Lasers in the Conservation of Artworks: Proceedings of the International Conference Lacona VII, Madrid, Spain, 17 - 21 September 2007 [1 ed.] 0415475961, 9780415475969

Citation preview

LASERS IN THE CONSERVATION OF ARTWORKS

PROCEEDINGS OF THE INTERNATIONAL CONFERENCE LACONA VII, MADRID, SPAIN, 17–21 SEPTEMBER 2007

Lasers in the Conservation of Artworks Editors Marta Castillejo Instituto de Química Física Rocasolano, CSIC, Madrid, Spain

Pablo Moreno Laser Facility, University of Salamanca, Salamanca, Spain

Mohamed Oujja Instituto de Química Física Rocasolano, CSIC, Madrid, Spain

Roxana Radvan National Institute of Research and Development for Optoelectronics, Bucharest, Rumania

Javier Ruiz Department of Applied Physics I, University of Málaga, Málaga, Spain

CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business © 2008 Taylor & Francis Group, London, UK Typeset by Charon Tec Ltd (A Macmillan Company), Chennai, India Printed and bound in Great Britain by Cromwell Press Ltd, Towbridge, Wiltshire All rights reserved. No part of this publication or the information contained herein may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying, recording or otherwise, without written prior permission from the publishers. Although all care is taken to ensure integrity and the quality of this publication and the information herein, no responsibility is assumed by the publishers nor the author for any damage to the property or persons as a result of operation or use of this publication and/or the information contained herein. Published by: CRC Press/Balkema P.O. Box 447, 2300 AK Leiden, The Netherlands e-mail: [email protected] www.crcpress.com – www.taylorandfrancis.co.uk – www.balkema.nl ISBN 13: 978-0-415-47596-9

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Table of Contents

Preface

XI

Committees

XIII

Sponsors

XV

Photonic restoration of marine artifacts and vessels of New Spain J.F. Asmus

1

Innovative Approaches in Laser Cleaning and Analysis Towards the restoration of darkened red lead containing mural paintings: A preliminary study of the β-PbO2 to Pb3 O4 reversion by laser irradiation S. Aze, P. Delaporte, J.M. Vallet, V. Detalle, O. Grauby & A. Baronnet

11

Prospective and applications of two-photon fluorescence in archaeology and art conservation D. Artigas, L. Serrado, I.G. Cormack, S. Psilodimitrakopoulos & P. Loza-Alvarez

15

Fast spectral optical coherence tomography for monitoring of varnish ablation process M. Góra, P. Targowski, A. Kowalczyk, J. Marczak & A. Rycyk

23

Study of laccaic acid and other natural anthraquinone dyes by Surface-Enhanced Raman Scattering spectroscopy M.V. Cañamares & M. Leona Potential of THz-Time Domain Spectroscopy in object inspection for restoration M. Panzner, Th. Grosse, S. Liese, U. Klotzbach, E. Beyer, M. Theuer, W. Köhler & H. Leitner Femtosecond laser cleaning of paintings: Modifications of tempera paints by femtosecond laser irradiation S. Gaspard, M. Oujja, M. Castillejo, P. Moreno, C. Méndez, A. García & C. Domingo Cleaning of paint with high repetition rate laser: Scanning the laser beam A.V. Rode, D. Freeman, N.R. Madsen, K.G.H. Baldwin, A. Wain, O. Uteza & P. Delaporte Removal of unwanted material from surfaces of artistic value by means of Nd:YAG laser in combination with Cold Atmospheric-Pressure Plasma C. Pflugfelder, N. Mainusch, W. Viöl & J. Ihlemann

29 35

41 49

55

Analytical Techniques Optical coherence tomography for structural imaging of artworks P. Targowski, M. Góra, T. Bajraszewski, M. Szkulmowski, M. Wojtkowski, A. Kowalczyk, B. Rouba, L. Tymi´nska-Widmer & M. Iwanicka Atmospheric Pressure Laser Desorption Mass Spectrometry based methods for the study of traditional painting materials M.P. Licciardello, R. D’Agata, G. Grasso, S. Simone & G. Spoto

V

61

67

Study of chromophores of Islamic glasses from Al-Andalus (Murcia, Spain) N. Carmona, M. García-Heras, M.A. Villegas, P. Jiménez & J. Navarro

73

Polychromed sculptures of Mercadante and Millán analysed by XRF non-destructive technique A. Križnar, M.A. Respaldiza, M.V. Muñoz, F. de la Paz & M. Vega

79

Litharge and massicot: Thermal decomposition synthetic route for basic lead(II) carbonate and Raman spectroscopy analysis M. San Andrés, J.M. De la Roja, S.D. Dornheim & V.G. Baonza Contamination identification on historical paper by means of the NIR spectroscopic technique M. Sawczak & A. Kaminska

89 95

Three dimensional survey of paint layer profile measurements E. Pampaloni, R. Fontana, M.C. Gambino, M. Mastroianni, L. Pezzati, P. Carcagnì, R. Piccolo, P. Pingi, R. Bellucci & A. Casaccia

101

Use of LA-ICP-MS technique with SEM-EDS analysis in the study of finishing layers L. Rampazzi, B. Rizzo, C. Colombo, C. Conti & M. Realini

109

Compositional depth profiles of gilded wood polychromes by means of LIBS A.J. López, A. Ramil, M.P. Mateo, C. Álvarez & A. Yáñez

115

Classification of archaeological ceramics by means of Laser Induced Breakdown Spectroscopy (LIBS) and Artificial Neural Networks A. Ramil, A.J. López, M.P. Mateo & A. Yáñez

121

Laser ablation- and LIBS-ranging by webcam and image processing during laser cleaning M. Lentjes, J. Hildenhagen & K. Dickmann

127

LIBS analysis of metal artefacts from Sucevita Monastery, Romania M. Oujja, M. Castillejo, W. Maracineanu, M. Simileanu, R. Radvan, V. Zafiropulos & D. Ferro

133

Comparative study of historic stained glass by LIBS and SEM/EDX K. Szelagowska, M. Szymonski, F. Krok, M. Walczak, P. Karaszkiewicz & J.S. Prauzner-Bechcicki

141

Portable Laser Systems for Remote and On-Site Applications Scanning hyperspectral lidar fluorosensor for fresco diagnostics in laboratory and field campaigns F. Colao, L. Caneve, R. Fantoni, L. Fiorani & A. Palucci

149

A lidar experiment for the characterization of photoautotrophic and heterotrophic biodeteriogens by means of remote sensed autofluorescence spectra V. Raimondi, L. Palombi, D. Lognoli, G. Cecchi & I. Gomoiu

157

Design and development of a new high speed performance fluorescence imaging lidar for the diagnostics of indoor and outdoor cultural heritage V. Raimondi, L. Palombi, D. Lognoli, G. Cecchi & L. Masotti

163

Remote fluorescence lidar imaging of monuments: The Coliseum and the Lateran baptistery in Rome J. Hällström, K. Barup, V. Raimondi, L. Palombi, D. Lognoli, G. Cecchi, R. Grönlund, A. Johansson, S. Svanberg & C. Conti Portable spectroscopic analysis of nitrates affecting to cultural heritage materials M. Maguregui, I. Martinez-Arkarazo, M. Angulo, K. Castro, L.A. Fernández & J.M. Madariaga A study of laser cleaning parameters using a portable system on a gargoyle of the Torres de Serranos in Valencia, Spain B. Sáiz & M. Iglesias

VI

169

177

183

Laser Cleaning of Monuments and Sculptures Castle of Quart, Aosta Valley: Laser uncovering of medieval wall paintings S. Siano, L. Appolonia, A. Piccirillo & A. Brunetto

191

Colour changes in Galician granitic stones induced by UV Nd:YAG laser irradiation A. Ramil, A.J. López, M.P. Mateo, C. Álvarez & A. Yáñez

199

Arch-collegiate church in Tum: Laser renovation of priceless architectural decorations A. Koss, J. Marczak & M. Strzelec

203

Laser cleaning of the Nickerson Mansion: The first building in the US entirely cleaned using laser ablation A. Dajnowski Laser cleaning of a set of 18th century ivory statues of Twelve Apostles A. Koss, D. Dre´scik, J. Marczak, R. Ostrowski, A. Rycyk & M. Strzelec

209 215

Laser Cleaning of Paintings and Polychromes Investigating the use of the Nd:YAG laser to clean ancient Egyptian polychrome artefacts C. Korenberg, M. Smirniou & K. Birkholzer

221

Laser cleaning as a more culturally appropriate treatment option for Native American pictographs and pictograms M. Abraham & C. Dean

227

The Arca Scaligera of Cansignorio della Scala by Bonino da Campione: Cleaning of the polychrome and gilded decorations V. Fassina, G. Gaudini, S. Siano & R. Cavaletti

231

Assessment of laser cleaning on a polychrome Islamic ceramic B. Sáiz, E. Aura, M.T. Domenech & A. Domenech

237

Laser cleaning of stucco’s fragments from an early middle age bas-relief A. Sansonetti, C. Colombo, M. Realini, M. Palazzo & M. De Marchi

243

Soot removal from artificial fresco models by KrF excimer laser J. Hildenhagen, K. Dickmann, W. Maracineanu & R. Radvan

249

The interaction of laser radiation at 2.94 µm with azurite and malachite pigments M. Camaiti, M. Matteini, A. Sansonetti, J. Striová, E. Castellucci, A. Andreotti, M.P. Colombini, A. deCruz & R. Palmer

253

Conservation of medieval polychromed wooden sculpture of Madonna and Child K. Chmielewski, A. Koss, M. Mazur, J. Marczak & M. Strzelec

259

Advanced laser renovation of old paintings, paper, parchment and metal objects J. Marczak, M. Strzelec, R. Ostrowski, A. Rycyk, A. Sarzy´nski, W. Skrzeczanowski, A. Koss, R. Szambelan, R. Salimbeni, S. Siano, J. Kolar, M. Strlic, Z. Márton, I. Sánta, I. Kisapáti, Z. Gugolya, Z. Kántor, S. Barcikowski, P. Engel, M. Pires, J. Guedes, A. Hipólito, S. Santos, A.S. Dement’ev, V. Švedas, E. Murauskas, N. Slavinskis, K. Jasiunas & M. Trtica

263

Comparative study of laser varnish removal from historical paintings Z. Márton, I. Sánta, É. Galambos, C. Dobai, Á. Dics˝o & Z. Kántor

271

VII

Laser Cleaning of Metal Objects Laser interactions with copper, copper alloys and their corrosion products used in outdoor sculpture in the United Kingdom M. Froidevaux, P. Platt, M. Cooper & K. Watkins

277

Investigating the laser cleaning of archaeological copper alloys using different laser systems C. Korenberg, A.M. Baldwin & P. Pouli

285

Investigating and optimising the laser cleaning of corroded iron C. Korenberg, A.M. Baldwin & P. Pouli

291

Nd:YAG laser cleaning of heavily corroded archaeological iron objects and evaluation of its effects J. Chamón, J. Barrio, E. Catalán, M. Arroyo & A.I. Pardo

297

Laser as a cleaning tool for the treatment of large-scale bronze monuments A. Dajnowski

303

Experimental study on the use of laser cleaning of silver plating layers in Roman coins A.A. Serafetinides, E. Drakaki, I. Zergioti, C. Vlachou-Mogire & N. Boukos

309

Morphological and colorimetric changes induced by UV laser radiation on metal leaves S. Acquaviva, E. D’Anna, M.L. De Giorgi, A. Della Patria & L. Pezzati

317

Application of Ion Beam Analysis (IBA) techniques for the assessment of laser cleaning on gilded copper M. Barrera, C. Escudero, M.D. Ynsa & A. Climent-Font Laser cleaning: Influence of laser beam characteristics M. Pires, C. Curran, W. Perrie & K. Watkins

323 329

Laser Cleaning of Documents and Textiles Study of laser cleaning of ancient fabric with femtosecond pulses C. Escudero, M.A. Martínez, P. Moreno, A. García, C. Méndez, C. Prieto & A. Sanz

337

Monitoring of the laser cleaning process of artificially soiled paper S. Pentzien, A. Conradi, J. Krüger & R. Wurster

345

Systematic case study on common cleaning problems on paper and parchment by using Nd:YAG laser (ω, 2ω, 3ω) J. Hildenhagen, M. Lentjes, K. Dickmann & B. Geller Use of laser and optical diagnostic techniques on paper: The Pomelnic from Sucevita Monastery (Romania) M. Strlic, J. Kolar, G. Pajagic Bregar, V. Ljubic, R. Radvan, M. Simileanu, W. Maracineanu, J. Hildenhagen, M. Castillejo, M. Oujja, W. Kautek, V. Zafiropulos, T. Sinigalia, O. Boldura & N. Melniciuc Laser cleaning of 19th century papers and manuscripts assisted by digital image processing G.M. Bilmes, C.M. Freisztav, N. Cap, H. Rabal & A. Orsetti Laser reduction of stamps from paper to avoid migration to the recto side: Case study based on illustrations from Jan Heesters M. Lentjes, K. Dickmann & P. van Dalen Laser cleaning and multi-method diagnostics of textile pieces of art W. Kautek, M. Oujja, M. Castillejo, J. Hildenhagen, V. Ljubic, M. Simileanu, W. Maracineanu, R. Radvan, V. Zafiropulos, N. Melniciuc, G. Pajagic Bregar & M. Strlic

VIII

353

357

361

367 371

Study of the effects of laser cleaning on historic fabrics: Review and results eight years after applications A. Martínez & C. Escudero

375

Structural Diagnosis and Monitoring Multifunctional encoding system for assessment of movable cultural heritage V. Tornari, E. Bernikola, W. Osten, R.M. Groves, M. George, T. Cedric, G.M. Hustinx, E. Kouloumpi, A. Moutsatsou, M. Doulgeridis, S. Hackney & T. Green

381

Digital preservation, documentation and analysis of heritage with active and passive sensors F. Remondino

387

Monitoring of changes in the surface movement of model panel paintings following fluctuations in relative humidity: Preliminary results using Digital Holographic Speckle Pattern Interferometry E. Bernikola, V. Tornari, A. Nevin & E. Kouloumpi

393

Integrated digital speckle based techniques for artworks monitoring D. Ambrosini, D. Paoletti & G. Galli

399

Laser-based structural diagnosis: A museum’s point of view E. Kouloumpi, A.P. Moutsatsou, M. Trompeta, J. Olafsdottir, C. Tsaroucha, A.V. Terlixi, R.M. Groves, M. Georges, G.M. Hustinx & V. Tornari

407

High-resolution 3D laser digitisation of the Maiano terracotta roundels for documentation and condition monitoring K. Hallett, Z. Roberts, S. Julien-Lees & A. Geary

413

An SLDV/GPR/IR-T integrated approach for structural and frescoes investigation in the medieval monasteries of Moldavia E. Esposito, A. Agnani, M. Feligiotti, A. del Conte & S. Goncalves Tavares

419

Development of an impact assessment procedure for artwork using shearography as a measurement tool R.M. Groves, W. Osten, S. Hackney, E. Kouloumpi & V. Tornari

427

Imaging and Documentation Ultra high-resolution 3D laser color imaging of paintings: The Mona Lisa by Leonardo da Vinci F. Blais, J. Taylor, L. Cournoyer, M. Picard, L. Borgeat, G. Godin, J.A. Beraldin, M. Rioux & C. Lahanier

435

Characterization and virtual reconstruction of polychromed alabaster sculptures A. Sarmiento, K. Castro, M. Angulo, I. Martinez-Arkarazo, L.A. Fernández, J.M. Madariaga, J.M. Gonzalez-Cembellín & M. Urrutikoetxea Barrutia

441

ITR: A laser rangefinder for cultural heritage conservation applications with multi-sensor data integration capabilities R. Ricci, M. Ferri De Collibus, G. Fornetti, M. Francucci, M. Guarneri & E. Paglia

447

Multispectral and multi-modal imaging data processing for the identification of painting materials A. Pelagotti, A. Del Mastio & V. Cappellini

453

Recovering colour and volume from relics in restoration tasks J. Finat, J.I. San José, J.J. Fernández, J.D. Pérez-Moneo, J. Martínez, F. Gutiérrez-Baños, L. Giuntini & F.M. Morillo

IX

459

Multi IR reflectography R. Fontana, M. Greco, M. Mastroianni, M. Materazzi, E. Pampaloni, L. Pezzati & P. Carcagnì

465

Miscellaneous Imaging and mass spectrometry of microparticles generated during surface decontamination of an ancient parchment sample by laser radiation R. Wurster, S. Pentzien & W. Kautek

473

Laser for removing remains of carbonated matrices from Pleistocene fossils L. López-Polín, A. Ollé, J. Chamón & J. Barrio

477

Environmental optical sol-gel sensors for preventive conservation of cultural heritage N. Carmona, E. Herrero, M.A. Villegas & J. Llopis

483

Author index

489

X

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Preface

Materials and technologies used in the creative artistic process or in the manufacture of historically relevant objects evolve with time reflecting the knowledge and uses of the society at the historic moment in which these works were created. Scientific research on Cultural Heritage, both for the study of its material aspects and for designing restoration and conservation strategies, faces a diversity of challenges due to the complexity and intrinsic value of the substrates and objects and their past history of exposure to degrading agents and to the passage of time. Fast developing laser systems and advanced optical techniques offer new solutions for Conservation scientists and provide answers to the challenges of Conservation Science. In this field, the international conference series LACONA, Lasers in the Conservation of Artworks, has been established as a reference meeting point for conservators, end users and scientists. LACONA series started in 1995 in Heraklion, (Greece) being followed every two years by editions in Liverpool (Great Britain), Florence (Italy), Paris (France), Osnabrük (Germany) and Vienna (Austria). The last edition, the 7th International Conference on Lasers in the Conservation of Artworks, LACONA VII, was celebrated in Madrid (Spain), 17–21 September, 2007 at the headquarters of Consejo Superior de Investigaciones Científicas (CSIC) the main research institution in Spain. This Volume of Proceedings of LACONA VII presents a selection of contributions on both emerging and well established applications of laser systems and techniques to real Conservation problems. Innovative Approaches in Laser Cleaning and Analysis are presented together with Analytical Techniques and Portable Laser Systems for Remote and On-Site Applications. A substantial number of contributions to this Volume deal with Laser Cleaning of monuments and sculptures, paintings and polychromes, metal objects and documents and textiles. Also included are developments on Structural Diagnosis, Monitoring, Imaging and Documentation of artworks. I would like to express my gratitude to the Co-Editors of this Volume, Pablo Moreno, Mohamed Oujja, Roxana Radvan and Javier Ruiz, for their enthusiasm and efficient work. Also, the organization of LACONA VII and the elaboration of this Volume of Proceedings has greatly benefited from the cooperative spirit of the Permanent Scientific Committee. Finally, I want to thank the sponsoring of LACONA VII by CSIC and its Thematic Network on Cultural Heritage, by the Instituto de Química Física Rocasolano, by the Spanish Ministry of Science and Education and by Consejería de Educación de la Comunidad de Madrid. Besides the official Institutions, I have very much appreciated the support of a numerous group of private sponsors and the competent technical assistance of Fase 20 Congresos.

Marta Castillejo, LACONA VII Chair, Madrid, May 2008.

XI

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Committees

PERMANENT SCIENTIFIC COMMITTEE Margaret Abraham, Los Angeles County Museum of Art, Los Angeles, USA John Asmus IPAPS, University of California, San Diego, USA Gerd v. Bally, Laboratory of Biophysics, University of Münster, Germany Giorgio Bonsanti, University of Florence and Centro Europeo di Ricerche sul Restauro (CERR) di Siena, Italy Marta Castillejo, Instituto de Química Física Rocasolano, CSIC, Madrid, Spain Martin Cooper, The Conservation Centre, National Museums Liverpool, United Kingdom Klaus Dickmann, Laserzentrum FH Münster, Germany Costas Fotakis, Foundation for Research and Technology Hellas, IESL, Heraklion, Crete, Greece Wolfgang Kautek, University of Vienna, Department of Physical Chemistry, Austria Eberhard König, Freie Universität Berlin, Germany Mauro Matteini, Istituto per la Conservazione e Valorizzazione dei Beni Culturali, CNR, Florence, Italy Johann Nimmrichter, Bundesdenkmalamt, Austrian Federal Office for the Care of Monuments, Centre of Art Conservation, Vienna, Austria Roxana Radvan, National Institute of Research and Development for Optoelectronics, Bucharest, Romania Renzo Salimbeni, Istituto di Fisica Applicata Nello Carrara CNR, Florence, Italy Manfred Schreiner, Academy of Fine Arts, Vienna, Austria Véronique Vergès-Belmin, Laboratoire de Recherche des Monuments Historiques, Champs-sur-Marne, France Kenneth Watkins, Department of Engineering, University of Liverpool, United Kingdom Vassilis Zafiropulos, Technological Educational Institute of Crete & Center for Technological Research – Crete, Sitia, Crete, Greece

ORGANIZING COMMITTEE Chair: Marta Castillejo, Instituto de Química Física Rocasolano, CSIC, Madrid Mónica Álvarez de Buergo, Instituto de Geología Económica, CSIC, Universidad Complutense de Madrid, Madrid Rocio Bruquetas, Spanish Group of IIC, Instituto del Patrimonio Histórico Español, Madrid Noemí Carmona, Centro Nacional de Investigaciones Metalúrgicas, CSIC, Madrid Manuela Casado, Instituto de Química Física Rocasolano, CSIC, Madrid Solenne Gaspard, Instituto de Química Física Rocasolano, CSIC, Madrid Fernando Guerra-Librero, CORESAL, Madrid Ana Laborde, Spanish Group of IIC, Instituto del Patrimonio Histórico Español, Madrid

XIII

Margarita Martín, Instituto de Química Física Rocasolano, CSIC, Madrid Pablo Moreno, Laser Facility, University of Salamanca, Salamanca Mohamed Oujja, Instituto de Química Física Rocasolano, CSIC, Madrid Javier Ruiz, Department of Applied Physics I, University of Málaga, Málaga Malgorzata Walczak, Instituto de Química Física Rocasolano, CSIC, Madrid

XIV

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Sponsors

OFFICIAL SPONSORS Consejo Superior de Investigaciones Científicas (CSIC)

CSIC Thematic Network on Cultural Heritage

Instituto de Química Física Rocasolano, CSIC

Ministerio de Educación y Ciencia

Consejería de Educación, Comunidad de Madrid

XV

PRIVATE SPONSORS Faro Technologies Inc., www.faro.com nub3d S.L., www.nub3d.com El.En Group, www.elengroup.com Lasertech Ibérica S.L., www.lasertechib.com CTS, www.ctseurope.com Erich Pummer Atelier, www.atelier-pummer.at Innova Sci., www.innovasci.com Renishaw plc, www.renishaw.com Lasing, www.lasing.com Linsinger, www.linsinger.at Iberlaser, www.iberlaser.com Clar, www.clar.es International Institute for Conservation of Historic and Artistic Works – Grupo Español GE-IIC, http://ge-iic.com/ Arespa, www.arespaph.com Geocisa, www.geocisa.es Artemon, www.artemon.es

XVI

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Photonic restoration of marine artifacts and vessels of New Spain J.F. Asmus University of California, San Diego, La Jolla, CA, USA

ABSTRACT: On September 28, 1524 Juan Rodríguez Cabrillo moored his vessels, San Salvador, Victoria, and San Miguel, at what is now Ballast Point at the entrance to San Diego Bay. The crew spent six days at this location performing naval maintenance and studying the customs of the natives, the marine life, and marine fossils embedded in the adjoining cliffs. When the Scripps Institution of Oceanography (SIO) was founded at the dawn of the 20th Century, Ballast Point was selected as the port for the research vessels. In contrast to Cabrillo’s crew, we had the opportunity to investigate in-situ underwater radiation ablation divestment to remove marine fouling from vessel surfaces and reactivate antifouling coatings. Cabrillo’s tradition of marine animal study and environmental science were continued and enhanced through the radiation cleaning of research-aquarium windows, fossils, native and Spanish artifacts, as well as the laser branding of whales.

1

INTRODUCTION

a Venetian holographic feasibility study was proposed. The objective was to demonstrate that even large (meters) Venetian statues could be recorded in holograms, in situ. Thus, a high-quality visual archival library would be produced of sculptural patrimony before further deterioration took place. In November 1971 Ente Nazionale Idrocaburi (ENI) funded the feasibility study (winter 1971–72) and more than 50 large-object large-format holograms of Venetian artworks were recorded and placed on display in the Academia Museum. In the course of the three-month investigation two unexpected discoveries were made. First, it was found that double-exposure holographic interferograms are able to reveal hidden defects in artworks and assess the success of subsequent conservation repairs. Second, a recent laser publication (Ready 1971) inspired the team to the discovery that pulsed-laser ablation could be adapted to advance art conservation technology in the attainment of very high-quality surface divestment. In a very few weeks it was found that self-limiting divestment of sulfation from stone, minerals from pottery, glue from canvas, corrosion from metals, fungi from leather, and soils from textiles was feasible (Asmus 1972). The conservation journal, “Studies in Conservation” immediately accepted a manuscript describing the holography. However, a manuscript describing the divestment results was rejected with the terse observation: “Laser cleaning is too hypothetical to be taken seriously” (even though 70 items had been laser cleaned at that point). Consequently, the SIO team returned to San Diego and addressed the divestment issues of foremost concern at the home facility, the

The introduction of laser technology into the field of art conservation began in January 1972 with the collaboration of three oceanographic institutions. For some years the Scripps Institution of Oceanography (SIO) of the University of California had been cooperating with a CNR Laboratory (Laboratorio per lo Studio della Dinamica delle Grandi Masse: Palazzo Papadopoli, Venice, Italy) and the Istituto di Geodesia e Geofisica of the University degli Studi di Trieste, Italy, in the computer modelling of the tides and currents of the Adriatic Sea and the Venetian Lagoon. The objective was guidance in the design of lagoon closure gates with the goal of alleviating the severity of “acqua alta” events. As the interdisciplinary team of theoretical and experimental scientists witnessed the progressive and accelerating deterioration of the city and its patrimony (Lorenzetti 1961), it became increasingly obvious that more immediate preservation measures were required due to the vigour and severity of the environmental assault. This observation was especially germane in light of the protracted time line for the approval, design, construction, and installation of the floodgates. The only proposal that was brought forward from the emergency deliberations was to record the overall appearance and geometry of the artistic patrimony with the highest possible fidelity. As pulsed ruby-laser in-situ 3D diffraction-limited holography had been demonstrated in the laboratory (and as enthusiasm for holography was running high in Italy because a resident of Tuscany, Dennis Gabor, was expected to win the Nobel prize for this discovery),

1

Figure 1. Depiction of Cabrillo’s landing at Ballast Point. Figure 2. Caption monument at Point Loma (left) and depiction of Kumeyaay native resident of the region (right).

Marine Physical Laboratory (MPL), at “Ballast Point”, Point Loma, San Diego. Whereas the radiation divestment studies reported in the following were performed during the interval 1975–78 at MPL, “Studies in Conservation” continued to reject the submissions and only fragments of the work were published in physics, chemistry, engineering, metallurgical, and oceanographic journals. A summary of the radiation divestment work performed at the Ballast Point marine facility is presented in the following sections. 2

HISTORICAL BACKGROUND Figure 3. Entrance channel to San Diego Harbour showing the MPL naval facility of SIO at Ballast Point in the foreground.

The crossbowman Juan Rodríguez Cabrillo arrived in the Americas in 1520 and joined Hernán Cortés in the conquest of the Aztec capital of Tenochtitlan. In gratitude for Cabrillo’s service the Viceroy of New Spain gave him command of three galleons and a commission to discover a route to Asia and the Spice Islands. On September 28, 1542 this flotilla entered a harbour that Cabrillo describes as “un puerto cerrado y muy bueno”. The vessels, San Salvador, Victoria, and San Miguel, were moored at a natural spit near the entrance to that harbour (known, today, as “Ballast Point”, within San Diego Bay). Before continuing to sail up the coast of North America, Cabrillo’s crew spent six days at this location (Fig. 1) performing naval maintenance and studying the customs of the local Kumeyaay natives (Fig. 2), the marine life, and the marine fossils embedded in the adjoining cliffs. When the Scripps Institution of Oceanography was founded at the dawn of the 20th Century, Ballast Point was selected as the port for its research vessels, bearing in mind the same considerations that lead Cabrillo to that site. Today, as in the 16th Century, there are no dry-dock facilities at Ballast Point, and routine maintenance of ship hulls is performed in-situ and/or underwater. In contradistinction to Cabrillo’s crew, SIO personnel had the opportunity to investigate underwater radiation ablation divestment to remove marine fouling such as barnacles and algae from vessel surfaces and marine research facilities as well as historic artifacts excavated from the adjoining terrain

(Fig. 3). It was found that ruby laser and Nd:YAG laser radiation as well as xenon flashlamp light have comparable cleaning effectiveness whether in air or underwater. Whereas seawater attenuates the laserbeam intensity reaching the surface being cleaned, a “witness plate” effect was found to enhance the strength of the interaction.

3

RADIATION ABLATIVE DIVESTMENT OF ARCHAEOLOGICAL ARTIFACTS AT BALLAST POINT

The excavations and dredging (60–70 years ago) that prepared Ballast Point for modern naval port activities uncovered prehistoric and native artifacts as well as items of New Spain dating from the 16th Century. When MPL was established neither museum nor conservation program existed, and the excavated artifacts were simply stored. The encouraging laser-divestment tests in Venice with the SIO lasers suggested pursuing a continuation of those conservation studies with those marine artifacts in storage. Figure 4 shows the results of laser divestment of a silver Spanish coin from the era of Cabrillo or one of his followers. Those areas where the marine fouling was removed with the ruby

2

Figure 4. Spanish silver coin covered with marine deposits (left) and laser cleaned (right). The bright areas on the cleaned coin resulted from CO2 TEA laser and the areas with a light grey patina were cleaned with a Q-switched ruby laser.

Figure 6. Fossil of a whale vertebrate bone showing original condition (right) and laser-cleaned area (left).

Figure 5. Laser removal (top) of a protective coating on wood.

Figure 7. Laser-cleaned (top) and dollar fossil.

holographic laser retained a pale grey patina, whereas the radiation from a CO2 TEA laser yielded a bright metallic finish. Wooden remains were also recovered from Ballast Point. Some may be the remains of ship repair work conducted by the Spanish explorers. When such materials were discovered during the construction of the port facility they were simply coated with polyurethane and placed in storage. Years later it was deemed desirable to remove the coating in order to study and display these pieces. Several laser types were tested in order to ablate the thick coating. Eventually it was learned that high-energy (50 J) normal-mode pulses from a Nd:glass laser soften the polyurethane and aerodynamic forces from the Laser-Induced Combustion (LIC) wave eject the coating residue in large molten droplets. It was determined that the laser ejection of condensed material is more efficient than radiation ablation (Fig. 5). Deeper excavations at Ballast Point site yielded marine fossils from earlier epochs. Figure 6 reveals a vertebrae fossil from a whale. Figure 7 is a sand dollar fossil, and Figure 8 shows the fossil of a dinosaur bone.

Figure 8. Fossil of a dinosaur bone. A laser cleaned the left side and upper portion.

4

LASER RESTORATION OF SIO SHIPS AND NAVAL HARDWARE

The apparent success of the exploratory laser cleaning tests on the marine archaeological specimens (Section 3) led to an expansion of the probative effort to

3

Figure 9. Barnacle removal from the steel hull of a ship (1000 ea. Ruby laser pulses at pulse energy of 10 J).

Figure 11. Laser removal of rust (left) from a portion of a marine engine-control linkage (steel).

Figure 10. Cylinder heads of a marine diesel engine. Carbon buildup is shown on left. Laser removal of carbon on right.

Figure 12. Deep ocean SIO submersible research vehicle.

include surface preparation of relevance to ship maintenance. Many of the conventional surface-cleaning methods employed in shipyards utilize or produce toxic substances that harm the environment and/or jeopardize worker health. Examples are the organic solvents commonly used to remove paints and greases. High-pressure blasting (abrasive, air, and/or water) is also widely employed in shipyard maintenance and frequently presents problems of debris containment. Figure 9 shows an example of laser barnacle removal from a ship hull. Figure 10 displays laser carbon removal for marine engine maintenance, Figure 11 is an engine-control linkage where corrosion removal by laser was demonstrated without resorting to the use of acids, toxic organic solvents, or blasting with abrasive media. Whereas these results have demonstrated the technical feasibility of laser surface preparation for ship maintenance, cost effectiveness, and utility are not assured. Particulate and vapour debris control may

be easier with laser divestment than with highpressure blasting. However, this supposition has yet to be proven. Further, at the time of these probative investigations (1975) laser performance (viz., power, efficiency, and reliability) had not reached levels commensurate with practical surface-coverage rates. In addition, even today, it is questionable as to whether the physical and optical durability of fibre optic beam delivery systems are able to meet the logistical requirements demanded by these applications. The submersibles in the SIO fleet (Fig. 12) present a number of extremely challenging maintenance problems. Because these vehicles dive to enormous depths, the most difficult of these issues is the adhesion of rubber insulating sheets on the exterior hull. Figure 13 shows high-power flashlamp irradiation of a typical rubber sheet in order to clean and etch the surface prior to the application of the adhesive and subsequent attachment to the submersible’s hull.

4

Figure 15. Pointing to area of coupon divested of marine fouling, underwater.

Figure 13. Xenon flashlamp system in use to clean and etch a rubber insulation panel from a deep-ocean submarine.

Figure 16. 1 MHz quartz ultrasonic transducer used to measure the stress wave amplitude under water and in air.

Figure 14. Water tank (white box) used for underwater flashlamp irradiation. The waterproof flashlamp/reflector housing is attached to the end of the immersion pole is lying on the tank. An algae-covered test coupon is leaning against the tank.

The logistics of marine-vessel maintenance could be further improved by performing the radiation ablation divestment and/or cleaning in situ (underwater). The initial submersible flashlamp test apparatus is shown in Figure 14. Subsequent underwater removal of marine fouling (algae) and reactivation of a spent organic copper antifouling coating was performed in bay water under an MPL pier (Fig. 15). Figures 16 and 17 show an ultrasonic transducer and its signals for laser irradiation both underwater and in air: with

Figure 17. Oscilloscope traces of laser-generated stress waves at a water interface (top) and when the target is in air (lower).

5

Figure 18. Photographs of freely expanding ablation blow-off plasma in air (top) and confined in underwater surface irradiation (bottom).

Figure 19. Aquarium observation window covered with a thin buildup of marine algae. An underwater xenon flash dislodged the algae on the left. A flash on the right from the airside went through the glass and dislodged the algae from behind.

pure water, a UV-grade fused silica window, and KrF excimer radiation (600 mJ/cm2 ). Figure 18 flashes show the manner in which underwater irradiation inhibits the expansion of the ejected plasma. This leads to a higher peak stress for such an underwater irradiation. The underwater stress wave was, on average, 52% stronger than that generated in air. Presumably, this is a consequence of the classic witness-plate effect, whereby the water blanket inhibits the free expansion of the blow off plasma. The peak stresses in water and air were 6.1 and 4.0 bars, respectively. 5

RESEARCH AQUARIUM FACILITIES

MPL’s seaside facility has large numbers of aquariums for the study of marine life. In addition there are numerous pools for marine mammals (viz., whales and dolphins). With a strong propensity for marine growth to inundate these facilities, it is a constant struggle to keep them clean and free of parasites and diseases. Figure 19 shows a view of one of these aquariumpool underwater glass windows. The waterproof xenon flashlamp head shown in Figure 14 was placed against the air side of the glass and flashed once. It was next placed against the algae on the water side. After the xenon flashes the algae debris could be seen drifting away from the glass. The circulation system eventually captured the residue in its filters. The flashlamp had a 10 cm arc length and a 7 mm bore. It was operated near its explosion limit at an electrical energy of 3 kJ. No flow tube or water jacket was employed as most of the irradiation of test specimens was performed underwater. Whereas this underwater xenon flashlamp approach to aquarium maintenance was a technical success, lamp life at the required electrical loading was only a few thousand shots. Consequently, a new lamp was

Figure 20. Aluminium aircraft wing panel used to demonstrate coating (paint) removal by means of a high-power flashlamp.

needed for each window (which negated any logistical or cost advantage). 6 AIRCRAFT MAINTENANCE There is an historic airfield across the channel from MPL. Some seventy years ago Charles Lindberg’s craft, “The Spirit of St. Louis” (Fig. 20, inset), was built and tested at that site. The flight to the start of Lindberg’s nonstop solo trans-Atlantic crossing began at this field on San Diego Bay. Thus, there is a well-established tradition of aviation research and engineering at Ballast point. Environmental concerns have complicated the routine maintenance of aircraft. Paint must be stripped regularly in order to inspect for stress-corrosion

6

Subsequently, the spectral camera images the scanned area. Next, the laser rescans the area; however, laser pulses are inhibited at those pixel locations having the character of the substrate composite spectrum. During the scan the laser only fires at pixel locations showing the character of the coating. This cycle repeats until the entire area is clear of the coating. Figure 22 shows the system in operation on an aircraft radome. This system functions by rotating the radome in order to achieve continuous surface scanning while the robot maintains a constant distance between the laser optics and the work surface as well as the lateral scan.

Figure 21. Feedback-controlled laser stripping of grey epoxy coating from advanced aerospace carbon composite test panel. The stripped area is the dark rectangle at the centre.

7

PHOTOCHEMICAL SURFACE PASSIVATION OF STEEL

Steel is the most frequently found material in naval vessels. Most of the effort that is expended in ship repair has to do with the removal of rust. An environmentally friendly approach to the removal of rust from steel and iron is citric acid chelation. Due to its non-toxicity, ease of handling, low cost, and ready biodegradability, hot aqueous solutions of citric acid have conventionally been used to effectively remove rust while maintaining the integrity of the base metal. The thermal mechanism by which it dissolves an iron oxide layer in solution is not complex. The acidity of the solution increases the solubility of the deposit followed by citrate anion chelation of the iron. Agitation and high temperatures accelerate the process. Solution additives (ammonia, triethanolamine) are needed to prevent precipitation of the iron-citrate salt. As effective as citric acid is, it still retains a number of drawbacks. The requirement of large solution volumes at high temperatures (95◦ C) poses constraints. Surfaces cleaned in hot citric acid baths demonstrate little passivation against flash rusting following the final rinse. The most serious drawback of this cleaning technique is that it is slow. As the rate of citric acid chelation is temperature dependent, a new experimental approach was undertaken to utilize superheated citric acid produced by high intensity light pulses. The hybrid photonic flash chelation process begins with the “painting” of the rust with a 6% aqueous citric acid solution. Then the moist rust is superheated with a xenon flashlamp pulse at an optical fluence of 3 J/cm2 . Early in the temporal evolution of the flashlamp pulse the surface temperature rises leading to gaseous release and material blowoff. Both oxygen vapour and hydration water from the rust are explosively removed. Simultaneously, dissolution of the iron oxide layer from the surface and iron-citrate anion chelation occurs. The mechanisms by which these processes progress (photochemical, photolytic, thermal, or some combination) are greatly accelerated by the high transient flashlamp-induced surface temperature. Subsequent sublimation and/or

Figure 22. Robot laser paint stripper removing epoxy coating from an aircraft radome fabricated from a composite material.

cracks. Unfortunately, chemical strippers as well as abrasive procedures (e.g., wet or dry blasting) can pollute the environment and present health hazards to workers. The xenon “flashblaster” shown in Sections 4 and 5 was adapted (through the addition of vacuum and air jet-accessories) to ablative paint stripping. Figure 20 shows the result of the flashlamp removal of the topcoat from an aluminium aircraft wing panel. The selective removal of coatings from aluminium is uniquely favourable due to its high optical reflectively and thermal conductivity. In contrast advanced composites are increasingly utilized in modern aircraft. Unfortunately, in contradistinction to aluminium, composites are low in thermal conductivity and in optical reflectivity. It was found that successful divestment of carbon composites calls for real-time process control. Figure 21 presents a carbon composite test panel that was stripped of its grey epoxy coating (central rectangle) with a pulsed CO2 laser. This was accomplished through system process control by means of multispectral imaging feedback to the scanner/laser machine controller. In this approach the laser performs a raster scan across the designated area.

7

Figure 23. Citric acid treatment with flash rerust (left), flashlamp/citric-acid treatment (centre), original rust (right).

Figure 24. Pacific dolphin named “Aquarius” that was the test animal for the laser branding experiments.

evaporation of the iron-citrate complex serves to etch oxide deposits away. As the surface temperature rises to its final value late in the pulse, a black oxide layer is formed by the reduction and dehydration of the initial layer of atmospheric rust (X-ray diffractometry has shown this black substance to be magnetite, a mixed Fe(II)-Fe(III) oxide). Upon conclusion of the flash pulse, a fresh layer of citric acid solution is applied to the surface and again flash treated. This cycle is repeated as many times as necessary to remove the rust completely. Figure 23 shows the comparative results of citric acid and citric acid plus flashlamp cleaning. Perhaps, the most far-reaching aspect of flashlamp/citric acid cleaning is its passivation. 8

Figure 25. Laser marking test on dorsal fin of a Pacific dolphin. The laser “brands” are the series of small white spots near the top of the figure. The large white spot near the centre of the frame is an earlier liquid nitrogen “freeze brand”.

MARINE MAMMAL RESEARCH AND CONSERVATION

One facility at MPL is the marine mammal research and training complex of large pools for large animals. The study and conservation of California grey and bowhead whales are two of the major activities. The identification and counting of individual whales during their annual migrations is a principal responsibility. The most challenging problem associated with the program is the identification of specific individuals. Today, identification depends upon an observer noting characteristics (size, colour, injuries, colour, and scars) of an animal in the wild (Fig. 24). This is difficult, labour intensive, and limited in reliability. For decades researchers have tested marking (or branding) techniques in order to enhance the individuality of whales and to facilitate easier counting. MPL assigned one of the captive dolphins for laser branding tests (Fig. 24). Figure 25 shows the tests being conducted on the dorsal fin of the animal. The laser radiation bleaches the melanin in the external tissue and may inactivate the malanocytes so that the brand

will not fade with time. When implemented, the visibility of the laser brands should permit counting from the Scripps Pier. 9

CONCLUSIONS

It is ironic that entirely new laser technologies for art conservation emerged from oceanographic research. That irony is compounded by the fact that these same technologies migrated back to oceanography. REFERENCES Asmus, J., Munk, W. & Wuerker, R. 1972. Lasers and Holography in Art Preservation and Restoration. IEEE NEREM Proceedings (November): 17–20. Lorenzetti, G. 1961. Venice and its Lagoon. Trieste: Lint. Ready, J. 1971. Effects of High-Power Laser Radiation. New York: Academic Press.

8

Innovative Approaches in Laser Cleaning and Analysis

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Towards the restoration of darkened red lead containing mural paintings: A preliminary study of the β-PbO2 to Pb3 O4 reversion by laser irradiation S. Aze & P. Delaporte LP3, Marseille, France

J.M. Vallet CICRP, Marseille, France

V. Detalle LRMH, Champs sur Marne, France

O. Grauby & A. Baronnet CRMCN, Marseille, France

ABSTRACT: Red lead pigment darkening in paintings is generally caused by the pigment alteration into plattnerite (β-PbO2 ). Plattnerite reversion into minium may be achieved by heating it over 375◦ C. This reaction may occur by using continuous-wave (cw) laser irradiation. Laser-induced photo-thermal reduction of plattnerite into minium was investigated through irradiation tests on both pure plattnerite powder samples and darkened red lead paint samples taken from experimental wall paintings. The effects of both visible (514 nm) and near infrared (1064 nm) sources were investigated. This study shows the high potentiality of an innovative technique based on a continuous wave-laser irradiation for the restoration of darkened red lead-containing paintings.

1

INTRODUCTION

Restoring blackened red lead containing paintings may consist in recovering the original red colour. In most cases, removing the black layer is inconceivable, considering the low amount of remaining pigment. The most sustainable solution might be based on the reversion of the alteration products into red lead. The main component of traditional red lead pigment is similar to minium, the mineral mixed valence lead oxide of formula Pb3 O4 . However, as a consequence of the manufacture process, significant amounts of lead monoxide may remain in the pigment. Both crystalline forms, i.e. orthorhombic (massicot) or tetragonal (litharge), have been identified in red lead-containing artworks. Either chemical or physical routes may achieve reversion of plattnerite into minium. Due to the complexity of the Pb-O system, which includes a number of distinct minerals, a chemical reduction of β-PbO2 into Pb3 O4 is challenging. Lead dioxide, however, is able to undergo thermal reduction through successive oxygen losses, which lead to the formation of lower oxides. In particular, plattnerite evolution into minium is spontaneous over ca. 375◦ C. This thermal transformation can be attained by using laser irradiation (Burgio et al. 2001). In order to profit from this property of a possible reversion of

Applications of laser beam techniques in the field of Art and Archaeology have been widely developed since a few decades. Both analytical and imaging methods, such as micro-Raman Spectroscopy, Laser Induced Breakdown Spectroscopy (LIBS), Laser Radar, Interferometry and Holography, provide useful information related to the object composition, conservation state and defects. Laser-based restoration techniques which mainly consist of surface cleaning (varnish removal from paintings, elimination of gypsum crusts from the rock materials) use either continuous wave or pulsed Q-switched laser radiations. This work attempts to apply continuous wave (cw) laser radiation for the restoration of darkened red leadcontaining paintings. Such an alteration phenomenon is commonly observed in mural paintings. Previous works showed that red lead darkening often result from the pigment transformation into black lead dioxide (plattnerite, β-PbO2 ). The degradation process is mainly controlled by both environmental factors, such as humidity, light, temperature or sulphur-containing pollutants. Besides, influence of intrinsic parameters, including composition and grain structure of the pigment, has been stated.

11

Table 1. Visual effects of Ar+ irradiation tests on pure plattnerite powder samples as a function of the laser power (P) and irradiation time (I). (R: reddishing, Y: yellowing).

darkened red lead, a series of laser irradiation tests were carried out. 2 2.1

P(W)

EXPERIMENTAL Samples

The effects of laser irradiation on plattnerite were investigated on both raw samples (powdered β-PbO2 , Merck, 99.95%) and blackened red lead-containing samples taken from experimental wall paintings (Morineau & Stefanaggi 1995, Aze 2005). Details related to both mural alterations and red lead discoloration have been described in previous papers (Aze 2006, Aze et al. 2007).

I(s)

0.75

0.90

1.00

1.10

1.25

1 5 10 30 60 120

None None None None None None

None Slight R Slight R Slight R Slight R Slight R

R R R R R R

R Y Y Y Y Y

Y Y Y Y Y Y

2.2 Methodology Irradiation tests were carried out with both visible and near-infrared radiation, using a cw-Ar+ laser (514 nm) and a cw-Nd:YAG laser (1064 nm), respectively. The laser beams were set-up using a set of optical devices, including both converging and diverging lenses, diaphragms and mirrors. The resulting laser beam size, measured with a CCD device, was approximately of 1.8 mm2 . The effect of the main irradiation parameters, namely, irradiation times and laser power density was investigated. The result of each irradiation test on the plattnerite samples was estimated through observations of the sample surface by means of optical microscopy. Figure 1. Raman spectra obtained from (a) the red areas, (b) the yellow areas resulting from plattnerite irradiation using Ar+ laser. The spectra match with the reference spectra of Pb3 O4 (minium) and β-PbO (massicot), respectively.

2.3 Characterisation of the method Local analyses of both sample surface and crosssections were carried out using a micro-Raman spectrometer. Spectra of ca. 1 µm2 areas were recorded using a Renishaw inVia system equipped with a Spectra Physics Ar+ Laser (514.5 nm, 20 mW) and a Renishaw GaAs diode Laser (785 nm, 300 mW) calibrated using the 520.5 cm−1 line of a silicon wafer. Laser power, optical magnification and irradiation times were selected so that no degradation occurs due to photo-thermal effects. Spectral separation of the scattered photons was performed using both a Notch filter and grating monochromators (1800 l/mm with 514.5 nm, 1200 l/mm with 785 nm). Photons were collected over the 100–3000 cm−1 range with a spectral resolution of 1 cm−1 using a Peltier-cooled charge-coupled device (CCD) detector. 3

A slight reddishing of the plattnerite grains was observed in these conditions of irradiation longer than 5 seconds. In the case of longer irradiation times (up to 120 seconds), no visible increase of the reddishing intensity was observed. At higher laser powers, the reddishing was noticeably as being more intense, until a strong yellowing took place. For irradiation times over 5 seconds, this phenomenon occurred at laser powers higher than 1.10 W. Micro-Raman spectroscopic analyses of the red and yellow phases showed the reduction of plattnerite grains into Pb3 O4 (minium) and β-PbO (massicot), respectively (Fig. 1). 3.2 Nd:YAG laser irradiation tests A slight reddishing of the plattnerite powder samples occurred at 0.15 W (Table 2). For long irradiation times (t > 5 seconds), the reddishing was more intense for laser powers between 1.0 and 3.6 W. According to microscopic observations, all visible plattnerite grains were transformed for laser powers higher than 2 W.

RESULTS

3.1 Ar+ laser irradiation tests Visible effects of the laser irradiation appeared for a laser power higher than 0.90 W (Table 1).

12

a consequence, the laser power density threshold for minium reduction into massicot is much higher with Nd:YAG irradiation than with Ar+ irradiation. Influence of the irradiation time appears to be negligible, over few seconds. We thus suppose that local temperature of the irradiated material quickly rises until it reaches a maximum corresponding to a stable thermal regime.

Table 2. Visual effects of Nd:YAG irradiation tests on pure plattnerite powder samples as a function of the laser power (P) and irradiation time (I). P(W) I(s)

0.10

0.15

1.00

3.00

3.60

1 5 10 30 60 120

None None None None None None

Slight R Slight R Slight R Slight R Slight R Slight R

Slight R R R R R R

R R R R R R

Y Y Y Y Y Y

5

CONCLUSION

Plattnerite irradiation using Ar+ laser (514 nm) leads to the formation of pure minium within a slight laser power density range. Due to the absorption of laser light by minium, massicot is readily produced. On the contrary, Nd:YAG irradiation (1064 nm) produces minium over a large power density range. Such irradiation tests show the high potentiality of continuous wave Nd:YAG laser for the possible reversion of blackened red lead pigment in paintings. ACKNOWLEDGEMENTS The authors wish to acknowledge the French Ministry of Culture and Communication for financial support, and the PLANI group from the LILM laboratory (CEA) which supplied the experimental set up.

Figure 2. Raman spectrum of the red phase obtained from plattnerite irradiated with cw-Nd:YAG laser at 2 W, compared to the reference Raman spectrum of minium Pb3 O4 .

REFERENCES

Yellowing of the plattnerite sample was observed over 3.60 W. According to micro-Raman analyses of the red phase obtained at 3.0 W, most of plattnerite grains have been reduced into minium (Fig. 2). The presence of an additional band near 342 cm−1 may be attributed to the lead sesquioxyde of formula Pb2 O3.33 (Aze 2005). 4

Aze, S. 2005. Altérations chromatiques des pigments au plomb dans les oeuvres du Patrimoine, PhD thesis, University of Marseille. Aze, S., Vallet, J.-M., Baronnet, A., Grauby, O. 2006. The fading of red lead pigment in wall paintings: tracking the physico-chemical transformations by means of complementary micro-analysis techniques. European Journal of Mineralogy 18: 835–843. Aze, S., Vallet, J.-M., Pomet, M., Baronnet, A., Grauby, O. 2007. Red lead darkening in wall paintings: natural ageing of experimental wall paintings versus artificial ageing tests. European Journal of Mineralogy 19: 883–890. Burgio, L., Clark, R. J. H., Firth, S. 2001. Raman spectroscopy as a means for the identification of plattnerite (PbO2 ), of lead pigments and of their degradation products. Analyst 126: 222–227. Ciomartan, D.A., Clark, R.J.H., McDonald, L.J., Odlyha, M. 1996. Studies on the thermal decomposition of basic lead (II) carbonate by Fourier-transform Raman spectroscopy, X-ray diffraction, and thermal analysis. Journal of Chemical Society – Dalton Transactions 18: 3639–3645. Clark, G. L., Schieltz, N. C., Quirke, T. T. 1937. A New Study of the Preparation and Properties of the Higher Oxides of Lead. Journal of the American Chemical Society 59: 2305–2308. Morineau, A., Stefanaggi, M. 1995. A statistical approach to the problem of frescoes:The French experience. Statistical Methods and Applications 4: 37–53.

DISCUSSION

Plattnerite reduction into minium spontaneously takes place over 375◦ C (Clark et al. 1937). On the other hand, minium itself is reduced into massicot over 512◦ C (Ciomartan et al. 1996). Depending on the irradiation parameters, plattnerite may thus be transformed into either minium or massicot. Irradiation of plattnerite by Ar+ laser (514 nm) initially leads to the formation of minium. Over a certain power density threshold, minium is reduced into massicot, due to the high absorption of the green laser light. When irradiated by near-IR laser beam (Nd:YAG, 1064 nm), a similar phenomenon occurs. Minium produced by plattnerite reduction, however, has a relatively low absorptivity in the IR spectral domain. As

13

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Prospective and applications of two-photon fluorescence in archaeology and art conservation D. Artigas & L. Serrado Department of Signal Theory and Communications, Universitat Politècnica de Catalunya, Barcelona, Spain

I.G. Cormack, S. Psilodimitrakopoulos & P. Loza-Alvarez ICFO-Institut de Ciències Fotòniques, Castelldefels, Spain

ABSTRACT: Two-Photon Excited Fluorescence (TPEF) is proposed as a new technique in archaeology and art conservation. In this paper, the fundamental basis of the technique and its main characteristics are explained. As an example of its use in archaeology, a recent work on the recovery of missing writings from an archaeological object is reviewed. In this case, the small quantity of paint and the necessity of ensuring low photodamage made TPEF the best choice. We also propose the use of TPEF, to perform optical sectioning in conservation studies for three dimensional characterization of the layer-distribution depth into the paint of an artwork.

1

INTRODUCTION

ideal technique in live science microscopy. However, the need of expensive ultrashort pulsed laser sources, usually a Titanium:sapphire oscillator and the relative complexity of these systems has, so far, restricted its wider application. The situation is changing rapidly. A new generation of cheap diode-pumped solid-state femtosecond oscillators is now under development (Droum et al. 2007). As a result, compact, low cost systems based on Yb:KYW crystals are being commercialized. This is making the use of TPEF a more attractive technique in biomedical applications than the conventional fluorescence techniques, which are currently routinely used. The objectives here are to understand how the materials were made and applied and to determine its origin and trade pattern. These objectives are normally achieved using spectroscopy techniques (Clark 2002), such as laser induced fluorescence (LIF) (Anglos et al. 1996, Weibring et al. 2001), Raman microscopy (Clark et al. 1998; Smith et al. 2004) or laser induced breakdown spectroscopy (LIBS) (Anglos 2001, Muller 2003). These are very powerful techniques which can be extensively used and combined in a number of situations (Castillejo et al. 2001). In addition to these techniques, we devise an application where the use of TPEF in archaeology and art conservation is considered to be more advantageous than the conventional techniques. A first application of TPEF is in those situations where the quantity of material to be identified is present in small amounts. Signals obtained with Raman microscopy from very low concentration of

Two photon absorption was first predicted by Maria Goeppert-Mayer in 1931 as a single quantum event. Here two photons with equal or different wavelengths are simultaneously absorbed by a molecule, thereby, exciting it from its ground state to an excited estate. This nonlinear effect is associated with the imaginary part of the third order susceptibility, and therefore, its absorption efficiency scales with the light intensity (Boyd 1992). The high intensity is usually reached by using mode-locked lasers, which provides the required high peak powers while at the same time maintains an energy low enough to preserve the sample. When the relaxation to the ground state is radiative the process is known as two-photon excited fluorescence (TPEF). Two-photon microscopy is one of the more interesting applications of two-photon fluorescence (Denk 1990). It offers some advantages over the widely used confocal fluorescence microscopy. In a two-photon microscope, the highest intensity is reached at the focal volume of the objective and TPEF is therefore generated only at this point. This provides optical sectioning without the need of a confocal aperture. In addition, the small confined excitation volume also minimizes the total photobleaching and photodamage. Since the excitation is produced with photons with half the frequency of the energy gap between the ground and the excited state of the molecule, laser sources in the near infrared are used. This further reduces the possibility of photodamage and increases the penetration depth in the sample. All these advantages make TPEF the

15

the target sample would be indistinguishable from the background signal. In LIF, high energy UV photons are needed for excitation. These can damage the valuable artwork material. Finally, LIBS is an intrinsically semi-destructive method, and therefore, not suitable for these purposes: low photodamage probability and photobleaching afforded by TPEF makes the technique highly suitable to address such problems. A second advantage of this technique is the potential to provide information about the depth of material within the sample surface, i.e., the painting layer morphology. This is possible due to the intrinsic optical sectioning of the nonlinear effect, providing 3D resolution, for both imaging and characterization, and to the use of IR light, which increases the penetration depth. In this paper we discus the potential of TPEF by analyzing these two applications. To do that, the remaining of the paper is organized as follows. In section 2, TPEF and other nonlinear techniques that have been successfully applied to microscopy are described. This is followed, in section 3, by an example of the use ofTPEF to detect painting in an archaeological sample. The use of 3D characterization of paints and determination of the layer morphology is proposed in section 4. Finally, in section 5, discussion and conclusions are presented.

Figure 1. Electronic energy level scheme showing the different excitation processes: one-photon (linear) absorption and two-photon (nonlinear) absorption.

to non-radiative relaxation between vibrational levels. However, the radiated wavelength is still much shorter than the long wavelength used to produce two-photon absorption. This large Stokes shift makes filtering of the near-IR excitation wavelengths from the fluorescent signal wavelengths very effective. The excitation of TPEF in microscopy requires a laser source able to provide the required high intensity wile at the same time, maintaining a low enough averaged power (Pav ) to preserve the sample. This can be achieved by using mode-locked lasers such as Titanium:sapphire laser. What makes this laser interesting for biological applications is its wide wavelength tuning range, λ = 720–920 nm. This spectral range allows pumping most of the dyes with linear absorption region in the green-blue-UV were specifically developed for fluorescence microscopy. Examples are: rodamin-B, fluorescein, coumarin, and the green-fluorescent protein present in genetically modified organisms (Xu 1996). A standard commercial system can deliver ultrashort pulses with duration in the range  tp = 50–200 fs, repetition rate around fR = 80 MHz, and average powers up to 2W. This results in peak powers above Pp = 100 KW. When the IR beam is focused into the sample, most light is mainly absorbed at the focus. The number of absorbed photons can be estimated by the following equation:

2 TWO-PHOTON FLUORESCENCE AND NONLINEAR MICROSCOPY 2.1 Two photon fluorescence Two-photon absorption is a nonlinear effect associated to the presence of the third-order susceptibility, χ(3) , which is present in any material (Boyd 1992). This effect can occur when the energy difference between electronics energy levels is similar to the added energy of two photons. In order to observe this effect, the two photons must coincide at the same position within the material. In addition, since the process lasts only for a very short period, in the femtosecond-attosecond scale, the two photons must effectively coincide also in the time domain. As a consequence, the absorption probability, and thus the efficiency, depend on the photon density, i.e., on the light intensity. The efficiency also depends on the wavelength (absorption bandwidth) and the material capability to perform two-photon absorption, features that are characterized by the two-photon absorption cross-section σ2p . Once an atom or molecule has been excited by two photons, the relaxation to the ground state is identical to the situation with one-photon (linear) excitation. When the process is in part radiative, the result is the expected fluorescence (see Fig. 1). Usually, the energy of the emitted photon is lower compared to the sum of the energy of the absorbed photons. This is due

where NA is the objective’s numerical aperture, h is the Plank constant and c the speed of light. The most important differences between this and the one-photon (linear) absorption are the quadratic dependence on the average power (linear in one-photon) and the

16

Figure 2. Volume at which fluorescence (white) is generated. (a) In one-photon fluorescence, it is generated mainly at the surface since the power of the excitation beam (at λ = 354 nm in this example) is depleted during propagation due to the linear absorption. (b) In two-photon fluorescence, it is generated only at the focal volume where the highest intensity is reached. Here, there is not absorption outside the focal volume and therefore there is not power depletion (in this case at λ = 708 nm the medium is transparent).

Figure 3. Scheme of an inverted nonlinear microscope simultaneously operating with TPEF (epi-detection), SHG and THG (forward detection). A CARS microscope would require at least an additional laser source and extra modifications.

different points of a three dimensional object with a precision that can be smaller than 1 µm3 . As mentioned before, virtually all the detected fluorescence arrives from the focal point. The entire generated signal therefore contributes to the detection, allowing the use of moderate to low excitation average powers. On the contrary, in confocal microscopy, due to the use of a pinhole lower detected signal are usually compensated by higher excitation powers, compromising the integrity of the sample. Another important advantage of TPEF in microscopy is the higher penetration depth of IR light. Rayleigh scattering is the main limitation to penetration length. It scales with the inverse of the wavelength, λ−n , with a power n that varies between 2.2 and 4, depending on the material. Therefore, scattering is in general an order of magnitude lower with TPEF than that the equivalent one-photon excitation in the visible-UV spectrum. Finally, photodamage and photobleaching in TFP is reduced to the focal volume, whereas in one-photon fluorescence the upper and lower regions of the excitation light cone are affected. This two negative factor are further reduced due to the use of a comparative lower intensity (since there is no pinhole) and the lower energy of the IR photons to produce the same excitation.

power-of-four dependence on the NA of the microscope objective (quadratic in one-photon). The result is that with standard values of the NA, around 80% of the photons are absorbed at the focus. This means that the fluorescence is mainly generated at this point, as shown schematically in Figure 2b, providing what is known as intrinsic optical sectioning for 3D imaging. On the other hand, with one-photon (linear) absorption, the fluorescence, is generated mainly at the surface, decreasing along the beam path (as shown in Fig. 2a). Therefore, optical sectioning can only be achieved by using a confocal pinhole.

2.2 Two-photon fluorescence scanning microscopy A two-photon fluorescent scanning microscope is depicted in Figure 3. It consists on the laser source, a scanning system (usually two galvanometric mirrors) in one of the inverted-microscope ports and a hot mirror leading the beam to the microscope objective. The fluorescence is then collected by the same objective and directed (crossing the hot mirror) to the detection system, coupled in one of the microscope output ports. Additional filters (BG39) can be used to stop the remaining IR light. For imaging purposes, the detection system is a standard photomultiplier. The excitation beam is then scanned at the sample plane using the two galvanometric mirrors. The generated signal is recorded and used to computer-reconstruct the 3D image. In spectroscopy applications, a spectrograph and a CCD camera can be coupled to the output port. This allows recording the fluorescence spectra at

2.3

Other nonlinear effects applied to microscopy

Two-photon scanning microscopy belongs to a broader family of microscopy techniques making use of nonlinear effects. This family of techniques is known as nonlinear or multiphoton microscopy. In this subsection, for completeness, we briefly comment on these techniques. The best known multiphoton techniques are second-harmonic generation (SHG) microscopy

17

(Gannaway 1978) and third-harmonic generation (THG) microscopy (Barad 1997). Another interesting nonlinear technique is also commented: coherent anti-stokes Raman scattering (CARS) microscopy (Zumbusch 1999). Since all these techniques are based on nonlinear processes they share one of the main characteristics of the TPEF microscopy: intrinsic optical sectioning. The first modality, SHG-microscopy is associated with the second order nonlinear susceptibility, χ(2) . This nonlinearity appears in non-centrosymmetric molecules or crystalline structures and is responsible for a variety of two wave mixing processes (Boyd 1992). SHG can also be generated in surfaces (that break the centrosymmetry) and nanocrystals (Wang 1996) and can be enhanced in structured patterns due to quasi-phasematching. Here, the generated SHG light can be used to distinguish material with large χ(2) or areas with a definite-ordered structure from the surrounding materials. THG is associated with the real part of χ(3) susceptibility. This nonlinear effect is present in all materials. However, due to the phase change in a Gaussian beam along the focus (Gouy phase) the THG signal is reconverted back to the fundamental wave after the focus. This THG signal only appears when the focus is in an interface between two media with different refractive indices. Then, a fraction of the generated signal before the focus is not reconverted. As a consequence, by detecting this signal, the corresponding interface can be imaged. Finally CARS again relays on the χ(3) nonlinear susceptibility. In this case it makes use of two laser sources, the first corresponding to the pump and the second to the Stokes wavelength, that interacts generating coherent anti-Stokes light. The generated light is resonantly enhanced when the difference in photon energy between the pump and the Stokes beam coincides with the energy difference between two vibrational states (a Raman resonance). The observed CARS signal is then related to the presence (and concentration) of species which resonantly react at the pump-stokes frequency difference.

Figure 4. Amphora under study, above (a), enlarged image showing the state of the paint (consular date), below (b).

war. At some moment they decided to take control of the country starting this foundation program. Among the first founded cities there were Empúries, Beatulo (Badalona) and Iluro (Mataró), together with others settlements not so well known as Iesso (Guissona) (Guitart et al. 2006). The exact foundation date of these cities is however still not known. Recently, in the archaeological site of Iesso, a piece of an amphora of the type Dressel A1 was found (Fig. 4a). In this piece there was some painted writing corresponding to the consular date (the name of the two consuls ruling in the year of wine production), a sign of good quality wine. In this case, it is believed that this amphora was brought by the first Roman Legions arriving to Iesso and that it was consumed during some ritual ceremony related with the city foundation (Guitard et al. 2004). The consular date in the amphora could therefore provide a terminus post quem for the city foundation of Iesso and indirectly, the starting of the Roman colonization of Hispania Citerior. Unfortunately, the past of time has deteriorated the painted area in the amphora (see Fig. 4b). At the moment of the discovery it was still possible to recognize some letters:

3 AN EXAMPLE OF PAINT DETECTION IN ARCHAEOLOGICAL ARTIFACTS 3.1 The archaeological problem Although the Roman colonization of Hispania is a well known process, there are still some events which have not been clearly determined. One of these cases is the starting of a city foundation program that took place at the end of the II century BC in Hispania Citerior. Previously to this foundation program, the Romans remained in Tarraco (Tarragona) as the only Roman possession since their arrival during the Second Punic

The first group of letters corresponds to the initials of the first consul, Quintus Fabius. The second group means consolibus (being consuls). The missing letters in the area between the two groups of letters correspond to the second consul. By examining the list of Roman consuls at the end of the 2nd and beginning of the 1st centuries BC, there were two occasions at which a consul had the name Quintus Fabius. The first

18

one, Quintus Fabius Maximus Allobrogicus was consul in the years 121 BC, together with Lucius Opimius. The second possibility was Quintus Fabius Maximus Eburnus, who was consul in the year 116 BC, together with Caius Licinius Geta. Therefore, the objective was to recover the missing letters by mapping the paint distribution using a laser scanning technique and to identify through fluorescence which of the two possible options was written in the amphora (Cormack et al. 2007). This would determine the year for the wine harvest contained in the amphora and thus provide a better indication of the city foundation date. 3.2

Paint detection

Figure 5. (a) Direct two-photon fluorescent image obtained after scanning the area shown in figure 4(b), (b) result after performing a Gaussian convolution. Squares show the area that could correspond to the Q and the F letters of the first consul Quintus Fabius, (c) equalization histogram of the same area. Squares show areas where paint could be observed by eye inspection resulting in no fluorescence.

One possibility to uncover the missing name is to detect small amounts of pigments and/or agglutinant to form an image of the missing letters. A method based on detecting the fluorescence generated by the paint was sought. As commented, TPEF was chosen to recover the missing letters in the amphora, since the use of IR excitation (when compared to UV excitation) helped to prevent damaging the paint. Thus, Cormack et al. (2007) used a Titanium:sapphire laser operating at 830 nm, delivering pulses of 150 fs at a 76 MHz repetition rate to excite the TPEF. Here, as high resolution was not required, a 125 mm focal length lens was used as the microscope’s objective (spot size 25 µm). Some preliminary experiments with fragments of amphora found in the same archaeological site showed that the only source of fluorescence was from painted areas. The main assumption was that the fluorescence would probably be generated by the agglutinant while the pigment (probably red ochre) would inhibit it. The collected fluorescence was centered at 570 nm with 80 nm bandwidth. This fluorescence was considerably shifted to the yellow-orange range, rather unusual for organic binding media (Nevin et al. 2006). It was therefore not possible to identify the agglutinant. These preliminary experiments also allowed determination of the averaged power thresholds for both, the fluorescence emission (20 mW) and damage (80–90 mW) measured at the sample plane. In addition, the reliability of the imaging system and some basic image processing techniques was tested using these fragments of amphora. After these preliminary results, the area corresponding to the consular date (Fig. 4b) was scanned. To do that, the total area was divided into three sections to better adapt to the curvature of the amphora. The scan was performed at three different depths to maximize the fluorescence. Unfortunately, the resulting image, shown in Figure 5a, did not give enough information to identify the missing letters. To do that, some more sophisticated image processing techniques and the use of historical information was required.

3.3 Identification of the missing letters As commented, in the preliminary experiments an image processing technique was tested. This was based on a Gaussian convolution and binarization that allowed determination of the painted area. Figure 5b shows the result after a Gaussian convolution. Here some regions at the beginning of the analyzed area could be associated with the letter Q and F of Quintus Fabius, the first consul. Nevertheless, the results were poor and no information was obtained from the second consul. A more sophisticated image processing technique based on equalization histogram (Castleman, 1996) was then used. Although the first attempt shown in Figure 5c was not a success, it allowed us to obtain very valuable information. First, it showed that areas in the amphora where the paint was visible by eye inspection resulted in black color (no fluorescence) areas while areas in white (larger fluorescence) were obtained outside painted areas. This is in agreement with previous reported results showing that the fluorescence observed in paint based upon metallic compounds is generated by the agglutinant, while the pigment actually inhibits the generation of fluorescence (Miyoshi, 1988). Second, we noticed that large fluorescence obtained from well preserved areas was decreasing the contrast in the region where the name of the second consul was. With the previous information, the equalization histogram was performed in areas corresponding to the size of a letter and changing the grey scale so that the white color corresponds either to a maximum or to a minimum of fluorescence. The results corresponding to the direct equalization histogram are shown in

19

surface of the paint. X-ray radiography (Schreiner 2004) and infrared reflectography (Van Asperen de Boer 1969) are extensively used to see below the surface, however they reduce the 3D information to a 2D plane, loosing information about the localization of a special feature in the vertical direction. The traditional procedure to solve this problem requires the extraction of a sample from the painting to be analyzed using different methods as can be scanning electron microscopy-energy dispersive X-ray spectroscopy (Ciliberto et al. 2000). Recently, two non destructive different alternatives have been proposed. The first one, optical coherent tomography (OCT), has been applied to analyze both archaeological samples (Yang et al. 2004) and in conservation studies (Targowski et al. 2004). This technique can accurately provide information of multilayer paint morphology and it is especially well suited to measure varnish thickness. However, it lacks the required characterization capability. The second technique is based on a sophisticated confocal X-ray fluorescence microscope able to characterize different layers with a total thickness around 500 µm (Woll 2006) with an axial precision up to 5 µm. LIF confocal microscopy, has also been proposed as alternative technique (Hogan 2007), although to our knowledge, no practical demonstration has been performed. In principle, this would be a more simple technique to characterize different paint layers. However, the use of visible-UV light would greatly reduce the penetration depth due to the high absorption and scattering at these wavelengths. This is the case in industrial application for polymeric paint analysis, where the poor penetration depth and photobleaching, have limited the scope of it use. For industrial applications, two alternative techniques have been proposed. The first one is a confocal aperture within a Raman microscope that can effectively characterize paint deep within the sample (Tabaksblat 1992). In this case, similarly to a fluorescent confocal microscope, only a small fraction of the generated signal (the one passing through the pinhole) is captured by the detector, reducing the sensitivity of the system. The second alternative is based on a TPEF microscope. Here, the entire signal is originated at the focus, making the detection system more simple and efficient, although it lacks the characterization specificity of Raman-based techniques. Two-photon microscopy has been used to characterize fluorophore-doped polymer blocks with submicron resolution, in multilayer structures (Bhawalkar 1997). In this work, the lower scattering and the greater transparency at IR wavelengths allowed characterizing structures up to a depth of 200 µm. The previous evidence shows the viability of twophoton microscopy to characterize the composition and structure of buried layers in artwork painting. The

Figure 6. (a) Histogram equalization corresponding to the letters forming the initials of the second consul, (b) result after performing a Gaussian convolution and a binarization to figure (a), (c) first option for the second consul: Lucius Opimius and in (d) second option, Caius Licinius Geta. In both cases, capitalis rustica letterform was used.

Figure 6a. In addition, after the equalization histogram, a Gaussian convolution and a posterior binarization was performed to help the eye, resulting in the image in Figure 6b. This figure shows clear, well defined areas where the original paint distribution can be identified. However, the immediate identification of the missing letters was still not possible and it was necessary to compare with the two possible options.The first option, OPI, corresponding to the surname for Lucius Opinius and the second option, LIC for Caius Licinius Geta, are written in Figure 6c, and 6d respectively. For a better evaluation, in both cases, the surnames initials were written using capitalis rustica, the informal cursive capital letterform used in those times (Lowe 1972). Comparing Figures 6b and 6c, the first letter can be identified as an O while the top part of the letter P in capitalis rustica clearly resembles the central panel in Figure 6b. Finally the last letter could be an I. Therefore it was assumed that the name of the second consul was Lucius Opinius, and the wine in the amphora was from the 121 BC harvest. With the results showed in that work, the archeologist must now determine the wine aging to more precisely determine the foundation date of Iesso and the starting of the roman colonization of Hispania Citerior. 4

PROSPECT OF TWO-PHOTON EXCITED FLUORESCENCE FOR PAINT MORPHOLOGY EXAMINATION

There are situations in which the knowledge of an artwork condition, the extent of a previous restoration or the working method of an artist requires a layer-by-layer measurement of the sample. LIF, Raman microscopy and LIBS give information mostly on the

20

Clark, R. J. H. 2002. Pigment identification by spectroscopic means: an arts/science interface, C. R. Chimie 5; 7–20. Cormack, I. G., Loza-Alvarez, P., Sarrado, L., Tomas. S., Amat-Roldan, I., Torner, L., Artigas, D., Guitart, J., Pera, J. & Ros, J. 2007. Lost writing uncovered by laser two-photon fluorescence provides a terminus post quem for Roman colonization of Hispania Citerior. J. Archaeol. Sci. 34: 1594–1600. Denk, W., Strickler, J. H., & Webb, W. W. 1990. Two-photon Laser Scanning Fluorescence Microscopy, Science 248: 73–76. Droum, F., Balembois, F. & George, P. 2007. C.R. Physique 8: 153–164. Gannaway, J.N. & Sheppard, C. 1978. 2nd-harmonic imaging in scanning optical microscope, Opt. Quantum Electron. 10: 435–439. Guitart, J., Pera, J. & Carreras, C. 2004. La presència del vi itàlic a les fundacions urbanes del principi del segle I aC a l’interior de Catalunya: l’exemple de Iesso. In J. Guitart & J. Pera (eds.), El vi a l’antiguitat. Econòmia, producció i comerç¸ al Mediterrani Occidental, Badalona (1998). Iesso I. Miscellània arqueològica, Barcelona-Guissona. Guitart, J., 2006. Iluro, Baetulo and Iesso and the establishment of the Roman town model in the territory of present-day Catalonia. In: Early roman towns in Hispania Tarraconense (2nd century BCe1st century AD), Portsmouth. Hogan, H., 2007. Photonics for art’s sake, Photonics Spectra 41: 46–53. Lowe, E.A., 1972. Codices Latini antiquiores.A Palaeographical Guide to Latin Manuscripts Prior to the Ninth Century. Oxford: The Clarendon Press. Miyoshi, T. 1988. Fluorescence from colours used for Japanese painting Ander N2 laser excitation. Jpn. J. Appl. Phys 27: 627–630. Muller, K. 2003. Evaluation of the analytical potential of laser-induced breakdown spectrometry (libs) for the analysis of historical glasses, Archaeometry 45: 421–433. Nevin, A., Cather, S., Anglos, D., Fotakis, C., 2006. Analysis of protein-based binding media found in paintings using laser induced fluorescence spectroscopy, Anal. Chim. Acta 573: 341–346. Schreiner, M., Frühmann, B., Jembrih-Simbürger, D. & Linke, R., 2004. X-rays in art and archaeology: An overview, Powder Diffraction 19: 3–11. Smith, G. D. & Clark, R. J. H. 2004. Raman microscopy in archaeological sciences. J. Archaeol. Sci. 31: 1137–1160. Tabaksblat R., Meier, R. J. & Kip, B. J. 1992. Confocal Raman microspectroscopy – Theory and application to thin polymer samples. Applied Spectroscopy 46: 60–68. Targowski, P., Rouba, B., Wojtkowski, M. & Kowalczyk, A. 2004. The application of optical coherence tomography to non-destructive examination of museum objects. Studies in Conservation 49: 107–114. VanAsperen de Boer, J.R.J. 1969. Reflectography of paintings using an infra-red vidicon television system. Studies in Conservation 14: 96–118. Wang, H., Yan, E.C.Y., Borguet, E. & Eisenthal, K.B. 1996. Second harmonic generation from the surface of centrosymmetric particles in bulk solution. Chem. Phys. Lett. 259: 15–20.

technique would consist of a spectrometer coupled to the output port of a nonlinear microscope. Then by scanning the sample and observing the emitted fluorescence spectrum, information about the paint structure and composition deep in the sample can be obtained.

5

DISCUSSION AND CONCLUSIONS

TPEF is a more sophisticated variant of the well known LIF technique widely used in archaeology and art conservation. In most of the standard situations, TPEF provides the same characterization features than LIF, but a more complicated ultrashort pulse laser source is required. Nevertheless, there are two situations where its unique characteristics makeTPEF the ideal option.The first one is when an extreme care is required and any kind of damage is not allowed. Here the use of low energy IR photons increases the photodamage threshold and at the same time, and in case this is produced it only occurs at the focal point. A second application has been proposed in those cases where information about the material depth in the sample or the paint layer structure is required. However, to demonstrate the viability of this technique, systematic studies are required to properly evaluate its properties, advantages and disadvantages for its application in cultural heritage conservation. REFERENCES Anglos, D., Solomidou, M., Zergioti, I., Zafiropulos, V., Papazoglou, T. G. & Fotakis, C. 1996. Laser-induced fluorescence in artwork diagnostics:An application in pigment analysis. Applied Spectroscopy 50; 1331–1334. Anglos, D. 2001. Laser-induced breakdown spectroscopy in art and archaeology. Applied Spectroscopy. 55; 186A– 205A. Barad,Y., Eisenberg, H., Horowitz, M. & Silberberg,Y. 1997. Nonlinear scanning laser microscopy by third harmonic generation. Appl. Phys. Lett. 70: 922–924. Boyd, R. W. 1992. Nonlinear Optics. Boston:Academic Press. Castillejo, M., Martin, M., Oujja, M., Silva, D., Torres, R., Domingo, C., Garcia-Ramos, JV. & Sanchez-Cortes, S. 2001. Spectroscopic analysis of pigments and binding media of polychromes by the combination of optical laserbased and vibrational techniques. Applied Spectroscopy 55; 992–998. Castleman, K.R. 1996. Digital Image Processing. New Jersey: Prentice May. Ciliberto, E., Spoto, G. 2000, Modern Analytical Methods in Art and Archaeology. New York: Wiley. Clark, R. J. H. & Gibbs, P. J. 1998. Analysis of 16th Century Qazwını Manuscripts by Raman Microscopy and Remote Laser Raman Microscopy. J. Archaeol. Sci. 25: 621–629.

21

Weibring, P., Johansson, T,. Edner, H,. Svanberg, S,. Sundnér, B., Raimondi, V., Cecchi, G. & Pantani, L. 2001. Fluorescence lidar imaging of historical monuments. Appl. Opt. 40: 6111–6120. Woll, A.R., Mass, J., Bisulca, C., Huang, R., Bilderback, D.H., Gruner, S. & Gao N. 2006. Development of confocal X-ray fluorescence (XRF) microscopy at the Cornell high energy synchrotron source. Appl. Phys. A 83: 235–238. Xu, C. & Webb, W., W. 1996. Measurement of two-photon

excitation cross sections of molecular fluorophores with data from 690 to 1050 nm. J. Opt. Soc. Am. B; 481–491. Yang, M.L., Lu, C.W., Hsu, I.J. & Yang, C.C., 2004. The use of optical coherence tomography for monitoring the subsurface morphologies of archaic jades. Archaeometry 46: 171–182. Zumbusch, A., Holtom, G.R., & Xie X.S. 1999. Threedimensional vibrational imaging by coherent anti-Stokes Raman scattering. Phys. Rev. Lett. 82: 4142–4145.

22

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Fast spectral optical coherence tomography for monitoring of varnish ablation process M. Góra, P. Targowski & A. Kowalczyk Institute of Physics, Nicolaus Copernicus University, Toru´n, Poland

J. Marczak & A. Rycyk Institute of Optoelectronics, Military University of Technology, Warszawa, Poland

ABSTRACT: Optical Coherence Tomography (OCT) is a new, fast-growing technique for non-contact and non-destructive imaging of semi-transparent objects. The application of OCT to monitor varnish ablation with the fourth harmonic of Nd:YAG laser is presented. It is shown that OCT may be used for recovering parameters like ablation rate and for in situ tracking of the process. Frames from OCT tomographic movies demonstrating dynamic processes during ablation are also presented.

1

INTRODUCTION

investigation, but it seems to be a promising one in certain circumstances. At present, laser techniques are well established in several branches of art conservation, e.g. for cleaning of stone sculptures and architectonic details (Bartoli et al. 2006). The choice of laser source, essential to the success of the treatment, mostly depends on the absorption properties of the removed layer. Since all varnishes absorb well both infrared and ultraviolet radiation, lasers emitting in these two spectral ranges are utilised for this purpose. In the ultraviolet range, excimer lasers (λ = 157 nm, 193 nm and 248 nm) and Q-switched Nd:YAG laser (fourth harmonic) are mostly tested (CRAFT 2001, Bordalo et al. 2006). The infrared radiation at λ = 2936 nm emitted by the Er:YAG (Erbium-doped Yttrium-AluminiumGarnet, Er : Y3Al5 O12 ) laser is also considered as a very promising one (de Cruz et al. 1999, 2000). Since the varnish layer is very inhomogeneous in thickness and often its susceptibility to ablation varies from point to point, it is important to monitor online and control in situ the process. For this purpose Laser Induced Breakdown Spectroscopy (LIBS) (Anglos et al. 1997) is generally used. With this technique, traces of elements from the underlying paint layer occurring in the plasma ablation plume are detected (CRAFT 2001). This method is fast and sensitive, but a small damage to the underlying paint layer cannot be avoided. Additionally, it cannot be used if partial removal (to a certain fraction of the original thickness) of varnish is desirable. On the contrary, Optical Coherence Tomography (OCT), that enables non-contact in situ continuous direct measurement of varnish layer thickness, should

OCT provides an art conservator with information about the structure of varnish, glaze and paint layers in a non-contact and non-invasive way (Targowski et al. 2004, Liang et al. 2005, Targowski et al. 2006). In this low-coherence interferometric technique a narrow beam of light is sent to the object. The light scattered and/or back reflected on its structural elements carries information about their locations within the object. The application of the technique is thus limited to imaging of transparent and semitransparent layers. So far, it is well suited for varnish layers imaging, especially in old pictures where the layer is thick enough to clearly distinguish boundaries even with moderate resolution OCT systems. Examinations of modern varnish layers often require a higher axial resolution. A varnish layer plays an important, protective role to the underlying paint layers. As a surface layer it also enhances colours of the paint. Unfortunately, it changes very often with time due to the long-term exposure to light or external pollutants and must be removed and replaced by a new one. It should also be removed in case it covers old retouches and the conservation process must be repeated. The traditional conservation methods utilize mechanical and chemical means. The mechanical method is the safest, but unfortunately, it is also timeconsuming. Chemical methods based on toxic solvents are very difficult to control, because of possible penetration of the solvent into the paint layer and different (often very limited) solubility of the old varnishes. The third method – laser ablation – is still under

23

of 18 s. The latter is limited by the size of the operation memory of the computer utilised. In case of real time tracking, the data processing must be performed in real time and the image quality is limited by the performance of the computer. For a 1.8 GHz system it is possible to imagine 4 B-scans/s, each composed of 200 A-scans. Despite of this low line density, the quality of images is sufficient for current process control. In all OCT tomograms presented here, the intensity of the light scattered and/or reflected from the internal structures within the sample is coded in a greyscale. Dark shades of grey indicate a high scattering level of the probing light, while the light ones indicate low scattering levels. In all presented tomograms light is coming from the top, the interface between air and varnish is always the uppermost line, varnish-paint layer interface is below.

allow continuous control of the remaining varnish layer during its removal. In this contribution, the applicability of OCT for monitoring the ablation process induced by UV radiation is discussed. The dynamical processes like ablation and detachment of varnish layer are investigated. In former reports (Góra et al. 2006, Targowski et al. 2007a) the implementation of OCT to monitor varnish ablation by Er:YAG laser was described. In the present study, the fourth harmonic of Nd:YAG laser was utilised. 2 2.1

INSTRUMENTATION Nd:YAG laser system

The fourth harmonic of Nd:YAG laser (λ = 266 nm) is strongly absorbed by the varnish and the penetration of light into the irradiated medium is very small. For the purpose of this study a ReNOVALaser 5 was chosen. It is a Q-switched system, generating pulses of duration 12 ns and output energy up to 120 mJ. The repetition rate of pulses was adjusted from 2 Hz to 10 Hz. During experiments, the output energy was kept between 10 and 55 mJ. Laser radiation was focused on the sample with a quartz lens (f = 50 cm) to achieve fluences from 1 to 7 J/cm2 . 2.2

3

EXPERIMENTS

3.1 Samples The determination of the ablation threshold and the studies of the process of laser ablation were performed with the specially prepared samples. For the purposes of the present study three different varnishes were tested: Maimeri Dammar Matt varnish, Talens Acrylic varnish (matt) 115 and Talens Picture varnish (ketone, glossy) 002. All samples were prepared with a cardboard support. For the first sample the support was covered with an acrylic paint, the others were made with an oil paint. When the paint layer dried out, all samples were covered with the respective varnish. Just before testing, all samples were sprayed with a thin layer of matt Schminke varnish, to reduce mirror reflections.

SOCT instrument

The Spectral OCT system used to obtain the data presented in this contribution is a prototype instrument constructed in our laboratory. It is described in detail elsewhere in this volume (Targowski et al. 2007b). Briefly, to illuminate the sample a superluminescent diode (SLD) with central wavelength of 840 nm and bandwidth of 50 nm, emitting high spatial but low temporal coherence light is used. The light is launched into a fibre optic Michelson interferometer. It consists of reference and object arm. In the latter, light is scanned across the object by galvoscanners. The optical power of the beam incident at the surface of the object is 800 µW. The in-depth (axial) resolution of the system is limited mostly by the bandwidth of the light source. In the system used in this study it is estimated to be 6 µm in varnish. The transverse resolution depends on the optical properties of the system and is kept below 20 µm. For the purpose of the study, the acquisition speed of the system is a crucial parameter. The exposure time per A-scan is 40 µs and the acquisition rate is 20000 Ascans/s. The instrument may work in two modes: registration and real time tracking. In the former, raw data are quickly stored and processed afterwards to create an OCT movie. To achieve high quality of imaging, every cross-sectional image (B-scan) must be composed at least of 1200 A-scans. This leads to a recording rate of 16 B-scans/s and a recording time

3.2 Experimental setup To monitor the varnish layer removal, the OCT system was combined with the Nd:YAG laser (Fig. 1). Both instruments were focused to the same spot, with the sample surface perpendicular to the OCT◦scanning beam. The laser beam was tilted at about 40 . For each sample, two kinds of experiments were performed. First, the ablation rate under different fluences and pulse repetition rates was determined by examining ablation craters created by the laser beam focused at the varnish layer. During this experiment the sample was held in a steady position and irradiated with a series of laser pulses. The energy of these pulses was determined by means of a power-meter. Simultaneously, the OCT cross-sections were recorded in the place of the ablation crater, as described in the preceding section. Finally, the position of the bottom of the ablation crater was recovered from these images as

24

to the OCT system

Nd:YAG Laser

Power meter Figure 1. The experiment setup, a top view.

Figure 3. Relative position of the bottom of ablation crater in Talens Acrylic varnish after cumulative number of laser pulses, fluence = 6.7 J/cm2 .

Figure 2. a) Determination of the position of the bottom of crater. b) Determination of the surface of the ablation crater.

shown in Figure 2a. To estimate the fluence, the diameter of the ablation crater was determined from the surface image (200 × 200 pixels, 4 × 4 mm) obtained with the same OCT system (Fig. 2b). Once the ablation rate and ablation threshold were determined, progressive ablation of the varnish layer was performed. For this purpose, the sample was shifted continuously with constant velocity. In all these experiments the diameter of the laser beam spot was approximately 2 mm and the scanning speed was 0.4 mm/s. Therefore, for a repetition rate of 2 Hz, each point on the sample was irradiated 10 times. If repetition rate is increased, the number of accumulated pulses increased accordingly. The ablation process was registered with the OCT device by recording the OCT movie as described in the previous section.

4

Figure 4. Frames from OCT movie registered during ablation of Acrylic varnish. Numbers in parenthesis show the amount of laser pulses accumulated. Arrow points to the exfoliated layer.

Results shown in Figure 3 were measured on frames extracted from the OCT movie, registered with 16 frames per second as described in Section 2.2. Selected frames from this movie are presented in Figure 4. Normal ablation process is illustrated in Figure 4a and 4b. When the remaining varnish layer becomes thin the exfoliation process starts – a separate layer is visible just above the paint layer (Fig. 4c, arrow). Then, immediately after the next pulse, the whole remaining varnish is exfoliated (Fig. 4d) and removed by the two following pulses (Fig. 4e, f).

RESULTS

4.1 Talens Acrylic varnish In this case, the ablation rate was measured for a fluence of 6.7 J/cm2 , well above the ablation threshold. It is worth noting that the varnish layer for a similar sample was only exfoliated rather than ablated when it was treated with an IR laser (λ = 2.936 µm, fluence of 2.2 J/cm2 ) (Targowski et al. 2007a). This was probably due to insufficient absorption of IR laser radiation by the varnish layer. On the contrary, in case of UV treatment, the layer is effectively ablated and the rate may be estimated to 4.3 µm/pulse (Fig. 3).

4.2 Talens Ketone varnish The results obtained for Talens Ketone varnish are shown in Figure 5. As in the previous case, data were collected with repetition rate 2 Hz. The ablation rate may be estimated to 6 µm/pulse.

25

Figure 5. Relative position of the bottom of ablation crater in Talens Ketone varnish after cumulative number of laser pulses, fluence = 3.6 J/cm2 . Figure 7. Frames from the OCT movies, showing tomograms of the Maimeri Dammar varnish layer ablated with laser repetition rate (a) 2 Hz and (b) 8 Hz. Other laser parameters were the same. Laser pulses were applied at the centre of the image and during the process the sample was moved to the right.

phenomenon was observed. During ablation with low repetition rate, an additional volumetric damage to the varnish layer occurred, well visible in Figure 7a. This additional scattering diminishes the transparency of the remaining layer and thus should be avoided. Similar effects were observed with OCT for IR ablation as well (Góra et al. 2006). When the repetition rate was increased to 8 Hz (Fig. 7b) this effect was not present. Interestingly, attentive inspection of frames from this movie revealed that, at the beginning of the laser treatment, when the number of pulses accumulated was small, the same destruction as for 2 Hz was present. But after accumulation of the following pulses, the trace of internal damage vanished, probably by local melting/resolidifying process.

Figure 6. Relative position of the bottom of ablation crater in Maimeri Dammar varnish after cumulative number of laser pulses obtained for 2 and 4 Hz pulse repetition rate. Fluences were 4.5 J/cm2 and 5 J/cm2 respectively.

4.3

Maimeri Dammar varnish

If a layer is significantly resistive to the laser treatment it may be convenient to increase the repetition rate. In Figure 6, the ablation rate of the Dammar varnish is presented for 2 different repetition rates. In both cases, the pulse energy was the same. Ablation rates were 3.2 µm/pulse and 3.7 µm/pulse respectively. Taking into account the slightly higher fluence in the second case, one may assume that the ablation rate does not depend on the repetition rate of the laser and indeed the thickness of ablated layer may be conveniently controlled this way. After determination of the ablation rates, the experiments simulating a cleaning process were performed (Fig. 7). As it can be seen, increasing the repetition rate allows effective control of the ablation depth. However, for this specific varnish another interesting

5

CONCLUSIONS

Optical Coherence Tomography can be conveniently used for imaging transparent varnish layers. This technique permits fast and non-invasive direct assessment of the condition (e. g. transparency) of the remaining layer and direct determination of its thickness. In this contribution, we present an application of the OCT to the monitoring of the varnish ablation process. First, the OCT technique may be used for this purpose as a tool for adjustment of the process parameters, like ablation rate. Second, fast modalities of OCT, like Spectral and Sweep Source OCT may be used for in

26

CRAFT Final Report project ENV4-CT98-0787. 2001. Advanced workstation for controlled laser cleaning of artworks. http://www.art-innovation.nl/. deCruz, A., Hauger, S.A. & Wolbarsht, M.L. 1999. The role of lasers in fine arts conservation and restoration. Optics and Photonics News 10: 36–40. deCruz, A. Wolbarsht, M.L. & Hauger, S.A. 2000. Laser removal of contaminants from painted surfaces, Journal of Cultural Heritage 1: 173–180. Góra, M., Targowski, P., Rycyk, A. & Marczak, J. 2006. Varnish ablation control by Optical Coherence Tomography. Laser Chemistry doi:10.1155/2006/10647, http://www.hindawi.com/journals/lc/. Liang, H., Cid, M., Cucu, R., Dobre, G., Podoleanu, A., Pedro, J. & Saunders, D. 2005. En-face optical coherence tomography–a novel application of non-invasive imaging to art conservation. Optics Express 13: 6133–6144. Targowski, P., Rouba, B., Wojtkowski, M & Kowalczyk, A. 2004. The application of optical coherence tomography to non-destructive examination of museum objects. Studies in conservation 49: 107–114. Targowski, P., Góra, M. & Wojtkowski, M. 2006. Optical Coherence Tomography for Artwork Diagnostics. Laser Chemistry doi:10.1155/2006/35373 http://www.hindawi. com/journals/lc/. Targowski, P., Marczak, J., Góra, M., Rycyk,A. & Kowalczyk, A. 2007a. Optical Coherence Tomography for Varnish Ablation Monitoring. Proc. of SPIE 6618: 661803-1– 661803-7. Targowski, P., Góra, M., Bajraszewski, T., Szkulmowski, M., Wojtkowski, M., Kowalczyk, A., Rouba, B., Tymi´nskaWidmer L. & Iwanicka M. 2007b. Optical coherence tomography for structural imaging of artworks. In this volume.

situ tracking of the process. Specifically, it is possible to estimate in real time the thickness of the ablated and remaining layers. The OCT method also allows evaluation of the quality of the remaining varnish layer. Collecting images every 60 ms allows monitoring of fast, dynamical processes like creation of exfoliations and melting resolidification inside the remaining layer. Since properties of paintings vary significantly, the adjustment of ablation conditions must be performed separately for every object. Moreover, in certain cases the choice of the wavelength and pulse duration may not be obvious. Therefore, the OCT monitoring may increase significantly the safety and efficiency of using lasers for varnish ablation and contribute this way to the propagation of this technique in future. REFERENCES Anglos, D., Couris, S. & Fotakis, C. 1997. Laser Diagnostics of Painted Artworks: Laser-Induced Breakdown Spectroscopy in Pigment Identification. Applied Spectroscopy 51: 1025–1030. Bartoli, L., Siano, S., Salimbeni, R., Pouli, P. & Fotakis, C. 2006. Characterization of laser cleaning of pollution encrustation on stonework by Nd:YAG lasers with different pulse duration, Laser Chemistry doi:10.1155/2006/81750, http://www.hindawi.com/ journals/lc/. Bordalo, R., Morais, P.J., Gouveia, H. & Young Ch. 2006. Laser Cleaning of Easel Paintings: An Overview. Laser Chemistry doi:10.1155/2006/90279, http://www.hindawi. com/journals/lc/.

27

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Study of laccaic acid and other natural anthraquinone dyes by Surface-Enhanced Raman Scattering spectroscopy M.V. Cañamares & M. Leona The Metropolitan Museum of Art, New York, NY, USA

ABSTRACT: The red dyes of natural origin, such as lac dye, cochineal and madder, are of great importance in the history of early textiles and lake pigments. The main colouring substances of these dyes (laccaic acid, carminic acid and alizarin, respectively) are similar in chemical composition: all of them are based on a hydroxyanthraquinonic structure. In this work, we present the Surface Enhanced Raman Scattering spectroscopy (SERS) study of laccaic acid at 514.5 nm. The SERS spectra were obtained at different pH values and on differently prepared Ag nanoparticles, by means of chemical reduction of silver nitrate with tri-sodium citrate and hydroxylamine hydrochloride. The aim of this work is to find the best experimental conditions for the vibrational study and detection of lac dye. A comparison between the SERS spectra of laccaic acid and other anthraquinone dyes (carminic acid and alizarin) was also carried out.

1

INTRODUCTION

Rubiaceae family, common to Europe and the Middle East (Scheweppe & Winter 1998). Madder dye was extensively used in Asia for dyeing materials since ancient times. In Egypt, however, it was employed in the 16th century for dyeing. Alizarin (AZ, Fig. 1c), the simplest anthraquinone dye, is the main colouring component of madder. The high fluorescence of these anthraquinone dyes seriously limits the application of normal Raman spectroscopy to study these molecules. However, FT-Raman and SERS can be successfully used to characterize this dye in aqueous media. The applicability of Fourier Transform-Raman spectroscopy to the study of LA is limited to the pure dye in the solid phase and in highly concentrated solutions, as the Raman effect is very weak at near-infrared excitation. SERS, on the other hand, can be successfully used to study very diluted aqueous solutions, and has the additional advantage of quenching the fluorescence due to the presence of a metal surface (Creighton 1988, Moskovits 1985). Recent articles have reported SERS identification and/or adsorption behaviour of several anthraquinone dyes (Cañamares et al. 2004, 2006, Chen et al. 2006a, b, Shadi et al. 2006, Whitney et al. 2006). In this work we present a SERS study of LA on silver nanoparticles prepared by chemical reductions with citrate and hydroxylamine as well as at different pH values. The aim of this study was to determine the best experimental conditions for detecting low concentrations of LA in samples from artifacts, such

The red dyes of natural origin are of great importance in the history of early textiles and lake pigments. Their identification on ancient and historical textiles can be crucial to answer questions of dating, provenance, types of degradation, original colour and appearance of the artwork (Hofenk de Graaff 2004). The main colouring matters of the most important red natural dyes, such as lac dye, cochineal and madder, are based on a hydroxyanthraquinonic chromophore structure. Lac dye is a resinous secretion developed by the insect Laccifer lacca, native to Bengal, Siam, southern India and the Moluccas. This dye was used in India and Japan since antiquity, and it seems to have been extensively used in European dyeing practice only in the late 18th century. The main colouring components of lac dye are laccaic acid A (LA, Fig. 1a) (71–96%) and B (0–20%) (Hofenk de Graaff 2004). Cochineal, another dye of scale origin, is extracted from the female beetles of Dactylopius coccus L. Costa, native to Mexico, Central America and regions of South America. Aztecs used this dye for dyeing and painting. It was brought to Europe in the 16th century following the Spanish conquest. It superseded kermes, the European red dye (Scheweppe & RoosenRunge 1986). The main colouring matter is carminic acid (CA, Fig. 1b), whose structure is similar to that of LA, as can be seen in Figure 1. Madder, a vegetal dye, is extracted from the roots of Rubia tinctorum L. and various other plants of the

29

CH2CH2NHCOCH3 OH

O

OH COOH CO OH

OH HO

OH OH

HO HO

O

OH

O

OH

O

CH3

OH

COO H OH OH

HO

O

OH

(a)

(b)

O

O

(c)

Figure 1. Structure of laccaic acid (a), carminic acid (b) and alizarin (c).

conditions. The colloidal suspension containing the dye was placed in a 1 cm-path cell.

as textiles or paintings. A comparison of the SERS spectra of LA, CA and AZ on the silver nanoparticles was also carried out.

3 2 2.1

EXPERIMENTAL

3.1

Materials

RESULTS AND DISCUSSION SERS spectra of laccaic acid with different pH and Ag nanoparticles

The SERS spectra of LA obtained at different pH values on SH and SC nanoparticles at 514.5 nm excitation are shown in Figure 2. Excitation at this wavelength is expected to result in the enhancement of the bands corresponding to the acidic species of LA (result not shown). The SERS spectrum obtained at pH 5 on SH colloid was too intense, therefore, the power had to be reduced to half of the value normally used (∼1 mW). For this reason, the equivalent global intensity of the spectra at pH 5 is approximately twice as large as that shown. A decrease in SERS intensity results from the pH changing from acidic to alkaline, the decrease in the resonance Raman effect and the increase of the electrostatic repulsion between LA and the Ag surface. As seen in Figure 2, the decrease in SERS intensity of LA upon increasing the pH of SC nanoparticles is higher than increasing that of the SH colloid. This is due to the higher negative zeta potential of the SC nanoparticle surface (Munro et al. 1995). The spectra undergo significant changes as the pH increased from 5 to 7. In the case of the SH colloids (Fig. 2c) there is a large increase in the intensity of the bands at 1532, 1344, 1291 and 1191 cm−1 and a shift of the bands at 1633, 1579, 1462 and 1227 to 1620, 1569, 1454 and 1233 cm−1 . On the other hand, when using SC nanoparticles (see Fig. 2d), an enhancement of the bands at 1534, 1350, 1300 and 1227 cm−1 takes place, together with a shift of the bands at 1670, 1585, 1462, and 1227 to 1659, 1571, 1453 and 1233 cm−1 . These observed differences can be attributed to the deprotonation of the hydroxyl and carboxylic groups. When an ionization process occurs, the charge distribution in the molecule is modified, leading to the changes observed in the relative intensities of the bands. As the chemical structure of LA, CA, and AZ is based on an anthraquinone skeleton, the first thought

LA was purchased from Tokyo Kaisei Co. (98%), CA from Sigma (96%) and AZ from Acros (97%). Stock solutions of the dyes (10−3 M) were prepared in water. All the reagents employed were of analytical grade and purchased from Sigma and Merck. The aqueous solutions were prepared by using 18 M ultra pure water (Millipore MilliQ). Silver colloids were prepared by reduction of silver nitrate with tri-sodium citrate dihydrate (SC) and hydroxylamine hydrochloride (SH) according to previously described procedures (Cañamares et al. 2005). Before adding the dye solutions, the citrate and hydroxylamine Ag colloids were activated by mixing 0.5 ml of each colloid with 12 and 20 µl of 0.5 M potassium nitrate solution, respectively. This activation is a prerequisite for observing SERS spectra at higher intensities, as it was demonstrated (Sanchez-Cortes et al. 1994, 1995). After activating the colloids, 55 µl of the dyes aqueous solution were added to 0.5 ml of each Ag colloid in order to reach a concentration of 10−4 M. Nitric acid and sodium hydroxide were employed to control the pH.

2.2 Instrumentation In a micro-Raman Renishaw RM1000 instrument, the SERS spectra were recorded at 514.5 nm with an Ar+ laser and a 50X objective. The maximum laser power at the sample was 1 mW. The resolution was set at 4 cm−1 and the geometry of micro-Raman measurements was 180◦ . Each spectrum was obtained by recording a single 30 s scan. The measurements of LA and CA were conducted with a micro-Raman setup where the laser beam was focused on a drop of the dye-nanoparticle suspension deposited on the surface of a glass slide. On the other hand, the SERS spectra of AZ were recorded in macro

30

1191

HO

-

O

O Ag

(d) SC pH 7

1300

1194

(c) SH pH 7

1427

1620

Figure 3. Adsorption mechanism of AZ on Ag surfaces.

(f) SC pH 11

1300

1087

x 1/2

(c) SH pH 7

(b) SC pH 5 1076

x 1/4 (a) SH pH 5 454

is to conclude that all of these dyes interact with the Ag surface in the same way. This interaction would be through the keto and hydroxyl groups, as shown in Figure 3 (Cañamares et al. 2004, 2006). The large increase of the band at 1344 cm−1 , assigned to the OH group adjacent to the C = O, supports this suggestion. However, LA could present an electrostatic interaction with the Cl− ions surrounding the silver nanoparticles by means of the NH+ of the amide group. This mechanism gains support from the vibration at 1620 cm−1 observed at neutral pH, which can be assigned to the amide group (Socrates 2001). On increasing the pH to 11 (Figs. 2e, f ) a decrease of the relative intensity of the bands at 1569, 1454, 1233 and 1010 cm−1 is observed. We suggest that these changes are due to another deprotonation process of LA. The downward shift undergone by higher frequency bands and the upward shift of the lower ones, originated as a result of the higher electronic delocalization in the molecule, supports this fact. The intensity of the SERS spectra obtained with the SH colloid is higher than with SC at every pH value. Thus, we can conclude that the SH colloid should be preferably used for the analysis of textiles suspected to contain lac dye. The existence of chloride ions on

1449

1640 1645

Figure 2. SERS spectra of LA (10−4 M) on SC and SH colloids at acidic, neutral and alkaline pH. Excitation at 514.5 nm. Spectrum (a) was obtained using half the laser power of the other ones.

(d) SC pH 7

1227

500

Wavenumber /cm -1

1327

x 1/2

1457

1000

SERS intensity

(a) SH pH 5

412

1011

(b) SC pH 5

454

1227

(e) SH pH 11

(e) SH pH 11

1100

1453

1347 1403 1350 1344 1300 1297 1291 1191 1194 1191 1101 1100 1013

1454

1368 1288 1462 1333

1625

1532

1606 1534

1659

1620 1569 1670 1585 1633 1579

SERS intensity

1500

O (f) SC pH 11

1500

1000

x 1/4

500

Wavenumber / cm −1 Figure 4. SERS spectra of CA (10−4 M) on SC and SH colloids at acidic, neutral and alkaline pH. Excitation at 514.5 nm. Spectra (a), (b) and (d) were obtained using 1/4, 1/4 and 1/2 of the laser power of the other ones, respectively.

the surface of the SH colloid would be responsible for the larger enhancement of the Raman bands if an electrostatic interaction between the Cl− and the NH+ of the amide group takes place. In the case of SC nanoparticles, the electrostatic interaction would happen between the citrate ions and the NH+ . This interaction is weaker than the Cl− -NH+ , so the SERS enhancement obtained would be weaker as well. 3.2 SERS spectra of carminic acid on different Ag nanoparticles SERS spectra of CA at pH 5 (on both Ag colloids, Figs. 4a, b) and 7 (on CT nanoparticles, Fig. 4d)

31

1422

1322 1290

1601

900

1047

1160

1500

1000

343

833

(a) SH pH 6.5 1012

1185

1287

1324

1452 1627 1557

SERS intensity

(b) SC pH 6.5

500

Wavenumber / cm−1 Figure 5. SERS spectra of AZ (10−4 M) on SH and SC nanoparticles at pH 6.5. Excitation at 514.5 nm.

Figure 6. Comparison of the SERS spectra on SH nanoparticles of 10−4 M solution of CA and LA at various pH. Spectrum (b) was obtained using 1/2 of the laser power of the other ones.

were too intense. As a result, laser power had to be reduced to 1/4 and 1/2 of the value normally used (∼1 mW), respectively. For this reason, the equivalent global intensity of those spectra is approximately 4 and 2 times larger than shown. The study to determine the best type of Ag nanoparticles to analyze CA in low concentration (10−4 M) by SERS is shown in Figure 4. The 514.5 nm excitation line was used. The intensity of the Raman bands slightly decreases when going from acidic (Figs. 4a, b) to neutral pH (Figs. 4c, d), while the SERS enhancement remains practically unchanged when going to an alkaline pH (Figs. 4e, f ). A high fluorescence background is seen in Figure 4, even at neutral pH. This shows that CA is weakly adsorbed on the Ag surface. In contrast to LA, the SERS spectra of CA show that the type of Ag nanoparticles used in the measurements is not especially significant in the detection of CA, as it interacts very weakly with the SERS substrate.

were obtained at acidic (Figs. 5a, b), neutral (Figs. 5c, d) and alkaline (Figs. 5e, f) pH on the SH colloid. The SERS spectra of LA are more intense than those of the CA in the pH range studied, which shows that the interaction of LA with the Ag surface is stronger. This is due to the existence of an electrostatic interaction between the Cl− ions that surrounds the SH surface and the NH of the amide group. Such an interaction is much stronger than the adsorption that occurs between CA and the silver nanoparticles. Consequently, if a mixture of LA and CA is found in a real sample, only LA features will be observed in the SERS spectra. Furthermore, there are some interesting differences between the SERS spectra pattern of LA and CA. The band at 1344 cm−1 shows a large increase in the LA spectrum at pH 7 and 11; yet is not present in the CA spectrum; the band at 1454 cm−1 decreases in the LA spectrum less than in the CA spectrum at pH 11; and finally, the three bands at 1010, 1058 and 1100 cm−1 are unique characteristics of LA.These features permit an easy differentiation of CA and LA by SERS, despite their similar chemical structures (Fig. 1). Figure 6 shows the SERS spectra of AZ obtained in macro conditions. Because the spectra of the other

3.3 Comparison of SERS spectra of laccaic acid with other anthraquinone dyes Figure 5 shows a comparison of the SERS spectra of various 10−4 M aqueous solutions of LA and CA employing the 514.5 nm excitation line. The spectra

32

acknowledged. We are also indebted to the National Institute of Justice (Department of Justice Award #2006-DN-BX-K034) and the City University Collaborative Incentive program (#80209).

anthraquinone dyes were measured in micro conditions, these spectra of AZ are presented in a separate graph. As it was seen with CA, no significant differences appear between the enhancement produced by SC and SH nanoparticles. This fact could be related to the type of interaction that takes place between these dyes and the surface. Thus, as the enhancement of the LA bands produced by the SH colloid is much higher than that gathered from SC, the electrostatic interaction proposed for LA can be supported. However, the different SERS profiles obtained show that the interaction of AZ with the Ag surface is much stronger than the CA one. 4

REFERENCES Cañamares, M. V., Garcia-Ramos, J. V., Domingo, C. & Sanchez-Cortes, S. 2004. Journal of Raman Spectroscopy 35: 921–927. Cañamares, M. V., Garcia-Ramos, J. V., Gomez-Varga, J. D., Domingo, C. & Sanchez-Cortes S. 2005. Langmuir 21: 8546–8553. Cañamares, M. V., Garcia-Ramos, J. V., Domingo, C. & Sanchez-Cortes, S. 2006. Vibrational Spectroscopy 40: 161–167. Chen, K., Leona, M., Vo-Dinh, K. C., Yan, F., Wabuyele, M. B. & Vo-Dinh, T. 2006a. Journal of Raman Spectroscopy 37: 520–527. Chen, K., Vo-Dinh, K. C., Yan, F., Wabuyele, M. B. & Vo-Dinh, T. 2006b. Analytica Chimica Acta 569: 234–237. Creighton, J. A. 1988. Selection Rules for Surface-Enhanced Raman Spectroscopy. In R. J. H. Clark (ed.), Spectroscopy of Surfaces. Chichester: Wiley. Hofenk de Graaff, J. H. 2004. The Colorful Past. Origins, Chemistry and Identification of Natural Dyestuffs. Riggisberg, London: Abegg-Stiftung and Archetype Publications. Moskovits, M. 1985. Review of Modern Physics 57: 783–826. Munro, C. H., Smith, W. E., Garner, M., Clarkson, J. & White, P. C. 1995 Langmuir 11: 3712–3720. Sanchez-Cortes, S., Garcia-Ramos, J. V. & Morcillo, G. 1994. Journal of Colloid and Interface Science 167: 428–436. Sanchez-Cortes, S, Garcia-Ramos, J. V., Morcillo, G. & Tinti, A. 1995. Journal of Colloid and Interface Science 175: 358–368. Scheweppe, H. & Roosen-Runge, H. 1986. Carmine. In R.L. Feller (ed.), Artists’ Pigments. A Handbook of their History and Characteristics, vol 1: 255–283. Cambridge: Cambridge University Press. Scheweppe, H. & Winter, J. 1998. Madder and Alizarin. In E.W. FitzHugh (ed.), Artists’ Pigments. A Handbook of their History and Characteristics, vol 3: 109–142. Oxford: Oxford University Press. Shadi, I. T., Chowdhry, B. Z., Snowden, M. J. & Withnall, R. 2004. Journal of Raman Spectroscopy 35: 800–807. Socrates, G. 2001. Infrared and Raman Characteristic Group Frequencies:Tables and Charts. Chichester: John Wiley & Sons, Ltd. Whitney, A. V., Van Duyne, R. P. & Casadio, F. 2006. Journal of Raman Spectroscopy 37: 993–1002.

CONCLUSIONS

A Surface-Enhanced Raman Scattering spectroscopy (SERS) study of lac dye was carried out using Ag nanoparticles. The surface enhancement of lac dye is strongly dependent on the pH and the nature of the nanoparticles used. When exciting at 514.5 nm, SERS spectra obtained at a pH above 5 show a decrease in the SERS enhancement. This is due to the decrease of the resonance Raman effect and the increase of the electrostatic repulsion between LA and the nanoparticles. Upon transitioning from acid to alkaline pH, successive deprotonation processes occur on the molecule and lead to changes in the charge distribution. However, when the SH colloid is employed, the enhancement of the SERS spectra of LA is higher overall than with SC nanoparticles. This is true to all pH values studied. Thus, the SH colloid is better than the SC one for detecting low concentration solutions of the dye. Lac dye interacts with Ag nanoparticles more strongly than carminic acid in the pH 3–11 interval. As a result, only laccaic acid will be detected when using SERS spectroscopy if a mixture of both red dyes is found in a textile. ACKNOWLEDGEMENTS Grants from the NSF (DMR-0526926), the Andrew W. Mellon Foundation and the David H. Koch Family Foundation in support of scientific research at The Metropolitan Museum of Art are gratefully

33

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Potential of THz-Time Domain Spectroscopy in object inspection for restoration M. Panzner, Th. Grosse, S. Liese, U. Klotzbach & E. Beyer Fraunhofer Institute Material and Beam Technology, Dresden, Germany

M. Theuer Fraunhofer Institute for Physical Measurement Techniques, Kaiserslautern, Germany

W. Köhler Labor Köhler, Potsdam, Germany

H. Leitner Hochschule für bildende Künste, Dresden, Germany

ABSTRACT: Teraherz-Time Domain Spectroscopy enables a parallel determination of tomographic and spectroscopic information. The electromagnetic pulses used for this method have pulse durations of approximately one picosecond. Thus, runtime measurements can be used to investigate the layered structure of materials with accuracy higher than geo-radar methods. While passing the material, the pulse shape changes because of dispersion. Rotational and vibrational resonances of polar molecular groups appearing as peaks in the Fourier spectrum serve as a useful means of identification of substances, as it was demonstrated with Lindane and Pentachlorophenol. THz-images can be made by scanning the object step by step as well. The method is tested on an oil painting. E(t) – and E(ω) – movies were calculated from the data at each point of the measuring grid. This method allows hidden paintings to be revealed. 1

INTRODUCTION

1.1 Teraherz-Technology In the last years, Teraherz (THz)-technology has developed to become a powerful tool for non-destructive material investigation (Bäumer 2002, Hoffmann et al. 2005, Mansner et al. 2007). Tomographic information as well as spectroscopic information can be deduced from experimental results of THz-Time Domain Spectroscopy (THz-TDS) by the runtime and spectrum of electromagnetic picosecond THz pulses. Such pulses can be generated by high acceleration of free electrons in a semiconductor. The free electrons are produced by photoelectric effect due to femtosecond laser irradiation. A voltage of about 100 V applied to electrodes on the semiconductor with a small distance in the micron range generates a strong electric field in the case of an Auston Switch emitter (Dorney et al. 2000, Shur 2005). For the acceleration of electrons, internal fields at the semiconductor surface (photoDember effect) can be used as well. Figure 1 shows a scheme of an experimental setup for THz-TDS. The spectrum of the picosecond THz-pulse (Fig. 2a) can be calculated by Fourier transformation (Fig. 2b).

Figure 1. Experimental setup used for THz-Time Domain Spectroscopy.

The width of the spectra is in the range of 3 to 10 THz, depending on the applied emitters and detectors. Figure 3 shows the comparison of a detected picosecond THz pulse with and without sample in

35

the beam line. Three characteristic differences can be seen: – delay of the sample pulse because of different signal velocities (real part of the material refraction index); – decay of the peak amplitude because of losses (imaginary part of the material refraction index); – changes in the pulse shape (broadening of the pulse, modulations) caused by the dispersion within the material and resonances of polar molecular groups. A wide variety of materials has been investigated by means of THz radiation (Roskos et al. 2004, Gorenflo et al. 2005, Smith 2005). The work in this field is fostered by promising applications in industry and specific problems of safety engineering. THz radiation is transmitted without important attenuation by many non-polar substances. This allows investigation of objects covered by such materials. Our work is focused on the utilization of these potentials for the investigation of artworks. 1.2 Detection of fungicides and insecticides In the past many objects of wooden and textile art were contaminated by such substances to avoid the invasion by different pests. Today the poisonousness of these substances is a well known issue and the objects have to be decontaminated. To apply a suitable technique, the type of substance has to be determined. For this identification, non-destructive methods are needed. Because of the low photon energy and the low power of the THzradiation used for THz-TDS, this method could solve the identification problem. This contribution is presenting the first THz-TDS measurements on Lindane and Pentachlorophenol (PCP).

Figure 2. (a) Electrical field of a THz picosecond pulse as a function of time. (b) Fourier transformation of the same electrical field.

1.3 Visualization of paintings by THz radiation Paintings and murals are valuable genuine pieces of the cultural heritage of mankind and efforts are devoted to its conservation (Möhlenkamp 2002, Pursche 2002). Up to now, revealing the images hidden below without damaging the likewise valuable top layers has been successfully done only in exceptional cases (Stewart 1991, Leitner 1994). Infrared light and X-rays are usually applied for this purpose (Humphries 2001, Mairinger 2003). The success of these techniques is restricted by the inherent limits to penetration and depth resolution. The use of THz radiation could solve this problem because of its interesting properties. As a first step, the visualization of a painting by THz-TDS in the transmission mode is presented. 2

EXPERIMENTAL

The experiments were carried out with the setup shown in the scheme of Figure 1. The picosecond THzpulses were generated by an In As-semiconductor

Figure 3. THz-pulses, before and after penetrating a sample of Lindane. E is the decay of the electrical field peak and t the delay of the pulse through the sample.

36

(photo-Dember effect) that was fired by a Ti: Sapphire laser “MaiTai” from Spectra Physics (pulse duration: 100 fs, repetition rate: 80 MHz, average power: 1.5 W). The THz-pulses were detected by an Auston Switch. The THz-beam was focused on the samples with metal mirrors to a diameter of about 1.5 mm. The fungicide Pentachlorophenol (PCP) – C6 Cl5 OH – and the insecticide Lindane – C6 H6 Cl6 – were available as powders in plastic bags. These bags were placed between two Teflon plates to homogenise the sample surface and to compress the powder. The sample used for the scanning experiments was a piece of an oil painting on canvas. During the measurement this segment (10 × 10 cm2 ) was fixed to a frame. The sample was systematically rastered through the focus of the THz beam in steps of 2.2 mm in X and Z direction. At each point, an E(t)-curve was measured by the described equipment and stored in a single file. So far, 2116 files containing 12801 data pairs each had to be considered for the calculation of graphics. 3

Figure 4. Structural formulae of the insecticide Lindane (left) and the fungicide Pentachlorophenol (right).

RESULTS

3.1

Spectroscopic investigation of Lindane and Pentachlorophenol

The molecules of both Lindane and Pentachlorophenol show aromatic structures (Fig. 4). The THz spectra were calculated by Fourier transformation of the E(t) data. The E(ω) curves of Lindane and Pentachlorophenol exhibit different minima caused by molecular resonances (Figs. 5, 6). This allows a fingerprint-like identification of these substances. Additionally, the spectra of the reference pulse (without sample) and that with the empty sample container are shown. 3.2

Figure 5. THz-spectrum of Lindane.

Investigation of paintings

To visualize the image of the painting, the E(t)-curves of each measured spot were used to calculate pictures containing the THz information. A complex Matlab routine was programmed to generate images of: – – – –

Figure 6. THz-spectrum of Pentachlorophenol.

the value of the electrical field in the maximum; the value of the electrical field in the minimum; the difference of maximum and minimum; the pulse delay with regard to the reference pulse.

at certain intervals of time and frequency. Figure 7 shows such characteristic frames of movie sequences for the E(t)- and the E(ω)-mode. The painting is only faintly reproduced by the THz pictures. The blurring is due to variations of thickness, complex refraction index, modulation by resonances, etc, on the THz transmission signals from the layer stack of oil painting. This includes effects of the surface and regions below the surface as well as the varying characteristics of the canvas. For solving the related problems, further investigations on samples with well defined materials

Additionally the program allows: – the calculation of an E(t)-movie using all values of E(t) at the measuring grid (x, y) as a function of time and, – the generation of an E(ω)-movie by Fourier transformation of E(t) at each point (x,y) of the scan. Particularly, by the movie procedure, image information could be visualized through structures seen

37

spot distance. The homogeneity as well as the resolution of the pictures could be improved by a finer measuring grid, which is, however, limited by the large focal diameter of 1–2 mm. A reduction of the focal diameter is limited by the rather long wavelength of THz radiation.

4

CONCLUSIONS

The THz-TDS measurements of the fungicide PCP and the insecticide Lindane demonstrate the existence of characteristic resonances within the THz spectral range. This allows a fingerprint-like identification of these substances. The present preliminary tests of the technique were done by penetrating 1–2 mm through the substance. For identification of biocides on contaminated art objects, experience has to be gathered with measurements on samples with lower concentration. The sensitivity of this method will be determined in forthcoming stages of our research. As an advantage of the THz-TDS technique, textile and thin wooden objects can transmit THz radiation without deterioration, which enables information to be obtained from inside without any damage. One of the next challenges of our work is the detection of different substances in the reflection mode for the case of non-THz-transparent objects like thicker wooden objects. The THz images of the oil painting represent partial information of the real image. This is promising to extend this method to reflective measurements. Hence, our ongoing investigations are focused on scanning of paintings with THz-TDS in the reflective mode. In case of success, “visualization of hidden wall paintings” could become real. A further advantage of the reflection method results from possible runtime measurements of THz-pulses. Thus, tomographic information, e.g. layer thickness, could also be deduced from these investigations. The spectroscopic analysis of the E(ω)-movie offers the possibility to display the distribution of substances. To get the frame showing the distribution of a certain substance, one has to stop the movie at the minimum of the spectrum which is related to the substance. This THz analysis method is not restricted to the investigation of paintings. Other artworks can be investigated even if they are hidden behind THz-transparent materials.

Figure 7. Selected frames of the E(t)-movie (top) and the E(ω)-movie (bottom) of the painting segment in the middle.

ACKNOWLEDGEMENTS as well as investigations on the individual materials and substances of paintings must be carried out. The images shown in Figure 7 are characterized by comparably large pixels due to the large measuring

We gratefully acknowledge Prof. Unger (RathgenForschungslabor) for helpful discussions and sample supply.

38

REFERENCES

Möhlenkamp, A., Kuder, U. & Albrecht, U. 2002. Geschichte in Schichten. Wand und Deckenmalerei im städtischen Wohnbau des Mittelalters und der frühen Neuzeit Int. Symposium 200 in Lübeck, Lübeck: Möhlenkamp, A., Kuder, U. & Albrecht, U. Pursche, J. 2002. Freilegen oder Verdecken? Erfahrungen aus Jahrzehnten. Geschichte in Schichten. Wand und Deckenmalerei im städtischen Wohnbau des Mittelalters und der frühen Neuzeit, Int. Symposium 200 in Lübeck: 204–219. Lübeck: S. Möhlenkamp, A., Kuder, U. & Albrecht, U. Roskos, H. & Löffler, T. 2004. Kurze Wellen, lange Wellen, Terawellen. Forschung aktuell (3–4): 45–48. Shur, M. 2005.Terahertz technology: devices and applications. Proceedings of ESSCIRC: 13–21. Smith, P. 2005. Pharmaceutical Analysis using Terahertz Spectroscopy. Innovation in Pharmaceutical Technology: 73–76. Stewart, S. 1991. The uncovering of wall paintings: Ethics and methods. Unpublished Diploma Thesis. London: Courtauld Institute of Art.

Bäumer, K. Terahertz durch dick und dünn – bei anderem Licht besehen. EMVU und Technik: 7–10. Dorney, T. et al. 2000. Imaging with Thz pulses: 763–767. Rice University, Houston. Gorenflo, S. et al. 2005. Terahertz – Time – Domain – Spektroskopie mit einem leistungsoptimierten elektrooptischen Detektionsverfahren. Technisches Messen 72: 435–437. Hoffmann, S. & Hofmann, M. 2005. Terahertz – Strahlung entgeht nichts. Rubin 1: 42–48. Humphries, H. 2001. Infrared and Thermal Testing for Conservation of Fine Art. Infrared and thermal testing. ASNT Non-destructive Testing Handbook 3 (3). Leitner, H. 1994. Die Freilegung der Landkartengalerie der erzbischöflichen Residenz in Salzburg, unveröffentlichter Restaurierbericht. Mairinger, F. 2003. Strahlenuntersuchung an Kunstwerken, Leipzig. Manser,A. & Battaglia, C. 2007. Prozessnah und berührungslos. MQ Management und Qualität 06: 37–39.

39

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Femtosecond laser cleaning of paintings: Modifications of tempera paints by femtosecond laser irradiation S. Gaspard, M. Oujja & M. Castillejo Instituto de Química Física Rocasolano, CSIC, Madrid, Spain

P. Moreno, C. Méndez & A. García Servicio Láser, Universidad de Salamanca, Salamanca, Spain

C. Domingo Instituto de Estructura de la Materia, CSIC, Madrid, Spain

ABSTRACT: Laser cleaning of paintings aims at removing oxidized varnish layers or overpaints. Removal of these coatings has to be performed without degradation of the light-sensitive paint layers. Femtosecond (fs) laser irradiation is being explored as a cleaning technique for different types of substrates due to the reduced thermal load as compared with nanosecond (ns) irradiation. In this work we present a study of the modifications induced by fs Ti:Sapphire laser pulses (795 nm, 120 fs, 1 KHz) on model unpigmented (egg yolk based binder) and pigmented tempera samples (azurite and zinc white). Irradiation with fluences below and above the ablation threshold, leads to various degrees of discoloration. Several analytical techniques were used to characterize the chemical and physical effects induced by fs laser irradiation including colorimetry, optical microscopy, laser induced fluorescence (LIF) at 266 nm and FT Raman at the laser excitation of 1064 nm.

1

INTRODUCTION

fluence at the specified wavelength (Chappé et al. 2003, Andreotti et al. 2007). In the last years, material processing with femtosecond (fs) laser pulses has attracted increasing attention. Compared to ns pulses, the main advantages of fs pulses are the reduction of the thermal diffusion and consequently the thermal degradation of the target, the reduction of shielding induced by the plasma plume and the improvement of the morphology of the ablated surface. In particular, fs laser cleaning can be advantageous for the treatment of light-sensitive substrates as artistic paintings (Fotakis et al. 2007 & Pouli et al. 2007). The study presented in this work aims to explore possible advantages that could offer the use of fs pulses for the cleaning of egg yolk based tempera paints by examining the physical and chemical effects taking place upon direct exposure of unvarnished samples to laser irradiation. Unvarnished aged model temperas of unpigmented, azurite and zinc white paints were irradiated by a Ti:Sapphire laser at 795 nm with pulses of 120 fs. We determined ablation thresholds of the systems and the effects induced by laser irradiation; physical and chemical modifications, were assessed by colorimetry, optical microscopy, laser induced fluorescence and Fourier transform FT-Raman. The results

Laser cleaning of paintings and polychromes was introduced as a novel conservation method few years ago (Teule et al. 2002, Castillejo et al. 2002, 2003, Gordon Sobott et al. 2003, Bordalo et al. 2006, Nevin et al. 2007). Different studies have been carried out on the laser removal of degraded varnishes or contamination layers from paintings and polychromes. These studies used pulses of nanosecond (ns) duration delivered by UV excimer (Georgiou et al. 1998), Q-switched Nd:YAG (Hildenhagen et al. 2003) and Erbium YAG lasers (Bracco et al. 2003). Due to the sensitivity to light of the components of pictorial artworks, careful studies are required to characterize the effects of laser irradiation on pigments, binders or varnishes. Investigation is necessary to identify the possible alterations induced by laser and several multianalytical investigations have been performed (Fotakis et al. 2005) to understand the mechanisms operating in the interaction with laser pulses, responsible of discoloration and degradation (Zafiropoulos et al. 2003, Pouli et al. 2003). In all, previous studies on the laser cleaning of paintings have highlighted the importance of the optimization of laser parameters, specifically pulse duration and

41

obtained are compared with those of a previous study performed with a KrF excimer laser at 248 nm, 25 ns pulses (Castillejo et al. 2002). 2 2.1

EXPERIMENTAL Samples

The samples of tempera paint on wood panels consist in a single paint layer of around 100 µm thickness applied on primed panels by using a stopping knife, (Castillejo et al. 2002). The pigments were previously mixed with egg yolk using a spatula and a glass plate. The samples were naturally aged for a period of four years in the dark. In this work, we present the results obtained in azurite (basic copper carbonate, 2CuCO3 · Cu(OH)2 ) and zinc white (zinc oxide, ZnO). Samples of unpigmented paint (egg yolk) were used as reference to study the modifications of the binding medium upon laser irradiation. Figure 1. Pictures of tempera samples and schemes of the different irradiated zones. Conditions of irradiation are given in Table 1: a) unpigmented, b) zinc white and c) azurite. The irradiated areas are of 1 cm2 .

2.2 Laser treatment Laser ablation was carried out in air using a commercial Ti:Sapphire oscillator (Tsunami, Spectra Physics) and a regenerative amplifier system (Spitfire, Spectra Physics) based on the chirped pulse amplification technique. The system produces linearly polarized 120 fs pulses at 795 nm with a repetition rate of 1 kHz. The pulse energy can reach a maximum of 1.1 mJ which is controlled by means of neutral density filters and measured with a powermeter. The transversal mode is gaussian TEM00. The beam was focused perpendicularly onto the target surface which was placed on a motorized XYZ translation stage. The pulses were focused on the surface by a cylindrical lens with focal length of 75 mm providing a spot size 6 × 9000 µm2 (1/e2 criterion). Homogeneous irradiation of the sample within a 1 cm2 , was achieved by using a squared mask placed on the surface of the sample and by scanning the sample along the direction of the smallest spot dimension of the beam with a scanning speed of 780 µm/s. At this speed, the pulses overlapped at an intensity of 87% of the maximum. After processing the whole square length, the sample was moved in the transverse direction by steps of 1500 µm, 2000 µm or 3000 µm, resulting in an overlap of 94.6, 90.6 and 82.6%. Motion in the Z axis helped to accurately focus the laser beam on the material surface. After determination of the ablation threshold fluences (Fth ) for each system (see 3.1), different irradiated zones were prepared with fluences below and above the threshold (Fig. 1, Table 1). For the unpigmented paint, seven zones were irradiated with fluences between 0.30 J/cm2 (F/Fth = 0.31) and 0.80 J/cm2 (F/Fth = 0.82). The azurite sample was irradiated with fluences between 0.28 J/cm2

Table 1.

Irradiation conditions of tempera paints.

Zones∗

Fluence J/cm2

F/Fth

Unpigmented tempera 1 (1500), 2 (2000) 3 (1500), 4 (2000) 5 (1500), 6 (2000) 7 (1500) Azurite tempera 1 (1500) 2 (1500) 3 (1500), 4 (3000), 5 (2000) 6 (1500), 7 (2000) 8 (1500), 9 (2000) 10 (1500), 11 (2000)

0.80 0.60 0.40 0.30

0.82 0.62 0.41 0.31

1.26 0.99 0.90 0.56 0.37 0.28

2.38 1.87 1.70 1.06 0.70 0.53

Zinc white tempera 1 (1500), 2 (2000) 3 (1500), 4 (2000)

0.65 0.53

1.67 1.00

∗ The displacement given in µm, is reported in parentheses (see text).

(F/Fth = 0.53) and 1.26 J/cm2 (F/Fth = 2.38). Finally, different areas were irradiated in zinc white tempera with fluences of 0.53 J/cm2 (F/Fth = 1.00) and 0.65 J/cm2 (F/Fth = 1.67). 2.3 Analytical techniques To record the reflectance spectra and characterize the chromatic properties and changes induced by laser irradiation, we used a Minolta CM-2500d

42

Table 2. Ablation thresholds and incubation factors. Ablation thresholds J/cm2

Unpigmented Azurite Zinc White

1 pulse

5 pulses

10 pulses

100 pulses

Incubation factor

1.5 ± 0.2 0.8 ± 0.1 0.57 ± 0.09

1.1 ± 0.2 0.58 ± 0.09 0.42 ± 0.06

0.90 ± 0.08 0.50 ± 0.04 0.37 ± 0.03

0.54 ± 0.04 0.31 ± 0.02 0.24 ± 0.02

0.78 ± 0.01 0.79 ± 0.02 0.81 ± 0.02

by an aperture. The diffraction pattern (Airy disk and rings) is focused on the target surface. The material is damaged where the laser fluence is above the threshold value. The diameters of the craters were determined by optical microscopy (ZeissAxio Imager Z1m) and scanning electron microscopy (Zeiss DSM940) (Moreno et al. 2006). Thresholds measured for 1, 5, 10 and 100 pulses and incubation factors ξ are reported in Table 2. The incubation factor is determined by Fth (N) = Fth (1) × N(ξ−1) , with Fth (1) and Fth (N) being the threshold fluences for 1 and N pulses respectively. It is important to underline that we clearly observed discoloration in the irradiated zones of the unpigmented sample upon irradiation with fluences below the ablation threshold. Thresholds of unpigmented system are higher than those of pigmented paints. In fact, in presence of pigment, the effective multiphoton absorption of the paints increases. The threshold drops for a higher number of pulses per spot. This can be correlated with incubation phenomena, however the high incubation factors calculated, similar in all systems studied (∼0.8), indicate a weak incubation effect and therefore a relative chemical stability of the systems upon multiple pulse irradiation (Krüger & Kautek 2004). As during the processing, a scanning speed of 780 µm/s was used and the pulses overlapped at an intensity of 87% of its maximum, 7.6 pulses participate in the formation of one crater. Therefore, ablation thresholds used therein are calculated by interpolation for this number of pulses.

portable spectrophotometer. The parameters were a standard illuminant D65 (average daylight) and a 10◦ standard observer, on an observation area of 1 cm diameter. Three spectra were recorded in each irradiated zone, and averaged to obtain one data point. Changes in the reflectance spectra were determined with the CIEL∗ a∗ b∗ colorimetric procedure. L∗ indicates lightness and a∗ and b∗ are the chromaticity coordinates. Colour changes are given in a three dimensions space (L∗ : + lighter, − darker; a∗ : + redder, − greener; b∗ : + yellower, − bluer). The magnitude of the colour change is given by E ∗ = [(L∗ )2 + (a∗ )2 + (b∗ )2 ]1/2 . Laser induced fluorescence (LIF) measurements of the virgin and irradiated areas were carried out using laser excitation at 266 nm (Q-switched Nd:YAG laser, 4th harmonic, 6 ns pulse) and a 0.30 m spectrograph with a 300 lines/mm grating (TMc300 Bentham)intensified charged coupled detector (2151 Andor Technologies) system. The temporal gate was operated at zero time delay and at a temporal width of 3 µs. The sample was illuminated at an incidence angle of 45◦ at a laser energy of about 0.1 mJ/pulse. For the results presented here, a 300 nm cutoff filter was installed in front of the spectrograph. Each spectrum resulted from the average of 20 measurements in 5 different points of each irradiated zone. FT-Raman spectra were recorded with an RFS 100/S-G Bruker spectrometer. The excitation source consisted of a Nd:YAG laser emitting at 1064 nm. Low laser power outputs, in the range of 10–20 mW, were used. Only 1/3 of this power illuminated the sample surface, preventing damage or laser-induced degradation of the samples during measurements. The light scattered from a surface of 0.01 cm2 was collected in backscattering (or 180◦ ) geometry. The wavenumber resolution was 8 cm−1 . Each data point was the result of the accumulation of 200 scans. 3

3.2 Colorimetric measurements Colorimetric measurements were performed on virgin and irradiated areas of the samples to characterize the discoloration resulting from laser irradiation of the tempera paint. Virgin zones of the samples were characterized in the CIEL∗ a∗ b∗ colorimetric space and the coordinates are reported in Table 3. In Figure 2, values of E ∗ are represented as a function of the irradiation conditions (F/Fth ). We previously reported (Gaspard et al., in press) an important degree of discoloration of the unpigmented sample upon irradiation with fluences below the ablation threshold (E ∗ = 46 at a fluence of 0.80 J/cm2 ).

RESULTS AND DISCUSSION

3.1 Ablation thresholds Ablation thresholds, for irradiation with fs pulses, were calculated using the method described by Dumitru et al. (2002) based on the diffraction of a laser beam

43

Table 3. CIEL∗ a∗ b∗ colorimetric coordinates of the virgin zones of the samples. Tempera paint

L∗

a∗

b∗

Unpigmented Azurite Zinc white

76.3 31.4 89.5

9.8 −8.6 −0.5

55.6 1.1 15.9 Figure 3. Optical microscopy pictures of the unpigmented tempera surface: (a) non irradiated zone and (b) irradiated zone at 0.80 J/cm2 .

Figure 2. Values of E∗ as a function of the irradiation conditions for the three tempera paints. Values correspond to zones irradiated with a displacement of 1500 µm.

Figure 4. LIF spectra at the excitation wavelength of 266 nm of virgin and irradiated zones of azurite tempera paint, with fluences of 0.28 and 0.90 J/cm2 (zones 10 and 3 respectively, see Table 1).

The main colour shift observed was due to changes in b∗ (−22, shift to bluer) and L∗ (−15, shift to darker). Furthermore, the observation of the irradiated zones of the unpigmented tempera by optical microscopy reveals the formation of bubbles below the ablated surface. (Fig. 3) Interaction of the laser radiation with the azurite system results in two different behaviours. Below the ablation threshold (F/Fth < 1), the colour of the sample was not altered. Above the ablation threshold, irradiation results in discoloration and the pigment acquires a white colour. A maximum E ∗ value of 9.6 at 1.26 J/cm2 is observed with L∗ (+ 8.2, shift to lighter) being the main factor of discoloration. Zinc white sample reacts differently, as this system remain practically unaltered upon laser irradiation with fluences above the threshold value. A maximum value of E ∗ = 1.8 at 0.65 J/cm2 was observed (Figs. 1–2).

the aromatic amino acids of the proteins tyrosine and tryptophan at 333 nm, the phospholipids in the 520– 570 range and derived crosslinked products of egg yolk (Wisniewski et al. 2007, Gaspard et al. 2008a). These products of photo-oxidation, combination and modification of amino acids, such as dityrosine, 3,4 dihydroxyphenylalanine (DOPA) or N-formylkynurenine (NFK) and kynurenine display fluorescence in the 400–500 nm region (Nevin et al. 2006). Other products of cross linking reactions between amino acids and sugar or lipids that are present in egg yolk can also contribute to the broad emission in the 400–650 nm region. LIF spectra of azurite paint are similar to those of the binder. They consist of two broad bands centred at 330 nm and at 470 nm with a shoulder at 445 nm (Fig. 4). In this case, the emission is narrower than the one of the binder and the emission from phospholipids is not observed. LIF spectra of zinc white tempera consist of a very intense and narrow band at 385 nm, attributed to the luminescence of semiconductor ZnO and a very broad band in the 400–500 nm region attributed to photodegradation products of the binder (Fig. 5).

3.3 Laser Induced Fluorescence LIF spectra were recorded on virgin and irradiated zones of the samples upon excitation at 266 nm. We previously reported LIF analysis on the unpigmented sample (Gaspard et al. in press). At 266 nm, the spectrum of the virgin zone consists of two broad bands centred at 333 nm and 520 nm with a shoulder at 450 nm. The emissions observed have their origin in

44

has been adequately subtracted. For pigmented samples, spectra of the virgin zones show bands of both pigment and binder and agree with the previously reported spectra for azurite and zinc white (Burgio et al. 2001). We previously reported the most relevant and characteristic bands of the FT Raman spectra of the unpigmented tempera (Gaspard et al. in press). In particular, we observed the C-H stretching region from 2700 to 3100 cm−1 , the C = O stretching at 1741 cm−1 , the Amide I and Amide III bands of the proteins backbone at 1653 and 1263 cm−1 respectively and the methylene groups of lipids at 1445 and 1302 cm−1 . From amino acids, only the phenylalanine band at 1003 cm−1 could be clearly identified. After irradiation, the binder bands showed no appreciable changes in the 1700–500 cm−1 region. New bands indicating a change in the chemical composition of the azurite pigment appear around 143 and 640 cm−1 as a result of laser irradiation. These two bands can be assigned to the reddish semiconductor cuprite Cu2 O (Castillejo et al. 2002). A very low shift to red (b∗ = 3 at 1.26 J/cm2 ) is observed by colorimetric measurements in the irradiated zone 1 of the sample. No changes were observed in the spectra of irradiated zones of zinc white paint. After irradiation, all spectra show an increase of the intensity in the C-H stretching region (not shown). This change related to the increase of CH3 groups is attributed to the degradation of lipids and proteins, also observed by LIF.

Figure 5. Normalized LIF spectra at the excitation wavelength of 266 nm of virgin and irradiated zones of zinc white tempera paint, with fluences of 0.53 and 0.65 J/cm2 (zones 1 and 3 respectively, see Table 1).

4

CONCLUSIONS

The modifications induced by 120 fs laser pulses at 795 nm were examined in unpigmented, azurite and zinc white tempera paints. Irradiation results in various degrees of discoloration and chemical changes as monitored by LIF and FT Raman. Results presented above can be discussed in relation with previous studies by some of us on KrF excimer laser irradiation (248 nm, 25 ns pulses) of similar tempera paint samples (Castillejo et al. 2002, 2003). Regarding colour changes, the degree of discoloration of the unpigmented tempera is significant below the ablation threshold while azurite tempera colour remains stable below the threshold and shows some shift to white above this value. In comparison with measurements under ns KrF laser treatment at equivalent irradiation fluences, a higher degree of discoloration (higher values of E ∗ ) is observed under fs irradiation. Nevertheless, in the presence of a pigment, the inverse phenomenon is observed. Zinc white tempera remains unaltered in the explored fluence range whereas the paint turned to black at 248 nm, 25 ns irradiation. LIF and Raman measurements allow the discussion of the chemical alterations induced in the binder and

Figure 6. FT Raman spectra of the azurite tempera paint sample in virgin and irradiated zone 1 (see Table 1) at 1.26 J/cm2 . Azurite bands are indicated by vertical bars.

LIF spectra recorded in the irradiated areas of the unpigmented and azurite tempera paints revealed an overall decrease of the fluorescence signal. A relative increase of the shoulder band at 450 nm was observed and attributed to enhanced contribution of photodegradation products of proteins participating in the composition of egg yolk which emissions are predominant in this region. For the zinc white paint, a comparable relative increase of the broad band in the 400–500 nm region was observed upon irradiation (see normalized spectra in Fig. 5). 3.4 FT-Raman spectra FT Raman spectra were recorded in virgin and irradiated zones of the temperas. Spectra of azurite sample are shown in Figure 6, once the spectrum of the panel

45

the pigments under fs laser irradiation. LIF spectral modifications observed upon irradiation indicate an enhancement of photodegradation compounds of proteins and lipids which are present in the egg yolk based binder. The effect of laser irradiation on the pigment itself is extremely dependent of the pigment composition, as illustrated in this work, in the comparison characteristic LIF and Raman results for azurite and zinc white pigments. Whitening of the azurite system, accompanied by formation of cuprite, is in contrast with unaltered colour and absence of chemical changes in zinc white paint. Excimer laser ablation thresholds for unpigmented, azurite and zinc white systems, 0.17, 0.24 and 0.34 J/cm2 respectively (Castillejo et al. 2002, 2003) are lower than those reported here for 795 nm, 120 fs irradiation (Table 2) which indicates a higher stability of tempera systems upon fs irradiation. In all cases no build-up of extra bands of amorphous carbon (indicative of carbonization or charring) takes place, in contrast with previous observations upon irradiation with 248 nm, 25 ns pulses (Castillejo et al. 2002, 2003). The differences with these previous studies illustrate the participation of mechanisms of diverse origin in the ns and fs domains. Work is in progress to characterize the mechanisms involved in the interaction of fs laser pulses with paints, aiming at establishing the advantages, related to the high degree of control over the induced modifications, offered by ultrashort laser pulses for the cleaning of paintings. ACKNOWLEDGEMENTS Funding from MEC (Project CTQ2007-60177-C0201/PPQ) is gratefully acknowledged. S.G. thanks EU for a Marie Curie contract (MESTCT-2004-513915). We thank R. Hesterman (Hesterman Restauratie Atelier voor Schilderijen, The Netherlands) for the preparation of the samples and M.I. Sanchez Rojas (Instituto Eduardo Torroja, CSIC) for the use of the spectrophotometer. We also acknowledge the support of the Red Temática de Patrimonio Histórico y Cultural, CSIC. REFERENCES Andreotti, A., Colombini, M. P., Nevin, A., Melessanaki, K., Pouli, P. & Fotakis, C. 2007. Multianalytical Study of Laser Pulse Duration Effects in the IR Laser Cleaning of Wall Paintings from the Monumental Cemetery of Pisa. Laser Chemistry Article ID 39046. Bordalo, R., Morais, P. J., Gouveia, H. & Young, C. 2006. Laser Cleaning of Easel Paintings: An Overview. Laser Chemistry Article ID 90279. Bracco, P., Lanterna, G., Matteini, M., Nakahara, K., Sartiani, O., de Cruz,A., Wolbarsht, M. L.,Adamkiewicz, E. & Colombini, M. P. 2003. Er:YAG laser: an innovative tool for controlled cleaning of old paintings: testing and evaluation. Journal of Cultural Heritage 4: 202s–208s.

46

Burgio, L. & Clark, R. J. H. 2001. Library of FT-Raman spectra of pigments, minerals, pigment media and varnishes, and supplement to existing library of Raman spectra of pigments with visible excitation. SpectrochimicaActa Part A 57: 1491–1521. Castillejo, M., Martín, M., Oujja, M., Silva, D., Torres, R., Manousaki A., Zafiropulos, V., Van den Brink, O. F., Heeren, R. M. A., Teule, R., Silva, A. & Gouveia, H. 2002. Analytical study of the chemical and physical changes induced by KrF laser cleaning of tempera paints. Analytical Chemistry 74: 4662–4671. Castillejo, M., Martín, M., Oujja, M., Santamaría, J., Silva, D., Torres, R., Manousaki, A., Zafiropulo, V., Van den Brink, O. F., Heeren, R. M. A., Teule, R. & Silva, A. 2003. Evaluation of the chemical and physical changes induced by KrF laser irradiation of tempera paints. Journal of Cultural Heritage 4: 257s–263s. Chappé, M., Hildenhagen, J., Dickmann, K. & Bredol, M. 2003. Laser irradiation of medieval pigments at IR, VIS and UV wavelengths. Journal of Cultural Heritage 4: 264s–270s. Dumitru, G., Romano, V., Weber, H. P., Sentis, M. & Marine, W. 2002. Femtosecond ablation of ultrahard materials. Applied Physics A 74: 729–739. Fotakis, C., Anglos, D., Zafiropulos, V., Georgiou, S. & Tornari V. (ed.) 2005. Lasers in the preservation of Cultural Heritage, Principles and Applications. In Series in Optics and Optoelectronics. NewYork: Taylor and Francis group. Fotakis, C., Zorba, V., Stratakis, E., Athanassiou, A., Tzanetakis, P., Zergioti, I, Papagoglou, D. G, Sambani, K., Filippidis, G., Farsari, M., Pouli, P., Bounos G. & Georgiou, S. 2007. Novel Aspects of Materials Processing by Ultrafast Lasers: From Electronic to Biological and Cultural Heritage Applications. Journal of Physics: Conference Series 59: 266–272. Gaspard, S., Oujja, M., Abrusci, C., Catalina, F., Lazare, S., Desvergne, J. P. & Castillejo, M. 2008a. Laser induced foaming and chemical modifications of gelatine films. Journal of Photochemistry and Photobiology A 193: 187–192. Gaspard, S., Oujja, M., Moreno, P., García,A., Domingo, C. & Castillejo, M. 2008b. Interaction of femtosecond laser pulses with tempera paints. Applied Physics A, in press. Georgiou, S., Zafiropulos, V., Anglos, D., Balas, C., Tornari, V. & Fotakis, C. 1998. Excimer laser restoration of painted artworks: procedures, mechanisms and effects. Applied Surface Science 127–129: 738–745. Gordon Sobott, R. J., Heinze, T., Neumeister, K. & Hildenhagen, J. 2003. Laser interaction with polychromy: laboratory investigations and on-site observations. Journal of Cultural Heritage 4: 276s–286s. Hildenhagen, J. & Dickmann, K. 2003. Nd:YAG laser with wavelengths from IR to UV (ω, 2ω, 3ω, 4ω) and corresponding applications in conservation of various artworks. Journal of Cultural Heritage 4: 174s–178s. Krüger, J. & Kautek, W. 2004. Ultrashort pulse laser interaction with dielectrics and polymers. Advance polymers Science 168: 247–289. Moreno, P., Méndez, C., García, A., Arias, I. & Roso, L. 2006. Femtosecond laser ablation of carbon reinforced polymers. Applied Surface Science 252: 4110–4119. Nevin, A., Cather, S., Anglos, D. & Fotakis, C. 2006. Analysis of protein-based binding media found in paintings

using laser induced fluorescence spectroscopy. Analytical Chimica Acta 573–574: 341–346. Nevin, A., Pouli, P., Georgiou, S. & Fotakis, C. 2007. Laser conservation of art. Nature Materials 6: 320–322. Pouli, P., Emmony, D.C., Madden, C. E. & Sutherland, I. 2003. Studies towards a thorough understanding of the laserinduced discoloration mechanisms of medieval pigments. Journal of Cultural Heritage 4: 271s–275s. Pouli, P., Bounos, G., Georgiou, S. & Fotakis, C. 2007. Femtosecond Laser Cleaning of Painted Artefacts; Is this the Way Forward? In J. Nimmrichter, W. Kautek & M. Schreiner (eds), Proceedings of Lasers in the Conservation of Artworks, LACONA 6, Vienna 21–25 September 2005: 287–293. Berlin: Springer. Teule, J. M., Ullenius, U., Larsson, I., Hesterman, W., van den Brink, O. F., Heeren, R. M. A. & Zafiropulos, V.

2002. Controlled laser cleaning of fire-damaged paintings. ICOM Committee for Conservation. Wisniewski, M., Sionkowskaa,A., Kaczmarek, H., Lazare, S., Tokarev, V. & Belin, C. 2007. Spectroscopic study of a KrF excimer laser treated surface of the thin collagen films. Journal of Photochemistry and Photobiology A 188: 192– 199. Zafiropulos, V., Balas, C., Manousaki, A., Marakis, Y., Maravelaki-Kalaitzaki, P., Melesanaki, K., Pouli, P., Stratoudaki, T., Klein, S., Hildenhagen, J., Dickmann, K., Luk’Yanchuk, B. S., Mujat, C. & Dogariu, A. 2003. Yellowing effect and discoloration of pigments: experimental and theoretical studies. Journal of Cultural Heritage 4: 249s–256s.

47

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Cleaning of paint with high repetition rate laser: Scanning the laser beam A.V. Rode, D. Freeman, N.R. Madsen & K.G.H. Baldwin Research School of Physical Sciences and Engineering, the Australian National University, Canberra, Australia

A. Wain Collection Services Section, the Australian War Memorial, Canberra, Australia

O. Uteza & P. Delaporte Lasers, Plasmas and Photonic Processes, CNRS – Mediterranean University, Marseille, France

ABSTRACT: Powerful ultrafast femtosecond laser pulses have the unique ability to ablate material with minimal collateral damage. This ability offers the potential for new applications of ultrafast lasers for removing surface contamination and unwanted surface layers in the conservation of artworks and heritage objects. In this paper, we concentrate on the problem of precise and fast scanning of the laser spot over the treated surface for cleaning relatively large surface areas. Preliminary results are presented for the removal of intrusive paint layers, using a 12 ps laser with 1.5 MHz repetition rate and 0.5 ps laser with 10 kHz repetition rate.

1

INTRODUCTION surgery, which are also very sensitive to collateral damage (Feit et al. 1998, Loesel et al. 1998, Juhasz et al 2000, Rode et al. 2002). Compared to long-pulse nanosecond lasers, femtosecond lasers offer higher etching resolution, minimal collateral damage and minimal photochemical modification of the surface. These qualities have opened up new possibilities in the cleaning of sensitive and technically demanding artworks and other objects of historical and cultural importance (Pouli et al. 2005). Currently, however, the application of femtosecond lasers to the cleaning of artifacts has been limited to small surface areas, and to a limited volume of the order of few mm3 at most. This is due to the fact that the low energy of the ultrashort laser pulse requires tight focusing to achieve the required ablation threshold intensity, which is of the order 1011 W/cm2 – 1012 W/cm2 . Moreover, the lower ablated mass per pulse, as compared with conventional long-pulse lasercleaning techniques, requires a significant increase in the laser repetition rate to achieve a reasonable rate of surface treatment. A technique is therefore required to provide fast and precise scanning of the laser beam over the surface area to be treated. In this paper, we analyse the potential advantages and challenges of using ultrafast laser pulses for precise removal of surface layers, and present preliminary

Laser ablation with ultrafast laser pulses possesses a unique ability to remove thin submicron surface layers without generating heat or shock waves in the bulk of the material. This is due to the very fast delivery of electromagnetic laser energy into the surface layer, which occurs too quickly for the absorbed laser energy to be transferred into the bulk material through heat conduction or shock wave propagation (Du et al. 1994, Stuart et al. 1995, Gamaly et al. 2002, Gamaly et al. 2007). As a result, a heated surface layer is ablated while the bulk of the material remains cold. The pulse duration required to remove a surface layer in this non-equilibrium, short-pulse regime of ablation depends on the properties of the material, such as heat diffusion, thermal capacity, ionisation energy and density, as well as electronic and optical properties (Gamaly et al. 2002, Gamaly et al. 2007). In general, the pulse should be shorter than the electron-lattice coupling time and the heat diffusion time from the surface layer. This usually lies in the picosecond time domain. The thickness of the layer removed at the laser fluence just above the ablation threshold is of the order of 100 nm per pulse. The advantages of using ultrafast lasers for micromachining (precise removal) of surfaces have already been proven in applications such as dentistry and eye

49

results on using ps and sub-ps lasers to remove paint for the conservation of artworks and heritage objects. 2 2.1

LASER SCANNING SYSTEM Scanning requirements

A key requirement for precise removal of surface layers of sub-micron thickness is a high degree of control over the scanning of the focused laser beam. There is a number of general requirements for scanning systems for the short-pulse high-repetition-rate laser ablation technique: – very precise control of the ablated depth at the level of 10 nm/shot requires that each laser shot be located at a fresh point on the surface; – scanning speed along the surface should be faster than the propagation of the heat wave, to avoid heat accumulation; – the beam should move with a constant speed over the target surface; – the scanning pattern should cover the surface uniformly, so that every surface point is given the same exposure; – the spot size and shape should be the same across the entire scanning pattern.

Figure 1. Scheme for a constant scanning velocity alternating spiral pattern. This is the most suitable for fast scanning of the laser beam due to the absence of sharp turning points.

placing the scanning mirrors after the focusing lens so that the difference in optical path of the deflected beam is within the lens caustic. This is, however, at expense of the tight focusing required for low-energy short pulses. 2.2 Constant velocity alternating spiral

The first and second requirements lead to a relatively high scanning speed v = dfoc × Rrep , where dfoc is the focal spot diameter, and Rrep is the laser repetition rate. For organic materials with thermal diffusivity D of the order of 10−2 cm2 /s and with 1 MHz repetition rate the speed is typically

The scanning pattern must be chosen so as to respect the limitations of each scanner; namely, the maximum acceleration (given by the available torque and the moment of inertia of the motor and mirror) and the maximum power dissipation in the motor windings (governed by the average magnitude of acceleration throughout the pattern). One should therefore avoid patterns that involve sharp corners, such as raster scanning, unless a fast beam blanker is available to allow “dead time” for a safer acceleration without uneven illumination of the target. While it is feasible to use a “polygon” scanner for the fast axis of a raster pattern, achieving even higher scanning speeds, there is still a problem at the edges of the scan lines. The beam is split between the end of one line and the start of the next, destroying the quality of the focused spot and reducing the fluence below the ablation threshold, so that a fast blanker would still be required. Blanking an ultrafast laser at high average and peak powers is not a trivial proposition, so we have used an alternative pattern for our work. A smooth and continuous spiral pattern, alternating between increasing and decreasing radius, has no corners and provides continuous scanning at a constant linear speed with nearly uniform coverage and no wasted dead-time (Figs. 1, 2). It suffers, however, from the disadvantage of a hole in the centre, which is required to limit the acceleration at small radii. The scanners were computer-controlled as follows. The path was traced at a constant linear speed vscan by

where lth is the thermal wave propagation length between two laser pulses. This scanning speed can be obtained with relative ease with an orthogonal pair of galvanometer scanners located at a suitable distance from the target. The third and forth requirements lead to an even coverage of the surface by the scanning laser spot, so that the number of laser pulses per unit area is constant. Galvanometer scanners typically consist of a mirror mounted onto an electric motor, with positional feedback supplied to the driver circuit. Under closed-loop operation, a return spring is unnecessary, since the drive electronics can accurately apply the required torque to start and stop the mirror even for a sudden jump in position. Proper tuning of the driver is essential, particularly when different mirrors are fitted. The fifth requirement, for a beam of constant size and shape, can be achieved with a telecentric scanning lens. Such a lens is designed so that the laser beam strikes normal (perpendicular) to the working surface over the entire scanning field. This can also be achieved with conventional long-focus lenses by

50

Figure 2. Laser scanning patterns: Lissajous figure (left), Lissajous figure superimposed on a circle (centre), and an expanding/collapsing spiral (right). Clearly, the spiral pattern provides the most uniform coverage, although it has a hole in the centre. The images are experimental results obtained by scanning onto a paper diffuser and acquiring an image using a CCD camera with a long integration time.

Figure 3. Racetrack scanning patterns at various laser fluences from 0.09 J/cm2 to 5.2 J/cm2 (the numbers under the patterns are fluences in J/cm2 ).

generating a pair of analogue waveforms, which were sampled at discrete time intervals (10 µs). The sample points were computed at the target, and converted to deflection angles for the scanners ±40◦ using the optical distance from each scanner to the target, and then output using 16-bit digital-to-analogue converters, resulting in angular steps of 80◦ /216 = 21.3 µrad. The outward spiral curve used, expressed in polar coordinates (r, θ), is:

uneven surface, the confocal parameter (the depth of focus) should be longer than the surface profile variations. The longer the focal length of the focusing optics and the smaller the numerical aperture, the longer the confocal parameter, which converts into a larger focal spot and lower laser fluence. To compensate the loss in fluence, which must be above the threshold value, one needs to increase the energy per pulse. This can only be achieved by increasing the average power of the laser or by reducing the repetition rate. 3 ABLATION EXPERIMENTS

where p is the pitch, the radial distance between successive turns; here, the pitch is larger than the focal spot. The value of θ at each sample time was computed using a discrete integration of the angular velocity determined at the preceding sample time. The angular velocity of this curve is given by:

Two laser systems were used to perform ultrafast laser ablation of paint. The aim of the tests was to analyse the effectiveness of ultrafast laser ablation in the removal of paint from firstly an underlying paint layer, and secondly a metal surface. One of the lasers used was a Nd:YVO4 laser, designed and built at the Australian National University for applications in micro-machining and the deposition of optical thin films (Luther-Davies et al. 2004). This laser produces an average power of 50 W in 12 ps pulses at a rate of 1.5 × 106 pulses per second and was converted to the second harmonic (λ2 = 532 nm). The second laser used was a commercialYb:YAG laser from ‘Amplitude Systems’ which generates 500 fs pulses at a rate 104 pulses per second; this laser was converted to the second harmonic (λ2 = 515 nm). The experiments were conducted with a light-grey ZnO based paint. The first series of experiments aimed to determine the required speed of scanning. Laser beams were directed to a painted metal target via a telescentric lens and focused to a spot size of 20 µm (532 nm) and 35 µm (515 nm). The laser was then scanned in a constant velocity ‘moving race track’ pattern with a size of 20 mm (532 nm) and 2 mm (515 nm) (Fig. 3a). The race-track pattern was chosen as a compromise to avoid the hole in the middle of the pattern but still preserve the constant scanning velocity over the linear tracks of the pattern.

When the radius reached the maximum allowed, the spiral changed to the inward type in which the radius decreased with further increase of angle, until the desired inner radius was reached and the next outward spiral began. Since the angular frequency (and therefore scanning mirror acceleration) is greatest at the centre of the pattern, the radius of the central hole must be large enough to respect the limitations of the scanners as described above, or else the pattern would become distorted, with non-uniform speed and area coverage at the centre. To ensure the individual spiral traces were not visible in the final result, a small arc θ was included at the outermost radius to produce a suitable angular offset between successive spirals, resulting in uniform coverage after a large number of spirals had been traced. This pattern has been the most successful to date in our laser deposition experiments (Luther-Davies et al. 2005, Gamaly et al. 2007). Keeping the laser parameters constant on the surface to be treated is another challenge. To treat an

51

Figure 4. Ablation of paint using 515 nm, 500 fs laser pulses with scanning speeds 1 shot/spot, 10 shot/spot and 100 shot/spot. Redeposition of debris can be clearly seen at the edges of the treated areas at the scanning speeds 10 shots/spot and 100 shots/spot.

Figure 5. An example of ablation rate measurements with 532 nm 1.5 MHz laser scanning with a speed of 10 m/s or 3 shots per spot.

It was clear from the experiments that a slow scanning speed, above 3 shots per spot, led to re-deposition of debris and decomposed vapour near the ablated area (see the dark rims around the ablated area in Fig. 4). This undesired effect with slow scanning was due to absorption of the laser radiation in the ablated vapour and possibly, due to accumulation of heat, as the effect was not observed with a fast scanning speed above 0.35 m/s corresponding to 1 shot/spot exposure with 104 Hz repetition rate laser and ∼20 m/s with 1.5 × 106 Hz repetition rate. The second series of experiments aimed at measuring the ablation threshold Fth . The ablation threshold was determined in a standard way (Stuart et al. 1995) by fitting the dependence of the ablated depth per pulse in a semi-logarithmic plot and approximating the curve to a ‘zero’ ablation thickness, which was taken to be a single atomic layer thickness of the order of 0.3 nm (Fig. 5). The ablation threshold was found to be 0.25 ± 0.1 J/cm2 (the error is due to the different samples used) for 12 ps pulses; and ∼0.10 ± 0.05 J/cm2 , for 500 fs pulses. We noted that, with both 500 fs and 12 ps pulses, discoloration is observed at the fluence close to, but lower the ablation threshold (see Fig. 3), second pattern at 0.21 J/cm2 . There was no discoloration at laser fluences below 0.5 Fth (see Fig. 3), patterns at 0.05 J/cm2 and 0.09 J/cm2 . 4

However, the need for tightly focused, low-energy laser pulses introduces new challenges in precise manipulation of the laser beam. Fast scanning of the beam over the treated surface is one of these challenges to be addressed. We demonstrate here that scanning patterns and the scanning speed of the laser beam are important issues in laser cleaning of surfaces with high repetition rate, ultrashort laser pulses. The morphology of the ablated areas strongly depends on the homogeneity of the scanning pattern, while the short time between pulses leads to a need for a relatively high scanning speed. To avoid thermal accumulation, the scanning speed should also be faster than the heat diffusion rate, which for most organic materials is above 1 m/s. We show that a constant velocity alternating spiral and its modification, a moving racetrack pattern, provide the best practical options for homogeneous coverage of the treated area. In spite of the obvious advantages of applying ultrafast, high-repetition-rate lasers to the conservation of artworks and heritage objects, there are still a few important issues to be addressed. For instance: – using a top-hat profile for the laser beam will further improve the uniformity of the surface illumination and the morphology of the treated area; – fast scanning and fast removal of the ablated vapour is required to avoid redeposition of ablated material; – laser beam delivery and precise focusing for cleaning of three-dimensional objects; – ultrafast lasers must be optimised for use in conservation, as to date they have only been used in experimental facilities and are prohibitively expensive.

CONCLUSIONS

High precision treatment of surfaces and minimum invasiveness are the main advantages of using shortpulse, high-repetition rate lasers for laser cleaning. Contaminant material can be removed in individual layers of precise depth, with sub-micron precision.

52

(ed), Pulsed Laser Deposition of Thin Films: Applications in Electronics, Sensors, and Biomaterials, 99–130. Hoboken, New Jersey: John Wiley & Sons. Juhasz, T., Djotyan, G., Loesel, F. H., Kurtz, R. M., Horvath, C., Bille, J. F. & Mourou, G. 2000. Laser Phys. 10: 495. Loesel, F. H., Fischer, J. P., Gotz, M. H., Horvath, C., Juhasz, T., Noack, F., Suhm, N. & Bille J. F. 1998. Appl. Phys. B 66: 121. Luther-Davies, B., Kolev, V. Z., Lederer, M. J., Madsen, N. R., Rode, A. V., Giesekus, J., Du, K.-M. & Duering M. 2004. Table-Top 50 W Laser System for Ultra-Fast Laser Ablation, Appl. Phys. A 79: 1051–1055. Luther-Davies, B., Rode, A. V., Madsen, N. R. & Gamaly, E. G. 2005, Picosecond high repetition rate pulsed laser ablation of dielectrics: The effect of energy accumulation between pulses, Optical Engineering 44: 051102. Pouli, P., Bounos, G., Georgiou, S. & Fotakis, C. 2005. Femtosecond Laser Cleaning of Painted Artefacts; Is this the Way Forward? LACONA VI Conference Proceedings, Vienna, Austria. Rode, A. V., Gamaly, E. G., Luther-Davies, B., Taylor, B. T., Dawes, J., Chan, A., Lowe R. M. & Hannaford, P. 2002. Subpicosecond laser ablation of dental enamel, Journ. Appl. Phys. 92: 2153–2158. Stuart, B. C., Feit, M. D., Rubenchik, A. M., Shore, B. W. & Perry M. D. 1995. Phys. Rev. Lett. 74: 2248–2251.

Nevertheless, with the fast development of powerful and compact femtosecond lasers, ultrafast laser ablation has the potential to become a standard tool in the conservation armoury and a key technique for conserving some previously untreatable artworks and heritage objects. ACKNOWLEDGEMENTS This work is supported by the Australian Research Council through the Linkage Project Scheme. REFERENCES Du, D., Liu X, Korn, G., Squier, J. & Mourou G. 1994. Appl. Phys. Lett. 64: 3071–3073. Feit, M. D., Rubenchik, A. M., Kim, B. M., Da Silva, L. B. & Perry, M. D. 1998. Appl Surf. Sc. 127–129: 869–874. Gamaly, E. G., Rode, A. V., Luther-Davies, B., & Tikhonchuk, V. T. 2002. Ablation of solids by femtosecond lasers: ablation mechanism and ablation thresholds for metals and dielectrics, Phys. Plasmas 9: 949–957. Gamaly, E. G., Rode, A. V., & Luther-Davies, B. 2007. Ultrafast laser ablation and film deposition. In R.W. Eason

53

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Removal of unwanted material from surfaces of artistic value by means of Nd:YAG laser in combination with Cold Atmospheric-Pressure Plasma C. Pflugfelder, N. Mainusch & W. Viöl HAWK, University of Applied Sciences and Arts, Göttingen, Germany

J. Ihlemann Laser-Laboratorium Göttingen, Göttingen, Germany

ABSTRACT: The removal of unwanted material such as spray lacquers (graffiti), organic residues or corrosion products is a very demanding task, especially in the case of sensitive surfaces and objects with historical value that conservators usually deal with. Since many years Nd:YAG lasers are applied to clear away specific substances. Recently the cleaning potentials of a commercially available plasma jet which emits a “cold” stream of exited gas were investigated at HAWK. The application of chemically reactive plasma species that decompose organic and inorganic compounds supplemented by laser pulses is a new approach to solve specific “cleaning problems”. Ongoing tests are being performed within the research project “PROKLAMO”, a collaboration of three scientific facilities and eight companies. The hybrid system permits the removal of organic binding media respectively varnishes and paint layers that are usually restricted to (expensive) UV laser ablation. Furthermore metallic corrosion products on archeological artifacts can be reduced and removed as well.

1

INTRODUCTION

of dissolved matter can complicate any thorough extraction. Furthermore whenever organic solvents, acids or alkali agents are to be used, danger to health and environment exists (Dignard et al. 2005). The use of an atmospheric-pressure plasma jet as a device for surface cleaning has been investigated in the course of a research project at HAWK between 2004–2006. Specific substances were removable with a plasma jet due to the thermal and especially the chemical impact of free radicals of the air-fed jet. On the one hand polymers like acrylic resins and polyester based sprays could be removed from smooth surfaces whereas other substances such as nitrocellulose were resistant (Dignard et al. 2005). Furthermore archaeological metal objects were treated by plasma with the goal to diminish oxidic crusts. As process gases, mixtures containing hydrogen were used. Especially in case of a strongly damaged silvered copper coin we have obtained good cleaning results: plasma reduces silver oxides very well. In contrast to this, copper and iron oxides proved to show weaker reaction to the plasma. In the case of completely corroded iron objects any plasma treatment proved to be dangerous because it destabilizes the structure and causes severe decomposition of the material (Pflugfelder 2005). As we sum up limitations and benefits of an atmospheric-pressure plasma jet and a Nd:YAG

A common goal of all cleaning procedures is the removal of unwanted material from a substrate. In conservation of archaeological metal artefacts, iron concretion or dark corrosion products have to be taken off in order to retain details from, for example, the coins original design. All such cleaning tasks have to be carried out selectively and without harming any substance of artistic value. Since the mid 1970’s lasers have been tested for a wide range of conservation activities (Asmus 1973). Many case studies prove short-pulse (5–10 ns) Nd:YAG-lasers to be a tool to overcome the aforementioned disadvantages (Cooper 1998). Essential for their successful employment is the appropriate absorption of laser energy from the unwanted material. Besides this, the underlying components with artistic value have to exhibit sufficient resistance to the laser impact so that selective ablation is possible and, in the most desirable case, a “self-limiting-cleaningprocess” can take place (Dickmann et al. 2005). Conventional cleaning methods include abrasive techniques and wet-chemical procedures. A disadvantage of mechanical intervention is minor accuracy and selectivity. Wet-chemical techniques bear the risk of the solvent’s uncontrolled permeation through a substrate due to capillarity. Unintended deposition

55

laser, it becomes obvious that the coupling of these devices should be an smart approach that offers new possibilities in dealing with specific cleaning problems. By means of initial investigations aimed at the removal of polymers we already specified cases in which the laser-plasma-coupling increases working efficiency (Mainusch 2006). Objective of this contribution now is the presentation of new results from cleaning attempts of metal artefacts.

2 2.1

EXPERIMENTAL PART Experimental setup Figure 1. Experimental setup.

In the experimental set-up the laser source was a solidstate, flash lamp-pumped Nd:YAG laser (LUMONICS, model MiniQ) that operates at 1064 nm. The intensity of the Q-switched laser beam (with pulse duration of 4 ns) was adjusted by means of an external trigger generator that provided 30 V signals for the activation of the Pockels Cell. These pulses were synchronized with a signal from the laser’s flash lamps. The time gap in between flash lamp signal and trigger pulse (Q-switch delay) was to be modified (125 µs to 250 µs). By this, fluences in the range from 200 mJ/cm2 to 800 mJ/cm2 and a laser spot size of approximately 0.04 cm2 could be generated. As depicted in Figure 1 the plasma stream is directed upon the target with an angle of incidence of around 30◦ . The plasma source (model “Plasma Blaster” with generator “V06”, manufactured by TIGRES GmbH, Rellingen) provides a potential-free plasma jet resulting from a gas discharge that is ignited between a centered electrode and a grounded nozzle. The plasma generator produces a voltage at about 10 kV that is applied with a frequency of 20–40 kHz to continuously feed the gas discharge. The currents are limited to 10–20 mA by means of a specific electric control. Therefore only transient arcing between the electrodes occurs and subsequently plasma with moderate escape temperatures is being generated. The plasma jet can be fed with either air or various process gases such as argon, argon-hydrogenmixtures, nitrogen etc.The gas flow rate was controlled by means of a variable area flowmeter. Throughout our experiments, the plasma jet was constantly driven at output power 140 W. A nozzle with a diameter of 1 mm was implemented. By this, luminance and extension of the stream as function of the gas flow rate 30 l/min could be investigated.

Figure 2. Original state of the coin.

containing up to 5% hydrogen. Thus a significant amount of free hydrogen radicals with the capability to chemically reduce oxidised metal surfaces is be generated.

2.3 Specimen A variety of heavily spoiled and corroded metal coins without any historical value were subjected to the combined laser-plasma jet treatment. Figure 2 represents such a coin, a “Dinara” from former Yugoslavia. The coin apparently consists of a copper alloy yielding a greenish copper patina and a white veil on its surface. After a series of tests, an archeological finding that approximately dates back to the year 1000 was treated as well. As to be seen in Figure 3, there is a leftover of crust in the upper part of the coin. White and green efflorescence with the brown corrosion products disfigure the surface. The pattern of its coinage is hardly readable.

2.2 Process gas As mentioned before, the plasma jet is designed for the use of various process gases. All through our tests, the plasma jet was supplied with argon based gas mixtures

56

Figure 5. State of archaeological metal object after cleaning process; plasma-laser combination on the left and only laser on the right.

Figure 3. Original state of the archaeological metal object.

loose efflorescence followed by laser ablation (right half) and laser-plasma jet application (left). We chose a low laser fluence of 0.3 J/cm2 , because the finding pieces were fairly fragile. Once again, the application of the coupled device results in different cleaning effects in comparison to only laser ablation.The slightly brownish discoloration as well as corrosion remains that are visible in the right half of the coin cannot be detected in the left zone anymore. Moreover the homogenous appearance of the laser-plasma jet treated area is striking. Figure 4. State of coin after cleaning process; plasma laser combination on the left, only laser middle and only plasma on the right.

3

4

CONCLUSIONS

According to the documented cleaning effects, we believe that there is a significant contribution of plasma species in the cleaning procedure of superficially corroded metals. By means of former spectrographic investigation on the plasma stream, a certain amount of hydrogen atoms could be detected, whenever generating plasma with hydrogen gas mixtures (Pflugfelder 2006). It is well known that hydrogen possesses the capability to chemically reduce metal oxides and thus in many cases our approach should be promising. From the conservator’s point of view the phenomenon “homogeneity” of the cleaned area due to the coupling as mentioned above is interesting since any cleaning intervention in conservation has to result in a homogenous appearance of the piece of art. The fact that the plasma generator provides a jet that continuously blows onto the surface has to be regarded positively: the jet gradually disintegrates respectively reduces unwanted material and this favours a sensitive and controllable intervention.

RESULTS

Figure 4 depicts the “Dinara” after the treatment. Three areas can be distinguished: at the very right side only plasma jet treatment was performed. Compared to the original state only a little change is visible. The adjacent area documents the effect of stand alone laser ablation. Ablation has been carried out with a fluence of 0.5 J/cm2 to 0.6 J/cm2 . The repetition rate was 10 Hz running a stage speed of 3 mm/s. As a cleaning result we note that by single scanning the superficial alteration could be diminished quite well. A grey veil is preserved. On the left half of the object an intensified cleaning result was obtained by means of a coupled laser-plasma jet application even though the laser fluence amounted to only 0.4 J/cm2 . In Figure 5 there are two zones to be distinguished that resulted from an initial (mechanical) removal of

57

REFERENCES

Mainusch, N. 2006. Plasma Jet Coupled with Nd:YAG Laser: A NewApproach to Surface Cleaning. Proceedings of 10th International Conference on Plasma Surface Engineering (PSE), S33-S38, WILEY-VCH, Weinheim. Pflugfelder, C. 2005. Einsatz von Atmosphärendruckplasma im Bereich der Restaurierung. Diplomarbeit, HAWKHHG. Pflugfelder, C. 2006. Cleaning Wall Paintings and Architectural Surfaces by Plasma, Proceedings of 10th International Conference on Plasma Surface Engineering (PSE), S516–S521, WILEY-VCH, Weinheim.

Asmus, J. F., Murphy, C. G. & Munk, W. H. 1973. Studies on the interaction of laser radiation with art artifacts. Proceedings of SPIE. Cooper, M. 1998. Laser Cleaning in Conservation: An Introduction. Butterworth-Heinemann. Dignard, C., Lai, W., Binnie, N., Young, G., Abraham, M. & Scheerer, S. 2005. Cleaning of Soiled white feathers using the Nd:YAG laser and traditional methods. Proceedings of Lasers in the Conservation of Artworks conference, LACONA V, Springer Verlag-Berlin, Heidelberg. Dickmann. K., Fotakis, C. & Asmus, J. F. eds. Proceedings of Lasers in the Conservation of Artworks conference, LACONA V, Springer Verlag-Berlin, Heidelberg.

58

Analytical Techniques

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Optical coherence tomography for structural imaging of artworks P. Targowski, M. Góra, T. Bajraszewski, M. Szkulmowski, M. Wojtkowski & A. Kowalczyk Institute of Physics, Nicolaus Copernicus University, Torun, Poland

B. Rouba, L. Tymi´nska-Widmer & M. Iwanicka Institute for the Study, Restoration and Conservation of Cultural Heritage, Nicolaus Copernicus University, Torun, Poland

ABSTRACT: The Optical Coherence Tomography (OCT) technique is applied to image internal structure of an oil painting in non-destructive and non-invasive way. Different varnish layers, glaze and paint layers are clearly visible. This technique can also determine the deep location of an artist’s signature providing a method of confirmation of its authenticity.

1

INTRODUCTION

(Liang et al. 2005). In that year, OCT was presented for the first time at LACONA conference series, applied to imaging of varnish and paint layers (Gorczy´nska et al. 2007, Szkulmowska et al 2007, Arecchi et al. 2007). All applications of OCT to artwork diagnostics were recently reviewed by Targowski et al. (2006). Lately, a new use of the OCT technique to monitor laser ablation of varnish on easel paintings has been reported (Góra et al. 2006, Targowski et al. 2007).

Determination of internal structure of the artwork is essential not only for the purpose of its documentation or technical and historical research, but also for understanding causes of the decay and planning a proper conservation-restoration treatment. Routine method of stratigraphic examination is based on collection of samples from the artwork structure. The obvious limitation of this procedure is its invasive and destructive nature. Additionally, it can be performed in painting regions of lesser importance, which obviously disqualifies its use in inviolable regions of painting such as the areas of signature. Moreover, because of the large variability of the structure, the physical sample may not represent its neighbourhood. Optical Coherence Tomography (OCT) is an alternative method which gives the possibility of sampling in a fast and non-invasive way, for an unlimited number of locations in any region of the painting surface. Its utilization is, however, limited to transparent or semitransparent structures of weak absorption and scattering of infrared light. This relatively new, but already well-established technology, mostly used in biological research and diagnostics, especially in ophthalmology (Huang et al. 1991, Srinivasan et al. 2006) was quickly transferred to materials science (Stifter 2007, for review). OCT was first applied to obtain images of internal structures within archaic jades (Yang et al. 2004). In the same year, the first application of OCT to examination of paintings, namely to varnish layer imaging and to profilometry of picture surface, was reported (Targowski et al. 2004). In the next year, OCT was used to image underdrawings and model paint layers

2

INSTRUMENTATION

The tomograms shown in this paper have been obtained with a prototype SOCT instrument based on an optical fibre Michelson interferometer setup (Fig. 1), constructed in the Nicolaus Copernicus University. A broadband (λ = 50 nm, central wavelength 845 nm) superluminescent diode (LS) was employed as the light source. The light of high spatial but low temporal coherence is launched into a single mode 50 : 50 fibre coupler (FC) through an optical isolator (OI). The optical isolator protects the light source from the light back reflected from the elements of the interferometer. Light beam is then divided by the fibre coupler into two arms of the interferometer. The light propagating in a reference arm passes through a polarization controller (PC) to provide the optimal conditions for interference, the neutral density filter (NDF) for adjustment of the power of light to achieve the shot noise limited detection and a block of glass acting as a dispersion compensator (DC). The light is then back reflected from the stationary reference mirror (RM) to the reference arm fibre and coupler (FC). The sample arm

61

Figure 2. A) “Portrait of Sir James Wylie”, oil on canvas, 19th century (fragment); B) UV induced luminescence of the same fragment. Restorer’s interventions are clearly visible as dark spots. Places where OCT tomograms were recorded are indicated by letters a–f.

Figure 1. A top view of the experimental setup.

in varnish and paint layer. The transverse resolution depends mostly on the optical properties of the system and is kept below 20 µm. The sensitivity of the system is 104 dB. The optical power of the incident beam at the surface of the object ranges between 500 and 2000 µW. In the tomograms presented, the intensity of scattered and/or reflected light from the internal structures within the sample is coded in a grey scale – the darker the shade, the higher level of reflectivity of the structure. Due to the fact that all originally measured axial distances are optical ones, a conversion to real lengths is required. To calibrate the in-depth axis, the group refractive index of the medium was assumed to be equal to 1.5.

comprises transversal scanners (X–Y) and lens which form the measuring head. The light beam is scanned across the object and backscatters and/or reflects from the elements of its structure and returns to the coupler FC. The light beams returning from the reference mirror and from the sample are brought to interference at the output of the interferometer and analyzed by a customized spectrometer. It consists of a volume phase holographic grating (DG) with 1200 lines/mm and achromatic lens (SL) which focuses the spectrum on a 12 bit line scan CCD camera (2048 pixels, 12 bit A/D conversion, Atmel). The spectral fringe pattern registered by this detector is then transferred to a personal computer (COMP). This signal, after Fourier transformation, yields one line of the crosssectional image (A-scan). The A-scan carries information about the location of structural interfaces in the object along the path of the penetrating beam. Scanning across the sample enables collecting 2D slice cross-sections (B-scan). Additional scanning in the perpendicular direction gives 3D information about a structure. The acquisition process and scanning protocols are controlled by a custom-designed compact electronic driving unit. To define precisely the examined area, an industrial camera registering the surface of the object is coupled to the OCT instrument. A high-density image (B-scan) is composed of at least 1000 A-scans per one millimetre and covers an area of 3.6 mm in depth. The exposure time (during which the object is illuminated) is 40 µs per singleA-scan and it will increase to 200 ms for typical B-scan comprising 5000 lines. The axial resolution of the system (equals to one half of the coherence length of the light) is 6 µm

3

RESULTS

3.1 Stratigraphy of an oil painting A 19th century Portrait of Sir James Wylie (Fig. 2) was chosen for the OCT examination due to its plain and typical painting techniques. The original varnish layer was preserved during the last conservation-restoration procedures which took place about 30 years ago and included vast reconstruction of the composition. Finally, a thin layer of synthetic final varnish was applied onto the whole surface of the painting. The presence of original glazes with two layers of varnish – the original thick one in combination with a thin but already aged restorer’s varnish – brought an opportunity to investigate the possibility of imaging these layers as separate morphological structures by means of OCT.

62

Figure 3. Tomograms of the inner structure of the painting “Portrait of Sir James Wylie”. Orientation of tomograms corresponds to direction of scanning. All images are scanned over 12 mm. Arrows point to following objects: 1: air-varnish interface; 2: re-storer’s final varnish layer; 3: restorer’s varnish–original varnish interface; 4: original varnish layers; 5: interface between layers of original varnish; 6: varnish–paint layer interface; 7: glaze layer; 8: opaque paint layer.

However, paint layers are in many cases transparent enough to enable deeper imaging. In Figures 3a, c–f one can observe the cross-sectional view of at least one layer of glaze underneath the varnish. The thickness of glaze layers in different parts of painting’s composition varies significantly, as may be noticed from the comparison of Figures 3c and 3e. This not only provides information about the artist’s original technique, but also may have great importance for planning conservation-restoration treatment. Detection and precise location of glaze under darkened or yellowed varnish will help decreasing risk of irreversible damage to the glaze layer during a process of varnish removal. In most of the tomograms, it proved to be possible to detect the presence of at least two films within the original varnish layer. Additionally, in Figures 3a and 3b, both taken across the boundary between the original and reconstructed paint layer, a thin superficial layer of restorer’s final varnish is revealed.

Figure 3 presents tomograms taken at locations indicated by letters a–f in Figure 2B. The light penetrates the object (Fig. 3) either from left (a–d) or from top (e, f ). As an example of diversity of the painting’s inner structure, the tomogram corresponding to Figure 3a may be considered. An interface between air and the surface of the painting is imaged as a thick black contour (1). The thin but recognizable light strip below is a layer of restorer’s final varnish (2). Then, two layers of original varnish (4) are visible as white strips due to its high transparency. The last visible layers are glaze (7) and opaque paint layer (8). Obviously, not all of the above described structures may be found in every tomogram, depending on the painting region examined. For instance, Figure 3b was taken from the area where the paint layer mostly consists of lead white, a pigment of great opacity and infrared impermeability. Therefore, there is no possibility of OCT imaging through this layer.

63

Such counterfeit, as shown in Figure 4, can be discovered by means of the OCT technique. In the OCT cross-sectional image, the exact location of the signature is clearly visible. It proves not to be lying directly on the paint layer, but rather “hovering” above it. The white gaps under the edges of the signature give evidence that it is localized between two varnish layers and, hence, it is probably not authentic.

4

CONCLUSIONS

In this paper, we have shown the applicability of the OCT technique to unravelling the sequence of semitransparent strata of painting in a non-invasive way. Varnish layers imaging by means of OCT has proven to deliver quite detailed information: the possibility of imaging and distinguishing two chemically and historically different types of varnish and up to three consecutively applied films within the original varnish layer. Beyond varnish, semitransparent layers of glazes are also visible in OCT tomograms. The last structure available for OCT examination is the upper boundary of opaque paint layer. Especially interesting and promising application of structural imaging of paintings seems to be examination of the inviolable region of artist’s signature in order to confirm its authenticity.

Figure 4. Example of OCT examination of the signature region. In the OCT cross-sectional image, the exact location of the signature is clearly visible inside the varnish layer and thus the counterfeit becomes obvious.

Ancient varnishes, since they consist of natural resins and oils, are in most cases inclined to form layers of greater thickness than contemporary synthetic resin varnishes. Due to the natural ageing process, these layers present more inhomogeneous structure and thus scatter light more efficiently than the contemporary ones. These features of varnish coatings make the OCT technique a suitable tool for quick and non-destructive analysis of the range of restorers’ interventions, which not always may, due to various reasons, be revealed during routine UV-fluorescence examination.

REFERENCES Arecchi, T., Bellini, M., Corsi, C., Fontana, R., Materazzi, M., Pezzati, L. & Tortora, A. 2007. Optical coherence tomography for painting diagnostics. In J. Nimmrichter, W. Kautek and M. Schreiner (eds.). Lasers in the Conservation of Artworks, LACONA VI Proceedings, Vienna/Austria, Sept. 21–25, 2005, Berlin-HeidelbergNew York: Springer Verlag. Gorczyñska, I., Wojtkowski, M., Szkulmowski, M., Bajraszewski,T., Rouba, B., Kowalczyk,A. &Targowski, P. 2007. Varnish Thickness Determination by Spectral domain Optical Coherence Tomography. In J. Nimmrichter, W. Kautek & M. Schreiner (eds.). Lasers in the Conservation of Artworks, LACONA VI Proceedings, Vienna/Austria, Sept. 21–25, 2005, Berlin-HeidelbergNew York: Springer Verlag. Góra, M., Targowski, P., Rycyk, A. & Marczak, J. 2006. Varnish ablation control by Optical Coherence Tomography. Laser Chemistry doi:10.1155/2006/10647, http://www.hindawi.com/journals/lc/. Huang, D., Swanson, E. A., Lin, C. P., Schuman, J. S., Stinson, W. G., Chang, W., Hee, M. R., Flotte, T., Gregory, K., Puliafito, C. A. & Fujimoto J. G. 1991. Optical coherence tomography. Science 254: 1178–1181 Liang, H., Cid, M., Cucu, R., Dobre, G., Podoleanu, A., Pedro, J. & Saunders, D. 2005. En-face optical coherence tomography–a novel application of

3.2 Examination of artist’s signature Another interesting and not yet well explored application of OCT to the analysis of paintings is examination of artists’ signatures of uncertain authenticity. If a forger draws a signature on an original varnish layer and then intentionally covers it with another varnish of high fluorescence, the routine UV examination may not reveal the fake. A model sample of oil painting was prepared and covered with dammar varnish (Maimeri). Subsequently, to imitate a common forgery technique, the signature was drawn on the varnish surface and then the whole sample was covered with a few layers of the same varnish.

64

non-invasive imaging to art conservation. Optics Express 13: 6133–6144. Srinivasan, V. J., Wojtkowski, M., Witkin, A. J., Duker, J. S., Ko, T. H., Carvalho, M., Schuman, J. S., Kowalczyk, A. & Fujimoto, J.G. 2006. High-definition and 3-dimensional imaging of macular pathologies with highspeed ultrahigh-resolution optical coherence tomography Ophthalmology 113: 2054 2065.e3. Stifter, D. 2007. Beyond biomedicine: a review of alternative applications and developments for optical coherence tomography. Appl. Phys. B DOI: 10.1007/s00340-0072743-2. Szkulmowska, A., Góra, M., Targowska, M., Rouba, B., Stifter, Breuer, E. & Targowski, P. 2007. The Applicability of Optical Coherence Tomography at 1.55 mm to the Examination of Oil Paintings. In J. Nimmrichter, W. Kautek & M. Schreiner (eds.).Lasers in the Conservation of Artworks, LACONA VI Proceedings, Vienna/Austria,

Sept. 21–25, 2005, Berlin-Heidelberg-NewYork: Springer Verlag. Targowski, P., Rouba, B., Wojtkowski, M & Kowalczyk, A. 2004. The application of optical coherence tomography to non-destructive examination of museum objects. Studies in conservation 49: 107–114. Targowski, P., Góra, M. & Wojtkowski, M. 2006. Optical Coherence Tomography for Artwork Diagnostics. Laser Chemistry doi:10.1155/2006/35373 http://www.hindawi. com/journals/lc/. Targowski, P., Marczak, J., Góra, M., Rycyk, A. & Kowalczyk,A. 2007. Optical CoherenceTomography forVarnish Ablation Monitoring.Proc. of SPIE 6618: 661803-1 – 661803-7. Yang, M. L., Lu, C. W., Hsu, I. J. & Yang, C. C. 2004. The use of optical coherence tomography for monitoring the subsurface morphologies of archaic jades. Archeometry 46: 171–182.

65

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Atmospheric Pressure Laser Desorption Mass Spectrometry based methods for the study of traditional painting materials M.P. Licciardello, R. D’Agata & G. Grasso Dipartimento di Scienze Chimiche, Università di Catania, Catania, Italy

S. Simone Dipietro Automazione S.r.l., C. da Cava Sorciaro, Priolo G. (SR), Italy

G. Spoto Dipartimento di Scienze Chimiche, Università di Catania, Catania, Italy Istituto Biostrutture e Bioimmagini, CNR, Catania, Italy

ABSTRACT: The study of ancient works of art is a very challenging task mainly due to the impossibility of applying experimental approaches that could damage anyhow the object from which analytical information has to be obtained. Spatially resolved analytical methods have significantly enhanced our capacity to study ancient materials since they cause minimal and at times no damage to the studied object. Unfortunately, only few analytical techniques operating within the requested spatial resolution are applicable for the investigation of the organic components of artistic and archaeological objects. In this scenario, Atmospheric Pressure/Matrix Assisted Laser Desorption Ionization-Mass Spectrometry (AP/MALDI-MS) has already proven to be a very valuable technique for the study of ancient materials as it combines the good spatial resolution of the conventional MALDI to the possibility of working in air. In this work we present results from the AP/MALDI-MS investigation of the indigo dye and the carmine and brazilwood lakes. All the studied systems can be found in traditional paintings and the experimental conditions used are aimed to simulate real ancient materials.

1

INTRODUCTION

tools. The destructive approach thus remains as a last resort for the extraction of analytical information from artistic and/or archaeological samples. Tremendous improvements have been made as regards the scope and efficiency of today’s analytical instruments. This has led to the development of new analytical methodologies that satisfy specific requirements to a greater degree such as microdestructiveness or non-destructiveness of the sample to be analysed (Ciliberto & Spoto 2000, Spoto 2007). A wider range of information is now available and a greater sensitivity and reproducibility of analyses is thus ensured. In this context, the use of spatially resolved analytical techniques have provided new opportunities for micro-destructive and, at times, completely non-destructive analyses thus opening up new diagnostic approaches for the study of samples of artistic and/or archaeological importance (Spoto et al. 2000). They have also amplified the range of analytical information obtainable from ancient and valuable objects (Spoto 2002). Spatial resolution allows analysing tiny fragments of samples scraped from the object of interest with minimal damage to the artefact itself. Moreover, the

All scientists involved in studies concerning works of art or samples of archaeological interest will recognize the importance of making an appropriate selection of the analytical method to be used in their studies. The main questions usually raised concern how the proposed analytical procedure will affect the integrity of the object to be examined. From this point of view, only those techniques which do not alter the integrity and appearance of artistic and/or archaeological objects are eligible as “ideal” techniques. Techniques which operate in situ, making sample-taking unnecessary, come close to this ideal. In the attempt to find a balance between the requirements of scientific methods and the need to maintain the integrity of the object under study, the only alternatives to in situ analysis require the object itself to be placed in the analysing chambers of the analytical instrument or tiny fragments of samples to be scraped from its surface. The former approach cannot be applied in all cases, since only small objects such as coins, certain jewellery and statuettes are of a size and shape which will fit those of common analytical

67

of artistic materials is the need of placing these samples under vacuum. The low-medium vacuum conditions required for the MALDI-MS analysis significantly limit its potential for in situ analysis of works of art and archaeological objects. The recently introduced Atmospheric Pressure (AP) MALDI-MS (Laiko et al. 2000) combines the advantages of the vacuum MALDI-MS analytical approach to the possibility of analysing samples in air. (Schneider et al. 2005) Two important advantages offered by AP/MALDI with respect to vacuum MALDI are the ability to produce lower internal energy molecular ions with minimal fragmentation (softness) (Gabelica et al. 2004) and the decoupling of the ion source from the mass analyser. In order to demonstrate the potential offered by AP/MALDI-MS in the study of ancient objects and materials, we recently presented an AP/MALDI study of the most important ink in western history: the irongall ink. (D’Agata et al. 2007) The study demonstrated that AP/MALDI can be used to identify the main gallotannic components of the ink by analysing in air strokes of ink directly from paper and parchment. To further expand our basic knowledge of the AP/MALDI potential in the identification of organic materials used in art, in this work we present results from the study of the indigo dye and the carmine and brazilwood lakes. All the studied systems were used in traditional painting. The study was conducted by using a commercially available AP/MALDI source previously modified in order to allow the direct analysis of real objects.

in situ analysis of microscopic areas of the artwork may also be accomplished by spatial resolution, thereby allowing the extraction of a wide range of valuable analytical information that can sometimes be imaged into 2D or 3D graphics. In spite of the powerful support provided by the today available analytical tools, the micro-destructive study of organic materials constituting the structure of works of art and archaeological objects is still a challenge. These organic materials are mainly natural products and, therefore, they are composed of complex mixtures of molecular and biomolecular components (Mills & White 1994). Detailed compositional information from such materials is obtained by making use of a wealth of instrumental methodologies. In this scenario, spatially resolved analytical methods offer valuable tools for the analysis of works of art with a micro-destructive approach. Unfortunately, only few analytical techniques operating within the requested spatial resolution are applicable for the investigation of the organic components of artistic and archaeological objects. Among them Raman microscopy (Vandenabeele et al. 2007, Clark 2007), secondary ion mass spectrometry (Darque-Ceretti & Aucouturier 2004) and Fourier Transform-IR micro-spectroscopy (Salvado et al. 2005) have been shown to be the most powerful analytical tools that are available today. However, the latter have limitations that prompt the quest for further, hopefully more versatile, spatially resolved analytical methods. In recent years, the potential showed in the above mentioned field by spatially resolved mass spectrometry (MS) techniques that make use of ionic sources based on direct laser desorption ionization (LDI) has been investigated (Boon & Learner 2002; Grim & Allison 2004). However, direct LDI is only effective in the study of a limited range of materials, while the use of matrices that assist the ionization process induced by the laser desorption (Matrix Assisted LDI, MALDI) has expanded the applicability of LDI-based MS methods to the field of spatially resolved studies of organic components from works of art. The matrix is a compound that absorbs light at a given laser wavelength and allows compounds that do not absorb laser light to be desorbed and ionized without much fragmentation. In other words, it protects and assists the analyte during the desorption and ionization processes and very often it is only by choosing the appropriate matrix for the particular system investigated that good quality mass spectra can be obtained (Williams et al. 2007). Examples of the applicability of MALDI-MS to analysis of pigments (Maier et al. 2004), siccative oils (Van Den Berg et al. 2004), proteinaceous binders (Tokarski et al. 2006, Kuckova et al. 2007), and varnishes (Zumbühl et al. 1998) are indeed reported in the literature. On the other hand, a significant disadvantage of using MALDI-MS for spatially resolved study

2

EXPERIMENTAL

Indigo, carmine naccarat, and brazilwood were purchased from Kremer (Germany). α-cyano-4hydroxycinnamic acid (CHCA), trifluoro acetic acid (TFA), acetonitrile (C2 H3 N), methanol (CH3 OH) were purchased from Sigma-Aldrich. All the AP/MALDI-MS measurements were carried out by using a Finnigan LCQ Deca XP PLUS (Thermo Electron Corporation, USA) ion trap spectrometer which was fitted with a MassTech Inc. (USA) AP/MALDI source. The source was modified in order to allow the direct analysis of real objects (Fig. 1). A home-built XYZ stage was used to place tridimensional objects on the focus of both the ion trap spectrometer and the laser source with a micrometric control. The sample positioning was controlled by home-built software operating in LabView framework that was also interfaced with two different digital cameras. A pulsed nitrogen laser (wavelength 337 nm, pulse width 4 ns, pulse energy 300 µJ, repetition rate up to 10 Hz, focus diameter ∼200 µm) was used as source. Laser power was attenuated to about 55%. Pulsed dynamic focusing (PDF) has been shown to improve S/N ratios in the AP-MALDI spectra,

68

Mulholland 2006) were carried out by applying a multipole offset voltage of +100 V for negative ions and of −100 V for positive ions. The matrix solution was a CHCA solution (1 mg/ml) in 30% TFA (0.1%) and 70% C2 H3 N. Typically AP/MALDI-MS experiments were carried out by mixing 1 µl of the pigment in CH3 OH with 5 µl of matrix solution. 1 µl of the final solution was spotted on theAP/MALDI plate and irradiated with the laser source after solvent evaporation. In situ experiments were carried out by spotting 0.5 µl of matrix solution (spot diameter about 1000–500 µm) or by depositing CHCA crystals directly on the surface to be analysed. Egg tempera was obtained by separating out the yolk of one chicken egg and by draining it into a container. Pigment was added to obtain the final tempera paint. Once ready, the paint mixture was applied on a flat glass surface and allowed to dry in sunlight. The paint layers were analysed by AP/MALDI-MS as both freshly prepared (after visual solidification of the paint medium, typically after two days) and after two months. To acquire the necessary fundamental knowledge about the AP/MALDI capability in the characterization of organic materials, organic dyes and pigments traditionally used in painting and dyeing were analysed. The studied organic systems were the indigo (a traditional organic dye also used as a pigment) and two lakes: carmine and brazilwood lake. It is worth noting that all the analyses were carried out by operating in air and with no previous chromatographic separation of the samples components. Similar experimental conditions must be assured when in situ spatially resolved analyses of solid objects are going to be planned.

Figure 1. Home-built XYZ stage used to place tridimensional objects on the focus of both the ion trap spectrometer and the laser source with a micrometric control. The sample positioning was controlled by home-built software operating in LabView framework that was also interfaced with two different digital cameras (a). Aluminum foil in contact with the sample ensured the application of the target voltage to the sample surface (b).

(Berkout et al. 2007) therefore a PDF module that imposes a delay of 25 µs between the laser pulse and the application of the high voltage to the AP/MALDI target was also used. The target voltage was applied to the sample surface by connecting an aluminum foil carefully placed on the sample surface in close proximity to the area to be analysed. The applied target voltage was 1.8 kV. The ion trap inlet capillary temperature was 200◦ C. Capillary and tube lens offset voltages of 30 and 15 V, respectively, were applied. Automatic Gain Control (AGC) was turned off and instead the scan time was fixed by setting the injection time to 220 ms and 5 microscans per scan. MS/MS scans were acquired using an isolation width of 5 m/z, activation qz of 0.250, activation time of 30 ms, and normalized collision energy (NCE) in the range 30–40%, dependent on the ion. (NCE is the amplitude of the resonance excitation RF voltage scaled to the precursor mass based on the formula: RF amplitude = [NCE%/30%] (precursor ion mass x tick amp slope + tick amp intercept), where the tick amp slope and tick amp intercept are instrument specific values. For our LCQ Deca, 35% NCE for m/z 1000 = 1.8 V.). In source collision-induced dissociation (CID) experiments (Baranov & Bohme 1996; Peterman &

3 3.1

RESULTS AND DISCUSSION Indigo

Natural indigo was traditionally obtained from a variety of plants. The most common indigobearing plants belong to the indigofera species (indigofera tinctoria). Independent of the plant source, the chemistry of dye extraction requires a fermentation stage during which an enzymatic hydrolysis of the original compounds present in the plant is followed by an oxidation process caused by the exposure to air. At the end a blue pigment is collected. Evidence of the use of indigoid dyes dates back to 2000 B.C. in Egypt. Indigo has been widely used as a dye or as colouring matter for blue paper in the renaissance or later drawings and, to some extent, as a pigment in paintings (Chiavari et al. 2005). It is stable when used with tempera medium while unstable with siccative oils.

69

O

H N

N H

O

Figure 2. Molecular structure of indigotin (MW = 262.2 Da). 50 40

Figure 4. Image of the indigo bearing paint layer. The brighter spots are due to the HCCA matrix crystals deposited on the different regions of the paint layer.

30

40

20

263.1

Relative Intensity (%)

Relative Intensity (%)

263.2

10 0 150

200

250

300

m/z Figure 3. Representative positive ion AP/MALDI spectrum obtained from the analysis of indigo.

30

20

10

0 200

The main chemical component of natural indigo is the indigotin (Fig. 2) (Andreotti et al. 2004). A representative AP/MALDI positive ion spectrum obtained from the natural indigo is shown in Figure 3. A quite evident signal at m/z 263.2 is present in the spectrum. The signal is attributed to the pseudomolecular species obtained from the protonation of the indigotin (MW = 262.2 Da). The spectrum shows a number of other signals most of which are due to ionic species formed by the CHCA matrix. To further verify the ability of AP/MALDI to identify organic pigments when present in complex matrices the indigo was mixed to an egg tempera (see Experimental section) and the obtained paint layer was analysed after its aging (Fig. 4). The mass spectra obtained from replicate analyses carried out by operating in air confirmed the presence of the protonated indigotin ion as shown in Figure 5 (m/z 263.1). The egg tempera was previously analyzed to check for the presence of ionic species having the same m/z observed for the indigo. A signal, due to unidentified ionic species formed by the tempera products, was observed at m/z 265.8 and did not affect the pigment identification (the experimental condition assured a spectral mass resolution lower that 0.5 Da).

265.8

220

240

260

280

300

m/z Figure 5. Representative positive ion AP/MALDI spectrum obtained from the analysis of indigo egg tempera paint layer. The signal at m/z 265.8 is attributed to ion species formed by the egg medium.

3.2 Carmine Plant parasites from the coccidea family have been traditionally used for dye extraction. The dried body of the egg-filled female scale insect was typically used in both Central Americas and Europe for dye extraction. Carmine used as a pigment in traditional painting was prepared by precipitating the aluminium complex of the cochineal extract dye (also known as crimson lake) (Schweppe & Roosen-Runge 1986). The chromophores in all scale insect dyes are derivatives of anthraquinone. The major constituent is carminic acid (MW = 492.2 Da, see Fig. 6) but the various species have characteristic fingerprints of anthraquinone minor components. Also in this case the AP/MALDI analysis provided a clear identification of the pigment, as ions at m/z 493.1 resulting from the protonation of the carminic acid were easily detected (Fig. 7).

70

CH3

O

OH

OH

HO

HOOC

Relative Intensity (%)

OH O

HO

OH

OH O

OH

Figure 6. Molecular structure of carminic acid (MW = 492 Da).

80 60 40

285.1

20

493.1

100

267.3

100

0 200

250

300

350

80

Figure 9. Representative positive ion AP/MALDI spectrum obtained from the analysis of the brazilwood pigment. The signal at m/z 285.1 is attributed to ions formed by the protonation of the brazilin ([M+H]+ ). The signal at m/z 267.3 is formed by dehydrated fragment ions ([M−H2 O+H]+ ).

60 40

10

20

Relative Intensity (%)

Relative Intensity (%)

m/z

0 420

450

480

510

540

570

m/z Figure 7. Representative positive ion AP/MALDI spectrum obtained from the analysis of the carmine lake pigment.

267.3

8 6 4

285.1 2

OH 0 200

HO O OH

300

350

Figure 10. Representative positive ion AP/MALDI spectrum obtained from the analysis of brazilwood egg tempera paint layer.

Figure 8. Molecular structure of brazilein (MW = 284.3).

3.3

250

m/z

O

TheAP/MALDI positive ion spectrum of the Brazilwood lake shows two intense signals at m/z 285.1 and m/z 267.3 (Fig. 9). The former is attributed to the pseudo-molecular ions formed by the protonation of the brazilin ([M+H]+ ). The latter is caused by dehydrated fragment ions ([M−H2 O+H]+ ). Both the signals allow a clear identification of the brazilwood dye also from the tempera paint film (Fig. 10).

Brazilwood

Soluble redwood dyes are extracted from various species of the genus caesalpinia and are collectively known as brazilwood. Their colouring principles are readily soluble in water and they are normally used as mordant dyes. Soluble redwoods were considered to have less dying properties than carmine or madder because they have poor fastness properties and therefore were usually used in combination with other dyes. Nevertheless, brazilwood has been reported to be used in European Renaissance paintings (Roy et al. 2004, Kirby et al. 2006). The main chromophore in brazilwoods is the brazilein (Fig. 8) which is obtained from the oxidation of the brazilin.

CONCLUSIONS AP/MALDI-MS investigation of traditional pigments was carried out in order to evaluate the applicability of such experimental approach to the analysis of organic

71

Gabelica, V. et al. 2004. Internal energy build-up in matrixassisted laser desorption/ionization. J. Mass Spectrom. 39: 579–593. Grim, D. M. & Allison, J. 2004. Laser Desorption Mass Spectrometry as a Tool for the Analysis of Colorants: The Identification of Pigments Used in Illuminated Manuscripts. Archaeometry 46: 283–299. Kirby, J. et al. 2006. Proscribed pigments in northern european renaissance paintings and the case of paris red. Preprints of the 21st IIC Congress: The object in context, crossing conservation boundaries. Munich. Kuckova, S. et al. 2007. Identification of proteinaceous binders used in artworks by MALDI-TOF mass spectrometry Anal. Bioanal. Chem. 388: 201. Laiko, V. V. et al. 2000. Atmospheric pressure matrixassisted laser desorption/ionization Mass Spectrometry. Anal. Chem. 72: 652. Maier, M. S. et al. 2004. Matrix-assisted laser desorption and electrospray ionization mass spectrometry of carminic acid isolated from cochineal. Int. J. Mass Spectr. 232: 225. Mills, J. S. & White, R. 1994. The Organic Chemistry of Museum Objects. Oxford: Butterworth-Heinemann, Ltd. Peterman, S. M. & Mulholland, J. J. 2006. A novel approach for identification and characterization of glycoproteins using a hybrid linear ion trap/FT-ICR mass spectrometer. J. Am. Soc. Mass Spectrom. 17: 168. Roy, A. et al. 2004. Raphael’s Early work in the National Gallery. National Gallery Technical Bulletin 25: 5–35. Salvado, N. et al. 2005. Advantages of the use of SR-FTIR microspectroscopy: applications to cultural heritage. Anal. Chem. 77: 3444–3451. Schneider, B. B. et al. 2005. AP and vacuum MALDI on a QqLIT instrument. J. Am. Soc. Mass. Spectrom. 16: 176. Schweppe, H. & Roosen-Runge, H. 1986. Carmine – Cochineal Carmine and Kermes Carmine. In R. Feller (ed.), Artists’ Pigments: A Handbook of their History and Characteristics: 255–283. Cambridge University Press. Spoto, G. et al. 2000. Probing archaeological and artistic solid materials by spatially resolved analytical techniques. Chem. Soc. Rev. 29: 429–439. Spoto, G. 2002. Detecting Past Attempts To Restore Two Important Works of Art. Acc. Chem. Res. 35: 652–659. Spoto, G. 2007. Chemical Methods in Archaeology. In KirkOthmer (ed.), Encyclopedia of Chemical Technology. 5th Edition. New York: John Wiley & Sons Inc. Tokarski, C. et al. 2006. Identification of proteins in renaissance paintings by proteomics. Anal. Chem. 78: 1494. Vandenabeele, P. et al. 2007. A Decade of Raman Spectroscopy in Art and Archaeology. Chem. Rev. 107: 675–686. Van Den Berg, J. D. J. et al. 2004. Effects of traditional processing methods of linseed oil on the composition of its triacylglycerols. J. Sep. Sci. 27: 181. Williams, T. I. et al. 2007. Effect of matrix crystal structure on ion abundance of carbohydrates by matrix-assisted laser desorption/ionization Fourier transform ion cyclotron resonance mass spectrometry. Rapid Communications in Mass Spectrometry 21: 807–811. Zumbühl, S. et al. 1998. A graphite-assisted laser desorption/ionization study of light-induced aging in triterpene dammar and mastic varnishes. Anal. Chem. 70: 707.

materials used in art. Indigo, carmine and brazilwood lakes were analysed either alone or from egg tempera paint films. MS results were interpreted so that the main chromophore species could be identified even in the case of tempera paint films, demonstrating the applicability of this approach for in situ analysis. A home-built apparatus was used to place tridimensional objects on the focus of both the ion trap spectrometer and the laser source with a micrometric control, opening the field to a wide range of in situ applications. We think that the obtained results contribute to the overall quest for the most applicable and least destructive experimental approachs for the examination of the organic components of ancient works of art. ACKNOWLEDGEMENTS We would like to thank Dr. Paolo Cremonesi for comments and helpful discussions. We also thank MIUR for financial support (S.I.D.ART. project, contract n. 12828/SSPAR/2001).

REFERENCES Andreotti, A. et al. 2004. Characterisation of natural indigo and shellfish purple by mass spectrometric techniques. Rapid Comm. Mass Spectrom. 18: 1213–1220. Baranov, V. & Bohme, D. K. 1996. In situ collisional dissociation in a selected-ion flow tube: A novel, inexpensive SIFT-CID operation. Int. J. Mass Spectrom. 154 (1–2): 71. Berkout, V. D. et al. 2007. Modeling of ion processes in atmospheric pressure matrix-assisted laser desorption/ ionization. Rapid Communications in Mass Spectrometry 21 (13): 2046–2050. Boon, J. J. & Learner, T. 2002. Analytical mass spectrometry of artists’ acrylic emulsion paints by direct temperature resolved mass spectrometry and laser desorption ionisation mass spectrometry. J. Anal. Appl. Pyrolysis 64 (2): 327–344. Chiavari, G. et al. 2005. Identification of Indigo Dyes in Painting Layers by Pyrolysis Methylation and Silylation. A Case Study: “The Dinner of Emmaus” by G. Preti. Chromatographia 61, April (No. 7/8): 403–408. Ciliberto, E. & Spoto, G. 2000. Modern Analytical Methods in Art and Archaeology. New York: John Wiley & Sons Inc. Clark, R. J. H. 2007. Raman microscopy as a structural and analytical tool in the fields of art and archaeology. J. Mol. Struct. 74: 834–836. D’Agata, R. et al. 2007. The use of atmospheric pressure laser desorption mass spectrometry for the study of iron-gall ink. Applied Physics A: Materials Science & Processing 89: 91–95. Darque-Ceretti, E. & Aucouturier, M. 2004. Secondary ion mass spectrometry. Application to archaeology and art objects. Compr. Anal. Chem. 42: 397–461.

72

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Study of chromophores of Islamic glasses from Al-Andalus (Murcia, Spain) N. Carmona, M. García-Heras & M.A. Villegas Instituto de Historia, Centro de Ciencias Humanas y Sociales, CCHS-CSIC, and Centro Nacional de Investigaciones Metalúrgicas, CENIM-CSIC, Madrid, Spain

P. Jiménez & J. Navarro Escuela de Estudios Árabes, EEA-CSIC, Granada, Spain

ABSTRACT: The recent discovery and excavation of 12th century AD urban glass workshops in the city of Murcia (Spain) have provided good evidences of glass production in the ancient Islamic territory of Al-Andalus. Among other findings, an important amount of bulk coloured and colourless glass fragments were unearthed during the archaeological works undertaken. This research presents the results obtained in the characterization of the chromophores responsible of the different colours found in the glass ensemble, namely turquoise blue, green bluish, emerald green, purple, yellow and red. The main goal of the study was to get some insights into the technology developed to obtain different colours in glasses. The resulting data have allowed the assignment of the ions responsible for each colour studied and have provided outstanding information on the colouring techniques used by the Islamic glassmakers of Al-Andalus.

1

INTRODUCTION

Up to now, little was known on technological aspects of Islamic glasses manufactured in the ancient territory of Al-Andalus (Southern Spain, AD 711–1492). However, the recent discovery and systematic archaeological excavation of urban glass workshops in the city of Murcia has changed this situation since, for the first time, it is possible to have glasses related to production sites with a chronology of mostly the 12th century AD (Jiménez Castillo et al. 2000). These workshops are the first evidence of glass production in the region of Murcia and are currently the only well documented inAl-Andalus, except the workshop of Pechina (Castillo & Martínez 2000). Throughout the 12th century and, particularly, during the kingdom of Ibn Mardanish, Murcia reached a high splendour to such as extent that it was one of the most prominent west Mediterranean cities of that time. For this reason, some Arab written sources reported that Murcia was an important glass production centre (e.g., De Gayangos 1984, Jiménez Castillo 2000). Ibn Mardanish fought against the Almohads, who came from north of Africa and rapidly dominated the most part of Al-Andalus, during more than two decades until the city fell in AD 1172 (Jiménez Castillo 2003). The location of Murcia and the approximate boundary between the Christian and Islamic territories

Figure 1. Map of the Iberian Peninsula showing the location of Murcia and the approximate boundary between the Christian and the Islamic territories around the 12th century AD.

of Al-Andalus around the 12th century AD are shown in Figure 1. One of these Murcian workshops is located in the Puxmarina street. The archaeological excavation of this site revealed a total of five well preserved furnaces and some remains of three others. These furnaces have been dated by archeomagnetism between AD 1100 and

73

of the glass ensemble. This selection encompassed the whole range of colours. In the second place, a sample of turquoise blue, green bluish, emerald green, purple, yellow and red glasses were taken to characterize their corresponding chromophores. In both cases, those fragments without a recognizable typological form were preferentially selected to undertake destructive analyses. 2.2 Analytical techniques Chemical analyses were carried out by X-ray Fluorescence (XRF) using a Philips PW-1404 wavelength dispersed X-ray spectrometer equipped with a tube of rhodium. Analytical determinations were obtained through the standard-less analytical software Uniquant 4.22 based on fundamental parameters. Once external deposits were removed by polishing to avoid contaminations, powder samples were prepared by grinding body glass fragments in an agate mortar. Then, pressed boric acid pellets, using a mixture of n-butylmethacrylate and acetone (10:90 wt %) as bonding medium, were made for the XRF analyses. The characterization of the glass chromophores was undertaken by UV/VIS absorption spectrophotometry using a Shimadzu 3100 spectrophotometer attached with an integrating sphere. Spectra were acquired in the 380–800 nm range on transparent glass samples of approximately 1 mm in thickness. The samples were obtained by polishing both sides of the glasses with a manual rotating polisher using an aqueous suspension of cerium oxide to remove external deposits. To the best of the authors’ knowledge UV/VIS absorption spectrophotometry has been little used in archaeological glasses despite its advantages to investigate their colours and chromophores (Sanderson & Hutchings 1987, García-Heras & Villegas 2004).

Figure 2. Some of the glass fragments from the Puxmarina workshop (Murcia, Spain) in the state as-received in the laboratory. Scales are in cm.

1200 (Gómez-Paccard et al. 2006). Contextual information suggested that at least three of the furnaces could have been used for glass melting. The excavation also provided a very fragmented ensemble of glasses, together with some glassworking waste evidences such as glass dribbles and threads, melts from batches and crucible remains (Jiménez Castillo et al. 2005). The main glass forming technique was blowing, even though some flat glass fragments were also present. The majority of glasses were colourless or slightly yellowish and, less frequently, turquoise blue, green bluish and purple. Only a few fragments of emerald green, yellow and red glasses were documented. All of them were transparent and bulk coloured in those cases in which they had colour. Due to the fragmentary state of the ensemble, a very few number of shapes could be reconstructed, including small vases or unguentaria and small necked bottles. Decoration is only present in a reduced number of fragments and is composed of black, white and red paints. Figure 2 shows some of the glass fragments recovered in the excavation. The relevance of such findings has been explored through a project focused on the archaeometric characterization of the glass productions, using different physical-chemical techniques (Carmona et al., in press). One of the key goals of the project was the characterization of the chromophores or chemical species responsible of the different colours exhibited by the glasses found in the Puxmarina workshop. Such a research is presented in this paper and was aimed at providing some insights into the technology developed by the glassmakers of Murcia, in order to shed new light on the general topic of the Islamic glass technology of Al-Andalus.

2 2.1

3

RESULTS AND DISCUSSION

3.1 Chemical analysis According to the chemical data obtained by XRF, the glasses studied can be classified into two distinct groups: 1) soda-lime-silicate glasses [Na2 O-CaOSiO2 ] and 2) soda-lime lead-silicate glasses [Na2 OCaO-PbO-SiO2 ], which are characterized by a high content of lead oxide. All the colourless and most of the bulk coloured glasses belong to the first group, whereas the second one is only represented by emerald green glasses. Mean and standard deviation of the 13 main components of both groups are displayed in Table 1. The major component of soda-lime-silicate glasses is the network-former SiO2 (58.83 wt %). The glass network-modifier Na2 O shows a relatively high concentration (19.28 wt %), while the content of the

EXPERIMENTAL Samples selected

In the first place, a total of 21 fragments, including glasses and remains of melt batches from furnaces, were selected to determine the chemical composition

74

Table 1. Results derived from the XRF chemical analysis of glasses (weight %). Soda-lime-silicate (n = 419)

Soda-lime lead-silicate (n = 2)

Mean

Mean

Na2 O MgO Al2 O3 SiO2 P2 O5 SO2 Cl− K2 O CaO TiO2 MnO Fe2 O3 PbO

19.28 4.89 3.77 58.83 0.27 0.15 1.06 2.16 7.27 0.19 0.34 0.96 0.83

Total

100.00

S.D. 1.88 0.89 1.47 2.64 0.06 0.04 0.17 0.48 0.98 0.08 0.26 0.22 1.35

12.46 2.61 3.00 49.77 0.12 0.16 0.84 1.52 4.92 0.14 0.10 1.39 22.97

S.D. 3.10 0.69 1.51 3.68 0.02 0.04 0.08 0.50 0.14 0.03 0.04 1.29 7.79 Figure 3. Visible absorption spectrum from a bulk emerald green soda-lime lead-silicate glass.

100.00

S.D. Standard deviation (±).

(2.61 wt %) than in the first group of glasses. The rest of components are otherwise very similar in both groups. Soda-lime lead-silicate glasses can be classified as Islamic high lead oxide glasses, following the terminology of Sayre & Smith (1961), and are documented in the same period of time.

network-stabilizer CaO is 7.27 wt %. The amounts of other network-modifiers such as MgO and K2 O are 4.89 and 2.16 wt %, respectively. The content ofAl2 O3 , which is also a network-former oxide, is 3.77 wt %. The percentages of P2 O5 and SO2 are not higher than 0.30 wt % and chloride ions range around 1 wt %. Minor components determined were transition metals such as TiO2 (0.19 wt %), MnO (0.34 wt %) and Fe2 O3 (0.96 wt %). Iron and titanium oxides can be considered as impurities of the raw materials. However, the manganese oxide was intentionally added as a chromophore to provide the purple colour as is discussed in the next section. This first group of glasses can be classified as high magnesia plant ash glasses (HMG) according to the terminology proposed by Sayre & Smith (1961). The use of plant ashes as a source of sodium oxide is documented throughout the Islamic world between the nine and fifteenth centuries AD and is strongly indicated by the high contents of Na2 O and MgO, as well as the noticeable concentration of K2 O (Tab. 1). These indicators suggest that natron was not used as alkali source because the concentrations of MgO and K2 O had to be then lower or around 1.00 wt %. In the second group, that of soda-lime leadsilicate glasses, the major component is also SiO2 (49.77 wt %). In this group the content of PbO is around 23.00 wt %, which at high concentrations can play the role of a network-former oxide (Götz et al. 1976, Fernández Navarro 2003). The percentage of Na2 O is 12.46 wt %, that of CaO is around 5.00 wt % and the content of K2 O is 1.52 wt %. On the other hand, the concentration of MgO is a little bit lower

3.2 Characterization of chromophores Figure 3 shows the absorption spectrum from a bulk emerald green glass. According to chemical analysis data, it seems that this colour was only produced in soda-lime lead-silicate glasses. The sample presents a unique wide absorption band of high intensity which can be assigned to Cu2+ ions. The band shifts towards lower wavelengths up to 740 nm. This can be attributed to the incorporation of high contents of lead oxide to the glass network since, as mentioned above, it can play the role of a network-former oxide at high concentrations. The high polarisability of the Pb2+ ions induces the glass to asymmetric structures able to be deformed (Fernández Navarro 2003: 453). This fact shifts the absorption band of Cu2+ ions from ∼800 to ∼740 nm, thereby changing the colour from blue to emerald green. The presence of copper oxide in this glass (4.22 wt %) was confirmed by chemical analysis data obtained by XRF. The UV absorption edge of this sample is around 425 nm and is probably due to the presence of Fe3+ ions which absorb at 380, 420 and 440 nm. Therefore, the intense emerald green colour was achieved by the synergic effect of the Cu2+ ions, which progressively shift their absorption band from blue to green due to the presence of a high lead oxide content in the glass, and the yellow colour

75

Figure 5. Visible absorption spectrum from a bulk green bluish soda-lime-silicate glass.

Figure 4. Visible absorption spectrum from a bulk turquoise blue soda-lime-silicate glass.

provided by the Fe3+ ions. On the other hand, in modern glasses the emerald green colouring is obtained through Cr3+ ions, which have a triple absorption band at 630, 650 and 675 nm (Bamford 1977). These bands do not appear in the spectrum of the emerald green glass. Figure 4 displays the absorption spectra obtained from a bulk turquoise blue soda-lime-silicate glass. It presents a unique wide band of moderate intensity, in comparison with the emerald green glass, between 780 and 810 nm produced by Cu2+ ions, which provide the characteristic bright turquoise blue colour.The presence of copper oxide in this glass was confirmed by XRF (1.70 wt %). In this case the band does not experience a shift towards lower wavelength, since the content of lead oxide is very low (1.19 wt % according to XRF data). It is important to note the absence of the triple absorption band produced by the Co2+ ions at 540, 590 and 640 nm (Bamford 1977), which gives rise also to a more intense blue colour. Cobalt oxide, therefore, was not used as a chromophore in the Puxmarina turquoise blue glasses. Figure 5 shows the absorption spectrum from a bulk green bluish soda-lime-silicate glass. It also presents a unique wide band between 780 and 810 nm which can be likewise assigned to Cu2+ ions. This band exhibits, however, a lower intensity in comparison with the turquoise blue glass. In addition, the absorption band shows a slightly shift up to approximately 760 nm, which produces that the turquoise blue colour turns into a light green hue. Such as shift can be also attributed to the noticeable amount of lead oxide in the glass (4.47 wt %). The concentration, however, is considerably lower than that determined in the emerald green glass and, therefore, the green colour is much

Figure 6. Visible absorption spectrum from a bulk purple soda-lime-silicate glass.

weaker. The presence of copper oxide in this sample (1.65 wt %) was also confirmed in this sample by means of XRF. As in emerald green or turquoise blue glasses, the absorption bands of Cr3+ and Co2+ ions do not appear either. The visible absorption spectrum from a bulk purple soda-lime-silicate glass is shown in Figure 6. The purple colouring presents a unique absorption band at around 499 nm. This band is produced by the Mn3+ ions, which provides an intense purple colouring to the glass. The purple colour is intensified in silica glasses as far as the glass alkalinity increases (Fernández Navarro 2003: 448) and, in fact, this sample was made from a type of glass with a high degree of alkalinity as is the case of the soda-lime-silicate glasses (Tab. 1). Chemical determinations by XRF confirmed

76

Figure 7. Visible absorption spectrum from a bulk silver yellow glass probably of the soda-lime-silicate type.

Figure 8. Visible absorption spectrum from a bulk ruby red soda-lime-silicate glass.

or yellow colouring. Therefore, from the technological point of view, obtaining such colours is difficult. It is important to point out that the solubility of both copper and silver nanoparticles rises as far as the alkalinity and the lead oxide concentration of the glass increase (Fernández Navarro 2003: 472). This fact could explain why a certain amount of lead oxide (2.84 wt % detected by XRF) was present in the bulk ruby red glass.

the optical absorption results, since the concentration of manganese oxide in this glass was 1.14 wt %. The bulk yellow colour is due to the formation of Ag silver colloidal nanoparticles, which are responsible of the absorption recorded at around 400 nm in the spectrum of Figure 7. The introduction of silver compounds in the glass gives rise to the formation of colloidal nanoparticles through the following three steps process: 1) dissolution of Ag+ ions and incorporation into the glass network, 2) thermal reduction of Ag+ ions to Ag atoms by means of a reducing atmosphere, and 3) precipitation and aggregation of Ag atoms which tend to form colloidal nanoparticles responsible of the yellow colouring (Fernández Navarro & La Iglesia 1994). Due to the reduced number of bulk yellow glass fragments found and the low weight available, it was not possible to analyze them chemically by XRF. Consequently, the content of silver could not be analytically confirmed. In any case, a very low concentration of silver (from 0.05 to 0.50 wt %) is enough to provide the characteristic silver yellow colour to the glass (Fernández Navarro 2003: 472). Finally, Figure 8 displays the absorption spectrum from a bulk red soda-lime-silicate glass. It shows an intense and well-defined absorption band at around 560 nm, which is characteristic of copper ruby red glasses. The red colour is due to the formation of Cu+ /Cu colloidal nanoparticles in a similar process to the silver ones but using copper compounds. The minimum concentration of copper oxide to produce the ruby red colour is estimated in ∼0.50 wt % (Fernández Navarro 2003: 468). The content of copper oxide determined in this sample by XRF was 0.59 wt %. Either in the case of copper or in the case of silver colloids it is necessary to produce critical reducing conditions during the glass melting to develop ruby red

4

CONCLUSIONS

The results of the present research have allowed the assignment by UV/VIS absorption spectrophotometry of the ions responsible for each glass colouring studied. The characterization of these chromophores indicated that Murcian glassmakers of the 12th century AD used copper oxide compounds to obtain emerald green, turquoise blue, green bluish and ruby red colouring in glasses. That is, by using the same copperbased chromophore, they were able to produce three different bulk colours varying the sensitive redox conditions of copper oxide during the melting process of glass. They also employed manganese oxide to obtain purple and some silver compounds to obtain the yellow colour. The presence of these ions was confirmed through chemical analysis by XRF, except in the case of the silver yellow in which there was not enough amount of sample available to be chemically analyzed. The resulting data suggest, therefore, a deep knowledge of glass colouring techniques. This implies a high degree of specialization in glass production in which control over different colours and glass compositions was achieved. In this sense it is important to emphasize the use of a soda-lime lead-silicate glass to specifically produce the emerald green colour, since both

77

en al-Andalus: 83–101. Madrid: Coeditions de la Casa de Velázquez. De Gayangos, P. 1984. The History of the Mohammadan Dinasties in Spain. Delhi: 2 vols. Fernández Navarro, J.M. 2003. El vidrio. Constitución, fabricación, propiedades. Madrid: CSIC (3rd edition). Fernández Navarro, J.M. & La Iglesia, A. 1994. Study of the red and yellow colour of glasses from the Cathedral of Toledo. Boletín de la Sociedad Española de Cerámica y Vidrio 33: 333–336. García-Heras, M. & Villegas, M.A. 2004. Notas para el estudio científico del vidrio antiguo. Zephyrus 57: 377–390. Gómez-Paccard, M., Chauvin, A., Lanos, Ph., Thiriot, J. & Jiménez Castillo, P. 2006. Archeomagnetic study of seven contemporaneous kilns from Murcia (Spain). Physics of the Earth and Planetary Interiors 157: 16–32. Götz, J., Hoebbel, D. & Wieker, W. 1976. On the constitution of silicate groupings in binary lead-silicate glasses. Journal of Non-Crystalline Solids 22: 391–398. Jiménez Castillo, P. 2000. El vidrio andalusí en Murcia. In P. Cressier (ed.), El vidrio en al-Andalus: 117–148. Madrid: Coeditions de la Casa de Velázquez. Jiménez Castillo, P. 2003. Murcia islámica. Una visión a través de la arqueología. Murcia: Ayuntamiento de Murcia. Jiménez Castillo, P., Muñoz, F. & Thiriot, J. 2000. Les ateliers urbains de verriers de Murcia au XIIè siècle (c. Puxmarina et pl. Belluga). In P. Pétrequin, P. Fluzin, J. Thiriot & P. Benoit (eds.), Arts du feu et productions artisanales, XX Rencontres Internationales d’Archéologie et d’Histoire d’Antibes: 433–452. Antibes: Editions APDCA. Jiménez Castillo, P., Navarro, J. & Thiriot, J. 2005. Taller de vidrio y casas andalusíes en Murcia. La excavación arqueológica del casón de Puxmarina. Memorias de Arqueología 13: 419–458. Sanderson, D.C.W. & Hutchings, J.B. 1987. The origin and measurement of colour in archaeological glasses. Glass Technology 28: 99–105. Sayre, E.V. & Smith, R.W. 1961. Compositional categories of ancient glass. Science 133: 1824–1826.

high contents of lead oxide and glass alkalinity favour the incorporation and solubility of copper oxide to the glass. However, it is unlikely that Murcian glassmakers were aware of this point beyond the empirical level. Overall, this research sheds new light on the Islamic glass technology developed in the ancient territory of Al-Andalus in which, up to now, little scientific evidence was available. ACKNOWLEDGEMENTS This work has been financed by projects 200510M068 co-founded by the General Office of Universities and Research from the Regional Government of Madrid and the Spanish National Research Council (CSIC), and CSIC-PIE 200610I031. The authors acknowledge the General Office of Culture from the Regional Government of Murcia and the Archaeological Museum of this city for providing glass samples and for their useful collaboration. Dr. N. Carmona acknowledges an I3P (CSIC-ESF) postdoctoral contract. Finally, the authors are indebted to the CSIC Thematic Network on Cultural Heritage for its professional support. REFERENCES Bamford, C.R. 1977. Colour generation and control in glass. Amsterdam: Elsevier Science Publishers. Carmona, N., Villegas, M.A., Jiménez Castillo, P., Navarro, J. & García-Heras, M. in press. Caracterización arqueométrica de vidrios andalusíes procedentes de talleres murcianos. Actas de las Jornadas sobre Vidrio de la Alta Edad Media y Andalusí 2006. La Granja: Fundación Centro Nacional del Vidrio. Castillo, F. & Martínez, R. 2000. Un taller de vidrio en Bayyana-Pechina (Almería). In P. Cressier (ed.), El vidrio

78

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Polychromed sculptures of Mercadante and Millán analysed by XRF non-destructive technique A. Križnar & M.A. Respaldiza Centro Nacional de Aceleradores, Universidad de Sevilla, Seville, Spain

M.V. Muñoz, F. de la Paz & M. Vega Museo de Bellas Artes de Sevilla, Seville, Spain

ABSTRACT: Lorenzo Mercadante de Bretaña and his pupil Pedro Millán belong to the most important medieval artists that worked in Andalusia in the second half of the 15th century. The Museum of Fine Arts of Seville has an exceptional collection of their big polychromed terracotta sculptures, among which are Mercadante’s “Virgin and Child” and three Milláns’s works, “Entombment of Christ”, “Christ Man of Sorrows” and “Christ bound to the column”. Their support and pigments were analysed by the X-ray Fluorescence (XRF) non-destructive technique. A high presence of Pb is related to Pb based compounds applied as a preparation/ imprimation, as a pigment or as a dryer. Red pigments are cinnabar, red earth and probably carmin. Blue pigments found are azurite, some organic blue and vivianite. Green colour is a Cu based pigment; in one case green earth was found. Brown colour is umbra. In several areas gold and silver were detected.

1

INTRODUCTION

the polychromy. They all form part of an exceptional collection of big terracotta sculptures from the late Gothic and Renaissance times and are permanently exhibited in the first room of the museum. Among them are Mercadante’s “Virgin and Child” (second half of the 15th Century) and three Millán’s works: “Entombment of Christ” (around 1490), “Christ Man of Sorrows” (1485–1503) and “Christ bound to the column” (second half of the 15th century). The last one was found broken in many pieces and had to be fully reconstructed. It lost most of the colour decoration as well.

Lorenzo Mercadante de Bretaña and his pupil Pedro Millán belong to the most important medieval artists that worked in Andalusia in the second half of the 15th and beginning of the 16th century. Their elaborated style and exceptional quality had a remarkable influence on the artistic production in southern Spain of that time. Lorenzo Mercadante could have been born in Italy (Moreno Mendoza et al. 1991), but he learned the arts in France, in Brittany, as reveals his name (de Bretaña). From there he came to Seville, where he is documented between 1454 and 1467, working on several orders for the town cathedral. Among the most important are the porticos of Nativity and Baptism. His style reveals a mixture of French court forms and Nordic spiritualism, as well as a realistic and meticulous modelling, inspired by the Flemish painting. Pedro Millán was his pupil and in the sixties his collaborator on the cathedral’s porticos. Millán carried on the master’s style, which he enriched with the traditional andalusian expression. He is documented between 1485 and 1507 and probably died before 1526. Both artists worked mostly in terracotta, but used also other materials. The terracotta sculptures had been polychromed, although a great part of the colour decoration is lost today, especially in the sculptures located outside. The Museum of Fine Arts of Seville holds four artworks from both masters that still preserve some of

2

OBJECTIVES

Until today, only few examples of polychromed terracotta sculpture have been preserved. The bibliography on this topic is also quite scarce. In this area, Seville has a special place, regarding a rich patrimony of big terracotta sculptures from the Late Gothic and Renasaince. Several studies on the cathedral sculptures of Mercadante and Millán have been carried out. Most of them are dealing with the ceramic composition and the environmental pollution (Arquillo Torres & Arquillo Torres 1992, Pérez-Rodríguez et al. 1995), while the technique of polychromy was not sufficiently studied (Jiménez del Haro et al. 2001). In all these mentioned four sculptures, the colour palette and the technique of

79

the polychromy are of interest, as well as the artistic relationship between the master and the pupil. Until now this relationship was based only on the style comparison, but not on the materials and techniques applied by both artists. This paper tries to cover this deficiency.

3

analysing some single elements reference materials. All the measurements were done with fixed instrumental conditions (29.5 kV of applied high voltage, 80 µA of cathode current and 300 s of preset lifetime). A semi-quantitative analysis was done by using the areas of XRF peaks obtained in the multi-channel analyser. These areas can give a semi-quantitative estimation of the element concentrations, because they are proportional to the weight concentrations and their square root can serve as a measure of experimental error. Therefore, comparisons between the content of one particular element in different samples of similar composition can be made directly through the respective peaks of that element. However, a comparison of the concentrations between different elements can be only possible if a complete calculation (that is, making the corrections for X-ray production cross-sections, self-absortion, etc.) of the concentrations are made. The pigments applied in the polychromed sculptures were recognized on the basis of characteristic chemical elements from the XRF spectra of analysed points. The elements are identified by the energies of their characteristic X-ray peaks. The comparison of the counts per second of the different elements in a particular point with regard to the background, gives the possibility to ascertain the presence or not of a particular element in that analysed point. The spectra were also compared with a pigment database that was elaborated at CNA, analysing commercial pure pigments from old traditional recipes. Because of the complex 3D modelling of the sculptures it was difficult or even impossible to reach with the XRF head to many points of interest, especially those inside the drapery folds, where pigments are usually better preserved. Besides, in many areas some later polychromies applied over the original ones were appreciated, as well as a thick hard layer of wax and resin over the surface of the sculptures. All these facts complicated the interpretation of the XRF spectra and made more difficult to distinguish original colour layers from the posterior restoration interventions. On the other hand, it is not always possible to identify the precise pigment applied, because the XRF analysis offers the elemental and not the compositional results. There exist several pigments with the same characteristic chemical elements. In this case, this technique alone can not distinguish between them. Such is the case, for example, of green pigments based on Cu. There are several types of Cu pigments; some of them are historic (have been used in the past), which narrows the choices in the case of studying old artworks. But it is still not possible to say with exactitude whether it is malachite, verdigris or maybe some copper resinate. Also the XRF technique does not serve to identify organic materials, because it does not detect elements with atomic number Z lower than 13 or 14. That is why

EXPERIMENTAL PROCEDURE

The sculptures are not in a restoration process, so it was not considered convenient to extract any microsamples. For this reason the non-destructive technique of X-Ray Fluorescence (XRF) was chosen for the analysis of support and pigments. The XRF equipment used has an X-ray tube of 30 kV with anode of W and one SDD detector with energy resolution of 140 eV. A 1 mm Al filter was coupled to the tube to suppress the characteristic peaks of the anode. This instrument was used directly in situ, in the exhibition room (Fig. 1), during the days when the museum is closed to the public. All four artworks were analysed on the front side and on both lateral sides, radiating many different points of interest. Furthermore, “Entombment of Christ” and “Christ bound to the column” were analysed also on the back side. Only these two sculptures could have been removed from the wall without any risk of damages. It should be pointed out that in all four sculptures, their back side was not meant to be presented to the public eye, so they did not use to be elaborated or polychromated. Before each new measurement session, the XRF equipment was calibrated in energy, radiating the air and showing characteristic peaks of Ar, the Zr peaks from an internal collimator of the detector and

Figure 1. In situ analysis of the sculpture by portable XRF.

80

dryer. According to Gómez (Gómez 2000), the use of minium was common in the technique of polychromy. Pacheco (Bassegoda i Hugas 1990) also speaks in his treaty about the use of a red lead substance in the imprimation, which he called azarcón. In the case of Mercadante’s and Millán’s sculptures there must be a combination of two or three of these uses, demonstrated by the variable quantities of Pb compounds in the analysed areas. Nevertheless, although the count numbers for Pb peaks in the XRF spectra vary widely (from 30 counts per second or cps to 720 cps), they always remain high if compared with other elements in each spectra. The high concentration of Pb in all polychromed areas reveals the possible existence of a preparation or an imprimation layer covering the whole surface. On the other hand, in the parts without polychromy, as on the back side of the sculpture “Entombment of Christ”, the count numbers of Pb are very low, almost insignificant. This difference confirms that lead exists above all in the polychromed areas.

organic pigments can not be detected by XRF. This fact complicates also the analysis of clay or terracotta with many chemical elements with a low Z (Na, Mg, Al, Si). Information on surface alteration processes can not be obtained by XRF. Nevertheless, the XRF technique is very important in the study of materials, especially in art, as it offers the first exam of the artwork, without touching it or damaging it in any way. With portable equipments, the tests can be run in situ, without the need to move or transport the piece of art. It is one of the best ways to obtain information about the materials applied in the artwork and to have the first overlook in the materials and technique of the artist. It also serves to discover possible later interventions revealing modern materials where there should only be traditional ones. Nevertheless, it is a good practice, when possible, to combine XRF with other complementary techniques to obtain more specific results. In the case of the four sculptures presented in this paper, there was no possibility to do so, as explained above. 4

4.3 Pigments

RESULTS AND DISCUSSION

A lot of literature has been published about historical pigments used in painting and in polychromy (Calvo 1997, Gómez 2000, Knoepfli et al. 1990, Montagna 1993, Serchi 1999, Schram & Herling 1995, West Fitzhugh et al. 1987–1997). The colour pallette found on the examined sculptures is very similar: white, carnations, red, blue, green, brown. Some decorative parts are gilded. Also the presence of silver was found in some areas, that could belong to some drapery decoration. Next, more detailed results for each sculpture are exposed.

4.1 Terracotta The analysis of the terracotta bulk of the four selected sculptures was limited to a number of small areas, where the colour layers are already lost. In all four of them, the spectra show two predominant peaks for Ca, Fe, and other peaks of lower intensities as Mg, Si and Mn. High peaks of Sr appear in all the spectra where also Ca peaks are of high intensity. Sr is associated to Ca and it is common to find them both together in the preparation layers of painted surfaces as well as in ceramic materials (Seccaroni & Moioli 2002). Zr is present in all XRF spectra due to an internal collimator of the SDD detector made with this material. 4.2

4.3.1 Lorenzo Mercadante: Virgin and Child The sculpture (Fig. 2) follows the iconographical type of gentle Madonna that was developed by Mercadante himself. Virgin Mary is standing; she wears a long red dress with a blue coat over it. Her hair is covered by a white wimple. She carries the Child on her left hand, her right hand is lost. The Child wears a blue tunic and his hands are also missing. On the other hand, another blue pigment seems to be used for the Virgin’s coat. The lack of characteristic chemical elements in the spectra indicates the use of some organic blue pigment or inorganic ultramar blue, which is difficult to detect by XRF because of the low Z of characteristic chemical elements of this pigment (Fig. 4). The inner part of the coat was decorated with a Cu based green and in parts with red earth (Ca, Fe). Gold (Au) as well as some traces of silver (Ag) were confirmed in the Virgin’s vestments, belonging to the decoration elements, mostly lost today. Silver

Lead compounds

A considerable quantity of lead was detected in all analysed sculptures. In all spectra obtained from the polychromed areas, the L peaks of Pb are the highest and the predominant ones as compared with those of bulk terracotta. Such an intense presence of lead shows an important role of some lead pigment(s) in the sculptures. Lead can be due to various lead compounds, not possible to distinguish by XRF: lead white, basic lead carbonate (2 Pb(CO3 )2 · Pb(OH)2 ), yellow litharge, lead oxide (PbO) or orange-red minium, also lead oxide (Pb3 O4 ). (Seccaroni & Moioli 2002, West Fitzhugh et al. 1987–1997, Dornheim & San Andrés Moya 2004). In painting and polychromed sculpture, they can be used as a preparation/imprimation, as a pigment (pure or added to another one), or as a

81

105

MERCADANTE: Virgin and child blue coat Virgin Pb Pb

Counts

104 Pb

Pb 103

Pb

Pb

Pb Fe Fe

Ca

Zr Zn

Pb

Zr

102

5

10

15

20

25

Energy (keV)

Figure 4. XRF spectrum of an analysed point from the Virgin’s blue coat (“Virgin and Child”). An organic blue pigment or inorganic ultramar blue was probably used.

Figure 2. Polychromed terracotta sculpture “Virgin and Child” by Lorenzo Mercadante (second half of 15th century). Height 1.34 m. 5 10

MERCADANTE: Virgin and child blue dress child Pb Pb

Counts

4 10

3 10

Pb

Pb

Fe

Pb

Pb

Zn Zn Pb Ca

Zr

Fe Pb

Cu

Zr

2 10

5

10

15

20

Figure 5. Polychromed terracotta sculpture “Entombment of Christ” by Pedro Millán (around 1490), a) front side and b) back side. Height 0,64 m.

25

Energy (keV)

4.3.2 Pedro Millán: Entombment of Christ A horizontal composition of seven figures standing around the dead Christ lying on the white sarcophagus is one of the most beautiful Millán’s works (Fig. 5). The artist’s signature appears in the centre of the tomb. The figures are dressed in blue, red and green vestments. The women and the two men figures behind the tomb have their head covered with wimples, while the both

Figure 3. XRF spectrum of one analysed point from the Child’s blue dress (“Virgin and Child”). The pigment used could be vivianite.

is detected on dark areas on the red dress and in some parts of the blue coat. High peaks of Zn in the Mary’s nose reveal some earlier intervention on the face, applying zinc white.

82

105

105

MILLÁN: Entombmentof Christ blue dress Pb Pb

Pb 104

104 Pb

Pb 103

Fe Pb Ca Ca

Pb

Cu Zn

Counts

Counts

MILLÁN: Entombment ofChrist blue back side Pb

Pb Zr

102

10

15

Pb Zr

Hg Pb

Pb

Fe

Zn Cu

Ca Zr

5

10

Pb

Fe

Pb

Pb Fe

3

Pb

Zr

2

10

20

5

25

10

15

20

25

Energy (keV)

Energy (keV)

Figure 7. XRF spectrum of an analysed point from the blue colour on the back side of the sculpture (“Entombment of Christ”). An organic blue pigment was used.

Figure 6. XRF spectrum of an analysed point from the blue tunic, left male figure (“Entombment of Christ”). Azurite was found to be the applied pigment.

105

men on the side with a cap. Christ lies on a sheet; his head is turned towards the spectator. This sculpture (0.64 m high) was examined in 45 points. These were chosen on both lateral figures, on the sarcophagus and on the Christ’s body. The figures behind the tomb could not be reached by XRF equipment from the front side, but they were analysed in some points on the back side, polychromed on the upper area of heads and shoulders. As said above, this sculpture was one of two that offered the possibility of analysis also on the back side. Lead white (Pb) was used for white colours. For carnations it was mixed with cinnabar (Hg). The Christ’s skin is paler than that of other figures in the presentation, that is why it contains less red pigment; the count numbers of Hg are lower and those of Pb much higher than in other analysed points of carnations. The red colour is a mixture of cinnabar (Hg) and red earth (Fe). Some interesting differences between the pigments on the front and on the back side were discovered. The blue pigment in the front is azurite, confirmed by the high peaks of Cu (Fig. 6), while in the back some cheaper, probably organic pigment was applied. There are no characteristic chemical elements for an inorganic blue pigment revealed by the spectra (Fig. 7). It is possible that the blue colour found on the back side belongs to some restoration work and not to the original palette. For the green colour, two different pigments were also used, identified by different characteristic chemical elements. In the front side a copper based green pigment was applied, revealed by high count numbers of Cu peaks (Fig. 8). It could belong to malachite, verdigris or some copper resinate. In the back side, the peaks of Si, Mn and Fe show the use of a green earth. The brown pigment on the hair is umbra (Mn, Fe). The edges of the sarcophagus are gilded (Au). On

MILLÁN: Entombment of Christ green dress Pb Pb

Cu 104 Counts

Cu Fe 103

Pb Pb

Ca Pb

Mn Ca

Fe

Pb Zr

Pb Hg

Pb Zr

2

10

5

10

15

20

25

Energy (keV)

Figure 8. XRF spectrum of an analysed point from the green tunic of the right male figure (“Entombment of Christ”). A copper based green pigment was applied.

the tools and in some azurite areas silver (Ag) was detected. Probably the tools were covered by a silver foil that oxidised. On the other hand, Ag peaks in the spectra of blue azurite areas could belong to the blue pigment itself (Seccaroni & Moioli 2002). 4.3.3 Pedro Millán: Christ Man of Sorrows This artistic work (Fig. 9) is organized in a sense of symbolic hierarchy with the dominant central figure of Christ, smaller figures of two angels at both sides and the smallest and iconographically less important kneeling donor. The sculpture presents a complex example of polychromy, although lost in parts. Christ wears a white shroud, a red coat over his back and shoulders and the spine crown on his head. He is defined by his five sores. Two angels are supporting the heavy coat of Christ with their inner hand, while with the outer one they carry

83

105

MILLÁN: Christ Manof Sorrows carnation angel face angel

Pb

Pb

Counts

104 Fe

10

Pb

Pb Hg Hg Pb

3

Zr Pb

Fe Cu

Ca

Pb

Zr

Pb Ca

Ti

102

5

10

15

20

25

Energy (keV)

Figure 10. Comparison of two XRF spectra of the left angel’s face: area with polychromy (straight line) and area without it (dot line) (“Christ Man of Sorrow”). 105

MILLÁN: Christ Man of Sorrows grey coat Christ Pb

Pb

Counts

104 Hg

10

Fe

3

Ca

2

10

golden attributes, related to the Passion of Christ. They both wear white tunics and a blue coat over it, golden strip is decorating their heads. The donor, kneeling at the left side of Christ, wears a blue dress and a red coat. The ground and the abundant vegetation seem of green-greyish colour. The sculpture (1.65 m high) was analysed in 84 points. They were chosen in diverse areas of carnations and vestments to compare results of different colours. A detailed study of this sculpture has been published elsewhere (Kriznar et al. 2008). The colour layers are very complex, with at least one re-touching belonging to some earlier restoration works. It was possible to establish the basic colour palette, but without stratigraphic sections it could not be determined whether they belong to the original or restored parts. The white pigment is lead white (Pb). The carnations are made with lead white (Pb) and cinnabar (Hg) (Fig. 10). Maybe also an organic red pigment as carmine was added, but it is not possible to identify it by XRF. In the Christ’s carnation also a low presence of copper based green pigment was detected. In the spectra of this area appear Cu peaks that are not present in the spectra of other carnations in this sculpture. The red colour is a mixture of cinnabar (Hg) and small amount of red earth (Fe). In some areas of the

Cu

Pb

Pb Pb Hg

Fe Hg

Pb

Figure 9. Polychromed terracotta sculpture “Christ Man of Sorrow” by Pedro Millán (1485–1503). Height 1,65 m.

Hg

Pb

Pb Zr

5

Ag Zr

Ca

10

15

20

25

Energy (keV)

Figure 11. XRF spectra of an analysed point of a grayish layer from the Christ’s coat (“Christ Man of Sorrow”). Detected silver probably belongs to already lost silver foil.

Christ’s coat, a dark red colour can be seen at a closer look in situ. The XRF spectra of selected analysed points do not show any characteristic chemical elements for any inorganic red pigment. In this case, the dark red layer is probably an organic colour. The blue pigment is azurite (Cu), maybe mixed with vivianite (Fe) in the angel’s coats. Fe could belong also to some red imprimation under the blue layer. On both coats blue over paint is clearly seen whose rests cover the gilded borders. The green colour of the plants is some Cu based pigment, malachite, some copper resinate or verdigris. With the XRF technique it is not possible to distinguish between them. The brownish hair colour is made with umbra (Mn, Fe). Gold (Au) was confirmed on gilded parts of the vestments, on Christ’s sores and on angels’s hair. Well defined peaks of silver (Ag) were detected in some areas of the Christ’s brooch, his coat and the

84

105

MILLÁN: Christbound to the coloumn carnation leg

104 Counts

Fe Pb 103

Ca Ti K Ca Ti

102

Pb Zr

Fe

Pb Pb Zn

Mn

Zr

Pb

Cu 101 5

10

15

20

25

Energy (keV)

Figure 13. XRF spectrum of an analysed point in the restored area from the Christ’s left leg (“Christ bound to the column”).

polychromy is conserved, except on Christ’s hands and face. For this reason only 20 points were chosen in order to get information about the basic material used that is terracotta. Some rests of the polychromy were analysed, as well. Also this sculpture is the highest of all four (1.88 m), could be removed from the wall and analysed on the back side. On the bases of the column, a small probe was found. It was previously made by the restorers of the museum, and allowed the examination of the original material of the sculpture. This was also possible on the damaged right elbow. The cleaning test in layers revealed that the whole surface of the artwork was covered with a unifying layer to hide contacts between the broken pieces. The presence of this new layer was confirmed also by the XRF results, which show an important difference between the areas that still conserve original polychromy and the ones with the new unifying layer. The basic difference is the high presence of Ti, Zn and Fe in the areas with the mentioned layer (Fig. 13), while in the ones with polychromy the predominant element is Pb. This element reveals the presence of lead white, together with a small amount of cinnabar (Hg) for the carnation (Fig. 14), conserved in both hands, on the face and on the left knee. The Ti and Zn elements belong to Ti-Zn white, mixed with ochre (Fe). With the help of the XRF analysis it was possible to determine which parts are covered with this layer and which are not.

Figure 12. Polychromed terracotta sculpture “Christ bound to the column” by Pedro Millán (2/2 15th century). Height 1.88 m.

coats of both angels. In the case of the Christ’s coat (Fig. 11) it probably belongs to an already lost silver foil which was later decorated with the organic red colour to render it shinier. The blue azurite and red cinnabar, also discovered in this coat, belong to later re-polychromies. Silver found on the angel’s coats could be a part of the blue pigment itself (Seccaroni, Moioli 2002) or of some drapery decoration that is not preserved anymore.

5

4.3.4 Pedro Millán: Christ bound to the column This sculpture (Fig. 12) was considered lost (Gestoso 1984), but it was found during excavations in the church of Santa Ana in Seville in 1971 (Moreno Mendoza et al. 1991). It was broken into 168 pieces and it had to be totally reconstructed. Almost no

CONCLUSIONS

The polychromed terracotta sculptures of Lorenzo Mercadante and Pedro Millán, exposed in the Museum of Fine Arts of Seville were analysed by the nondestructive technique of XRF. The support and the

85

pigments applied are of interest, as well as the technical connection between the master and the pupil. The results, summarised in Table 1, revealed that both of them used the same bulk terracotta base, not much elaborated at the back sides of the sculptures. The polychromy was applied only on the front side and not at the back. On all the sculptures the results showed the use of similar inorganic pigments: lead white (Pb), red cinnabar (Hg), red earth (Ca, Fe), copper based green pigments (Cu) and umbra (Mn, Fe). In the choice of blue pigments the two artists differed. Mercadante preferred to use an organic blue colour or an inorganic

105

Counts

10

ultramar, difficult to determine with XRF. Also vivianite was detected in his work. On the other hand, Millán preferred to apply azurite. In all four sculptures an organic red pigment, such as carmine (which can not be identified by XRF), could be present. In some cases, it is difficult to know for sure if the colour layers are the original ones or belong to some later retouching. In all analysed artworks, gold was confirmed for the decorative elements. The presence of silver, that is not seen with the naked eye but is present in the spectra, is difficult to understand. It can belong to some already lost silver foil or could form part of a blue pigment. REFERENCES

MILLÁN: Christ bound to the coloumn hand

4

Pb

Arquillo Torres F. & Arquillo Torres J. 1992. ¡¡Salvemos los “Mercadantes” de las portadas del Bautismo y del Nacimiento en la catedral de Sevilla!! IX Congreso de conservación y restauración de bienes culturales: 403–421. Bassegoda i Hugas B. (ed.) 1990. Francisco Pacheco: El Arte de la Pintura. Madrid: Cátedra. Calvo A. 1997. Conservación y Restauración, Materiales, técnicas y procedimientos, de la A a la Z. Edición del Serbal: Barcelona. Dornheim S.D. & San Andrés Moya M. 2004. Litargirio y masicote. Terminología, propiedades y usos. Reproducción a escala de laboratorio de algunos de sus procesos de obtención. XV. Congreso de conservación y restauración de bienes culturales 2004, Actas, Vol. I: 533–546. Gestoso J. 1984. Sevilla Monumental y Artística. Sevilla: Monte de Piedad y Caja de Ahorros Sevilla. Gómez M. L. 2000. Examen científico aplicado a la conservación de obras de arte. Madrid: Cátedra, Instituto del Patrimonio Histórico Español.

Pb

Fe 103 Fe

Ca

Pb Pb

Pb 2

10

K Ca Ar Ti

Zr

Zn Hg Mn

5

Pb Zr Pb Pb

Cu

10

15

20

25

Energy (keV)

Figure 14. XRF spectrum of an analysed point in an original polychromated area from Christ’s left hand (“Christ bound to the column”) showing the use of lead white (Pb) and cinnabar (Hg) pigments.

Table 1. Comparison of the used pigments in different areas of the four analysed sculptures.

Virgin and Child Terracotta

Terracotta (Ca, Fe/Mg, Si, Mn) White pigment Lead white (Pb) Carnations Lead white (Pb)+ cinnabar (Hg)

Entombment of Christ

Christ Man of Sorrows

Terracotta (Ca, Fe/Mg, Si, Mn) Lead white (Pb) Lead white (Pb) + cinnabar (Hg)

Terracotta (Ca, Fe/Mg, Si, Mn) Lead white (Pb) Lead white (Pb) + cinnabar (Hg) (+Cu based green pigment − Christ) Cinnabar (Hg) + red earth (Fe) Carmine? Azurite (Cu) vivianite? (Fe) − angel’s coats Cu based pigment (Cu)

Red pigment

Cinnabar (Hg) + red Cinnabar (Hg) + red earth (Fe) Carmine? earth (Fe) Carmine? Blue pigment Vivianite? (Fe, Zn) − Azurite (Cu) − front side Virgin Organic/ ultramar Organic blue − back side blue − Child Green pigment Cu based pigment (Cu) Cu based pigment − front side Green earth (Si, Mn, Fe) − back Brown pigment Umbra (Mn, Fe) Umbra (Mn, Fe) Umbra (Mn, Fe)

86

Christ bound to the column Terracotta (Ca, Fe/Mg, Si, Mn) Lead white (Pb) Lead white (Pb) + cinnabar (Hg) Not preserved Not preserved

Not preserved

Not preserved

Jiménez del Haro M.C., Pérez Rodríguez J.L. & Justo A. 2001. La técnica del brocado para decorar las cerámicas de las puertas de la catedral de Sevilla. III. Congreso Nacional de Arqueometría: 325–333. Knoepfli A., Emmenegger O., Koller M. & Meyer A. (eds) 1990. Reclams Handbuch der künstlerischen Techniken. Vol 1–3. Stuttgart: Philipp Reclam jun. Kriznar A., Muñoz M.V., De la Paz F., Respaldiza M.A. & Vega M. 2008. Pigment identification using X-ray fluorescence in a polychromated sculpture by Pedro Millán. X-ray Spectrometry. Vol. 37. Chichester: Wiley Publisher. (in press) Montanga G. 1993. I pigmenti. Prontuario per l’arte e il restauro. Firenze: Nardini editore. Moreno Mendoza A., Pareja López E., Sanz Serrano M. & Valdivieso Gonzáles E. 1991. Museo de Bellas Artes de Sevilla. Sevilla: Ediciones Gever, S.L.

Pérez-Rodríguez J.L., Jiménez de Haro M.C., Justo A., Maqueda C. & Ruiz Conde A. 1995. Study of the ceramic sculptures of the birth and baptism porticos of Seville Cathedral. The Ceramics Cultural Heritage: 635–642. Seccaroni C. & Moioli P. 2002. Fluorescenza X. Prontuario per l’analisi XRF portatile applicata a superfici policrome. Firenze: Nardini editore. Serchi M. (ed) 1999. Cennino Cennini: Il Libro dell’Arte. Firenze: Felice Le Monnier. Schram H.P. & Herling B. 1995 Historische Malmaterialen und ihre Identifizierung. Stuttgart: Ravensburg Buchverlag. West Fitzhugh E., Feller R. & Roy A. (eds.) 1987-1997. Artists’ pigments. A Handbook of their history and characterisation. New York, Oxford: National Gallery of Art: Washington, Oxford University Press.

87

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Litharge and massicot: Thermal decomposition synthetic route for basic lead(II) carbonate and Raman spectroscopy analysis M. San Andrés, J.M. De la Roja & S.D. Dornheim Facultad de Bellas Artes, Universidad Complutense, Madrid, Spain

V.G. Baonza Facultad de Ciencias Químicas, Universidad Complutense, Madrid, Spain

ABSTRACT: Litharge (α-PbO) and massicot (β-PbO) are historical lead-based pigments which have been identified in paintings. In this work we reproduce in the laboratory the synthetic route of both by roasting lead white. The resulting products have been analyzed using X-ray Diffraction (XRD) and Raman microscopy. These analyses indicate that after 45 minutes of roasting at 873 K, lead white is transformed into massicot and this composition remains unchanged after 2 hours and 15 minutes. However, when the roasting is at 603 K, the transformation depends on heating time. In the first stages the roasting process gives rise to the formation of mixtures of (xPbCO3 · yPbO), with x and y = 1 or 2. Litharge appears after 2 hours and after 20 hours the compound formed is red lead (Pb3 O4 ) mixed with other lead oxides. The presence of different compounds influences the final colour of the pigment

1

INTRODUCTION

historical pigments, in which the properties of the products are described. For instance, litharge tonality is defined within a range varying from clear yellow to reddish yellow. Merrifield (1967) and Gettens (1996) define the colour range of litharge as yelloworange; Burgio et al. (2001) oscillating from yellow to red; Bearn (1923) describe three litharge forms: canary litharge (pale yellow), ordinary litharge (reddish yellow) and litharge in grudges (reddish). The wide chromatic variation found for this pigment very likely responds to changes in composition that could be related to the synthetic process. The aim of our study work is to reproduce, at laboratory scale, the synthetic process of these pigments and to study its evolution. In studies made by other authors, the effect of temperature was studied (Ciomartan et al. 1996); in our case, temperature was kept constant and the evolution of the process on terms of the roasting time is studied. Given the specificity of Raman microscopy, we use it to identify lead white and all the compounds formed at different stages of the transformation. Our results have been complemented with X-ray diffraction (XRD) experiments. In addition, we have carried out a colorimetric characterization of the products formed at each stage (De la Roja et al. 2007). Considering the previous studies (Burgio et al. 2001), our samples have been excited at low power with a He-Ne laser (632.8 nm) as excitation source.

There exist a large variety of lead pigments, all are of artificial origin and their synthetic procedures were known in the Antiquity. The less studied are litharge (α-PbO) and massicot (β-PbO) with the same elemental composition (like lead monoxide PbO) but differ in their crystalline structure. The old terminology used to refer to both is confusing and only a few recipes appear in the old texts (Dornheim et al. 2006). In general, the two varieties are rather reactive. However, in spite of these disadvantages, they have been identified in paintings. The characterization by Raman microscopy of litharge (α-PbO) and massicot (β-PbO) as pure chemical compounds is well defined in several available sources (Bell et al. 1997, Bouchard et al. 2003). Studies on the synthesis process of these pigments from roasting the lead white reveal that the evolution of the process is determined by different factors: amount of lead white, surface of sample directly exposed to heat and, of course, the temperature and duration of the roasting process. Although all these experimental variables can be precisely controlled nowadays, in the Antiquity it was not an easy task. Therefore the results obtained in the synthesis process could vary in relation to the composition of the product and its colorimetric characteristics. This fact is reflected in several literature sources on

89

2 2.1

EXPERIMENTAL Samples

We used lead white samples purchased to Panreac© to obtain the desired pigments. XRD analysis indicates that the pigment consists of lead (II) dihydro-biscarbonate [Pb3 (CO3 )2 (OH)2 ] and lead (II) carbonate (PbCO3 ). Litharge was obtained from lead white subjected to 603 K during 20 h, while the synthesis of massicot requires application of 873 K and this value was kept for 2 hours and 15 minutes. In both cases, we took samples at selected time intervals along the transformation process for analysis. When the processes were completed, the samples were removed, stirred and kept at ambient conditions. 2.2 Analytical techniques The resulting products have been analyzed by combined XRD and Raman spectroscopy, two techniques well-established in the literature (Ciomartan et al. 1996). The Raman spectrum was excited using the 632.8 nm line of a He-Ne laser. The equipment is composed of a 100X microscope, an ISA HR460 monochromator, and a CCD detector (1024 × 256 pixels). The spectral resolution is less than 4 cm−1 using a 600 grooves/mm holographic grating. The XRD study was performed with a Philips X‘PERT diffractometer using a voltage of 45 kV and intensity of 40 mA. This uses two slits, a 1◦ divergence slit for primary optics and a 1◦ anti-scatter slit (receiving slip 0.5 mm) for secondary optics. A curved Cu monochromator was used to eliminate the contribution of the Kβ line. 3

Figure 1. Pigments obtained by thermal decomposition of lead white at 873 K as a function of time. Table 1. Results of XRD analysis of thermal decomposition of lead white at 873 K as a function of roasting time. Sample (roasting time in hours)

RESULTS AND DISCUSION

Figure 1 shows the products obtained after roasting of lead white at 873 K. The analysis of the XRD patterns revealed that after 15 minutes treatment a mixture of litharge (α-PbO), massicot (β-PbO), and lead oxide carbonate [Pb3 O2 (CO3 )] is formed, together with unreacted lead white. However, if the treatment is extended up to 45 minutes, only the massicot phase is obtained as a primary product, together with a mixture of lead oxides (Pb2 O3 ) and (Pb5 O8 ). When the treatment is extended to 75 minutes, the final product is a mixture of massicot and Pb5 O8 . This composition remains unaltered after 135 minutes of treatment (see Table 1). The comparison of our Raman results with existing spectra (Ciomartan et al. 1996, Bell et al. 1997) confirms the results obtained by X-ray analysis. After 15 minutes the characteristic bands can be assigned to litharge (339 cm−1 ) and massicot (289 and 384 cm−1 ). Other bands can be attributed to the

Ref. Code (JCPDS)

Score

Crystalline species identified

0

00-047-1734 00-013-0131

60 42

PbCO3 Pb3 (CO3 )2 (OH)2

1/4

03-065-0398 03-065-0129 00-019-0681 00-001-0680 00-001-0687

51 45 38 20 18

α-PbO (litharge) β-PbO (massicot)

3/4

03-065-0129 00-052-0772 00-036-0725

85 17 12

1 1/4

03-065-0129 00-052-0772 03-065-0129 00-052-0772 03-065-0129 00-052-0772

68 12 67 11 56 48

1 3/4 2 1/4

Pb3 O2 CO3 Pb10 (CO3 )6 (OH)6 O 2PbCO3 · Pb(OH)2

β-PbO (massicot) Pb5 O8 Pb2 O3

β-PbO (massicot) Pb5 O8 β-PbO (massicot) Pb5 O8 β-PbO (massicot) Pb5 O8

lead (II) hydroxycarbonates [xPbCO3 · yPbO · (OH)z ] [1033, 1051 and 1056 cm−1 ]. After 45 minutes the bands assigned to massicot and a new band appears around 422 cm−1 . These spectral features remain unaltered after 135 minutes (288, 384, 422 cm−1 ) (see Table 2 and Fig. 2). The band appearing at 422 cm−1 might be related to the formation of the Pb5 O8 compound identified by X-ray analysis. Figure 3 shows the products obtained after a roasting treatment of lead white at 603 K. It is obvious that

90

Table 2. Raman bands (wavenumber/cm−1 ) of thermal decomposition of basic lead (II) carbonate at 873 K as a function of roasting time and reference compounds. Samples

0h – – – – – – – 414 w(br) – – – 678 vw – 694 vw – 835 vw

Reference Compounds

1/4 h

– 289 m – – 339 m – 384 vw – – – – – – – – – – – 1033 vw 1052 vs 1051 vw – 1056 vw(sh) 1369 w(br) – – – 1476 vw(br) – – – – – – –

3/4 h

2PbCO3 . Litharge Massicot Pb(OH)2 (Bell et al. (Bell et al. (Ciomartan 1 1/4 h 1 3/4 h 2 1/4 h 1997) 1997) et al. 1996)

Red Lead PbCO3 (Bell et al. (Ciomartan 1997) et al. 1996)

– 288 m – – – – 385 m – 422 w – – – – – – – – – 1051 vw – – – – – – –

– 289 s – – – – 384 m – 423 w – – – – – – – – – – – – – – – – –

– – 313 w – 340 vw – 390 w – – 480 vw 548 vs – – – – – – – – – – – – – – –

– 288 s – – – – 384 m – 424 w – – – – – – – – – – – – – – – – –

– 288 s – – – – 384 m – 422 w – – – – – – – – – – – – – – – – –

– 285 vw – – 336 w – – – – – – – – – – – – – – – – – – – – –

– 289 s – – – – 385 w – – – – – – – – – – – – – – – – – – –

267 vw – – 321 vw – – – 411 w – – – 679 vw – 693 vw 707 vw 837 vw 862 vw – 1050 vs – 1365 m – 1467 vw – – 1731 vw

– – – – – – – – – – – 673 w 682 w 694 vw 715 vw 837 w – – – 1054 vs 1364 m 1425 w 1476 m 1679 vw – 1735 vw

Very weak (vw); weak (w); medium (m); strong (s); very strong (vs); shoulder (sh); broad (br)

Figure 3. Pigments obtained by thermal decomposition of basic lead (II) carbonate at 603 K as a function of time.

Figure 2. Raman spectra of pigments obtained by thermal decomposition of lead white at 873 K as a function of time.

91

Table 3. Raman bands (wavenumber/cm−1 ) of thermal decomposition of basic lead (II) carbonate at 603 K as a function of time samples and reference compounds. Samples

Reference Compounds

0h

1/2 h

1h

2h

3h

4h

20 h

Litharge (Bell et al. 1997)

Massicot (Bell et al. 1997)

2PbCO3 . Pb(OH)2 (Ciomartan et al. 1996)

Red Lead (Bell et al. 1997)

PbCO3 (Ciomartan et al. 1996)

– – – – –

261 w 285 w – – –

262 vw 283 vw – – –

– 280 w – – 338 w

– 284 w – – 340 m

– 286 vw 310 vw – –

– 285 vw – – 336 w

– 289 s – – –

267 vw – – 321 vw –

– – 313 w – 340 vw

– – – – –

– – – 414 w (br) – – – – – 678 vw – 694 vw – 835 vw – – 1052 vs

360 w – – –

360 w – – –

261 vw 283 vw – – 338 vw (sh) 360 w – – –

359 m – 407 vw –

359 w – 406 vw –

– 389 vw – –

– – – –

– 385 w – –

– – – 411 w

– 390 w – –

– – – –

467 w – 494 vw – – 674 vw – 699 vw – – – 1034 vw 1052 s (sh) 1056 vs – – 1372 w (br) 1420 vw (br) –

467 w – – – – 675 vw – 697 vw – – – 1033 vw 1051 s (sh) 1055 vs – – 1373 w

465 w – – – 654 vw 674 vw – 698 vw – – – 1033 vw 1050 s (sh) 1057 vs – – 1371 w

463 m – – – 656 w 674 w – 696 vw – – – 1033 s 1049 vs

463 m – – – 655 w 675 w – 696 vw – – – 1035 s 1049 vs

– 475 vw – 545 vs – – – – – – – – –

– – – – – – – – – – – – –

– – – – – – – – – – – – –

– – – – – 679 vw – 693 vw 707 vw 837 vw 862 vw – 1050 vs

– 480 vw – 548 vs – – – – – – – – –

– – – – – 673 w 682 w 694 vw 715 vw 837 w – – –

– 1284 vw 1327 vw –

– – – –

– – – –

– – – –

– – – 1365 m

– – – –

1054 vs – – 1364 m

1425 vw (br) –

1428 w (br) –

– 1282 vw 1325 vw 1370 vw (br) 1431 m

1430 m

–

–

–

–

–

1425 w

–

–

–

–

–

1467 vw

–

1476 m

– – –

– – –

– – –

1679 vw 1693 vw –

1679 vw 1693 vw –

– – –

– – –

– – –

– – 1731 vw

– – –

1679 vw – 1735 vw

– – – 1369 w (br) – 1476 vw (br) – – –

Very weak (vw); weak (w); medium (m); strong (s); very strong (vs); shoulder (sh); broad (br)

the roasting time leads to important colour variations. Our analyses confirm that this phenomenon is intimately related to the composition of the transformed products (Tables 3 and 4, Fig. 4). The analysis of the XRD patterns (Table 4) indicate that the roasting process gives rise to the formation of mixtures of (xPbCO3 · yPbO), with x and y = 1 or 2. Litharge, mixed with lead oxides carbonates, is obtained after 3–4 hours of roasting time. After 20 hours the product obtained is a mixture of red lead (Pb3 O4 ) and Pb2 O3 , as confirmed

by XRD. Previous studies suggested that red lead already appears after 12 hours (De la Roja et al. 2007). The comparison of our Raman spectra and those reported by Ciomartan (1996) reveals that some characteristic bands can be attributed to lead oxide carbonates: 359, 463, 656, 676, 697 and 1431 cm−1 . The bands of litharge: 285 and 338 cm−1 appear after 2 h and after 20 h the compound formed is red lead (Pb3 O4 ): 310, 389, 475 and 545 cm−1 (Fig. 4 and Table 3).

92

Table 4. Results of XRD analysis of thermal decomposition of lead white at 603 K as a function of roasting time. Sample (roasting time) (h)

Ref. Code (JCPDS)

Score

Crystalline species identified

00-047-1734 00-013-0131 00-017-0730 00-048-1888 00-019-0681

60 42 46 28 23

PbCO3 Pb3 (CO3 )2 (OH)2 Pb3 C2 O7 Pb2 OCO3 Pb3 O2 CO3

1

00-017-0730 00-019-0681 00-048-1888

44 37 31

2

00-019-0681 00-017-0730 00-019-0682

46 45 22

Pb3 C2 O7 Pb3 O2 CO3 Pb2 OCO3 Pb3 O2 CO3 Pb3 C2 O7 Pb2 OCO3

3

00-019-0681 00-017-0730 00-048-1888 03-065-0401 00-019-0681 03-065-0400 00-008-0019 00-001-0654 01-076-1832

53 26 21 18 54 36 52 25 47

Pb3 O2 CO3 Pb3 C2 O7 Pb2 OCO3 α-PbO (litharge) Pb3 O2 CO3 α-PbO (litharge) Pb3 O4 (red lead) Pb3 O4 Pb2 O3

0 1/2

4 20

depends strongly on both the temperature and the roasting time.The results for the two temperatures used in this work (603 and 873 K) indicated that the early stages of the process leads to the formation of complex mixtures of lead oxide carbonates. After 45 minutes of roasting at 873 K massicot (β-PbO) is formed along with other lead oxides. At still higher roasting times the composition seems to stabilize and the transformed product is a mixture of massicot and Pb5 O8 . Longer roasting times (3–4 hours) are required at lower temperatures to obtain litharge (β-PbO). Finally, if the roasting time is extended up to 20 hours the product recovered is red lead (Pb3 O4 ) mixed with other lead oxides. Our study is directly related to the conservation scientific and art work documentation, as our results will help to interpret future analysis on pictorial samples. For example, those analyses in which litharge or massicot are mixed with other compounds (lead white, lead carbonate, etc) should lead to the conclusion that the presence of the secondary compounds might be associated to the synthetic process. Furthermore, in the case of litharge, we have verified the wide chromatic range derived from the roasting process (from yellowish to reddish brown). These observations based on our results points to the question that, in the Antiquity, other pigments could be obtained from lead white (for instance, the pigment known as “pardillo de albayalde” according to the Spanish terminology.

ACKNOWLEDGEMENTS This work was financed under Project BHA200202085, supported by the MCYT. We also express our gratitude to the Centre for X-ray Diffraction of the Universidad Complutense de Madrid.

REFERENCES Bearn, J.G. 1923. The Chemistry of Paints, Pigments and Varnishes, London, Ernest Benn Limited. Bell, I.M., Clark, R.J.H., Gibbs, P.J. 1997. Raman spectroscopy of natural and synthetic pigments (pre- ∼1850AD), Spectrochimica Acta Part A 53: 2159–2179. Bouchard, M. & Smith, D.C. 2003. Catalogue of 45 reference Raman spectra of minerals concerning research in art history or archaeology, especially on corroded metals and coloured glass, Spectrochimica Acta Part A, 59: 2247–2266. Burgio, L. Clark, R.J.H. & Firth, S. 2001. Raman spectroscopy as a means for the identification of plattnerite (PbO2 ), of lead pigments and their degradation products, Analyst, 126: 222–227. Ciomartan, D. A. Clark, R.J.H. McDonald, L.J. & Odlyha, M. 1996. Studies on the thermal decomposition of basic(II)

Figure 4. Raman spectra of pigments obtained by thermal decomposition of lead white at 603 K as a function of time.

Overall, our these results confirm the XRD analysis listed in Table 4, except for the presence of Pb2 O3 , which could be identified in our Raman experiments.

4

CONCLUSIONS

Our results demonstrate that the thermal decomposition of lead white is a quite complex process that

93

carbonate by Fourier-transform Raman spectroscopy, Xray diffraction and thermal analysis. J. Chem. Soc., Dalton Trans., 3639–3645. De la Roja, Sancho, N, San Andrés, M., Baonza V.G. 2007. Obtención de litargirio a partir de la tostación de blanco de plomo. Caracterización cromática de los productos obtenidos. Proceedings VIII Congreso Nacional del Color, Madrid 19–21 September, 117–118. Dornheim, D. & San Andrés, M. 2006. Litargirio y masicote. Terminología, propiedades y usos. Reproducción

a escala de laboratorio de algunos de sus procesos de obtención. In Proceedings XV Congreso de Conservación y Restauración, Murcia, 21–24 Octubre, 2004, Vol. 1: 535–546. Gettens, R. J. & Stout, G. L. 1996. Paintings Materials, a Short Encyclopedia, New York: Dover. Merrifield, M.P. 1967. Medieval and Renaissance treatises on the arts of painting: original texts with English translations, New York, Dover Publications.

94

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Contamination identification on historical paper by means of the NIR spectroscopic technique M. Sawczak Polish Academy of Sciences – IFFM, Gdansk, Poland

A. Kaminska Agency for Integration of Conservation Activities, Gdansk, Poland

ABSTRACT: The soiling on historical documents and artworks on paper represents one of the most serious conservation problems. The selection of an effective and secure cleaning method preceded by a reliable examination of the contamination chemical composition is a key for recovering and preservation of the historical and aesthetical value of the paper object. The soiling and substrate analysis has therefore to be considered as an important part of the conservation project. In this work the potential usefulness of Near Infrared (NIR) Spectroscopy for analysis and classification of different types of stains on paper documents is studied.

1

INTRODUCTION

relatively deep, even few millimetres into the sample. As a consequence of low absorption, NIR is not a particularly sensitive technique but it can be very useful in probing bulk material with little or no sample preparation what makes this technique non-invasive. The molecular overtone and combination bands seen in the near IR are very broad and it can be difficult to assign absorption bands to specific chemical components. For analyzing the results obtained from NIR measurements the multivariate calibration techniques, e.g. Principal Components Analysis (PCA), Partial Least Squares (PLS), etc. are often employed to extract the desired chemical information. In this work, the absorption spectra in the near IR region were obtained for a variety of original and model stains and adhesives on historical and contemporary paper samples.

The application of analytical techniques is the essential part of evaluating the collection condition as well as an important part of conservation or restoration practice. These techniques can detect a wide range of issues including evidence of damage, vulnerability, etc. These evaluations lead to conservation decisions or maintenance guidance for the curators. The majority of the analytical techniques used for paper analysis are destructive and require a relatively great amount of samples that in the case of valuable pieces of art or historical documents is almost impossible. The necessity of implementation of non-destructive analytical techniques enabling examination of historical papers seems to be a crucial issue. Scientific interest is currently focused on the applicability of NIR for non-destructive and fast evaluation of paper condition (Lichtblau & Anders 2004, Trafela et al. 2007). The purpose of this research project is to establish whether it is possible to implement NIR for classification of contamination on historical papers. We expect a sufficient level of soiling recognition can be achieved with those techniques. Furthermore, we expect that soiling recognition by NIR techniques could enable objects assessment by providing reliable data and be an alternative for the chromatographic and spectroscopic techniques which demand sampling. NIR is based on molecular overtone and combination vibrations. Since the transitions are forbidden by the selection rules of quantum mechanics, the absorption in the near IR region is small and the radiation can penetrate

2

EXPERIMENTAL METHOD

NIR spectra were collected in the reflectance mode using a 0.5 m spectrometer equipped with a broadband radiation source (quartz tungsten bulb, 3000 K) and a PbS detector. Spectral data were acquired in a diffuse reflection by scanning between 1300 and 2500 nm with a step of 2 nm. Data of each sample were collected at five different locations. The absorbance spectra represented as log(1/R), where R is reflection coefficient, were obtained by averaging measurements and comparing them to the reference sample (Spectralon). The statistical analysis of the

95

experimental data was carried out using Principal Component Analysis (PCA) method. Spectral data were processed using own procedures written in Scilab software. 3

RESULTS AND DISCUSSION

Typical NIR spectra consist of broad overlapping bands which are very difficult to assign to specific chemical compounds. Figure 1 presents representative NIR spectra recorded for non-contaminated, contemporary and historical papers. The analysed papers are listed in Table 1. The curves have been offset for clarity. The spectra are visually very similar. Indicated selected bands are characteristic of C=O, C-C and C-H bonds of cellulose molecules. A global PCA performed on all data enable to separate them into groups. The main objective of PCA is to reduce the dimension of the matrix of data so that useful information can be extracted from the overlapped chemical information. Principal components (PC) can be taken as the projection of the original data in a new space. Results presented on Figure 2 show that spectra, represented by single points in the PC space, can be classified into separate groups. Historical papers having very similar composition are located in one group while contemporary papers characterized by different chemical composition are classified in separate groups. Applying PCA to historical papers enable to split them depending on individual chemical and physical properties. Building appropriate calibration model for PCA, the qualitative analysis of papers can be carried out. 3.1

Figure 1. Representative NIR spectra recorded for non-contaminated, contemporary and historical papers. Table 1.

Description of paper samples.

contemp_1 contemp_2 contemp_3 archiv_6, 8, 15, 20, 22

Whatman filter paper contemporary newspaper offset print paper (1986) historical rag papers 17th, 18th century

Model contamination

For the experiment, a set of samples using three contemporary papers (Table 1) covered with model contaminants was prepared. Substances often found on historical papers such as fats, waxes, tea, coffee, milk, were used as contaminants. Methylcellulose, starch, etc., as adhesives. In most cases, contaminants and residues responsible for the soiling and discoloration of paper cannot be determined by characteristic spectral bands. Typical examples of soiling resulting from inappropriate maintenance of historical paper indicate that the data can be grouped depending of chemical composition similarities. Global PCA analysis performed to differential spectra obtained by subtracting the spectra of contaminated and clean paper indicated that substances can be split into three groups: group A – fats (mainly: triglycerides, glycosides, alkane hydrocarbons); group B – adhesives: long chain carbohydrate, polymers; C – substances containing mainly carbohydrates (honey, fruit juice, etc). In

Figure 2. First two PC scores for spectra presented in Figure 1.

Figure 3, representative spectra of samples covered with different fats and paraffin are presented. There is a clear split in the group A (Fig. 4): margarine and oil gather together away from the principal components as an effect of their chemical composition similarities, mostly characterized by monosaturated and polysaturated fatty acids. The difference in the chemical nature of the other two substances: milk (4% fat: saturated, monosaturated and polysaturated fatty acids) and paraffin (paraffin wax, C25 H52 ) is

96

8

differential spectra

0,5

1725 1755

2295 2340

paraffin

Absorbance [a.u.]

6

Absorbance [a.u.]

margarine

oil

0,0 4

2

clean paper contaminated area

-0,5

0

-2 milk

1400

1600

1800

2000

2200

2400

λ [nm]

1400

1600

1800 λ [nm]

2000

2200

Figure 5. Spectra recorded for finger print near the pagination of one page of the book.

2400

Figure 3. Differential spectra of model samples (Whatman filter paper) contaminated with fats and paraffin. 8

1732 1766

differential spectra

Absorbance [a.u.]

2432

4 2310 2348

6

2 0

4

-2

clean paper contaminated area

2

-4 -6

0

-8 -2 1400

1600

1800

2000

2200

2400

λ [nm]

Figure 6. Wax stain from a page of a old print.

gelatine

methylcellulose

Absorbance [a.u.]

Figure 4. Result of PCA analysis of spectra recorded for model paper contaminated with fats and paraffin.

remarkable, although they are commonly considered as a fatty soiling on paper, too. Ages of candle lighting for reading left significant amounts of traces in the form of greenish, yellowish or brownish wax stains on book pages. Traces of fat and wax are likewise recognizable on the historical papers. Spectra of a trace of a finger print (usually containing traces of fat, lactic acid, particular matter, etc) found near the pagination of one page of the book and of a wax stain from a page of an old print are shown in Figures 5 and 6. The typical C-H bonds located near 1730, 1760, 2300 and 2340 nm can be observed distinctly. In Figure 7, representative differential spectra measured for samples covered with typical adhesives

starch

polyvinyl acetate

glue stick

1400

1600

1800

2000

2200

2400

λ [nm]

Figure 7. Differential spectra of model samples (Whatman filter paper) contaminated with adhesives.

97

Figure 8. First two PC scores for spectra presented in Figure 7.

Figure 9. First two PC scores for spectra recorded for samples contaminated with substances containing carbohydrates. Figure 10. Result of PCA analysis of spectra recorded for three contemporary papers covered with model contaminants.

applied on paper are shown. Each of the adhesives split into different space of PC1 and PC2 (Fig. 8). It should be expected according to their chemical nature: polysaccharides, protein, polymers as vinyl polyacetate and vinyl polypyrolidone (glue stick). This gives us the possibility to distinguish different kinds of adhesives used in paper conservation and book binding. In Figure 9, results of PCA analysis of group C, substances containing carbohydrates, are presented. The main compounds of these substances are saccharides and polysaccharides. The group shows the highest similarities for honey and sugar. The split into the other groups is probably caused by the difference in the fructose level in the fruits and additional components. Preparation of model samples sets for PCA procedure enables to classify unknown contaminants regarding its chemical composition. 3.2

the set of samples was prepared by covering three different contemporary papers (Table 1) with four model contaminants (milk, stick glue, methylcellulose and cream). In Figure 10, results of the PCA analysis performed on differential spectra of contaminated and clean paper are presented. Despite considerable differences in the composition of model papers used in the experiment, distinction between substances used as contaminants is still possible.

4

CONCLUSIONS AND FUTURE RESEARCH

The application of NIR spectroscopy for nondestructive, user-friendly recognition of contamination on paper documents promises to be a powerful tool for performing fast, in situ sample examination. The purpose of this paper was to demonstrate the potential usefulness of this technique for classification of contaminants, residues, adhesives, etc. Results of this work evidence that the technique can be used to classify groups of substances of similar chemical composition. The investigation is in the initial stage and will be continued for a large variety of materials and contaminants in order to obtain comprehensive and reliable characterization.

Influence of paper in PCA results

Changes in the spectra of paper due to contamination are very small. Moreover, in the case of historical papers, variability of paper parameters results in big differences of spectra. The question is whether it will be possible to classify different contaminants and distinguish compounds of very similar chemical composition using the NIR technique. For the experiment,

98

ACKNOWLEDGEMENTS

the International Conference, Durability of Paper and Writing, November 16–19, 2004, Ljubljana, Slovenia. Scilab scientific software homepage [www.scilab.org]. Trafela, T., Strlic, M., Kolar, J., Lichtblau, D. A., Anders, M., Pucko Mencigar, D. & Pihlar, B. 2007. Nondestructive Analysis and Dating of Historical Paper Based on IR Spectroscopy and Chemometric Data Evaluation. Anal. Chem. 79: 6319–6323.

Work is supported by the Ministry of Science and Higher Education under the project No 2059/B/T02/2007/33 REFERENCES Lichtblau, D. & Anders, M. 2004. Characterization of paper by near infrared spectroscopy. Proceedings of

99

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Three dimensional survey of paint layer profile measurements E. Pampaloni, R. Fontana, M.C. Gambino, M. Mastroianni & L. Pezzati Istituto Nazionale di Ottica Applicata (CNR – INOA), Firenze, Italy

P. Carcagnì, R. Piccolo & P. Pingi Istituto Nazionale di Ottica Applicata (CNR – INOA) – Sez. di Lecce, Arnesano (LE), Italy

R. Bellucci & A. Casaccia Opificio delle Pietre Dure (OPD), Firenze, Italy

ABSTRACT: The quantitative morphological analysis of a painting surface allows to evidence form defects and thus to study their influence on the stability of the paint and preparatory layers, as well as on the support. Therefore a three-dimensional survey can be very useful in planning the restoration intervention of a painting. In this work we present the results of the surface analysis carried out on the painting “Ultima Cena” by Giorgio Vasari. This panel painting is severely affected by paint film wrinkling produced as a consequence of the flood that occurred in Florence in 1966. Our analysis, accomplished to quantify the lengthening of the paint layer with respect to the one of the support, was made in order to plan the restoration intervention, which was performed on 25 profiles separated each by 10 cm to cover the whole painting surface. A data analysis, based on morphological filtering named “Rolling Ball” transformation, was used to evaluate the length difference between the paint layer and the panel support along each profile.

1

INTRODUCTION

Shape represents highly significant data for the historical and artistic evaluation, the diagnostic analysis and the conservation of an artwork. The possible scenarios involved in the use of 3D digital models range from the monitoring of deterioration due to pollutants, to the realization of digital archives, from reverseengineering to fast-prototyping, and from the analysis of the conservation state to the monitoring of restoration interventions. Moreover, it is possible to monitor the form of variations by computing the differences between measurements at different times. At present, a variety of instruments is available on the market for surface measurement. The commonly used techniques for in situ roughness measurements are contact techniques and they make use of stylus profilometers. The sample surface is investigated by means of a stylus or needle (often a diamond point) that is moved along the surface, its profile is then recorded. The system is usually calibrated with a known flat surface and the depth information is obtained by calculating the difference between sample and reference measurements. These profilometers have a very good axial resolution (up to some tens of nanometers), whereas lateral resolution depends on stylus diameter. The typical measured areas extend to

a few tens of square millimeters. This method is suited for measuring hard surfaces, but it is not applicable for surveying frail and precious objects like paintings, as stylus sharpness can damage the surface causing micro-scratches. In the diagnostics of paintings the non-contact characteristic is a mandatory step; this requirement makes the optical techniques particularly suitable for this purpose. That is why they are widely used and extremely well received in the field of conservation together with their effectiveness and safety (Bertani et al. 1990, Fabbri et al. 2000, Fontana et al. 2003, Carcagnı et al. 2007, Bellucci et al. 2007). Optical techniques for shape measurements are often derived from industrial metrology but the peculiarity of an artwork does not allow for a straightforward application. In fact, industrial manufacture is generally regular in shape, with uniformly coloured surfaces. On the contrary, artworks are unique in their shape, having polychrome and highly contrasted surfaces such as painting surfaces. Conoscopic micro-profilometry is particularly well suited for surveying the surface of paintings due its unsensitivity to color contrast, it enables measurements on surfaces with almost any reflectivity and it allows the survey of microscopic details working with an incident angle very close to grazing incidence.

101

Figure 1. The working principle of the conoscopic system.

Figure 2. The laser scanner micro-profilometer during the measurements.

2

INSTRUMENTS

Our conoscopic micro-profilometer is composed of a rangefinder (Conoprobe 1000 by Optimet) mounted on a high-precision scanning system. The Conoprobe is substantially a video camera within which, between objective and CCD, is placed the conoscopic module, consisting in a uniaxial birefringent crystal sandwiched between two circular polarizers (Fig. 1). The probe working principle is as follows: a light beam, projected by a diode laser, is both reflected and back scattered by the sample surface and then is collected by the lens. The conical light beam, after impinging on the crystal is split into two beams, the ordinary and extraordinary one. These two beams running along slightly different optical paths produce an interference pattern that depends on the beam aperture angle and is related to the distance of the object. By measuring the fringe spacing, the distance of the investigated point from the conoscopic probe can be retrieved (Sirat et al.

1985, 1988, Charlot 1988). The probe is equipped with a 250 mm lens setting an axial resolution better than 15 µm and a measurement range of ±90 mm at a standoff distance of about 240 mm. The overall accuracy is better than 100 µm and the transversal resolution of about 100 µm. The system allows measurements on a maximum area of about 1.5 × 1 m2 . The device has a maximum acquisition rate of 800 points/s, but due to downtimes and scanning parameterization, the typical acquisition rate ranges from 100 to 500 points/s. The whole system is computer controlled (Fig. 2). 3 3.1

MEASUREMENTS AND DATA ANALYSIS Painting surface measurement

We performed a profile analysis on the surface of a panel painting severely affected by paint film wrinkling and colour raisings. This crumpling of the paint layer is the consequence of the flood caused

102

Figure 3. The “Ultima Cena” by Giorgio Vasari (1546), 263 × 660 cm2 . Photo executed in 1956. The analyzed panel is enclosed by the dashed rectangle.

Figure 4. The measured panel: a) photograph in 1956, b) recent photograph with velinatura, c) raking light photograph, d) photograph of the back.

by the Arno river in Florence in 1966. The painting entitled “Ultima Cena” realized in 1546 by Giorgio Vasari for the monastery “Le Murate” in Florence, is currently under restoration at the laboratory of the Opificio delle Pietre Dure in Florence, where it arrived in 2004 after a period characterized by a long wandering in several restoration laboratories without success. The conservation aim is both the historical reconstruction of unlucky vicissitudes suffered by the panel painting and its restoration. Our analysis, accomplished to measure the amount of lengthening of the paint layer with respect to the support, allows the restorers to plan an appropriate restoration intervention.

Because of the painting huge dimensions, we performed our surface analysis only on the second panel constituting the artwork (Figs. 3, 4), that is representative of the conservation status of the whole painting. In order to achieve information useful for planning the restoration intervention of the painting surface and not excessively time-consuming, we acquired 25 profiles every 10 cm along the panel length (Fig. 5), with a sampling step of 0.25 mm (4 points/mm). Microprofilometry is a surface measurement that gives information on the painting layer and not on the morphology of the support. Measuring the rear part of the panel will not give the correct information because the two surfaces of the support are

103

Figure 5. The measured panel (raking light photograph) and the 25 measured profiles.

not supposed to be identical. In order to evaluate the support length we need an appropriate data processing. The application of morphological transformations, like the “Rolling Ball” filtering, allows to discriminate the contribution of the wrinkled paint layer from that one of the support in the acquired profile curve. We implemented the “Rolling Ball” algorithm in the MatLab computing environment. 3.2 The Rolling Ball transformation The “Rolling Ball” transformation is a mathematical morphological algorithm first proposed by Sternberg (Sternberg 1983) to minimise image background noises. It consists in the application of morphological openings or closings (Sternberg 1986, Hashim 1996, Lee et al. 2005) to grayscale images by using a spheric structuring element. To best understand of this algorithm, we consider a grayscale image as a surface where bright areas are hills or peaks and dark areas are pits or valleys, and then we consider a large sphere rolling over the grayscale surface tracing a path as it rolls. This path represents the set of points where the ball fits the surface. This new surface is smooth relative to the original. By taking the grayscale difference image, we obtain an image of all the places where the ball could not fit into crevaces in the surface because it is too large. In this example, the “Rolling Ball” algorithm is a morfological transformation (closing, dotted line in Fig. 6a) followed by an image subtraction. Similarly, the morphological opening can be visualized as a ball rolling under the grayscale surface. In this case, any

Figure 6. The Rolling Ball algorithm representation: closing dotted line, opening dashed line, b) the top hat transformation.

protrusions of high curvature, such as sharp edges and ridges, are lost by the procedure (opening, dashed line in Fig. 6a). Therefore the “Rolling Ball” transformation describes the smoother features of a surface. The grayscale difference image, obtained by subtracting the opening (closing) “Rolling Ball” transformation from the original surface, is called a “Top Hat” (“Bottom Hat”) transformation (Fig. 6b). The main problem when using these morphological filters is to choose the right ball radius R to seize the desired surface features. In our case it was necessary to discriminate the crests of the wrinkled paint surface from the smooth surface of the wooden support and thus we used the “Top Hat” transformation. In order to choose the appropriate ball radius, we calculated the length of the curve L for each

104

Figure 7. a) The length L as function of the ball radius R, b) the derivative dL/dR as function of R.

profile as a function of R, by means of the opening “Rolling Ball” transformation. The typical result for the profile is shown in Figure 7a. Obviously, for an increasing R, the length L decreases and for R → ∞, L(R) tends to the limit represented by the length of the projection of the profile on a horizontal line or equivalently by the sampling step length multiplied by the number of samples. But, for particular R values, L(R) decreases more slowly and presents a flex indicated by the arrow in Figure 7a. This, can be seen more clearly in the derivative dL/dR, where the curve presents a local maximum (the arrow in Fig. 7b). For this value of R the Opening “Rolling Ball” algorithm extracts all the profile features corresponding to the smooth surface of the wooden support. For instance, Figure 8 shows the 22nd profile, compared to the curve resulting by the Opening “Rolling Ball” algorithm for R = 1250 mm, equivalent to a points number N = 5000 (indicated by the arrows in Fig. 7). Figure 4d shows that the panel support is constituted by four planks linked by wooden crossbars, as is also visible in Figure 8. In fact, there are three local minima in the “Rolling Ball” filtered curve corresponding to the union points between two adjacent planks. These minima are located near the abscissas at 400, 650 and 880 mm.

4

RESULTS

For each profile the lengths of the paint layer and the wooden support were computed, and their differences L are shown in Figure 9. As the painting panel is composed by four planks, for restoration purpose, its important to evaluate not only the length difference L between paint layer and the wooden support over the whole panel, but we have also to consider how it is distributed on each plank. So all the profiles were

Figure 8. The 22nd profile (continuous line) compared to the “Rolling Ball” filtered curve(dashed line) for R = 1250 mm. The arrows indicate the union points between two adjacent planks. The ordinate scale is 10 times the abscissa scale.

subdivided into four parts, one for each plank, and the length differences between paint layer and support were calculated for each part. Figures 10a–d, show the results obtained by this calculation. The first plank variations range between 0.3 and 2.9 mm for the first ten acquired profiles, however in the other profiles it ranges between 4 and 8 mm. The maximum value for the deformation in this plank is 8.2 mm, for the profile number 21, which is the profile with the maximum dimensional variation between pictorial film and support (near 20.03 mm). In the second plank, dimensional variation is ranging between 1.5 and 4 mm (there is a maximum of 6 mm in the profile number 18 where detachments are more easily discernible). The third plank shows variations ranging between 0.4 and 3 mm which are nearly uniform along the whole plank. The fourth plank presents a more complex situation, the profiles from 1 to 10 have a dimensional variation ranging between 1 and 3 mm, however, the profiles from 11 to 16 have a variation ranging between 3 to 4 mm. Finally the profiles from 17 to 23 have a deformation ranging between 3 to 7 mm. Clearly we have more variations in the first (28.5 cm long) and in the fourth (38.5 cm long) plank; probably due to the different working process used for their preparation and to their spatial position which

105

Figure 9. Length differences between the paint layer and the support along the painting.

Figure 10. Paint-support length differences for the first a), second b), third c) and fourth d) planks.

allows them to deform more easily. The two planks in the middle, being linked to the others, present less deformations (Figs. 10a–d). From the acquired profiles it is possible to obtain not only quantitative information about dimensional variations of the support with respect to the preparatory layer and to the painting layer, but also about the heights of the wrinkled paint layer. Considering the results shown in Figure 9, we estimate the difference between the paint layer length and the whole surface of the plank in contact with the preparatory layer. This difference is about 21.03 mm and its probably due to the planks shrinkage after the drying process. Wood dimensional variation is not constant along each plank due to the natural anisotropy of this material and to its different permeability. When the amplitude of wrinkling is greater, the shrinkage of the planks is also greater. Moreover the lower part of the panel is more affected by wrinkling (near 15 mm). These bigger deformations are surely due to the different water absorption of each plank

and consequently to the wrinkling of the paint layer. 5

CONCLUSIONS

In this work we have presented the results of a surface analysis performed on 25 profiles of the “Ultima Cena” by Giorgio Vasari, a panel painting severely affected by paint wrinkling, consequence of the flood occurred in Florence in 1966. These profiles were surveyed by means of an optical high resolution micro-profilometer. The results of the data analysis, based on the “Rolling Ball” morphological filtering, were used to evaluate the difference length between the paint layer and the panel support along each profile. This allowed also the evaluation of the height of each single wrinkle. This investigation produced a lot of useful informations to plan the restoration intervention in order to decide whether to substitute integrally the original panel support or to keep it with a few interventions.

106

REFERENCES Bertani, D., Cetica, M., Poggi, P., Puccioni, G., Buzzegoli, E., Kunzelman, D. & Cecchi, S. 1990. A scanning device for infrared reflectography. Studies in Conservation 35: 113– 119. Bellucci, R., Carcagnı, P.L., Della Patria, A., Fontana, R., Frosinini, C., Gambino, M. C., Greco, M., Mastroianni, M., Materazzi, M., Pampaloni, E., Pezzati, L., Piccolo, R. & Poggi, P. 2007. Integration of image data from 2D and 3D optical techniques for painting conservation applications. The Imaging Science Journal 55: 80–89. Carcagnı, P., Della Patria, A., Fontana, R., Greco, M., Mastroianni, M., Materazzi, M., Pampaloni, E., Pezzati, L. 2007. Multispectral imaging of paintings by optical scanning. Opt. Lasers Eng. 45: 360–367. Charlot, D. 1988. Holographie conoscopique – reconstructions numeriques. Annales des Telecommunications 9: 23–26. Fabbri, F., Mazzinghi, P. & Aldrovandi, A. 2000. Tecnica di identificazione di materiali pittorici attraverso l’acquisizione di immagini digitali multispettrali in fluorescenza UV. Quaderni di Ottica e Fotonica 6: 94–104.

Fontana, R., Gambino, M.C., Greco, M., Pampaloni, E., Pezzati, L. & Scopigno, R. 2003. High-resolution 3D digital models of artworks. Proc. SPIE 5146: 34–43. Fontana, R., Gambino, M.C., Greco, M., Marras, L., Materazzi, M., Pampaloni, E. & Pezzati, L. 2003. Thediagnostics of statues: a high-precision surface analysis of roughness of Michelangelo’s David. Proc. SPIE 5146: 236–243. Hashim, M. 1996. New Texural Extraction Method Using Rolling Ball and Riping Membrane Transforms. Proceeding of the 17th Asian Conference on Remote Sensing ACRS. Lee, J.R.J., Smith, M.L., Smith, L.N. & Midha, P.S. 2005. A mathematical morphology approach to image based 3D particle shape analysis. Machine Vision and Applications 16: 282–288. Sirat, G. Y. & Psaltis, D. 1988. Conoscopic holograms. Opt. Comm. 9: 243–245. Sirat, G.Y. & Psaltis, D. 1985. Conoscopic holography. Optics Letters 10: 4–6. Sternberg, S. R. 1983. Biomedical image processing. IEEE Computer Society 16: 22–34. Sternberg, S. R. 1986. Grayscale morphology. Comp. Vision, Graphics, Image Process. 35: 333–355.

107

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Use of LA-ICP-MS technique with SEM-EDS analysis in the study of finishing layers L. Rampazzi & B. Rizzo Dipartimento di Scienze Chimiche e Ambientali, Università dell’Insubria, Como, Italy

C. Colombo, C. Conti & M. Realini Istituto per la Conservazione e la Valorizzazione dei Beni Culturali, CNR, Milano, Italy

ABSTRACT: The investigation of the finishing layers in artworks is of major concern, as important information on the artistic technique can be suggested. The study is usually carried out with Scanning Electron Microscopy equipped with Energy Dispersive Spectroscopy (SEM-EDS), but in those cases where very similar compositions are detected, more sensitive techniques are called for. Laser Ablation-Mass Spectrometry equipped with Plasma source (LA-ICP-MS) allows very sensitive element analysis of solid surfaces sampled by means of a laser. It is an emerging methodology, applied only in a few cases in the field of cultural heritage. The integration of the two techniques has been studied and applied on the finishing layers of Baroque stucco decorations, in order to define the artistic technique in terms of raw materials, stratigraphy and working tools.

1

INTRODUCTION

The surfaces of artworks may answer important questions about the artistic technique such as: “Which materials have been used?. How?” Stratigraphy provides all the information on the raw materials, the technological process, and the finishing layers applied in the past. In some cases, shedding light on the technique could be very problematic: the materials detected in the decoration of mortars and stuccoes are, as an example, quite the same. In particular, the technique of stucco decorations is difficult to define, since: – the finishing layers are often complex, as many layers are present; – the layers may be very similar from the compositional point of view, being the binder mainly composed by lime; – the thickness of the layers can be very small, around tens of micron. The structure, the morphology and the element composition of stuccoes are usually determined by means of Scanning Electron Microscopy equipped with Energy Dispersive Spectroscopy (SEM-EDS) (Toniolo et al. 1998), characterised by a high visual resolution and a good sensitivity. Those techniques are usually resolutive and successful in defining

the artistic technique. When a complex stratigraphy or a very similar composition is observed, trace components become the qualifying and peculiar elements for the classification of the samples and conservation scientists call for sensitive analyses. Laser Ablation-Mass Spectrometry equipped with Plasma source (LA-ICP-MS) is an emerging analytical technique which allows a very sensitive element analysis of solid samples. The surface is observed with an optical microscope and the area of interest ablated by means of a laser beam and analysed. The technique has been often employed in environmental investigations of solid surfaces and only in a few cases in the field of cultural heritage, particularly for ceramic, metal and glass objects (Manson & Mank 2001, Robertson et al. 2002, Sanborn & Telmer 2003, Dussubieux & Van Zelst 2004, Wagner & Bulska 2004). The element analysis is much more sensitive than the one performed by SEM-EDS technique, but the resolution of the microscope is usually lower, thus preventing the precise mapping of elements. The authors have set up a methodology which has taken advantage of SEM-EDS high resolution and of LA-ICP-MS high sensitivity, in order to overcome the critical points of both techniques. The new methodology has been tested during a project called L’arte dello Stucco nel Parco dei Magistri Comacini – Valorizzazione, conservazione

109

e promozione (The art of stucco in the Magistri Comacini Park – Enhancement, conservation and promotion), funded by EU Interreg IIIA Program and aimed to enhance the Baroque stucco decorations in the borderline area between Como (Italy) and Lugano (Switzerland). The aim of the integrated investigation was to corroborate decorating phases, ‘recipes’ or hypotheses about building sequences, previously suggested, on the basis of historical and artistic data, for the stucco decorations inside three churches: San Lorenzo in Laino, Santa Maria dei Ghirli in Campione d’Italia and Santa Maria del Restello in Castiglione Intelvi. The stucco cycles date back to 17th and 18th century and have been made by the Magistri Comacini, among the most important artists of Northern Italy.

2

Optical microscopy

Polished cross-sections of the samples were observed in reflected light using a Leitz Ortholux microscope with Ultropack illuminator equipped with a digitalisation image system. Fragments and cross sections were observed using Leitz Wild M420 stereomicroscope equipped with a digitalisation image system.

2.2

Analyses were performed by ICP-MS (Thermo X-serie due, 10 ms dwell time, 1 channel of mass, sweep number 100, power 1300 W) equipped with a Laser Ablation system (UP266, New Wave Research). The ablation step was optimised as follows: laser output 1.56 mJ, surface power (55%), repetition rate 20 Hz, spot size 20 µm, scanning speed 2 µm/s in continuous mode. A pre-ablation step was performed to clean the sample surfaces using a weak energy laser spot. Element signals were corrected with an internal standard, as used in LA-ICP-MS methodology to normalize the data toward different local ablation efficiencies and general instrumental drift. In this study, Cu was selected due to the homogeneous distribution on the samples surface.

MATERIALS AND METHODS

Sampling was carried out in areas suggested by art historians, supplying a number of representative stratigraphies. Several micro samples were collected using a scalpel according to the indication of UNI-Normal rules 3/80 (Raccomandazione Normal 3/80 ‘Campionamento’ ICR-CNR 1980). Polished cross-sections of the samples were prepared, observed by an optical microscope and analysed by SEM-EDS, in order to determine the major and minor elements and to define in detail the stratigraphy, i.e. morphology, number and thickness of the layers. Then, LA-ICP-MS element analysis was carried out moving the laser beam from the outer layer to the inner one by means of a motorised stage. Thus, the distribution of elements could be precisely linked to the single layers of the stratigraphy. Fourier Transform Infrared (FTIR) analyses were performed to determine the presence of organic compounds.

2.1

2.3 Laser Ablation Mass Spectrometry equipped with Plasma source (LA-ICP-MS)

Scanning Electron Microscope

SEM investigations were carried out by a JEOL 5910LV microscopy equipped with an X-ray spectrometer IXRF System/EDS 2000. Observations were carried out on polished cross sections.

2.4 Fourier Transform Infrared Spectroscopy FTIR spectra were recorded in transmission through a diamond cell by an FTIR spectrophotometer Nicolet Nexus, equipped with microscope Continuµm (400 to 4000 cm−1 , 4 cm−1 resolution). In order to avoid contamination by the substrate, samples were carefully collected under a stereomicroscope by means of a needle-sampler.

3

RESULTS AND DISCUSSION

The analyses distinguished three typologies of stucco decorations: simple and complex stratigraphies and gilding finishing layers. As far as the raw materials are concerned, the stucco and the finishing layers were mainly made of magnesiac lime mortar and only in a few cases of gypsum. For centuries, most of the conservation works affected most of the stuccoes and as a consequence, an overlapping of many white-finishing layers was observed in the samples. Finally, in some of the fragments gilding traces were detected among the white layers or on the top of the decorations. In case of simple stratigraphies, only one or two layers were observed onto the bulk. As an example, the samples coming from the church of Santa Maria del Restello showed the presence of a single layer on the gypsum bulk, constituted of magnesiac lime lacking in aggregate and many round clots of lime in the external layer (Fig. 1). The backscattered observation showed the different morphology of the external portion due to the re-crystallisation of rich in magnesium components (Fig. 2); EDS analysis detected only calcium and magnesium (Fig. 3). On the contrary LA-ICP-MS analyses pointed out also the presence of traces of lead, as showed in Figure 4.

110

Figure 1. Optical microscopy image of polished cross section: only one white layer is perfectly adhered on the gypsum bulk (bar = 100 µm).

Figure 3. EDS spectrum of the external white layer.

Figure 4. LA-ICP-MS spectrum showing the presence of lead in the external surface.

Figure 2. Backscattered image of polished cross section: in the upper part of the external layer rich in Mg crystals are visible (bar = 100 µm).

The presence of lead can hint the use of metal working tool. First, lead is abundant only in the external portion of white layers. Second, lead was detected only in decorative elements, whose repeated shape could be achieved only with the use of moulds. As regards the gilding decorations, optical microscope investigations determined the presence of a gold leaf on a yellow preparation layer (Fig. 5), and SEM morphological observation pointed out the presence of a thin homogeneous and continuous layer between them (Fig. 6). The signal of lead detected by EDS analysis should be ascribed only to the preparation layer composition, i.e. chrome yellow (PbCrO4 , Fig. 7). The FTIR analysis carried out on this layer did not show the presence of lead white, easy detectable through the characteristic peaks at 1402, 836 and 683 cm−1 , while XRD analyses showed the presence of crocoite (PbCrO4 ). Because of the absence of lead sulphate, as showed

Figure 5. Optical microscopy image of polished cross section: gold leaf applied on a yellow preparation layer (bar = 100 µm).

by the X-ray diffraction (XRD) analysis, it is possible to suppose that the pigment corresponds to the middle (PbCrO4 ) or (PbCrO4 PbO) orange shade of a chrome compound according to the classification

111

Figure 8. LA-ICP-MS spectrum showing the distribution of lead between gold leaf and mordant layer.

Figure 6. Backscattered image of polished cross section: a thin homogeneous and continuous layer is well distinguishable under a gold leaf.

Figure 9. FTIR spectrum showing the presence of calcite, gypsum and oil.

Figure 7. EDS spectrum acquired from the yellow layer.

of lead chromate pigments of Schiek (Kuhn & Curran 1986). The LA-ICP-MS analyses showed a wide abundance of lead in the yellow layer and between this one and the gold leaf (Fig. 8). The presence of lead just under the gold leaf is probably due to the employ of lead oxide as a dryer for lipidic adhesive (Figs. 6, 8). Literature corroborates this hypothesis, dealing with the use of lipidic compounds for the application of the gold leaf with the employ of a mordant (Gettens & Stout 1966). FTIR analyses of other gold finishing layers analysed in this work, revealed the presence of absorbance peaks at 1739, 2856, 2927 cm−1 , which can be ascribed to lipidic substances (Fig. 9). So, at the light of FTIR and SEM analyses, the presence of lead pointed out in the above discussed LA-ICP-MS results may indirectly suggest the presence of a lipidic adhesive. Coming to the study of stratigraphies characterized by many white layers (Fig. 10), SEM-EDS distinguished well the morphological differences among the layers and the different calcium to magnesium ratios (Figs. 11–13).

Figure 10. Optical microscopy image of polished cross section: three layers are applied onto the bulk (bar = 200 µm).

Once again LA-ICP-MS detected some trace elements, which discriminated the layers and suggested their application had been carried out at successive times.

112

Figure 11. Backscattered image of polished cross section: the different morphological structure of layers is well distinguishable (bar = 100 µm).

Figure 13. EDS map of magnesium distribution in the area showed in Figure 11.

Figure 12. EDS map of calcium distribution in the area showed in Figure 11.

In particular, iron and lead distribution stressed the peculiarity of the external layer (Fig. 14). Iron and lead distribution are different: iron is randomly scattered in the whole layer while lead is mainly located close to the surface. Iron has to be considered as an impurity (such as iron oxide, hematite, hydroxide and limonite) of the raw material, carbonated stones, employed for the preparation of the lime. Differently, the presence of lead in the external portion of white finishing layer should be correlated to metal working tools, such as broad knives, trowels and scrapers employed by the artists to mould the lime. 4

CONCLUSIONS

The possibility of using SEM-EDS and LA-ICPMS integrated methodology for the investigation of stucco finishing layers emerged, in particular during

Figure 14. LA-ICP-MS spectrum showing the distribution of iron (top) and lead (bottom) in the stratigraphy.

the compositional, chronological and technological characterisation. The sensitive LA-ICP-MS analysis clarified the element composition in the stratigraphy previously defined by SEM-EDS.

113

Here, we list the main critical points of the methodology: – a deep stratigraphic investigation is required; – LA-ICP-MS investigations have to be focused on well-defined areas, as layer’s thickness is very variable. The marking of cross-sections is under investigation in order to overcome these problems. REFERENCES Dussubieux, L. & Van Zelst, L. 2004. LA-ICP-MS analysis of platinum-group elements and other elements of interest in ancient gold. Applied Physics A 79: 353–356. Gettens, J. & Stout, G. L. 1966. Painting materials: 33 and 132. Kuhn, H. & Curran, M. 1986. Chrome yellow and other chromate pigments. Artist’s pigments, 1: 187–217.

Manson, P. R. D. & Mank, A. J. G. 2001. Depth-resolved analysis in multi-layered glass and metal materials using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). Journal of Analytical Atomic Spectrometry 16: 1381–1388. Robertson, J. D., Neff, H. & Higgins, B. 2002. Microanalysis of ceramics with PIXE and LA-ICP-MS. Nuclear Instruments and Methods in Physics Research B 189: 378–381. Sanborn, M. & Telmer, K. 2003. The spatial resolution of LA-ICP-MS line scans across heterogeneous materials such as fish otoliths and zoned minerals. Journal of Analytical Atomic Spectrometry 18: 1231–1237. Toniolo, L., Colombo, C., Bruni, S., Fermo, P., Casoli, A & Palla, G. 1998. Gilded stuccoes of the Italian Baroque, Studies in Conservation 43: 201–208. Wagner, B. & Bulska, E. 2004. On the use of laser ablation inductively coupled plasma mass spectrometry for the investigation of the written heritage. Journal of Analytical Atomic Spectrometry 19: 1325–1329.

114

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Compositional depth profiles of gilded wood polychromes by means of LIBS A.J. López, A. Ramil, M.P. Mateo, C. Álvarez & A. Yáñez Departamento de Enxeñaría Industrial II, CIT, Universidade da Coruña, Ferrol (A Coruña), Spain

ABSTRACT: Gilded wood polychromes are important artistic expressions in Galicia (NW Spain) during the baroque period. The variety of techniques applied in gilded ornamentation and the possible existence of products of previous interventions make difficult the restoration process of these artworks which frequently include the removal of a brass-based paint (purpurin) added to cover the lack of gold leaf. The aim of this work is the characterization by means of LIBS of the different layers in gilded wood polychromes. For this purpose small pieces taken from two baroque altarpieces were analyzed using a Nd:YAG laser source operating at the wavelength of 355 nm to obtain characteristic LIBS spectra and compositional depth profiles.

1

INTRODUCTION

During the baroque period, polychromes on wood were the most important forms of sculpture in Spain. One of the most genuine Spanish baroque creation was the “retablo” which is a large, architectural panel divided into compartments which contains religious statues. These artistic workpieces were frequently polychromed and decorated with gold. In Galicia (NW of Spain) the baroque style appeared in the first half of the 17th century and it lasted during the whole 18th century in a great number of “retablos” located in rural churches and chapels. In polychromes the wood substrate is usually covered with different layers; ground, paint layers, gold leaf, varnish and so on. Characterization of the different layers becomes of great importance before any restoration work can be attempted. Several studies have shown the capability of laserinduced breakdown spectroscopy (LIBS) to identify the elemental composition of pigments and to characterize the different layers in painted artworks (Anglos et al. 1997, Burgio et al. 2000, Castillejo et al. 2000, Castillejo et al. 2001, Melessanaki et al. 2001, Clark 2002, Oujja et al. 2005b, Kaminska et al. 2006). LIBS technique is based on the spectral analysis of the emission from the plasma produced during laser ablation. In this sense, when a pulsed laser beam is focused onto the sample surface, it induces not only the ejection and vaporization of material from the sample surface, but also the formation of a plasma plume which emits light at wavelengths characteristic of the elemental composition of the removed layer.

Therefore, the analysis of the emission spectra can provide detailed information about surface composition at each pulse, that is, the in-depth compositional profile of the sample. On the other hand, in the last years, laser cleaning has been increasingly applied to artworks. The process must be controlled to avoid the damage of the substrate (gold leaf and paint layers in our case). In this context, LIBS appears as an adequate diagnostic tool (Gobernado-Mitre et al. 1997, Maravelaki et al. 1997, Tornari et al. 2000, Castillejo et al. 2002, Acquaviva et al. 2004, Colao et al. 2004, Oujja et al. 2005a, Acquaviva et al. 2005). The aim of this work is to test the capability of LIBS for the characterization of the different layers in gilded wood polychromes and to obtain compositional depth profiles which can help, not only in the knowledge of the structure and composition of these materials, but also in a controlled laser cleaning of these artworks. For these purposes LIBS has been applied to the identification of the different layers in real samples of gilded wood polychromes taken from two baroque altarpieces: Antiguo Retablo de Santa María la Mayor (17th century), nowadays at the Museo de Pontevedra, and Retablo de la Capilla del Pazo de San José de Vistalegre (18th century) Tui, Pontevedra; both involved in restoration processes. 2

EXPERIMENTAL

The results reported here were obtained with a Q-switched Nd:YAG laser source (Quantel, model

115

Brilliant B) operating at the third harmonic, 355 nm. The samples were irradiated in air at room temperature and pressure. The laser beam was focused onto the sample surface by a plano-convex quartz lens with a focal length of 300 nm situated on a ruled rail that allowed to change the distance between the lens and the sample in a controlled way. X–Y translation stages and an alignment system consisting of two He-Ne lasers was used to help in sample positioning in terms of laser focal point situation and spectral collection. The emission of the plasma was collected and guided to the spectrograph (Oriel, model MS257) with a fiber optic. Light was dispersed by using the 600 grooves/mm grating of the spectro-graph. An intensified coupled charge device, ICCD (Andor, model DH5H7-18F03) detected the spectral resolved emission from the plasma. The depths reached by the laser ranged from 40 to 60 µm, depending on the sample, with a typical etch per pulse of 5 µm and a crater area of around 0.2 mm2 . This procedure allows to distinguish by LIBS the different layers which constitute the polychrome without causing an important damage to the piece. In addition to LIBS a stratigraphic analysis with the optical microscope was performed over a minuscule specimen taken with scalpel from the altarpiece and embedded in epoxy resin. 3

RESULTS AND DISCUSSION

As it has been previously pointed out, the carved wooden substrate is covered with a series of layers in polychromes; the number and characteristics of each one depend on the artistic technique used. In the case of gilded work, there are many different techniques (González-Alonso 1997); in brief, the carved wooden substrate is usually covered with a white ground layer or gesso (generally a mixture of hide glue and chalk), a layer of red bole (a mixture of clay and glue), gold leaf and in some cases a paint layer above. The quality and characteristics of the gold leaf can be different; the gold usually is alloyed with Ag and Cu and even Pt or other metals. 3.1 Antiguo Retablo de Santa María la Mayor The ancient altarpiece of Santa María la Mayor (Pontevedra) consists of five wooden panels showing important episodes of the life of The Virgin. They are bas-reliefs carved by Jácome de Prado (1623– 1626) and stylistically could be classified as an early and popular baroque piece (Filgueira-Valverde 1991). In Figure 1 one of the panels corresponding to the “Natividad de La Virgen” is shown. The vestments and other parts of the panel were ornamented using the method of sgraffito on gold which consists in covering

Figure 1. Natividad de la Virgen (153 × 76.5 cm2 ). One of the five panels which constitute the altarpiece Antiguo Retablo de Santa María la Mayor (17th century), nowadays at the Museo de Pontevedra (Spain). The square in the upper right side indicates the white zone where LIBS analysis were performed.

the gold leaf with a paint layer which is incised or scratched through, to reveal the gold underneath. Figure 2 depicts the optical micrography of a stratigraphic section (cross section) of a white-colored area of the polychrome. Different layers can be observed: on the top a white pigment layer (≈30µm depth), underneath a thin layer of gold (2–3 µm) and finally the layer of red bole (25–30 µm), a mixture of red clay and glue. LIBS depth profiles were carried out by subsequent ablation of the sample surface at the same irradiated spot. Figure 3 shows LIBS spectra obtained in

116

Figure 2. Embedded cross section of a small fragment removed from the altarpiece Antiguo Retablo de Santa María la Mayor. On the top, a layer of white pigment followed by a thin layer of gold leaf and a layer of red bole underneath, can be appreciated.

Figure 4. Variation of intensity of () Au (I) 312.278, (
) Pb (I) 282.319, and (∗ ) Fe (I) 274.356 lines as a function of the number of laser shot; i.e. sample depth in Antiguo Retablo de Santa María la Mayor. Lead is the characteristic element of the external layer (lead white), gold indicates the intermediate layer and finally iron content is characteristic of the red bole.

The in-depth variation of the intensity of all the elements which appeared in successive LIBS spectra has been obtained after normalization of signals to take into account the decrease of the plasma signal with depth (López et al. 2005). Figure 4 shows the variation of intensity as a function of the number of laser shots, i.e. sample depth, for the peaks Pb (I) 282.319, Au (I) 312.278 and Fe (I) 274.356, characteristics elements of each layer (lead white pigment, gold leaf and bole). Lead content decreases in the first four laser pulses whereas gold content increases, taking its higher value around the third pulse. The iron content, characteristic of the red bole, remains practically constant after the fourth pulse. 3.2 Retablo del Pazo de San José de Vistalegre

Figure 3. LIBS spectra corresponding to: a) 1st pulse and b) 3rd pulse of a series of ten laser shots delivered at the same point of the sample taken from the Antiguo Retablo de Santa María la Mayor.

the 1st and 3rd pulses of a series of ten laser shots. The first two laser pulses produced a spectrum in which emission lines due to lead can be distinguished (Fig. 3a) confirming the use of lead white pigment (2PbCO3 · Pb(OH)2 ) in the external layer of the polychrome. The next two pulses result in an increase of the relative intensities of peaks attributed to Au, Cu and Ag (Fig. 3b) characteristics of a gold alloy.

The altarpiece of the chapel in Pazo de San José de Vistalegre (Tui-Pontevedra) is shown in Figure 5. It was built in 18th century by an unknown author and represents an example of baroque altarpieces located in rural churches or chapels in Galicia (NW Spain). It was made of gilded and polychromed wood. The altarpiece presents a central niche with the image of San José surrounded by six small statues of other saints. The process of restoration of this altarpiece included the removal of a brass-based paint (purpurin) probably added in previous interventions to cover the lack of gold leaf. Purpurin turned opaque due to the oxidation process of its constituents; furthermore, oxidation products increased the adhesion to the substrate making difficult the mechanical or chemical elimination of such substance. This problem is quite frequent for restorers because purpurin was used extensively throughout the last century to cover loss of gold leaf. In this case, the capability of LIBS as a control tool

117

Figure 5. The altarpiece of the chapel in Pazo de San José de Vistalegre (Tui-Pontevedra), 18th century. Gilded and polychromed wood. The square in the bottom left side indicates the zone studied. Figure 7. LIBS spectra of a) gold leaf and b) purpurin layers in the spectral window from 260 to 340 nm. Gold used was alloyed withAg and Cu and purpurin is a brass based material.

Figure 6. Two optical microscopy images of cross sections obtained at different points of the Retablo del Pazo de San José de Vistalegre. In (a) no purpurin was added. In (b) a thick pile of purpurin can be appreciated.

in a laser cleaning process focused on the elimination of purpurin in gilded polychromes was tested. Specifically, LIBS was used to distinguish between original materials (gold leaf) and purpurin added in previous interventions. Figure 6 shows the cross sections obtained by means of optical microscopy of specimens taken from the altarpiece. Figure 6a corresponds to a sample free of purpurin showing a top layer of gold leaf ≈3 µm, a layer of bole ≈ 20 µm and the layer of white ground > 50 µm. Conversely, Figure 6b depicts a sample where purpurin was added. A thick and mat pile of purpurin over the layer of bole can be appreciated. LIBS spectra, in the range 260 nm to 340 nm, of gold leaf and purpurin layers, respectively, are shown in Figure 7. As depicted in the plots, gold was alloyed with silver and copper and purpurin consists basically of copper and zinc. Where purpurin was added (Fig. 8b), LIBS depth profiles show that copper (peak Cu (I) 327.396) characteristic of the purpurin and iron (peak Fe (I) 274.356) are present for the 8 first pulses, which, in addition to stratigraphy in Figure 6b, indicates that the purpurin is probably mixed with the bole. The intensity of Ca (II) 315.887 increases in the last two pulses indicating

118

ACKNOWLEDGMENTS The authors are grateful to Museo de Pontevedra, specially Carlos Valle, director, and Sonia Briones, conservator; and Galicia Proarte S.L., for providing the samples to be analyzed. This work was partially supported by Xunta de Galicia, Project PGIDIT06CCP00901CT.

REFERENCES

Figure 8. LIBS depth profiles obtained in areas of the altarpiece (a) free of purpurin and (b) with purpurin added. The intensity of the peaks Au (I) 274.356, Cu (I) 327.396, Fe (I) 274.356 and Ca (II) 315.887, which characterize the different layers in polycromed samples from the altarpiece of the chapel in Pazo de San José de Vistalegre, are shown as a function of the number of laser shots.

that the ground layer has been reached. These results confirm the structure of the polychromes obtained by means of optical microscopy and demonstrate the capability of LIBS, not only to give information about the structure and composition of the different layers in polychromes, but also to monitor the laser cleaning of these artistic materials; specifically, the elimination of purpurin added to cover the losses of gold leaf.

4

CONCLUSIONS

Laser Induced Breakdown Spectroscopy, LIBS, has allowed to perform the elemental characterization of the different layers in gilded wood polychromes and to obtain compositional in-depth profiles of real samples from two baroque altarpieces. Specifically LIBS has allowed to characterize the different layers in a polychrome decorated with the technique of sgraffito on gold and to distinguish added layers of purpurin, a brass based pigment, extensively used during the last century to cover the losses of gold leaf. For these reasons LIBS appears as an adequate technique not only for the knowledge of the structure and composition of gilded wood polychromes, but also as a control tool in the laser cleaning of these materials.

Acquaviva, S., De Giorgi, M. L., Marini, C. & Poso, R. 2004. Elemental analyses by Laser Induced Breakdown Spectroscopy as restoration test on a piece of ordnance. Journal of Cultural Heritage 5: 365–369. Acquaviva, S., De Giorgi, M. L., Marini, C. & Poso, R. 2005. A support of restoration intervention of the bust of St. Gregory the Armenian: Compositional investigations by Laser Induced Breakdown Spectroscopy. Applied Surface Science 248: 218–223. Anglos, D., Couris, S. & Fotakis, C. 1997. Laser diagnostics of painted artworks: Laser-Induced Breakdown Spectroscopy in pigment identification. Applied Spectroscopy 51: 1025–1030. Burgio, L., Clark, R. J. H., Stratoudaki, T., Doulgeridis, M. & Anglos, D. 2000. Pigment identification in painted artworks: A dual analytical approach employing Laser-Induced Breakdown Spectroscopy and Raman microscopy. Applied Spectroscopy 54: 463–469. Castillejo, M., Martin, M., Oujja, M., Silva, D., Torres, R., Domingo, C., Garcia-Ramos, J. V. & Sanchez-Cortes, S. 2001. Spectroscopic analysis of pigments and binding media of polychromes by the combination of optical laserbased and vibrational techniques. Applied Spectroscopy 55: 992–998. Castillejo, M., Martin, M., Oujja, M., Silva, D., Torres, R., Manousaki, A., Zafiropulos, V., van den Brink, O. F., Heeren, R. M. A., Teule, R., Silva, A. & Gouveia, H. 2002. Analytical study of the chemical and physical changes induced by KrF laser cleaning of tempera paints. Analytical Chemistry 74: 4662–4671. Castillejo, M., Martin, M., Silva, D., Stratoudaki, T., Anglos, D., Burgio, L. & Clark, R. J. H. 2000. Analysis of pigments in polychromes by use of Laser Induced Breakdown Spectroscopy and Raman microscopy. Journal of Molecular Structure 550: 191–198. Clark, R. J. H. 2002. Pigment identification by spectroscopic means: an arts. Comptes Rendus Chimie 5: 7–20. Colao, F., Fantoni, R., Lazic, V., Caneve, L., Giardini, A. & Spizzichino, V. 2004. LIBS as a diagnostic tool during the laser cleaning of copper based alloys: experimental results. Journal of Analytical Atomic Spectrometry 19: 502–504. Filgueira-Valverde, J. 1991. La basílica de Santa María de Pontevedra. A Coruña: Fundación Pedro Barrié de la Maza. Gobernado-Mitre, I., Prieto, A. C., Zafiropulos, V., Spetsidou, Y. & Fotakis, C. 1997. On-line monitoring of laser cleaning of limestone by Laser-Induced Breakdown Spectroscopy and Laser-Induced Fluorescence. Applied Spectroscopy 51: 1125–1129.

119

González-Alonso, E. 1997. Tratado del dorado, plateado y su policromía. Tecnología, conservación y restauración. Valencia: Servicio de Publicaciones de la Universidad Politécnica de Valencia. Kaminska, A., Sawczak, M., Oujja, M., Domingo, C., Castillejo, M. & Sliwinski, G. 2006. Pigment identification of a XIVwooden crucifix. Journal of Raman Spectroscopy 37: 1125–1130. López, A. J., Nicolás, G., Mateo, M. P., Piñon, V., Tobar, M. J. & Ramil, A. 2005. Compositional analysis of Hispanic Terra Sigillata by Laser Induced Breakdown Spectroscopy. Spectrochimica Acta Part B-Atomic Spectroscopy 60: 1149–1154. Maravelaki, P. V., Zafiropulos, V., Kilikoglou, V., Kalaitzaki, M. & Fotakis, C. 1997. Laser Induced Breakdown Spectroscopy as a diagnostic technique for the laser cleaning of marble. Spectrochimica Acta Part B-Atomic Spectroscopy 52: 41–53. Melessanaki, K., Papadakis, V., Balas, C. & Anglos, D. 2001. Laser Induced Breakdown Spectroscopy and

hyper-spectral imaging analysis of pigments on an illuminated manuscript. Spectrochimica Acta Part B-Atomic Spectroscopy 56: 2337–2346. Oujja, M., Rebollar, E., Castillejo, M., Domingo, C., Cirujano, C. & Guerra-Librero, F. 2005a. Laser cleaning of terracotta decorations of the portal of Palos of the Cathedral of Seville. Journal of Cultural Heritage 6: 321–327. Oujja, M., Vila, A., Rebollar, E., Garcia, J. F. & Castillejo, M. 2005b. Identification of inks and structural characterization of contemporary artistic prints by Laser Induced Breakdown Spectroscopy. Spectrochimica Acta Part B-Atomic Spectroscopy 60: 1140–1148. Tornari, V., Zafiropulos, V., Bonarou, A., Vainos, N. A. & Fotakis, C. 2000. Modern technology in artwork conservation: a laser-based approach for process control and evaluation. Optics and Lasers in Engineering 34: 309–326.

120

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Classification of archaeological ceramics by means of Laser Induced Breakdown Spectroscopy (LIBS) and Artificial Neural Networks A. Ramil, A.J. López, M.P. Mateo & A. Yáñez Departamento de Enxeñaría Industrial II, CIT, Universidade da Coruña, Ferrol (A Coruña), Spain

ABSTRACT: The aim of this work is to analyze the feasibility of Artificial Neural Networks (ANN) for the classification, in function of the provenance, of archaeological ceramics Terra Sigillata analyzed by means of Laser Induced Breakdown Spectroscopy (LIBS). An ANN algorithm which is fed with the whole LIBS spectra of the ceramic samples was proposed and an analysis of the network architecture as a function of the number of hidden neurons and number of epochs of training was carried out in order to optimize the performance of the network. Finally, the correct classification of Terra Sigillata pieces from their LIBS spectra has been achieved in a systematic and objective way.

1

INTRODUCTION

This work deals with a provenance study of archaeological ceramics Terra Sigillata, a Roman fineware characterized by a red sintered slip. The production of Terra Sigillata began in Central Italy in the 1st century BC and from there it spread over many areas of the Roman Empire. Due to extensive trading in the ancient world, findings of Terra Sigillata from an excavated site may include products from different workshops and periods which frequently differ in technological features. The reconstruction of trade is important for archaeologists and the ability to determine the source of ancient ceramics becomes of great interest (Maggetti 2001, López-Pérez 2004). Different analytical techniques have been used for provenance studies of archaeological materials; among them, Laser-Induced Breakdown Spectroscopy (LIBS) provides some significant advantages: This technique is almost non-destructive and the study can be performed over the whole piece in air at room temperature without preparation or fragmentation requirements (Fotakis et al. 2007). In a previous work (López et al. 2006) we have shown that it is possible to obtain a reliable classification of Terra Sigillata shreds as a function of the provenance by considering the whole LIBS spectrum as a fingerprint of the sample and using linear correlation techniques for quantifying the degree of similarity among them. In this work, however, a different approach based in artificial neural networks is tried.

Artificial neural network (ANN) algorithms are now being used in a wide variety of data processing applications giving solutions to classification and prediction problems with high degree of accuracy. An advantage of this technique is the fast pattern recognition with an appropriate training (Bishop 1994, Peterson 2000). There is a number of papers in the field of archaeometry concerning the use of ANN (Remola et al. 1996, Bell & Croson 1998, López-Molinero et al. 2000, Maa et al. 2000, Fermo et al. 2004). However, despite of the fact that artificial neural networks have found great impact in spectral analysis, up to now there is little work concerning the application of ANN in LIBS spectroscopy (Sattmann et al. 1998, Samek et al. 2001, Inakollu 2003, Sirven et al. 2006). This paper is focused on the use of neural networks for the classification of Terra Sigillata samples as a function of the provenance. The advantage of ANN over traditional computing techniques like linear regression is the adaptability to unknown situations (generalization), the ability of learning, ease of use and strong tolerance to errors or noise (Haykin 1999, Ham & Kostanic 2001). The proposed ANN algorithm processes the whole LIBS spectrum of the analyzed samples. The optimum configuration of the neural network is analyzed and the performance for the correct assignation of provenances is evaluated and compared with results obtained from linear correlation techniques.

121

2 2.1

EXPERIMENTAL Experimental setup

A standard LIBS configuration was used in this study consisting of a Q-switched Nd:YAG laser (Quantel, model Brilliant B, pulse length (FWHM) 6 ns) operating at the third harmonic, 355 nm. The laser beam was focused onto the sample surface at normal incidence with a quartz plano-convex lens (300 mm focal length) leading to an analyzed area of 0.5 mm2 . The emission of the plasma was collected and guided to the spectrograph (Oriel, model MS257) with an optic fibre. Light was dispersed by using the 600 grooves/mm grating of the spectrograph. An ICCD (Andor, model DH5H718F-03) which consisted of 512 × 512 pixels detected the spectral resolved emission from the plasma.

Figure 1. Characteristic LIBS spectra of Terra Sigillata samples from Hispanic, Gaulic and African workshops, in the spectral window between 260 and 340 nm.

2.2 Samples

subtle variations in the ratios of line intensities which are not obvious in a visual examination.

A total of 36 pieces of Terra Sigillata from different provenances (Hispanic, Gaulic and African workshops) and different periods (Higher Roman Empire and Lower Roman Empire) were analyzed by LIBS. On the basis of archaeological information, they were classified as follows: – 24 pottery samples (from S1 to S24) were assigned to Hispanic workshops in Tricio (La Rioja) and Andújar (Andalucía). Most of them dated from 1st – 2nd century AC. However, only four samples correspond to the Lower Roman Empire (4th–5th century AC). – 6 pieces (from S25 to S30) were attributed to the Gaulic center of La Graufesanque (1st – 2nd century AC). – 6 samples correspond to the North-African workshops (from S31 to S36). 2.3 LIBS data Each ceramic shred was analyzed at different random points of a fresh fracture (10 positions in the case of samples used for training the network, and 3 for the rest) and ten laser shots were recorded over the same sample position. In all the cases, the emission signal reached higher values of intensity at the first shot than in the next nine, in which it remained stable. Due to this the first LIBS spectrum was always disregarded for the analysis. The spectral window between 260 and 340 nm was selected as the most adequate for the analysis of these materials (López et al. 2005). Characteristic LIBS spectra of samples from the Hispanic, Gaulic and African workshops are shown in Figure 1. All the samples analyzed present similar elemental composition in terms of major constituents (López et al. 2006). Consequently, as depicted in the plots, the differences between their LIBS spectra consist of

3

DATA PROCESSING

3.1 Designing the networks ANN are computational algorithms appropriate for complex classification and pattern recognition problems. Inspired by human mind, the neural networks are implemented using simple nonlinear or linear processing elements named neurons, which may interconnect with other ones by means of transfer functions, forming complex processing networks. These networks may be trained adjusting the interconnection branch loads (synapse weights) through training algorithms, being the back-propagation one of the most common. In this algorithm, a number of example data whose outputs are known are used as input data. The calculated output is then compared with the known output (target). The difference between the two outputs is then back-propagated to recalculate the weights. The mean square error mse given by Equation 1, where N is the number of outputs used to quantify such difference. Such an iterative procedure is continued until the difference becomes small enough (Haykin 1999).

After training, the ANN can be used to perform certain tasks depending on the particular application. Different types of neural networks are described in the literature; the most commonly used is the multilayer feed-forward ANN. This kind of network is composed of one input layer that receives the vector to be classified, one output layer that presents the classification results and a set of intermediate layers named hidden layers. The number of intermediate layers and the

122

Table 1. Network desired outputs targets in neurons n1 , n2 and n3 of the output layer for Hispanic (H), Gaulic (G) and African provenances.

Hispanic Gaulic African

n1

n2

n3

Neurons

Epochs

mseT

mseO

1 0 0

0 1 0

0 0 1

0-3 10-3 20-3 25-3 30-3

5000 4479 1399 1797 990

0.228782 0.005000 0.004997 0.004997 0.004983

0.154623 0.161673 0.071905 0.031622 0.095472

number of neurons in each intermediate layer depend on the complexity of the problem to be solved. One of the goals of the classification by means of ANN is the selection of a low complexity neural network having a good ability for generalization. Given that, in our case the aim is the classification of Terra Sigillata samples in three groups, Hispanic, Gaulic and African; a network with three neurons in the output layer, n1 , n2 and n3 , seems to be the most appropriate. In order to have a good discrimination among the group outputs and an easy classification rule, the logsigmoid function, which gives values between 0 and 1, was chosen as transfer function. The network desired outputs (targets) were selected according to Table 1. The input vector consisted of 512 intensity values of the normalized LIBS spectra in the spectral window between 260 and 340 nm. The set of LIBS spectra were divided into two groups; one for the calibration or training of the network and the other one for validation purposes. The training set T consisted of samples S5 (Hispanic), S28 (Gaulic) and S32 (African) which were selected because they presented the lower dispersion in five LIBS spectra (López et al. 2006). The networks were trained by means of a gradient decrease with momentum and adaptive learning rate algorithm (Haykin 1999, Cirovic 1997). The complete procedure was designed and performed using the Matlab Neural Network Toolbox (Demuth et al. 1992). 3.2

Table 2. Values of the mean squared error for training (mseT ) and optimization (mseO ) sets obtained as a function of the number of neurons in hidden layer and epochs of training.

Figure 2. Mean squared error surface for the optimization set, mseO , as a function of the number of neurons in hidden layer and training epochs.

Optimizing the networks

In order to optimize the design of the network to obtain a reliable classification of the Terra Sigillata shreds, different number of neurons in hidden layer were tried. For the optimization process one sample from each provenance group was taken from the validation set; this optimization subset O comprises the samples S10, S26 and S33. The performance of ANN as a function of the number of hidden neurons: 0, 10, 20, 25 and 30, was analyzed. The training was programmed to stop if the mse dropped below 0.005 or the number of the cycles of training (epochs), reached 5000. As it can be appreciated in Table 2 a network with no hidden layer reaches 5000 epochs without achieving the desired performance and may be rejected. Clearly, increasing

Figure 3. Mean squared error for training set (mseT ) and optimization set (mseO ) as a function of the number of training epochs.

the number of hidden neurons decreases the training epochs necessary to attain the same value of mse. However a network with a large number of weights could cause overfit. For this reason, mean-squared error of the optimization set, mseO , must be taken into account to select the optimum network design (Haykin 1999). From the data in Table 2, ANN (25:3) i.e. a network with 25 neurons in the hidden layer and 3 neurons in the output layer can be considered as the best choice in terms of mse.

123

The variation of mean squared error, mseO , represented as a function of the number of neurons in hidden layer and training epochs (Hernández-Caraballo & Marcó-Parra 2003) is depicted in Figure 2. mseO reaches a minimum for a number of hidden neurons around 25 (Fig. 3). A decrease in mseO by increasing the number of epochs can be appreciated. However the network could become overtrained and would lose the ability for generalization. To avoid this problem and to improve the network output i.e. the classification results, determination of when the training should stop, or early stopping criterium, is necessary. The procedure was as follows: ANN (25:3) was retrained for 100.000 epochs and the variation of mse for training and optimization sets as a function of the number of epochs is shown in Figure 3. mseT curve decreases monotonically for an increasing number of epochs. In contrast, mseO decreases monotonically to a minimum and then it starts to increase as the training continues. This behavior indicates an overtraining of the network and the minimum point in optimization learning curve must be used as criterion for stopping the training session (Haykin 1999). In the case of ANN (25:3) the minimum occurs at 20.200 epochs (see Fig. 3). 3.3 Classification efficiency Once the network was optimized and the stopping criteria established, the capability of classification was tested. By using as inputs the average spectrum of each sample in the validation set, the output of the network was represented in Figure 4 as a function of the sample number. The output of ANN (25:3) which takes the highest value in neuron n1 for Hispanic samples, in n2 for Gaulic samples, and in n3 for African samples, attains a correct identification of all the samples (100% of success). It can be observed that the outputs of some Hispanic samples are less precise than Gaulic or African samples; especially for S11 and S20 the output of neuron n2 is close to the value of neuron n1 . The ability of the designed network for the classification of Terra Sigillata samples with a single-shot spectrum was also evaluated, and the percentages of correct identification for each sample are shown in Figure 5. Averaging over the 35 Terra Sigillata shreds studied, a mean value of 91% was obtained by ANN (25:3) and 86% by linear correlation method. Moreover, taking into account all the LIBS spectra used for the analysis of the 35 samples, the ANN failed in 9% of the total cases and linear correlation in 18%, which demonstrates that neural networks present higher tolerance than linear correlation to small variations in spectra from the same provenance. Artificial neural networks may appear as a sophisticated method in comparison with linear correlation. However, an important advantage of ANN method lies

Figure 4. Values of the neurons n1 , n2 and n3 in the output layer of the neural network ANN (25:3) for the samples in the validation set. Hispanic group ranges from S1 to S24, Gaulic group from S25 to S30 and African group from S31 to S36.

Figure 5. Percentage of single shot correct identification of samples for both linear correlation (CORR) and artificial neural network (ANN). A mean value of 91% of success is attained in the case of ANN, and 86% in the case of linear correlation.

on the capability of automatization; i.e. once ANN (25:3) has been designed as the suitable network for the classification of Terra Sigillata it can be applied as a black box by non specialized users. Moreover, the possibility of being retrained with new samples to refine or expand the classification to new provenances, enhance the advantages of ANN over other conventional techniques.

4

CONCLUSIONS

The results of the present study are indicating that the use of neural networks to automate the classification of Terra Sigillata in function of the provenance by means of LIBS spectra is feasible and efficient. A feed-forward back-propagation algorithm network which works with the whole LIBS spectrum as input data was selected as the most appropriate for the classification of Terra Sigillata shreds. Simple topological configuration consisting of one hidden layer and one output layer was analyzed as a function of the number of hidden neurons and number of epochs of training. The optimum configuration obtained for the classification of Terra Sigillata in 3

124

provenances: Hipanic, Gaulic and African was ANN (25:3) consisting of 25 neurons in the hidden layer and 3 neurons in the output layer. By means of ANN (25:3) correct classification of Terra Sigillata shreds using the average spectrum of each sample amounted 100%. When a single shot spectrum was used for classification, ANN (25:3) failed in 9% of the total cases and linear correlation in 18% demonstrating that neural networks present higher tolerance than linear correlation to small variations in spectra from the same provenance. To conclude, these results demonstrate that a network as simple as ANN (25:3) allows to classify with high precision Terra Sigillata shreds in function of the provenance by means of their LIBS spectra, achieving even better results than those obtained by linear correlation. ACKNOWLEDGMENTS Special thanks to M.C. López Pérez (archaeologist) and Museo de Prehistoria e Arqueoloxía de Vilalba (Lugo, Spain) for providing Terra Sigillata samples. This work was partially supported by Xunta de Galicia through Project PGIDIT06CCP00901CT. REFERENCES Bell, S. & Croson, C. 1998. Artificial neural networks as a tool for archaeological data analysis. Archaeometry 40: 139–151. Bishop, C. M. 1994. Neural networks and their applications. Review of Scientific Instruments 65: 1803–1832. Cirovic, D. A. 1997. Feed-forward artificial neural networks: Applications to spectroscopy. Trac-Trends in Analytical Chemistry 16: 148–155. Demuth, H., Beale, M. & Hagan, M. 1992. Neural Network Toolbox User’s Guide (2007 ed.). The Mathworks Inc. Fermo, P., Cariati, F., Ballabio, D., Consonni, V. & Gianni, G. B. 2004. Classification of ancient Etruscan ceramics using statistical multivariate analysis of data. Appl. Phys. A 79: 299–307. Fotakis, C., Anglos, D., Zafiropulos, V., Georgiou, S. & Tornari, V. 2007. Lasers in the Preservation of Cultural Heritage. Principles and Applications. Taylor & Francis. Ham, F. M. & Kostanic, I. 2001. Principles of neurocomputing for science and engineering. McGraw Hill. Haykin, S. (1999). Neural Networks: A comprehensive foundation. New Jersey: Prentice Hall. Hernández-Caraballo, E. A. & Marcó-Parra, L. M. 2003. Direct analysis of blood serum by total reflection

X-ray fluorescence spectrometry and application of an artificial neural network approach for cancer diagnosis. Spectrochimica Acta Part B-Atomic Spectroscopy 58: 2205–2213. Inakollu, P. 2003. A study of the effectiveness of neural networks for elemental concentration from LIBS spectra. Master’s thesis, Faculty of Mississippi State University, Mississippi. López, A. J., Nicolás, G., Mateo, M. P., Piñon, V., Tobar, M. J. & Ramil, A. 2005. Compositional analysis of Hispanic Terra Sigillata by Laser-Induced Breakdown Spectroscopy. Spectrochimica Acta Part B-Atomic Spectroscopy 60: 1149–1154. López, A. J., Nicolás, G., Mateo, M. P., Ramil, A., Piñon, V. & Yáñez, A. 2006. LIPS and linear correlation analysis applied to the classification of Roman pottery Terra Sigillata. Applied Physics A-Materials Science & Processing 83: 695–698. López-Molinero, A., Castro, A., Pino, J., Pérez-Arantegui, J. & Castillo, J.R. 2000. Classification of ancient Roman glazed ceramics using the neural network of selforganizing maps. Fresenius Journal of Analytical Chemistry 367: 586–589. López-Pérez, M. 2004. El comercio de Terra Sigillata en la provincia de A Coruña. Brigantium. Museo Arqueolóxico e Histórico Castelo de San Antón, A Coruña 16. Maa, Q., Yana, A., Hu, Z., Lib, Z. & Fanc, B. 2000. Principal component analysis and artificial neural networks applied to the classification of Chinese pottery of Neolithic age. Anal. Chim. Acta 406: 247–256. Maggetti, M. 2001. Chemical analyses of ancient ceramics: What for? Chimia 55: 923–930. Peterson, K. L. 2000. Artificial neural networks and their use in chemistry. Reviews in Computational Chemistry 16: 53–140. Remola, J. A., Lozano, J., Ruisánchez, I., Larrechi, M. S., Rius, F. X. & Zupan, J. 1996. New chemometric tools to study the origin of amphorae produced in the Roman Empire.Trac-Trends inAnalytical Chemistry 15: 137–151. Samek, O., Telle, H. H. & Beddows, D. C. 2001. LaserInduced Breakdown Spectroscopy: a tool for real-time, in vitro and in vivo identification of carious teeth. BMC Oral Health 1. Sattmann, R., Monch, I., Krause, H., Noll, R., Couris, S., Hatziapostolou, A., Mavromanolakis, A., Fotakis, C., Larrauri, E. & Miguel, R. 1998. Laser-Induced Breakdown Spectroscopy for polymer identification. Applied Spectroscopy 52: 456–461. Sirven, J. B., Bousquet, B., Canioni, L., Sarger, L., Tellier, S., Potin-Gautier, M. & Hecho, I. L. 2006. Qualitative and quantitative investigation of chromium-polluted soils by laser Induced Breakdown Spectroscopy combined with neural networks analysis. Analytical and Bioanalytical Chemistry 385: 256–262.

125

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Laser ablation- and LIBS-ranging by webcam and image processing during laser cleaning M. Lentjes, J. Hildenhagen & K. Dickmann Laser Center (LFM), Münster University of Applied Sciences, Steinfurt, Germany

ABSTRACT: Pulsed lasers with manual articulated beam delivery systems are usually used for cleaning of various artworks. The advantage of spatial manoeuvrability can be a disadvantage as well, since the exact irradiation position is unknown. During laser cleaning the coordinates of each laser pulse impinging on the artwork can be obtained by a camera placed above the sample in combination with image processing. This method was applied to obtain the corresponding coordinates of Laser Induced Breakdown Spectroscopy (LIBS) measurements whilst laser cleaning. If the position of the measurements is known, the results can be displayed as a transparent overlay with discreet colour variations on top of the picture of the artwork. The colour value corresponds to the result obtained at that coordinate. In this way, any kind of single value result can be displayed in a picture, e.g. element line intensity, intensity ratio, linear correlation coefficient, etc. We calculated the linear correlation coefficient r between the current measured spectrum and a pre-stored reference spectrum, and displayed the value of r by the corresponding colour in the picture of the sample.

1

INTRODUCTION

Due to the extension of laser cleaning to various conservation fields, the diversity of laser-cleaned artworks is also increasing. Whereas the cleaning process of encrusted marble by Nd:YAG laser is self-limiting (when using the suitable parameters), this is not the case for all polluted artworks. This means that in these cases, on time stopping of the cleaning process is essential to avoid over-cleaning. Without intervention, the ablation process continue through the surface to be preserved. In general, the pollution layers are not uniform. Therefore, the number of laser pulses at a certain position has to be adapted to the degree of pollution in this position to ensure an appropriate cleaning of the object. The ablation process is manually controlled by the restorers in most of the practical laser cleaning projects. They observe the laser cleaning process through a laser safety goggle and manually stop the laser emission as soon as the pollution is removed. However, there are different methods applied to avoid over-cleaning during laser treatment of artworks. Some existing methods are: – spectroscopic analysis of the plasma emission induced during laser cleaning (LIBS) (GobernadoMitre et al. 1997, Klein et al. 1999, Scholten et al. 2000, Teule et al. 2003)

– plasma intensity measurement (Hildenhagen & Dickmann 2003, Lentjes et al. 2005) – spectroscopic analysis of laser induced fluorescence (LIF) (Anglos et al. 1995, Gobernado-Mitre et al. 1997) – measuring the transmitted laser radiation (Chaoui et al. 2003) – acoustic monitoring of the induced snapping sound (Jankowska & Sliwinski 2003, Lee & Watkins 2000) – chromatic monitoring (Lee & Watkins 2000, Lee & Steen 2001). In general, these methods are applied to control/ monitor laser cleaning processing with systems equipped with scanners or automatic translation stages. In this paper, a procedure is presented for the use of laser cleaning monitoring methods in combination with laser systems with articulated beam delivery. In the case of cleaning with manual articulated beam delivery, monitoring the process can be used to support a human decision. The developed method is based on the correlation of LIB spectra with pre-stored reference spectra. Identifying the layers during laser cleaning by means of LIB spectra and correlation analysis is described in more detail in a former paper (Lentjes et al. 2007a).

127

2

MONITORING LASER CLEANING WITH ARTICULATED BEAM DELIVERY

Nd:YAG laser systems used for cleaning of artworks are often equipped with an optical fibre or articulated mirror arm for beam delivery. The advantage of the spatial manoeuvrability can also be a disadvantage, since the exact irradiation point is not known and the beam cannot be manipulated automatically. Therefore, cleaning systems with manual articulated beam delivery are generally not adapted for closed-loop cleaning. However, during laser cleaning the coordinates of each laser pulse on the artwork can be obtained by a camera above the sample in combination with image processing (Figs. 1, 2). In the first attempts this method was applied to obtain the corresponding coordinates of LIB spectra measurements whilst laser cleaning. The position of

the interaction area of each pulse was achieved shortly after the laser pulse and plasma emission by analysing the position of a coaxial He-Ne laser spot on the sample. The He-Ne laser spot was used since the Nd:YAG laser pulse and the plasma emission saturated the CCD of the webcam. Since the positions of the laser pulses are known, the results of the corresponding LIBS measurements can be displayed as a transparent overlay with discreet colour variations on top of the picture of the artwork. The colour value corresponds to the result obtained at that coordinate. In this way, any kind of single value result can be displayed in a picture, e.g. element line intensity, intensity ratio, linear correlation coefficient, etc. In this research, the linear correlation results, calculated by correlating the current measured spectrum with a pre-stored reference spectrum, were displayed by the corresponding colour in a picture of the sample. On-line visualisation of the artwork with a current measurement-overlay on a (computer) screen during laser cleaning can be used to monitor/control the process (Figs. 1, 2). By means of the colour values, the restorer can recognise the parts of the artworks that are sufficiently cleaned to avoid over-cleaning. 3

Figure 1. Photograph of the webcam-LIBS set-up.

EXPERIMENTAL SETUP

The laser system used (by Thales, see Table 1), is equipped with an articulated mirror arm enabling manual spatial manipulation of the laser beam (Fig. 1). The laser system is a flash lamp pumped Q-switched Nd:YAG laser based on oscillator/amplifier principle with the possibility of frequency doubling, tripling and quadrupling. In this research project, only the first harmonic was applied. The hand piece of the articulated arm embodies a telescope consisting of

Figure 2. Schematic of the webcam-LIBS setup used to measure LIB spectra and the interaction area coordinates. The right picture shows the sample with ablation spots while the left picture shows the same picture with the transparent overlay. The sample in the figure is iron covered with a thin rust layer.

128

Table 1. Characteristics of the Nd:YAG cleaning laser. Type Wavelengths

Repetition rate Max. Pulse energy Pulse duration

Thales Saga 220/10 1064 nm (Nd:YAG) 2ω, 532 nm 3ω, 355 nm 4ω, 266 nm 1–10 Hz 1500 mJ @ 1064 nm 10 ns

Figure 3. Cross section of the mirror holder developed for in-process plasma emission measurement.

a positive lens ( f = 102.3 mm), followed by a negative lens ( f = −101.0 mm). The distance between both lenses can be varied in order to adjust the working spot diameter and consequently the energy density on the object surface is adjusted as well. Since the output beam was not parallel but converging, a distance holder was applied to secure a constant energy density on the object surface (Fig. 1). The implemented He-Ne laser was coaxial with the Nd:YAG beam. The HR2000 spectrometer system is a userconfigured miniature fibre optic spectrometer from Ocean Optics. For collecting the plasma radiation into the spectrometer, a 2 m long and diameter 600 µm optical fibre, in combination with a f = 10 mm collimator with diameter 5 mm was used. The groove density (300 grooves/mm) and the entrance aperture of 25 µm result in a spectral range of 200–1100 nm with a resolution of 2 nm. The exposure time of the electronic “shutter” is factory-set on 2 ms. The optical fibre with collimator was implemented in the last mirror holder of the articulated arm to allow in-process plasma emission measurement (onaxis plasma emission collection). Fig. 3 shows a cross section of this especially constructed mirror holder. The camera is a FireWire 1394 Webcam with 640 × 480 VGA resolution. The process was controlled by in LabView 7.1 software. The time between attaining the coordinate, firing the laser and acquiring a LIB spectrum was arranged in the LabView program in combination with a DAQ-card with onboard counters. The standard LabView 7.1 software was extended with the LabView

Figure 4. Flow diagram of the webcam and LIBS based laser cleaning monitoring.

Vision Development Module which enables image processing. The sequence of the software for monitored laser cleaning with manual articulated beam delivery is shown in Figure 4.

4

RESULTS AND SIDE EFFECTS

First trials of this method applied to 3 different samples (black on white paint, polluted marble gravestone and rusty steel) showed the basic workability of this method. It was possible to acquire LIB spectra and the position of the measurements after the webcam and image processing settings were optimized for each sample. The image processing settings that could be changed were detection threshold, sharpness, saturation, brightness and post processing image filter. The results were repeatable at different sample positions under the condition that the LIBS experimental parameters were kept constant. This was achieved by applying a distance holder and constant irradiation angle (Fig. 1). Figure 5 shows the trend of the correlation coefficients obtained with the polluted marble sample at constant experimental parameters (the results of the remaining two samples are not displayed as they featured the same trend). The LIB

129

Figure 5. Variation of the correlation coefficient versus ablation pulse number per spot calculated during first trials with the webcam Nd:YAG cleaning laser monitoring setup (constant experimental parameters).

beam delivery. The standard deviation will increase which results in a spread of the single distributions. This means that the acquired spectra, corresponding to the different layers, have to provide a higher difference in comparison to cleaning with constant experimental parameters. This was the limiting effect during the experiments with the low resolution spectrometer and articulated beam delivery. The differences between the correlation coefficients corresponding to the second and third pulse at spot 1 to 4 in Figure 5 were < 0.005 at constant experimental condition. This was too low for in-process identifying during cleaning without distance holder. If the output beam is parallel, a perpendicular object hand piece distance variation will not change the energy density. With these pre-conditions in-process identifying during cleaning without distance holder with the applied setup should be possible.

5

Figure 6. LIB-spectra acquired at the first and third laser ablation pulse on spot 1 (sample polluted marble). The delay between the laser pulse and recording the plasma emission was 0 µs.

spectra of pulse one and three at spot 1 are shown in Figure 6. Since the resolution of the screen image with overlay is too low for printing it is not shown in this paper. In practice, experimental parameters like lens object distance, spot size and irradiation angle can vary from pulse to pulse when the beam is manually manipulated. This influences the form and intensity of the acquired spectra and therewith the distribution of correlation coefficients of the different layers (Lentjes et al. 2007b). If the correlation coefficients distributions of the different layers barely overlap when using a distance holder and a constant irradiation angle, this method cannot be applied to identify layers when cleaning with complete manual articulated

CONCLUSIONS

In a first trial, the LIBS-correlation method has been used to monitor laser cleaning with a manual articulated mirror arm. Therefore, a webcam was placed above the sample to ascertain the irradiation position. The obtained correlation coefficients were visualized in a current picture of the sample by colour marks representing the progress of the cleaning. The obtained results for the tested samples (black on white paint, polluted marble gravestone and rusty steel) were repeatable, provided that the experimental parameters remain constant. This was achieved by applying a distance holder and constant irradiation angle. It was feasible to acquire LIB-spectra and the position of the measurements after the webcam and image processing settings were optimized per sample. REFERENCES Anglos, D., Couris, S., Mavromanolakis, A., Zergioto, I., Solomidou, M., Liu, W. Q., Papazoglou, T. G., Fotakis, C., Doulgeridis, M. & Fostiridou, A. 1995. Artworks Diagnostics Laser Induced Breakdown Spectroscopy (LIBS) and Laser Induced Fluorescence (LIF) Spectroscopy. In E. König & W. Kautek (eds.), Lasers in the Conservation of Artworks, Restauratorenblätter, Sonderband – LACONA. 1: 113–118. Chaoui, N., Solis, J., Afonso, C. N., Fourrier, T., Muehlberge, T., Schrems, G., Mosbacher, M., Bäuerle, D., Bertsch, M. & Leiderer, P. 2003. A high-sensitivity in situ optical diagnostic technique for laser cleaning of transparent substrates. Applied Physics A 76(5): 767–771. Gobernado-Mitre, I., Prieto, A. C., Zafiropulos, V., Spetsidou, Y. & Fotakis, C. 1997. On-Line Monitoring of Laser Cleaning of Limestone by Laser-Induced Breakdown Spectroscopy and Laser-Induced Fluorescence. Applied Spectroscopy 51(8): 1125–1129.

130

Hildenhagen, J. & Dickmann, K. 2003. Low-cost sensor system for online monitoring during laser cleaning. Journal of Cultural Heritage 4: 343–346. Jankowska, M. & Sliwinski, G. 2003. Acoustic monitoring for the laser cleaning of sandstone. Journal of Cultural Heritage 4: 65–71. Klein, S., Stratoudaki, T., Zafiropulos, V., Hildenhagen, J., Dickmann, K. & Lehmkuhl,T. 1999. Laser-induced breakdown spectroscopy for on-line control of laser cleaning of sandstone and stained glass. Applied Physics A 69: 441–444. Lee, J. M. & Watkins, K. G. 2000. In-process monitoring techniques for laser cleaning. Optics and Lasers in Engineering 34: 429–442. Lee, J. M. & Steen, W. M. 2001. In-Process Surface Monitoring for Laser Cleaning Processes using a Chromatic Modulation Technique. The International Journal of Advanced Manufacturing Technology 17: 281–287. Lentjes, M., Klomp, D. & Dickmann, K. 2005. Sensor Concept for Controlled Laser Cleaning via Photodiode. In K. Dickmann, C. Fotakis & J. F. Asmus (eds.), Laser in the Conservation of Artworks, Springer Proceedings in Physics. 100: 427–433.

Lentjes, M., Dickmann, K. & Meijer, J. 2007a. Influence of process parameters on the distribution of single shot correlation coefficients obtained by correlating LIB-spectra. Applied Physics A 88: 661–666. Lentjes, M., Dickmann, K. & Meijer, J. 2007b. Low Resolution LIBS for Online-Monitoring During Laser Cleaning Based on Correlation with Reference Spectra. In H. Nimmrichter, W. Kautek & M. Schreiner (eds.), Laser in the Conservation of Artworks, Springer Proceedings in Physics. 116: 437. Scholten, J. H.,Teule, J. M., Zafiropulos,V. & Heeren, R. M.A. 2000. Controlled laser cleaning of painted artworks using accurate beam manipulation and on-line LIBS-detection. Journal of Cultural Heritage 1: 215–220. Teule, R., Scholten, H., Brink, O. F. v., Heeren, R. M. A., Zafiropulus, V., Hesterman, R., Castillejo, M., Martin, M., Ullenius, U., Larsson, I., Guerra-Librero, F., Silva, A., Gouveia, H. & Albuquerque, M. B. 2003. Controlled UV laser cleaning of painted artworks: a systematic effect study on egg tempera paint samples. Journal of Cultural Heritage 4: 209–215.

131

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

LIBS analysis of metal artefacts from Sucevita Monastery, Romania M. Oujja & M. Castillejo Instituto de Química Física Rocasolano, CSIC, Madrid, Spain

W. Maracineanu, M. Simileanu & R. Radvan National Institute for Research and Development for Optoelectronics INOE 2000, Bucharest, Romania

V. Zafiropulos Technological Educational Institute of Crete, Sitia, Crete, Greece

D. Ferro Istituto per lo Studio dei Materiali Nanostructurati, CNR, Rome, Italy

ABSTRACT: This work presents the results obtained in June 2006 during a campaign within the frame of an EU-funded 2000 Culture Project. Laser induced breakdown spectroscopy (LIBS) has been used to determine the elemental composition of different parts of two metal objects kept in Sucevita Monastery in Suceava, Romania: a cache-pot from the 18th century and a ligneous cross from the 19th century. The results obtained by LIBS were validated by the energy dispersive spectroscopy/scanning electron microscopy technique (EDS/SEM). The integration of results on compositional microanalyses obtained by both techniques on small areas of the objects reveals interesting aspects of the working process and of ancient restoration phases. The comparison of the LIBS results with the EDS/SEM characterisation demonstrates the advantages of this spectroscopic technique for the in situ practice of metal artefact analysis. Finally, optical microscopy was used to observe the fingerprint of LIBS effect and to assess the microstructure of the studied metal objects.

1

INTRODUCTION

Systematic chemical and structural analysis of ancient artistic objects provide important insight into the techniques available and used for processing materials in the manufacture of an art object. Various analytical techniques have been applied extensively in the study of art objects providing important physical and chemical information about the materials and the structure of objects. Among these techniques, scanning electron microscopy (SEM), X-ray fluorescence (XRF), proton induced X-ray emission (PIXE), Xray diffraction (XRD), energy dispersive spectroscopy (EDS) and Raman microscopy (Clark et al. & Bell et al. 1997, Mantler et al. 2000, Burgio et al. 2001, Mandrino et al. 2004, Elfwing et al. 2005), can be cited. However, most of these laboratory techniques require special sample preparation and handling procedures. In addition, transportation of artworks to specialised laboratories is often subject to strict regulations and requires lengthy procedures. Laser Induced Breakdown Spectroscopy (LIBS) offers several analytical advantages and in some cases can be a potential alternative to the mentioned techniques. LIBS is a

practically non-destructive as well as rapid elemental analysis technique with the critical advantage of being applicable in situ, thereby avoiding sampling and sample preparation. LIBS is currently used to characterise the elemental composition of different materials (Anglos & Rinaldi et al. 2001, Bustamante et al. 2002). Qualitative analyses of the elemental composition of stones (Klein et al. 1999), metals (Melessanaki et al. 2002), inks (Oujja et al. 2005), and paintings (Scholten et al. & Castillejo et al. & Burgio et al. 2000) have been carried out. Recent approaches to quantitative analysis by LIBS have lead to the determination of absolute concentration values for each element of the analysed material (Colao et al. 2002, Kuzuya et al. 2003, Carmona et al. 2005 & 2007, Corsi et al. & Yaroshchyk et al. 2006). Some of us have used LIBS in combination with XRF and SEM/EDX in previous works to determine the qualitative and quantitative composition of glasses and inks (Oujja et al. 2005, Carmona et al. 2005 & 2007). In this work, we present the results of the studies performed on two different historical objects kept in Sucevita Monastery, in Suceava, Romania. Analytical information from LIBS analyses was used in order

133

Figure 1. General view of the ligneous cross broken in two parts (a) and cache-pot (b) kept in Sucevita Monastery, Romania; c), d) and e) localisation of zones analysed by LIBS and where the material has been extracted for EDS analyses.

to characterise the metal objects, and to compare these results with those obtained with the conventional analytical technique EDS/SEM. 2

OBJECTS AND METHODS

The study was concentrated on two metal objects; a ligneous cross and a cache-pot. The ligneous cross, 15 cm high, 10 cm wide and 2 cm thick (Fig. 1a) was kept in Sucevita Monastery but it originates from Balinesti, Suceava County (tagged with inventory n◦ 492, dated 19th century). The cross is composed of three parts carved in wood which represent the cycle of Christ’s life in bas-relief and is encapsulated in a metal box made using a mould and assembling separate parts.The lateral decorations and the base were soldered to the cross. Hand refining traces are indicative of the style of the manufacture. To make the conical base of the metal container the metal was moulded with a particular shape without any further chisel work. Regarding the state of conservation, the ligneous material appears

in good condition presenting only some mechanical fractures. The left beam of the cross seems to be made of different wood and with different structure; furthermore figures appear to be carved using a different less accurate style. The right beam is inverted. The metallic box containing the cross is separated from its base and some of its components are missing. Apart from the slight oxidation strata, most of the damage is probably due to mechanical stress. Some points of interest for further study were identified.These include mechanical deformations and the presence of additive materials.These aspects were considered indicative for the understanding of the history of the object. The second metallic object studied was a cachepot (Fig. 1b, tagged with inventory n◦ 590, dated end of 18th century, Parochia Zvoristea, Suceava county). This object has an oval section with axes of 33 and 22 cm and is 13 cm high. It consists of a metallic strip decorated by stamping with amphorae and floral motives. Four legs are soldered to the bottom part of the body and these are decorated with motives of leaves made by casting. Inside the object, a metallic layer covers the whole internal surface. The object is in a discrete state of conservation; the internal layer of the cache-pot is completely corroded and broken in some parts near the edge. Areas with soldering material are evident, particularly in the zones where the four legs are attached to the main body of the object. The peculiarity of the object, composed of different parts, each of them manufactured in a different way, induced the choice of a series of zones for analyses, aiming at individualising the characteristics of each part of the manufacture. The areas selected for LIBS analyses at the base of the cross, the cross and in the cache-pot are indicated in Figures 1c, 1d and 1e. For each object, samples were taken to be analysed by EDS/SEM in order to provide evidence of the elemental quantitative composition and to obtain qualitative data for comparison with the results obtained by LIBS. The sampling consisted in extracting very small amounts of material by scraping with a quartz dust abrasive card. Sample taking was done with the consent of the project conservators. For LIBS measurements, laser irradiation was carried out with the fundamental harmonic of a Q-switched Nd:YAG laser (Quanta Systems, pulses of 6 ns, repetition rate of 10 Hz, 1064 nm). The samples were irradiated by the focused laser using an f = 100 cm lens allowing to achieve fluences up to 9 J cm−2 . The plume emission was collected with a quartz optical fibre. LIB spectra were recorded at the 300– 700 nm wavelength range with a Mechelle spectrograph (ME5000) coupled to a time gated ICCD camera (iStar, Andor Technologies). The temporal gate was operated at 500 ns time delay (in order to discriminate the atomic emission from the continuum background)

134

qualitative/quantitative determination (0.2 wt resolution-limit) by observing the backscattered electrons to get atomic contrast and to distinguish metal grains from the abrasive card components. For each sample, the material extracted by scrapping was extended to a thin layer. Analyses were done on several positions of the extended material layer.

3

Figure 2. LIB spectra in positions BC1 and BC2 on the base of the cross (Figure 1c).

Figure 3. LIB spectrum of the area C2 of the cross as indicated in Figure 1d.

and width of 3 µs. The spectra were obtained by using commercial software by Andor Technologies for a single pulse. The assignment of lines is based on the information taken from the NIST database (NIST Electronic database, at http://physics.nist.gov). The analyses can be characterised as practically nondestructive because the laser beam is focused on a spot of typical diameter in the range of 400–500 µm. The optical microscopy measurements were carried out using an optical microscope which can provide us images at 5 different magnifications: 3x, 6x, 16x, 32x and 50x, these images are acquired with a CCD camera and transmitted to the PC software. The EDS/SEM measurements were performed using a microanalysis system INCA 300 for

RESULTS AND DISCUSSION

LIB spectra on different zones analysed at the base and on the cross are shown in Figs. 2 and 3 respectively and the main results are summarised in Table 1. The characteristic lines of the most representative elements could be assigned. At the base of the cross, the areas labelled BC1, BC2 and BC3 were analysed (Fig. 1c, Table 1). The LIB spectrum taken in the area BC1 (Figure 2) reveals the presence of emissions attributed to copper (Cu I at 324.75, 327.39, 465.11, 510.55, 515.32, 521.82 and 578.21 nm), zinc (Zn I at 328.23, 330.29, 334.50, 468.01, 472.21 and 481.05 nm), silver (Ag I at 328.07, 338.29, 520.90 and 546.55 nm), nickel (Ni I at 341.47, 345.84, 352.45 and 361.93 nm), calcium (Ca II at 393.36 and 396.84 nm, and Ca I at 422.67 nm) and sodium (Na I at 588.99 and 589.59 nm). The analysis done in BC2 (Fig. 2) and BC3 revealed the same emissions as in BC1, except for the absence of zinc and nickel in BC2 and silver in BC3. Figure 3 shows the spectrum taken in the area C2 of the cross (Fig. 1d); there are emissions characteristic to copper, silver, calcium and sodium. The area C1 gives rise to the same emissions as in C2. LIB spectra collected in different areas of the cachepot (Fig. 1e, Table 1) allow the characterisation of the different materials used in the manufacture of this object. LIBS analyses of the inner flat of the cachepot (area CP-IF1) indicate that it is mainly composed of zinc as evidenced by lines of Zn I. In addition, there are emissions attributed to calcium (lines of Ca II), lead (Pb I at 405.78 nm) and sodium (lines of Na I). The soldering material between the cachepot and the legs (area CP-SM6) gives rise to emissions attributed to lead (PbI at 357.27, 363.95, 367.14, 368.34, 373.99 and 405.78), tin (Sn I at 317.50, 326.23 and 380.10 nm), vanadium (V I at 437.92 nm), calcium (lines of Ca II) and sodium (lines of Na I). The qualitative analysis made in the area CP-L2 (the back part of legs), the area CP-SP3 (soldering material of the decorated strip) and the area CP-SP4 (the metal of the decorated strip) of the cache-pot indicate the presence of emissions attributed to zinc, copper, calcium and sodium. In addition to the emissions observed in these points, the analyses in point CP-BS5 (back side of the cache-pot) indicate the presence of iron (Fe I at 385.99, 407.17and 438.35 nm). Figure 4 shows the LIB spectrum obtained for the area CP-BS5.

135

Table 1. Elements identified by LIBS and EDS/SEM performed in different areas of the metallic studied objects (Figure 1). The elements present in higher concentration, as determined by EDS, are marked in bold. Analysed areas Cross Metal box Lateral decorations Base of the cross Top (soldering material) Back side of the base Central part Cache-pot Inner flat part Metal legs Decorated strip Back side Soldering material legs-main body

LIBS

EDS

C1 C2

Cu, Ag, Ca, Na Cu, Ag, Ca, Na

Cu, Ag, Zn, Ti

BC1 BC2 BC3

Cu, Zn, Ag, Ni, Ca, Na Cu, Ag, Ca, Na Cu, Zn, Ni, Ca, Na

Cu, Zn, Ag, Ni, Si, Ti, Cl, Ca Cu, Ag, Br, Mg, K, Ti, Ca, Na

CP-IF1 CP-L2 CP-SP3 CP-SP4 CP-BS5 CP-SM6

Zn, Pb, Ca, Na Zn, Cu, Ca, Na Zn, Cu, Ca, Na Zn, Cu, Ca, Na Zn, Cu, Fe, Ca, Na Pb, Sn, V, Ca, Na

Zn, Cu, Al, Si, Cr, Fe, Ca, Na Zn, Cu, C, Al, Si, As, Ca, Na

Table 2. EDS/SEM qualitative and quantitative composition given in wt% (±0.20%) in different regions of the material extracted from the area BC1 of the base of the cross. Positions tested Si 1 2 3 4 5 6 7

Figure 4. LIB spectrum of the area CP-BS 5 of the cache-pot as indicated in Figure 1e.

EDS determination on the sample obtained from the top of the base of the cross (Fig. 1c, area BC1) as shown in Table 1 reveals the presence of different elements. Apart from the elements coming from the composition of the scrapping card (Si) and from impurities (Cl and Ca), the other elemental compo nents (Cu, Zn, Ni and Ag) are in correspondence with those detected by LIBS (Fig. 2 and Table 1). The elements O and C, observed by EDS, and not indicated in Table 2 are not interesting for our purposes. It appears that Cu is in constant ratio with Zn, indicative of a brass alloy with an average composition Zn (40%)/Cu(60%), while Ag also observed by LIBS, seems to be present in discrete amounts in some areas. The fabrication techniques of Zn, Cu and Ag alloys were not available at the time the object was dated (19th cc). Other alternative is considered in which a silvered brass lamina was prepared by pressing a fine layer of Ag on the brass substrate by the technique of

Cl

33.88 5.75 33.75 5.10 25.17 7.32 31.36 6.85 29.84 7.76 31.19 6.51 22.39 10.68

Ca

Ti

Ni

Cu

Zn

Ag

4.82 5.75 3.27 5.28 6.55 4.64 4.58

1.99 2.05 1.68 2.21 3.81 2.45 4.05

4.70 5.53 4.30 4.74 6.16 4.15 3.93

23.98 23.05 26.86 23.42 21.37 22.86 24.44

17.99 17.49 20.03 18.83 17.48 18.21 23.89

3.60 3.72 6.33 3.61 2.35 5.91 2.05

lamination. Subsequent polishing by abrasion would remove the silver layer from most parts of the surface. On the other hand the anachronistic presence of Ti/Ni not in a constant ratio suggests the probable use of a material constituted by these two elements during a modern restoration intervention to repair the damage caused by the separation of the cross from its base. This kind of welding is very fragile and not suitable to join two heavy parts. This is the reason why the cross was detached from its base. The elements identified by EDS in the material extracted from the back side of the base of the cross (Fig. 1c, area BC2) are indicated in Table 3. From this analysis, Cu andAg appear as the main metallic species in good agreement with the LIBS results. The absence of zinc is very strange while the presence of silver could be explained by the diffusion of this element through copper during the lamination process. The other elements detected (Ca, Na, K, Mg and Br) are attributed to impurities adhered to the surface of the base of the cross. Titanium is also present in this

136

Table 3. As in Table 2 from the area BC2 of the base of the cross.

Table 6. As in Table 2 from the area CP-SP3 of the strip of cache-pot.

Positions tested Na

Mg

K

Ca

Ti

Cu

Br

Ag

Positions tested C

Na

Al

Si

Ca

Cu

Zn

As

1 2 3 4 5 6 7

11.67 7.74 14.60 9.43 9.66 12.72 10.30

2.91 1.77 3.09 3.24 1.56 2.70 2.26

11.34 11.94 11.25 10.59 9.42 10.28 11.01

6.34 3.66 8.68 6.65 6.23 8.00 9.02

16.28 21.31 15.28 18.70 24.87 16.47 14.98

18.11 17.25 18.17 18.33 17.26 20.33 17.51

13.19 14.36 9.77 12.97 14.43 14.03 10.62

1 2 3 4 5

0.83 0.72 1.26 0.90 0.67

1.84 1.48 1.78 1.34 1.16

1.90 2.18 2.53 2.31 1.94

0.16 0.24 0.40 0.34 0.22

37.40 35.84 41.41 33.45 36.79

16.71 14.68 1.66 21.59 19.34

0.17 0.18 0.38 0.06 0.22

16.10 18.03 14.84 14.99 12.04 12.04 15.09

37.35 40.09 44.98 36.00 35.33

Table 4. As in Table 2 from the area C1 of the cross. Positions tested

Ti

Cu

Zn

Ag

1 2 3 4 5

1.6 2.0 6.9 5.7 7.4

59.3 62.7 58.3 62.6 54.5

1.6 2.5 1.4 0 3.5

37.5 32.8 33.4 31.7 34.6

Table 5. As in Table 2 from the area CP-L2 of cache-pot. Positions tested Na

Al

1 2 3 4 5 6

0.46 3.58 0.44 0.04 0.39 3.89 0.48 0.01 0.33 3.74 0.50 0.01 0.28 20.05 2.76 0.04 0.29 2.63 0.38 0.06 0.34 5.20 0.54 0.08

1.67 1.49 1.54 5.44 1.58 1.08

Si

Ca

Cr

Fe

Cu

Zn

0.22 0.23 0.19 0.47 0.25 0.18

33.33 32.31 31.23 1.15 29.69 33.63

10.84 11.21 10.87 0.46 10.21 11.23

Figure 5. Micrography of the laser spot corresponding to position C2 on the cross (Fig. 1d).

area of the object in a constant proportion without any Ni. Carbon, oxygen and silicon are also detected and not indicated in Table 3 due to their less importance. Table 4 presents the elements observed by EDS in the metal extracted from the metallic box of the cross (Fig. 1d, area C1). Copper and silver are the main elements in good agreement with the results obtained by LIBS (Table 1). A silvered lamination after the fabrication of the cross is again considered. Zinc and titanium were also observed in discrete amounts. Table 5 presents the elemental composition of the sample extracted from the back side of the legs of the cache-pot (Fig. 1e, area CP-L2). The presence of Zn and Cu (observed also by LIBS) with average composition 25–75% indicates that the nature of the material of the cache-pot legs is a brass alloy. The other elements observed are typically the contamination constituents.

Finally, the constituents of the strip of the cachepot obtained by EDS analyses of the material extracted from the area CP-SP3 reveal the same composition as the legs (Table 6), with different percentages of Zn and Cu (30–70%). The qualitative composition of the decorated strip mainly based in Cu and Zn indicates the use of a brass alloy to manufacture this material with a low Cu content. The use of a different brass alloy for the decorated strip than for the legs could have been intentional for the embossing working process. In the region corresponding to the union between the two strip ends (Table 1, area CP-SP4), no other elements different than those used in the composition of the strip itself, were found by LIBS analysis. This suggests that an autogenic welding has been employed for connecting the two strip sides. Differently, the soldering between the body of the object and the legs was based on the brazing technique, using a Pb-Sn alloy determined by LIBS (Table 1, area CP-SM6). The use of this alloy is justified by the necessity of providing a strong connection to support the weight of the object. Optical microscopy was used to observe the fingerprint of LIBS effect, i.e. the crater created by laser ablation needed for LIBS analysis. Figure 5 shows

137

European medieval cultural heritage). Discussions with Teresa Sinigalia (INMI, Romania), Oliviu Boldura (CERECS ART, Romania) and Octaviana Marincas (Universitatea de Arte “G. Enescu”, la¸si, Romania) are acknowledged. MO thanks CSIC-ESF I3P program for a postdoctoral contract. We also acknowledge the support of the Red Temática de Patrimonio Histórico y Cultural, CSIC. Figure 6. Optical micrographs of different areas of the cache pot showing the microstructure of the metal used: a) strip and b) leg.

the dimensions of the mark left in the area C2 of the metallic box of the cross after a single laser pulse. The diameter of the laser impact did not exceed 400 µm. Optical microscopy was also performed on cache- pot strip and legs to assess the microstructure of the metals that were used. Different morphologies were observed that lead to the conclusion that the strip was made by lamination (Fig. 6a) and the legs by casting (Fig. 6b). This choice was surely related to the different function of the two parts. The strip had to be decorated by moulding, while for the legs it was necessary to use a strong metal structure to stand mechanical stresses. 4

CONCLUSIONS

LIBS analysis has allowed the identification of the elemental composition of two metallic historical objects kept in the Sucevita Monastery in Suceava, Romania. The qualitative and quantitative analysis Romania. The qualitative and quantitative analysis of the samples, extracted from different zones of the metallic objects, by conventional techniques (EDS/SEM) allowed the comparison of the obtained results with those achieved by LIBS. The main components detected by LIBS are in correspondence with those observed by EDS/SEM. These results show the potential of LIBS for the in situ determination of the elemental composition of metallic objects without previous preparation and by consuming a micrometric amount of sample and all that is required is optical access to the material to produce the plume and collect the emitted light. LIBS measurements performed on the cross at its base and on the cache-pot allowed the characterisation of the objects from a technological point of view. Indications for the restoration and conservation of the items have been suggested to conservators and restorers. ACKNOWLEDGEMENTS Work funded by 2000 Culture Project (CLT 2005/A1/CHLAB/RO-488, Saving sacred relics of

REFERENCES Anglos, D. 2001. Laser-induced breakdown spectroscopy in art and archaeology, Applied Spectroscopy 55: 186–205. Bell, I. M., Clark, R. J. H. & Gibbs, P. J. 1997. Raman spectroscopic library of natural and synthetic pigments (P re- N 1850 AD), Spectrochimica Acta Part A 53: 2159–2179. Burgio, L., Corsi, M., Fantoni, R., Palleschi, V., Sialvetta, A., Squarcialuppi, M. C. & Tognoni, E. 2000. Self-calibrated quantitative elemental analysis by laser-induced plasma spectroscopy: application to pigments analysis, Journal of Cultural Heritage 1: 281–286. Burgio, L. & Clark, R. J. H. 2001. Library of FT-Raman spectra of pigments, minerals, pigment media and varnishes, and supplement to existing library of Raman spectra of pigments with visible excitation, SpectrochimicaActa Part A 57: 1491–1521. Bustamante, M. F., Rinaldi, C.A. & Ferrero, J. C. 2002. Laserinduced breakdown spectroscopy characterization of Ca in soil depth profile, Spectrochimica Acta B 57: 303–309. Carmona, N., Oujja, M., Rebollar, E., Römich, H. & Castillejo, M. 2005. Analysis of corroded glasses by laser induced breakdown spectroscopy, Spectrochimica Acta B 60: 1155–1162. Carmona, N., Oujja, M., Gaspard, S., García-Heras, M., Villegas, M. A. & Castillejo, M. 2007. Lead determination in glasses by laser-induced breakdown spectroscopy, Spectrochimica Acta B 62: 94–100. Castillejo, M., Martín, M., Silva, D., Stratoudaki, T., Anglos, D., Burgio, L. & Clark, R. J. H. 2000. Analysis of pigments in polychromes by use of laser-induced breakdown spectroscopy and Raman microscopy, Journal of Molecular Structure 550–551: 191–198. Clark, R. J. H., Curri, L., Henshaw, G. & Laganara, C. 1997. Characterization of Brown–Black and Blue Pigments in Glazed Pottery Fragments from by Castel Fiorentino (Foggia, Italy) Raman Microscopy, X-Ray Powder Difractometry and X-Ray Photoelectron Spectroscopy, Journal of Raman Spectroscopy 28: 105–109. Colao, F., Fantoni, R., Lazic, V. & Spizzichino, V. 2002. Laserinduced breakdown spectroscopy for semiquantitative and quantitative analyses of artworks-application on multilayered ceramics and copper based alloys, Spectrochimica Acta B 57: 1219–1234. Corsi, M., Cristoforetti, G., Hidalgo, M., Legnaioli, S., Palleschi, V., Salvetti, A., Tognoni, E. & Vallebona, C. 2006. Double pulse, calibration-free laser-induced breakdown spectroscopy: A new technique for in situ standardless analysis of polluted soils, Applied Geochemistry 21: 748–755.

138

Elfwing, M. & Norgren, S. 2005. Study of solid-state sintered fine-grained cemented carbides, Int. J. Refract. Met. Hard Mater. 23: 242–248. Klein, S., Stratoudaki, T., Zafiropulos, V., Hildenhagen, J., Dickmann, K. & Lehmkuhl, Th. 1999. Laser-induced breakdown spectroscopy for on-line control of laser cleaning of sandstone and stained glass, Applied Physics A 69: 441–444. Kuzuya, M., Murakami, M. & Maruyama, N. 2003. Quantitative analysis of ceramics by laser-induced breakdown spectroscopy, Spectrochimica Acta B 58: 957–965. Mandrino, Dj., Godec, M., Skraba, P., Sustarsic, B. & Jenko, M. 2004. AES, XPS and EDS analyses of an ironbased magnetic powder and an SMC material, Surface Interface Analysis 36: 912–916. Mantler, M. & Schreiner, M. 2000. X-ray fluorescence spectrometry in art and archaeology, X-ray Spectrometry 29: 3–17. Melessanaki, K., Mateo, M., Ferrence, S.C., Betancourt, P.P. & Anglos, D. 2002. The application of LIBS for the analysis of archaeological ceramic and metal artefacts, Applied

Surface Science 197-198: 156-163. Rinaldi, C. A & Ferrero, J. C. 2001. Analysis of Ca in BaCl2 matrix using laser-induced breakdown spectroscopy, Spectrochimica Acta B 56: 1419–1429. NIST Electronic Database.Available at http://physics.nist.gov Oujja, M., Vila, A., Rebollar, E., García, J. F. & Castillejo, M. 2005. Identification of inks and structural characterization of contemporary artistic prints by laser-induced breakdown spectroscopy, Spectrochimica Acta B 60: 1140– 1148. Rinaldi, C. A. & Ferrero, J. C. 2001. Analysis of Ca in BaCl2 matrix using laser-induced breakdown spectroscopy, Spectrochimica Acta B 56: 1419–1429. Scholten, J.H., Teule, J.M., Zafiropulos, V. & Heeren, R.M.A. 2000. Controlled laser cleaning of painted artworks using accurate beam manipulation and on-line LIBS-detection, Journal of Cultural Heritage 1: 215–220. Yaroshchyk, P., Body, D., Morrison, R. J. S. & Chadwick, B. L. 2006. A semi-quantitative standard-less analysis method for laser-induced breakdown spectroscopy, Spectrochimica Acta B 61: 200–209.

139

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Comparative study of historic stained glass by LIBS and SEM/EDX K. Szelagowska & M. Szymonski Institute of Physics, Jagiellonian University, Krakow, Poland

F. Krok Institute of Physics, Jagiellonian University, Krakow, Poland & Faculty of Conservation and Restoration of Fine Arts, Academy of Fine Arts, Krakow, Poland

M. Walczak∗, P. Karaszkiewicz & J.S. Prauzner-Bechcicki Faculty of Conservation and Restoration of Fine Arts, Academy of Fine Arts, Krakow, Poland

ABSTRACT: Medieval stained glass windows from St. Mary’s Church and Corpus Christi Basilica in Krakow (14th–16th century), as well as glass originating from several Polish historical buildings (18th–20th century) have been studied by means of Laser-Induced Breakdown Spectroscopy and Scanning Electron Microscopy with Energy Dispersive X-Ray spectrometer. The results are compared in order to obtain chemical composition of the samples and correlate them with the sample morphologies. Investigated glass samples can be divided into two groups: soda-lime-silicate glasses (modern glasses) and potash-lime-silicate glasses (historic glasses). Furthermore, the analysis of a stained glass sample of unknown dating acquired from a small Polish town, Grodziec, is presented. It is demonstrated that Grodziec stained glass has the characteristic potash–lime–silicate chemical composition, indicating that it belongs to the historic group of samples.

1

INTRODUCTION

Glass can be described as a solid substance of unordered structure composed of long chain molecules. Historic glasses, as being main constituents of some outstanding works of art, are a valuable part of cultural heritage. The inorganic glass, widely used since about 3000 BC, consists of several components (García-Heras et al. 2003): oxides that are network formers, oxides as network modifiers (flux and stabilizers) and colouring elements (chromophores). The main glass network former is silicon oxide (SiO2 ), however, in some glasses other oxides, such as aluminium oxide (Al2 O3 ), phosphorus oxide (P2 O5 ), etc, are used. Flux components decrease the silica melting temperature. Nowadays sodium oxide (Na2 O) is mostly used. However, in the medieval period potassium oxide (K2 O) was more common, since it was easier to obtain from wooden ashes. Stabilizers such as calcium oxide (CaO), magnesium oxide (MgO), lead oxide (PbO) and others are added to the glass batch to enhance their general chemical durability. Finally transition metal oxides provide colour to the glass

∗

Present address: Instituto de Química Física Rocasolano, CSIC, Madrid, Spain

(Co- blue, Cu- red or green, Cr- yellow, Fe- brown colour, Mn- pink colour, etc.). Detailed analysis of glass composition (qualitative and quantitative characterization) can play an important role in dating and locating the glass origin; e.g. the proportions of concentration of three main types of oxides can suggest the period of glass production. For instance, in the medieval glasses, a high concentration of K2 O usually suggests that it originates from Western and Central Europe, where wood ash, rich in potassium compounds, was used for the glass production (Newton & Davison 1989). Therefore, it is clear that application of new analytical methods and instruments give enhanced opportunities for glass characterization, so important for art and glass historians as well as for conservators. Moreover, it must be stressed that historic glasses, as a valuable part of cultural heritage, should be investigated in a non destructive way. In the field of glass research Scanning Electron Microscopy (SEM) with Energy Dispersive X-ray spectrometer (EDX) has been widely used for many years (Schreiner 1988). Although it is a very useful technique to obtain information about the major glass components, the sensitivity of SEM/EDX is often insufficient to measure the concentration of trace impurities and light elements. Alternatively, LaserInduced Breakdown Spectroscopy (LIBS) has been

141

developed as a very promising diagnostic technique that permits real-time, qualitative and under specific circumstances also semi-quantitative measurements of elemental composition of solids, liquids and gases (Burgio et al. 2000, Anglos 2001, Melessanaki et al. 2002, Müller & Sterge 2003). The main advantage of this method is related to its micro destructive character (a laser spot diameter that is used for analysis is smaller than 0.2 mm), sensitivity and possibility of carrying the in situ analysis, which is very important for glass objects. Recently, semi-qualitative analysis of the elemental composition of stones, metals, inks, and wall paintings have been carried out (Anglos 2001, Melessanaki et al. 2002, Oujja et al. 2005). Also chromophores and trace elements have been identified by this technique (Burgio et al. 2000, Castillejo et al. 2000, Carmona et al. 2005). Recent approaches to quantitative LIBS measurements in minerals, ceramics and soils have been reported (Colao et al. 2002, Kuzuya et al. 2003, Carmona et al. 2007, Müller & Sterge 2003). In this paper it is reported on the successful application of two complementary methods mentioned above (SEM/EDX and LIBS) for comparative analysis of medieval stained glass windows originating from several Polish historical buildings and churches (18th–20th century). The obtained results allow to determine the manufacturing period of a glass sample of unknown origin.

2

Table 1. List of the investigated glass samples. Period

Origin

1 2 3

Modern Modern Modern

4

19th c (?)

5

18th c. (?)

6

16th c.

7

14th c.

8

Unknown

Model glass Hand blown, Jaslo, Poland Machine made window glass, Krakow, Poland Hand blown, Brzesko Nowe, Poland Hand blown, Brzesko Nowe, Poland Corpus Christi Basilica, Krakow, Poland Stained glass, St. Mary’s Church, Krakow, Poland Stained glass, Grodziec, Poland

Table 2. Chemical composition of the selected glass samples as determined by SEM/EDX. Component (wt. %)

EXPERIMENTAL

In this study 5 samples of historic glass from Polish stained glass windows from different periods and for comparison 3 samples of modern (20th century) glass have been investigated. The samples are listed in Table 1. Glasses 2 and 6 had a blue colour, glasses 7 and 8 were green, and the rest of the samples were colourless. To analyse properly the bulk composition, each sample was cut and all the measurements were carried out on the fresh made cross-section area. For SEM investigation samples of 1 cm2 size were introduced to the microscope chamber (JOEL 5550). The imaging was performed with 20 keV electron beam (the size of the spot was in the range of 20 nm). Prior to the SEM imaging, the glass samples were covered with a thin carbon layer to avoid the charging of the sample surface. Simultaneously with the SEM imaging, the EDX technique was used allowing to study the elemental composition of the samples. For these measurements the silicon detector cooled down to liquid nitrogen temperature was used (IXRF Systems). The X- ray photons of energy up to 20 keV were acquired. LIBS analysis was carried out by means of a LIBS2000+ system (Ocean Optics) with a Q-switched

Sample no.

Sample no.

Si

Na

K

Ca

Al

Mg

Mn

1 2 3 4 5 6 7 8

28.8 21.2 22.5 16.6 23.0 22.8 19.0 20.5

7.5 8.7 7.5 0.1 0.5 0.2 0.3 0.1

0.5 0.6 0.1 6.7 15.5 14.6 26.8 13.6

3.2 4.1 4.5 8.2 17.7 8.2 14.9 11.8

0.7 0.6 0.4 0.5 0.8 0.3 0.4 0.8

1.9 – – 0.6 1.1 1.3 1.0 1.9

– – – 0.5 1.3 0.3 1.1 0.5

Nd:YAG laser (Big Sky Laser Technologies). Samples were irradiated at the fundamental wavelength of 1064 nm (pulse duration of 6 ns, 1 KHz) with a fluence of 2.6 J/cm2 . The plasma emission from the neutral and ionized atoms was collected by a bundle of glass fibres placed near to the analysed surface and transmitted to the optical analyser consisting of a set of 7 spectrometers. The final spectrum includes the 200–900 nm range.

3

RESULTS AND DISCUSSION

Table 2 shows the results of SEM/EDX analysis of the glass samples listed in Table 1. It is seen that modern glass samples belong to the soda-lime-silicate class, while the historic glass belongs to the potassium-lime silicate class. To further differentiate samples within the groups, SEM/EDX data were compared with the LIBS results. A typical example of the LIBS spectrum obtained for sample no. 8 is shown in Figure 1. Characteristic

142

Ca I

Ca I, II 300

400

500

600

Ca II 700

KI

Ca I Ca I

KI

Si I Si I Si I

Na I Na I Si I Si I

Mg I

Si I

Mg I, Si II Ca I

Ca I

Ca II

Ca II

Si I

Si I

Mg I

Si I Ca II

Intensity [a. u.] 200

800

900

Wavelength [nm]

I (Na 589.59 nm) / I (Si 612.40nm)

Figure 1. LIBS spectrum obtained for sample no. 8.

14

2

1

12 Modern

10 8 3

6

Historic

4 5 2

7

4 6

0 0

1

2

8 3

4

5

6

7

I (K 766.49 nm) / I (Si 634.71 nm) Figure 2. Cluster analysis of the intensity ratios: I (Na 589.59 nm)/I (Si 612.40 nm) (LIBS) versus I (K 766.49 nm)/I (Si 634.71 nm) (LIBS).

lines of main elements, such as Si, Ca, K, Mg and Na, were undoubtedly identified and denoted in the figure. The corresponding wavelengths of those lines are: Si (I) at 243.52, 250.69 251.60, 288.18, 390.60, 393.39, 612.40 and 614.35 nm; Si (II) at 412.80, 413.08, 504.10, 505.59, 634.71 and 637,14 nm; Ca (I) at 422.78, 429.98, 430.32, 431.95, 443.60, 445.51, 458.59 and 558.87 nm; Ca (II) at 315.88 317.93, 393.36, 396.84, 528.53 and 866.2 nm; K (I) 766.49

and 770.11 nm; Mg (I) 278.14, 517.27 and 552.84 nm; Na (I) 588.99 and 589.59 nm (NIST, Müller & Sterge 2003). For comparative analysis of the investigated samples, the following scatter plots have been prepared (Figs. 2–4) displaying dependences for selected pairs of data. In Figure 2 the intensity of sodium line 589.59 nm versus the intensity of potassium line 766.49 nm is shown. Both intensities are normalized with respect to intensities of relevant silicon lines in order to get semi-quantitative values of sodium and potassium concentration. Sodium line is normalised with respect to silicon line 612.40 nm and potassium line with respect to silicon line 634.71 nm. As it can be seen, modern glass samples (no. 1–3) and historic glass samples (no. 4–8) form two distinct groups, although some differences among each of the two groups are present. First of all, modern glasses have higher sodium content, as expected (Garcia-Heras et al. 2003). However, 3rd sample seems to be well separated from the other two modern samples towards lower sodium content. Furthermore, it is clear that samples no. 4, 5 and 6 create their own group among the historic glass group, while samples 7 and 8 differ from that group and from each other. It is known that historic glasses, on their own, are divided into several groups and in order to investigate this aspect with LIBS technique further studies are required and undertaken.

143

I (Na 589.59 nm) / I (Si 612.40 nm)

14

samples 5 and 6 are fairly similar. Those discrepancies are due to the existence of different glass groups among the historical glasses, as mentioned above.

2

1

12 Modern

10

4

8 3

6 4

Historic 4

2

5

7 8 6

0 0

2

4

6

8

10

Na [wt %]

Figure 3. Cluster analysis of intensity ratios: I (Na 589.59 nm)/ I (Si 612.40 nm) (LIBS) versus concentration of Na in wt. % (SEM/EDX).

I (K 766.49 nm) / I (Si 634.71 nm)

7

8

6 5

Modern 7

4 3

2

4

6 5

1

2 1

Historic

3 0 0

5

10

15 K [wt%]

20

25

30

Figure 4. Cluster analysis of intensity ratios I(K 766.49 nm)/ I(Si 634.71 nm) (LIBS) versus concentration of potassium in wt. % (SEM/EDX).

In Figure 3 the intensity of the sodium Na line at 589.59 nm normalized with respect to the intensity of the silicon Si line at 612.40 nm is compared with the sodium content (in wt. %) obtained with SEM/EDX. As can be seen, points gather into two distinct groups corresponding to two classes of glass samples. Once again, sample no. 3 differs from other modern samples, whereas historic glasses are fairly close to each other. In Figure 4 results of measurements of potassium content are shown. Intensity of the potassium line 766.49 nm is normalized with respect to the intensity of silicon line 634.71 nm and plotted versus potassium content obtained with SEM/DEX. Contrary to sodium, potassium concentrations show high similarity for the modern glass samples (although sample 3 still differs from other samples). Rather large differences between historic glass samples are seen. Only

CONCLUSIONS

Summarizing, a comparative analysis of different glass samples (historic and modern) has been performed by means of LIBS and SEM/EDX techniques. From results obtained some conclusion may be drawn. Namely, in the modern glass group, being sodalime-silicate glasses, different composition of the 3rd sample may suggest that this particular sample is of different origin than the others (for instance imported from outside of the particular region). Furthermore, for the historic glass group, being potash-lime-silicate glasses, samples 4, 5 and 6 show clear similarities. This similarity is particularly interesting result, if one notes that the sample no. 6 is from 16th century, while samples no. 4 and 5 are dated to supposedly be from 18th/19th centuries. This may suggest that samples 4 and 5 could be older than expected before. This hypothesis, however, needs to be confirmed by further investigations. Furthermore, it is clear that glass sample no. 8, of unknown origin, belongs to the historic glass group (potash-lime-silicate). It means that this glass (and at least parts of the stained glass panel from which the sample was collected) has characteristic properties of the medieval stained glass. Finally, the study demonstrates that parallel application of both analytical methods, LIBS and SEM/EDX, can provide complementary material for comparative investigations of historic stained glasses allowing the identification of the sample origin in time and space.

ACKNOWLEDGEMENTS Authors thank the Polish Ministry of Science and Higher Education for the financial support. REFERENCES Anglos, D. 2001. Laser-induced breakdown spectroscopy in art and archaeology. Appl. Spectrosc. 55: 186A–205A. Burgio, L., Corsi, M., Fantoni, R., Palleschi, V., Sialvetta, A., Squarcialuppi, M.C., Tognoni, E. 2000. Self-calibrated quantitative elemental analysis by laser-induced plasma spectroscopy: application to pigments analysis. J. Cultural Heritage 1: 281–286. Carmona, N., Oujja, M., Rebollar, E., Römich, H., Castillejo, M. 2005. Analysis of corroded glasses by laser induced breakdown spectroscopy. Spectrochimica Acta B 60: 1155–1162. Carmona, N., Oujja, M., Gaspard, S., García-Heras, M., Villegas, M.A., Castillejo, M. 2007. Lead determination in glasses by LIBS. Spectrochimica Acta B 62: 94–100.

144

Castillejo, M., Martín, M., Silva, D., Stratoudaki, T., Anglos, D., Burgio, L., Clark, R.J.H. 2000.Analysis of pigments in polychromes by use of laser-induced breakdown spectroscopy and Raman microscopy. J. Molec. Struct. 550–551: 191–198. Colao, F., Fantoni, R., Lazic, V., Spizzichino, V. 2002. Laserinduced breakdown spectroscopy for semiquantitative and quantitative analyses of artworks-application on multilayered ceramics and copper based alloys. Spectrochimica Acta B 57: 1219–1234. García-Heras, M., Gil, C., Carmona, N., Villegas, M.A. 2003. Weathering effects on materials from historical stained glass windows. Mater. Construct. 270: 21–34. Kuzuya, M., Murakami, M., Maruyama, N. 2003. Quantitative analysis of ceramics by laser-induced breakdown spectroscopy. Spectrochimica Acta B 58: 957–965. Melessanaki, K., Mateo, M., Ferrence, S.C., Betancourt, P.P., Anglos, D. 2002. The application of LIBS for the analysis of archaeological ceramic and metal artifacts. Appl. Surf. Sci. 197–198: 156–163.

Müller, K., Sterge, H. 2003. Evaluation of the analytical potential of laser-induced breakdown spectroscopy (LIBS) for the analysis of historical glasses. Archaeometry 45: 421–433. Newton, R., Davison, S. 1989. Conservation of Glass. London: Butterworths. NIST Electronic Database.Available at http://physics.nist.gov/ cgi-bin/AtData/lines-form. Oujja, M., Vila, A., Rebollar, E., García, J.F., Castillejo, M. 2005. Identification of inks and structural characterization of contemporary artistic prints by laser-induced breakdown spectroscopy. Spectrochimica Acta B 60: 1140–1148. Schreiner, M. 1988. Deterioration of stained medieval glass by atmospheric attack. Part 1. Scanning electron microscopic investigations of the weathering phenomena. Glastech. Ber. 61: 197–204.

145

Portable Laser Systems for Remote and On-Site Applications

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Scanning hyperspectral lidar fluorosensor for fresco diagnostics in laboratory and field campaigns F. Colao, L. Caneve, R. Fantoni, L. Fiorani & A. Palucci ENEA, FIM-FISLAS, Frascati RM, Italy

ABSTRACT: A scanning hyperspectral system based on Laser Induced Fluorescence (LIF) has recently been developed for optical characterization of surfaces relevant to cultural heritage. This paper describes its field application for frescoes diagnostics in some monasteries in Bucovina, a Romanian region. The LIF system provides information on the present conservation status, identifies areas with biological attack and gives details on the restoration methods of the frescoes under study.

1

INTRODUCTION

The recent technological development of optical devices, able to measure spectral components in fast and accurate, way represents one of the major breakthroughs in the field of work art investigation. At present, a number of imaging multispectral and hyperspectral sensors as well as specialized software have been conceived (Anglos 1999, Carcagni 2007), making accessible the tools to take advantage of imaging technologies to a continuously increasing number of experts. Algorithms for digital image analysis, mainly matured in the field of remote sensing of land and vegetation, have been developed and their application to artworks investigation represents one of the most powerful, promising and fast growing technology. This paper describes the results obtained by the application of Laser Induced Fluorescence (LIF) technique for diagnostics of frescoes during the 2006 campaign in Bucovina. In particular, the sites under study were the Resurrection Church in the Suceviþa Monastery, the Saint Nicholas Church in the Popˇau¸ti Monastery near Boto¸sani and the Saint Nicholas Church in Bˇaline¸sti, all currently under restoration. Among a lot of techniques offering the possibility of pigments and biodeteriogens identification in frescoes, the laser based methods were applied successfully for in situ or remote characterization of artwork surface, both for diagnosis (Asmus 2003) and for maintenance (Klein 2001). In particular, the LIF technique has been recently applied to the remote sensing of fresco (Comelli 2004) to detect characteristics invisible with the naked eye without moving samples from their original location. Large images can be collected once a fluorescence lidar system equipped with a scanning device is utilized (Lognoli 2003).

However the unmixing of pixel spectral information still remains as one of the most challenging tasks to carry out in order to complete data pre processing thus allowing for an appropriate data interpretation (Coma 2000). Indeed the area pertaining to a single pixel contains different materials all of them contemporarily excited by the incident laser beam: the measured signal is then a combination of the fluorescence induced on all the different material layers. While some pigments as lime white have distinct LIF emission, other pigments, as for example red ochre, have a much less distinct signature (Anglos 1996, Nevin 2006). Nonetheless they can be distinguished when the signal emitted by plaster and substrates laying underneath is modulated both spectrally and in intensity by the absorption and re-emission of pigments located close to the superficial layers. By this rationale we try to present in this paper how to interpret the LIF images, although they are characterized by a very complex and extremely rich spectral signatures which at a first glance might appear as just undesired details. A compact scanning lidar fluorosensor apparatus has been designed and built in the ENEA laboratory and it has been used to perform the field measurements of this campaign. Several multispectral images were obtained and their combination was attempted in order to reveal the occurrence of surface biodegraded area.

2

EXPERIMENTAL

The hyperspectral scanning LIF system here used is based on a previous instrument’s version developed during former projects (Aristipini 2004). All the mechanical and optical elements have been renewed

149

and installed in an aluminum box of 58 × 43 × 36 cm3 , weighting less than 15 kg. Its small size and light weight allow an easy transport of the system and its operation from scaffoldings, in the case of surfaces out of the current maximum range for remote operation (10 m). The layout of the system is given in Figure 1. The Thomsom DIVA is a diode pulsed Nd:YAG laser used as light source to generate radiation at 266 or 355 nm, producing laser pulses of 10 ns duration at a fluence of 1 to 50 mJ/cm2 . The spectrometer entrance is protected from the intense backscattered radiation by means of an appropriate dichroic filter. Proper selection of the energy per pulse content is a critical issue, especially in the case of precious frescoes which might be damaged by bleaching processes induced by UV light excitation (Athanassiou 2000, Sansonetti 2000). A systematic study on this topic has not yet been completed and the laser energy was kept as low as possible to ensure a signal to noise ratio of the order of few units. Since no optical elements are used to collimate the laser beam, the overall spatial resolution is a function of the operation distance and in the present case we may infer a resolution of approximately 1 to 2 mm from the spot size on the target. The digitized spectrum is transferred to a portable computer where a LabView program allows the user to set experimental parameters, to control data acquisition and to perform a preliminary data analysis. 2.1

Figure 1. Optical layout of hyperspectral scanning LIF system.

burning deposits etc.) and biological agents as well. Previous study on biodeterioration agents mainly identified the presence of fungi, while chlorophyll, easily detectable by the LIF system (Colao 2005), has not been found, probably because the low level of natural light inside the churches under study prevented the development of photosynthetically active microorganisms.

Data analysis

The hyperspectral LIF scanner here described is able to measure the fluorescence radiation induced on a surface at a series of narrow and contiguous spectral bands. In every scanned image the light emitted by each pixel is decomposed in its components, providing the maximum extent of spectral information today conceivable in the UV to VIS wavelength range (200–800 nm). After the completion of an exhaustive reference spectral library for all the substances and pigments present in the sample under study, the analysis of hyperspectral images allows for distinguishing among different spectrally similar materials. However the image analysis and subsequent data interpretation still remain one the most difficult and challenging tasks, since for a variety of reasons the reference database is far from being completed. To work out this problem we resort to a less accurate data analysis, based on the attempt to isolate and identify spectral characteristics of selected area directly on the LIF images and validating a posteriori with complementary analyses. As it will be detailed in the following, we direct our efforts to the identification of deteriorated areas on fresco and mural paintings by environmental factors (wall humidity, salt efflorescence, beeswax

2.2 Data processing Various factors affect the signal measured by the LIF scanner; just to cite a few, it is worth mentioning (1) the thermal drift caused by the long times needed to complete an acquisition scan, (2) the radiometric distortion introduced by optical elements between the first collection mirror and the spectrometer entrance, (3) the spectral transfer function of the spectrometer itself. Also (4) the geometrical effects due to the change in incidence angle should be considered. Accurate data processing requires the correction for all of these effects. Actually the overall radiometric response of the apparatus has been experimentally measured by using traceable reference sources: a deuterium lamp for the 200 to 380 nm spectral region and a tungsten lamp for the region from 400 to 800 nm. As the cross talk between adjacent spectral band and the effects of finite spectral resolution of the spectrometer is concerned, we estimated them to affect the LIF signal for less than 10% and consequently they were ignored. The geometrical effect correction has been carefully considered, and it can be introduced in the data pre

150

Figure 2. LIF spectra of selected fungi’s strain excited at λex = 355 nm. Strains starting from upper left are SCH = Schizophyllum commune, BJC = Paecylomices varioti, TRICH = Trichoderma viride, CHAET = Chaetomium, ALT = Alternaria, HORM = Hormodendron.

processing chain once a LRF scan of the same fresco’s portion is available. Details on this correction may be found in (Colao 2005). Finally the Spectral Angle Mapper (SAM) algorithm has been used to analyse images (Colao 2007). According to SAM, the spectrally resolved intensities can be treated as vectors’ components. Then it is possible to compute the angle between a given pixel spectrum and a reference spectrum: the smaller is the angle, the higher will be the similarity between pixel and reference spectra. To improve the diagnostic readability of images and to give relevance to selected areas, a threshold has been introduced for the spectral angle, thus giving a black and white version of the original scanned image. Whenever the reference spectra pertains to a bio contaminated sample, this processing marks well the areas relevant for the bio deterioration diagnostic. 3

RESULTS AND DISCUSSION

LIF measurements taken with the scanning hyperspectral lidar fluorosensor are grouped in two sections, the first concerning with excitation at third harmonic of the Nd:YAG laser (355 nm) and the second with excitation at fourth harmonic of the Nd:YAG laser (266 nm). The first wavelength is optimal for excitation of pigments and has a strong excitation efficiency in case of chlorophyll, thus allowing for identification of algae or autotrophic micro organisms; the second is better suited for identification of heterotrophic micro organisms like fungi and also for identification of organic binders lying over the pigments.

3.1 Laboratory data bases Reference fluorescence spectra of biological samples and of acrylic resins used as consolidants were measured in order to support the assignment of features emerging in LIF analysis of frescoes. 3.1.1 Microorganisms identification Pure cultures of several fungi strains were excited at 355 nm. LIF acquired spectra (Fig. 2) sometimes show common broadband peaks, while in other cases show specific and distinctive features. Moreover some samples are more intensively fluorescent than others, while few samples with marginal fluorescence intensity were found once excited at this wavelength. As an example of common feature we might consider the UV/blue fluorescence peaked at 400–450 nm, which is present in almost all of the considered samples. Similarly the green fluorescence, with a maximum at 500 nm, is observed in almost all of the samples. Quite different is the case of red fluorescence with maximum at 600–650 nm, which has a significant intensity only in few samples. A preliminary evidence from the experimental findings is that the fluorescence intensity and bands ratios do depend on a number of factors like fungi strains, physiological state and environmental conditions as well (for example the temperature of the specimen under study). Data analysis made also evident that the classification of organisms could be possible only by using full spectral information, thus requiring for a hyperspectral detection system. However it might be possible that a double or multiple excitation wavelengths

151

LIF Intensity (a.u.)

1.0 0.8 0.6 0.4 0.2 0.0 300

400

500

600

700

Wavelength [nm]

Figure 4. Processed images of a scan in Sucevita church (90 × 80 cm2 ). Left image (a) was taken with a conventional photo camera; in the right pane (b) it is shown a black and white LIF intensity on selected spectral bands.

Figure 3. LIF spectra of two consolidants on plaster: Calaton (dash dot), Mowilith (solid line).

could reduce the requirements on spectral information needed for unambiguous fungi strain identification. Further laboratory experiments are currently carried out to investigate this point. 3.1.2 Consolidants identification The identification of organic materials as protective varnishes and or consolidants in works of art is of particular interest, since they strongly influence the deterioration processes (Domenech 2001). Here we report some preliminary results obtained by applying the fluorescence induced by UV laser excitation at 266 nm to distinguish different consolidants on plaster laboratory samples. For the purpose of the present experiment three different consolidants were taken into account: Calaton, soluble nylon, Mowilith, vinyl acetate and Paraloid, ethyl acrylate (Cappitelli 2004, Abdel-Kareem 2000). In Figure 3, as example, we report the UV fluorescence spectra for Calaton (dashdot line), and for Mowilith (solid line). The spectra of Figure 3 were then used as references to apply SAM algorithm to the analysis of images acquired with the hyperspectral LIF scanner. The identification of areas treated with different consolidants types was excellent (not shown here) being independent from several factors including (1) the plaster used, (2) the pigment in the underlying layer and (3) the amount of consolidant sprayed on the surface (at least in the concentration range used for the present experiment). In all the tested cases we obtained an unambiguous and successful identification. Measurements of fluorescence emission from the same consolidant samples excited at 355 nm are in progress in our laboratory, in order to check the possibility to distinguish them meanwhile acquiring information on pigments and biodeterioration. No many data, in fact, are available in literature with respect to the fluorescence of consolidants, but the possibility of their identification by applying PCA to LIF spectra obtained upon excitation at 355 nm was demonstrated (Ballerini 2001).

3.2 Frescoes at 355 nm excitation The 355 nm laser excitation wavelength was used to excite fluorescence in Sucevita Monastery on four different portions of a painted dome. All the scanned areas (70 × 80 cm2 ) are partially restored and their preservation status is generally poor showing a blackish superficial deposit encompassing almost all of the frescoes. Restoration started from the gilded areas and actually it is not yet completed, thus it does not extend on other portion of the frescoes. Figure 4 shows details on the first scanned image. In the left panel (Fig. 4a) we can see a black and white image of the area taken with a standard photocamera (Canon PhotoShot 600), and in the right pane (Fig. 4b) the image of the same fresco region acquired by LIF. The acquired fluorescence spectral data have been combined to generate a false RGB color image using the spectral channel intensities at 340 nm, 480 nm and 560 nm respectively for Blue Green and Red (Colao 2005); this image is characterized by a high spectral contrast reproducing icon details with an exceptional accuracy. Spectral changes appear as a weak modulation over a mean LIF signal dominated by the plaster contribution. The superficial pigments and deposits act as quencher of the plaster fluorescence leaving their signatures as an entangled combination of (1) laser excitation absorption (2) plaster fluorescence absorption and (3) possibly pigments fluorescence emission. Gilded details appear in several different parts of the frescoes: in the centre the aureole of the main character at left side (Jesus), as well as in stars in the upper central part. From a qualitative point of view the gilding fluorescence represents a case quite different from the rest of the scanned area: despite that metal fluorescence is generally very low and frequently completely absent, we observe a strong spectral intensity and unique spectral features. On one hand this is explained by the fact that to prevent superficial layer detachment, during restoration gilded areas have been partially treated with organic consolidants like Paraloid.

152

Figure 5. Processed images of a scan in Sucevita church. Left panel (a) is a SAM image with bio-attack obtained with a reference spectrum; in the right panel (b) a SAM image obtained with a reference spectrum from a portion of a cleaned fresco surface.

Areas with bio attack by fungi occur on upper left regions; the similarity map with SAM and a reference spectrum taken in a partially contaminated region is shown in Figure 5(a). To enhance the bio contaminated areas, this image has been processed with a threshold filter set at a spectral angle of approximately 5◦ , thus eliminating intermediate intensity gray levels. As result we obtain an easy recognition of contaminated and cleaned portions. In this case the heavily attacked areas appear in black, while non or slightly contaminated portions are in white. As a further check, Figure 5b reports areas with small biological attack; in this case the SAM similarity map has been computed using a reference spectrum taken in a cleaned region. 3.3 Frescoes at 266 nm excitation The 266 nm laser excitation wavelength was used to excite fluorescence from fresco contained in Ballinesti Church, where a strong biodeterioration by fungi is visible on the Church walls sometimes also by the naked eye. To face with biological attack, the restorers working in this site were planning to use biocides and prudently were testing their effects on a small area of a church wall. While they were using standard methods to assess potential unwanted deterioration caused by the biocide, it was also of great interest to characterize the used chemical agents by means of LIF. Figure 6a shows a black and white image of a detail of a fresco on the church wall taken with a standard photo camera, the same area was then scanned with the hyperspectral LIF system. Three regions can be recognized in Figure 6a: the first one marked as i) is typical of a strong biological attack, a second one marked as ii) only treated with biocide and a third area marked as iii) treated with biocide and subsequently cleaned by restorers. SAM algorithm was used to identify selected regions in the scanned area.The three specified regions with different spectral characteristics were identified respectively for biological attack (from Penicillium

Figure 6. Processed images of a scan in Ballinesti church (35 × 35 cm2 ). Upper left (a) image by a conventional photo camera. SAM maps obtained with reference spectrum of (b) a bio-attack; (c) a biocide, (d) a cleaned area.

crysogenum, independently identified), treated with biocide, and areas which were treated with biocide and subsequently cleaned by restorers. SAM similarity map obtained using biological attack reference spectrum is shown in the upper right (Fig. 6b), Figures 6c and 6d show SAM mapping obtained respectively with biocide and cleaned reference spectra. SAM algorithm is able to correctly identify the treated and untreated regions, while Figure 6d shows that cleaned areas are identified less precisely as is made evident from the non negligible number of misidentified pixels in treated but not cleaned portion. Several considerations must be made to explain the partial failure of identification on cleaned areas. First of all we must observe that the spectral differences in treated only and treated plus cleaned areas are quite small. To quantify this observation we computed the average spectral correlation of these areas. It resulted that regions i) and ii) have a correlation coefficient of 0.96, meaning that a part for floral decoration appearing in the central part of the regions, they have been covered with the same kind of pigment. The spectral differences are then due only to the presence of the biocide chemicals, which evidently do not have any special and distinctive feature. A second point deserving some attention is related to the fact that the treatment and cleaning in region iii) was made at least six months before the LIF analysis. Since in the meantime the appropriate actions to avoid fungi proliferation were still to be completed, it happened that fungi started again to colonize the wall, thus contaminating with hyphae and mycelium also regions ii)

153

ACKNOWLEDGEMENTS

Figure 7. Photo picture of a Sucevita fresco detail (12 × 12 cm2 ) (a) and corresponding SAM image for identification of consolidant (b).

and iii). Nonetheless SAM proves to be an effective and powerful tool for identification. As found in the laboratory database construction, the excitation at 266 nm is especially suitable to reveal the presence of specific consolidants utilized during the restoration. This could be verified in field measurements on frescoes with consolidated gilded details. In Sucevita monastery, a particular of the same image already examined at 355 nm and shown in Figure 4 (Jesus’ head picture) has been scanned at 266 nm in order to trace the extent of the restorer’s action (see Fig. 7). The presence of Paraloid at specific locations on the aureole has been revealed by its characteristic fluorescence emission band peaked at 310 nm. The SAM mapping image reconstruction of the scanned area, shown in Figure 7b, highlights the not uniform distribution of the consolidant due to the restorer’s intention to cover only the detaching gold leaf.

4

CONCLUSIONS

The new compact laser scanning LIF apparatus has successfully applied to the painted walls investigation of relevant cultural heritage. The advantages of the LIF technique as diagnostic tool for cultural heritage are mainly related to the capability of space resolved remote measurements with minimal invasiveness and of a rapid acquisition of data. The presented set-up is compact and solid, the optical system is in 58 × 43 × 36 cm3 , light enough (about 20 kg) and relatively cheap. Moreover, the technique gives additional, valuable and complementary information with respect to conventional visible or infrared imaging and the results presented demonstrate that hyperspectral LIF gives useful information to identify biological attacks areas on painted surfaces. Fluorescence emissions can be also related to the different materials and methods applied during the fresco realization. It is indeed an effective tool for specific diagnostics of cultural heritage.

The contribution of I. Gomoiu for the selection, handling and growing the fungi strains is gratefully acknowledged. Also we thank the contribution of I. Nemec for the preparation of plaster samples with pigments and consolidants. For the field campaign the invaluable support of R. Radvan, coordinator of the CULTURE project “Saving Sacred Relics of European Medieval Cultural Heritage” is gratefully acknowledged. Work partially supported by the European Union in the framework of the “CULTURE 2000” program (project CLT 2005/A1/ CHLAB/RO-488).

REFERENCES Abdel-Kareem 2000. Microbiological testing of polymers and resins used in conservation of linen textiles. 15th World Conference on Nondestructive Testing, Rome. Anglos D., Solomidou, Zergioti, Zafiropulos V., Papazoglou, Fotakis C., 1996. Laser-Induced Fluorescence in Artwork Diagnostics: An Application in Pigment Analysis, Appl. Spectroscopy, 50:1221–1337. Anglos D., Balas C., Fotakis, C., 1999. Laser spectroscopic and optical imaging techniques in chemical and structural diagnostic of painted artwork. American Laboratories 31: 60–67. Aristipini, P., Colao, F., Fantoni, R., Fiorani, L. & Palucci, A. 2004. Compact scanning lidar fluorosensor for cultural heritage diagnostics, Proceedings of SPIE 5880: 196–203. Asmus, J.F. 2003. Non-divestment laser applications in art conservation. Journal of Cultural Heritage 4: 289–293. Athanassiou, A., Hill, A.E., Fourrier, T., Burgio, L. & Clark, R.J.H. 2000. The effects of UV laser light radiation on artists pigments. Journal of Cultural Heritage 1: S209–S213. Ballerini, G., Bracci, S., Pantani, L. & Tiano, P. 2001. Lidar remote sensing of stone cultural heritage: detection of protective treatments. Optical Engineering 40: 1579–1583. Cappitelli F., Zanardini E., Sorlini C., 2004. The biodeterioration of synthetic resins used in conservation. Macromol. Biosci. 4:399–406. Carcagni, P. 2007. Multispectral imaging of paintings by optical scanning. Optics and Lasers in Engineering, 45: 360–367. Colao, F., Fantoni, R., Fiorani, L., Palucci, A. & Gomoiu, I. 2005. Compact scannino lidar fluorosensor for investigations of biodegradation on ancient painted surfaces. Journal of Optoelectronics and Advanced Materials 7: 3197–3208. Colao, F., Fantoni, R., Fiorani, L. & Palucci, A. 2007. In press. Scanning hyperspectral lidar fluorosensor: a new tool for fresco diagnostics. Proceedings of Conference Conservation Science 2007. Coma, L., Breitman, M., Ruiz-Moreno, S. 2000. Soft and hard modelling methods for deconvolution of mixtures of Raman spectra for pigment analysis. A qualitative and quantitative approach. Journal of Cultural Heritage 1: S273–S276.

154

Comelli D., D’Andrea C., Valentini G., Cubeddu R., Colombo, C., Toniolo L., 2004. Fluorescence lifetime imaging and spectroscopy as tools for nondestructive analysis of works of art. Applied Optics 43: 2175–2183. Domenech-Carbo M.T., Gimeno-Adelantado J.V., Bosh-Reig F., 2001. Identification of synthetic resins in works of art by Fourier transform infrared spectroscopy. Applied Spectroscopy 55: 1590–1602. Klein, S., Fekrsanati, F., Hildenhagen, J., Dickmann, K., Uphoff, H., Marakis, Y.& Zafiropulos, V. 2001. Discoloration of marble during laser cleaning by Nd:YAG laser wavelengths. Appl. Surf. Sci. 171 : 242.

Lognoli, D., Cecchi, G., Mochi, I., Pantani, L., Raimondi, V., Chiari, R., Johansson, T., Weibring, P., Edner, H. & Svanberg, S. 2003. Fluorescence lidar imaging of the cathedral and baptistery of Parma, Applied Physics B 76: 457–465. Nevin A., Cather S., Anglos D., Fotakis C., 2006. Analysis of protein-based binding media found in paintings using laser induced fluorescence spectroscopy, Anal. Chim. Acta., 573-574C: 341–346. Sansonetti, A. & Realini, M. 2000. Nd:YAG laser effects on inorganic pigments. Journal of Cultural Heritage 1: S189–S198.

155

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

A lidar experiment for the characterization of photoautotrophic and heterotrophic biodeteriogens by means of remote sensed autofluorescence spectra V. Raimondi, L. Palombi, D. Lognoli & G. Cecchi Institute for Applied Physics ‘Nello Carrara’ – National Research Council, Sesto Fiorentino, Florence, Italy

I. Gomoiu Institute of Biology, National Arts University of Bucharest, Splaiul Independentei, Bucharest, Romania

ABSTRACT: Remote laser-induced autofluorescence measurements have been performed on a set of biodeteriogens samples selected in an archaeological site, Tropaeum Traiani, near Constanta, Romania. Both photoautotrophic (lichens) and heterotrophic biodeteriogens (pure cultures of fungi and bacteria) were examined with a high resolution fluorescence lidar system featuring a UV laser (XeCl at 308 nm) as excitation source. The measurements were carried out on the in vivo samples placed at a distance of about 25 m in the outdoor, in full sunlight. The results confirm the fluorescence lidar technique as a powerful method for the remote detection and characterisation of photoautotrophic biodeteriogens and also open good prospects for remote, non-invasive monitoring of heterotrophic biodeteriogens on non-movable objects, also outdoors.

1

INTRODUCTION

Biodeterioration is a typical process that affects stone cultural heritage, especially outdoors. Its monitoring, treatment and timely prevention is thus essential for the conservation of cultural heritage. This, however, preliminarily requires the identification of the microorganisms affecting the surface of monuments and an assessment of the extension of the contaminated areas. Extensive diagnostics of the stone outdoor cultural heritage can become quite demanding in terms of time and costs when traditional techniques are applied. In this context, remote sensing can offer several advantages to assess the overall distribution of biodeteriogens over extended surfaces and to routinely monitor its modifications for a timely prevention. In particular, a preliminary assessment of the extension and type of biological growth on monuments’ surfaces can be exploited for the identification of the most suitable sampling areas for possible applications of specific analytical techniques. In vivo fluorescence spectral signatures of algae and cyanobacteria have been studied for a long time (Yentsch et al. 1979, Bazzani et al. 1992) and the fluorescence properties of these photoautotrophic biodeteriogens have been also exploited as a diagnostic tool for the monitoring of biodeterioration on stone cultural

heritage with remote sensing techniques (Raimondi et al. 1998, Weibring et al. 2001, Lognoli et al. 2002). However, although visual observation of fluorescence emitted from lichens under UV light has been carried out for a long time (e.g. observation with epifluorescence microscopy techniques or confocal microscopy; see e.g. Mathey et al. 2001), spectral signatures of lichens fluorescence have been not studied to a great extent up to now and only few works on this subject are available (Mathey et al. 2001, Hidalgo et al. 2002). As far as the autofluorescence of heterotrophic biodeteriogens is concerned, only few papers are available in the literature about their spectral signatures (Arcangeli et al. 1997, Bengtsson et al. 2005, Colao et al. 2005, Raimondi et al. 2007). This work illustrates the results of a set of measurements aimed at a fluorescence-based characterisation of different types of biodeteriogens and carried out during a joint experiment within a EU-funded Culture 2000 project at a Roman archaeological site, Tropaeum Traiani, near Constanta, Romania. Laser induced autofluorescence spectra acquired with a mobile lidar system featuring a UV laser as an excitation source were the measurements collected. The system was deployed at the site and both photoautotrophic and heterotrophic biodeteriogens were examined during the experiment in uncontrolled environmental conditions.

157

Figure 2. Picture of the pure culture of the fungal strains of Verticillium sp. (left) and of Aureobasidium pullulans (right).

Figure 1. Picture of the stone samples affected by biological growth. Most stones showed the presence of lichens as listed in Table 1.

Table 2. Pure cultures of fungal and bacterial strains, and relevant nutrient media, examined during the experiment (second session).

Table 1. Lichens on stone substrate examined during the experiment (first session). Label

Description

Sample A Sample B Sample C Sample D

Parmelia sp. Caloplaca sp., Parmelia sp. Caloplaca sp., Physcia sp. White crust, not identified

2 2.1

MATERIALS AND METHODS Samples description

The samples consisted of two different sets: – a selection of stones affected by the growth of photoautotrophic biodeteriogens, particularly lichens; – a set of pure cultures of different fungal and bacterial strains, previously isolated from the samples collected in the archaeological site. All the samples were selected in the Tropaeum Traiani archaeological site. Figure 1 shows the set of stones affected by biological growth, mainly lichens, that were selected for the first session of measurements. A description of the biodeteriogen-affected stone samples presented in this paper can be found in Table 1. Figure 2 shows a picture of pure cultures of the two fungal strains – Verticillium sp. (left) and Aureobasidium pullulans (right) – among those collected at the archaeological site and isolated in the laboratory. A list of the bacterial and fungal strains examined during the measurement session is reported in Table 2. 2.2

Instrumentation

Fluorescence lidar remote sensing allows to transfer the Laser Induced Fluorescence (LIF) technique to the

Label

Description

F-blank F1 F2 B-blank B1-a B1-b B2-a B2-b

Nutrient medium Aureobasidium pullulans Verticillium sp. Nutrient medium Bacillus sp.1 Bacillus sp.2 Pseudomonas sp.1 Pseudomonas sp.2

outdoor environment where uncontrolled environmental conditions are met. If the lidar is provided with a scanning system to scan the laser beam over the target, a fluorescence map can be acquired. A further description of the technique can be found in e.g. Weibring et al. (2001), Lognoli et al. (2003) and Svanberg (2005). The fluorescence measurements were carried out with a high spectral resolution fluorescence lidar that has been in-house developed at CNR-IFAC and has been operative aboard a mobile laboratory since 1991. Although originally intended for the monitoring of marine environment and afterwards of vegetation, the CNR-IFAC lidar system has already shown great potential for the remote non-invasive monitoring of stone cultural heritage since first experiments carried out in the mid 1990’s (Raimondi et al. 1998, Lognoli et al. 2003). The CNR-IFAC lidar system features a XeCl excimer laser emitting at 308 nm as an excitation source. The signal is collected with a 25 cm-diameter telescope and the detection system consists of a 275 mm focal length spectrometer coupled to a 512 channel photodiode array detector. The acquired fluorescence spectrum covers the 300–800 nm spectral window and is achieved by combining a measurement of the 300–600 nm range with a measurement of the 500–800 nm range to avoid second order superposition. The detailed description of the system can be found in e.g. Cecchi et al. (1992, 1994).

158

2.3

Experimental setup

The measurements were performed during two sessions carried out in the frame of a EU-funded Culture 2000 project at the Roman archaeological site of Tropaeum Traiani (Romania). The first session was devoted to the acquisition of fluorescence spectra on stones affected by several biodeteriogens; the second was devoted to the measurements of autofluorescence from the pure cultures of fungal and bacterial strains. Both measurement sessions were carried out with the CNR-IFAC mobile lidar system. The mobile lidar system was deployed at a distance of about 25 m from the sample holder where samples were placed before each measurement. The sample holder was covered with non-fluorescent material to avoid spurious contributions to the collected fluorescence signal. The measurements were performed on the in vivo samples outdoors, in full sunlight, to reproduce as much as possible the experimental conditions during a lidar field campaign. In these experimental conditions the area of the target actually measured at each laser shot was a spot of about 2 cm diameter. Full-sunlight operation was possible by applying an electronic temporal gating (of the order of µsec) to the detector, in order to increase the signal-to-background ratio. In addition, a background spectrum was automatically acquired between each laser pulse and the following one and subtracted from the laser-excited spectrum to remove the possible residual background.

Figure 3. Autofluorescence spectra on stones affected by biodeteriogens. Fluorescence spectrum from the front and rear part of the stone (sample A): the rear part shows the typical fluorescence peak of phycocyanin at 660 nm.

Figure 4. Autofluorescence spectra on three different samples of stones affected by lichen growth.

3 3.1

RESULTS Lichens on stone substrates

Laser induced autofluorescence spectra of the examined biodeteriogens on stone substrates are shown in Figures 3–5. Figure 3 shows the fluorescence spectra of the front and rear section of the same stone: the fluorescence spectrum from the rear part of the stone shows a spectral profile typical of several cyanobacteria (see, e.g. Lognoli et al. 2002) characterised by the fluorescence peak of phycocyanin at about 660 nm. This leads to infer the presence of cyanobacteria on the rear surface of the stone, as it was also supported afterwards by visual and optical microscopy inspection of the rear surface of the stone by the biologists. Figure 4 shows the fluorescence spectra of different samples of stone with lichens on their surface (samples B, C and D; refer to Table 1 for their description). In particular, the areas effectively examined on the samples refer to: – sample B: area affected by Parmelia sp. (greyish lichen); – sample C: area affected by Caloplaca sp. (orange lichen);

Figure 5. Autofluorescence spectra on a stone affected by lichen growth (sample B) with the typical fluorescence peak due to Chl a at 680 nm after scratching the lichen surface.

– sample D: area affected by a white encrustation, probably due to by-products of the lichen acids. All the three samples show a typical fluorescence shape which is remarkably different from each other and could be used for their detection and mapping of the surface of a monument. Sample B shows the typical fluorescence peak of Chlorophyll a (Chl a) at about

159

x 1012

x 1011 F1 F2 F-blank

12

4. 5 4 Intensity (a.u.)

10 Intensity (a.u.)

B1-a B1-b B2-a B2-b B-blank

5

8 6

3. 5 3 2. 5 2 1. 5 1

4

0. 5 0

2

350

400

450

500

550

600

650

700

Wavelength (nm)

350

400

450

500

550

600

650

Figure 7. Autofluorescence spectra of pure cultures of four different bacterial strains (B1-a, B1-b, B2-a, B2-b) and of the relevant nutrient medium (B-blank).

700

Wavelength (nm)

Figure 6. Autofluorescence spectra of pure cultures of two different fungal strains (F1, F2) and of the relevant nutrient medium (F-blank).

3.2

Pure cultures of fungal and bacterial strains

The second session of measurements was devoted to the acquisition of autofluorescence spectra on pure cultures of fungal and bacterial strains, as listed in Table 2. Several measurements were peformed for each culture. Figures 6–7 show the autofluorescence spectra for the examined fungal and bacterial strains, respectively. The spectra shown in the figures are obtained by operating a mean over the measurements performed on the

F1/F-blank F1/F-blank F1/F-blank F-blank'/F-blank F-blank/F-blank

5

4 Ratio

680 nm, although not very intense. Sample C features a very intense fluorescence band at 650 nm, while sample D has a sharp-rising fluorescence band peaked at 450 nm which could be due to calcium oxalates. Similar fluorescence spectra due to calcium oxalates in lichens, excited at 250 nm, are reported for example in Clark et al. (2001). Figure 5 shows the fluorescence spectrum obtained on the same sample B (whose spectrum is already shown in Fig. 4) after scratching the surface of the lichen: after the operation, the Chl a fluorescence at 680 nm is definitely more apparent, both because of the partial removal of the crust responsible for the fluorescence between 500–600 nm and because of a more efficient excitation of the algal layer partially covered by the metabolites crust and the hyphal network on the surface of the colonies. Some by-products have in fact the effect to protect the algae from UV radiation and effectively decrease the amount of radiation reaching the algal layer (Fernandez et al. 1996, Clark et al. 2001).

6

3

2

1

0 350

400

450

500

550

600

650

700

Wavelength (nm)

Figure 8. The ratio between three spectra relative to the F1 sample and the F-blank spectrum. The ratio between another blank measurement (F-blank’) and the F-blank spectrum is reported as well.

same sample. The fluorescence spectra of the relevant nutrient media are also shown. Apart from some strain (e.g. B1-b bacterial sample) which features a noticeably different spectral shape, most fluorescence spectra are quite similar to the fluorescence spectrum of the nutrient medium. However, a preliminary analysis based on a very simple data processing, such as the ratio between the fluorescence spectra obtained on the strain and the relevant nutrient medium, reveals differences in the fluorescence contributions due to the presence of the bacterial or fungal strains. As an example, Figures 8–9 show the ratios between the fluorescence spectra obtained on the fungal strains

160

10

6

B-blank'/B-blank B1-a/B-blank B1-b/B-blank B2-a/B-blank B2-b/B-blank

F2/F-blank 5

8

F2/F-blank F2/F-blank

6

F-blank'/F-blank Ratio

4 Ratio

F-blank/F-blank 3

4

2

2

0

1 350

400

450

500

550

600

650

700

Wavelength (nm)

0 350

400

450

500

550

600

650

700

Wavelength (nm)

Figure 9. The ratio between three spectra relative to the F2 sample and the F-blank spectrum. The ratio between another blank measurement (F-blank’) and the F-blank spectrum is reported as well.

and the fluorescence spectrum of the relevant nutrient medium. Data in both graphs are shown with the same X–Y axis scale to make the comparison easier. Noise at higher wavelengths (>650 nm) is due to very low intensities in the fluorescence spectra. While the sample F1 does not show significant different contributions to the fluorescence spectral shape with respect to the fluorescence of the nutrient medium (Fig. 8), the ratios between the fluorescence spectra from the F2 sample and the nutrient medium point out differences in the fluorescence contributions around 400 nm and 600 nm (Fig. 9). Similarly, Figure 10 shows the ratios between the fluorescence spectra obtained on the bacterial strains and the fluorescence spectrum of the relevant nutrient medium. For comparison, the graph also shows the ratio between the fluorescence spectrum of the only nutrient medium (B-blank) and another measurement (B-blank’) taken on the same nutrient medium at the end of the set of measurements on the bacterial strains. From Figure 10 it can be inferred that all the examined bacterial strains show peculiar contributions to the fluorescence spectral shape in the 400–600 nm spectral range with respect to the nutrient medium (B-blank’/B-blank). In addition, the graph shows how the ratios referring to B2-a and B2-b strains have a very similar spectral shape, making them not distinguishable from each other. These different contributions to the fluorescence spectral shape can be exploited, by using more refined processing methods like multivariate analysis, to characterise different strains. Fluorescence data were also analysed with multivariate statistical techniques, specifically Principal

Figure 10. The ratio between the fluorescence spectra of the bacterial strains and that of the nutrient medium (B-blank). The ratio between another blank measurement (B-blank’) and the B-blank spectrum is reported as well.

Component Analysis and Cluster Analysis (see, e.g. Rencher 2002), to investigate the possibility to distinguish the fluorescence of the fungal and bacterial strains from that of the relevant nutrient medium as well as to distinguish one strain from the other one. The results, presented in a separate publication (Raimondi et al. 2007), allowed to characterise all the examined strains on the basis of their fluorescence features, except for the Aureobasidium pullulans that did not allow its discrimination from the nutrient medium. The proposed data processing techniques allow to differentiate among genera and even strains opening good prospects for the differentiation of heterotrophic organisms with remote sensing techniques in the field, at least when the fluorescence background, due to e.g. the stone substrate, has a known, relatively homogeneous spectral shape as in the case of the nutrient medium.

4

CONCLUSIONS

Remote laser-induced autofluorescence measurements have been peformed on both photoautotrophic and heterotrophic biodeteriogen samples selected in an archaeological site with a high spectral resolution fluorescence lidar from a 25 m distance in full sunlight, under uncontrolled environmental conditions. Most examined photoautotrophic and heterotrophic biodeteriogen samples showed peculiar fluorescence features that can be exploited for their characterisation during on-site remote non-invasive monitoring of monuments outdoors. In particular, the examined pure cultures of fungal and bacteria strains showed fluorescence features that allow their characterisation. This opens good prospects for the remote fluorescence

161

mapping of heterotrophic organisms on outdoor monuments. Remote fluorescence imaging-based documentation and diagnostics could actually be exploited to provide a preliminary, global assessment of the extension and typology of biodeterioration on the vast stone cultural heritage with the aim to plan at its best cleaning, maintenance and preventive conservation interventions of large-scale objects. ACKNOWLEDGEMENTS The authors want to thank all people that made it possible the realisation of this experiment. In particular, they are grateful to Roxana Radvan and Roxana Savastru for their helpful support all through the measurement campaign. They acknowledge the EU-funded Culture 2000 Project (Contract No. CLT 2003 A1 RO 515) for funding this experiment. REFERENCES Arcangeli, C. et al. 1997. Fluorescence study on whole Antarctic fungal spores under enhanced UV irradiation. Journal of Photochemistry and Photobiology B: Biology 39: 258–264. Bazzani, M. et al. 1992. Phytoplankton Monitoring by Laser Induced Fluorescence. EARSeLAdvances in Remote Sensing 1: 106–110. Bengtsson, M. et al. 2005. Fungus covered insulator materials studied with laser-induced fluorescence and principal component analysis. Applied Spectroscopy 39: 1037–1041. Cecchi, G. et al. 1992. FLIDAR: a multipurpose fluorosensorspectrometer. EARSeL Advances in Remote Sensing 1: 72–78. Cecchi, G. et al., 1994. Remote sensing of chlorophyll a fluorescence of vegetation canopies: I. near and far field techniquesRemote Sensing of Environment 47: 18–28.

Clark, B.M. et al. 2001. Characterization of mycobiont adaptations in the foliose lichen Xanthoparmelia chlorochroa (Parmeliaceae). American Journal of Botany 88: 1742– 1749. Colao, F. et al. 2005. Compact scanning lidar fluorosensor for investigations of biodegradation on ancient paited surfaces. Journal of optoelectronics and advanced materials 7: 3197–3208. Fernandez, E. et al. 1996. Photoprotector capacity of lichen metabolites against UV-A and UV-B radiation. Cosmetics Toiletries 111: 69–74. Hidalgo, M.E. et al. 2002. Photophysical, photochemical, and thermodynamic properties of shikimic acid derivatives: calycin and rhizocarpic acid (lichens). Journal of Photochemistry and Photobiology B: Biology 66: 213–217. Lognoli, D. et al. 2002. Detection and characterization of biodeteriogens on stone cultural heritage by fluorescence lidar. Applied Optics 41: 1780–1787. Lognoli, D. et al. 2003. Fluorescence lidar imaging of the cathedral and baptistery of Parma. Applied Physics B. 76: 457–465. Mathey,A. et al. 2001. Spatial distribution of perylenequinones in lichens and extended quinones in quincyte using confocal fluorescence microscopy. Micron 32: 107–113. Raimondi, V. et al. 1998. Fluorescence lidar monitoring of historic buildings. Applied Optics 37: 1089–1098. Raimondi, V. et al. 2007. Remote detection of laser-induced autofluorescence on pure cultures of fungal and bacterial strains and their analysis with multivariate techniques. Optics Communications 273: 219–225. Rencher, A. C. 2002. Methods of Multivariate Analysis. New York: Wiley Interscience. Svanberg, S. 2005. Fluorescence imaging of lidar targets. InT. Fujii & T. Fukuchi (eds), Laser Remote Sensing: 433–467. Boca Raton: CRC Press. Weibring, P. et al. 2001. Fluorescence lidar imaging of historical monuments. Applied Optics 40: 6111–6120. Yentsch, C.S. et al. 1979. Fluorescence spectral signature: the characterization of phyotoplankton populations by the use of excitation and emission spectra. Journal of Marine Research 37: 471–483.

162

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Design and development of a new high speed performance fluorescence imaging lidar for the diagnostics of indoor and outdoor cultural heritage V. Raimondi, L. Palombi, D. Lognoli & G. Cecchi Institute for Applied Physics ‘Nello Carrara’ – National Research Council, Sesto Fiorentino, Florence, Italy

L. Masotti Electronics and Communications Department, University of Florence, Italy

ABSTRACT: Fluorescence lidar imaging is a remote sensing technique that can be used to obtain hyperspectral fluorescence images on a monument from a distance by using a low-energy laser beam. The technique represents a particular economic advantage (in terms of time, money and specialised personnel) for an extensive monitoring of the outdoor stone cultural heritage. In addition, fluorescence lidar imaging data can also be presented in an easy-to-read output format, like false-colour thematic maps, which can become an effective support for restorers, historians and decision-makers. This paper presents the main technical features of a new fluorescence imaging lidar system specifically developed for the remote, non-invasive diagnostics on the cultural heritage by CNR-IFAC in collaboration with a private company, the El.En. Group S.p.A., in the frame of the SIDART National-Funded Project.

1

INTRODUCTION

The fluorescence lidar technique has been applied to the investigation of the cultural heritage since the mid 1990s (Raimondi et al. 1995). Since then, several experiments have proved it as a useful tool for the remote non invasive diagnostics of monuments, providing helpful information for the assessment of the state of conservation of monuments and for the characterisation of masonry materials. Up to now, main investigated issues include the detection and characterisation of different stones, mortars and other masonry materials (Raimondi et al. 1995, Cecchi et al. 1996b, Raimondi et al. 1998, Cecchi et al. 2000), of protective treatments (Ballerini et al. 2001), of frescoes (Colao et al. 2005), of biodeteriogens (Cecchi et al. 1996a, Lognoli et al. 2002, Colao et al. 2005 & Raimondi et al. 2007) and the analysis of the effects of biocide treatments (Lognoli et al. 2002). A step forward in the application of the fluorescence lidar technique to the monitoring of the cultural heritage has been achieved in 1997 with the use of lidar hyperspectral imaging during an experiment carried out on the Cathedral of Lund, Sweden (Weibring

et al. 2001). The results showed good potential for achieving thematic maps aimed at the characterisation of the lithotypes and the detection and characterisation of biodeteriogens. In the following years hyperspectral fluorescence lidar imaging has been applied on several other monuments, such as: the Cathedral and Baptistery of Parma (Lognoli et al. 2003), the roman archaeological site of Adamclisi in Romania (Cecchi et al. 2004) and the Coliseum and the Baptistery of San Giovanni in Laterano in Rome (Hällstrom et al. 2007). All these experiments were conducted by using multipurpose fluorescence lidar sensors, specifically the one developed at CNR-IFAC (Cecchi et al. 1992), the other developed at the Lund Institute of Technology, Sweden (Weibring et al. 2003). Both lidars were initially developed for different applications, such as atmospheric studies and sea, natural waters and vegetation monitoring, and were then applied to the diagnostics on the cultural heritage. A compact scanning lidar fluorosensor has been also developed for indoor operation on artworks and especially frescoes inside tombs where only a limited space is available (Colao et al. 2005). This prototype has been utilized to investigate painted walls of a Byzantine crypt in Constanta (Romania).

163

This paper introduces a new fluorescence lidar sensor, specifically developed for cultural heritage applications, which has been built at the CNR-IFAC in collaboration with a private company, the El.En. Group S.p.A.. Design criteria have taken into account the following issues:

Figure 1. Block diagram of a fluorescence lidar system.

n-columns

m-rows

These characteristics make the sensor able to operate at different sites, e.g. outdoor as well as indoor, and also aboard a small-size van. The sensor is provided with a computer-controlled target scanning system and with a pointing laser system in the visible to reference the acquired hyper-spectral images on the target. The following sections briefly describe the hyperspectral fluorescence lidar imaging technique and the main features of the new prototype, specifically designed for cultural heritage applications, together with the results of a demonstrative test in the laboratory.

Intensity

– High scanning speed of the target for a quick image acquisition also on large areas; – Relatively high spatial resolution capabilities for the investigation of monument details; – High spectral resolution for a thorough spectral shape analysis by refined data processing techniques, such as multivariate statistical analysis; – Wide field of view and reduced minimum operational distance to enhance the sensor’s operational capabilities; – Compactness and on-site transportability.

Laser spot

(a) Scanning of the target to obtain an image with (mxn) pixels

Wavelength

(b) Detection of a high resolution fluorescence spectrum for each image pixel

(c) Realisation of thematic maps by analysing the fluorescence spectra with specific processing techniques

Figure 2. Operating principles of hyperspectral fluorescence lidar imaging.

2

HYPER-SPECTRAL FLUORESCENCE LIDAR IMAGING

A fluorescence lidar is essentially composed of a laser, a telescope, a dispersion system and a detector (Fig. 1). The laser beam, which can also be collimated by using a beam expander, is sent to the target and interacts with its constituents. The backscattered signal, including the fluorescence emitted by the target and containing information about its chemical-physical characteristics, is collected by the telescope and then fed to the dispersion and detection system, usually featuring high spectral resolution. The data are finally stored in a PC for the analysis of the signal. Hyper-spectral fluorescence lidar imaging technique essentially exploits a computer-controller scanning system to operate a scan of the target and thus acquire a complete spectrum for each scan position. In this way, at the end of the scan, a hyper-spectral fluorescence image of the investigated area can be retrieved. The measurement process is synthetically sketched in Figure 2; the investigated area on the surface of the monument can be ideally divided into m × n ‘squares’ (Fig. 2a) and then scanned with the lidar system to obtain a fluorescence image of the target. For

each ‘square’ of the target image, or ‘image pixel’, a full high spectral resolution fluorescence spectrum is acquired (Fig. 2b). The m × n set of fluorescence spectra is finally processed to obtain thematic maps that outline specific fluorescence features of the target (Fig. 3c). Fluorescence-based thematic maps are particularly attractive for the monitoring of monuments: firstly, they provide a comprehensive assessment on the status of the whole monument and a spatial definition that cannot be obtained by means of mere sampling. Moreover, the opportunity of recording time-dependent, repetitive fluorescence images opens new prospects for reliable monitoring, repeated in time, of the status changes of the monument. Another important aspect of thematic maps is that they make it easier to transfer information gained with sophisticated data processing to the conservation specialist to sustain the action of the decision maker. The hyper-spectral image analysis and then the thematic maps can be achieved in different ways, such as calculating a ratio between two selected spectral bands of the fluorescence spectra, applying Principal

164

Figure 3. Layout of the lidar arrangement. Figure 4. The fluorescence lidar prototype during operation.

Component Analysis (PCA) or Cluster Analysis (CA) methods to the fluorescence data set (Rencher 2002). The results can be plotted as a function of the corresponding (x, y) position in a false-colour coded map.

Table 1.

3 THE NEW PROTOTYPE OF FLUORESCENCE IMAGING LIDAR 3.1

New lidar sensor main features.

Features

Description

Excitation laser source Optical reference system laser source Detector

3ω Nd:YAG, Q-Switched (@ 355 nm) 8 mJ @ 16 Hz or 40 mJ @ 50 Hz CW Laser @ 532 nm, 5 mW (IIIA class)

General description

A general layout of the new prototype is showed in Figure 3. The sensor has a UV laser for target fluorescence excitation and a laser emitting in the visible to reference the fluorescence image to the target. The two laser beams are coaxial to the optical axis of a telescope. A movable folding mirror is used as a pointing and scanning system to send the laser beam on the target and to scan the area. The telescope provides the collection of the backscattered radiation from the target. The collected radiation is focused on the entrance of an optical fibres bundle. The exit of the fibre bundle is mechanically coupled to the entrance slit of a spectrometer coupled to a 512 × 512 pixel matrix detector. The whole system is managed by a personal computer. The lidar sensor is arranged within an aluminium frame. The laser sources and relevant conditioning optics are placed on an aluminium plate, while the laser folding mirrors, the telescope and the optical fibre bundle support are fixed under the plate. The pointing and scanning system is also fixed on the aluminium frame. The dimension of the lidar sensor are 250 cm × 85 cm × 75 cm (l × h × w) with a weight of about 150 kg. The sensor has been designed so as to allow the transport aboard a small size van and its operation from the side door. Figure 4 shows a picture of the new lidar sensor during operation. Table 1 summarises the main features of the new lidar sensor.

Spectrometer Spectrometric linear resolution Telescope Telescope distance range Telescope Far field of view Pointing system Lidar Sensor Pointing field of view Pointing accuracy (both axes) Imaging single pixel acquisition time Weight and dimension

Intensified CCD 512 pixel × 512 pixel QE >10% (in 150 nm–870 nm range), Minimum gate width < 5 ns 300 mm focal length Three gratings (150 gg/mm, 600 gg/mm, 2400 gg/mm) 0.51 nm/pixel, 0.12 nm/pixel, 0.02 nm/pixel Newtonian Layout, 1 m focal length, 250 mm diameter 4 m up to inf 1 mrad Two axis motorized folding mirror Primary Axis: −30◦ up to +300◦ from the reference direction Secondary Axis: ±45◦ from the orthogonal direction to telescope axis 0.2 mrad 20

532 18.6

355 3

1064 7

532 5

355 0.9

−

>> 9

–

–

3.5

0.9

to the surface was observed with the naked eye. The sample was then removed and examined with a microscope to check if the damage indeed occurred and to determine exactly where. Then, the sample was placed again in the laser beam path in a position shifted by 5 to 10 mm from initial position, where damage was detected, and the sample was moved 1 mm each step in the direction of the beam waist plane in order to locate the exact position, and therefore exact laser fluence, where damage occurs. It has to be mentioned that every sample was also moved crosswise to the axis of laser beam in order to expose a fresh unprocessed sample surface to the laser pulse and, on the other hand, to average on the surface, thus minimising the influence of potential heterogeneity or the presence of surface cracks on the determined damage thresholds. Results of measurements are presented in Table 2. Results of laser cleaning are shown in Figures 4, 5. Soil clusters inside 0.1–1 mm slots were several times cleaned using focused laser beam, in all cases not exceeding a fluence of 4 J/cm2 . 4

Figure 4. Statue of St. John (a) and pedestal of St. Philip (b) after laser cleaning. The small photograph at the top shows the pedestal before restoration (Fig. 1).

DIAGNOSTICS

It is known, that encrustation is non-homogeneous across the whole object surface, and does not possess the same thickness, structure and even colour. One of the few physical parameters allowing description of encrustation characteristics is the average reflection-backscattering coefficient of white light (or laser light). Spectrometric measurement of the amplitude of the backscattered white light as a function of the wavelength represents synonymous and objective

Figure 5. View of the collection of the ivoryTwelveApostles after laser cleaning.

colorimetry, frequently used for fast and suitable determination of the cleaning level of different substrates, included ivory (e.g. Marczak 2001). The task of the fibre optics spectrometer shown in Figure 6 was to detect the amplitude of the

217

depending on, for example, the level of cleaning of the investigated surface (Fig. 7). 5

Figure 6. Scheme of the diagnosis system with fibre optics spectrograph for the investigation of reflection (scattering) coefficient of superficial layers.

CONCLUSIONS

The results of laser cleaning of the Twelve Apostles collection were surprisingly good. The use of a laser beam allowed to remove soil from almost all hard to get relief areas. The condition of the ivory surface after laser treatment has been tested using a standard microscope (50× magnification). Thanks to earlier measurements of ivory damage thresholds and careful adjustment of laser fluence, there were no observable destructions of the primary ivory substrate. It is difficult to compare the obtained results with the results of other methods of ivory cleaning, although comparative chemical tests with detergents and mechanical cleaning with the use of glass fibre stick were both arduous and time consuming. It should be stated that proposed laser technology is fast, ecological (no solvents) and safe for the object, if we take into consideration the problems connected with the hygroscopicity and anisotropy of ivory, signalised in the introduction of this paper. Another advantage is the absence of contact of the cleaning tool (laser beam) with the delicate object surface. ACKNOWLEDGEMENTS Work has been supported by the Ministry of Science and Higher Education, Poland, project 120/E-410/ SPB/EUREKA/KG/DWM 97/2005-2007.

Figure 7. Results of measurements of light amplitude scattered from ivory for different laser cleaning levels.

REFERENCES

backscattered light where cleaning was carried out. Light emitted by an halogen lamp (mercury, sodium lamp or even another laser) is delivered to the cleaned surface of the object by the central optical fibre. Backscattered light is collected by six other fibres surrounding the central one. Collected light is then transmitted through the optical system to the diffraction grating of the spectrometer and, after dispersion, to the linear matrix of CCD detectors. Finally, it is displayed on the monitor of a computer. The distance of the measurement tip from the examined surface was selected in such a way to obtain a maximum of scattered light for ivory fracture. In the present case, fresh ivory determined a reference for comparison to other values of light scattering

Landucci, F. et al. 2000. Laser cleaning of fossil vertebrates: a preliminary report. J. Cult. Heritage 1: S263–S267. Landucci, F. et al. 2003. Toward an optimized laser cleaning procedure to treat important paleontological specimens. J. Cult. Heritage 4: S106–S110. Madden, O. et al. 2003. Removal of dye-based ink stains from ivory: evaluation of cleaning results based on wavelength dependency and laser type. J. Cult. Heritage 4: S98-S105. Mann, E. O. & Espinoza, M. J. (ed.) 1992. Identification Guide for Ivory and Ivory Substitutes. 2nd edition. Baltimore: WWF Publications. Marczak, J. 2001. Surface cleaning of art work by UV, VIS and IR pulse laser radiation. Proceedings of SPIE 4402: 202–209. Ostrowski, R. et al. 2007. Laser damage thresholds of bone objects. Proc. SPIE 6618. In print. Strzelec, M. et al. 2005. Results of Nd:YAG laser renovation of decorative ivory jug. Springer Proceedings in Physics 100: 163–168.

218

Laser Cleaning of Paintings and Polychromes

Lasers in the Conservation of Artworks – Castillejo et al. (eds) © 2008 Taylor & Francis Group, London, ISBN 978-0-415-47596-9

Investigating the use of the Nd:YAG laser to clean ancient Egyptian polychrome artefacts C. Korenberg, M. Smirniou & K. Birkholzer The British Museum, London, United Kingdom

ABSTRACT: The aim of the present study was to investigate whether laser cleaning using a Nd:YAG laser would be a suitable technique to treat ancient Egyptian polychrome artifacts. Red ochre, realgar, calcite, gypsum, yellow ochre, orpiment, green frit, malachite and Egyptian blue pigments were mixed with gum arabic. Samples were prepared to investigate whether parameters such as the binder concentration or the nature of the substrate affect the damage threshold of the paints. For most of these paints, none of these factors was observed to have a significant effect. The ablation thresholds of lamp black and five consolidants were measured and found to be higher than the damage thresholds of most paints. Also, when removing lamp black, yellowing was observed on some of the substrates studied. It was concluded that laser cleaning at these laser parameters would not be a suitable method to remove consolidants or soot from ancient Egyptian polychrome artefacts.

1

INTRODUCTION

The British Museum holds an extensive collection of ancient Egyptian polychrome coffins and wall paintings. These artifacts usually have a surface that is absorbent and vulnerable, attracting dust particles into the porous painted layer. Deposits on wall paintings and painted coffins can include: dust, soot, salts, pollutants, old conservation materials and residues from previous restorations. Conservators employ different methods to clean painted surfaces, depending on the physical structure and stability of the pictorial layers and the composition and thickness of the deposits to be removed. Dry erasers and sponges, such as vulcanised latex or smoke sponges of vulcanised natural rubber, are used to remove dry dust and loose soot or grit, but they are not appropriate for heavy deposits. Mechanical cleaning using scalpels or dental tools is a common way to remove solid encrustations and salts. However, this can cause scratches and surface damage and the outcome depends on the skills and experience of the conservator. Wet cleaning is used to remove soot, grime, soluble salts or organic deposits from surfaces that tolerate wet treatment. This technique is not entirely satisfactory as applying a solution may redeposit dirt instead of removing it. Many polychrome Egyptian artifacts in the British Museum are very difficult to clean using the conventional methods mentioned above as these risk damaging the delicate pictorial layers. These artifacts include an Egyptian coffinAES 6690, which is covered

Figure 1. Fragment of the Nebamun mural, EA 37976, on which a past conservation coating has become dark and unsightly.

by soot from fire damage, and a fragment of the Nebamun mural, EA 37976, on which a past conservation coating has become dark and unsightly (Fig. 1). As reviewed by Fotakis et al. (2007), there have been several case studies reporting successful laser cleaning of painted surfaces covered with contaminants, such as soot or aged resins. However, laser irradiation risks altering the colour of certain paints due to phase changes or decomposition reactions (Chappe et al. 2003, Pouli et al. 2003, Sansonetti & Realini 2003) and it is necessary to assess the effect of laser irradiation on every paint on which it is to be used. The effect of laser irradiation on ancient Egyptian paints has not been previously studied and the

221

aim of the present research was to investigate whether laser cleaning using a Q-switched Nd:YAG laser would be suitable for cleaning these artifacts. As binder concentration in ancient Egyptian paints is probably variable and the paints were diluted to different extents and applied on different substrates, it was first investigated whether these factors have an effect on the damage thresholds of the paints. Further, the ablation thresholds of lamp black and five consolidants that are currently used or were used in the past to treat painted surfaces were determined to assess whether such contaminants could be removed safely from ancient Egyptian polychrome artifacts using laser light. Some samples were also thermally aged to investigate the effect of ageing on the values of the damage thresholds of the paints and ablation thresholds of the consolidants. The results are presented and discussed below.

2 2.1

EXPERIMENTAL Paint samples

This project is focussed on the main paints used by the ancient Egyptians before the Roman period and the following pigments were selected for study: red ochre, realgar, calcite, gypsum, yellow ochre, orpiment, ‘green frit’ and Egyptian blue. Although malachite was rarely used as a pigment by the ancient Egyptians, it was included in this study as it is present on polychrome artifacts as a degradation by-product of green frit and Egyptian blue (Lee & Quirke 2000). All the pigments except green frit were purchased from Kremer and Cornelissen artists materials suppliers; green frit was synthesised in the laboratory following the method published by Pages-Camagna & Colinart (2003). The pigments were analysed prior to use using Raman spectroscopy to confirm their composition and the spectra obtained were found to match the reference spectra of the corresponding pigments. Some pararealgar was detected in the realgar pigment, probably formed upon exposure to light during the manufacturing process -realgar is known to transform into pararealgar when exposed to light (Douglass et al. 1992). When investigating the laser cleaning of painted artifacts, it is necessary to conduct tests on paints made with the same binder as the nature of the binder has an effect on the damage threshold of paints (Hildenhagen et al. 2005, Schnell et al. 2005). Gum arabic dissolved in water was used in the present work as this was the binder the most widely used by the ancient Egyptians (Newman & Serpico 2000). It should be noted that the effect of the binder on the damage threshold of paints is complex. It has been shown that there is no common principle that allows predictions to be

made for the damage threshold of a paint made with a specific binder and pigment (Hildenhagen et al. 2005). Paints were prepared with different concentrations of gum arabic to determine the effect of the binder concentration on the damage threshold. The paints were also diluted with increasing amounts of water to investigate whether thin layers of paint have a different damage threshold than thicker layers. 2.2 Substrates The nature of the substrate has been reported to affect the damage threshold of paints and this is thought to be due to the different thermal conductivity and diffusivity of different substrates (Gordon Sobott et al. 2003). In the present study, different substrates representative of Egyptian artifacts were used. Paints on ancient Egyptians artifacts were generally not applied directly on stone or on wood, but on a ground of gypsum or crushed calcite (Lee & Quirke 2000, Middleton 1999, Middleton 2000). Unlike gypsum, calcite cannot be used on its own as a ground and it was probably mixed with a binder. To date, no information is available on the nature of the binder in calcite grounds of ancient Egyptian artifacts and in the present study gum arabic was used to make calcite and gypsum grounds. Thin layers of paints were applied on slabs of gypsum, cardboard covered with gypsum mixed with gum arabic and cardboard covered with calcite mixed with gum arabic. To investigate the effect of the laser irradiation on the paints with a negligible contribution from the substrate, thick and opaque layers of paints made with a low binder concentration were also applied on cardboard. 2.3 Materials to be removed The removal of lamp black and five consolidants was investigated. The consolidants selected for the present project were animal glue, soluble nylon (N-methoxymethyl nylon), microcrystalline wax, Primal AC33 (ethylacrylate methyl methacrylate copolymer), Paraloid B72 (ethyl methacrylate copolymer) and Mowital B30H (polyvinyl butyral polymer). Animal glue, soluble nylon and microcrystalline wax were commonly used in the middle of the past century for consolidating flaking paint. Since then, it has been shown that these materials are not suitable for long-term conservation treatments and should be removed from artifacts. For example, soluble nylon yellows upon ageing (Sease 1981), while animal glue discolours, shrinks and becomes brittle resulting in potential damage to paint layers. Upon ageing these consolidants tend to become very difficult to remove using conventional techniques. Primal AC33, Paraloid B72 and Mowital B30H are currently used in conservation and, as it may be desirable to remove them in some instances, it was

222

assessed whether laser cleaning would be a suitable technique. Lamp black was applied on a slab of gypsum, cardboard covered with gypsum mixed with gum arabic and cardboard covered with calcite mixed with gum arabic, while the consolidants were applied on slabs of gypsum. Paraloid B72 was prepared as a 2.5% (by weight) solution diluted in a 1:1 acetone/IMS (industrial methylated spirits) mixture, Primal AC33 as a 5% (by weight) dispersion in water and Mowital B30H as a 3% (by weight) solution in a 1:1 acetone/IMS mixture.

Table 1. Damage threshold at 1064 nm of the red ochre paint prepared with increasing amounts of gum arabic solution and diluted to different extent. The paints were applied on cardboard strips covered with either gypsum or calcite grounds.

2.4 Ageing tests To assess the effect of thermal ageing on the values of damage and ablation threshold, sets of samples of paints and consolidants were aged at 60◦ C and 70◦ C for 28 days in a dessicator in which the relative humidity was controlled at approximately 50% using a glycerol solution. 2.5 Laser cleaning tests Tests were conducted using a Lynton Phoenix Q-switched Nd:YAG laser emitting 5–10 ns pulses at wavelengths 1064 nm and 532 nm. The average fluence was calculated by dividing the energy per pulse by the area of the laser spot. The size of the laser spot was estimated by taking a burn pattern on a photographic paper. Results were assessed visually, using a magnifier and using an optical microscope. 3

EFFECT OF THE BINDER CONCENTRATION, DEGREE OF PAINT DILUTION, SUBSTRATE AND AGEING

Composition of the paint

Gypsum J/cm2

Calcite J/cm2

0.1 g in 0.2 mL solution 0.1 g in 0.4 mL solution 0.1 g in 0.8 mL solution 0.1 g in 1.6 mL solution 0.1 g in 3.2 mL solution 0.1 g in 6.4 mL solution Same as above, but diluted Same as above, but diluted Same as above, but diluted

0.24 0.24 0.20 0.27 0.16 0.14 0.33 0.32 0.22

0.34 0.34 0.30 0.24 0.23 0.24 0.32 0.21 0.16

Table 2. Damage thresholds of the paints applied on different substrates at 1064 nm. The substrates employed were A: slabs of gypsum, B: cardboard covered with gypsum mixed with gum arabic, C: cardboard covered with calcite mixed with gum arabic and D: cardboard. Damage threshold at 1064 nm J/cm2 Paints

A

B

C

D

Red ochre Realgar Yellow ochre Orpiment Egyptian blue Malachite

0.45 * 0.46 0.42 0.71 0.25

0.24 * 0.30 0.23 0.64 0.23

0.26 * 0.44 * 0.40 0.15

0.21 * 0.24 * 0.60 0.32

* Alteration was noted at the lowest fluence allowed by the set up (approximately 0.14 J/cm2 ).

3.1 Effect of the binder concentration and degree of paint dilution As illustrated in Table 1 for red ochre, there was no discernible trend in the value of the damage threshold when the paints were prepared with various concentrations of gum arabic and diluted to different concentrations. Similar results were obtained for the other paints. This suggests that the binder concentration and degree of dilution of the paint had no detectable effect on the damage threshold. It was noted that there was a large variability in the value of the damage threshold of the paints. This was thought to be due to a certain extent to the heterogeneity of the paints which were mixed and applied by hand, but also to the laser itself. The fluence of the Nd:YAG laser is not uniform over the area of the beam with some hot spots present in the beam and damage to the paints seemed to occur at the location of the hot spots. The performance of the employed laser is optimized by

the manufacturer at the maximum energy output such that the beam quality is relatively poor at low energy levels. Since the energy used in the experiments was very low for most of the paints, it is suspected that at a given average fluence, the local fluence at the location of the hot spots was quite variable from pulse to pulse, contributing to the observed variability of the damage threshold. 3.2 Effect of the nature of the substrate The values of the damage threshold of the paints applied on different substrates are given in Tables 2 and 3. These values varied between different substrates, but the variations were within the experimental variations noted earlier and it was concluded that the nature of the substrate had no detectable effect on the damage thresholds of the paints.

223

Table 3. Damage thresholds of the paints applied on different substrates at 532 nm. The substrates employed were A: slabs of gypsum, B: cardboard covered with gypsum mixed with gum arabic, C: cardboard covered with calcite mixed with gum arabic and D: cardboard.

Table 4. Damage thresholds of unaged and thermally-aged paints applied on cardboard. (Ageing was conducted at 60◦ C.) 1064 nm J/cm2

532 nm J/cm2

Damage threshold at 532 nm J/cm2

Paints

Unaged

Aged

Unaged

Aged

Paints

A

B

C

D

Red ochre Realgar Yellow ochre Orpiment Egyptian blue Malachite

0.04 0.09 0.06 * 0.27 0.23

* 0.03 0.04 0.06 0.39 0.10

* 0.05 * 0.06 0.43 0.13

* 0.03 0.04 0.03 0.25 0.12

Red ochre Realgar Yellow ochre Orpiment Malachite

0.21 * 0.24 * 0.32

0.16 * 0.24 * 0.34

0.02 0.03 0.04 0.03 0.12

0.02 0.02 0.04 0.03 0.20

* Alteration was noted at the lowest fluence allowed by the set up (approximately 0.01 J/cm2 ).

However, it was noticed that discolouration was perceived more easily on some substrates for some of the paints and this may affect the experimental determination of the damage threshold. For instance, orpiment tends to become paler when irradiated at 1064 nm and it was easier to perceive fading on the white calcite ground than on the off white gypsum ground. Also, for some paints, such as Egyptian blue, laser irradiation caused the removal of pigment particles rather than discolouration and this was in general more easily perceived on the thick paint layers applied on cardboard than on the thin layers applied on the other substrates. This highlights the limitations of visual assessment and future work should focus on developing methods to assess laser-induced damage on paints in a systematic way.

3.3 Effect of thermal ageing To investigate the effect of thermal ageing on the damage thresholds of the paints, two sets of samples made with the same concentrations of gum arabic and diluted to the same extents were applied on cardboard and one set of samples was aged at 60◦ C for 28 days. The appearance of the paints did not change upon ageing. The values of the damage thresholds for the unaged and aged paints are shown in Table 4. No significant change in the damage threshold was observed. To further investigate the effect of thermal ageing, two additional sets of paint samples were prepared for ageing at 70◦ C. Since cardboard may be affected at this higher temperature, the paints were applied on gypsum slabs. The Egyptian blue and malachite paints acquired a brown tinge upon ageing. Egyptian blue paints on ancient artifacts are often found to have become brown with time and this has been attributed to the browning of gum arabic combined with the poor hiding power

* Alteration was noted at the lowest fluence allowed by the set up at 1064 nm (approximately 0.14 J/cm2 ).

Table 5. Damage thresholds of unaged and artificially aged paints applied on gypsum slabs. (Ageing was conducted at 70◦ C.) 1064 nm J/cm2

532 nm J/cm2

Paints

Unaged

Aged

Unaged

Aged

Red ochre Realgar Yellow ochre Orpiment Egyptian blue Malachite

0.45 * 0.46 0.42 0.71 0.25

0.34 * 0.28 0.29 0.68 0.29

0.04 0.09 0.06 0.01 0.27 0.23

0.02 0.04 0.04 0.02 0.28 0.21

* Alteration was noted at the lowest fluence allowed by the set up (approximately 0.14 J/cm2 ).

and transparency of the pigment (Daniels et al. 2004). It is likely that the same factors are responsible for the browning of the malachite paint observed here. The appearance of the other paints did not change upon ageing. The values of the damage thresholds for the unaged and aged paints are shown in Table 5. The damage threshold values were observed to decrease after ageing for some of the paints, however these changes were within the experimental variations. From these two series of tests, it was concluded that thermal ageing did not affect the damage thresholds of the paints detectably.

4

EVALUATING THE SUITABILITY OF LASER CLEANING FOR EGYPTIAN POLYCHROME ARTEFACTS

4.1 Threshold values of the paints Table 6 compiles the values of the damage thresholds of the paints from all the tests. The damage thresholds are generally higher when using the 1064 nm wavelength than the 532 nm wavelength.This trend has been

224

Table 6. Damage threshold ranges of the paints measured at 1064 and 532 nm.

Table 8. Ablation thresholds of the thermally-aged consolidants.

Paints

1064 nm J/cm2

532 nm J/cm2

Consolidants

1064 nm J/cm2

532 nm J/cm2

Red ochre Realgar Yellow ochre Orpiment Egyptian blue Malachite Green frit* Calcite* Gypsum*

0.14–0.45