 16.8 MPa m1/2 at 300 °C, 19.4 MPa m1/2 at 450 °C and 31.0 MPa m1/2 at 700 °C. Then, the fully crystallographic cracking, covering the entire thickness of specimen, takes place along the slip planes 3 and 4. The fracture surfaces in specimens are shown in Fig. 2b. At RT (Fig. 2b-(i)), the crack propagates along plane 1 (blue) and plane 4 (yellow) without any deflections. The crystallographic crack observed here is so-called factory roof crack with Mode I and III components. At 300 °C (Fig. 2b-(ii)), the transition from Mode I to crystallographic cracking starts from pre-crack and finishes at DKI = 15.9 MPa m1/2. At 450 °C (Fig. 2b-(iii)) and 700 °C (Fig. 2b-(iv)), the crack initially propagates in Mode I manner after the pre-crack. Then, crystallographic crack is initiated at DKI = 16.3 MPa m1/2 and 23.0 MPa m1/2 at 450 °C and 700 °C, respectively. They expand to the interior after DKI = 22.6 MPa m1/2 and 31.0 MPa m1/2, and fully crystallographic cracking occurs at DKI = 23.6 MPa m1/2 at 450 °C and DKI = 52.0 MPa m1/2 at 700 °C. In the primary orientation (Fig. 3(c) and (d)). This trend is attributed to a unique c/c′ microstructure aligned in < 100 > orientation as discussed in the previous study [8]. Comparing the da/dN during crystallographic cracking between and specimen, shows higher da/dN at RT, 300, and 450 °C. This trend generally agreed with the result of the previous study using specimens of a single crystal NKH-304 [7, 9]. This result means that the driving force for the crystallographic cracking largely depends on the crystal orientation of the specimen even when the stress intensity factor range at the crack tip is set to the same level. An evaluation of the intrinsic driving force is the main target of the following sections based on a crystal plasticity analysis.

Crystal Plasticity Analysis The fatigue crack propagation tests in the previous section revealed that crystallographic cracking along slip planes contains all Mode I, II, and III components, and the resultant da/dN was influenced by crystallographic primary and secondary orientations. The da/dN could not be explained in a unified manner by the conventional Mode I stress intensity factor range. To evaluate the temperature-dependent cracking behavior and quantify their intrinsic driving forces, precise examination of slip activities is required to take account of (1) the actual geometry of the 3D inclined crystallographic crack, (2) crystallographic elastic anisotropy, and (3) anisotropic plastic deformation localized at the specific slip systems at the crack front [13, 14]. To examine these factors, a finite element (FE) analysis coupled with a crystal plasticity framework has been considered. The previous proposed constitutive equations for crystal plasticity (CP) are adopted and given as [15, 16]: Fig. 2 Fracture surfaces and corresponding 3D schematic illustrations at RT, 300, 450, and 700 °C for four kinds of specimens (a) , (b) , (c) , and (d)

328

X. Chen et al.

Fig. 3 Crack propagation rate in the four specimens (a) , (b) , (c) , and (d) at RT, 300, 450, and 700 °C as a function of DKI

r

r¼C :D

n  X

 C : Pk þ W k
r  r
W k c_ k

ð2Þ

k¼1

 k  k 1=m s s  ; ð3Þ s0  s0      here Pk ¼ 1=2 nk bk þ bk nk and W k ¼ 1=2 nk bk  bk nk . c_ k ¼ c_ 0 sgn

r

k represents twelve slip systems (k = 1, 2,…, 12). r, r, C, D, nk and bk are Jaumann stress rate, stress, tensor of elastic moduli, deformation rate tensor, unit direction vector of slip plane normal, and unit vector of slip direction, respectively. c_ k , sk , c_ 0 , s0 and m are shear strain rate of slip system k, shear stress of slip system k, reference shear strain rate, yield shear stress, and rate sensitivity coefficient, respectively. A user-defined material subroutine UMAT [17], which is programmed on the basis of above constitutive equations, is

incorporated with Abaqus 2017 for finite element implicit analysis. Figure 4 shows FE models for the and specimens during crystallographic cracking at RT, with careful consideration of actual 3D crack tip geometries. Due to symmetry along the thickness direction, half the thickness of the specimen is modeled, namely 0.5 mm. In all models, a symmetry boundary condition is applied to the back surface. For the model (Fig. 4a), the crystallographic crack plane is modeled as along a single yellow slip plane, which is inclined 45° to both the propagation and thickness directions. For the model (Fig. 4(b)), the crystallographic crack plane is modeled as along stripe-like yellow and blue slip planes from the initial notch. These cracks are modeled as a “seam crack” assuming that the crack propagates along the slip plane without deflection. For the models at 300 °C,

Crystal Plasticity Mechanism of Temperature Dependent …

329

Fig. 4 Finite element meshes for crystal plasticity analysis during crystallographic cracking in and specimens at RT

450 °C, and 700 °C, crystallographic cracks are reproduced based on the fracture surface as in Fig. 2. The number of elements and nodes in all the models are approximately 60,000 and 240,000, respectively. The minimum mesh size in front of the crack tip is 2 lm. This has been confirmed to be fine enough to keep the calculation stable and convergent. The element type is chosen as C3D20R to ensure high precision of the calculation results. Critical resolved shear stresses (CRSS), s0 , at RT, 300, 450, and 700 °C are shown in Table 2 which accords to [18]. Stiffness at RT, 300, 450, and 700 °C is shown in Table 3, which are obtained from [19]. For the rate sensitivity coefficient m in FCC single crystal like Al and Cu, the order of magnitude is usually −2 and the simulated stress-strain response is not quite affected by c_ 0 in this case. Due to difficulties in determining actual c_ 0 and m, m is chosen as 1  10−2 and c_ 0 is chosen as 10−5 s−1 for the Ni-based single crystal superalloy in this study. In this study, work hardening is not considered for simplicity. A fatigue damage parameter, namely slip plane activity, proposed in a previous study [9] is also adopted in this study. This parameter evaluates fatigue damage on the slip plane as    X 3i Z  j j Fi ¼  ð4Þ c b dr; i¼ 1; 2; 3; 4;  j¼3i2

Fig. 5 Relationship between crack propagation rate and Fmax at RT, 300, 450, and 700 °C during the crystallographic cracking in the and specimens

where Fi represents fatigue damage on slip plane i. cj and bj are corresponding slips and slip directions on slip plane i, respectively. The summation is regarded as slip plane activity, which considers the activities of all the three slip directions in a slip system [9]. Fi for slip plane i can be calculated by integration along the distance from the crack tip on the slip plane. Both four types of slip planes and three types of slip directions are individually considered in Fi.

Results of the Crystal Plasticity Analysis Evaluations of da/dN Based on the Slip Plane Activity (a) Fully crystallographic cracking The driving force of crystallographic cracking can be correlated with the maximum value among the four types of slip planes (Fmax), because the crack propagation along slip

Table 2 Critical resolved shear stress at RT, 300, 450, and 700 ° C [18]

Temperature

RT

300 °C

450 °C

700 °C

CRSS (MPa)

389

390

398

402

Table 3 Stiffness C11, C12, and C44 at RT, 300, 450, and 700 °C [19]

Temperature

C11 (GPa)

C12 (GPa)

C44 (GPa)

RT

248.8

158.5

129.8

300 °C

233

147.5

121.5

450 °C

231.6

151.6

115.4

700 °C

212.2

140.3

106.6

330

X. Chen et al.

plane is governed by slip activity of the most activated slip plane with little influence from the other slip planes. Figure 5 shows relationship between computed Fmax values and experimental da/dN in and specimens under RT, 300, 450, and 700 °C. It can be seen in Fig. 5 that the plots of and specimens at RT, 300, and 450 °C are evaluated within a very narrow band, which indicates that da/dN–Fmax relationship is independent of temperature and crystallographic orientation up to 450 °C. Therefore, crack propagation rate during crystallographic cracking is essentially governed by the Fmax on the corresponding slip plane. Generally speaking, crack propagation rate is a result of competing balance between the driving force and propagation resistance, which can be roughly expressed in the following Eq. (5); da ¼ CðFmax  Rcr Þm : dNcr

ð5Þ

Here, C and m are the material parameters. Rcr is the propagation resistance to the crystallographic cracking, which should be distinguished from the resistance to the Mode I cracking, RI. Based on the computed result in Fig. 5 where da/dNcr at RT, 300, and 450 °C is almost the same under the same driving force Fmax, it can be concluded that the propagation resistance to crystallographic cracking is considered as the same between RT, 300, and 450 °C. On the other hand, data plots at 700 °C in Fig. 5 are smaller than the data band for lower temperatures. This can be attributed to a very different mechanism of dislocation motion at 700 °C, which is influenced by activation of

cube-slip system and climb motion associated with creep deformation [20]. They can lower slip plane activity along the octahedral slip system and reduce the driving force Fmax in Eq. (5). These complicated slip mechanisms should be considered in the CP constitutive equation in Eqs. (2) and (3) to evaluate the crystallographic cracking at higher temperatures. Due to that at present the effect of oxidation is unclear, the effect of oxidation on crack propagation is not discussed, which need to be analyzed in the future with fatigue crack propagation test at high temperature in vacuum environment. (b) Mode I cracking In this subsection, the da/dN during pure Mode I cracking at lower DKI regime is evaluated based on the computed slip activity on the Mode I plane. Based on the multiple slip mechanism by Forsyth [21], a damage parameter for Mode I cracking can be given as: X

F¼

4 X

Fi ;

where Fi is the same of that in Eq. (4), but the integration is on the Mode I crack plane. Multiple slip activity is considered in Eq. (6) because the activation of multiple slip systems should contribute to Mode I cracking. This damage P parameter F considers a summation of 12 slip systems on the Mode I crack plane, in which the stress state is basically equivalent to the conventional Mode I stress fields. This P means F has a similarity to DKI, except for the fact Pthat that F considers elastic anisotropy and the individual activities of 12 slip systems. Figure 6 shows the relationship between the experimental P da/dN and the computed F for different specimens during Mode I cracking at different temperatures. It is found that plots for different specimens P are evaluated within a data band. With the increase of F, the crack propagation rate for Mode I cracking also increases regardless of crystalloP graphic orientation. Thus, F can explain the da/dN of Mode I cracking with an accuracy, better than that of DKI, as shown in Fig. 3. Similar to Eq. (5), a relationship between P da/dN and F for Mode I cracking can be given as: X n da ¼D F  RI ; dNI

Fig. 6 Relationship between crack propagation rate and the Mode I cracking in all specimens

P

F during

ð6Þ

i¼1

ð7Þ

where D and n are material parameters, and RI is the propagation resistance to the Mode I cracking. For Ni-based single crystal superalloy, Mode I cracking occurs mainly within the c phase [8, 10]. With the increase of temperature, the strength of the c phase decreases and propagation resistance RI decreases, which should result in higher da/dN.

Crystal Plasticity Mechanism of Temperature Dependent …

331

Fig. 7 Distribution of computed Fmax values for , and at the critical condition that induced the transition at 450 °C

Transition from the Mode I to Crystallographic Cracking In the three specimens of , and , the transition from the Mode I to crystallographic cracking took place at certain DKI levels as shown in Fig. 2. This transition DKI at 450 °C was 17.3 MPa m1/2 in , 22.6 MPa
m1/2 in , and 29.0 MPa m1/2 in . Because the resistance to the transition from Mode I to crystallographic cracking is related only to the intrinsic slip property of a FCC crystal, rather than the crystal orientation of the specimen, the critical Fmax value at the transition should be identical in all the specimens. CP analysis during the transition process is also conducted based on the FE models, carefully simulating the actual geometry of crack. Figure 7 shows Fmax values at the critical DKI points for , and specimens. Here, Fmax values are computed in the crack front of the FE model and plotted as a function of the distance from the specimen surface (Fig. 7a). It is found in Fig. 7 that the Fmax values in three specimens are of the same order of Fmax ¼ 1  10  105 m. This value might be the critical condition that induces the transition and also be the intrinsic resistance to the crystallographic cracking, namely Rcr in Eq. (4). It should be noted in Fig. 7 that the Fmax values in the specimen at the critical point are the largest in all of the cases. In front of the inclined crack in the specimen, crystallographic cracking can be initiated along both slip planes 3 and 4, which are equivalent due to the symmetry condition. In contrast, in front of the inclined crack in the specimen,

crystallographic cracking can only be initiated along a single slip plane due to the geometry of the Mode I plane and slip planes. As the intersections of non-coplanar dislocations are the main cause of work hardening, work hardening in front of the inclined crack in the specimen can be more severe than that in the and specimens due to severe dislocation intersections between two possible slip planes in the specimen. Because no work hardening has been considered in the CP constitutive equation in this study, the actual Fmax in the specimen is relatively smaller.

Conclusions Fatigue crack propagation tests at RT, 300, 450, and 700 °C and corresponding crystal plasticity finite element analysis were conducted for a Ni-based single crystal superalloy to study the temperature-dependent fatigue crack propagation. Conclusions are summarized as follows: (1) The fatigue crack propagated in the crystallographic cracking mode at room temperature, while the crack transitioned its propagation mode from the Mode I to crystallographic cracking at 300, 450, and 700 °C. The transition of cracking mode occurred at higher DKI values at higher temperature. (2) Crystal plasticity analysis was conducted for the Mode I cracking, the transition, and the crystallographic cracking to investigate the slip activity of an octahedral slip system in front of the crack tip. A damage parameter, which accounted for the fatigue damage on

332

X. Chen et al.

the individual slip plane, rationalized the fatigue crack propagation rates during Mode I cracking and crystallographic cracking. It has been found that crack propagation resistance for crystallographic cracking is more or less the same at low temperature, while crack propagation resistance for Mode I cracking decreases with the increase of the temperature. This damage parameter also quantified the critical condition that induces the transition from Mode I to crystallographic cracking in a single crystal superalloy.

References 1. Telesman J, Ghosn LJ (1989) The unusual near-threshold FCG behavior of a single crystal superalloy and the resolved shear stress as the crack driving force. Eng. Fract. Mech. 34(5):1183–1196. 2. Lerch BA, Antolovich SD (1990) Fatigue Crack-Propagation Behavior of a Single Crystalline Superalloy. Metall. Trans. A 21 (8):2169–2177. 3. Telesman J, Ghosn LJ (1996) Fatigue Crack Growth Behavior of PWA 1484 Single Crystal Superalloy at Elevated Temperatures. J. Eng. Gas Turb. Power 118(2):399–405. 4. Liu L, Husseini NS, Torbet CJ, Lee WK, Clarke R, Jones JW, Pollock TM (2011) In situ synchrotron X-ray imaging of high-cycle fatigue crack propagation in single-crystal nickel-base alloys. Acta Mater. 59(13):5103–5115. 5. Sakaguchi M, Tsuru T, Okazaki M (2012) Fatigue crack propagation in thin-wall superalloys component; Experimental investigation via miniature CT specimen. Paper Presented at the 12th International Symposium on Superalloys, Champion, Pennsylvania, 9–13 September 2012. 6. Okazaki M, Sakaguchi M, Yamagishi S (2013) Subcritical crack growth on crystallographic planes in a Ni-base superalloy: Relevance to orientations. Procedia Eng. 55:677–684. 7. Sakaguchi M, Komamura R, Hosaka R, Inoue H (2016) Stage I fatigue crack propagation in a single crystal and a directional solidified Ni-base superalloy. Paper Presented at the 13th International Symposium on Superalloys, Champion, Pennsylvania, 11– 15 September 2016.

8. Suzuki S, Sakaguchi M, Inoue H (2018) Temperature dependent fatigue crack propagation in a single crystal Ni-base superalloy affected by primary and secondary orientations. Mater. Sci. Eng. A 724(2):559–565. 9. Sakaguchi M, Komamura R, Chen X, Higaki M, Inoue H (2019) Crystal plasticity assessment of crystallographic Stage I crack propagation in a Ni-based single crystal superalloy. Int. J. Fat. 123:10–21. 10. Neu RW (2019) Crack paths in single-crystal Ni-base superalloys under isothermal and thermomechanical fatigue. Int. J. Fat. 123:268–278. 11. Suzuki S, Sakaguchi M (2020) Fatigue crack retardation associated with creep deformation induced by a tension hold in a single crystal Ni-base superalloy. Scripta Mater. 178:346–350. 12. Chen X, Sakaguchi M (2020) Transition behavior from Mode I cracking to crystallographic cracking in a Ni-base single crystal superalloy. Int. J. Fat. 132:105400. 13. Sabnis PA, Forest S, Cormier J (2016) Microdamage modelling of crack initiation and propagation in FCC single crystals under complex loading conditions. Computer Methods in Applied Mechanics and Engineering 312:468–91. 14. Sabnis PA, Mazière M, Forest S, Arakere NK, Ebrahimi F (2012) Effect of secondary orientation on notch-tip plasticity in superalloy single crystals. International Journal of Plasticity 28:102–23. 15. Peirce D, Asaro RJ, Needleman A (1982) An analysis of nonuniform and localized deformation in ductile single crystals. Acta Metall. 30(6):1087–1119. 16. Asaro RJ (1983) CRYSTAL PLASTICITY. J. Appl. Mech.-T. ASME 50(4B):921–934. 17. Huang Y (1991) A user material subroutine incorporating single crystal plasticity in the ABAQUS finite element program, Harvard University: Cambridge, MA, USA. 18. Kagawa H, Mukai Y (2012) The Effect of Crystal Orientation and Temperature on Fatigue Crack Growth of Ni-based Single Crystal Superalloy. Paper Presented at the 12th International Symposium on Superalloys, Champion, Pennsylvania, 9–13 September 2012. 19. Sieborger D, Knake H, Glatzel U (2001) Temperature dependence of the elastic moduli of the nickel-base superalloy CMSX-4 and its isolated phases. Mat. Sci. Eng. a-Struct 298(1–2):26–33. 20. Fedelich B (2002) A microstructural model for the monotonic and the cyclic mechanical behavior of single crystals of superalloys at high temperatures. Int. J. Plast. 18:1–49. 21. Forsyth P (1962) A two stage process of fatigue crack growth. Paper Presented at the Crack Propagation Symposium, Cranfield, UK 1962.

Micro-mechanisms of Cyclic Plasticity at Stress Concentrations in a Ni-Based Single-Crystal Superalloy Alessandro Piglione, Jian Yu, Jinqian Zhao, Chengbo Xiao, Fionn Dunne, and Minh-Son Pham

Abstract

Ni-based single-crystal superalloys are high-temperature materials used for turbine blades in jet engines. Fatigue damage can pose a major threat to the integrity of such components in operation. Traditionally, TEM-based studies on the fatigue behaviour of superalloys has been studied by investigating cyclic plasticity in the bulk of the material. When the cyclic loads are nominally elastic, however, such investigation may not contribute to the understanding of the alloy’s fatigue behaviour, since plastic micro-strains are confined to regions near stress raisers such as microstructural defects and are therefore randomly distributed. In turn, the plastic micro-strains near the ‘critical’ stress raiser, i.e. the one that acts as nucleation site for the dominant crack, govern fatigue life by inducing early crack initiation, and are therefore the key to capture the material’s fatigue behaviour. Hence, this work is concerned with the experimental characterisation of cyclic plasticity at the initiation site in a Ni-based single-crystal superalloy at 800 °C tested with nominally elastic cyclic loads. Such investigation was carried out by focused ion beam (FIB) lift-outs and subsequent transmission electron microscopy (TEM) studies. It is shown that deformation is significantly more pronounced near the ‘critical’ stress concentration; in addition, deformation is rather homogeneous across large regions surrounding the stress raiser, with remarkably different deformation modes compared to those observed in the bulk of the specimens and from those expected in superalloys tested in similar conditions. The investigation A. Piglione (&)
F. Dunne
M.-S. Pham Department of Materials, Imperial College London, Exhibition Road, London, SW7 2AZ, UK e-mail: [email protected] J. Yu
J. Zhao
C. Xiao Science and Technology on Advanced High Temperature Structural Materials Laboratory, Beijing Institute of Aeronautical Materials, Beijing, 100095, China

of local cyclic plasticity at stress concentrations promises therefore to provide new insight into fatigue crack initiation in Ni-based superalloys. Keywords







Single-crystal superalloys Fatigue concentrations Cyclic plasticity




Stress

Introduction Ni-based single-crystal superalloys are widely used for turbine blades in jet engines thanks to their excellent mechanical properties at elevated temperatures and their resistance to aggressive environments. In operation, turbine blades may experience cyclic loads, both in the high- and low-cycle fatigue regimes. Thus, the fatigue behaviour of single-crystal superalloys has been extensively studied; in particular, their cyclic deformation mechanisms have been thoroughly investigated in previous studies, along with their temperature dependence [1–4]. Traditionally, however, the cyclic deformation of superalloys has been investigated in unspecified regions (the ‘bulk’) of the test pieces, under the assumption that the observed mechanisms would be representative of the deformation of the test pieces in their entirety. Whilst this may be true for large imposed cyclic strains, it is arguable whether this assumption holds when cyclic strains are considerably smaller. In fact, when the test pieces (or the components in operation) are subjected to loads that nominally induce elastic deformation, plastic deformation only occurs in the vicinity of stress raisers, where the stresses become locally greater than the yield strength of the alloy. The accumulation of such plastic micro-strains near the most severe stress raiser, i.e. the ‘critical’ stress raiser, eventually leads to the nucleation of the dominant crack. Stress concentrations are induced by microstructural inhomogeneities, such as cast micro-pores,

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_32

333

334

A. Piglione et al.

which have a variety of shapes and dimensions, thus causing variability in the magnitude and distribution of stresses. In turn, such variable stresses may translate into different deformation micro-mechanisms, which directly affect fatigue crack initiation. It follows that, in order to understand the fatigue behaviour of a superalloy subjected to elastic cyclic loads, deformation mechanisms must be investigated near the critical stress raiser. Whilst plenty of work has been conducted on the bulk cyclic deformation of Ni-based single-crystal superalloys [1–9], deformation microstructures in the vicinity of stress concentrations have not been extensively investigated yet. Hence, this work provides an in-depth study of local cyclic plasticity near the critical stress raiser in a single-crystal superalloy tested in tension/ compression fatigue at 800 °C. A notch was introduced on one of the test pieces to act as the critical stress raiser and facilitate the investigation of local plasticity. Such investigation was carried out by means of focused ion beam (FIB) lift-outs and subsequent transmission electron microscopy (TEM) studies.

Materials and Methods Material The alloy selected for this work is the DD6 Ni-based single-crystal superalloy; its nominal composition is shown in Table 1 [10]. The castings were cylindrical bars measuring 15 mm in diameter and 150 mm in length. Such bars were subjected to the following heat treatment in vacuum: 1290 ° C/1 h + 1300 °C/2 h + 1315 °C/4 h + air cooling (AC), 1120 °C/4 h + AC, 870 °C/32 h + AC. The microstructure of the alloy after heat treatment is shown in Fig. 1a; it consists of an ordered distribution of coherent cuboidal c′ precipitates in the c matrix. In particular, the alloy is characterised by a c′ volume fraction of *70%, and the cube edge length of the precipitates is *450 nm.

Experimental Procedure The tests in this work were carried out on two single-crystal cylindrical test pieces with a gauge diameter of 5.5 mm and a gauge length of 15 mm. The gauge section had a roughness average Ra of 0.4 µm. These test pieces were machined

Table 1 Nominal composition of the DD6 single-crystal superalloy [10]

so as to have their longitudinal loading axis within 10° of the [001] crystallographic orientation of the single-crystal. To facilitate the investigation of localised deformation at stress concentrations, a surface notch was created on one of the test pieces before the test using a multi-axis computer numerical control (CNC) electrical discharge machine (EDM). Such notch was cylindrical in shape, with nominal depth and diameter of 300 µm. Fully reversed (R = −1) fatigue tests were carried out in air at 800 °C, with the load applied uniaxially along the longitudinal direction of the test pieces. The tests were performed in strain-control imposing a triangular waveform with a strain amplitude of 0.6% at a strain rate of 5  10−3 s−1 using a MTS Landmark Servohydraulic Test System equipped with a 12 mm high-temperature extensometer. The imposed strain amplitude was lower than the yield strain of the alloy at 800 °C, resulting in nominally elastic loading. Load/displacement data was recorded for selected cycles in every test. Data from the first and half-life cycles of the smooth test piece is shown in Fig. 1b; it should be noted that the imposed strain amplitude reached the nominal value of 0.6% after a few cycles at slightly lower strain amplitudes. The mechanical behaviour of the notched test piece (not shown here) was not affected by the notch and was therefore analogous to that of the smooth test piece.

Characterisation Post-mortem fractographic analyses were carried out using a Zeiss Auriga scanning electron microscope (SEM) operating at 20 kV. To study microstructural deformation in the bulk of the test pieces, thin foils for transmission electron microscopy (TEM) were prepared by twin-jet electropolishing in a solution of 10% perchloric acid in methanol applying 20 kV at −5 °C with a Struers TenuPol-5. In addition, to investigate deformation mechanisms in the initiation region of the notched specimen, thin lamellae for TEM were prepared performing focused ion beam (FIB) lift-outs from the test piece’s fracture surface, using a FEI Helios. The lift-outs were carried out at least 30 µm away from the notch surface to avoid regions that may have been affected by EDM. TEM and scanning TEM (STEM) investigations were carried out using a Jeol 2100F microscope operating at 200 kV.

Element

Ni

Co

Cr

Mo

W

Al

Ta

Re

Nb

Hf

%wt.

Bal.

9.0

4.3

2.0

8.0

5.6

7.5

2.0

0.5

0.1

Micro-mechanisms of Cyclic Plasticity at Stress Concentrations …

335

Fig. 1 a Transmission electron microscopy (TEM) bright field (BF) micrograph showing cuboidal c′ precipitates orderly dispersed in the c matrix. b Stress/strain data for the first and half-life cycles of the smooth test piece

Results Fatigue Lives and Fractography The fatigue lives of the smooth and notched test pieces were 57,786 and 2786 cycles, respectively. It is therefore apparent that the notch induced a significant reduction in the number of cycles to failure. SEM micrographs of the fracture surfaces of the smooth and notched test pieces are shown in Fig. 2. In particular, Fig. 2a, b shows that crack initiation occurred at a subsurface cast micro-pore in the smooth test piece, in agreement with what is reported in the literature for Ni-based single-crystal superalloys tested in similar conditions [1, 2, 11, 12]. The minimum distance of such pore from the surface was found to be *640 µm. On the other hand, initiation occurred at the notch in the notched specimen (Fig. 2c, d). Hence, it can be argued that the notch reduced the fatigue life of the notched specimen by inducing early crack initiation. In both cases, a large portion of the fracture surface surrounding the initiation region appears rather flat and normal to the loading direction, suggesting that the crack grew on the (001) plane in the first stages of crack propagation. A number of small trenches can be seen around the notch (Fig. 2d); these were created when TEM foils were extracted by FIB lift-outs and will be discussed in detail later in the paper. The actual notch dimensions were measured from Fig. 3d, yielding a maximum notch depth of 329 µm and a cross-sectional area of 86,110 µm2, which accounts for 0.36% of the total cross-sectional area of the test piece.

Deformation in the Bulk TEM investigations carried out in the bulk of the test pieces revealed very inhomogeneous deformation. The vast majority of the regions observed exhibited little to no evidence of deformation, with very low dislocation densities.

Only few regions, arguably in close proximity to stress concentrations (e.g. cast pores, carbides), were characterised by significant dislocation activity. A representative STEM micrograph in this regard is shown in Fig. 3a, where plastic deformation is only observed in the region surrounding the carbide at the centre of the image. In the regions where plastic deformation was found to be more pronounced, dislocation activity was observed to be confined entirely to the c channels, with no evidence of c′ shearing (Fig. 3b). The dislocations that appear to be ‘on top’ of the precipitates are likely to be gliding in the horizontal channels between the c′ cuboids.

Deformation at the Notch Two TEM micrographs obtained via a FIB lift-out in close proximity to the notch are shown in Fig. 4; both micrographs show the same region, imaged using two distinct two-beam conditions. It is apparent that the deformation near the notch is significantly different from that observed in the bulk of the same test piece (Fig. 3), in that high and rather homogeneous dislocation densities are observed across large regions in the TEM foil. In particular, dense dislocation tangles are observed in the vertical c channels (i.e. the ones that run parallel to the loading direction LD); in contrast, the c′ precipitates appear to be generally dislocation-free, with the exception of a few dislocations that run diagonally with respect to the cubes’ cross-section. Such ‘oblique’ dislocations are observed to be all parallel to one another (Fig. 4). Moreover, the comparison between the micrographs shows that there are in fact two sets of such dislocations, with only one set visible in each image. Although the micrographs in Fig. 4 enable the observation of deformation near the notch, their interpretation is made difficult by the overall poor quality of the TEM images. In addition to the contrast arising from dislocations, easily recognisable as thin black lines, a number of features appear black in each micrograph; such

336

A. Piglione et al.

Fig. 2 SEM micrographs of the fracture surfaces of the smooth (a, b) and notched (c, d) test pieces

Fig. 3 STEM-BF micrographs from the bulk of the test pieces. Two distinct regions are imaged using the same two-beam condition; the diffraction vector is shown in (a), as well as the loading direction LD. The beam direction was [001] for both micrographs

Micro-mechanisms of Cyclic Plasticity at Stress Concentrations …

features are unrelated to the deformation of the alloy and are rather associated with the damage induced in the foil by the ion beam during the lift-out. Ion damage is a known issue in TEM foils prepared by FIB lift-outs [13]. To improve the quality of the micrographs, further imaging was conducted in scanning TEM mode (STEM). A STEM micrograph of the same region imaged in Fig. 4 obtained with the same diffraction condition is shown in Fig. 5a. It should be noted that STEM enables the rotation of the micrographs, so that the loading direction (labelled as LD) in Fig. 5 is rotated with respect to that in Fig. 4. A careful examination of Figs. 4 and 5a reveals that the dislocation contrast is not modified in the STEM image, whilst the contrast arising from ion damage has been excluded from the image. The use of STEM enables therefore a clearer observation of deformation in FIB-prepared thin foils, whilst allowing scans of larger areas. Figure 5 includes STEM micrographs obtained from three further FIB-prepared foils, along with an SEM micrograph of the fracture surface (Fig. 5b) showing the locations such foils were extracted from; in particular, the label of each micrograph matches the letter assigned to the respective lift-out site on the fracture surface. Investigating foils extracted from several regions around the notch allows for a complete study of how the notch affects deformation in the initiation region. Figure 5c was obtained from a TEM foil with the same normal as the one used for Fig. 5a and imaged using the same diffraction condition; it shows that the dislocation arrangements previously observed in Fig. 4 are also seen in this region, with thick dislocation tangles in the vertical c channels and a number of parallel dislocations ‘on’ the c′ precipitates. Interestingly, the high-magnification inset of Fig. 5a shows that these ‘oblique’ dislocations are in fact dislocation pairs rather than individual dislocations. These arrangements are further observed in Fig. 5d, where another reflection is used for imaging to confirm the presence of the second set of ‘oblique’ dislocations. Finally, Fig. 5e was obtained from a foil with a different normal and imaged using a cube reflection; the same dislocation

Fig. 4 TEM-BF micrographs from a foil obtained via a FIB lift-out near the notch. The same region is imaged with two distinct two-beam conditions; the respective diffraction vectors are shown in each

337

configurations are once again confirmed, and both sets of ‘oblique’ dislocation pairs can be observed at the same time.

Discussion When the imposed cyclic strains are nominally elastic, plastic strains take place locally in regions of sufficient stress concentration. Once significant local fatigue damage has accumulated, fatigue crack initiation occurs near the most severe stress concentration. For the smooth test piece in this study, in the absence of a surface notch, fatigue crack initiation was linked to the stress concentration induced by a cast pore (Fig. 2a, b). This observation is not surprising, as cast pores are often found to induce fatigue crack initiation in Ni-based single-crystal superalloys tested both in similar conditions [1, 2, 11, 12] and in rather different ones [14–17]. However, the notch was found to act as the preferential site for early crack initiation in the notched specimen, therefore significantly reducing its fatigue life. This is arguably a consequence of the fact that the surface notch was significantly larger compared to the subsurface pore responsible for initiation in the smooth test piece (Fig. 2); hence, the stress was locally raised to a higher level, inducing more pronounced local plasticity and earlier fatigue crack initiation, leading to an overall shorter fatigue life. In other words, the notch played the role of the ‘critical’ stress raiser due to its large size compared to that of the process defects and microstructural inhomogeneities in the test piece. It should be mentioned, however, that the fabrication of the notch via EDM might have induced damage and recrystallisation in a thin layer of material around the notch itself; this, in turn, may have also contributed to the shorter fatigue life of the notched test piece. The nominally elastic loading explains why the vast majority of the regions in the bulk of the test pieces were found to be free of dislocations: plastic deformation occurred only in regions where the stress was locally raised above the

micrograph. The loading direction LD is shown in (a). The beam direction was [110] for both micrographs

338

A. Piglione et al.

Fig. 5 STEM-BF micrographs showing dislocation arrangements in proximity of the notch (a, c, d, e), obtained from four foils extracted via FIB lift-outs in the locations shown by the red circles in the fracture surface micrograph (b); the labels of the red circles in (b) match those of the respective STEM micrographs. Each image was obtained after

tilting the foils to a two-beam condition; the diffraction vector is shown in each image. The beam direction was [110] for (a, c, d) and [100] for (e). The direction of the far-field load is the same for all STEM micrographs and is therefore only shown in (a), labelled as LD

yield strength by process defects or microstructural inhomogeneities. Figure 3a provides an example of one of these occurrences, with plastic deformation developing in the region surrounding a carbide. Although a number of such scattered plastically deformed regions were found, their in-depth investigation would not contribute to the understanding of the fatigue behaviour of this alloy with the test conditions adopted in this work. Indeed, cyclic plasticity must be investigated in the region where it actually developed into the nucleation of the dominant crack, i.e. near the ‘critical’ stress concentration, to study the cracking mechanisms and hence capture the alloy’s fatigue behaviour. Figure 5 shows local plasticity in the regions close to the notch. It is apparent that the alloy’s deformation is remarkably different from that observed in the bulk of the test pieces: firstly, in contrast to very scattered and inhomogeneous deformation, the FIB-prepared foils extracted near the notch were plastically deformed in their entirety. This observation suggests that the notch induced a significant stress concentration in the regions surrounding it, severe enough to induce a rather large plastic zone. Secondly, the deformation observed in each foil was quite homogeneous and evenly distributed across the entire foil. The same deformation modes, in addition, are observed in four FIB-prepared foils extracted in different locations around the notch. This is an observation of significant importance; as previously mentioned, the notch not only raises the stress in

its surroundings, but also locally modifies the stress state. Rather than the nominal uniaxial stress state, the regions close to the notch experience a triaxial stress state, the components of which are dependent on the size and shape of the notch itself. Such modification to the stress state is expected to have a profound influence on the deformation modes in these regions, and hence on crack initiation. Due to the shape of the notch, a spatial variation of such triaxial stress state could be expected; the observation of the same deformation modes in four distinct locations close to the notch, however, indicates that the local stress distribution surrounding the notch was rather homogeneous or, equivalently, that the spatial variation of the stress state was not significant enough to affect the underlying deformation mechanisms. It has so far been shown that the notch acted as crack initiation site and that the stress concentration it generated was such to locally induce homogeneous plastic deformation with well-defined deformation modes. Since crack initiation is intimately related with plastic flow [18], the in-depth characterisation of such deformation modes acting in the initiation region is the key to understand crack initiation mechanisms in this test conditions. It is well known that the yield stress of c′ increases with temperature [19] up to *800 °C, whilst the c matrix becomes increasingly weaker. Thus, at 800 °C it is significantly easier for dislocations to glide in the c matrix,

Micro-mechanisms of Cyclic Plasticity at Stress Concentrations …

bypassing c′ precipitates by thermally assisted mechanisms such as climb and cross-slip [1]. This is found to be particularly true in fatigue, since cyclic plastic strains are normally small and hence easily accommodated in the c channels. For the above reasons and since the imposed cyclic strains in this work were nominally elastic, it would have been expected to only observe plastic deformation in the c channels. Indeed, the c channels in all micrographs shown in Fig. 5 appear heavily deformed, with very dense tangles of dislocations. All micrographs, however, also show the presence of two sets of consistently aligned dislocations within the c′ precipitates. It could be argued that such dislocations may be gliding in the horizontal c channels ‘on top’ of the precipitates, as the ones observed in the bulk in Fig. 3b. This argument can be disputed by observing the micrographs in Fig. 5 and relying on the available literature on superalloys deformation. Indeed, the inset of Fig. 5a shows that these features are actually dislocation pairs. Were such pairs to be gliding in the c channels, this observation would be explained with the well-known dislocation dissociation in fcc lattices on {111} planes, where an a/ 2 dislocation dissociates into two Shockley partials (a/6 and a/6 ) separated by a stacking fault. However, the fringe image characteristic of stacking faults was never observed, regardless of the diffraction condition selected and despite the fact that most studies were carried out with the beam aligned with the [110] orientation, the most suitable one in this regard since two {111} planes are viewed edge-on. Moreover, the simultaneous invisibility of the two dislocations in each pair suggests that they are characterised by the same burgers vector. Hence, it can be argued that the observed pairs are constituted by two a/ 2 dislocations, which paired up in order to shear the c′ precipitates and are separated by an anti-phase boundary (APB). Although c′ shearing is not uncommon in superalloys (in particular at low temperatures/high strain rates or with large imposed strains) [15, 20], shearing is not expected at elevated temperatures and low strain rates, particularly with small (nominally elastic) imposed cyclic strains. Indeed, c′ shearing was not observed in regions far from the notch (Fig. 3). However, precipitate shearing appears to be occurring rather extensively in the regions surrounding the notch. The observed shearing is believed to be associated with the stress concentration induced by the notch. Since the dominant fatigue crack was initiated from the notch (Fig. 2c, d), c′ shearing in regions near the notch may play a major role in the crack initiation mechanisms in this test piece.

339

As previously mentioned, fatigue crack initiation at cast pores is not unusual in Ni-based single-crystal superalloys [1, 2, 11, 12, 21]. As a consequence, significant efforts have recently been directed towards the characterisation of plasticity and microstructure evolution near pores during cyclic loading [15, 16, 22]. To the best of the authors’ knowledge, however, this study presents the first investigation of the mechanisms of cyclic plasticity at the ‘critical’ stress concentration conducted by means of TEM. By showing distinct deformation modes in the bulk of the test piece compared to the regions near the notch, this study highlights the necessity of the investigation of micro-plasticity in regions near stress raisers. The mechanisms described above are intimately related to fatigue crack initiation for the test piece in this study; their thorough investigation is therefore crucial to develop an in-depth understanding of how cracks nucleate at stress concentrations, and further relate these mechanisms to the overall fatigue behaviour of the alloy.

Conclusions This work presents the results of the experimental characterisation of the mechanisms of local cyclic micro-plasticity in regions of stress concentrations in a Ni-based single-crystal superalloy tested in tension/compression fatigue at 800 °C. It was shown that, when far-field stresses are nominally elastic, plastic deformation in the bulk of the test pieces is highly inhomogeneous. By contrast, the cyclic deformation near the ‘critical’ stress raiser, i.e. the one that induced the nucleation of the dominant crack, was found to be homogeneously distributed across large regions, with analogous dislocations arrangements observed at several different locations. Moreover, along with the expected accommodation of plastic strain in the c channels, extensive c′ shearing was observed in the regions near the stress raiser, in contrast with what was observed in the bulk of the test piece and with what is expected for superalloys tested with low cyclic strains at elevated temperatures and low strain rates. This work presents the first TEM-based investigation of local cyclic plasticity near the critical stress raiser in Ni-based superalloys; these results may contribute to an improved understanding of fatigue crack initiation at stress concentrations in Ni-based single-crystal superalloys. Acknowledgements The authors gratefully acknowledge the support received from the Beijing Institute of Aeronautical Materials (BIAM). The research was performed at the BIAM-Imperial Centre for Materials Characterisation, Processing and Modelling at Imperial College London.

340

References 1. Leverant GR, Gell M (1975) The influence of temperature and cyclic frequency on the fatigue fracture of cube oriented nickel-base superalloy single crystals. Metall Trans A 6:367– 371. https://doi.org/10.1007/BF02667291. 2. Milligan WW, Jayaraman N, Bill RC (1986) Low cycle fatigue of MAR-M 200 single crystals with a bimodal c′ distribution at 760 and 870 °C. Mater Sci Eng 82:127–139. https://doi.org/10.1016/ 0025-5416(86)90101-1. 3. Brien V, Décamps B (2001) Low cycle fatigue of a nickel based superalloy at high temperature: Deformation microstructures. Mater Sci Eng A 316:18–31. https://doi.org/10.1016/S0921-5093 (01)01235-7. 4. Li P, Li QQ, Jin T, Zhou YZ, Li JG, Sun XF, Zhang ZF (2014) Comparison of low-cycle fatigue behaviors between two nickel-based single-crystal superalloys. Int J Fatigue 63:137–144. https://doi.org/10.1016/j.ijfatigue.2014.01.018. 5. Pineau A, Antolovich SD (2009) High temperature fatigue of nickel-base superalloys - A review with special emphasis on deformation modes and oxidation. Eng Fail Anal 16:2668–2697. https://doi.org/10.1016/j.engfailanal.2009.01.010. 6. Wang XG, Liu JL, Jin T, Sun XF, Zhou YZ, Hu ZQ, Do JH, Choi BG, Kim IS, Jo CY (2015) Deformation mechanisms of a nickel-based single-crystal superalloy during low-cycle fatigue at different temperatures. Scr Mater 99:57–60. https://doi.org/10. 1016/j.scriptamat.2014.11.026. 7. Glatzel U, Feller-Kniepmeier M (1991) Microstructure and dislocation configurations in fatigued [001] specimens of the nickel-based superalloy CMSX-6. Scr Metall Mater 25:1845– 1850. https://doi.org/10.1016/0956-716X(91)90315-R. 8. Gabb TP, Welsch G, Miner R V. (1987) The characteristics of c′ dislocation pairs in a nickel-base superalloy. Scr Metall 21:987– 992. https://doi.org/10.1016/0036-9748(87)90140-2. 9. Liu L, Meng J, Liu J, Jin T, Sun X, Zhang H (2017) Effects of crystal orientations on the cyclic deformation behavior in the low cycle fatigue of a single crystal nickel-base superalloy. Mater Des 131:441–449. https://doi.org/10.1016/j.matdes.2017.06.047. 10. Li JR, Zhong ZG, Liu SZ, Tang DZ, Wei P, Wei PY, Wu ZT, Huang D, Han M (2000) A Low-Cost Second Generation Single Crystal Superalloy DD6. In: Superalloys 2000. pp 777–783. 11. Gabb TP, Gayda J, Miner R V. (1986) Orientation and temperature dependence of some mechanical properties of the single-crystal nickel-base superalloy Rene N4: Part II. Low cycle fatigue behaviour. Metall Trans A, Phys Metall Mater Sci 17:497–505. https://doi.org/10.1007/BF02643956.

A. Piglione et al. 12. Evans WJ, Lancaster R, Steele A, Whittaker M, Jones N (2009) Plain and notched fatigue in nickel single crystal alloys. Int J Fatigue 31:1709–1718. https://doi.org/10.1016/j.ijfatigue.2009.03.010. 13. Man J, Vystavěl T, Weidner A, Kuběna I, Petrenec M, Kruml T, Polák J (2012) Study of cyclic strain localization and fatigue crack initiation using FIB technique. Int J Fatigue 39:44–53. https://doi. org/10.1016/j.ijfatigue.2011.05.002. 14. Jiang R, Bull DJ, Evangelou A, Harte A, Pierron F, Sinclair I, Preuss M, Hu XT, Reed PAS (2018) Strain accumulation and fatigue crack initiation at pores and carbides in a SX superalloy at room temperature. Int J Fatigue 114:22–33. https://doi.org/10. 1016/j.ijfatigue.2018.05.003. 15. Cervellon A, Hémery S, Kürnsteiner P, Gault B, Kontis P, Cormier J (2020) Crack initiation mechanisms during very high cycle fatigue of Ni-based single crystal superalloys at high temperature. Acta Mater 188:131–144. https://doi.org/10.1016/j. actamat.2020.02.012. 16. Cervellon A, Cormier J, Mauget F, Hervier Z (2017) VHCF life evolution after microstructure degradation of a Ni-based single crystal superalloy. Int J Fatigue 104:251–262. https://doi.org/10. 1016/j.ijfatigue.2017.07.021. 17. Zhao Z, Zhang F, Dong C, Yang X, Chen B (2020) Initiation and Early-Stage Growth of Internal Fatigue Cracking Under Very-High-Cycle Fatigue Regime at High Temperature. Metall Mater Trans A 51:1575–1592. https://doi.org/10.1007/s11661020-05633-3. 18. Stroh AN (1954) The formation of cracks as a result of plastic flow. Proc R Soc London Ser A Math Phys Sci 223:404–414. https://doi.org/10.1098/rspa.1954.0124. 19. Pope DP, Ezz SS (1984) Mechanical properties of Ni3Al and nickel-base alloys with high volume fraction of y’. Int Met Rev 29:136–167. https://doi.org/10.1179/imtr.1984.29.1.136. 20. Milligan WW, Antolovich SD (1987) Yielding and deformation behavior of the single crystal superalloy PWA 1480. Metall Trans A 18A:85–95. https://doi.org/10.1007/BF02646225. 21. Rémy L, Geuffrard M, Alam A, Köster A, Fleury E (2013) Effects of microstructure in high temperature fatigue: Lifetime to crack initiation of a single crystal superalloy in high temperature low cycle fatigue. Int J Fatigue 57:37–49. https://doi.org/10.1016/j. ijfatigue.2012.10.013. 22. Bortoluci Ormastroni LM, Mataveli Suave L, Cervellon A, Villechaise P, Cormier J (2020) LCF, HCF and VHCF life sensitivity to solution heat treatment of a third-generation Ni-based single crystal superalloy. Int J Fatigue 130. https://doi.org/10.1016/ j.ijfatigue.2019.105247.

Tensile, Low Cycle Fatigue, and Very High Cycle Fatigue Characterizations of Advanced Single Crystal Nickel-Based Superalloys Luciana Maria Bortoluci Ormastroni, Satoshi Utada, Jérémy Rame, Lorena Mataveli Suave, Kyoko Kawagishi, Hiroshi Harada, Patrick Villechaise, and Jonathan Cormier

Abstract

Keywords

Tensile and fatigue life variabilities are investigated for new-generation single crystal Ni-based superalloys: the 3rd generation CMSX-4 Plus, the 6th generation TMS-238, and a newer Ni-based superalloy, TROPEA containing Pt. Consistently from the results of previous research, very high cycle fatigue (VHCF) properties at the chosen condition of T = 1000 °C/Re = −1/f = 20 kHz are mainly influenced by the solidification/homogenization pore size and position. TROPEA alloy has the best low cycle fatigue (LCF) life among all tested alloys at 650 °C and Rr = 0.05/f = 0.5 Hz. To better understand the influence of chemical composition on the LCF endurance, tensile properties were investigated using nine different single crystalline alloys at 650 °C with the strain rate of 5
10−4 s−1. The yield stress is directly affected by the chemical composition of the Ni-based superalloys, and alloys with high contents of Ti and Ta have a higher yield stress, due to an increased shearing resistance of c′ precipitates. Hence, the yield stress is the main control parameter of LCF at the selected condition. No influence of chemical composition on VHCF life durability has been observed, in good agreement with previous studies.

Ni-based superalloys Single crystal Low cycle fatigue Very high cycle fatigue Tensile properties

L. M. Bortoluci Ormastroni (&)  S. Utada  P. Villechaise  J. Cormier Physics and Mechanics of Materials Department, Institut Pprime, UPR CNRS 3346, ISAE-ENSMA, 1 avenue Clément Ader, BP 40109, 86961 Futuroscope-Chasseneuil Cedex, France e-mail: [email protected] S. Utada  J. Rame SAFRAN Aircraft Engines, Site de Villaroche, Rond-Point René Ravaud - Réau, 77550 Moissy-Cramayel, France L. Mataveli Suave SAFRAN Tech, PFX, 171 Boulevard Valmy, 92700 Colombes, France K. Kawagishi  H. Harada National Institute for Materials Science (NIMS), 1-2-1 Sengen, Tsukuba, Ibaraki 305-0047, Japan







Introduction With a global increase in air traffic, a growth of 7.4 pct. in 2018 and another 5 pct. projected growth in 2019 [1], cleaner in terms of NOx, CO, and CO2 emissions, and cost-effective aero-engines, in terms of fuel consumption, are required. Advanced single crystal (SX) Ni-based superalloys are necessary to achieve higher operating temperatures, especially in the first stage blades and vanes, in addition to more efficient internal cooling systems and advanced thermal barrier coatings [2–4]. Since fatigue is responsible for most of the crack initiation events and failure for internally cooled blades [5], the present study is especially focusing on the fatigue life durability of three recently developed SX Ni-based superalloys: the 3rd generation CMSX-4 Plus [6], the 6th generation TMS-238 [7], and TROPEA [8, 9]. CMSX-4 Plus has been chosen in this study as a reference alloy, given the fact that currently most advanced aero-engine blades are manufactured with 3rd generation SXs or eventually with 4th generation having very good creep properties [6, 10, 11]. TMS-238, a 6th generation that probably presents the best compromise between (very) high-temperature creep properties and oxidation resistance, has been chosen since very few fatigue data exist for this alloy [7]. TROPEA is a new-generation Pt-containing superalloy which has recently been developed between ISAE-ENSMA/Institut Pprime and SAFRAN in France. It is considered as a potential alloy for future airfoils [8, 9]. The unique features of this alloy are a very high Ta content to increase the resistance to c′ shearing, a reduced Re content to preserve density and cost, and an addition of 1.95 wt. pct. Pt. The main idea behind the development of this alloy was to have a good environmental

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_33

341

342

resistance, to have a low c/c′ lattice misfit by increasing the c′ lattice parameter, and to maximize the strength and the stability of the c′ phase above 1200 °C by forming a (Ni, Pt)3 (Al, Ti, Ta) L12 structure [5, 8, 12, 13]. Pt is a very effective solid solution strengthener as a substrate element [14], and it is aimed to reduce the complexity of Ni-based superalloy coating process eliminating a bond coating step as the superalloy itself presents high environmental resistance [15, 16]. It is known that the Pt is not as much effective as Ta or Mo to strength the material at temperatures under 650 °C [17, 18]. However, the influence of the Pt element on the tensile and fatigue properties in SX Ni-based superalloys is not well understood. For the fatigue properties evaluations, very high cycle fatigue (VHCF) at T = 1000 °C/Re = −1/f = 20 kHz and low cycle fatigue (LCF) at T = 650 °C/Rr = 0.05/f = 0.5 Hz were chosen to simulate service conditions of the profile and the blade root, respectively. In addition to tensile tests at 650 °C for the three targeted alloys, other tests were performed using MAR-M200 + Hf, AM1, AM3, René N4, MC2, René N5, PWA 1484, René N6, and MCNG to analyze the influence of chemical compositions on yield stress (YS). The aim of these additional tests will be to better understand the role of chemistry on the fatigue durability at low temperature of the three alloys mainly studied in this article. Finally, it has to be noted that HCF/VHCF durability is a certifying criterion of airfoils according to airworthiness authorities (e.g., EASA and FAA) [19–22].

Material and Experimental Techniques

L. M. Bortoluci Ormastroni et al.

Mechanical Tests Two different fatigue conditions were investigated, all performed with a sinusoidal waveform. The fatigue testing conditions were chosen to cover different temperature/stress states encountered in high-pressure turbine blades. Strain-controlled VHCF tests were performed at 1000 °C, f = 20 ± 0.5 kHz, Re = −1 with an ultrasonic fatigue machine used in previous studies [11]. The specificities and details of the machine and the specimen geometry are detailed in the literature [19, 21]. Load-controlled LCF tests were conducted using an electro-mechanic Instron 8562 machine at 650 °C, f = 0.5 Hz, Rr = 0.05. Specimens used for these tests have a 13 mm gauge length, a *4.3 mm gauge diameter, and a 56 mm total length. All specimens tested in fatigue were low-stress polished up to a mirror finish with 1 µm diamond paste to remove residual stresses introduced by the machining process and remaining scratches. Tensile tests were performed using an electro-mechanical Instron 8562 machine at 650 °C. Strain measurements were ensured with an extensometer, equipped with ceramics arms, positioned onto the gauge section of the specimens. The tensile tests were performed using a strain rate of 5
10−4 s−1. Cylindrical specimens having a 14 mm gauge length, a 4 mm gauge diameter, and a total length of 42 mm were used. Low-stress mechanical polishing up to a 4000 grade SiC paper was performed on the cylindrical specimens to remove remaining scratches and the plastically deformed layer before tensile tests. The polishing procedure avoids recrystallization during high-temperature monotonic tests.

SX Ni-Based Superalloys Chemical Composition Microstructural Characterizations SX bars of CMSX-4 Plus and TROPEA were prepared at SAFRAN Tech PFX in Gennevilliers, France, while TMS-238 bars were prepared at National Institute for Materials Science (NIMS), Japan. The bars were solidified with a 14-mm diameter and the longitudinal direction close to [001]. The misorientation from the perfect [001] orientation is less than 13°. CMSX-4 Plus and TROPEA were solutioned and aged at Institut Pprime, and TMS-238 was solutioned and aged at NIMS. To better understand the influence of chemical composition on YS at 650 °C, the authors chose nine different SX Ni-based superalloys. The bars have been solidified with nearly the same solidification parameters. Misorientation of the bars was less than 5° from [001]. These alloys are fully heat-treated to optimize the microstructure. The chemical compositions in wt. pct. of the investigated alloys are presented in Table 1. The heat treatment (HT) conditions of TROPEA, CMSX-4 Plus, and TMS-238 are described in Table 2.

A combination of optical and scanning electron microscopy (OM and SEM, respectively) observations was used to characterize the microstructure of each alloy. The pore size distribution and the area fraction were determined by OM using unetched mirror-polished specimens, up to a 1-lm diamond paste polishing grade, sliced from heat-treated bars. Metallographic observations were performed roughly on (001) plane, i.e., perpendicular to the solidification direction. The c′ size was determined from scanning electron microscope (SEM) observations using JEOL JSM-7000F field emission gun microscope operating at 25 kV, at a working distance of *10 mm. For such SEM observations, specimens were previously polished up to a mirror finish and were etched for 8 to 10 s using aqua regia (1/3 HNO3 + 2/3 HCl, vol. parts) at *4 °C. Pore size, c′ size distribution, and c/c′ eutectics area fraction were determined by image analyses using ImageJ software and specifically developed algorithms [23, 24].

Tensile, Low Cycle Fatigue, and Very High Cycle Fatigue …

343

Table 1 Nominal chemical composition of SX Ni-based superalloys examined (in wt. pct.) Alloy

Ni

Cr

Mo

Co

W

Re

Al

Ti

Ta

Pt

Ru

Hf

C

B

Nb

Zr

TROPEA

Bal.

6.5

0.6

9.0

6.0

1.0

5.6

1.0

9.0

2.0

/

0.1

/

/

/

/

CMSX-4 Plus

Bal.

3.5

0.6

10

6.0

4.8

5.7

0.85

8.0

/

/

0.1

/

/

/

/

TMS-238

Bal.

4.6

1.1

6.5

4.0

6.4

5.9

/

7.6

/

5.0

0.1

/

/

/

/

MAR-M200 +Hf

Bal.

9.0

/

10.0

12.5

/

5.0

2.0

/

/

/

1.6

0.15

0.015

1.0

0.05

AM1

Bal.

7.8

2.0

6.5

5.7

/

5.2

1.1

7.9

/

/

0.05

/

/

/

/

AM3

Bal.

8.0

2.2

5.5

5.0

/

6.0

2.0

3.5

/

/

/

0.15

/

/

/

René N4

Bal.

9.8

1.5

7.4

5.9

/

4.2

3.5

4.7

/

/

0.1

0.06

0.004

0.5

/

MC2

Bal.

8.0

2.0

5.0

8.0

/

5.0

1.5

6.0

/

/

/

/

/

/

/

René N5

Bal.

7.0

2.0

8.0

5.0

3.0

6.2

/

7.0

/

/

0.15

0.05

/

0.2

/

PWA 1484

Bal.

4.9

2.0

9.8

5.9

3.0

5.7

0.02

8.6

/

/

0.1

/

/

/

/

René N6

Bal.

4.0

1.0

11.7

6.0

5.23

5.7

0.04

6.9

/

/

0.2

/

/

/

/

MCNG

Bal.

4.0

1.0

/

5.0

4.0

6.0

0.5

5.0

/

4.0

0.1

/

/

/

/

Table 2 Heat treatment conditions for TROPEA, CMSX-4 Plus, and TMS-238

Alloy

TROPEA

CMSX-4 PLUS

TMS-238

Solution heat treatment

1300 °C/24 h/AQ (heating rate 2 ° C/min from 1200 °C)

1340 °C/15 h/AQ (heating rate 2 ° C/min from 1200 °C)

1300 °C/1 h +1310 °C/1 h + 1335 °C/3 h + 1345 °C/20 h/AQ

AGING 1

1200 °C/1 h/AQ

1163 °C/3 h/AQ

1150 °C/2 h/AQ

AGING 2

870 °C/16 h/AQ

1100 °C/4 h/AQ

870 °C/20 h/AQ

AGING 3

/

870 °C/20 h/AQ

/

After the mechanical testing, fractographic observations were performed using JEOL JSM-7000F SEM, using the secondary (SEI) and backscattered (BSE) electron imaging modes. Investigation of the c/c′ microstructure evolution after fatigue tests was also performed in longitudinal cross section, with the same polishing procedures and equipment described above.

Results Metallurgical Defect Characterization The microstructural characterization parameters are summarized in Table 3. Given the narrow solution window of TROPEA [9], specimens were left with a 10–13% remaining eutectics fraction. All of the superalloys present a similar pore area fraction of 1.3–1.6% and a maximum pore diameter between 80 and 120 µm. The c/c′ microstructures in primary dendrite arms after full HTs are presented in Fig. 1. A regular c/c′ cuboidal microstructure is obtained. The results for CMSX-4 Plus and TROPEA show a comparable value of c′-precipitate average edge length around 500 nm and a c channel width around 60–70 nm. TMS-238 presents a c′-precipitate average edge length around 235 nm and a c channel width around 50 nm.

Very High Cycle Fatigue at 1000 °C The S-N diagram gathering all VHCF results obtained at 1000 °C is presented in Fig. 2a. Results of CMSX-4 Plus with the HT described in Table 2, CMSX-4 Plus with a standard homogenization (STD), and CMSX-4 from a previous study [11] are inserted in the same figure. Their maximum pore size is *100 µm [11]. At high stresses, TMS-238 presented a VHCF lifetime one order of magnitude higher than the lifetime presented by TROPEA and CMSX-4 Plus. As already documented by Cervellon et al. [20], for R = −1 and a limit of 109 cycles, the crack initiation always starts from casting pores. Similar results have also been obtained in the present study for all three main alloys, as illustrated in Fig. 2b, c. The rough region present around the crack initiating casting pore is a characteristic of VHCF crack initiation mechanism [20, 22].

Low Cycle Fatigue at 650 °C The S-N diagram at 650 °C/Rr = 0.05 and f = 0.5 Hz is presented in Fig. 3a. Results obtained from a previous study of Bortoluci et al. [11] have been added in this plot using CMSX-4 Plus specimens in the same testing conditions and machine. CMSX-4 Plus and TROPEA have a comparable

344

L. M. Bortoluci Ormastroni et al.

Table 3 Stereological characterizations of SX Ni-based superalloys examined c′-Precipitate average edge length (nm) Primary dendrite arm

c Channel width (nm)

Pore’s area fraction (Pct.)

Pore’s maximum diameter (µm)

Eutectic’s area fraction (Pct.)

Interdendritic region

TROPEA

471 ± 91

482 ± 71

71 ± 15

1.5

120

10 - 13

CMSX-4 – Plus

525 ± 91

533 ± 157

60 ± 12

1.3

100

0

TMS-238

234 ± 24

293 ± 53

43 ± 8

1.6

80

0

MAR-M200 +Hf

405 ± 81

/

81 ± 25

AM1

465 ± 62

360 ± 52

52 ± 19

AM3

440 ± 113

/

60 ± 20

René N4

418 ± 91

440 ± 88

98 ± 42

MC2

415 ± 77

477 ± 18

67 ± 17

René N5

288 ± 31

369 ± 26

13 ± 7

PWA 1484

316 ± 46

404 ± 64

46 ± 16

René N6

380 ± 75

384 ± 46

54 ± 16

MCNG

375 ± 70

349 ± 69

64 ± 20

Fig. 1 c/c′ microstructure along a (001) plane before mechanical testing. The images were taken inside the primary dendrite arm in CMSX-4 Plus (a), TMS-238 (b), and TROPEA (c). Detail of the c-like particles within c′ precipitates in TROPEA specimens is shown as the insert in (c)

LCF durability within the rmax range of 930–1000 MPa. TMS-238 has lower lifetime variability. Under these LCF fatigue conditions, crack initiation occurred at casting pores, Fig. 3b, c. This observation is in good agreement with previous LCF results from Steuer et al. using AM1 under similar conditions [25]. Neither c′-precipitate rafting nor recrystallization layer was identified in the internal microstructure of specimens after failure.

Tensile Tests The main objective of tensile tests at 650 °C is to interpret the LCF behaviors as the maximum applied stress is approaching YS.

Tensile curves are presented in Fig. 4, and corresponding YS defined at 0.2 and 0.05 pct. of plastic offsets, ultimate tensile stress, strain at failure, and a partition coefficient is presented in Table 4. The targeted superalloys have a specific yielding behavior at this condition. TROPEA shows a higher YS of 1027 MPa and a higher ultimate tensile stress compared to CMSX-4 Plus. Nevertheless, both of them exhibit a good elongation which is a typical tensile behavior in SX Ni-based superalloys at this temperature and strain rate [26, 27]. TMS-238 presents a particular tensile behavior with lower YS of 883 MPa. However, this superalloy shows a spectacular hardening, leading to an ultimate tensile stress equivalent to the ones observed for CMSX-4 Plus and TROPEA alloys.

Tensile, Low Cycle Fatigue, and Very High Cycle Fatigue …

345

Fig. 2 S-N diagram at 1000 °C, Re = −1/f = 20 kHz. The alternating stress ra is plotted as a function of the number of cycles to failure (a), the corresponding maximum pore diameters (micrometer) are shown in black. Typical crack initiation sites in TMS-238 (b), and TROPEA (c) specimens after failure at ra = 180 MPa. Pores serving as main

crack initiation sites have been highlighted using white dotted curves and white arrow, respectively. The black dotted curves highlight the rough region. CMSX-4 Plus, CMSX-4 Plus STD, and CMSX-4 data are extracted from the literature [11, 20]

Fig. 3 S-N diagram at 650 °C, Rr = 0.05/f = 0.5 Hz. The maximum stress rmax is plotted as a function of the number of cycles to failure. Typical crack initiation sites after LCF tests at rmax = 950 MPa. Crack initiation from an internal pore for TMS-238 (b) and from an internal

pore connected to a c/c′ eutectic for TROPEA (c). The white arrows highlight the casting/homogenization pores. The black arrows in (c) highlight the presence of eutectics. CMSX-4 Plus data are extracted from the literature [11]

346

L. M. Bortoluci Ormastroni et al.

Fig. 4 Tensile curves representing the true stress versus true strain at 650 °C/5.0
10−4 s−1 for CMSX-4 Plus [11], TMS-238, and TROPEA (a) and a magnified plot around the YS (b). The nine additional SX Ni-based superalloys investigated (c) and a magnified plot around the YS (d)

Table 4 Tensile properties at 650 °C/5
10−4 s−1 of SX Ni-based superalloys examined

Superalloy

Yield stress 0.2 pct. of plastic offset (MPa)

Yield stress 0.05 pct. of plastic offset (MPa)

TROPEA

1058

CMSX-4 Plus [11]

979

TMS-238 MAR-M200 +Hf

Strain at failure (Pct.)

1047

1370

13.5

0.37

965

1160

8.5

0.30

883

736

1310

3.0

0.19

991

981

1400

12.0

0.25

AM1

1164

1162

1460

7.5

0.35

AM3

1058

1056

1320

10.0

0.29

René N4

1074

1024

1385

2.5

0.67

MC2

1000

975

1240

10.5

0.17

René N5

882

876

1220

10.0

0.17

PWA 1484

1106

1089

1200

3.5

0.23

René N6

1006

947

1095

1.5

0.19

876

874

1230

15.0

0.36

MCNG a

XTi þ XTa a XAl

Ultimate tensile stress (MPa)

X concentration in atomic percent

Tensile, Low Cycle Fatigue, and Very High Cycle Fatigue …

347

Fig. 5 True stress (MPa) versus true strain (%) diagram at 650 °C, Rr = 0.05/f = 0.5 Hz/rMax = 950 MPa (a). True stress (MPa) versus true strain diagram at 650 °C, Rr = 0.05/f = 0.5 Hz Normalized

(rMax/YS) = 1.05 (b). The cycles and the specimens lifetime are labeled in the figure. CMSX-4 Plus data are extracted from the literature [11]

Note that these specific tensile behaviors have been observed in several specimens machined out from different bars with different orientations close to the [001] crystallographic orientation.

A special attention will then be paid to the relation between chemistry and the mechanical properties of the superalloys in the following discussion.

database of Institut Pprime for the same VHCF conditions [19], the TMS-238 performance is higher, almost reaching the lifetime of AM1/MCNG HIPed specimens [11, 19]. The chemical composition of the superalloys seems to have no impact on the VHCF lifetime variability under fully reversed conditions, in good agreement with our previous results [20]. A contribution of the chemistry of the alloys to the VHCF lifetime should, however, be expected at high temperatures (>850°C) and positive stress ratio, where creep damage coexists with fatigue damage, or even dominates [19, 28].

Performance of TMS-238, CMSX-4 Plus, and TROPEA in VHCF

Performance of TMS-238, CMSX-4 Plus, and TROPEA at LCF Conditions

As already documented by Bortoluci et al. [11], and more specifically investigated by Cervellon et al. [20], the effect of pore size is much stronger than the presence of other metallurgical defects in the VHCF domain under fully reversed conditions. Hence, the durability of the investigated superalloys under VHCF conditions at Re = −1/1000 °C is mainly dependent on the pore size and position with respect to the surface, as described in Table 3. From Fig. 2a, TMS-238 presents a superior lifetime, mainly at very high stresses. Moreover, pores in TMS-238 are 20% smaller than CMSX-4 Plus and 40% smaller than TROPEA. These differences in pore size probably result from minor variations in withdrawal rates/thermal gradients during the solidification process and/or casting mold thickness since SX bars had the same diameter. From past investigations [11, 20], the solidification parameters are the most important criteria controlling the VHCF lifetime variability. In fact, comparing the results obtained in this project with all the

The difference in LCF life between CMSX-4 Plus, TMS-238, and TROPEA results from a clear difference in the first LCF loops, as shown in Fig. 5a. TMS-238 has a pronounced plastic deformation during the first cycle, while TROPEA and CMSX-4 Plus present almost no plasticity under the same temperature and maximum applied stress of 950 MPa. Although an almost fully elastic loop is obtained since cycle 2 for TMS-238 under this condition, the large plastic deformation at cycle 1 leads to a very pronounced slip localization, overall leading to an earlier crack initiation from pore (see Fig. 3b), serving as notches. At rMax = 950 MPa, TMS-238 has a LCF life nearly two times smaller than the one obtained for TROPEA and CMSX-4 Plus. To better understand the LCF durability, and more specifically the contribution of YS to the LCF life, the cyclic behavior of TMS-238, CMSX-4 Plus, and TROPEA was compared at iso-maximum stress with respect to the YS value, i.e., same rMax/YS ratio, see Fig. 5b. The LCF

Discussion

348

L. M. Bortoluci Ormastroni et al.

Fig. 6 Yield stress, measured at 0.05 pct. plastic offset, plotted as a function of the c′-precipitate size (nm) (a) and as a function of the (XTi + XTa)/ XAl ratio (X in at. pct.) (b). A black dotted main trendline is added in each plot

endurance is almost the same. Moreover, it is observed that TMS-238 stabilizes since cycle 2 while TROPEA and CMSX-4 Plus are reaching elastic loops after at least five cycles. This faster elastic accommodation of TMS-238 is a result of its very steep hardening, see Fig. 4b. The differences in the YS are hence the main criteria controlling the LCF life in this condition.

Influence of Chemical Composition on Tensile Properties at 650 °C A difference in c′-precipitate size in primary dendrite arms, and interdendritic region, as well as the channel width has been noticed for TMS-238. The c′-precipitate average size is around 55% smaller, and the channel width is around 30% smaller than those reported for CMSX-4 Plus. This difference in size is known to impact the mechanical properties of Ni-based superalloys [29]. Sengupta et al. have documented that the YS increases with the increasing of c′-precipitate size for CMSX-4 at different temperatures, with a peak in the 700–800 °C temperature range [30]. On the contrary, Shah et al. [31] have showed for PWA 1480 that the YS is almost insensitive to the c′-precipitate size in the same temperature range. However, for temperatures around 650 °C, the YS decreases with the increasing of c′-precipitates. In the present study, a correlation between the YS (MPa) and the c′precipitate size (nm) for all superalloys tested is detailed in Fig. 6a. The results induce the conclusion that the microstructure is the main parameter to control the YS. Conversely, Caron et al. have proved that the alloy composition strongly influences the YS and the hardening behavior of SX Ni-based superalloys [32]. In the same study, a great care was taken to assess the effect of chemical

composition of various SX Ni-based alloys having the same precipitate size and crystalline orientation. They concluded that the YS variation is mainly related to the ratio in the concentrations in atomic percent of Ti + Ta over the concentration of Al, results in Table 4. These elements are well known to be c′ former. They increase the APB energy consequently contributing to the resistance to precipitate shearing. TMS-238 does not contain Ti, which should be replaced in the c′ precipitate by other elements such as Ta (*2.7 at. pct.). The absence/low concentration of Ti/Ta compared to Al influenced the c′ resistance to shearing, decreasing the YS at 650 °C [31–33]. Regarding the results presented in Table 4, the ratio for TMS-238 is around 0.19, while it is of 0.30 for CMSX-4 Plus and 0.37 for TROPEA. It hence seems that the higher the ratio, the higher the YS at 650 °C. A similar conclusion was obtained by Caron et al. [32] at fixed precipitate size. They suggested that the superalloys chemistry, especially the c′ one, is the key to this evolution. Figure 6b also shows that the YS is highly influenced by the superalloys chemistry. It is worth noting that both superalloys outside the main trend in this figure (TMS-238 and René N4) present a strong hardening (see Fig. 4) and low ductility. This may result from a higher density of sub-grain/low angle boundaries at the dendrite/interdendritic interfaces if a very fast cooling rate (greater than 500 °C/min) at the end of the solution heat treatment has been used, leading to pronounced dendritic stresses [34]. If this is the case, a higher density of sub-grain boundaries may trigger earlier yielding, more hardening, and lower ductility due to a lower mean free path for dislocation/slip bands. This assumption remains to be checked objectively. Among all alloys subjected to tensile test at 650 °C, TMS-238 showed a particular behavior with a very low YS

Tensile, Low Cycle Fatigue, and Very High Cycle Fatigue …

and a very spectacular work hardening. The specific chemistry and high misfit of the alloy can be the reasons for such an atypical behavior. Compared to other alloys, TMS-238 is estimated to have a *5% lower fraction of c′ due to its slightly lower content of main c′ forming element (sum of Ti + Ta + Al contents). Although it has a high Ta content, overall Al-substitution element that partitions into c′ phase is lower than other alloys that performed well in the tensile test, such as CMSX-4 Plus, TROPEA, and PWA1484. Ta and Ti are known to increase anti-phase boundary (APB) energy that increases precipitate’s resistance to shearing by ½ pair dislocations. Lower strength of c′ phase is the most probable reason of lower YS for TMS-238 [17, 35]. On the other hand, partition of alloying elements, particularly very high Re content in c phase in TMS-238, decreases stacking fault energy of c matrix [35]. Activation and interaction of multiple slip systems by APB shearing and interaction of stacking faults in the matrix both may be contributed to this distinctive work hardening. Based on the literature results [33], a possibility to improve fatigue performances of TMS-238 at fixed chemistry would be an increase in the c′ size keeping the regularity in c′ morphology inherited from the very high misfit of the alloy. This can only be achieved by using slower cooling rates and/or longer/hotter aging HTs [36], without affecting the coherency of the precipitates with the matrix.

TROPEA Chemistry and Its Mechanical Performance TROPEA has the precious metal element Pt in its composition. It is known to stabilize the c′ precipitate and to increase its solvus temperature [13, 17]. This new scenario with a c′ composition of (Ni, Pt)3(Al, Ta, Ti) seems to be beneficial to the tensile properties of TROPEA. Indeed, according to Table 4, TROPEA holds the best compromise in terms of YS/elongation at failure. This very high YS and strain at failure have been obtained with all the room temperature, i.e., 850 °C temperature range. Even though TROPEA presents *13% of eutectics due to its high Ta content and the introduction of Pt, overall closing the solution HT window, the presence of such metallurgical defects has no detrimental impact on the LCF properties of the material. This result is in fact in good agreement with a previous comprehensive study on the role of metallurgical defects such as pores, c/c′ eutectics pools, incipient melting, and chemical homogeneity across the dendritic structure on the fatigue life variability of CMSX-4 Plus [11]. Further information about TROPEA’s chemical development and creep properties is detailed in Rame et al. [9].

349

While the cost of TROPEA and TMS-238 is comparable, TROPEA has shown a good compromise between the YS, elongation at failure, and more importantly, a very good performance in LCF at low temperatures. However, the exact role of Pt remains to be understood, especially in terms of YS and how its partitioning coefficient evolves as a function of temperature. TROPEA alloy, in itself, is a very first trial to investigate how minor additions of Pt in the chemical composition could impact mechanical properties. Still, the present results are showing potential of Pt bearing SX Ni-based superalloy that exhibits very good tensile behavior for turbine blade material, with an acceptable density [9].

Conclusions Tensile and LCF at 650 °C, and VHCF at 1000 °C have been investigated for three main SX Ni-based superalloys: CMSX-4 Plus (3rd generation alloy and chosen as a reference), TMS-238 (6th generation alloy), and TROPEA (a newly developed Pt-containing alloy). The influence of chemistry on tensile properties was investigated adding nine other SX Ni-based superalloys to the analysis. The following main conclusions have been obtained: • The solidification parameters and consequently the casting pore size are still the main parameters controlling the VHCF life at 1000 °C/20 kHz/Re = -1. • The yield stress is directly affected by the chemical composition of the SX Ni-based superalloys. Ti and Ta are very effective elements to increase the resistance of c′ precipitates to shearing. • The yield stress is the main parameter controlling the low cycle fatigue (at 650 °C, Rr = 0.05/f = 0.5 Hz) life. • Despite having a low yield stress, TMS-238 presents a remarkable strain hardening, overall leading to a fast cyclic stabilization in LCF at 650 °C. • TROPEA alloy has superior fatigue endurance in LCF at 650 °C compared to CMSX-4 Plus and TMS-238 alloys due to its higher yield stress, mainly resulting from its high Ta content. • LCF and VHCF properties do not seem to be impacted directly by the presence of Pt in the chemical composition.

Acknowledgments SAFRAN Tech/PFX is gratefully acknowledged for providing CMSX-4 Plus and TROPEA SX bars. NIMS is gratefully acknowledged for providing TMS-238 SX bars, especially Dr. Toshiharu Kobayashi, Dr. Tadaharu Yokokawa, and Mr. Yuji Takata. LMBO, SU, and JC gratefully acknowledge Dr. Pierre Caron (formerly

350 at ONERA) for fruitful discussions. Likewise, they wish to thank Mrs. Florence Hamon, Mr. Florent Mauget, and Mr. Jacques Lefort (Physics and Mechanics of Materials Department, Institut Pprime, France) for their help and advices during mechanical testing and microstructural observations.

References 1. (2019) Statista. In: Annu. growth Glob. air traffic Passeng. demand from 2006 to 2019. https://www.statista.com/statistics/193533/ growth-of-global-air-traffic-passenger-demand/. Accessed 9 Jul 2019 2. Reed RC (2005) The superalloys: fundamentals and applications. 363 3. Pollock TM, Tin S (2008) Nickel-Based Superalloys for Advanced Turbine Engines: Chemistry, Microstructure and Properties. J Propuls Power 22:361–374. https://doi.org/10.2514/1.18239 4. Caron P (2000) High gamma prime solvus new generation Nickel-based superalloys for single crystal turbine blade applications. Superalloys 2000 737–746. https://doi.org/10.7449/2000/ Superalloys_2000_737_746 5. Zhang WJ (2016) Thermal mechanical fatigue of single crystal superalloys: Achievements and challenges. Mater Sci Eng A 650:389–395. https://doi.org/10.1016/j.msea.2015.10.078 6. Wahl JB, Harris K (2016) CMSX-4 Plus Single Crystal Alloy Development, Characterization and Application Development. Superalloys 2016 25–33. https://doi.org/10.1002/9781119075646. ch3 7. Kawagishi K, Yeh AC, Yokokawa T, et al (2012) Development of an Oxidation-Resistant High-Strength Sixth-Generation Single-Crystal Superalloy TMS-238. Superalloys 2012 189–195. https://doi.org/10.1002/9781118516430.ch21 8. Rame J, Cormier J (2019) Ni-based superalloy for single crystalline turbine French Patent. (A1) 9. Rame J, Utada S, Bortoluci Ormastroni LM, et al (2020) Platinum containing new generation nickel-based superalloys for single crystalline applications. Superalloys 2020 10. Harris K, Wahl JB (2016) High strength single crystal superalloy. US Patent. US9518311B2. https://patents.google.com/patent/ US9518311B2/en 11. Bortoluci Ormastroni LM, Mataveli Suave L, Cervellon A, et al (2020) LCF, HCF and VHCF life sensitivity to solution heat treatment of a third-generation Ni-based single crystal superalloy. Int J Fatigue 130. https://doi.org/10.1016/j.ijfatigue.2019.105247 12. Cormier J (2016) Thermal Cycling Creep Resistance of Ni-Based Single Crystal Superalloys. Superalloys 385–394 13. Tin S, Zhang L, Ofori AP, Miller MK (2007) Atomic partitioning of platinum and ruthenium in advanced single crystal Ni-based superalloys. Mater Sci Forum 546–549:1187–1194. https://doi.org/ 10.4028/www.scientific.net/msf.546-549.1187 14. Corti CW, Coupland DR, Selman GL (1980) Platinum-Enriched Superalloys. Platin Met Rev 2–11 15. Chen Y, Zhao X, Bai M, et al (2015) Effect of platinum addition on oxidation behaviour of c/c′ nickel aluminide. Acta Mater 86:319–330. https://doi.org/10.1016/j.actamat.2014.12.023 16. Gleeson B, Wang W, Hayashi S, Sordelet D (2004) Effects of platinum on the interdiffusion and oxidation behavior of Ni-Al-based alloys. Mater Sci Forum 213–222. https://doi.org/ 10.4028/www.scientific.net/msf.461-464.213 17. Vattré A, Devincre B, Roos A (2009) Dislocation dynamics simulations of precipitation hardening in Ni-based superalloys with high c′ volume fraction. Intermetallics 17:988–994. https:// doi.org/10.1016/j.intermet.2009.04.007

L. M. Bortoluci Ormastroni et al. 18. Van Sluytman JS (2010) Microstructure and High Temperature Creep of Platinum Group Metal Modified Nickel Base Superalloys. University of Michigan. https://deepblue.lib.umich.edu/ handle/2027.42/77722 19. Cervellon A (2018) Propriétés en fatigue à grand et très grand nombre de cycles et à haute température des superalliages base nickel monogranulaires. PhD Thesis - Ecole Nationale Superieure de Mecanique et D’Aerotechnique. http://theses.fr/ 2018ESMA0009 20. Cervellon A, Cormier J, Mauget F, et al (2018) Very high cycle fatigue of Ni-based single crystal superalloys at high temperature. Metall Mater Trans A 49:3938–3950. https://doi.org/10.1007/ s11661-018-4672-6 21. Cervellon A, Cormier J, Mauget F, Hervier Z (2017) VHCF life evolution after microstructure degradation of a Ni-based single crystal superalloy. Int J Fatigue 104:251–262. https://doi.org/10. 1016/j.ijfatigue.2017.07.021 22. Cervellon A, Hemery S, Kürnsteiner P, et al (2020) Crack initiation mechanism during very high cycle fatigue of Ni-based single crystal superalloys at high temperature. Acta Mater 188:131–144. https://doi.org/10.1016/j.actamat.2020.02.012 23. Vaunois J-R, Cormier J, Villechaise P, et al (2010) Influence of both c’ distribution and grain size on the tensile properties of UDIMET 720Li at room temperature. Superalloy 718 199–213 24. Karunaratne MS., Carter P, Reed R. (2000) Interdiffusion in the face-centred cubic phase of the Ni–Re, Ni–Ta and Ni–W systems between 900 and 1300°C. Mater Sci Eng A 281:229–233. https:// doi.org/10.1016/S0921-5093(99)00705-4 25. Steuer S, Villechaise P, Pollock TM, Cormier J (2015) Benefits of high gradient solidification for creep and low cycle fatigue of AM1 single crystal superalloy. Mater Sci Eng A 645:109–115. https:// doi.org/10.1016/j.msea.2015.07.045 26. Diologent F (2002) Comportement en fluage et en traction de superalliages monocristallins à base de nickel. PhD Thesis Universite Paris XI UFR Scientifique d’Orsay. http://theses.fr/ 2002PA112300 27. Mataveli Suave L (2017) High temperature durbaility of DS200 + Hf alloy. PhD Thesis - École Nationale Superieure de Mecanique et d’Aerotechnique 28. Cervellon A, Yi JZ, Corpace F, et al (2020) Creep, fatigue, and oxidation interactions during High and very high cycle fatigue at elevated temperature of nickel-base single crystal superalloys. Superalloys 2020 29. Long H, Mao S, Liu Y, et al (2018) Microstructural and compositional design of Ni-based single crystalline superalloys —A review. J Alloys Compd 743:203–220. https://doi.org/10. 1016/j.jallcom.2018.01.224 30. Sengupta A, Putatunda SK, Bartosiewicz L, et al (1994) Tensile behavior of a new single crystal nickel-based superalloy (CMSX-4) at room and elevated temperatures. J Mater Eng Perform 3:664–672. https://doi.org/10.1007/BF02645265 31. Shah DM, Duhl DN (2012) The Effect of Orientation, Temperature and Gamma Prime Size on the Yield Strength of a Single Crystal Nickel Base Superalloy. 105–114. https://doi.org/10.7449/1984/ superalloys_1984_105_114 32. Caron P, Diologent F, Drawin S (2011) Influence of chemistry on the tensile yield strength of nickel-based single crystal superalloys. Adv Mater Res 278:345–350. https://doi.org/10.4028/www. scientific.net/AMR.278.345 33. Wang-Koh YM (2017) Understanding the yield behaviour of L12-ordered alloys. Mater Sci Technol (United Kingdom) 33:934– 943. https://doi.org/10.1080/02670836.2016.1215961 34. Epishin A, Link T, Brückner U, et al (2004) Effects of segregation in Nickel-base superalloys: dendritic stresses. Int Symp Superalloys 537–543. https://doi.org/10.7449/2004/Superalloys_2004_537_543

Tensile, Low Cycle Fatigue, and Very High Cycle Fatigue … 35. Yuan Y, Kawagishi K, Koizumi Y, et al (2014) Creep deformation of a sixth generation Ni-base single crystal superalloy at 800 °C. Mater Sci Eng A 608:95–100. https://doi.org/10.1016/j.msea.2014. 04.069

351 36. Steuer S, Hervier Z, Thabart S, et al (2014) Creep behavior under isothermal and non-isothermal conditions of AM3 single crystal superalloy for different solutioning cooling rates. Mater Sci Eng A 601:145–152. https://doi.org/10.1016/j.msea.2014.02.046

Competing Mechanism of Creep Damage and Stress Relaxation in Creep-Fatigue Crack Propagation in Ni-Base Superalloys Shiyu Suzuki, Motoki Sakaguchi, Ryota Okamoto, Hideaki Kaneko, Takanori Karato, Kenta Suzuki, and Masakazu Okazaki

Abstract

Keywords

Effect of creep deformation at crack tip on fatigue crack propagation behavior in a single crystal and a directionally solidified superalloys at 900 °C was investigated. Creep-fatigue crack propagation tests with single tension hold introduced into cyclic fatigue loading were conducted. In specimen extracted from the single-crystal superalloy, ICMSX-4, when the cyclic fatigue loading was restarted after the tension hold, nascent crack was immediately initiated followed by significant crack retardation. This crack propagation behavior was ascribed to mechanisms based on two different concepts of residual compressive stress and crack closure. From the viewpoint of mechanism based on the residual stress concept, material degradation at crack tip induced by the tension hold was investigated using scanning electron microscope, while stress relaxation and the resultant residual compressive stress at crack tip were quantified by elastic-plastic-creep finite element analysis coupled with digital image correlation technique. Finally, crack propagation behavior in polycrystalline specimen extracted from the directionally solidified superalloy, MGA1400, was investigated focusing on effect of grain boundary. An insight into criteria of a transition from crack retardation to accelerated intergranular cracking was suggested based on a relative grain size to “creep affected zone.”

Ni-based superalloys Creep-fatigue crack propagation Creep damage Stress relaxation Crack closure Digital image correlation Finite element analysis Grain boundary Grain size

S. Suzuki
M. Sakaguchi (&)
R. Okamoto Tokyo Institute of Technology, 2-12-1 O-okayama, Meguro-ku, Tokyo, Japan e-mail: [email protected] S. Suzuki e-mail: [email protected] H. Kaneko
T. Karato
K. Suzuki Mitsubishi Heavy Industries, Ltd., 2-1-1 Shinhama, Arai-cho, Takasago-shi, Hyogo, Japan M. Okazaki Nagaoka University of Technology, 1603-1 Kamitomioka, Nagaoka, Niigata, Japan















Introduction Extensive studies have been done to investigate creep-fatigue crack propagation (CFCP) in cast Ni-base superalloys in the past few decades [1–3]. Trapezoidal loading with holds at max/minimum load of fatigue loading has been common waveform for CFCP test, and the effects of load hold and hold time have been investigated. Typical results have been that crack propagation rate on cycle basis (CPR, hereafter) becomes higher with the dwell under load than that under pure fatigue loading, and longer hold time also increases CPR [1, 2]. The increase in CPR has been ascribed to several damage mechanisms, including nucleation and growth of creep voids, depletion of the c′ precipitates, c/c′ rafting and grain boundary (GB) embrittlement caused by oxidation. By contrast, Palmert et al. [3] found that the longer hold time resulted in decrease in CPR in a single-crystal superalloy. Similar crack retardation has also been reported for wrought superalloys [4, 5]. These retardation behavior has been ascribed to stress relaxation and crack tip blunting caused by creep deformation, and oxidation-induced crack closure. Fracture mechanics parameters that can quantify the CFCPR in Ni-base superalloy have attracted special interests as well. Generally, in the case where the hold time is short and creep deformation at crack tip is sufficiently small (small-scale creep condition), the linear elastic fracture mechanics parameter, K, is applicable [1]. If the hold time is longer and the transition to large-scale creep condition occurs at crack tip, it is appropriate to apply creep J-integral,

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_34

352

Competing Mechanism of Creep Damage and Stress Relaxation …

353

Jc, which takes energy dissipation due to the inelastic deformation into account [6]. This parameter can be experimentally obtained from hysteresis loop of load versus crack tip opening displacement, and is applicable to evaluate CFCPR universally from small-scale to large-scale creep conditions [6]. However, crack propagation behavior which cannot be explained by the conventional fracture mechanics parameter has also been reported, where CFCPR decreased with longer hold time [3], for example. The parameters such as K and Jc are generally obtained by the macroscopic specimen’s geometry or load-displacement curve, and there is a limit to accurately reflect time variation of stress/strain state and damage accumulation at crack tip. In order to understand CFCP behavior in Ni-base superalloy, it is necessary to clarify how the stress-strain evolution and the damage accumulation due to creep deformation control the crack propagation behavior. In a recent publication, we experimentally investigated the effect of creep on the crack propagation in a single-crystal superalloy, by conducting CFCP test with single tension hold introduced into cyclic fatigue loading instead of applying the trapezoidal loading [7]. In the present paper, the experimental investigation in [7] is firstly reviewed, after which microscopic observation and computational simulation are described aiming to clarify the CFCP mechanism. Finally, CFCP test using polycrystalline superalloy is described aiming to investigate the effect of GB.

cuboidal c′ precipitates were about 0.42 lm and 66% in ICMSX-4, and 0.36 lm and 51% in MGA1400.

Experimental Procedure

Specimen Preparation Compact (CT) specimens of 1 mm thickness were extracted from ICMSX-4 and MGA1400 by wire electric discharge machining (EDM) techniques. The proportion of the specimen’s dimensions was designed according to ASTM E647 [8], as described in detail in [9]. For CT specimens manufactured from single-crystal ICMSX-4 (SC specimen, hereafter), two types of crystal orientations were prepared, as shown in Fig. 2a. The one was specimen with orientation in both loading and crack propagation directions. The other was specimen with and orientations in loading and crack propagation directions, respectively. Most of the tests for SC specimens were conducted using specimens. This specimen shows non-crystallographic crack propagation perpendicular to the loading direction [9–11], thus, Mode I symmetric stress distribution can be readily created at crack tip. By contrast, in specimen, crack often propagates on crystallographic slip planes inclined to the loading direction, so that the symmetric stress distribution at crack tip cannot be expected. Polycrystalline CT specimen manufactured from the directionally solidified MGA1400 (PC specimen, hereafter) was extracted from plane normal to the solidification direction, as shown in Fig. 2b. Thus, the PC specimen had two-dimensional grain distribution with average grain size of 1.3 mm, so that effect of GB was readily evaluated.

Materials Two types of cast Ni-base superalloys were employed in this study. The one was a single-crystal superalloy, ICMSX-4, which is slightly modified CMSX-4, and the other was a directionally solidified superalloy, MGA1400. Table 1 shows their chemical compositions. For ICMSX-4, following heat treatments were applied: solution treatment in argon atmosphere at 1277 °C for 2 h, 1288 °C for 2 h, 1296 °C for 3 h, 1304 °C for 3 h, 1313 °C for 2 h, 1316 °C for 2 h, and 1277 °C for 2 h, followed by aging treatment at 1140 °C for 6 h, and 871 °C for 20 h. For MGA1400, solution, stabilizing and aging heat treatments were conducted. Figure 1 shows microstructures of ICMSX-4 and MGA1400 after the heat treatments. Average size and volume fraction of the

Table 1 Chemical compositions of ICMSX-4 and MGA1400 (wt %)

Creep-Fatigue Crack Propagation Test CFCP tests were conducted at 900 °C in air, using an electro-hydraulic machine and an induction heating system. Temperature on the specimen’s gauge section was controlled within ±5 °C using type K thermocouples. Following loading sequences with different conditions of Mode I stress intensity factor, KI, were applied, as shown in Fig. 3; 1. Cyclic fatigue loading of constant DKI condition (DKI = 20 MPa m1/2). 2. Tension hold of different KI values and hold times.

Co

Cr

W

Al

Ti

Ta

Mo

Re

Ni

ICMSX-4

9.7

6.5

6.4

5.7

1.0

6.6

0.6

3.0

Bal.

MGA1400

10

14

4.3

4.0

2.7

4.7

1.5

–

Bal.

354

S. Suzuki et al.

3. Cyclic fatigue loading of DKI increasing condition from DKI = 20 MPa m1/2.

Fig. 1 c/c′ microstructures of ICMSX-4 and MGA1400

In the loading sequence (2), KI value was 23.3 or 33.3 MPa m1/2, which correspond to mean (Kmean) or maximum (Kmax) value of the cyclic fatigue loading of constant DKI = 20 MPa m1/2, respectively. The tension hold time was 30, 90 or 180 min. Five tests were conducted using SC specimens with different crystal orientation, KI value and hold time of the loading sequence (2), as shown in Table 2. In tests (1)–(4), specimens were employed. Kmean tension hold for 90 min was applied in test (1), whereas Kmax tension holds for 90, 30 and 180 min were applied in test (2), (3) and (4), respectively. In test (5), specimen was employed, and Kmax tension hold for 90 min was applied. Only one test (test (6)) was conducted using PC specimen, where Kmax tension hold for 90 min was applied (Table 2). For all tests, load ratio R and loading frequency were 0.4 and 10 Hz, respectively. Crack length was measured by direct current potential drop method. Additionally, crack tip on the specimen surface was observed in situ using an optical microscope, KEYENCE VHX-5000. In all tests, creep crack propagation during the tension hold was not observed.

Experimental Results Effect of KI Value of Tension Hold in SC Specimen

Fig. 2 a Two types of crystal orientations of SC specimen and b PC specimen extracted from directionally solidified MGA1400

Fig. 3 Schematic illustration of loading sequences applied in CFCP tests. Unit of stress intensity factors: MPa m1/2

In this section, effect of the KI value of tension hold on the subsequent fatigue crack propagation (FCP) is evaluated, by comparing results of test (1) and (2) where specimens were used. Kmean and Kmax tension holds were applied for 90 min in test (1) and (2), respectively. Figure 4a shows FCPR, da/dN, in test (1) and (2). The horizontal axis indicates projected crack length, a, from load line. Tension holds were introduced at a = 6.5 mm, as indicated by arrows in the figure. Note that fatigue crack propagated from these points after the restart of cyclic fatigue loading, since creep crack propagation was not observed during the tension holds in this study. In test (1) with Kmean tension hold, there was no obvious effect of the tension hold on FCP, i.e., da/dN gradually became higher as DKI increased from 20 MPa m1/2 after the restart of cyclic fatigue loading. By contrast, in test (2) with Kmax tension hold, the restart of cyclic fatigue loading resulted in da/dN almost comparable to that before the tension hold at first. Immediately afterward, da/dN rapidly decreased to about one-hundredth and remained low for a distance of 0.40 mm despite increasing DKI, followed by its rapid acceleration.

Competing Mechanism of Creep Damage and Stress Relaxation … Table 2 Testing conditions of six tests

355

Test No.

Specimen

KI during tension hold (MPa m1/2)

(1)

Kmean (23.3)

90

Kmax (33.3)

90

(2) (3)

Hold time (min)

30

(4)

180

(5)

90

(6)

Polycrystalline

90

Figure 5a, b shows crack tips observed in situ by the microscope at the end of tension holds in test (1) and (2), respectively. In test (2) with Kmax tension hold, localized strain was clearly observed at the crack tip, as typically shown in Fig. 5b. By contrast, nothing noteworthy was observed in test (1) with Kmean tension hold (Fig. 5a). Figure 5c, d show the crack tip observed at 200th and 1000th cycle after the cyclic fatigue loading was restarted in test (2). At the 200th cycle (Fig. 5c), a nascent crack of 36 lm was initiated with da/dN of 1.8  10−7 m/cycle during these 200 cycles. At the 1000th cycle (Fig. 5d), the nascent crack length was 59 lm, and da/dN during these 800 cycles was rapidly decreased to 2.9  10−8 m/cycle.

Effect of Crystal Orientation in SC Specimen Figure 7 shows crack path and FCPR, da/dN, obtained in test (5) using specimen, where Kmax tension hold for 90 min was applied. Scales in horizontal axes of upper picture and bottom graph are equivalent. For a comparison, da/dN obtained in test (2) using specimen under the same load condition is also plotted in the figure. Although da/dN in specimen was basically higher than that in specimens, overall phenomenon induced by the tension hold seems comparable. It is noteworthy that the crack path was inclined to the loading direction, which implies stress distribution at crack tip are not symmetric as of typical Mode I crack.

Effect of Tension Hold Time in SC Specimen

Discussion In this section, effect of the tension hold time on the subsequent FCP is evaluated, by comparing results of test (2)– (4) where specimens were used. Kmax tension holds were introduced for 90, 30 and 180 min in test (2), (3) and (4), respectively. Figure 4b shows FCPR, da/dN, obtained in test (2) and (3). In test (2) with 90 min tension hold, da/dN rapidly decreased to one-hundredth of that before the tension hold, and remained low for 0.40 mm as described in Fig. 4a. In test (3) with 30 min tension hold, da/dN decreased to one-tenth after the cyclic fatigue loading was restarted. A distance where da/dN decreased was only 0.16 mm, after which it was rapidly accelerated. Figure 4c shows da/dN obtained in test (2) and (4). In test (4) with 180 min tension hold, a magnitude of the decrease in da/dN was similar to that in test (2), whereas the distance where da/dN decreased was longer, 0.60 mm. Figure 6 shows da/dN under the cyclic fatigue loadings after the tension holds in test (2)–(4) as functions of DKI. Here, data during the significant crack retardation are excluded, and only those after the rapid acceleration are plotted. For a comparison, da/dN obtained in a pure FCP test is also plotted in the figure. Interestingly, in all test (2)–(4), da/dN after the rapid acceleration converged on that of pure FCP.

Mechanisms of Transient Crack Propagation Induced by Tension Hold The experimental results in SC specimens are summarized in Table 3. Here, following three characteristics are drawn; 1. When Kmean tension hold was applied, no effect on crack propagation was found. 2. When Kmax tension hold was applied, nascent crack was immediately initiated after the restart of cyclic fatigue loading. Subsequently, FCPR rapidly decreased and remained low for certain distance despite increasing DKI, followed by its rapid acceleration and convergence on CPR of pure fatigue. 3. The longer hold time resulted in larger magnitude of decrease in FCPR and longer distance of crack retardation. To explain the phenomena described above, two possible mechanisms can be proposed. The first mechanism is based on the residual stress concept (Fig. 8a). During the tension hold, (1) material degradation due to void nucleation/growth, c′ depletion and c/c′ rafting and (2) stress relaxation may occur at crack tip. Here, (1) the material degradation lowers

356

S. Suzuki et al.

Fig. 5 Crack tips observed in situ a after 90 min tension hold in test (1), and b in test (2). c Crack tips at 200th cycle and d 1000th cycle after the restart of cyclic fatigue loading in test (2)

Fig. 6 Comparison of da/dN in test (2), (3) and (4) with that of pure FCP

Fig. 4 da/dN a in test (1) and (2), b in test (2) and (3), c in test (2) and (4)

crack propagation resistance, whereas (2) the stress relaxation results in residual compressive stress after the tension hold and reduces crack driving force. Thus, when the cyclic fatigue loading is restarted, at first, nascent crack initiation is

promoted by (1) the material degradation. After the crack propagates through the area of (1) material degradation, the effect of (2) stress relaxation and resultant residual compressive stress becomes dominant and retards crack propagation. The second mechanism is based on the crack closure concept (Fig. 8b). During the tension hold, crack tip blunting occurs accompanying with tensile creep deformation. When the cyclic fatigue loading is restarted, stress intensity at crack tip opening, Kop, is lowered because of the crack tip blunting. Thus, effective DK, DKeff, is temporarily increased

Competing Mechanism of Creep Damage and Stress Relaxation …

357

amount of creep deformation, hence result in the larger decrease in FCPR and the longer distance of retardation, as described in characteristic (III). The retardation was not observed after Kmean tension hold as described in characteristic (I) possibly because the load was not large enough to generate sufficient creep deformation at crack tip, as shown in Fig. 5a. In below sections, from the viewpoint of the mechanism based on residual stress concept, (1) the material degradation at crack tip was investigated using scanning electron microscope (SEM), while (2) the stress relaxation was evaluated by finite element analysis (FEA) coupled with digital image correlation (DIC) technique.

SEM Observation on Material Degradation at Crack Tip

Fig. 7 a Crack path and b da/dN obtained in test (5) using specimen

(C and D in Fig. 8b), and promotes nascent crack initiation. Once the nascent crack is initiated, Kop gradually becomes higher due to an enhanced crack face contact caused by a residual creep deformation in crack wake (E and F in Fig. 8b). This reduces the DKeff and retards FCP. Either or both of the two mechanisms may be responsible for the phenomena in test (2)–(5), where the nascent crack was immediately initiated when the cyclic fatigue loading is restarted, after which FCPR rapidly decreased, as described in characteristic (II). Crack tip blunting and oxidationinduced crack closure do not seem to contribute to the crack retardation because the retardation occurred after the nascent crack initiation. FCPR was accelerated and it converged on that of pure FCP possibly because crack propagated through the area where the effect of creep deformation prevails. Furthermore, longer hold time should generate a larger

Table 3 Summary of experimental results using SC specimens

To investigate the material degradation at crack tip, another CFCP test with Kmax tension hold for 180 min was conducted using specimen. The test was interrupted at the end of tension hold, and the crack tip was observed by SEM (KEYENCE VE-9800). Figure 9a shows SEM image at the crack tip. A void can be clearly observed at about 2 lm ahead of the crack tip. Figure 9b, c shows distribution of Oxygen and Aluminum near the crack tip analyzed by energy dispersive X-ray spectroscopy. Bright area in the Oxygen map (Fig. 9b) and dense distribution of Aluminum in the corresponding area (Fig. 9c) indicate thick Al-rich oxides on the crack surface. Behind the oxides layer, there was a depleted zone of Aluminum of about 2 lm thickness implying c′ depletion. Although some evidence of the material degradation such as the void and the c′ depletion were observed, their scales were only few micrometers from the crack tip and did not seem to cause the nascent crack initiation of tens of micrometers as observed in the CFCP tests. Several authors have reported that tension hold at high temperature induces c/c′ rafting at crack tip resulting in faster crack propagation along the planar microstructure [2, 12], which implies further investigation is necessary in the present study.

Test No.

Crystal orientation

KI during tension hold (MPa m1/2)

(1)

Kmean (23.3) Kmax (33.3)

FCPR after tension hold

Retardation (mm)

90

Not changed

Not observed

90

1/100

0.40

(3)

30

1/10

0.16

(4)

180

1/100

0.60

90

1/100

0.36

(2)

(5)

Hold time (min)

358

S. Suzuki et al.

Fig. 8 Schematic illustration of two mechanisms based on a residual stress concept and b crack closure concept

was in loading direction and in crack propagation direction, and Mode I crack of a = 6.5 mm was introduced as “seam crack,” which correspond to testing conditions in test (1)–(4). Mesh size at very vicinity of the crack tip was about 10 lm. Element type was 20-node reduced-integration brick elements, C3D20R. Numbers of elements and nodes were 25,648 and 116,171, respectively. Two types of load conditions, A and B, were applied in the FEA. The condition A was monotonic tension hold of Kmax = 33.3 MPa m1/2 for 180 min, followed by one cycle of fatigue loading of DKI = 20 MPa m1/2 at 10 Hz. The condition B was only one cycle of pure fatigue loading of DKI = 20 MPa m1/2 at 10 Hz. Elastic-plastic anisotropy of single-crystal superalloy at 900 °C were defined by material properties referred from a work on CMSX-4 by Kagawa and Mukai [13] as listed in Tables 4 and 5. For the plastic anisotropy, Hill yield criterion was applied. Creep law in this study was defined as following equations aiming to simulate transient and steady-state creep deformation. Fig. 9 SEM observation on crack tip after Kmax tension hold for 180 min

Elastic-Plastic-Creep FEA Coupled with DIC Three-dimensional FE model of the CT specimen used in the CFCP tests was created by commercial FEA software ABAQUS (abaqus/2017). Material orientation of the model

e_ ¼ C_ expðueÞ

ð1Þ

where e_ is creep strain rate, C_ is initial strain rate and u is hardening coefficient. According to Reed et al. [14], C_ can be expressed as a function of stress and temperature:   Q C_ ¼ a exp br  ð2Þ RT

Competing Mechanism of Creep Damage and Stress Relaxation … Table 4 Independent elastic constants of CMSX-4 at 900 °C [13] C11 (GPa)

C12 (GPa)

C44 (GPa)

225

150

88

Table 5 0.2% Yield strength of CMSX-4 at 900 °C [13]. Strain rate was 7.5%/min 0.2% Yield strength (MPa)

1029

913

Table 6 List of literature data of uniaxial tensile creep test of CMSX-4 in orientation Author Svoboda et al. [15] Kakehi et al. [16] MacLachlan et al. [17] Epishin et al. [18]

Temperature (°C) 750

Stress (MPa) 650

750

800

900

400

950

250

1288

10

Table 7 Calibrated values of a, b, Q and u used in creep law a (/s)

8:73  105

b (/Pa)

1:65  108

Q (J/mol)

3:5  105

u

111

where a, b and Q are constants, R is gas constant. In this study, the constants a, b, Q and u were calibrated using uniaxial creep data of CMSX-4 in direction under different stress and temperature conditions from literatures [15–18] listed in Table 6. The calibrated values were shown in Table 7. The creep law was incorporated into ABAQUS via user subroutine “CREEP.” For model verification, DIC technique was employed to experimentally obtain strain field at crack tip. Owing to high temperature at 900 °C, special attentions were paid as follows: • Random pattern was created using heat resistant paints. • Optical band-pass filter which transmits blue light was used to suppress heat radiation effect. • Air flow near the specimen surface was controlled using air nozzle to suppress heat haze effect. • Luminance values were averaged over 9 s for each digital image to minimize the heat haze effect.

359

Monochrome CMOS camera (XIMEA MQ042-MG-CM) and single focus lens (Nikon AF MICRO NIKKOR 200 mm) were used for optical system. An open source DIC program “Ncorr v1.21” built by Blaber et al. [19] was employed. Figure 10a shows strain along the loading direction on the model surface at 0 min and 180 min of Kmax tension hold under load condition A. The horizontal axis indicates distance from the crack tip. For comparison, the results of DIC under the same conditions of crack length, load condition and crystal orientation are also plotted. In a region less than 0.2 mm from the crack tip, the strain measured by DIC became smaller as it came closer to the crack tip compared with the FEA result. This was possibly because of a limitation of the random pattern to follow large strain in the vicinity of crack tip. By contrast, in the other region, the discrepancy between FEA and DIC was quite small. Figure 10b shows the strain at 0.03, 0.20, 0.40 and 1.00 mm from the crack tip as function of time, which also indicates good agreement between FEA and DIC. Thus, we could say the elastic-plastic-creep model in this study can fairly simulate the deformation of SC specimen. Figure 11a shows stress along the loading direction in mid-plane of the model from 0 to 180 min during Kmax tension hold under load condition A. The horizontal axis indicates distance from the crack tip. In a region less than 0.5 mm from the crack tip, the stress became lower with time, which indicates the stress relaxation at crack tip during the tension hold. Figure 11b shows the stresses at Kmax after one cycle of fatigue loading under both load conditions A and B. Under load condition A, the stress value was equivalent to that at 180 min of the tension hold (Fig. 11a) indicating that the relaxed stress remained as residual compressive stress. More importantly, at very vicinity of the crack tip, the stress under load condition A was 20% lower than under condition B of pure fatigue loading, which implies reduction in crack driving force. Although the stress relaxation during tension hold was evaluated in this study, quantification of FCPR during the retardation was not accomplished. Furthermore, solid evidence of the material degradation at crack tip to cause the nascent crack initiation was not discovered. To explain the transient crack propagation induced by tension hold, further investigation is necessary taking also possible c/c′ rafting at the crack tip and the crack closure concept into consideration.

Effect of Grain Boundary in PC Specimen The phenomena described above are what is observed in SC specimen. It is expected that, in PC specimen where GBs

360

S. Suzuki et al.

Fig. 10 Comparison of strain along the loading direction between FEA and DIC as a function of a distance from crack tip and b time. Kmax tension hold for 180 min was applied as load condition A

Fig. 11 Stress along the loading direction as a function of distance from crack tip in FE model. a Stress at 0, 30, 90 and 180 min during Kmax tension hold under load condition A. b Comparison of stresses at Kmax after one cycle of fatigue loading between load conditions A and B

exist, the effect of tension hold would be very different, thus is investigated in this section. Figure 12 shows crack path and FCPR, da/dN, obtained in test (6) with Kmax tension hold for 90 min using PC specimen. Scales in horizontal axes of upper picture and bottom graph are equivalent. Contrary to the expectation, the crack retardation was observed after the restart of cyclic fatigue loading, as similar to SC specimen. This result contradicts a general understanding that GB is responsible for acceleration of crack propagation at high temperature. In wrought superalloys, during tension hold, GBs ahead of crack tip are heavily oxidized and embrittled being assisted by stress concentration, as denoted by “stress assisted grain boundary oxidation (SAGBO)” and “dynamic embrittlement (DE).” In this study, PC specimen had the average grain size of about 1.3 mm, which was significantly larger than micron-order grain size of the wrought superalloys. Thus, it can be speculated that the GB does not necessarily play the harmful role on crack propagation if the grain size is large enough, and instead, the creep deformation inside grains plays dominant role to retard the crack propagation, as schematically illustrated in Fig. 13a. If the effect of creep at

Fig. 12 a Crack path and b da/dN obtained in test (6) using PC specimen

Competing Mechanism of Creep Damage and Stress Relaxation …

Fig. 13 Schematic illustration of crack propagation behavior when creep affected zone a is small compared to grain size and b extends over sufficient numbers (n) of grains

crack tip extends over sufficient numbers (n in Fig. 13) of grains, accelerated intergranular cracking may occur due to GB embrittlement (Fig. 13b). A study to evaluate the “creep affected zone, rcr” in different materials of various grain sizes is now undergoing, and criteria of the transition from retardation to acceleration is going to be clarified based on a relative grain size to rcr, considering the effects of temperature, stress intensity and hold time.

Conclusion In this study, the effect of creep deformation at crack tip on fatigue crack propagation behavior in a single-crystal superalloy, ICMSX-4, and a directionally solidified superalloy, MGA1400, at 900 °C was investigated. Creep-fatigue crack propagation tests with single tension hold introduced into cyclic fatigue loading were conducted. Effects of KI value of tension hold, hold time, crystal orientation and grain boundary (GB) were evaluated. Following conclusions can be drawn: 1. In single-crystal specimens extracted from ICMSX-4, when the cyclic fatigue loading was restarted after the tension hold, nascent crack was immediately initiated followed by significant crack retardation. This crack propagation behavior was ascribed to mechanisms based on two different concepts of residual compressive stress and enhanced crack closure induced by the tension hold. 2. From the viewpoint of residual stress concept, material degradation at crack tip was investigated using scanning electron microscope (SEM), while stress relaxation and the resultant residual compressive stress were quantified by elastic-plastic-creep finite element analysis (FEA) coupled with digital image correlation (DIC) technique. Little evidence of the material degradation such as creep

361

void and c′ depletion was found by SEM. FEA and DIC revealed 20% of stress was relaxed due to the tension hold, however, quantification of crack propagation rate was not accomplished. Further investigation is necessary taking also possible c/c′ rafting at the crack tip and the crack closure concept into consideration. 3. In polycrystalline specimen extracted from MGA1400, despite the presence of GBs, the crack retardation was observed after the tension hold. It was speculated that crack propagation is retarded by the creep deformation inside grains when a relative grain size to “creep affected zone, rcr” is large enough, whereas accelerated intergranular cracking may occur due to GB embrittlement if rcr extends over sufficient numbers of grains.

References 1. Shahinian P, Sadananda K (1984) Creep and Fatigue Crack Growth in Several Cast Superalloys. Paper Presented at the 5th International Symposium on Superalloys, Champion, Pennsylvania, 7–11 October 1984 2. Okazaki M, Yamazaki Y (1999) Creep-fatigue small crack propagation in a single crystal Ni-base superalloy, CMSX-2, Microstructural influences and environmental effects. Int. J. Fat. 21 (1):S79–S86. https://doi.org/10.1016/s0142-1123(99)00058-4 3. Palmert F, Moverare J, Gustafsson D, Busse C (2018) Fatigue crack growth behaviour of an alternative single crystal nickel base superalloy. Int. J. Fat. 109:166–181. https://doi.org/10.1016/j. ijfatigue.2017.12.003 4. Sadananda K, Shahinian P (1978) Hold-time effects on high temperature fatigue crack growth in Udimet 700. J. Mater. Sci. 13 (11):2347–2357. https://doi.org/10.1007/bf00808048 5. Liu X, Kang B, Chang KM (2003) The effect of hold-time on fatigue crack growth behaviors of WASPALOY alloy at elevated temperature. Mater. Sci. Eng. A 340(1–2):8–14. https://doi.org/10. 1016/s0921-5093(02)00074-6 6. Taira S, Ohtani R, Komatsu T (1979) Application of J-Integral to High-Temperature Crack Propagation: Part II—Fatigue Crack Propagation. J. Eng. Mater. Technol. 101(2):162–167. https://doi. org/10.1115/1.3443669 7. Suzuki S, Sakaguchi M (2020) Fatigue crack retardation associated with creep deformation induce d by a tension hold in a single crystal Ni-base superalloy. Scr. Mater. 178: 346–350. https://doi. org/10.1016/j.scriptamat.2019.11.058 8. ASTM International (2008) Standard Test Method for Measurement of Fatigue Crack Growth Rates. Annual Book of ASTM Standards E647-08, ASTM International, West Conshohocken, Pennsylvania, pp 669–713 9. Suzuki S, Sakaguchi M, Inoue H (2018) Temperature dependent fatigue crack propagation in a single crystal Ni-base superalloy affected by primary and secondary orientations. Mater. Sci. Eng. A 724(2):559–565. https://doi.org/10.1016/j.msea.2018.03.090 10. Sakaguchi M, Komamura R, Chen X, Higaki M, Inoue H (2019) Crystal plasticity assessment of crystallographic Stage I crack propagation in a Ni-based single crystal superalloy. Int. J. Fat. 123:10–21. https://doi.org/10.1016/j.ijfatigue.2019.02.003 11. Chen X, Sakaguchi M (2020) Transition behavior from Mode I cracking to crystallographic cracking in a Ni-base single crystal

362

12.

13.

14.

15.

S. Suzuki et al. superalloy. Int. J. Fat. 132:105400. https://doi.org/10.1016/j. ijfatigue.2019.105400 Ott M, Mughrabi H (1999) Dependence of the high-temperature low-cycle fatigue behaviour of the monocrystalline nickel-base superalloys CMSX-4 and CMSX-6 on the c/c′-morphology. Mater. Sci. Eng. A 272(1): 24–30. https://doi.org/10.1016/s09215093(99)00453-0 Kagawa H, Mukai Y (2012) The Effect of Crystal Orientation and Temperature on Fatigue Crack Growth of Ni-based Single Crystal Superalloy. Paper Presented at the 12th International Symposium on Superalloys, Champion, Pennsylvania, 9–13 September 2012 Reed RC, Matan N, Cox DC, Rist MA, Rae CMF (1999) Creep of CMSX-4 Superalloy Single Crystals: Effects of Rafting at High Temperature. Acta Mater. 47(12):3367–3381. https://doi.org/10. 1016/s1359-6454(99)00217-7 Svoboda J, Lukáš P (2000) Creep deformation modelling of superalloy single crystals. Acta Mater. 48(10):2519–2528. https:// doi.org/10.1016/s1359-6454(00)00078-1

16. Kakehi K, Takahashi S (2005) Influence of Aging Heat Treatment on Creep Strength of CMSX4. J. Soc. Mater. Sci., Jpn. 54(2):136– 142. https://doi.org/10.2472/jsms.54.136 17. MacLachlan DW, Wright LW, Gunturi S, Knowles DM (2001) Constitutive modelling of anisotropic creep deformation in single crystal blade alloys SRR99 and CMSX-4. Int. J. Plast. 17(4):441– 467. https://doi.org/10.1016/s0749-6419(00)00058-9 18. Epishin A, Fedelich B, Nolze G, Schriever S, Feldmann T, Ijaz MF, Viguier B, Poquillon D, Bouar YL, Ruffini A, Finel A (2018) Creep of Single Crystals of Nickel-Based Superalloys at Ultra-High Homologous Temperature. Metall. Mater. Trans. A 49 (9):3973–3987. https://doi.org/10.1007/s11661-018-4729-6 19. Blaber J, Adair B, Antoniou A (2015) Ncorr: open-source 2D digital image correlation matlab software. Exp. Mech. 55:1105– 1122. https://doi.org/10.1007/s11340-015-0009-1

Part IV Component Manufacture and Repair

Microstructure and Material Properties of Alloy 718/713LC Joints Using Orbital Friction Welding Björn Hinze

Abstract

Introduction

This work demonstrates on flat test specimens that orbital friction welding is a suitable process to join a typical disk alloy 718 and a typical blade alloy 713LC. The capability to join nickel-based blades on disks enables the manufacture of blisks instead of the typically used bladed disks. Blisks have the potential to save weight within an aero engine. Orbital friction welding is an alternative process to linear friction welding to enable blade and disk joints. The advantage of orbital friction welding is that the movement is homogenous and the amplitude (or eccentricity) is smaller compared to linear friction welding. This allows joints in areas which are geometrically challenging and impossible to join for linear friction welding. Due to the solid state joining process, there is basically no material mixing within the weld, but within both alloys, recrystallization can be seen near the weld line. The result is a fine-grained microstructure within both alloys adjacent to the bond line. While this means only a small microstructural change for alloy 718, it is a major microstructural change for the cast coarse grain alloy 713LC. Consequently, the mechanical properties are changed compared to the base material due to the friction welding process. The microstructural changes of orbital friction welding and their influence on the material properties are presented within this work. Keywords

Dissimilar weld Recrystallization

















Friction welding Turbine blisk Tensile Stress rupture Fatigue

B. Hinze (&) Rolls-Royce Deutschland, Eschenweg 11, 15827 Blankenfelde-Mahlow, Germany e-mail: [email protected]

In today’s aero engines, linear friction welded blisks are state of the art. Such blisks are usually located within the compressor section and are made of titanium alloys. In general, there are studies to transfer the linear friction welding technology to nickel components [1–3] or to use orbital friction welding as alternative manufacturing process [4]. This work combines both approaches and focuses on nickel-based orbital friction welding joints for turbine applications. The difference between compressor and turbine components is that due to the temperatures within the hot gas path, turbine disks and blades are usually manufactured from different alloys. In this study, alloy 718 represents the disk, while cast alloy 713LC represents the blade. Therefore, turbine blisks require dissimilar welds between a disk alloy and a blade alloy. This means that the blade platform cannot be simply integrated within the disk as it is state of the art for compressor blisk applications. Since the exposure of the disk alloy to the hot gas would cause temperatures beyond the temperature capability of the disk alloy. Therefore, the weld line of a turbine blisk would probably be below the platform which is still part of the blade. The gaps between blade platforms depend on the engine but are usually 40%), commonly produced by cast and wrought processes. Conventional ingot-to-billet conversion is an expensive and very complex process, requiring a significant amount of steps to break up the coarse as-cast structure and interdendritic regions. Due to the size of conventional ingots, it is difficult to achieve a uniformly high level of strain for recrystallization, resulting in non-recrystallized regions that retain large unrecrystallized (UnRex) grains with characteristic intragranular precipitates [2]. Non-uniform grain distributions strongly affect the ultrasonic inspectability response, which is used to find flaws such as cracks, porosity, and large inclusions in billet, and also defects in the final component.

© The Minerals, Metals & Materials Society 2020 S. Tin et al. (eds.), Superalloys 2020, The Minerals, Metals & Materials Series, https://doi.org/10.1007/978-3-030-51834-9_43

441

442

M. Pérez et al.

During cogging at subsolvus temperatures, primary c′ precipitates play a key role in both grain size control and in recrystallization mechanisms which operate in Ni-based superalloys. For the 720Li alloy, a significant amount of work has been published on the recrystallization behaviour and the interaction with c’ precipitates [3–6]. However, little work has been done to study the evolution of c′ precipitates during forging operations. Processes such as dissolution, coarsening and coalescence of primary c′ precipitates takes place during processing at subsolvus temperatures, affecting subsequently the recrystallization and microstructural evolution of these alloys at different stages of the cogging process [7]. In addition, intragranular precipitates within large unrecrystallized (UnRex) grains and undissolved secondary c′ precipitates also play an important role on the microstructural evolution of c–c′ Ni-based superalloys [2, 8]. Developing understanding about the evolution of c′ precipitates is key to developing accurate microstructural models for complex materials such as c–c′ Ni-based superalloys during cogging and subsequent hot forging operations. The present work studies the recrystallization mechanisms which operate in the alloy 720Li during hot forging at subsolvus temperatures, paying a special emphasis on the evolution and interaction of intragranular and secondary c′ precipitates on the removal of large UnRex regions.

Material and Experimental Methodology 720Li material was supplied by Aubert & Duval (Les Ancizes, France) in the form of ∅256 mm diameter billet. The chemical composition is given in Table 1. Double truncated cones (DTCs) of ∅120  90 mm were machined along the cogging direction. Hot forging trials were carried out in a 500T hydraulic press at subsolvus temperatures (1100 °C), at a constant forging speed (2 mm/sec) and with die temperatures of 435 °C. The DTCs were preheated by isothermal holding for 1 h at 1100 °C. The transfer time from the furnace to the press was 3 s and the resting time on the bottom die was about 10 s prior to hot forging. In this work, two forging approaches were considered:

intermediate heat treatment. In a second blow, an additional 30% reduction was introduced. Finally, the DTC was cooled in air. For the metallurgical analysis, sections of DTCs were cut by Electrical Discharging Machining (EDM). Surfaces were prepared for metallurgical analysis by conventional grinding and polishing techniques. No etching was applied on the metallographic samples. Microstructural analysis of the hot forged DTCs was conducted at 8 positions along the central section, covering a large range of strains [0.3–2 mm/mm] [8]. Backscattered-electron (BSE) images were taken from the 720Li samples. In addition, combined Electron Backscatter Diffraction technique (EBSD) and Energy Dispersive X-Ray Spectroscopy (EDX) analysis were performed using TSL EBSD system attached to a FEI Quanta 650 scanning electron microscope (FEG-SEM) operating at 20 kV. EBSD maps were collected from the processed samples using a step size of 0.2 lm. Channel 5 software provided by HKL Technology was used for data analysis. EDX analysis was used to identify the primary c′ precipitates as those regions with low chromium content [9]. EBSD images are mainly represented by band contrast (BC), local misorientation (LM) and inverse pole figures (IPF) by using Channel 5—Tango software. BC is an electron backscatter diffraction pattern (EBSP) quality parameter that indicated the sharpness of Kikuchi bands, where deformed regions and grain boundaries are associated with low BC values. LM component displays the average misorientation among consecutive data points, discarding misorientations over a certain value associated with subgrain or grain boundaries. IPF maps show the crystallographic direction of the orientation in colour codes. In some cases, BC, LM and IPF figures are combined with EDX images and grain boundary maps (GB), where red, black, and blue lines correspond, in most cases, to low angle (LAGB, 2° < h < 15°), high angle (HAGB, h > 15°) and annealing twin boundaries (R3, 60°), respectively.

Results As-Received Condition

i. Single-step forging operation: 60% height reduction was introduced in one single blow, followed by air cooling (AC); ii. Two-step forging operation: An initial 30% reduction was introduced in the first blow followed by an

Table 1 Chemical composition of 720Li billet material

720Li in the as-received condition (billet material) is characterized by the presence of retained large unrecrystallized grains with a fine distribution of intragranular c′ precipitates strongly aligned in certain directions, see Fig. 1. Significant

Element

C

Cr

Mo

W

Al

Ti

Co

Ni

wt%

0.014

16.0

3.0

1.3

2.5

5.0

14.5

57.5

Gamma Prime Precipitate Evolution During … Fig. 1 SEM analysis of 720Li in the as-received condition (billet material) showing the presence of regions with a fine distribution of c′ precipitates (large UnRex regions)

443

1 2

50 μm

a)

b)

Fig. 2 Hetero-epitaxial grains on 720Li in the as-received condition (billet material)

differences in the size and distribution of primary c′ precipitates can be detected between these large UnRex grains and the recrystallized (Rex) regions. At higher magnifications, Fig. 2a shows the microstructure of 720Li billet material, where coherent c-shells around primary c′ precipitates were found. Each primary c′ precipitate is associated with only one c-grain (embedded). The secondary c′ precipitates show a butterfly-shape morphology, associated with slow cooling rates of the billet cooled in air (Fig. 2b). In addition, elongated (fan-type) secondary c′ precipitates were observed. Coupled EBSD-EDX analysis was conducted to separately analyse the Rex (Fig. 3a–d) and UnRex regions (Fig. 3e–h) for 720Li alloy in the as-received condition. These Rex and UnRex regions correspond to the square No. 1 and 2, respectively (Fig. 1a): • In the Rex regions, significant bulging and serrated grain boundaries can be seen (yellow arrows, Fig. 3b) in a fully recrystallized structure, as confirmed by the LM map (Fig. 3d). The EBSD-EDX analysis also confirms that the primary c′ precipitates are not incoherent particles located in the grain boundaries, but they are embedded within c grains, sharing the same crystallographic orientation (hetero-epitaxial pairs), even annealing twin boundaries (R3) are extended to the matrix grain (white arrows, Fig. 3c).

• Concerning the large UnRex regions, a mixture of Rex grains with R3 in their interior, and a well-developed substructure formed by subgrains with a high density of LAGB around intragranular c′ precipitates was found (Fig. 3f). The LM map (Fig. 3h) clearly suggests high recrystallization fractions in the interior of these apparent UnRex regions.

As-Forged Condition From the 720Li DTC forged in 1 blow, Fig. 4 shows the EBSD-EDX analysis of 4 selected regions, covering a large range of strain distribution (e  0.3 − 2). High fractions of strain accumulation and deformed structures can be seen in the LM maps (Fig. 4m–p) with a large density of LAGBs and MAGBs (BC maps, Fig. 4e–h), especially in the region with a lowest strain level (e  0.3). Large fractions of undissolved secondary c′ precipitates were found within the deformed structures (e  0.3, Fig. 4a), in contrast with the highly recrystallized structures, where much smaller fractions of secondary c′ precipitates can be found (e  2, Fig. 4d). The recrystallized regions are characterized by small recrystallized grains, devoid of secondary c′ precipitates, but with embedded small primary c′ precipitates. A significant larger density of primary c′ precipitates can be

444

M. Pérez et al.

Fig. 3 EBSD-EDX analysis of 720Li in the as-received condition (billet material): Recrystallized (Rex) and apparent unrecrystallized regions (UnRex) with intragranular c′ precipitates

seen in the as-forged condition (Fig. 3f–h) in direct comparison to both the Rex regions in the as-received condition (Fig. 3a–d) and the DTC region with the lowest level strain (e  0.3, Fig. 4e). The microstructure of 720Li DTC forged in 2 blows (Fig. 5) shows fully recrystallized structures (Fig. 5m–p), without any internal substructure and LAGBs (Fig. 5e–h). No apparent dependency of the microstructural evolution with the strain level was found. In contrast with the DTC forged in 1 blow, no presence of secondary c′ precipitates was found (Fig. 5a–d), in good agreement with the observed fully recrystallized structures. Similarly to the DTC forged in 1 blow, a larger number of fine primary c′ precipitates can be observed (Fig. 5f–h) as compared to the as-received condition (Fig. 3a–d) but also to the region with the lowest strain level (e  0.3, Fig. 5e). However, the fine c′ precipitates are organized into clusters of c′ precipitates with different crystal orientation, being located at grain boundaries and triple points of equiaxed grains (Fig. 5f–h). Hetero-epitaxial grains, easily detected in the as-received condition, have been highly consumed by new Rex grains. Note that the largest grains correspond to those grains either free of primary c′ precipitates or with incoherent intragranular c′ precipitates (Fig. 5i–l).

Discussion In the as-received condition, 720Li billet material is characterized by the presence of apparent large UnRex structures with a fine distribution of (primary) intragranular c′ precipitates, see Fig. 1. These UnRex regions are inherited from the ingot-to-billet conversion (cogging) process [10]. Significant differences in the size and distribution of primary c′ precipitates can be seen between Rex and UnRex regions (Fig. 1). These observations were reported previously for 720Li [8] and for a similar alloy AD730 [2]. In the present work, evidence that hetero-epitaxial recrystallization

mechanism is operating in 720Li was found (Fig. 2), in good agreement with reported literature [11, 12]. As commented previously, coherent c-shells around primary c′ precipitates were found in the as-received condition (Fig. 2a). Fan-type secondary c′ precipitates within hetero-epitaxial grains were found, indicating that the grain growth process of these grains can occur at relatively low temperatures, during the slow cooling of the billet. It is well known that the morphology of secondary c′ precipitates is influenced by the cooling rate. For 720Li, dendritic and fan-type morphologies are reported to be formed by a coupled growth of the c and c′ rods at low cooling rates (  0.12 °C/s) [13]. In addition, significant bulging and serrated grain boundaries, pinned by secondary c′ precipitates of adjacent grains, indicate the occurrence of grain boundary migration (grain growth process). The growth of hetero-epitaxial grains seems to require the dissolution of the secondary c′ precipitates during grain boundary migration. The subsequent re-precipitation behind the grain boundary could explain the formation of the observed fan-type secondary c′ precipitates. Coupled EBSD-EDX analysis on Rex regions (billet material) confirms that the primary c′ precipitates cannot be described as incoherent particles located on the grain boundaries, in fact, they are embedded within c-grains, sharing the same crystallographic orientation (heteroepitaxial pairs). Only a few incoherent primary c′ precipitates of globular morphology and smaller size, separated from the matrix by high angle boundaries, were found (black arrow, Fig. 3b). They are located in the interior of grains (intragranular) with a different crystal orientation. These results could indicate that these c′ precipitates were surpassed during grain boundary migration (recrystallization or grain growth processes), leaving behind the precipitate incoherent within the matrix (grain), tending to adopt a globular morphology, and potentially, being gradually dissolved into the matrix. Concerning the apparent large UnRex regions, a high density of LAGB around the fine primary c’ precipitates was

Gamma Prime Precipitate Evolution During …

445

Fig. 4 EBSD-EDX analysis of 720Li DTC forged in 1 blow (60% reduction)

found (Fig. 3e). A very low dislocation density and no apparent internal misorientations are inferred from the LM and IPF maps, respectively (Fig. 3h–g), indicating a well-developed substructure formed by subgrains (Fig. 3f). Recovery processes of deformed structures promote the formation of the observed subgrain structures [14], being the LAGBs associated with dislocation substructure induced by dynamic recovery [15]. Subsolvus temperatures during long periods of time (cogging processes) can explain the inferred low levels of dislocation density (rearrangement) in the apparent UnRex regions (Fig. 3h). However, it seems that competing annealing phenomena, such as recovery and recrystallization processes, are operative in the apparent large UnRex regions [14]. Annealed twin boundaries (R3) in the interior of Rex grain, see Fig. 3f, were also found. Grain boundary migration is frequently accompanied by twinning

or multiple twinning [14]. These results are in contrast to those reported for AD730 billet material, where high levels of strain accumulation were found for UnRex regions with a similar fine distribution of primary c′ precipitates [2]. In 720Li, the small c′ precipitates also share the same crystallographic orientation with the Rex c-grains (with R3 boundaries) and subgrains (Fig. 3g), denoting the occurrence of hetero-epitaxial nucleation/recrystallization in UnRex regions. In addition, evidence that continuous dynamic recrystallization mechanisms (CDRX) are also operating in 720Li alloy in both Rex and UnRex regions was found in the as-received condition. A progressive subgrain rotation can explain the transition from LAGBs to MAGBs and HAGBs around the fine intragranular c′ precipitates in UnRex regions (Fig. 3f) [16]. In Rex regions, the presence of LAGB and

446

M. Pérez et al.

ε ≈ 0.3 (0.05 + 0.25)

ε ≈ 2 (0.9 + 1.1)

ε ≈ 1.2 (0.5 + 0.7)

ε ≈ 0.6 (0.2 + 0.4) b)

c)

d)

e)

f)

g)

h)

i)

j)

k)

l)

m)

n)

o)

p)

LM + primary γ’

IPF + primary γ’

BC + primary γ’

SEM image

a)

Fig. 5 EBSD-EDX analysis of 720Li DTC forged in 2 blows with an intermediate heat treatment (30% + HT + 30% reduction)

MAGBs around the primary c′ precipitates (red arrows, Fig. 3b) indicates the rotation of the primary c′ precipitates with respect to the original hetero-epitaxial grain during hot forging (cogging) operations, resulting potentially into the splitting of the original hetero-epitaxial pair into two new grains: (1) A refined hetero-epitaxial grain (c′ precipitate/cgrain pair), and; (2) A grain devoid of the primary c′ precipitate. In the as-forged condition, remarkable differences in the recrystallization behaviour were found between the DTCs forged in 1 and 2 blows with an intermediate heat treatment, see Figs. 4 and 5, respectively. The microstructure of the DTC forged in 1 blow is characterized by the presence of high fractions of deformed structures (Fig. 4); unrecrystallized structures can be seen even in those regions subjected to high levels of deformation (e  2). A strong dependency

of the strain with the microstructural evolution was observed. In the region of the lowest deformation (e  0.3), a fully unrecrystallized structure was found. As the strain increases, higher fractions of recrystallization levels were observed [8]. In contrast with the as-received condition (Fig. 3a–d), the recrystallization structures in the as-forged condition (1 blow) are formed by small grains with embedded fine c′ precipitates. The deformed grains, instead, tend to be of larger dimensions than the Rex grains. They are elongated in the forging plane and devoid of primary c′ precipitates. Another main difference is the presence of large fractions of undissolved secondary c′ precipitates of globular morphology found within deformed structures, in contrast with the highly recrystallized structures, free of secondary c′ precipitates (Fig. 6). In the range of 1040–1130 °C, the dissolution of secondary c′ phases is expected [7], but it

Gamma Prime Precipitate Evolution During … Fig. 6 Presence of non-dissolved secondary c′ precipitates on unrecrystallized regions from the 720Li DTC forged in 1 blow (60%)

447

ε ≈ 1.2

ε ≈ 0.6

a)

b)

seems that was not the case in this work. These results suggest that the observed grain refinement is attributed to the strong interaction (Zener pinning effect) of the undissolved c′ secondary precipitates, retarding recrystallization mechanisms such as DDRX during hot forging, and therefore, promoting strain accumulation and the occurrence of CDRX. The latter is confirmed by the fully Rex structures (100%) found on the DTC forged in 2 blows (Fig. 5). Equiaxed grain structures were observed, with most of the primary c′ precipitates located on grain boundaries. No significant microstructural differences as a function of strain (e  0.3 − 2) were found. These results are in good agreement with reported literature, where insensitivity of grain microstructures to deformation parameters was found on 720Li during hot deformation below c′ solvus temperatures [3, 7]. Figure 7 shows the SEM micrographs of the regions with the lowest and highest level of deformation (e  0.3 and 2) for both 720Li DTCs (1 vs. 2 blows), showing remarkable differences in terms of undissolved secondary c′ precipitates. Only a small number of secondary c′ precipitates, located mainly on grain boundaries, can be seen on the DTC forged in 2 blows (Fig. 7c, d) in contrast with the large density of intragranular c′ precipitates observed in the DTC forged in 1 blow (Fig. 7a, b). The large structural differences and recrystallization behaviour found between both DTCs can be explained by the dissolution of secondary c′ precipitates prior to the second blow of deformation. Mechanisms such as dynamic coarsening and dissolution of the secondary c′ precipitates during hot deformation (strain-assisted) could play an important role [17], especially in those regions with significant levels of deformation. In principle, dynamic dissolution can be accelerated further in those UnRex regions because of large density of grain boundaries (shorter diffusion paths). However, the observed preconditioning of 720Li microstructure prior to the second blow of deformation can only be explained by the dissolution of the secondary c′ precipitates in the course of the intermediate heat treatments, promoting static recrystallization processes [8]. A microstructure free of secondary c′ precipitates would

require much lower dislocation densities (lower strain levels) and associated driving forces to activate DRX mechanisms by strain-induced boundary migration (SIBM). Significant differences in the distribution, size and location of primary c′ precipitates were also found across the as-received and as-forged conditions for both DTCs forged in 1 and 2 blows, requiring a separate explanation. In the as-received condition, clear differences between Rex and UnRex were found, based on the distribution and size of primary c′ precipitates. However, after forging, such differences are far less evident. In the DTC forged in 1 blow, Rex and UnRex regions mix together, making it difficult to conduct a separate analysis of the microstructural evolution of Rex and UnRex regions from billet material during subsequent forging operations. With exception of the region with the lowest level of deformation (e  0.3, fully UnRex), the microstructure of the DTC forged in 1 blow shows a fine distribution of primary c′ precipitates (Fig. 4) embedded within small Rex grains, resembling the apparent UnRex regions in the as-received condition (Fig. 3e–h). In the DTC forged in 2 blows, fully recrystallized structures were found, regardless of the strain level. The presence of regions with a fine distribution of small primary c′ were not evident in this case, or at least they are far less obvious than in the as-received condition. However, clusters of fine primary c′ precipitates, located mainly on grain boundaries of fully Rex grains, were observed (Fig. 5). Mechanisms of coalescence of primary c′ precipitates with different crystal orientation are clearly operative at this stage. Based on the significant similarities found between the fine primary c′ precipitates in UnRex regions in the as-received and those found in the as-forged conditions, it is possible to explain the evolution and removal of the apparent large UnRex regions as follows; During the first blow of deformation, the misorientation of the subgrain structure in UnRex regions increases from LAGB/MAGB to HAGB, with the consequent formation of small new grains with embedded fine c′ precipitates in their interior. As discussed previously, the presence of undissolved secondary c′

448

ε ≈ 2.0

a)

b)

c)

d)

DTC forged 1 blow

ε ≈ 0.3

DTC forged 2 blows

Fig. 7 Differences in the presence of non-dissolved secondary c′ precipitates between the 720Li DTCs forged in 1 and 2 blows with an intermediate heat treatment

M. Pérez et al.

precipitates promotes the occurrence of CDRX mechanism at the expense of DDRX. This latter one can also explain the observed shape change and strain accumulation on the original Rex grains of larger dimensions, as those observed in the as-received condition. In the as-forged condition, these deformed grains do not contain apparently primary c′ precipitates. This provides support to the already described mechanism where the primary c′ precipitates can split from the original hetero-epitaxial grains as a result of the strain accumulation/primary c′ precipitates rotation. During intermediate heat treatments, almost full dissolution of the secondary c′ precipitates and the occurrence of static recrystallization and grain growth processes take place [8]. This results in a preconditioned microstructure prior to second blow of deformation, promoting recrystallization mechanisms by SIGB. The reorganization of the fine distribution of small primary c′ precipitates into clusters of coalesced c′ precipitates, increasing the inter-particle distance and therefore the formation of coarser equiaxed grains, represents a clear transition from the apparent UnRex regions to the fully Rex regions. Reported mechanisms for low driving forces, such as grain boundary migration together with second particles and displacement of second particles through the grain boundary, can explain the formation of the observed clusters of c′ precipitates [7, 14]. Uneven distribution of primary c′ precipitates increase the possibilities of the occurrence of recrystallization and grain boundary by SIBM [16]. In the present work, no apparent coarsening of the fine c′ precipitates was observed. It seems that further holding times at subsolvus temperatures would

be required to complete the coarsening of the observed clusters of small primary c′ precipitates, resulting in larger c′ precipitates. Recrystallization mechanisms along with the subsequent deformation step (2nd blow) could promote the coalescence of the fine primary c′ precipitates, originally formed in the interior of the UnRex regions (intragranular c’ precipitates). However, the little or no microstructural differences found across the DTC forged in 2 blows suggests that strain plays a secondary role on the transition from UnRex to fully Rex grains. Intermediate heat treatment is clearly playing an instrumental role on the recrystallization behaviour for 720Li alloy with the presence of undissolved secondary c′ precipitates, and on the removal of undesired UnRex regions.

Conclusion • 720Li in the as-received condition (billet) is characterized by the presence of large UnRex structures with a fine distribution of (primary) intragranular c′ precipitates. The UnRex regions present a combination of fully recrystallized grains and a well-developed subgrain structure. The intragranular c′ precipitates share the same crystallographic orientation with the Rex c-grains and/or subgrains. • Coherent c-shells around primary c′ precipitates, sharing the same crystallographic orientation, were found in the as-received condition denoting the occurrence of

Gamma Prime Precipitate Evolution During …

hetero-epitaxial nucleation/recrystallization. The presence of fan-type secondary c′ precipitates within heteroepitaxial grains suggests that grain boundary migration/ grain growth occurs at relatively low temperatures during slow cooling, strongly interacting with secondary c′ precipitates. From these results, the dissolution and re-precipitation of secondary c′ precipitates behind the grain boundary is inferred. • Non-dissolved secondary c’ precipitates play a key role on recrystallization mechanisms which operate during hot forging of 720Li, promoting the retention of high fractions of deformed structures in the DTC forged in 1 blow. Significant differences in the distribution of secondary c′ precipitates between Rex and UnRex regions were found, indicating that the dissolution of the secondary precipitates takes place during recrystallization/grain boundary migration processes. • Intermediate heat treatments strongly affect the microstructural evolution of 720Li, by promoting static recrystallization processes and dissolving the secondary c′ precipitates. The latter one results in a strong microstructural conditioning for 720Li prior to the second blow of deformation. Fully recrystallized structures, free of UnRex regions, were observed in the DTC forged in 2 blows, regardless of the strain level. The reorganization of the fine distribution of small primary c′ precipitates into clusters of coalesced c′ precipitates represents the transition from the UnRex to the fully Rex regions.

Acknowledgements The corresponding author would like to acknowledge Aubert & Duval for help provided in terms of supplied material, advice and fruitful discussions. The corresponding author also wants to acknowledge the Advanced Forming Research Centre (AFRC) for its support.

References 1. Forbes Jones RM, Jackman L.A (1999) The Structural Evolution of Superalloy Ingots during Hot Working. JOM 51:27–31. https:// doi.org/10.1007/s11837-999-0007-9. 2. Pérez M, Dumont C, Nodin O, Sebastien N (2018) Impact of forging direction on the recrystallization behaviour of nickel base superalloy AD730 billet material at subsolvus temperatures. Mater Charact 146:169–181. https://doi.org/10.1016/j.matchar.2018.10. 003. 3. Yu QY, Yao Z, Dong JX (2016) Deformation and recrystallization behavior of a coarse-grain, nickel-base superalloy U720 720Li ingot material. Mater Charact 107:398–410. https://doi.org/10. 1016/j.matchar.2015.07.035.

449 4. Liu F, Chen J, Dong J, Zhang M, Yao Z (2016) The hot deformation behaviors of coarse, fine and mixed grain for Udimet 720Li superalloy. Mater Sci Eng A 651:102–115, 2016. https:// doi.org/10.1016/j.msea.2015.10.099. 5. Monajati H, Jahazi M, Taheri AK (2005) Deformation Characteristics of Isothermally Forged Udimet 720 Nickel-Base Superalloy. Metall Mater Trans A 36:895–905. https://doi.org/10.1007/ s11661-005-0284-z. 6. Lindsley B et al. (2000) Sub-solvus recrystallization mechanisms in Udimet alloy 720Li. Paper presented at the 9th Intenational Symposium of Superalloys (Superalloys 2000), Pennsylvania, USA, 17–21 September 2000. 7. Chen J, Dong J, Maicang Z, Yao Z (2016) Deformation mechanism in a fine-grained Udimet 720LI nickel-base superalloy with high volume fractions of gamma prime phases. Mater Sci Eng A 673:122–134. doi:0.1016/j.msea.2016.07.068. 8. Pérez M, Dumont C, Sebastien N - Impact of the forging strategy on the recrystallization behaviour of nickel base superalloy Udimet 720Li billet material at subsolvus temperatures. Paper in progress. 9. Berment-Parr I et al. (2016) Inhomogeneous grain coarsening behavior in supersolvus heat treated nickel-based superalloy RR1000. Paper presented at the 13th Intenational Symposium of Superalloys (Superalloys 2016), Pennsylvania, USA, 11–15 September 2016. 10. Vernier S, Franchet JM, Dumont C, Vennéguès P, Bozzolo N (2018) c′ precipitates with a twin orientation relationship to their hosting grain in a c-c′ nickel-based superalloy. Scripta Mater 153:10–13. https://doi.org/10.1016/j.scriptamat.2018.04.037. 11. Charpagne MA, Billot T, Franchet JM, Bozzolo N (2016) Heteroepitaxial recrystallization: A new mechanism discovered in a polycrystalline c-c’ nickel based superalloy. Alloys Compd 688:685–694. https://doi.org/10.1016/j.jallcom.2016.07.240. 12. Charpagne MA et al. (2016) Heteroepitaxial recrystallization observed in Rene 65 and Udimet 720: A new recrystallization mechanism possible occurring in all low lattice mismatch gamma-gamma’ superalloys?. Paper presented at the 13th International Symposium of Superalloys (Superalloys 2016), Pennsylvania, USA, 11–15 September 2016. 13. Furrer DU, Fecht HJ (1999) c′ formation in superalloy U720LI. Scripta Mater 40:1215–1220. https://doi.org/10.1016/s1359-6462 (99)00094-9. 14. Humphreys FJ, Hatherly M (2004) Recrystallization and related annealing phenomena. In: Pergamon 2004. 15. Lin YC, Wu XY, Xiao-Min C, Chen J, Wen DX, Zhang JL, Ti LT (2015) EBSD study of a hot deformed nickel-base superalloy. Alloys Compd 640:101–113. https://doi.org/10.1016/j.jallcom. 2015.04.008. 16. Mandal S, Bhaduri AK, Subramanya Sarma V (2011) A study on microstructural evolution and dynamic recrystallization during isothermal deformation of a Ti-modified austenitic stainless steel. Metall. Mater. Trans. A 42:1062–1072. https://doi.org/10.1007/ s11661-010-0517-7. 17. Semiatin SL, Shank JM, Shiveley AR, Saurber WM, Gaussa EF, Pilchak AL (2014) The effect of forging variables on the supersolvus heat-Treatment response of powder-metallurgy nickel-base superalloys. Metall Mater Trans 45:6231–6251. https://doi.org/10.1007/s11661-014-2572-y.

Dynamic and Post-dynamic Recrystallization During Supersolvus Forging of the New Nickel-Based Superalloy—VDM Alloy 780 Juhi Sharma, Masood Hafez Haghighat, Bodo Gehrmann, Charbel Moussa, and Nathalie Bozzolo

Abstract

The need of developing new high temperature materials has increased significantly in the last decades owing to the demand of higher engine operating temperatures. This requires improved microstructural stability of the polycrystalline nickel-based superalloys used for turbine disks. The microstructure of VDM Alloy 780 consists of c′ strengthening precipitates in addition to the needle/plate-shaped particles (of d and/or η phase) to pin the grain boundaries. The present article aims at discussing the recrystallization behavior of the new VDM Alloy 780 in the supersolvus domain. Both dynamic and post-dynamic microstructural evolutions are reported. The dynamic recrystallization (DRX) kinetics was found to be rather sluggish. For a plastic strain of 1.3 at 1050 °C applied at a strain rate of 0.01 s−1, the microstructure of VDM Alloy 780 is only 50% recrystallized. The DRX grain sizes are quite close for the two applied strain rates —0.01 and 0.1 s−1. Despite slow DRX kinetics, the fast post-dynamic evolutions allowed to achieve fully recrystallized microstructures with a grain size of 26 lm within 5 min of post-deformation holding at 1050 °C. The post-dynamically recrystallized grain sizes were predominantly temperature dependent and were not sensitive to strains and strain rates within the applied range. Deformation followed by 5 min holding at temperatures below 1050 °C but still in the single-phase domain could eventually generate finer grain sizes (50 nm). In the René 65 material, the distribution is narrower and centered at less than half of the grain size with a lower density of long and intense bands. There are therefore more short slip bands of low intensity in the René 65 material. This is consistent with the reduction of plastic localization induced by the presence of γ  precipitates located on twin boundaries as shown in Fig. 6 and the higher density of small fatigue cracks reported in Sect. 3.1.

Processing Pathways for Improved Fatigue Resistance: Dual-Phase Grain Boundary Engineering Following the results presented above, long coherent annealing twins favorably oriented for parallel slip are the preferred sites for crack nucleation induced by plastic localization and accumulation [25]. In situ tensile loading with HR-DIC measurements provides evidence of the plastic localization along specific twin boundaries in a parallel slip configuration. It is also observed that the primary γ  precipitates can disrupt the plastic localization along these twin boundaries.

478

M. A. Charpagne et al.

Fig. 9 Two-step forging process applied to the René 65 material

Fig. 8 Number density of slip bands in a parallel slip configuration, according to their average shearing and length normalized by the average grain size in each material for a René 88DT, b René 65

Thermomechanical processing pathways for improving mechanical properties by disrupting plastic localization involve breaking either the twin coherency and/or the apparent length of coherent twins. While the twin coherency will likely be lost at high strains, the apparent twin boundary length can be reduced by either reducing the overall grain size, or increasing the number density of precipitates located on twin boundaries in René 65. As industrial components exhibit microstructure gradients due to different local deformation conditions, global processing conditions (temperature and strain rate) that yield optimized microstructures across a variety of strains need to be determined. Three-dimensional characterization of the René 65 twin structure has shown that a high density of primary γ  precipitates on the twin boundaries can be reached after critical recrystallization [8]. This phenomenon happens during annealing in the sub-solvus domain following deformation at small strains, e.g., 1065 ◦ C for one hour [8], when start-

ing from a microstructure of low stored strain energy. While the mechanisms responsible for the formation of twin boundaries at the location of precipitates are not fully understood, it requires the passing of precipitates by the migrating recrystallization front, implying grain sizes that exceed the average spacing between precipitates (12 µm). However, such large grains would dramatically reduce the fatigue life. One can thus think of a two-step forging process designed to first increase the density of precipitates on twin boundaries by triggering critical recrystallization, and then a second forging step aimed at decorating the boundaries of the first grains with smaller recrystallized ones, for an overall reduced grain size and disrupted slip pathways. A processing path to achieve this is schematically presented in Fig. 9. Cylindrical compression specimens measuring 5 mm in height and diameter were machined from the heat-treated René 65 material. They were deformed at high temperature using a MTS frame equipped with a resistive furnace and a thermocouple placed onto the specimen surface. Specimens were air cooled after deformation. Microstructures were characterized by EBSD collected in the center of the samples under the voltage conditions described in Section “Strain Localization During Monotonic Loading”, using a step size of 100 nm. Corresponding BSE images were also collected for identification of primary γ  precipitates and aligned using the method described in [10]. Keeping in mind that triggering critical recrystallization is the goal, several temperatures, strain and strain rates were tested for the first forging pass, followed by annealing for one hour at 1065 ◦ C, the standard solution heat-treatment conditions for the René 65 material. The average grain sizes as well as the number density of primary precipitates are reported in Fig. 10. All deformation conditions yielded relatively fine grains (