Solar Cells and Their Applications 1118024052, 9781118024058

Citation preview

SOLAR CELLS AND THEIR APPLICATIONS Second Edition Edited by LEWIS FRAAS LARRY PARTAIN

A JOHN WILEY & SONS, INC., PUBLICATION

SOLAR CELLS AND THEIR APPLICATIONS

WILEY SERIES IN MICROWAVE AND OPTICAL ENGINEERING KAI CHANG, Editor Texas A&M University A complete list of the titles in this series appears at the end of this volume.

SOLAR CELLS AND THEIR APPLICATIONS Second Edition Edited by LEWIS FRAAS LARRY PARTAIN

A JOHN WILEY & SONS, INC., PUBLICATION

Copyright © 2010 by John Wiley & Sons, Inc. All rights reserved Published by John Wiley & Sons, Inc., Hoboken, New Jersey Published simultaneously in Canada No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/permission. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com. Library of Congress Cataloging-in-Publication Data: Solar cells and their applications / [edited by] Lewis Fraas, Larry Partain.—2nd ed. p. cm.—(Wiley series in microwave and optical engineering) ISBN 978-0-470-44633-1 (cloth) 1. Solar cells. I. Partain, L. D. II. Fraas, Lewis M. TK2960.S652 2010 621.31'244—dc22 2010000196 Printed in Singapore 10 9 8 7 6 5 4 3 2 1

Contents Preface

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

ix

Contributors .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

xiii

INTRODUCTION TO SOLAR CELLS

.

.

.

.

.

1

.

3

.

17

Solar Cells, Single-Crystal Semiconductors, and High Efficiency . . . . . . . . . . Lewis Fraas

43

PART I

.

Chapter

1

Solar Cells: A Brief History and Introduction . Lewis Fraas and Larry Partain

Chapter

2

Solar Cell Electricity Market History, Public Policy, Projected Future, and Estimated Costs . Larry Partain and Lewis Fraas

Chapter

Chapter

3

.

.

.

67

PART II TERRESTRIAL SOLAR CELL ELECTRICITY

.

.

111

Chapter

5

Crystalline Silicon Solar Cells and Modules Leonid Rubin

.

.

113

Chapter

6

Thin-Film Solar Cells and Modules . Robert Birkmire

Chapter

7

Terrestrial Module Fabrication and Assembly Technologies . . . . Christopher Bunner

Chapter

4

8

Solar Cell Device Physics . Larry Partain

Chinese Solar Cell Status. Wang Sicheng

.

.

.

.

.

.

.

.

.

.

.

137

.

.

.

.

159

.

.

.

.

171 v

vi

Chapter

CONTENTS

Tracking the Sun for More Kilowatt Hour and Lower-Cost Solar Electricity . . . . . . Ron Corio, Michael Reed, and Lewis Fraas

.

207

Solar Cell Systems: Definition, Performance, and Reliability. . . . . . . . . . Jason Strauch, Larry Moore, and Elmer Collins

.

219

Levelized Cost of Energy for Utility-Scale Photovoltaics . . . . . . . . . Matthew Campbell

.

.

251

.

.

271

Chapter 12

Low-Concentration Crystalline Silicon Systems . Lewis Fraas

.

273

Chapter

High-Concentration, III–V Multijunction Solar Cells . . . . . . . . . . Geoffrey Kinsey

.

.

293

High-Concentration Fresnel Lens Assemblies and Systems . . . . . . . . . . Gerhard Peharz and Andreas Bett

.

313

High-Concentration Cassegrainian Solar Cell Modules and Arrays . . . . . . . . Michael Ludowise and Lewis Fraas

.

337

Concentrator Solar Cell Installations at the University of Nevada, Las Vegas . . . . Suresh Sadineni and Robert Boehm

.

.

361

Concentrator Photovoltaic Field Installations . Francisca Rubio, María Martínez, and Pedro Banda

.

377

Chapter

Chapter

9

10

11

PART III TERRESTRIAL CONCENTRATOR SOLAR CELL SYSTEMS . . . . . . . . .

Chapter

Chapter

Chapter

Chapter

PART IV Chapter

13

14

15

16

17

SOLAR CELLS IN SPACE . 18

.

.

.

.

.

.

.

395

Space Solar Cells and Applications . Sheila Bailey and Ryne Raffaelle

.

.

.

.

397

CONTENTS

vii

PART V

OTHER ASPECTS AND CONSIDERATIONS .

Chapter

19

Solar Resource for Space and Terrestrial Applications . . . . Christian A. Gueymard and Daryl Myers

.

.

.

.

425

.

.

427

.

463

Chapter

20

Solar Energy Costs: The Solar Advisor Model . Paul Gilman, Nathan Blair, and Christopher Cameron

Chapter

21

Challenges of Large-Scale Solar Cell Electricity Production . . . . David Faiman

.

.

.

.

483

.

.

.

.

497

Market Overview of Flat Panel Detectors for X-Ray Imaging . . . . . . . . . Carl LaCasce, Larry Partain, and Chuck Blouir

.

499

Amorphous Silicon Transistors and Photodiodes . . . . . Robert Street

PART VI THIN FILMS AND X-RAY IMAGER TECHNOLOGIES . . . . . . Chapter

22

Chapter

23

.

.

.

.

.

.

511

Chapter

24

Amorphous Silicon Digital X-Ray Imaging . Richard Colbeth

.

.

533

Chapter

25

Photoconductor Digital X-Ray Imaging. George Zentai

.

.

.

559

PART VII Chapter

Index .

SUMMARY 26

.

.

.

.

.

.

.

.

.

581

Summary, Conclusions, and Recommendations. . . . Lewis Fraas and Larry Partain

.

.

.

.

.

.

583

.

.

.

.

.

.

613

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Preface This Second Edition is intended to be a comprehensive survey, review, and analysis of all the major factors related to the continuing technical development of solar cell electricity and its market development into a major worldwide source of electric power in response to powerful political and economic influences. It is divided into six major sections plus a Summary section including conclusions and recommendations. In contrast to the First Edition, Part I contains three initial chapters written so that nonspecialists and the more general readers and investors and policy makers can follow their contents without the need for specialized training or understanding. The goal is to allow a broad spectrum of readers to at least comprehend the market history, the influence of public policy, the likely costs of solar cell-generated electricity, and the special role that near-perfect, single-crystal semiconductor fabrication materials can have on overall performance. Chapter 4 in this part, on Solar Cell Device Physics (like the First Edition), is again aimed at advanced undergraduate and graduate college courses and other technical professionals involved in teaching, research, and commercial development. It not only covers the traditional abrupt p/n junction configuration of the First Edition but also expands into the very non-abrupt p-i-n geometries that characterize a whole new class of high-performance solar cells including interdigitated back-contact cells, point-contact cells, and heterojunction-with-intrisinsic-thin-layer (HIT) cells. It further addresses the special resistive restrictions that can limit p-i-n-type device performance as well as proposed paths to performance levels well beyond 50% efficiency levels. However, to maintain a reasonable length, this physics chapter does use the First Edition as a reference. Part II addresses the current state of terrestrial solar cell electricity technology and development programs. This includes the dominant crystalline silicon abrupt p/n junction devices and their large-scale fabrication and the emerging thin-film amorphous and polycrystalline semiconductor cells and modules. The amazing recent growth of the Chinese terrestrial solar cell program is presented in some detail. The potential advantages of tracking the sun are explored along with a detailed description of 3 years of field experience with fixed-axis crystalline silicon modules of 12% efficiency (under standard test conditions) in the Arizona desert. The emerging utility-scale installations are summarized along with their important cost-determining characteristics. Part III attempts to present a comprehensive overview of the terrestrial concentrator approach to solar cell electricity production and its special advantages ix

x

PREFACE

and challenges. This includes both low and high sunlight concentration levels with various system approaches as well as early results of small field tests at the University of Nevada and of substantial utility-scale field tests of multiple and varied concentrator systems in Spain. Part IV takes a broad look at space systems and all of the unique approaches, needs, accomplishments, plans, and future needs for space. Part V contains a chapter giving precise descriptions of the solar resource both terrestrially and in space. It also contains a chapter describing a sophisticated and detailed cost and performance model from the National Renewable Energy Laboratory (NREL). This Solar Advisor Model is reviewed and summarized. Finally, the special challenges of large-scale solar electricity production are explored. Part VI is a special four-chapter addition of the Second Edition that discusses how thin-film solar cells can be transformed into X-ray imaging devices when devices are reduced to submillimeter sizes and are aligned in columns and rows that are covered by a scintillator film that converts X-ray photons into visible light photons. If these are then attached to an array of the thin-film transistor switches, a flat-plate X-ray imager is produced. The market analysis of this whole X-ray imager field shows that its current market size of $2 billion per year should continuously evolve into a $15 billion per year wholesale market over the next 10 years or so as these devices continually improve in performance and drop in price. The final chapter summarizes the amazing growth of this solar cell electricity technology and market over the 15 years since the publication of the First Edition. It provides recommendations for how major countries and unions can play major roles from both technology and public policy perspectives and how continuing cost reduction and improved performance demands should be met under both near- and medium-term time frames. In summary, this book describes today’s baseline planar solar cell power systems as well as innovations in high-efficiency solar cells and concentrated sunlight systems that have occurred in the last 15 years, which now promise lower cost electricity competitive with other mainstream electric power sources. In addition to describing these technical breakthroughs in clear and simple terms, this book also describes the path from research breakthrough to high-volume production, emphasizing the cooperation required between government and private enterprise. Given this cooperation, solar cells can be a major contributor to the electric power production mix within the next 10 years. This book has been written for a large audience, not just a technical audience. It is hoped that any educated reader will find this book interesting, especially any reader who seeks to understand how the world’s energy supply problems can be increasingly addressed by exploiting direct solar energy resources available within a country’s borders. It further describes how most countries can start moving away from increasingly intense competition for decreasing depletable energy supplies and how they can continue moving toward a long-term, sustainable solution with inherently positive attributes.

PREFACE

xi

The thesis of this book is that solar energy can be cost competitive with other forms of electric power production and that the technical innovations required for this have already been made. Incentives for investment are needed to bring these innovations into high-volume production. It is hoped that this book will help educate the public, possible investors, as well as policy makers worldwide about the potential for a bright sunny energy future. Lewis Fraas Issaquah, WA Larry Partain Mountain View, CA

Contributors Sheila Bailey, NASA Glenn Research Center at Lewis Field, Space Environments and Experiments Branch, Cleveland, OH; email: [email protected] Pedro Banda, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Andreas Bett, Fraunhofer Institut für Solare Energiesysteme (ISE), Freiburg, Germany; email: [email protected] Robert Birkmire, Institute of Energy Conversion, University of Delaware, Newark, DE; email: [email protected] Nathan Blair, National Renewable Energy Laboratory, Golden, CO; email: Nate_ [email protected] Chuck Blouir, Varian Medical Systems, Cleveland, OH; email: chuck.blouir@ varian.com Robert Boehm, Center for Energy Research, Department of Mechanical Engineering, University of Nevada, Las Vegas, NV; email: rboehm@unlv. nevada.edu Christopher Bunner, Spire Corporation, Bedford, MA; email: cbunner@spirecorp. com Christopher Cameron, Sandia National Laboratories, Albuquerque, NM; email: [email protected] Matthew Campbell, Utility Power Plant Products, SunPower Corporation, Richmond, CA; email: [email protected] Richard Colbeth, Varian Medical Systems, Mountain View, CA; email: richard. [email protected] Elmer Collins, Sandia National Laboratories, Microsystems Science & Technology, Albuquerque, NM; email: [email protected] Ron Corio, Array Technologies, Inc., Albuquerque, NM; email: rcorio@wattsun. com David Faiman, Department of Solar Energy and Environmental Physics, Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Israel; email: [email protected] Lewis Fraas, President, JX Crystals Inc., Issaquah, WA; email: lfraas@jxcrystals. com Paul Gilman, Consultant, Chicago, IL; email: [email protected] Christian A. Gueymard, Solar Consulting Services, Colebrook, NH; email: [email protected] xiii

xiv

CONTRIBUTORS

Geoffrey Kinsey, Amonix, Inc., Seal Beach, CA; email: [email protected] Carl LaCasce, Varian Medical Systems, Salt Lake City, UT; email: carl.lacasce@ varian.com Michael Ludowise, SolFocus, Inc., Mountain View, CA; email: mike_ludowise@ solfocus.com María Martínez, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Larry Moore, Sandia National Laboratories, Microsystems Science & Technology, Albuquerque, NM Daryl Myers, National Renewable Energy Laboratory, Golden, CO; email: Daryl. [email protected] Larry Partain, Varian Medical Systems, Mountain View, CA; email: larry. [email protected] Gerhard Peharz, Fraunhofer Institut für Solare Energiesysteme (ISE), Freiburg, Germany; email: [email protected] Ryne Raffaelle, U.S. Department of Energy, National Center for Photovoltaics, National Renewable Energy Laboratory, Golden, CO; email: Ryne.Raffaelle@ nrel.gov Michael Reed, Array Technologies, Inc., Albuquerque, NM; email: mreed@ arraytechinc.com Leonid Rubin, Day4 Energy Inc., Burnaby, BC, Canada; email: lrubin@day4 energy.com Francisca Rubio, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Suresh Sadineni, Center for Energy Research, Department of Mechanical Engineering, University of Nevada, Las Vegas, Las Vegas, NV; email: [email protected] Wang Sicheng, Energy Research Institute, National Development and Reform Commision, Beijing, China; email: [email protected] Jason Strauch, Sandia National Laboratories, Integrated Microdevice Systems, Microsystems Science & Technology, Albuquerque, NM; email: jestrau@ sandia.gov Robert Street, Palo Alto Research Center, Palo Alto, CA; email: [email protected] George Zentai, Ginzton Technology Center, Varian Medical Systems, Mountain View, CA; email: [email protected]

PART I INTRODUCTION TO SOLAR CELLS

1 SOLAR CELLS: A BRIEF HISTORY AND INTRODUCTION LEWIS FRAAS1 AND LARRY PARTAIN2 1 JX Crystals Inc., 2Varian Medical Systems

1.1

BRIEF HISTORY

The history of the solar cell is really quite interesting [1]. In 1839, Edmond Becquerel found that two different brass plates immersed in a liquid produced a continuous current when illuminated with sunlight. We now believe that he had made a coppercuprous oxide thin-film solar cell. Later in the 1870s, Willoughby Smith, W. G. Adams, and R. E. Day discovered a PV effect in selenium. A few years later, an American named C. E. Fritts placed a sheet of amorphous selenium on a metal backing and covered the selenium with a transparent gold leaf film. He reported that this selenium array produced a current “that is continuous, constant, and of considerable force—with exposure to sunlight.” At the time, there was no quantum theory and there was considerable skepticism about his claim of converting sunlight into electricity. So he sent a sample to Werner Siemens in Germany, who was one of the most respected experts in electricity at the time. Siemens’s observation verified Fritts’s claims. However, the conversion efficiencies of both the thin-film cuprous oxide and the amorphous selenium solar cells were less than 1%. Around 75 years passed while quantum mechanics was discovered, the importance of single-crystal semiconductors was recognized, and p/n junction behavior was explained (see Chapter 3). By 1954, Chapin et al. [2] at Bell Labs had discovered, invented, and demonstrated the silicon single-crystal solar cell with 6% efficiency. Over the few following years, researchers brought the silicon solar cell efficiency up to 15%. The timing was fortunate because Sputnik was launched in 1957 and solar cells were the perfect lightweight low-maintenance remote electric power source. Today, silicon solar cells are being used to power the space station. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

3

4

A BRIEF HISTORY AND INTRODUCTION

The solar cell industry remained small until the first Arab oil embargo in 1973. Up until that time, the solar cell industry established a firm foothold with low-level but consistent cell and array production and performance. During those first 20 years, reliability was the driver and cost was not as important. After 1973, the flat-plate silicon module was brought down to earth and modified for weather resistance. This transition also included major improvements in cell and module fabrication that brought down costs dramatically (Fig. 2.3, chapter 2). Flat-plate “champion” silicon cell efficiencies (defined in Section 2.1, Chapter 2) have improved to values as high as 25%. Production module efficiencies have improved from around 10% for early modules to as high as 19% today (SunPower Corporation). Most important, annual production quantities have grown dramatically. Worldwide production exceeded 1 GW/year in 2002 and rose to over 3.8 GW/ year by 2006 (Fig. 2.1, Chapter 2). In the late 1970s, it was discovered that good cells could be made with multicrystalline wafers as long as the crystal size is at least 20 times larger than the optical absorption length [3]. Only those carriers within an optical absorption length from the crystal boundaries are lost. This is less than 5% of the carriers. Typical production quantity multicrystalline cell efficiencies are around 14%, whereas comparable single-crystal cells have efficiencies around 15%. By 2007, modules with multicrystalline cells accounted for about 45% of sales and modules with single-crystal cells accounted for about 40% of sales. Planar silicon cell modules dominated the market in 2007 because of their early well-funded foundation years for space satellites and their huge learning curve support (Fig. 2.3, Chapter 2) from single-crystal silicon and integrated circuit technology development. While silicon-based cells still dominate the solar cell electricity market today, several other cell types have now entered the market. (Solar cells are also known as PV cells.) These newer cell types have added diversity in potential applications as well as offered alternate paths to lower-cost solar electric power. These alternate cell types include hydrogenated amorphous silicon, cadmium teluride and CIGS thin-film cells (Chapter 6), as well as concentrator cells with efficiencies as high as 41% (Chapters 13–17).

1.2

APPLICATIONS AND MARKETS

In the late 1970s and early 1980s, the traditional solar cell electricity applications [4] were at remote locations where utility power was unavailable, for example, campers and boats, temporary power needs for disaster situations, and power for remote communication station repeaters. In the late 1980s and early 1990s, solar cells began to be routinely used to provide site-specific energy for urban and suburban homes, office buildings, and a multitude of other mainstream grid-connected applications. Also, solar cell electricity systems have become very important sources of energy in the developing world. Today, for an increasing number of power needs, solar cell electricity is the cheapest and best way to generate electricity.

APPLICATIONS AND MARKETS

5

In addition to the solar power arrays on space satellites, there are now many different types of PV systems used here on Earth including 1. 2. 3. 4. 5. 6.

remote stand-alone without battery storage, remote stand-alone with battery storage, small modules for calculators and toys, residential grid connected with DC to AC inverter, commercial grid connected with inverters, and PV fields for utility power generation.

Remote solar water pumping is a nice example of stand-alone solar cell electricity where batteries are not needed. Solar water pumping is very desirable for crop irrigation, livestock watering, and clean water for remote villages. Solar water pumping systems are now installed around the world. The nice thing about this application is that underground water is pumped when the sun is shining. It can be immediately used for crop irrigation. In other areas, it can be pumped into tanks for livestock to drink. In third world countries, pumping underground water for people to drink provides cleaner water than surface water thereby limiting disease. This application is quite economical because the system is simple. Battery storage or DC to AC conversion is not necessary. Simple solar trackers are used to maximize pumping time. The electric motors driving the pumps have a threshold current that must be provided before they will operate. By tracking the sun, this power is provided from dawn to dusk, not just at around noon as would be the case without tracking. Another application where there is a good match with demand is for air conditioning in developed countries like the United States. For many remote applications, storage is needed to store electric energy for when it is needed. Examples of these applications include off-grid cabins and remote communication repeater stations. For most solar cell applications where storage is needed, secondary or storage batteries are the best alternative. Generally, batteries should be deep discharge batteries such as marine batteries or motive power batteries. Forklift trucks and golf carts use large-capacity deep discharge batteries that are designed for long life and many discharge cycles. In addition to batteries, combination systems can be used to compensate for the fact that the sun does not always shine. A solar/wind combination is particularly good since quite often, either one or the other is available. Another combination system can be a solar–thermal cell electricity system. In this case, solar cells are located on your roof for generating electricity in the summer and infrared-sensitive PV cells (also known as TPV cells) are integrated into your heating furnace to generate electricity when it is cold and dark outside and you need heat to keep warm. In a TPV cell electricity system, a ceramic element is heated in the furnace flame and its glow in the infrared is converted to electricity by infraredsensitive TPV cells [5]. Solar-powered calculators are another familiar application for solar cells. While the efficiencies of amorphous silicon solar cells are much lower than either single or multicrystalline cells, an advantage for thin-film cells is that they can be made with cell interconnections built into the process. This means that for applica-

6

A BRIEF HISTORY AND INTRODUCTION

tions like powering calculators where voltage but little current is required to run the calculator, amorphous silicon circuits are preferred to save on the cost of interconnecting multiple cells to provide voltage. Credit is due to the Japanese for recognizing this advantage and to the inventors of the amorphous silicon solar cell for making solar cells a common household item [6]. Today, more and more homes on the grid are using solar cell arrays to generate electricity to save on costs of peak electric power. The passage of the PURPA by the U.S. Congress made it possible for a small producer to install generating systems and to sell the power to the utility at a favorable price without the enormous amount of red tape usually required of a new electric power producer. Most states have now also passed net metering laws that allow the electric meter at a home to run both directions. However, at least in California, the utility charge can at most be reduced to zero and they never pay any net money to their customers who produce more electricity than they consume. This allows homeowners generating solar cell electricity to send energy to the grid if they are producing excess electricity with a credit from the utility so that they can use electric power from the grid on days without sufficient sunlight. An example of real cost savings with a solar cell electricity installation for a homeowner in San Jose, California, is shown in Figure 1.1 [7]. Current Monthly KWH 2,500 2,000

Over 300% $0.26 300% Baseline $0.24

1,500

200% Baseline $0.19

1,000

130% Baseline $0.14

500

100% Baseline $0.13

e

Ju ly Au g Se u pt st em be r O ct ob N ov er em b D ec er em be r

ay

ril

Ju n

M

Ap

Ja nu a Fe ry br ua ry M ar ch

–

Monthly KWH With PV System 2,500 2,000

Over 300% $0.26 300% Baseline $0.24

1,500

200% Baseline $0.19

1,000

130% Baseline $0.14

500

100% Baseline $0.13

e

Ju ly Au g Se u pt st em be r O ct ob N ov er em be D r ec em be r

ay

Ju n

M

Ap ril

Ja nu a Fe ry br ua ry M ar ch

–

Figure 1.1. When electric utility rates are staged, a homeowner with solar can displace electricity at the peak power rate as illustrated here. This example was originally presented by Akeena Solar on their web site in 2003 and then published in reference [7].

APPLICATIONS AND MARKETS

7

Figure 1.1 is for an actual case in 2003. Note in this figure that the utility electric rates are staged. While the homeowner pays a base rate of 13¢ per kilowatt hour that in itself is well above the national average. More importantly, the homeowner is paying twice that or 26¢ per kilowatt hour for his peak power. So his solar electric system is saving him money at the 26¢ per kilowatt hour rate. While the grid-connected solar cell electricity market started with residential customers, commercial customers are now starting to use solar arrays on their flat building rooftops. Figure 1.2 shows a photograph of two 1-kW solar cell arrays on a flat rooftop in Spokane, Washington. These arrays are mounted on carousel solar trackers (Chapter 9). People have been dreaming of the potential of solar cell electricity systems as a major electric power source for over 100 years. Now with the existence of solar power fields such as the one in China shown in Figure 1.3, this dream is becoming reality.

Figure 1.2. Two-kilowatt PV array from JX Crystals Inc on a commercial building flat rooftop.

Figure 1.3. Solar cell electricity generating field in Shanghai, China. System designed by JX Crystals Inc.

8

1.3

A BRIEF HISTORY AND INTRODUCTION

TYPES OF SOLAR CELLS AND MODULES

Unfortunately, solar cell electricity is still too expensive for widespread economical use (Section 2.4, Chapter 2). While it is hoped that traditional crystalline silicon module prices will continue to fall, there are other alternatives under development as shown in Figure 1.4. Figure 1.4 shows the three types of solar modules in use today [8]. The upper section (Figure 1.4a) of this figure shows the planar single-crystal silicon modules and fabrication procedure. This approach dominates the solar market today with over 85% of solar modules sold. As shown in Figure 1.5, retail module prices have

(a) Standard Silicon Single Crystal Module Fabrication Crystal to Ingot to Wafer to Module

(b) Concentrator Module Fabrication Smaller Single Crystal Cells With Mirrors (shown) or Lens Array

(c) Thin Film Module - Spray-on Successive Non-Crystalline Films

Figure 1.4. Alternate PV module types: (a) standard silicon single-crystal module fabrication, crystal to ingot to wafer to module; (b) concentrator module fabrication, smaller single-crystal cells with mirrors (shown) or lens array; and (c) thin-film module, spray-on successive noncrystalline films.

TYPES OF SOLAR CELLS AND MODULES

9

8

Module cost (US$/W)

7 6 5 4 3

Small Medium Large

2 1 0 1980

1985

1990

1995

2000

2005

2010

2015

2020

Year

Figure 1.5. Solar module prices for small, medium, and large volumes from 1985 through 2009. All values in then current dollars without inflation adjustments (from Photovoltaics World, September 2009).

been falling dramatically recently. Wholesale module prices are substantially lower than retail prices. The silicon cell cost accounts for about 75% of the module cost with the cost of the glass, frame, junction box, and labor accounting for the remaining approximately 25%. The lower section of Figure 1.4c shows a thin-film module. This concept is attractive because thin films require up to 100× less semiconductor material and offer a promise of lower costs per watt. Since single-crystal material is expensive, why not replace it with inexpensive thin films? The challenge is accommodating their lack of crystallinity. The latter degrades conversion efficiency, which, if too severe, limits their abilities to compete economically in the marketplace (Figs. 2.8 and 2.9 and accompanying text, Chapter 2). An appeal of multicrystalline silicon solar cells is that they offer lower manufacturing costs while still maintaining a conversion efficiency at least two-thirds that of the single-crystal ones [9] of similar Jet Propulsion Laboratory-like configurations (see Chapter 2). However, there are other useful thin-film applications, particularly for amorphous silicon, where its unique properties offer particular advantages and where high quantum efficiency but not high light conversion efficiency is a dominant factor. An example of this is use of amorphous silicon cells in medical imaging (Chapters 22–25) as shown in Figure 1.6. Here, the complete absence of crystallinity in amorphous silicon provides strong radiation damage resistance, and its higher bandgap (than crystalline silicon) gives lower dark currents. These are two strong advantages in the field of flat-plate, digital X-ray imagers that have almost totally replaced analog X-ray film. Recently, amorphous silicon imagers have also begun to displace many of the vacuum tube-based image intensifiers traditionally used in X-ray fluoroscopy. Both X-ray film and intensifier fluoroscopy replacements typically use a thin scin-

10

A BRIEF HISTORY AND INTRODUCTION Kilovoltage X-Ray Source Megavoltage X-ray Source

a-Si Kilovoltage Flat Plate X-ray Imager a-Si Megavoltage Flat Plate X-ray Imager (retracted)

Figure 1.6. Medical imaging system using amorphous silicon solar cell modules.

tillator film to convert the incident X-ray photons into visible light that the underlying amorphous silicon cells efficiently convert into electronic signals that are readily digitized. Frequently, the amorphous silicon solar cells (or pixels) measure a few hundred microns on a side, and millions of them form the rows and columns of a single X-ray imager plate. Such plates provide the digitized X-ray images at up to 30 frames/s and higher. The difficulty with module approaches (a) and (c) in Figure 1.4 is that one tries to obtain both low cost and high efficiency with the same element. In the approach shown in Figure 1.4b, one separates the two requirements of low cost and high performance into two separate elements. The single-crystal cells are the high-efficiency converters used sparingly, while mirrors or lenses are used to concentrate the sunlight onto the cells. The aluminum mirrors (or alternately glass or plastic lenses) are relativley inexpensive. For the case shown in Figure 1.4b, the cell cost is halved. The aluminum mirrors cost at least 10 times less than the singlecrystal cells. In this approach, the sunlight is concentrated onto the expensive high-efficiency single-crystal cells diluting their cost. This approach is now termed CPV. In Figure 1.4b, the sunlight intensity on the cell is doubled; that is, the concentration ratio is 2. Chapter 12 describes a configuation similar to the mirror configuration in Figure 1.4b with a concentration ratio of 3. Various concentration ratios are possible up to as high as 1000. A negative for this approach is that the modules must be aimed at the sun using solar trackers. Trackers by themselves are not a negative as the additional kilowatt per hour per installed kilowatt pays for the trackers. However, when high-concentration optical elements are used, only the direct sunlight is collected. This limits CPV to very sunny locations. However, in any case, solar cell electricity in general will be most economical first in very sunny locations such as the Southwestern United States.

ARGUMENTS FOR SOLAR CELL ELECTRIC POWER

1.4

11

ARGUMENTS FOR SOLAR CELL ELECTRIC POWER

While solar cell electricity is still expensive today, there are three strong arguments for national programs to accelerate its transition into a mainstream power source. The first argument is that there is a logical path for future lower costs for solar electricity. There are three simple steps that will lead to lower cost given development and manufacturing scale-up. These steps are based on technical breakthroughs that have now been made. In step 1, given that the cost of solar electricity today (August 2009) is about 20¢ per kilowatt hour (Solarbuzz) for commercial-sized systems for fixed flat-plate systems in the sunny Southwestern United States, by implementing solar trackers where the modules continuously point at the sun, one can gain 1.3 times more kilowatt hour per installed kilowatt, reducing the cost of solar electricity to about 16¢ per kilowatt hour. This is already being done as evidenced in Figures 1.2 and 1.3 [10]. Step 2 is then to decrease the module cost while maintaining its performance by using lower-cost optical elements as shown in Figure 1.4b. This CPV approach by itself can potentially reduce the system cost for solar electricity to under 10¢ per kilowatt hour (see Chapter 12) [10]. In step 3, one then increases the module efficiency in the CPV approach to well over 20%. As described in Chapter 3, this should reduce the cost of solar electricity still further. “Champion” CPV module efficiencies as high as 31% [11] have now been demonstrated including the one shown in Figure 1.7. While logic suggests these lower costs, this will depend on funding for manufacturing scale-up and government top-down commitment. Actually, there are multiple approaches for CPV ranging from LCPV systems using linear mirrors with silicon cells as shown in Figure 1.4b to HCPV systems

Table III: Performance Summary Packaged Projected Measure Measure Cells at STC with at Module STC 90% Operate at STC Optical Temp (April 28) Effic (April 28) DJ Cell 17.4 W 15.7 W 14.4 W 15.1 W Power DJ Cell 31.5% 28.4% 26.1% 27.3% Effic. IR Cell 3.64 W 3.28 W 2.6 W 3.1 W Power IR Cell 6.6% 5.9% 4.7% 5.6% Effic. Sum 21 W 19 W 17 W 18.7 W Power Sum 38.1% 34.3% 30.8% 32.9% Effic. NIP DNI = 0.92; Area = 600 cm2; Input Power = 55.2 W

Figure 1.7. Prototype CPV module with demonstrated outdoor module efficiency of 31%.

12

A BRIEF HISTORY AND INTRODUCTION

with newer semiconductor materials [12] such as the one shown in Figure 1.7. These LCPV and HCPV systems will be described in more detail in Chapters 12–17. Of course, while the above three steps can be implemented, this still requires investment and political commitment. This leads us to our next two arguments in favor of national programs to accelerate the penetration of solar cell electricity into the mainstream energy mix. The second reason relates to the fact that oil and natural gas resources are being depleted. Quoting from Kenneth Deffeyes’s [13] book titled Hubbert’s Peak: The Impending World Oil Shortage, “In 1956, the geologist M. King Hubbert predicted that U. S. oil production would peak in the early 1970s. Almost everyone inside and outside the oil industry rejected Hubbert’s analysis. The controversy raged until 1970 when the U.S. production of crude oil started to fall. Around 1995, several analysts began applying Hubbert’s method to world oil production, and most of them estimated that the peak year for world oil will be between 2004 and 2008. These analyses were reported in some of the most widely circulated sources: Nature, Science, and Scientific American” [14]. The 2008 peaking of world oil prices to record levels above $140 per barrel seems to support these predictions. The war in Iraq that began in 2003 was likely influenced, at least in part by the shortage of proven U.S. oil and natural gas reserves that could only last 3.0 and 7.5 years, respectively, should the United States have to depend only on its own reserves [15]. The consequence of this “impending world oil shortage” is that electricity prices are going to be rising probably abruptly within the next 5–10 years. This affects the economics of solar cell electricity as solar modules based on semiconductor devices will last for 25 years or longer. Today’s cost competition calculations for solar cell electricity usually assume a short-term payback and nonescalating energy prices. The third argument in favor of bringing solar cell electricity into the mainstream is the environmental and moral argument. It is desirable to avoid global warming as well as oil related war. When one thinks about conventional electric power production, one thinks about oil, natural gas, nuclear, and coal as fuel sources. Solar cell electricity is not on this list because it is currently too expensive. However, these conventional fuel sources have hidden unintentional costs. For example, nuclear fuels are coupled with nuclear waste management and nuclear weapons. Then nuclear waste and nuclear weapons are coupled with the cost of homeland security and our fear of weapons of mass destruction. There are hidden costs involved in attempting to guarantee that nuclear materials do not find their way into the hands of terrorists. Another example of hidden costs is the world’s dependence on oil from the Middle East that is linked unavoidably, particularly in the United States and in other developed countries, with terrorists from the Middle East. It can arguably be claimed that wars have now been fought in the Middle East to secure oil supplies.

ABOUT THIS BOOK

13

In contrast to the unintended costs just enumerated, consider solar energy. Solar energy is inevitable on the larger scale of time. Solar energy is really already a primary energy source through wind and hydroelectricity. Solar energy generated our coal, oil, and natural gas via photosynthesis a hundred million years ago. Solar cells are very much more efficient than plants at converting sunlight to useful energy. Finally, solar energy is benign and will benefit the whole world.

1.5

ABOUT THIS BOOK

The first edition of this book [16] was published in 1995 and can serve as a reference for this second edition. This second edition is divided into four main parts. Part I is an introduction to the current markets, cell and module types, and the physics of solar cell operation. The solar cell electricity market has grown appreciably over the last 14 years as described in Chapter 2. The basics of solar cell operation are presented for single-crystal cells and for thin-film cells in Chapters 3 and 4. Part II of this book focuses on the status of solar cell systems today. Single and multicrystalline silicon and thin-film cells and modules are described in Chapters 5 and 6. Over the last 3 years, silicon module automated manufacturing is coming online with the promising major cost reductions. The traditional and currently dominant silicon module manufacturing, now with automation, is described in Chapter 7. Also, over the last 3 years, China has made a major commitment to solar module manufacturing, and the status of solar electricity in China is described in Chapter 8. A major cost reduction for solar cell electricity comes through the use of solar trackers as described in Chapter 9. Large multi-MW solar cell field installations are then described in Chapters 10 and 11. Part III of this book then describes newer concentrated solar cell and system developments. Chapters 12–17 describe various concentrator solar cell electricity (also known as CPV) modules and system types and installations. Major developments have been taking place here over the last 3 years and that potentially could lead to major cost reductions over the next 5 years. While it remains to be seen if thin-film solar modules can produce electricity at rates competitive with other mainstream electricity generating technologies, nevertheless, amorphous silicon thin-film panels have found a place in other applications and in major markets like flat panel displays and medical imagers. The fourth part of this book describes successful applications of thin film technology as a spin-off from solar cell electricity. Chapters 22–25 then discuss the newest and rapidly growing applications of amorphous silicon thin films in X-ray imaging. The issue of the cost of solar cell electricity, solar modules, and solar systems is a very important subject addressed from various points of view in Chapters 2, 3, 10, 11, 20, and 26. Many believe that solar electricity prices will drop as a result of the recent investments in thin-film module manufacturing, and it is true that thin-film module prices have fallen. However, conversion efficiencies for these commercially available thin-film modules are still under 10%, and this means that

14

A BRIEF HISTORY AND INTRODUCTION

2.5–3.0 times more module area needs to be deployed relative to modules with over 19% conversion efficiencies. So, costs need to be compared at the system level, not just at the module level. Furthermore, the focus needs to be on the cost of the electricity produced in cents per kilowatt hour, not just the hardware cost. A summary and conclusions are presented in Chapter 26. Features that distinguish this second edition from the first edition are the much larger number of solar field installations and the major advances in the concentrator arena using very high-efficiency cells as well as the advances in other novel uses of thin-film modules. The current magnitude and momentum of the solar cell electricity market development (as summarized in Chapter 2) makes its eventual success both inevitable and unstoppable. This is due to the certainty of its future development path with its inherent, major advantages along with the worldwide spread of the scientific knowledge and the manufacturing and engineering know-how (covered in Chapters 3–18), plus the national commitments (alluded to in the Public Policy section of Chapter 2) that will turn promise into reality. World fossil fuel energy production rates will decline in the near to medium time frames, and the current lifestyles of the developed world cannot continue without appropriate replacements. Thus, it is no longer a question of whether this solar cell electricity power transition will occur but one of who will lead this process, who will reap the most benefits, and on what time scale it will occur. ABBREVIATIONS AC—alternating current CIGS—copper indium gallium deselenide CPV—concentrating photovoltaic DC—direct current HCPV—high-concentration photovoltaic LCPV—low-concentration photovoltaic MW—megawatt PURPA—Public Utilities Regulatory Power Act PV—photovoltaic TPV—thermophotovoltaic

REFERENCES [1] [2] [3] [4]

J. Perlin. From Space to Earth, the Story of Solar Electricity. Ann Arbor, MI, AATEC Publications (1999). D. M. Chapin, C. S. Fuller, and G. L. Pearson. A new silicon p-n junction photocell for converting solar radiation into electrical power. J. Appl. Phys. 25, 676 (1954). H. C. Card, and E. S. Yang. IEEE-TED 24, 397 (1977). R. J. Komp. Practical Photovoltaics, 3rd edition, revised. Ann Arbor, MI, AATEC Publications (2001).

REFERENCES [5] [6] [7] [8] [9] [10] [11] [12]

[13] [14] [15] [16]

15

L. M. Fraas, J. E. Avery, and H. X. Huang. Thermophotovoltaic furnace-generator for the home using low bandgap GaSb cells. Semicond. Sci. Technol. 18, S247 (2003). D. E. Carlson and C. R. Wronski. Appl. Phys. Lett. 28, 671 (1976). L. M. Fraas. Akeena Solar example cited in Chapter 2, p. 15, in Path to Affordable Solar Electric Power and the 35% Efficient Solar Cell, Issaquah, WA, JX Crystals Inc. (2004). L. M. Fraas. Path to Affordable Solar Electric Power & The 35% Efficient Solar Cell. Available at http://www.jxcrystals.com/ (2004). M. A. Green. Solar cell efficiency tables (version 29), Prog. Photovolt. Res. Appl. 15, 15–40 (2007). L. M. Fraas, J. Avery, L. Minkin, H. X. Huang, A. Gehl, and C. Maxey. Carousel trackers with 1-Sun or 3-Sun modules for commercial building rooftops. Presented at the Solar 2008 Conference, May 6, 2008, San Diego, CA (2008). L. M. Fraas, J. Avery, H. Huang, L. Minkin, and E. Shifman. Demonstration of a 33% efficient Cassegrainian solar module. Presented at 4th World Conference on PV, May 7–12, Hawaii (2006). L. M. Fraas and R. C. Knechtli. Design of high efficiency monolithic stacked multijunction solar cells. In IEEE Photovoltaic Specialists Conference, 13th, Washington, D.C., June 5–8, 1978, Conference Record (A79-40881 17-44). New York, Institute of Electrical and Electronics Engineers, Inc., pp. 886–891 (1978). K. S. Deffeyes. Hubbert’s Peak: The Impending World Oil Shortage. Princeton, NJ, Princeton University Press (2001). C. A. Campbell and J. H. Laherrere. The end of cheap oil. Sci. Am. March, 78 (1998). www.BP.com web site has a section entitled “BP Statistical Review of World Energy 2003.” This site has country and regional proven reserves and consumption data for both oil and natural gas. L. D. Partain, ed. Solar Cells and Their Applications, 1st edition. New York, John Wiley & Sons (1995).

2 SOLAR CELL ELECTRICITY MARKET HISTORY, PUBLIC POLICY, PROJECTED FUTURE, AND ESTIMATED COSTS LARRY PARTAIN1 AND LEWIS FRAAS2 1 Varian Medical Systems, 2JX Crystals Inc.

2.1

MARKET HISTORY

The worldwide solar cell (or photovoltaic) electricity generation market has grown dramatically over the past 30 years, both in terms of gigawatts per year (Fig. 2.1) (1 GW = 1 billion watts) and in billions of U.S. dollars per year sales (Fig. 2.2) [1]. Worldwide production exceeded 1 GW/year in 2002 and rose to over 3.8 GW/ year by 2006, and worldwide sales increased from US$(2007)1.5 to 9.7 billion over this same time period. The US$(2007)390 billion sales of the world’s largest oil company, Exxon Mobile, are shown in Figure 2.2 for reference [2]. Since 1975, this progress has provided a 1000-fold increase in gigawatts per year produced and almost a 100-fold increase in billions of U.S. dollars per year in sales. From 1975 to about 1985, the gigawatt per year growth rate was about a factor of 2 every 2 years. From about 1985 to 1995, it dropped to about half that rate before returning to approximately doubling every 2 years from about 1995 through 2006. Another 1000-fold increase in gigawatts per year production would place the solar cell electricity generation capacity in the range of both the total 2006 U.S. installed electricity generation capacity of 1075 GW [3] and the installed 2005 world electricity generation capacity of 3889 GW [4] even after allowing for the “capacity factor” of terrestrial sunlight being available only about 15–22% of the time [5, 6], neglecting storage issues. Since most commercial solar cells for terrestrial use

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

17

18

SOLAR CELL ELECTRICITY

World Solar Cell Production (gigawatts)

10000

US PURPA

1000

Net Metering

JPL FSA

100

06 US Generating 05 World Capacity Generating Capacity

10 1

Rate To Double Every 2 Years

Cumulative

0.1 Annual

0.01 0.001 1970

1980

1990

2000

2010

2020

2030

2040

Year

Figure 2.1. The world solar cell production levels, on both an annual and cumulative basis, since 1975, with comparisons to the U.S. and world total electricity generating capacities in 2006 and 2005, respectively, and with support and public policy program time intervals for the JPL FSA, the U.S. PURPA, and the net metering programs. 1000 Exxon

World Solar Cell Market Size (billions 2007 US$/yr)

US PURPA

100

Net Metering

JPL FSA

Rate To Double Every 2 Years

10

1

0.1 1970

1980

1990

2000 2010 Year

2020

2030

2040

Figure 2.2. The world annual solar cell market size since 1975 with a comparison to the annual sales of the world’s largest oil company, Exxon Mobile, in 2007.

are designed for 20 years or greater lifetimes, the actual solar cell generation capacity lies between the annual and cumulative curves of Figure 2.1. The reason for the difference in the gigawatt and dollar growth rates is the constantly falling cost of terrestrial solar cells over the past 30 years as plotted in Figure 2.3. The 1975 technology was essentially slightly modified space satellite technology that cost about a hundred U.S. dollars per watt in 2007. By 2006, this had steadily dropped 25-fold to about four 2007 US$/watt. Like many high-technology and electronic products, this cost reduction roughly followed a learning

MARKET HISTORY

19

Solar Cell Module Cost (2007 US$/watt)

100 2X Cost Drop For Every 10X Increase In Cumulative Production Growth

10

8.6 gigawatts by 2006 1 0.001

0.01

0.1

1

10

100

1000

Cumulative Solar Cell Production (gigawatts)

Solar Cell Production (megawatts/yr)

Figure 2.3. The decrease in solar cell module costs since 1975 as a function of the cumulative solar cell production power through 2007. The dotted line indicates a learning curve decrease in cost by a factor of 2 with every 10× increase in cumulative production. 10000

Japan Europe China

1000 Total

Taiwan

100

Others

US

India

10

1 1994

1996

1998

2000

2002

2004

2006

2008

Year

Figure 2.4. The solar cell annual production rate from 1995 through 2006, broken down by the United States, Japan, Europe, China, Taiwan, India, and all others.

curve of dropping about a factor of 2 for every 10-fold increase in cumulative market production [7]. However, this price reduction has pretty much stalled during the last 10-fold increase in cumulative market. Many analysts attribute this cost plateau to a shortage of feedstock silicon required to manufacture the vast majority of current terrestrial solar cells [8]. The United States led in the production of solar cells from 1975 until the late 1990s. As shown in Figure 2.4, the Japanese production rate exceeded that of the United States in 1999, followed by Europe in 2003 and by China in 2006. In 2006,

20

SOLAR CELL ELECTRICITY

the United States’ fourth place position was essentially shared with Taiwan, whose rate of growth was higher. As shown in Figure 2.5 the United States held the third position in the yearly installation of solar cells in 2007, barely ahead of the rest of Europe (i.e., all but Germany) and with the rest of Asia (all but Japan) quite close behind. By 2004, Germany’s installation rate surpassed that of Japan. There are three major options to try and resume the module cost reduction process plotted in Figure 2.3. The first is to fundamentally improve crystalline cell technology and performance well beyond that of the original 1970s “modified space cell” configuration. Two potential candidates for this are the commercial SunPower interdigitated back contact cell and the SANYO HIT cell (see Chapters 3 and 4) both with prototype modules of independently confirmed “champion” efficiencies over 20% [9]. The second is thin film technologies that require much less (typically 100X) material than crystalline silicon solar cells (see Chapter 6). By 2006, the thin film annual production rate had grown to within a factor of a 1000 of the mainly silicon, total solar cell production rate, led by the United States and Japan as shown in Figure 2.6. In 2007, thin films’ growth rate was greater that that of the total market for the first time (compare with Fig. 2.1). This latest growth was mainly due to thinfilm CdTe cells that have taken the thin film lead from amorphous silicon. The decreased thin film production in Japan was mainly due to reduced activity in amorphous silicon. The third major option, for continued cost per watt reduction, is concentrator solar cell systems where the sunlight is focused down to a small area so that much smaller amounts of expensive solar cell material are also required for electricity 10000 World Solar Cell Installations (megawatts per year)

Europe Feed-In Tariffs

1000

Germany 100,000 Roofs

Germany Total

Japan

US

100

Rest of Asia

10 Rest Of Europe

1 1998

2000

2002

2004

2006

2008

Year

Figure 2.5. The annual solar cell installation rate from 2000 through 2007, broken down by the United States, Germany, Japan, the rest of Europe, and the rest of Asia, with time interval comparisons to Germany’s 100,000 Roofs Program and Europe’s Feed-In Tariff Program.

MARKET HISTORY

21

World Thin Film Solar Cell Production (gigawatts/yr)

10000 1000 100 10 Rate To Double Every 2 Years

1 0.1

Europe Japan

Total US

0.01

Rest Of World

0.001 1994

1996

1998

2000

2002

2004

2006

2008

Year

Figure 2.6. The annual world production of thin-film solar cells from 2003 through 2006, broken down by the United States, Europe, Japan, and the rest of world, and with the world total extended through 2007. The dashed line indicates a growth rate to double every 2 years.

Cumulative World Solar Cell Sales (billions 2007 US$)

1000 Rate To Double Every 2 Years

100

$36 billion by 2006

10 Germany 100,000 Roofs

1 JPL FSA

0.1 1970

1980

1990

2000

2010

2020

Year

Figure 2.7. The world cumulative solar cell sales from 1975 through 2006, with comparisons to the total expenditures of the JPL FSA Program through 1985 and Germany’s 100,000 Roofs Program through 2003.

generation (see Chapters 12–17). The latter are in their earliest, larger-scale testing phases, with most installation efforts currently centered in Spain ([10]; Banda and Rubio, Chapter 17). Figure 2.7 shows the cumulative sales totals of solar cells that have grown to US$(2007)36 billion by 2006. This is a classic learning curve-type response to a total integrated investment, which has overwhelmingly been into crystalline silicon solar cells, over this 1975–2006 time period. This total magnitude serves

22

SOLAR CELL ELECTRICITY

as a significant and growing barrier to any new technology that essentially has to become competitive at a greatly accelerated growth rate (comparatively) but with much less total integrated investment. Until recently, the thin-film growth rate had not been able to duplicate that of crystalline silicon (Figs. 2.1 and 2.6). Compounding any new solar cell approach’s problems is crystalline silicon’s constantly improving efficiency, reliability, ruggedness, acceptance, and costs as it continues down its learning curve, with ever-growing totals of integrated sales investment. The classic response strategy for this situation is a disruptive one where the new technology enters a niche market where the dominant technology does not participate and proceeds independently down its own learning curve largely unchallenged until its new concepts suddenly become competitive [11]. A potential example of the latter is the very high-efficiency, but also very high-cost, triplejunction space solar cells, based on III–V compounds (e.g., GaAs) that now dominate the space solar cell market (see Chapters 13–17).

2.2

PUBLIC POLICY

Over the past 30 years, there have been multiple significant public policy decisions, which have directly driven and strongly influenced the rate of development and the geographical location of major segments in the world total solar cell electricity marketplace. Here, five primary ones are highlighted that can be readily correlated with major market responses evident in the historical statistical market data. The earliest is the JPL FSA from 1975 through 1985 [12]. It corresponds to the initial rapid rate of market growth as indicated in Figure 2.1. It transformed delicate and expensive space silicon cells into rugged, reliable, and affordable terrestrial ones through five increasingly large (hundreds of kilowatts) and rigorously competitive block purchases of silicon cell modules whose properties were monitored and meticulously tested and reported [13]. The last two of these blocks produced the two early peaks in market growth shown on the left side of Figure 2.2. Its US$(1985)235 million (US$[2007]403 million) [14] total program size was equivalent to about 30% of the total integrated total market sales between 1975 and 1983 as is evident from Figure 2.7. Both its magnitude and sharp focus (on transforming fragile, expensive space silicon cell technology into robust and affordable terrestrial products) made it the primary driver of this early market development period. At each stage, these modules had to pass increasing stiff environmental reliability, stability, and performance standards. By the end, the project had met all of its major goals except for the price of the modules. The emerging results of singlecrystal, sawed silicon wafers, screen printed with a top metal grid contact, sandwiched between a plastic back plane and a top glass plate with polyvinyl butyral adhesive or EVA, set the standard for the next 20 years of solar cell market leadership by the United States. Following World War II, the U.S. economy was the world’s largest and it could best afford such a market generating and leading investment into this new breakthrough renewable energy technology. According to Figure 2.7, the growth

PUBLIC POLICY

23

in the world market equaled the total US$(2007)403 million JPL investment by 1988, with geometrical larger annual revenues the following years into the 1990s when U.S. companies still led in the world marketplace. For the next 10 years (about 1985–1995), the growth rate dropped by about a factor of 2, supported but only rather weakly by the PURPA that had passed in 1978. PURPA allowed “independent” electricity producers to be paid utility wholesale “avoided costs” under long-term (15–30 years), fixed-price contracts at least early on [15]. This was strongly led by California Public Utility Commission’s interpretation and requirements that ordered utilities to offer standardized contracts, most notably Standard Offer No. 4, with fixed prices. In the 1980s, PURPA was estimated to have produced up to 12 GW in non-hydro renewable energy system installations in the United States. Over half (6.1 GW) were in California, but most (over 90%) were for wind, geothermal, and biomass. Unfortunately, PURPA was not restricted just to renewables. By the early 1990s, its support of renewables stagnated with most new contracts going for nonrenewable natural gas “cogeneration” plants (producing both electric and steam) due to changing conditions including relatively low natural gas costs. Nevertheless, a significant aspect of PURPA was its first demonstration of a “feed-in”-type tariff, where renewable energy sources received “above-market” prices for their power output under longterm, fixed-price contracts, resulting in a large market response and many installations, although the majority of these were not for solar cell electricity. The next big positive stimulus for solar cell electricity was net metering, which started in the early 1990s [15]. It became prevalent enough in the United States to restore the rapid growth of solar cell production by about 1995 (as indicated on Fig. 2.1). Net metering allows independent producers, such as rooftop systems on residences and commercial buildings, to be credited by their utility for the retail avoided cost of utility-delivered electricity to their site (as opposed to wholesale avoided costs of PURPA). Typically, the utility only credits for the siteproduced power, up to the net power delivered to the site from the utility each year. Thus, any excess site electricity production is not accrued and no actual payment goes from the utility to the customer. The benefit is that the utility customer, producing local power, can offset up to 100% of the retail value of the power delivered to his or her site each year. From Figure 2.4, it can be seen that the Japanese solar cell production exceeded that of the United States by 1999 and it has maintained this world leadership production position since that time. The cost of Japanese solar cell modules decreased by over a factor of 4 from 1994 to 2003 (from ¥1733 per watt to ¥583 per watt or from US$[2007]23.62 to 4.32 per watt [8, 14, 16]) and went from being not cost competitive to being very competitive. During this same period, the efficiency of their modules increased from 12% to the 15–17% range, while the thickness of the silicon wafers used decreased from 380 to 180 microns. The performance increase came from continuous incremental improvements in light trapping (textured surfaces, double antireflection coating, reduced contact shadow). It also came from lower losses (good-quality substrates, passivation, heterojunctions) and from lower resistance (thicker electrodes, fine contact pattern). Cost reductions came

24

SOLAR CELL ELECTRICITY

from thinner substrates, larger cell sizes, simpler and lower-temperature processes, non-vacuum technology, and from making substrates directly from molten silicon. This was all largely developed directly at large commercial Japanese companies, but in close coordination with Japanese universities and with the Japanese government agencies with a clear mandate and a government resources commitment sufficient for success. The latter program was essentially that of the NEDO as part of the New Sunshine Program under the MITI [8] and its successor organizations. Approximately 8% of Japanese corporate funding (including NEDO support) goes to basic research, and this has been enough to provide the high-performance crystalline silicon HIT cell from SANYO (see Chapters 4 and 5). Since the late 1980s, the Japanese economy has been one of the world’s largest. In 2007, it was the world’s second largest economy, exceeded only by that of the United States [17]. With its comparatively small commitment to military expenditures, Japan has been particularly free to focus heavily on its internal hightechnology industrial development. “… It was in the middle of that decade [the 1990s] that the country [Japan] realized the importance of innovation and redesigned its whole strategy” [18]. “The size of NEDO’s budget illustrates this point. Japan’s corporations spend a combined 111 billion [2008] US$ on research and development every year, but more than 90% of that actually goes to incremental improvements of existing products, with 7 billion [2008] US$ going to basic research. That makes the more than 1 billion [2008] US$ that NEDO pours into its 140 research projects equal to 16% of the total corporate [Japanese] investment into basic research … Take NEDO’s long-term commitment to solar cell research. On the one hand, Japanese manufacturers like Sharp and Kyocera have won a significant share of the global market using technologies developed through NEDO projects. … With … sinister place names like ‘Devil River’, ‘Valley of Death’ and ‘Darwinian Sea’, … the illustration on the table [of the] … director general of … NEDO looks like some kind of map … It’s a vivid visual metaphor created by the U.S. National Institute of Standards and Technology to express the pitfalls that lie waiting for inventors trying to turn their research into commercial products … In Japan … [the NEDO director general] is the man charged with helping the scientific community to catapult technology and know-how past these obstacles and into the marketplace. … Since being reorganized in 2003 … [NEDO has] complete control of the national projects … [to] cancel projects that aren’t going well and direct more money into promising-looking research.” There are no details of the amounts of NEDO funding of solar cell R & D programs over the 2000–2005 time period, but when combined with Japanese corporate R & D expenditures, these combined totals are plausibly on the order of 30% of the integrated cumulative sales of solar cells of ∼10 billion (US$[2007]) by that time (see Fig. 2.7). Such NEDO-coordinated and well-supported programs in Japan, centered in large commercial operations, are reminiscent of the scale and methods of the U.S. technology development efforts termed the military–industrial complex by U.S. President Eisenhower [19]. The latter evolved dramatically from the cauldron of pressures leading up to and into World War II [20], and their implementations can be reasonably credited with keeping the U.S. military technology more than com-

PUBLIC POLICY

25

petitive with Germany’s and Japan’s early on in the World War II years and later on with the former Soviet Union’s. Its time-tested and hard-proven U.S. Department of Defense’s Science and Technology program (of US$[2007]10.8 billion) is split with 13% to basic research, 41% to applied research, and 46% to advanced technology development [21]. The latter two guide new military technology through “the Valley of Death”-type obstacles, identified by the former Advanced Technology Program of the U.S. National Institute of Standards and Technology and as highlighted and accommodated in current NEDO programs. Survivors are then ready for standard procurement or sales processes. U.S. public policy since the 1980s has largely blocked comparable programs in the United States for any technologies and industrial developments other than those focused or directed at U.S. military applications. The U.S. military–industrial complex approach was a dramatic change and improvement in the R & D process and in the rapid translation of research success into large-scale production and availability [20]. Its only major missing part, relevant to the widespread commercial introduction of solar cell-based energy, is the economic viability and market evaluations like those of the former Advanced Technology Program of the U.S. National Institute of Standards and Technology. By the mid-1990s, the EU countries and their increasingly integrated industries and markets began to build an expanding and major role for themselves in the world economy, not only for the EU in general but for Germany in particular, which was the world’s sixth largest economy by 2007 [17]. By 2007, the combined economies of the EU actually exceeded that of the United States. From 1999 until 2003, the German government instituted the 100,000 Roofs Program, targeted only at solar cell electricity, with 1.7 billion euros (US$[2007]2.2 billion) [14, 16] in favorable loans and feed-in law tariffs of up to 0.574 euros per kilowatt hour (US$[2007]∼0.75 per kilowatt hour) for up to 20 years [22]. This US$(2007)2.2 billion level investment again represents about 30% of the total world integrated sales of solar cells up to 2003 as shown in Figure 2.7. This program generated 145 GW of installed solar cell panels in Germany in 2003 to make it the leading installer since that time (see Fig. 2.5). It is estimated that the Roofs Program generated 10,000 German jobs and created 800 million euros (US$[2007]1.0 billion) in German industry revenues in 2003 alone. With this rate of revenue growth, this total loan investment was duplicated in German sales revenues within about the three following years. In comparison, the size of the California economy ranks eighth in the world [23]. Its and Germany’s (ranked sixth) have both the ability and track record of early demonstration of breakthrough technology and market trends including ones in renewable energy. Expanding on the PURPA concept, Germany introduced its version of feed-in tariffs in 1990, which was refined into its successful form by 2000 when it became a federally managed program [24, 25]. With such feed-in tariff law programs instituted in multiple European Community nations, Germany, Spain, and Denmark each provide significant portions of their countries’ electricity totaling 5%, 9%, and 20%, respectively, in 2007 from renewable sources, which were under long-term (20 year) contracts in Germany and in Spain [26]. The 2007 solar cell

26

SOLAR CELL ELECTRICITY

electricity installation totals for Germany and Spain were 47% and 23% (for a combined 70%) of the world market. These latest feed-in tariffs are justified by rationale that can include the following [24]. Long-term contracts (up to 20 years) pay solar cell electricity “tariffs” up to five times the going commercial rate (e.g., US$[2007]0.75 per kilowatt hour) [22], as opposed to representative rates (of US$[2007]0.16 per kilowatt hour typical of Italy in 2006 [27]). When 3% of a country’s electric power is produced under such agreements and the costs are spread among all the users, the 500% premium becomes a ∼15% increase in each user’s electricity bill. The integrated production that supplies these countries’ installations drives down the cost of solar cell modules, according to its learning curve (see Fig. 2.3). For a learning curve of 2X cost reduction for every 10X increase in cumulative production, the cost premium would be reduced by a factor of 2 when the cumulative production grows 10-fold, and a factor of 4 for 100-fold. After that, the cost to the customers for the next 3% of a country’s solar cell electricity installation reduces the price premium by a factor of 2–4 down to 7.5–3.75%, respectively (compared to the initial 15%). With integrated solar cell production growth doubling every 2 years (see Fig. 2.1), these 10- and 100-fold increases would occur in about 3.3 and 13.0 years, respectively. Three percent of a large user market can be quite significant particularly in the early penetration stages of solar cells into the total electricity market. For instance, 3% of the 2007 U.S. installed electricity capacity would be over 32 GW, more than eight times the total size of the world solar cell production total in 2006. The benefits that could accrue from such a program are a nation’s electricity energy supply becoming substantially less dependent on imported oil and gas, from a domestic source of sunlight that never decreases and that is delivered on-site free of charge, where the electricity conversion process itself (excluding the feedstock refining, manufacturing, installation, recycling, and disposal of solar cells) produces no pollution and generates no greenhouse gases. As a widely distributed source, it would be less subject to disruption by terrorist or military acts or by natural disasters. An orderly transition into this major new energy resource could be made steadily over many years, before world oil and gas production peak and begin to decline, which otherwise could generate sharp price fluctuations and major delivery disruptions with little or no time for orderly change. The resulting industries generated within countries could earn sales revenues equal to such subsidies within about three or four following years (if historical trends repeat) while at the same time creating large numbers of new jobs and major positive contributions to each country’s international balance of trade. Early participation in the process could well lead to world economic market leadership positions and geographic shifts of the centers of activity that historical evidence suggests can last for many years. It is the opinion of many economists [28, 29] that China is well on its way to becoming the world’s largest economy within the next few years or decades, with India likely to vie with the United States for second place perhaps 10–20 years even further out. From indications like the very high rate of expansion of its solar cell production capacity (Fig. 2.4), China may well be in the best position to invest

PUBLIC POLICY

27

the tens to hundreds of billions of 2007 U.S. dollars that historical data indicate could thrust it into the world’s leading position, at least in the area of solar cell production. And if history repeats itself, China would rapidly equal this total investment amount within 3 or 4 years of sales from its industries while also benefiting from the jobs generated and national sales revenue as well as from further increases to its international balance of payments surplus. There have been many other government-sponsored programs, some involving even larger monetary investments at their time, than the five cited above, but none with the comparably large and demonstrable effects on the world’s solar cell electricity market development history. Hence, it is not only the magnitude of a government program’s investments that provides the impetus for whole new technology applications and for relocating the geographical centers of leading activity, but it also takes wise selection of program focus and strong program execution. It is instructive to note that only the first (the JPL FSA) and the third (the Japanese NEDO development of crystalline silicon) of the five government programs covered above was directly focused on a single technology approach. Most, if not all of the other three, were essentially solar cell technology area neutral and were economic and technology stimulus programs. To a large degree, these all have mainly made use of the original JPL crystalline silicon design, with only this technology’s steady and incremental improvements, developed mostly by the manufacturers themselves (with large government subsidies at least in Japan) in response to the growing market. Over the last 30 years, the U.S. government’s (and many other countries’) renewable energy programs have had their major focus on the development of new breakthrough technology. Unfortunately, to date, the results have had little measureable impact on the solar cell electricity market and its development trajectory. However, this may be about to change; as Figure 2.3 indicates, the last 10X increase in cumulative solar cell production has produced little or no decrease in module costs, essential to the strategies covered above, for making solar cell electricity into a major alternative to depletable oil, natural gas, and coal energy supplies. A relevant public policy success example for technology implementation is the U.S. Interstate Highway System that was initiated in the Eisenhower administration beginning in 1956 [30]. It was justified as a defense-related program and was ∼90% federally funded by highway user taxes because it was considered “vitally important to national goals.” The Interstate Highway segments cost many times that of the then typical highways, and most of the highway users taxed had no direct Interstate Highway access in the early years of the program. The eventual positive impact of this system on the employment and housing opportunities of a large fraction of the U.S. population and on the U.S. economy as a whole has arguably been even greater than that of the federally funded U.S. National Aeronautics and Space Administration and its space programs. An analogue for this in solar cell electricity would be a federal tax on the users of depletable energy, namely, oil, natural gas, and coal. For example, this could pay a fraction of the initial capital costs (starting at, say, 75%) of installed solar cell electricity generating systems until approximately 30% of the then current cumulative world solar cell electricity

28

SOLAR CELL ELECTRICITY

market has been subsidized. These depletable energy users would then be directly paying for the development and deployment of renewable energy systems “vitally important to national goals” to the point that the latter becomes self-sustaining. For scale and reasonableness comparisons to the above past and potential future government programs, the following historical cost numbers may be useful. The cost of petroleum and petroleum products imported into the United States was US$327 billion in 2007, an increase of US$27 billion from the 2006 costs [31]. The U.S. 3-month balance of payment (current account) deficit increased by US$(2007)10.2 billion largely due to petroleum and petroleum product imports, which brought the total U.S. deficit to US$(2007)176 billion by the end of March 2008 [32]. Should military actions be needed at some future time to avoid major disruptions in imported gas and oil to any country, the ongoing Iraq war could provide a rough order-of-magnitude cost estimate from its 5 year costs to the United States of US$450 billion through the end of September 2007 [33]. The total final estimated cost in 1991 of the 42,795 miles of the U.S. Interstate Highway System at that point in time was $128.9 billion with $114.3 billion (or 88.7%) paid for from U.S. government funds at an average cost of $3 million per mile [34].

2.3

PROJECTED FUTURE

The historical market data above allow analysis of known quantitative results and comparatively objective evaluations of probable causes of, and contributions to, major past events and developments. Future projections require extrapolations beyond what is presently known and require a greater degree of speculation and unavoidably involve more subjective judgments and opinions. Nevertheless, important and accurate parallels can often be drawn between past history and future results, when past lessons learned are well applied, using well-accepted scientific principles and limitations combined with sound mathematical engineering and market analyses as attempted below. A continued growth rate of doubling every 2 years of production (and appropriately lower sales growth due to learning curve price reductions in Fig. 2.3) for solar cell electricity products and systems could lead to a sizeable fraction of the total U.S. and world electricity generating capacity and energy sales being provided by solar cell electricity in the decade between 2020 and 2030, as seen from Figure 2.1. One risk that could impede this progress is continuing stagnation of the Figure 2.3 solar cell module cost reduction with cumulative production. Another risk is that reasonable approaches may not be readily found to match the production of solar cell electricity to the demand schedule of electricity customers (e.g., night use of electricity when the sun is not shining). An accelerating factor for an even faster rate of growth is the rise in the real cost of energy (after correction for inflation) derived from depletable fossil fuel sources (i.e., oil, natural gas, and coal). The growing fossil fuel energy demands of the rapidly growing economies of China and India, plus the steady growth from the other leading world economies, strongly suggest that this real price increase will continue inexorably

PROJECTED FUTURE

29

for the medium term and precipitously over the longer term when the world production rates of oil and natural gas energy supplies eventually peak and begin to fall. A parallel case of a similar solid-state electronic technology growth is the approximate doubling of the calculating capacity on computer-integrated circuit chips every 2 years, which is termed Moore’s law [35]. Although all agree that this computing power growth rate cannot continue indefinitely due to fundamental and mathematical limitations, this empirical Moore’s law progress rate has essentially remained on track for more than 40 years. In comparison to solar cell electricity’s potential, 2030 is less than 25 years away. The terrestrial solar resource available is not a fundamental limitation to producing solar cell electricity each year at the level of the world generation capacity in 2005 (shown in Fig. 2.1) or even higher. The intensity of sunlight striking the outer boundary of the earth’s atmosphere is 1367 W/m2 (±4%) 24 h/day and 365 days/year as shown in the First Edition, Append B [36]. The diameter of the earth (at the equator) is 12,756 km [37] and the intensity of sunlight striking the earth’s surface (through an average atmospheric path length—i.e., AM1.5G) is 0.704 of its intensity at the atmosphere edge as given in the First Edition, Append A [36]. If one combines these with the conservative assumptions that (1) only land-based systems will be used; (2) one-third of the earth’s surface is land; (3) the availability of sunlight only matches demand 20% of the time (i.e., the capacity factor = 0.2); (4) 1% of the land is appropriate for solar cell electricity generation; and(5) the average system conversion efficiency is 10%, then the worldwide solar cell electricity resource size is 8193 GW or more than twice the 2005 world electricity generation capacity of 3889 GW shown in Figure 2.1. As the world’s dominant energy sources have evolved over time, from wood to coal and then to oil and gas, the total energy market size has increased substantially with each transition due to inherent advantages following each step. The solar input resource is large, so that the major challenge is to match solar cell electricity supply with demand. With natural hydro capacities at 121 and 923 GW for the United States and the world, respectively, in 2005 [38], this renewable storage resource can only cover demand mismatches of up to these same orders of magnitude. At Figure 2.1 extrapolated growth rates, such world levels could be reached as early as 2025. Unless battery or some other storage alternatives become practical, available and cost-effective in the meantime, solar cell electricity could be limited to supplying a fraction of the total world electric energy demand. To become the major electric energy resource, some fundamental change would likely be needed in electric energy use patterns. One of the most important and valuable U.S. exports is agricultural crops. This crop production essentially occurs only in the daytime, peaks in the summer months, with a virtual shutdown for most of the winter. Time of day and time of year utility rates could start to encourage business and personal lifestyles to take full advantage of such time availability of solar cell-generated electricity. Electric energy intense industries, such as aluminum refining and others, might well be induced to shift major fractions of their yearly

30

SOLAR CELL ELECTRICITY

production toward the daylight hours of the summer, if there were financial incentives like lower electricity rates.

2.4

COST ESTIMATES

The levelized cost of energy, L, is a detailed calculation of the effective cost of energy provided, such as 2007 U.S. dollar per kilowatt hour, considering all of the life cycle factors and expenses that contribute to such effective costs [39]. The intent of this chapter is to be as free of involved mathematical expressions and derivations as possible. Hence, Table 2.1 is used to separately summarize simple and approximate expressions that can be readily used to estimate L values and what effects them most based on parameter input values in the range of those listed in Table 2.2. This starts with baseline estimates taken from recent field experience and then proceeds to projected future cost reductions in the near- and medium-term time frames. The less mathematically inclined can thus skip the next two paragraphs and proceed directly to the results that follow. A more sophisticated and rigorous cost analysis is the complete topic of Chapter 20 (Solar Advisor Model) of this book. As listed in Table 2.1, the levelized cost of energy (L) is obtained by taking all the capital costs incurred for installing a solar cell electricity system, converting these initial capital costs into an annualized charge using fixed charge rates, and then dividing the results by the electric energy produced by such a system in a year. The result is an L value in dollar per kilowatt hour. Here, the simple approach developed by the Electric Power Research Institute (EPRI) was used [39]. The annual energy produced per unit area is just the solar energy density SEY (in kilowatt hour per square meter) that strikes the surface of the system per year, which is then adjusted for all of the relevant conversion efficiencies. The derating efficiency accounts for the mismatch and interconnect losses inevitably encountered in system assembly. A major component of fixed charge rates is the loan amortization rate, which is the per dollar cost per year of obtaining a loan at a given rate for a given period of time. This is easily obtained from loan amortization tables or from use of the Microsoft Excel PMT function. In locations where there are tax write-offs for the interest particularly on home loans, this reduces the fixed cost rate, proportionally to the tax bracket B and the fraction f of the loan payment that is interest. Additional annual financing costs are covered in the factor Δ. In a residential home loan, this Δ includes the annual taxes and insurance cost (per total loan dollar value) that adds to the amortized “interest and principal” payment (i.e., the principal, interest, taxes, and insurance [PITI] payments). The L value is conveniently expressed in terms of area-related capital costs per unit area and power-related capital cost per unit power (including the DC to AC inverter costs for the latter). One just multiplies these by the area A and the peak power P needed to generate 1 kWh in 1 year, and this gives the expression for L. For the power part, one needs to realize that module efficiency values are based on the assumption of a standard peak value of solar energy striking a system

COST ESTIMATES

31

TABLE 2.1. Levelized Cost of Energy (L) L ($ kWh ) =

TS FS + Ti Fi Annual AC energy production

Annual AC Energy Production per unit area = ηDηiηmSEY

L = (1 + r )

(Cm + Cb ) Fs A ηD ηi ηmSE Y

+ (1 + r ) Ci Fi P

AC kWh produced per year = ηDηiηmSEYA so A = 1/(ηDηiηmS) Peak output power P of inverter in kW needed for array of area A with a module efficiency ηm where the nominal peak solar input power is 1000 W/m2 = 1 kW/m2 P = ηDηiηm 1 kW/m2 A = ηDηiηm/(ηDηiηm) = 1/SEY and Cm($/m2) = Cw($/W)ηm1000(W/m2) so that (1000ηm [Cw − I ] + Cb ) FS + Ci Fi / SE Y L = (1 + r ) ηD ηi ηmSE Y where I—government incentives in $/W and η η C b FS Cw = D i {L SE Y (1 + r ) − Ci Fi } + I − 1000 1000ηm

T—total capital investment F—fixed charge rate (converts initial investment into annualized charge) Subcript S—total system except for DC to AC power inverter Subscript i—DC to AC power inverter ηi—DC to AC inverter power conversion efficiency ηm—module efficiency of converting sunlight into DC power ηD—system derating efficiency SEY—sunlight energy density striking module surface over year in kWh/m2 where FS = mS – BfSmS + Δ and Fi = mi – Bf imi + Δ r—indirect cost rate (including yearly operations and maintenance costs) Cm—module cost in $/m2 Cb—area-related balance of systems costs in $/m2 (including installation) m—loan amortization rate per year f—fraction of m that is interest B—tax bracket Δ—additional fixed charge rate A—area required to produce 1 kWh/ year in m2 Ci—DC to AC inverter cost in $/kW P—peak inverter power output rating in kW required to produce 1 kWh/year

TABLE 2.2. Parameter Values for Baseline, Near-Term, and Medium-Term Time Frames Parameter

Symbol

Value

Units

Baseline

Near Term

Medium Term

Module efficiency

ηm

0.135 (13.5%)

0.16 (16%)

0.20 (20%)

— (%)

Module cost

CW

4

2.2

1.25

$/W

Sunlight energy density/year

SEY

2435

2435

2435

kWh/m2/year (Phoenix)

Area-related BOS

Cb

307

155

150

$/m2

DC to AC inverter cost

Ci

900

690

300

$/kW

Inverter efficiency

ηi

0.95 (95%)

0.96 (96%)

0.97 (97%)

System lifetime

LTS

30

35

35

Years

Inverter lifetime

LTi

5

10

20

Years

System derating efficiency

ηD

0.95 (95%)

0.95 (95%)

0.95 (95%)

— (%)

I

2.50

1.21

0.55

$/W

0.06 (6%)

0.06 (6%)

0.06 (6%)

1/year (%/year)

Government incentive Loan rate

— (%)

Loan amortization rate system

mS

0.0726 (7.26%)

0.069 (6.90%)

0.069 (6.90%)

1/year (%/year)

Loan amortization rate inverter

mi

0.237 (23.7%)

0.136 (13.6%)

0.0872 (8.72%)

1/year (%/year)

Interest fraction of amortization system

fS

0.826 (82.6%)

0.870 (87.0%)

0.870 (87.0%)

— (%)

Interest fraction of amortization inverter

fi

0.253 (25.3%)

0.442 (44.2%)

0.688 (68.8%)

— (%)

Tax bracket

B

0.28 (28%)

0.28 (28%)

0.28 (28%)

— (%)

Additional fixed charge rate

Δ

0.0442 (4.24%)

0.0442 (4.24%)

0.0442 (4.24%)

1/year (%/year)

Fixed charge rate system

FS

0.100 (9.82%)

0.0964 (9.64%)

0.0964 (10%)

1/year (%/year)

Fixed charge rate inverter

Fi

0.265 (26.3%)

0.163 (13.7%)

0.115 (11.8%)

1/year (%/year)

Indirect cost rate

r

0.225 (22.5%)

0.225 (22.5%)

0.225 (22.5%)

— (%)

Levelized cost of energy

L

0.32

0.15

0.09

$/kWh

COST ESTIMATES

33

is 1000 W/m2. It is a standard accounting technique to multiply direct costs by an indirect rate, r, to include overhead and general and administrative costs including yearly operations and maintenance costs. For the expressions used here, it was assumed that the majority of the balance-of-systems (BOS) costs (including installation costs but excluding inverter costs) were area related. Hence, any BOS given in power-related form was converted into area-related form using the conversion relationship that is four lines from the bottom of Table 2.1. It is noted that any first-year government incentives I offered in terms of dollar per watt rebates simply subtract directly from the capital cost of the modules CW in dollar per watt. The final L expression is given three lines from the bottom of Table 2.1. The module cost CW expression on the bottom line came just from solving the L expression for CW in terms of all the other parameters. The baseline calculated costs for residential solar cell system electricity of $0.32 per kilowatt hour in Phoenix, Arizona, are given on the third column bottom line of Table 2.2. The baseline input parameter values for module efficiency, module costs, BOS costs, inverter costs, inverter efficiency, system derating efficiency, system lifetime, and inverter lifetime came from the U.S Solar Energy Technology Multiyear Program Plan 2007–2011’s [40] 2005 benchmark numbers obtained from 200 residential installations between 2000 and 2005. The loan rate and tax bracket were taken from the 2008–2012 plan [41]. The indirect cost rate and the system fixed charge rate were taken from the EPRI values [39] and its 22.5% value includes the 0.5% operations and maintenance charge used in the Solar Advisor Model. The Δ value (adding to indirect charge rate) was chosen so that the calculated FS value equaled the EPRI value of 10%. The Phoenix annual solar energy density SEY was taken from Riordan’s Chapter 20 of the First Edition [36]. As a reality check, the fixed charge rate for a 2007 home loan in Mountain View, California, was calculated as shown in Table 2.3, and its value of 9.13% (for PITI) is in this general range. The government incentive of $2.50 per watt was chosen to give the same $0.32 per kilowatt hour L value as reported in the SETP 2007–2011 plan [40]. It is in the approximate range of the most beneficial rebates offered by the state of California.

TABLE 2.3. Fixed Charge Rate F for the Purchase of a $400,000 Home with 20% Down and a $320,000 30-Year Mortgage at a 6.375% Loan Rate in Mountain View, California, in 2007 Parameter

Monthly Cost ($)

Yearly Cost ($)

Normalized Yearly Cost per Loan ($)

1995.0

23,935.0

0.0748 (7.48%)

375.0

4500.0

0.0141 (1.41%)

766.0

0.0024 (0.24%)

Principal and interest Taxes Insurance Fixed charge rate F (total)

63.85

0.0913 (9.13%)

34

SOLAR CELL ELECTRICITY

Levelized Cost of Energy L ($/kW-hr)

One of the advantages of the simple Table 2.1 expressions is the ease with which they can be used for sensitivity analyses. Figure 2.8 shows such a sensitivity plot of the levelized cost L to different module efficiency values. For the baseline case with a module efficiency of 13.5% and an L value of $0.32 per kilowatt hour (shown by a vertical line), one can see the beneficial effects of greatly increased module efficiency rapidly saturate with a sharper increase in L costs with decreased module efficiency. Similarly, Figure 2.9 shows the sensitivity of allowed module 0.5

0.4

0.3 Base Line

0.2 Near Term

0.1 Medium Term

0 0

10

20

30

40

50

60

Module Efficiency ηm (%)

Figure 2.8. The levelized cost of energy L as a function of module efficiency ηm for the baseline, near term, and the medium term obtained from the expressions of Table 2.1 using the input parameter values of Table 2.2. 6

Base Line $0.32/kW-hr

Module Cost CW ($/W)

5 4 Near Term $0.15/kW-hr

3 2

Medium Term $0.9/kW-hr

1 0 -1 -2 0

10

20

30

40

50

Module Efficiency ηm (%)

Figure 2.9. The allowed module cost CW obtained by holding all the parameters constant, except for module efficiency ηm and module cost CW, using the other parameter values of Table 2.2 for the baseline, near term, and the medium term calculated with the expression at the bottom of Table 2.1.

COST ESTIMATES

35

costs in dollar per watt (around the baseline $4 per watt value indicated by the vertical line) with module efficiency but with all the other baseline parameters held constant including L at $0.32 per kilowatt hour. Here, higher module costs can be supported with higher module efficiencies, but again this rapidly saturates. At lower module efficiencies, the module cost that can be justified drops sharply. For module efficiencies below 5%, the constant energy cost value L cannot be supported even if the module costs went to zero. This effect is due to the BOS cost described by the Cb parameter in the Cw expression at the bottom of Table 2.1. One conclusion from these sensitivity analyses is that there is no one parameter whose improvement can make the cost of solar cell electricity competitive with current electricity rates. Hence, two more cases were considered and they are the “near-term” and “medium-term” cases shown in Table 2.2 and in Figures 2.7 and 2.8 that respectively correspond to L values of $0.15 per kilowatt hour and $0.09 per kilowatt hour, module efficiency values of 16% and 20%, and module costs of $2.20 per watt and $1.25 per watt plus all the other optimistic parameter values listed in Table 2.2. These parameter values were taken from the “2011” and “2020” projections of the SETP 2007–2011 tables [40] plus similar listed adjustments to the remaining parameters. It should be noted that the SETP 2008–2012 table and plans [41] accelerated such improvements in to the “2010” and “2015” time frames. In this context, “near term” corresponds to the 2010–2011 time frame and “medium term” corresponds to the 2015–2020 time frame. Again, the corresponding curves of Figure 2.8 illustrate the positive but saturating benefits of increased module efficiencies and the sharper penalties for lower module efficiencies. Here, Figure 2.9 shows the even stronger penalties for reduced module efficiencies with the L costs held constant at increasingly competitive values. For the specific $0.09 per kilowatt hour case considered here, module efficiencies below 8% could not be justified even if their costs were zero. Further, at a module price of $2 per watt, the calculated minimum module efficiency required to be competitive is 6% for the baseline case and 14% for the near-term cases but is not achievable for the medium-term case. Similarly, at a module price of $1 per watt, the calculated minimum efficiencies required to be competitive are 5%, 6%, and 15%, respectively, for the baseline, near-term, and medium-term cases. A bottom line here is that decreased module costs are essential for achieving competitive system performance but that lower-cost modules still have to have efficiencies relatively near the more expensive modules of higher performance to provide competitive, levelized cost values. All of these emphasize the importance of resuming the Figure 2.3 module cost learning curve reductions that most recently have become quite flat. The breakdown of the three major system components and their contributions to the levelized costs L are shown in Table 2.4 for the baseline, near-term, and medium-term cases. For all three, the module costs contribute in the 70s percent range; the area-related BOS contributions are in the mid-30s to low 40s percent range, with the inverter contributions in the 30s–40s percent range. The resulting L values are only possible due to government incentives that pick up 30–40% of the total component contributions. This is the case even though the Table 2.2

36

SOLAR CELL ELECTRICITY

TABLE 2.4. Individual Parameter Contributions to the Levelized Cost of Energy L Parameter

Symbol

Contribution to L ($/kWh and %) Baseline

Near Term

Medium Term

Module cost ($/W)

CW

0.235 (73.6%)

0.117 (77.8%)

0.066 (72.7%)

Area-related BOS ($/m2)

Cb

0.134 (41.8%)

0.052 (34.2%)

0.039 (43.6%)

DC to AC inverter cost ($/kW)

CI

0.098 (30.6%)

0.046 (30.8%)

0.014 (15.6%)

Government incentive ($/W)

I

–0.147 (−46.0%)

–0.064 (−42.8%)

–0.029 (−32.0%)

Levelized cost of energy

L

0.320 (100%)

0.150 (100%)

0.090 (100%)

parameter assumptions show the actual government contributions in dollar per watt falling by more than a factor of 3 over this period. The SETP 2008–2012 Plan [41] and its Solar Advisor Model actually have the government incentive going to zero by the medium term. If these government incentives were all zero, the respective L values would be $0.467 per kilowatt hour, $0.214 per kilowatt hour, and $0.119 per kilowatt hour for the baseline, near-term, and medium-term cases (from the addition of the top three contribution rows of Table 2.4). It is such amounts that would be appropriate for “feed-in tariff ” situations. A reality check of this zero-incentive “2005” baseline value is that it is on the order of the $0.75 per kilowatt hour long-term contracts that Germany’s programs signed in the early 2000s to make their “feedin” programs financially attractive. Indeed, it is such feed-in tariff scenarios that market history indicates have been the most effective in stimulating rapid adoption and in promptly moving new geographic regions into higher world market ranking positions. The above near- and medium-term “zero-incentive” L values seem particularly reasonable for this latter type of public policy support strategy.

2.5

CONCLUSIONS

The 1000-fold growth of the world’s production of solar cell electricity in the 30 years since 1975 resulted from a growth rate that doubled output every 2 years for the first and last 10 years of this period. The middle 10 years had a growth rate reduced by about a half. Another 1000-fold growth in production would provide an electricity production level equivalent to the world’s total installed generation capacity of 3889 GW in 2005. Should a growth rate of doubling every 2 years be maintained, the latter capacity levels would be reached in the decade between 2020 and 2030.

CONCLUSIONS

37

The technical and manufacturing expertise required to support such developments is now spread throughout the world. The challenges to achieving these results are likely on the same order as those faced by the Manhattan Project that produced the first nuclear weapons or by the National Aeronautics and Space Administration’s program that placed the first men on the moon. These challenges include matching the availability of solar cell electric power with demand, continued reductions in the costs of producing electricity from solar cell systems (probably by another factor of 4 from baseline values) in ways that do not deplete basic resources needed for their construction and deployment, that does not seriously degrade the environment, and in a time frame that meets the inevitable peaking and decline of energy provided from fossil fuels. The leadership position in this technology originated in the United States but has largely shifted to the European Community and to Japan and to China over the last 10–15 years. Although the remaining challenges are daunting, the expertise to overcome them is now developing throughout the world. Changes in the rankings of dominating regions leading this technology will get increasingly expensive as the cumulative world market continues to expand. Those with the wisest and boldest strategies will likely be the major benefactors. The political entities with the highest capabilities for playing major roles in the rapidly evolving world market for solar cell electricity are those backed by the largest economies. In current order of size, these economies are those of (1) the EU, (2) the United States, (3) China, (4) Japan, and (5) India [17]. The market history-demonstrated and the best-estimated public policy approaches for promptly changing rankings among the leaders in the world market, in descending order of effectiveness, are (1) feed-in tariffs; (2) depletable energy user taxes applied to renewable energy development and deployment; (3) government-funded incentives that pay a substantial portion of the capital costs in the first year of installation; (4) long-term (30 years), low-interest rate loans often funded with tax-free bonds; (5) net metering particularly when extended to total metering so that the electricity excess delivered to a utility is paid for at least at 30–40% of the retail delivered costs; and (6) other tax incentives that include loan interest write-offs at constant interest rates over 30-year time periods. Several of these six listed approaches contain features that overlap. Of the three major options for resuming module cost reductions with market growth, key to the large-scale adoption strategy described in this chapter, is the concentrator option that has been the least explored so far and that needs the greatest assistance in making the transition from research into large-scale commercial availability and into clear feasibility demonstrations. Concentrator technology’s failure so far to transition into commercial availability is likely due in part to the “Valley of Death” syndrome described above in the Public Policy section, and that is endemic to all new technologies trying to break into large established markets. The EU is beginning to demonstrate that different countries with adjacent borders, speaking various languages, and with a spectrum of individual cultures, economic development levels, and technology competencies can join into cooperative trade, economic, consumer, and political arrangements that thrust such unions

38

SOLAR CELL ELECTRICITY

into a world leadership position, as ranked by economic metrics in general and by renewable energy implementations and deployments in particular. At its peak, the British Empire demonstrated that a dominant economic power base and trading entity could be assembled from widely spread countries whose participations were established and maintained through military “support.” A key challenge of the twenty-first century may be whether other “adjacent” and “nonadjacent” but diverse countries can cooperate through similar unions motivated primarily by economic self-interest and mutual well-being. For the latter unions to remain competitive with the large and rapidly growing economies like China’s and India’s, the involved “union” populations likely need similar sizes and mobilities and similar lack of restrictions. This latter likely includes ready access to nearly “comparable” education and training for the involved populations plus the latter’s ability to promptly relocate to job and professional opportunities unfettered by overly restrictive or artificial constraints such as those based on nationality, class, race, religion, or sex. As a nation composed essentially of diverse immigrant populations, the United States may have some advantages in the latter areas but with a remaining question of how much immigration is needed, desirable, and politically acceptable. A key to competitive solar cells and systems manufacturing in more developed countries like the EU, the United States, and Japan is automation. The fabrication equipment will likely cost about the same no matter where it is located. A Pacific Rim advantage is lower-cost labor that automation can offset to a large degree. Competitive plant construction costs including land in developed countries can likely be accommodated by appropriate tax breaks and incentives. This would offer reduced employment opportunities in manufacturing, but it would retain the product sales and factory construction revenues so that the involved country could avoid balance of payment deficits that would otherwise develop. “Champion” cells and modules are very important demonstrations of what is possible with current technology, but their performance numbers must be used with caution. When a fabricator has made and tested 10,000 devices, the very “best” by definition occurs with a yield of one in 10,000. Such a performance extreme does not necessarily provide accurate estimates of what can actually be achieved on a large scale in the near term at reasonable costs. In the past, such differences between the average achieved in large installations and in solitary “champion” devices have been as high as a factor of 4X to 5X with the differences exacerbated with attempts to provide low costs using less understood or developed materials systems (see Chapter 1 in Reference 36). When the price paid for solar cells is high using very well understood and developed materials systems, such as with space solar cells that can range in price from $300 per watt to $1800 per watt, the difference between “champion” and average cell performance can drop into the 30% or even lower range (see Chapters 4 and 23 in Reference 36). The terrestrial challenge is to simultaneously provide relatively low costs and high performance. Concentrators are an approach to using “expensive” and well-developed cell materials technologies with potentially low-cost optics to provide such a simultaneous result. However, this approach has not yet been well developed or tested

ABBREVIATIONS

39

experimentally. Early data show that the output of a two-axis tracking module of a high 17% efficiency produces 99% more kilowatt hour per square meter per year in a sunny Palo Alto, California environment in a side-by-side comparison with a near horizontal (5 ° fixed tilt) 12% efficient module (Fraas and Partain, Chapter 26). Recent tests of a 3X concentrator, two-axis tracking module of crystalline silicon interdigitated back contact cells showed that side by side, it had comparable output in kilowatt hour per square meter to a fixed axis module of similar cells in Shanghai, China, but only in measurements for a single day (L. Fraas, pers. comm.). Unfortunately, most of the current concentrator demonstration sites do not include non-concentrator modules, for direct side-by-side comparisons in terms of kilowatt hour per square meter per year outputs. Theoretical calculations of the differences are beginning to become available (Gueymard, Chapter 19). However, such calculations have not yet been carefully validated with experimental data. Their accuracy is unlikely to be trusted until such validations have been completed. Sensitivity analysis illustrates that there is no one parameter whose improvement will lead to the widespread practical implementation of solar cell electricity. However, module cost reductions (in dollar per watt) and efficiency improvements (in percent) will play dominant roles. However, major reductions in inverter costs and improvements in their lifetimes (by factors of 3X to 4X according to the Table 2.2 estimates) will also be needed. Among the other major challenges are economical means for energy storage in combination with time-of-use schemes that will be required for large-scale penetration into the traditional electricity energy market over the longer term. Most of the basic scientific understanding and principles needed to accomplish this are essentially already known. It is the engineering applications and their optimizations and scale-up that remain Herculean in scope and in magnitudes. This a situation largely shared with that of the U.S. Manhattan Project and the U.S. Man-on-the Moon Program in their early days. However, with sufficient forethought, the extremely short time frames and major diversion of resources of the Manhattan Project should not be required. While never predictable, research breakthroughs will accelerate at least some small parts of this effort. It is no longer a question whether this new energy source transition will occur. It is only a question of who will lead the process and who will reap the most benefits.

ABBREVIATIONS AC—alternating electric current CdTe—cadmium telluride DC—direct electric current EIA—U.S. Energy Information Administration EU—European Union EVA—ethylene vinyl acetate FSA—Flat-Plate Solar Array Project

40

SOLAR CELL ELECTRICITY

GaAs—gallium arsenide HIT—heterojunction with intrisic thin-layer solar cell JPL—U.S. Jet Propulsion Laboratory MITI—Ministry of International Trade and Industry (Japan) NEDO—New Energy and Industrial Technology Development Organization (Japan) PMT—name of the Microsoft Excel function that calculates mortgage payments as a function of interest rates and mortgage periods for given loan amounts PURPA—U.S. Public Utility Regulatory Policies Act R & D—research and development SETP—U.S. Solar Energy Technology Mulityear Program X—multiplier REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

J. G. Dorn. Solar Cell Production Jumps 50 Percent in 2007, Earth Policy Institute. Available at http://www.earth-policy.org/indicatiors/solar/2007.htm. Accessed March 9, 2008 (2007). Exxon Mobile Corporation. Annual Report. Available at http://thomson.mobular. net/thomson/7/2677/3201/. Accessed September 6, 2008 (2007). Energy Information Administration, U.S. Government, Existing Generating Units in the United States. Available at http://www.eia.doe.gov/cneaf/electricity/epa/ epat2p2.html. Accessed September 6, 2008 (2007). Energy Information Administration, U.S. Government, Table H1, World Total Installed Generating Capacity. Available at http://www.eia.doe.gov/oiaf/ieo/ieoecg. html. Accessed September 6, 2008 (2008). L. Stoddord and R. Pletka. CEC Workshop on Renewable Energy. Available at http://www.energy.ca.gov/2004_policy_update/documents/2004_06_08_BLACK_ VEATCH.pdf. Accessed September 6, 2008 (2004). R. Wiser and M. Bolinger. Projecting the Impact of State Portfolio Standards on Solar Installaltions, slide 4. Available at http://www.cleanenergystates.org/library/ ca/CEC_wiser_estimates.0205.pdf. Accessed September 6, 2008 (2005). W. Wallace. Government terrestrial acceleration programs. In Solar Cells and Their Applications, L. Partain, ed., p. 495. New York, Wiley (1995). T. Tomita. Prog. Phototovolt Res. Appl. 13, 471–479 (2005). M. A. Green, K. Emery, and D. L. King. Solar cell efficiency tables (version 29). Prog. Photovolt. Res. Appl. 15, 15–40 (2007). V. Salas and E. Olias. Overview of the photovoltaic technology status and perspective in Spain. Renewable and Sustainable Energy Reviews. Elsevier, London. doi:10.1016/j.rser.2008.03.011 (2008). C. Christensen. The Innovator’s Dilemma. Boston, Harvard Business School (1997). E. Christensen. Flat plate solar array project. US Department of Energy Report, Jet Propulsion Laboratory, Pasadena, CA (1985). W. Callahan and R. McDonald. Flat-Plate Solar Array Project Final Report, JPL Publication 86-31, 5101-289, DOE/JPL 1012-125, Jet Propulsion Laboratory, Pasadena, CA (1986). Bureau of Economic Analysis, U.S. Government, Table 1.19 Implicit Price Deflators for Gross Domestic Product. Available at http://www.bea.gov/national/nlpaweb/ TableView.asp#Mld. Accessed September 8, 2008 (2008).

REFERENCES [15] [16] [17]

[18]

[19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

41

E. Martinot, R. Wiser, and R. Hamrin. Renewable Energy Policies and Markets in the United States. Available at http://www.resource-solutions.org/lib/librarypdfs/ IntPolicy-RE.policies.markets.US.pdf. Accessed September 13, 2008 (2005). New York Federal Reserve Bank, US$ Foreign Exchange Rates. Available at http:// www.ny.frb.org/markets/fxrates/historical/fx.cfm. Accessed September 10, 2008 (2008). Central Intelligence Agency, U.S. Government, The World Fact Book, Rank Order—GDP (purchasing power parity). Available at http://www.cia.gov/library/ publications/the-world-factbook/rankorder/2001rank.html. Accessed September 18, 2008 (2008). Fortune magazine. A catalyst for change, the Japan of the future, pp. S10-S11, Special Advertising Section, July 21. Available at http://www.timeinc.net/fortune/ services/sections/customprojects/sections/071126_Japan2.pdf. Accessed September 16, 2008 (2008). Military-Industrial Complex Speech, Public Papers of the Presidents, Dwight D. Eisenhower, 1960, pp. 1035–1040. Available at http://coursesa.matrix.msu. edu/∼hst306/documents/indust.html. Accessed September 27, 2008 (1961). D. E. Kash. Perpetual Innovation, Chapter 6. New York, Basic Books (1989). FY 2008 Budget Request for Defense S&T, FYI. AIP Bulletin of Science Policy News, February 13, 2007, No. 23, pp. 1–2 (2007). SolarBuzz. German PV Market. Available at http://www.solarbuzz.com/fastfactsgermany.htm. Accessed September 6, 2008 (2007). A brief history of the California economy. Available at http://www.dof.ca.gov/ HTML/FS_DATA/HistoryCAEconomy/index.htm. Accessed November 22, 2008 (2008). World Future Council, Feed-in Tariffs. Available at http://worldfuturecouncil.org/ fileadmin/user_upload/Maja/Feed-in_Tariffs_WFC.pdf. Accessed September 16, 2008 (2008). Wikipedia. Feed-in tariff. Available at http://en.wikipedia.org/wiki/Feed_in_Tariff. Accessed November 22, 2008 (2008). SolarBuzz. 2007 World PV Industry Report Highlights. Available at http://www. solarbuzz.com/Marketbuzz2008-Intro.htm. Accessed September 7, 2008 (2008). NEI Nuclear Notes, Italy Nuclear Update. Available at http://www.neinuclearnotes. blogspot.com/2006/02/italy-nuclear-update.html. Accessed September 15, 2008 (2006). China to become world’s largest economy by 2010, Finance Daily. Available at http://financedialy.co.uk/News/Chinatobe WorldsLargestEconomyby2010_448. html. Accessed September 27, 2008 (2008). The World’s Largest Economies in 2050, University of Phoenix, based on 2003 Goldman Sachs study. Available at http://www.everthing2.com/index.pl?node_ id=1756826. Accessed September 27, 2008 (2005). 50th Anniversary of the Interstate Highway System. Available at http://www.fhwa. dot.gov/interstate.history.htm. Accessed August 22, 2008 (2006). US oil import bill to top $400 billion this year. Available at http://www.reuters.com/ article/pressRelease/idUS236508+07-Mar-2008+BW20080307. Accessed September 7, 2008 (2008). Bureau of Economic Analysis, U.S. Government, U.S. International Transactions: First Quarter 2008. Available at http://www.bea.gov/newreleases/rels.htm. Accessed September 7, 2008 (2008). The Cost of Iraq, Afghanistan, and Other Global War. Available at http://zfacts.com/ metaPage/lib/CRS-Belasco-2006-09-Iraq-Costs-RP33110.pdf. Accessed September 7, 2008 (2008).

42 [34] [35] [36] [37] [38] [39] [40]

[41]

SOLAR CELL ELECTRICITY Interstate Cost. 50th Anniversary of the Interstate Highway System Frequently Asked Questons. Available at http://www.fhwa.dot.gov/interstate/faq.htm# question6. Accessed November 20, 2008 (2006). S. Thompson and S. Pathasarathy. Materials Today 9, 20–25 (2006). L. D. Partain, ed. Solar Cells and Their Applications. New York, Wiley (1995). Earth radius. Available at http://en.wikipedia.org/wiki/Earth-radius. Accessed November 22, 2008 (2008). Energy Information Administration, U.S. Government, Table H6, World Installed Hydroelectric and Other Renewable Generating Capacity. Available at http://www. eia.doe.gov/oiaf/ieo/ieoecg.htm. Accessed September 6, 2008 (2008). F. R. Goodman, J. C. Schaefer, and E. A. DeMeo. Terrestrial, grid connected systems. In Solar Cells and Their Applications, Chapter 16, L. D. Partain, ed. New York,Wiley (1995). US Department of Energy, Solar Energy Technologies Program, Multi Year Program Plan 2007-2011. Available at http://www.pv-era.net/doc_upload/documents/218_0 084aSolarEnergyTechnologiesProgram2007-2011_proof.pdf. Accessed October 12, 2008 (2007). US Department of Energy, Solar Energy Technologies Program, Multi Year Program Plan 2008-2012. Available at http://www1.eere.energy.gov/solar/pdfs/ solar_program_mypp_2008-2012.pdf. Accessed October 26, 2008 (2008).

3 SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS, AND HIGH EFFICIENCY LEWIS FRAAS JX Crystals Inc.

3.1

INTRODUCTION

How efficient can a solar cell be and how much power and energy can it produce and at what cost? These are the fundamental questions addressed in this chapter. In order to address these questions, it will be necessary to first describe the nature of sunlight and then the nature of semiconductors. It will be shown that singlecrystal semiconductors are necessary for high sunlight conversion efficiency and that high conversion efficiency is important for lower-cost solar electricity. In addition to silicon solar cells, a new class of semiconductors based on LED materials will also be discussed. The reader is familiar with LEDs and the fact that they come in different colors. It is noted herein that LEDs are just solar cells running in reverse. In other words, in a LED, electricity goes in and light comes out. In a solar cell, light goes in and electricity comes out. It will be shown that the LED class of semiconductors can be used to make multicolor or multijunction solar cells where the multicolor feature is critical for making solar cells with energy conversion efficiencies as high as 40%. While it is true that high-efficiency single-crystal solar cells are more expensive to make than amorphous or small grain size polycrystalline thin-film cells, fortunately, it is possible to use inexpensive mirror or lens materials to collect the sunlight and to focus the solar energy on these single-crystal converters thereby diluting their cost. This is the concentrated sunlight PV or CPV approach. The newer 40% cells are even more complex and expensive than single-crystal silicon Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

43

44

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

cells requiring higher-concentration HCPV systems, whereas the silicon singlecrystal cells can use lower concentration LCPV simpler systems. This chapter will then conclude with a discussion of the future cost potential for the various solar PV options comparing the potential future system-level cost for the conventional planar silicon cell and thin-film cell approaches with the concentrated sunlight approaches.

3.2

SUNLIGHT, RAINBOWS, AND PHOTONS

How much energy is in sunlight? The answer to this question requires a definition of some terms. The three terms are power, solar intensity, and energy. Power is in watts and solar intensity is in watts per unit area. Energy is in kilowatt hours. Homeowners pay for energy in kilowatt hours. Solar modules and arrays produce power in watts. The relevant measure of solar radiation is solar intensity. On a nice, sunny day at noon, the solar intensity is usually around 1 kW/m2. One square meter is close to 11 ft2. How does a scientist describe sunlight? The observation of rainbows proves that sunlight can be divided into different colors. Also, when a group of very fine parallel lines are scribed close to each other to make a grating, it is observed that the colors can be correlated with line spacing. This means that there is a wavelength connected to each color. So, light is an electromagnetic wave as shown in Figure 3.1, just like radio waves and microwaves with a wavelength, λ, and electric field, E. The history of the study of light is interesting. Actually, Newton thought of light as particles. However, with grating experiments in the 1800s, it was decided that light was an electromagnetic wave. Then, in about 1905, Einstein looked at the photoelectric effect and said that light comes in small energy packets called

E

Wave length

Figure 3.1. Electromagnetic wave.

SUNLIGHT, RAINBOWS, AND PHOTONS

45

quanta, which behave somewhat like particles. Einstein won the Nobel Prize for this work [1], not for his theory of relativity. Einstein noted that when the light of a certain color hits a metal, all of the electrons that escape the metal have the same peak energy and that when the light intensity is increased, the number of electrons increases, not the energy of each electron. It is like the energy in wave theory is E2, but the energy also equals nλ × eλ, where nλ is the number of photons and eλ is the energy of each photon. So, one can think of the electromagnetic wave intensity as getting smaller and smaller until one discovers that it is coming in discrete particles. It occurs in very fine steps. So now, the sun’s intensity spectrum can be divided into color or wavelength intervals as in Figure 3.2. To understand how solar cells work, first note that a photon’s energy is inversely proportional to its wavelength. This just means that shorter wavelength photons at the left in this curve have more energy than the longer wavelength photons at the right. Electric power is the product of voltage and current. In a solar cell, the power results when electrons pass through a voltage. It is the nature of semiconductors that each semiconductor material has a threshold absorption energy that then controls the voltage the electrons see. It is convenient then to talk in units called electron volts. An electron volt is the energy that an electron produces when it moves through a potential of 1 V. Photon energies and material threshold absorption energies are measured in electronvolt. For example in Figure 3.2, a red photon with an energy of 2 eV has a wavelength of about 0.6 microns, and the absorption threshold energy for silicon, called its bandgap energy, is at 1.1 eV, which equates to 1.1 microns (1 micron = 1 micrometer = 1 millionth of a meter = 1 μm). Now the efficiency limits can be seen. First, in Figure 3.2, all photons with energies less than 1.1 eV are lost; that is, all solar energy with wavelengths greater 1600

Watts/m2/micron

1400 1200 1000 800 600 400 200 0

0

0.5

1

1.5

2

2.5

Wavelength (microns)

Figure 3.2. Sun’s spectrum at AM1.5. The gray region is the usable photon energy for a silicon solar cell.

46

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

than 1.1 microns is lost because these photons are not absorbed. Next, for photons with energies large enough to be absorbed (left side of figure), each photon absorbed must create an electron that can move to generate voltage. The yield here should be one electron per photon, but the actual number is less and is called the quantum efficiency or QE. QEs of 90% occur in pure single-crystal materials. QEs in non-single-crystal materials are lower. So, single-crystal semiconductors are important for high efficiency as will be discussed in the next sections of this chapter. For now, high QEs in single-crystal materials are assumed. These moving electrons create the current. To calculate the current, the photons absorbed are simply counted. However, note that while a photon might have 2 eV of energy, it can only create an electron moving through a voltage of no more than the bandgap energy. For silicon, this is 1.1 eV. So, 2-eV photons loose at least 2.0 – 1.1 = 0.9 eV. The gray region in Figure 3.2 shows the energy that can be captured from the higher-energy photons. This now leads to a way to make higher-efficiency solar cells using multiple materials. Two stacked materials can absorb the lower-energy photons in one material generating a lower voltage and the higher-energy photons in a second material generating a higher voltage. The result is more photons better used. This concept is shown in Figure 3.3. The gray regions in Figure 3.3 show that more photon energy can now be captured. To implement this two-junction or two-color cell concept, one must choose materials wisely. The GaAs/GaSb two-color cell (2) demonstrated in 1989 uses two simple materials with bandgap energies of 1.4 eV (0.9 micron) and 0.7 eV (1.8 micron). Theoretically, the efficiency for this pair can be as high as 41%. Thirtyfive percent has been demonstrated [2].

1600

Watts/m2/micron

1400 1200 1000 800 600 400 200 0

0

0.5

1 1.5 Wavelength (microns)

2

2.5

Figure 3.3. GaAs/GaSb and solar spectrum. The GaAs absorbs higher-energy photons at the left, generating higher voltage. The GaSb absorbs lower-energy photons at the right (1 micron = 1 micrometer = 1 millionth of meter).

ELECTRONS IN ATOMS AS WAVES AND THE PERIODIC TABLE

47

For the more general reader who may not care about the chemical names, note that the top GaAs cell actually looks blue and the bottom GaSb cell actually looks red. So, one can also refer to the cells in multicolor cell stacks as blue cells glued on top of red cells. The chemical terms will be used through most of this book for the benefit of technical readers because the chemical names are more precise. More information on multijunction concentrator cells will be presented later in this chapter and in Chapter 13.

3.3 ELECTRONS IN ATOMS AS WAVES AND THE PERIODIC TABLE OF THE ELEMENTS There are several different types of solar cells made from materials ranging from single crystals to amorphous silicon. The goal here is to describe the different types of solar cells and their advantages and limitations. A fundamental description of the nature of semiconductors is presented, beginning with electrons in atoms as waves. The discussion of electrons as waves then leads to a description of semiconductors as single crystals. The theory of single-crystal semiconductors is then used to describe how diodes and solar cells work. A discussion of the effects of various defects in semiconductor materials on solar cell performance follows. The reader will see that the performances enumerated are consistent with the simple concepts presented. This chapter explains why high-efficiency cells require good single-crystal materials. In the last section, it was noted that the sun’s rays are really electromagnetic waves with varying wavelengths. Electromagnetic radiation includes radio waves, microwaves, and infrared, visible, and ultraviolet waves. When one thinks about longer wavelength radiation like radio waves, one always thinks about waves. However, for the shorter wavelengths associated with infrared and visible light, physicists start to talk about photons. A photon is like a particle or wavelet having a specific wavelength and energy. A photon is a quantum of energy or discrete packet of energy. Now, is radiation a wave or a particle? The answer is both! This is the wave–particle duality, a subject called quantum mechanics [3], a subject normally taught in graduate school physics classes along with a lot of mathematics. However, the key ideas can actually be described in simple nonmathematical terms, and these ideas are important to the understanding of solar cells. While electromagnetic radiation is normally thought of as waves, one generally thinks of electrons as particles circling an atomic nucleus just as planets circle the sun. However, an atom is really extremely small, so small that in crossing a human hair, one will pass by 200,000 atoms. Intuition based on everyday experience fails at this small size. It turns out that electrons around atomic nuclei are described by wave functions. Here is the wave–particle duality again. However, the rules that govern electrons in atoms and solids can be described in fairly simple terms. In Figure 3.4, start with the simple hydrogen atom [4, 5] with a single negatively charged electron and a single positively

48

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Nucleus

D Px, Py, Pz

4103 A 4342 A 4563 A

S

E X

6565 A

Figure 3.4. Left: potential well for an electron around the nucleus in an atom with energy level S, P, and D wave functions. Right: a spectral line sequence for hydrogen.

charged proton. The oppositely charged proton and electron attract each other and as they get closer and closer to each other, it is harder and harder to pull them apart. The electron is said to be in an energy well or a potential well as shown on the left of Figure 3.4. The question then is can the electron collapse down and sit on the proton? The answer is no. Why not? Scientists have observed the electromagnetic spectra emitted by atoms and find discrete wavelengths and energies as shown on the right in Figure 3.4. Not all energies are possible. How is this explained? Scientists hypothesize that the electron position is described by a wave function that then gives its probable position. Since one knows that the electron cannot be outside the potential well, one knows the wave functions have to be zero outside the well. Now, observe that the waves will have to have one, two, and three nodes as is shown in the wells at the left in Figure 3.4. For historical reasons, the state with one peak node is labeled S, and the states with two nodes are labeled Px, Py, and Pz. (x, y, and z are the three directions in three-dimensional space.) The next rule is that electrons can have positive and negative spins, and only one electron can occupy each state. So there will be two S states with opposite spins and two Px, two Py, and two Pz states for a total of eight state configurations possible. This wave hypothesis has proven to be very successful as it explains atomic spectra and the periodic table of the elements [6] and all of chemistry. The rule of eight including S and P orbits explains the second and third rows of the periodic table. Table 3.1 summarizes important features of the periodic table including the common commercial semiconductor materials. The D level transition metals are not shown since they are not relevant here. Compounds are formed from the elements in the periodic table. For example, table salt is a compound containing Na (sodium) and Cl (chlorine) and is written as NaCl. The two semiconductor

SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY

49

TABLE 3.1. Periodic Table of the Elements I

II

III

IV

V

VI

VII

H Hydrogen

VIII He Helium

Li Lithium

Be Berilium

B Boron

C Carbon

N Nitrogen

O Oxygen

F Fluorine

Ne Neon

Na Sodium

Mg Magnesium

Al Aluminum

Si Silicon

P Phosphorus

S Sulfur

Cl Chlorine

Ar Argon

Ga Gallium

Ge Germanium

As Arsenic

In Indium

Sb Antimony

compounds making up the two color solar cell stack described in Figure 3.3 are gallium arsenide (GaAs) and gallium antimonide (GaSb).

3.4 SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY Why is it important to know about electrons as waves? The answer is that waves are intrinsically periodic as are the atom locations in single crystals. It is this periodicity that makes semiconductors special. Historically, the semiconductor revolution started 80 years ago with the discovery of the importance of high-purity single crystals and the wave theory of solids. The first application of the electron wave theory was by A. Sommerfeld in 1928 [7]. As shown in Figure 3.5a, he described the motion of electrons in metals by assuming the electrons moved in a flat-bottom energy well bounded by the metal surfaces. Any wavelength would be possible, leading to a set of conduction band energy levels. The real breakthrough came in 1928 with F. Bloch [8], who then modeled a periodic single crystal with a set of periodic energy wells representing the atom positions within the larger energy well as shown in Figure 3.5b. He showed that there are then two or more energy bands separated by energy gaps. The states with wave functions centered on the atomic wells represent valence electrons, and the states with wave functions centered between the atomic wells represent conduction band electrons. A key point is that any electron can be near any atom in the crystal. A. H. Wilson [9] in 1931 then provided the elementary theory of semiconductors by noting that if the material is pure enough and if the valence band is completely full and the conduction band is completely empty, then one has a semiconductor. An intrinsic semiconductor is defined as having zero conductivity at zero temperature with increasing conductivity as the temperature increases. The increasing conductivity comes about as electrons are thermally excited into the conduction band. This is opposite from metals where the number of conduction

50

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Ec (a) Metal

Ec Ev (b) Semlconductor

Figure 3.5. Solid potential energy and energy band models for metal and semiconductor.

band electrons is fixed and collision frequency increases with temperature decreasing electrical conductivity as the temperature increases. A key point here is that there is an energy gap between the valence band and the conduction band and that this energy gap derives from the periodicity, which derives from the single crystal. This history is fascinating. However, the goal here is to explain the reasons why single crystals are important to solar cells and to probe the question of how pure and how perfect solar cell materials need to be. Most importantly, how can solar cells be integrated into systems to generate electricity on a large scale at a competitive price? Before describing semiconductors in more detail, let us return to the periodic table and contrast the semiconductors with metals and insulators to see why semiconductors are special and why they are needed to make solar cells. To preview the answer, note that in order to deliver electric power, a solar cell needs to generate both current and voltage. Generating current requires electron mobility and generating voltage requires a gap between electron energy states. Metals have electron mobility and insulators have gaps between energy states, but only semiconductors have both. The metals like sodium and magnesium are on the left in the periodic table. These atoms have only a few loosely bound electrons each, and they can be tightly packed with up to 12 nearest neighbors. Because the atoms are closely packed, the potential energy well for a metal looks like a flat-bottom well with the well extending to its surfaces. As shown in Figure 3.5a, the metal surfaces form the energy barriers confining the electrons. Because this well is so large compared to one atom, all electron wave function wavelengths and energies are possible. Electrons are then free to move around in the metal, but there are no energy gaps between energy states. Since the electrons hardly feel the metal atom core positions with the flat-bottom potential well, crystallinity is not important to metallic properties.

SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY

51

The elements at the right of the periodic table like oxygen and chlorine have tightly bound electrons and are hungry to grab more. They readily form ionic compounds like salt (sodium chloride) and glass (silicon dioxide). The energy levels in these compounds are much like those of atoms in that the electrons only are excited between atomic energy states. There are gaps in energy, but the electrons are not mobile. Crystallinity is not very important since electrons are localized on ions. This brings us to the group IV elements like silicon. The structure of silicon in a silicon crystal is shown in Figure 3.6. Silicon has four electrons and forms four tetrahedral bonds as shown. Looking at a row of silicon atoms along the diagonal in a silicon crystal, note the alternating bonded and nonbonded spaces between silicon atoms. The energy potential well profile for this row is shown in the middle of this figure along with two wave patterns, one drawn as a solid line and one drawn as a dashed line. The peaks in the solid line wave pattern localize the electrons in the bonded regions with lower-average energy potential. Meanwhile, the peaks in the dashed line wave pattern are localized in the nonbonded regions with higher-average energy. However, both waves allow the electrons to be near any silicon pair in the crystal, implying electron mobility throughout the crystal. Because of the periodic nature of the atomic positions in a single crystal, the wave functions allowed describing the electrons in a single crystal must have a

E

Eg

c v

Eg

c v

E x Si

Si

Si

Si

Si

Si

Si

E

E x As

Ga

As

Ga

As

Ga

As

Ga

Figure 3.6. Top: tetrahedrally bonded silicon atoms in groups along cube diagonal in silicon crystal showing alternate bonded and nonbonded pairs. Middle: energy potential for top atom sequence with valence band bonding wave function as solid line and conduction band antibonding wave function as dashed line. Bottom: the potential and wave functions for a GaAs crystal.

52

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

corresponding wavelength. Thus, the two types of states with bonding and antibonding electron locations between nearest silicon pairs or farthest silicon pairs are the only states allowed. There is an energy gap between these states because no other electron wave functions are allowed. The states representing the bonding states form what is called the valence band and the states representing the antibonding states form what is called the conduction band. Figure 3.6 also shows the energy potential and wave functions for a group III–V semiconductor. The III–V semiconductors are the materials used to make LEDs. In this case, a group III element like gallium can form tetrahedral bonds with a group V element like arsenic where the result is the sharing of four electrons per atom as in silicon. Group III–V is a rich class of semiconductors. It turns out that because of the crystal periodicity, there is both energy gap and electron mobility in semiconductors. Figure 3.7 allows one to visualize this more easily. In this figure, both connected bonded regions and open channels in between can be seen. One can imagine electrons traveling in the bonded regions or separately in more energetic states in the open channels. Propagating electrons in the bonded region have energies in a valence band, and propagating electrons in the open channels have energies in a conduction band. The separation between these regions provides the energy gap. Looking at Figure 3.7, one can also imagine a large foreign atom or a crystal boundary or defect interfering with flow in the channels or a total disruption of the channels smearing the two sets of energy states into each other.

Figure 3.7. A view of a channel open for conduction electron movement in a GaAs single crystal. Small sphere: gallium atom; large sphere: arsenic atom; white cylinders: valence bonds.

JUNCTIONS AND DIODES

53

Figure 3.7 suggests intuitively that electrons will have higher mobility in single crystals than in amorphous or small crystal size thin films. This is in fact true quantitatively. Electron mobility is easily and routinely measured. The electron mobility in single-crystal silicon is typically 1500 cm2/Vs, and in single-crystal GaAs, it is 4500 cm2/Vs. However, in amorphous silicon and copper indium diselenide (CIS), two common thin-film solar cell materials, it is only 4 cm2/Vs. This is a difference by a factor of 1000 consistent with our intuitive expectations based on Figure 3.7.

3.5

JUNCTIONS AND DIODES

It has now been established that carriers are mobile, allowing current to flow in solar cells. How does one use an energy gap to create voltage? A P/N junction (P = positive, N = negative) is needed. In the above description of electron movement in semiconductors, one now needs to note that it is important to count electrons. If the semiconductor is very pure (a state called intrinsic), then all of the bonding states will be occupied by electrons and there will be no electrons to move in the conduction band. Electrons cannot move in the valence band either because there are no empty spaces to move to. Substituting a small number of phosphorus atoms for silicon atoms can rectify this problem (one in a million). Since phosphorus is from group V, it has one more electron than silicon. The resultant material is labeled N-type because the extra electrons are negatively charged. Alternately, as a complement to the N-type material, one can substitute an aluminum atom for a silicon atom leaving the bonding or valence band one electron deficient because aluminum from group III has one less electron than a silicon atom. Now instead of thinking about a million electrons in the valence band, we talk about the missing electrons in the valence band. We call this a hole. It is like watching a bubble move in water. The hole has a positive charge and we call this material P-type. Now what happens when N- and P-type materials are brought together? The result is a P/N junction diode [10–12] as shown in Figure 3.8. The band edge diagrams at the bottom of this figure describe how a diode works. When the P and N regions first come together, the electrons and holes from each side diffuse together, eliminating each other, leaving an electric field region in the junction. This happens until the valence band edge (v) in the P material almost lines up with the conduction band edge (c) in the N material as shown on the left in this figure. At this point, the free electrons and holes on both sides of the junction have the same energy as shown by the dashed horizontal line. This is the zero voltage band diagram (Fig. 3.8A). Now notice that there is an energy hill for electrons to climb in order to move from the N to the P side of the junction. An applied voltage can either decrease this hill or energy barrier for forward bias (Fig. 3.8B) or increase it in reverse bias (Fig. 3.8C). If the hill is made small enough by a forward voltage about equal to two-thirds (67%) of the bandgap energy, Eg, then current starts to

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Current (Amps)

54

Current PN y

z

x

c Eg v E

c

P v

x (a) Zero voltage

c Eg v

(A)

(C)

Voltage (V)

Forward current

c Eg v

Metal Contacts

N

(B)

P

Electron c Applied N voltage v

P

N

c v

(b) Forward voltage

(c) Reverse voltage

Figure 3.8. Upper left: P/N junction diode; upper right: current versus voltage for P/N diode; lower left (A): conduction band minimum and valence band maximum positions through P/N junction at zero applied voltage. Lower middle (B): forward voltage band diagram—reduced barrier for high current flow. Lower right (C): reverse voltage—barrier blocks current flow.

flow. This corresponds to the knee in the diode current versus voltage curve shown at the top right in this figure. In reverse bias, no current flows because the barrier just gets bigger. Thus, a diode is a rectifier allowing current flow in only one direction.

3.6

SOLAR CELL BAND DIAGRAMS AND POWER CURVES

Referring now to Figure 3.9, a solar cell is just a large P/N junction diode with a metal grid on its front side facing the sun. A solar cell converts the energy in sunrays to electric power. Now we shall refer to the sunrays as photons. In Figure 3.9, the now familiar band edge diagrams are shown at the bottom. These band edge diagrams show how a solar cell works. First, a photon is absorbed exciting an electron from the ground state or valence band in the P material to an excited conduction band state. It is mobile in the conduction band and if it lives long enough in this excited state, it can diffuse to the junction and fall down the potential barrier. Another way of thinking about this potential barrier is simply that it represents an electric field region created by the initial separation of electron and holes when the junction was formed. Anyway, when an electron enters a field region, it gains electrical energy. This can be converted to a voltage and current to do work.

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

Sun ray Current (Amps)

Voltage (V)

P N

x

Eg

c c P

x

Max power point

Light generated current Forward dark current

c

E

Voc

Isc

Light generated current

v

55

Eg

N

Voc

v v

Short circuit current (Isc) at zero voltage

P Open circuit voltage (Voc) at zero current

Figure 3.9. Upper left: P/N junction solar cell with metal grid on top. Lower left: photon absorption excites electron into conduction band. Electron then falls through junction potential. Upper and lower right: current versus voltage curve for solar cell is diode I versus V curve moved down by light-generated current.

3.7

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

How efficient can a solar cell be and how do we achieve these high efficiencies? Theoretically, a solar cell efficiency of 70% is possible. However, no one believes that, in practice, this can be achieved. Still, a 35% efficient solar cell has been demonstrated and 45% is probably an achievable target. What needs to be done to achieve high efficiencies is a more interesting question. In fundamental terms, three things need to be done. First, for each photon absorbed, the excited state carrier generated needs to last long enough to be collected at the junction. Second, while the sun’s spectrum contains photons of different energies, the energy available in each photon must be used as wisely as possible. And third, the voltage a cell generates should be as close as possible to the bandgap energy. We will discuss each of these requirements in succession in the following paragraphs. The first requirement of one electron collected for every photon absorbed implies single-crystal material and high-purity material. The measure of electrons collected per photon absorbed is called quantum efficiency. Figure 3.10 provides a semiquantitative answer to the semiconductor purity question. To understand Figure 3.10, let us go back to the crystal channels shown in Figure 3.7. First, how far will an electron move through one of these crystal channels? The answer is about 100 atomic spaces. This is because the atoms are not really stationary but

56

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Random walk

Light absorption

Junction

Figure 3.10. A light-generated carrier diffuses to the junction in a random walk sequence.

are vibrating small distances around their home positions because they have thermal (heat) energy. This vibration energy is small, however, so that the excited electron does not return to the valence band but just gets deflected into another channel. We think of this deflection as a step in a random walk diffusion problem. This brings us back to Figure 3.10. The next question is how far is the excited-state electron away from the junction? This depends on the photon absorption distance. This absorption distance depends on the material and on the rules for photon absorption. Now we shall divert for a minute to the rules for photon absorption. This will be important because, as we will see, silicon is fundamentally different from the III–V semiconductors in its photoelectric properties. Looking back to the hydrogen atom in Figure 3.4, a rule for photon absorption is that the wave functions involved have to have different symmetries. For example, note that the S and D wave functions are symmetric around the position of the nucleus, while the P functions are antisymmetric. Thus, absorption between S to P and P to D are allowed, but S to D is not allowed. Now let us look at the wave functions for silicon and GaAs in Figure 3.6. Note that both wave functions for silicon are symmetric around the point between two silicon atoms. This means that photon absorption in silicon is not allowed to first order. In GaAs, however, photon absorption is allowed. So the photon absorption length in GaAs is about 10,000 atomic spaces. In reality, photons are also absorbed in silicon but in about 100,000 atomic spaces. This second-order absorption in silicon results because of atomic thermal vibrations. Now, returning to the purity question and the random walk diffusion problem, remember that a step length is about 100 atomic spaces. So a carrier in GaAs will

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

57

be about 100 steps away from the junction, and a carrier in silicon will be about 1000 steps away. However, in a random walk problem, the number of steps required to move N steps away from the start is N × N steps. So, the distance an excited electron must travel to the junction in GaAs will be 10,000 steps or one million atomic spaces. If it were to see a large impurity in a channel on this path, it could return to the valence band and be lost. So the purity requirement for GaAs is about 1 ppm. The analogous argument for silicon suggests a purity requirement of 10 ppb. This is 1/100,000,000. In fact, silicon solar cells lose performance given transition metal impurities in the range of several parts per billion. The above argument has been a little tedious, but the goal is to impress the reader with this purity requirement. By analogy, it should also be clear that good single-crystal quality without defects is as important as purity. The above purity specification is routinely met in commercial single-crystal silicon solar cells today as well as in various other single-crystal silicon-based devices that have revolutionized our lives over the last 50 years. While the reader is probably not aware of it, various single-crystal III–V devices have penetrated our everyday lives as well in the last 10 years. As the above argument about the difference in photon absorption for GaAs versus silicon suggests, III–V are often a better choice for photoelectric and optical-electronic applications. Referring to the periodic table, there are a large number of III–V materials available including GaAs, InP, InSb, and GaSb. Additionally, alloys of these materials are available including AlGaAs, GaAsP, GaInAs, InGaP, and InGaAsP. This makes a large set of bandgaps and electron mobilities available. Single-crystal III–V devices can now be found in cell phones, satellite receivers, CD music players, CD-ROMs in personal computers, taillights in cars, traffic stoplights, and military weapon systems. Single-crystal III–V devices are also key components in fiber optic phone communication and the Internet. In fact, the most efficient solar cells are made using III–V materials. This brings us back to our second requirement for making high-efficiency solar cells. We need to use the energy in the sun’s varied colored rays as efficiently as possible. A problem with sunlight is that the photons come in different colors with different associated energies. If we wanted to maximize the efficiency of a photodiode, we would illuminate it with only photons with a single energy with an energy equal to the bandgap energy Eg. Then, if the crystal quality and purity were sufficient, all of the excited carriers would be collected at the junction with 67% of the photon energy being delivered as a voltage. The energy conversion efficiency would be roughly 67%. However, referring to Figure 3.2, photons from the sun come with different energies. Some of the photons have too little energy to be absorbed, and some of the photons have energy considerable in excess of the bandgap energy. For the sun’s spectrum, this limits the single-junction solar cell efficiency to less than 30%. Fortunately, group III–V offers a solution because various materials with various bandgap energies are available. Specifically, one can stack a visible light-sensitive GaAs solar cell with metal grids on its front and back on an infrared-sensitive GaSb solar cell to arrive at the two-color or two-junction solar cell shown at the right in

58

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Figure 3.11. In this way, one absorbs the high-energy photons first in the top material generating a high voltage, while the low-energy photons pass through the top cell to be converted in the bottom cell. More photons are used and they are used more wisely. This then is the 35% efficient GaAs/GaSb two-color or two-junction solar cell demonstrated by L. Fraas et al. in 1989 [2].

Silicon Eg=1.1 eV

Ga Sb Eg=0.7 eV

Ga As Eg=1.4 eV

Figure 3.11. Left: for single-junction solar cell, sunlight contains high-energy photons with excess energy and low-energy photons with too little energy. Right: solar spectrum can be more efficiently utilized by stacking two different junctions together. Three junction cell

First Second Third

n+ n p p++ n n p+ p+ n n

Window InGaP GaInAs

Ge

p

From p. 888 (at 300 suns AM1.5) 13.3 mA/cm2 x Vop η = --------------------------------84 mW/ cm2 Vop= 2.55 V Efficiency = 40%

Active solar cell junction Shorting tunnel interconnect

Figure 3.12. Monolithic triple-junction InGaP/GaInAs/Ge CPV cell as first described by L. Fraas and R. Kinechtli in the 13th IEEE PV Specialist Conference, predicting 40% at 300 suns AM1.5 [13].

PV MODULE AND SYSTEM COST TRADES

59

Concentrating lens

Solar Cell Heat spreader back plane

Light Generated Current (Amps)

Sun light Sunlight concentrated

Sunlight not concentrated

Voltage (V)

Figure 3.13. Solar cells are more efficient with concentrated sunlight because both current and voltage increase.

As first predicted in 1978 [13], an efficiency of 40% has recently been demonstrated [14] for the monolithic three-junction InGaP/GaInAs/Ge CPV cell as shown in Figure 3.12. This brings us to the third way of increasing solar cell efficiency. For a given bandgap energy, we want to generate more voltage. Concentrating the sunlight onto the cell can do this. This is shown in Figure 3.13. Sunlight can be concentrated using a lens as is shown at the left in this figure. The resulting currents versus voltage curves with and without a lens are shown at the right. As is customary for solar cells, the diode curves here have been flipped over. Note that the higher current concentrator cell has a higher efficiency. This is because the diode is being driven harder to a higher current and voltage. In other words, if the light level goes up by 10, the current also goes up by 10 but at the same time, the voltage also goes up. In practice, the open-circuit voltage can go up from about two-thirds of Eg to about three-quarters of Eg under concentrated sunlight.

3.8

PV MODULE AND SYSTEM COST TRADES

So it is clear that single-crystal cells have higher efficiencies than amorphous or small grain size thin-film cells and that multijunction cells have still higher efficiencies. However, single-crystal cells are more expensive than thin-film cells and multijunction cells are very complex. It is also clear that CPV systems see only direct sunlight, which is less than global sunlight. So, it is necessary to look at the cost of complete PV systems. This is shown in Table 3.2. Note that the attached table has five columns comparing system-level costs measured in terms of simple payback times. The columns summarize future (ca.

60

TABLE 3.2. Comparison of Future Economics for Planar PV and Concentrated Solar PV Module Type

Thin Film

SC Silicon

LCPV

HCPV

HCPV JXC

Cell Efficiency @ temperature

9%

19%

22%

35%

44%

Annual available irradiancea

2336 kWh/m2 (fixed tilt)

2905 kWh/m2 (tracking 1-axis)

2382 kWh/m2 (1-axis; 87% global; 94% optical efficiency)

2178 kWh/m2 (2-axis; DNI = 78% global; 85% optical efficiency)

2178 kWh/m2 (2-axis; DNI = 78% global; 85% optical efficiency)

Annual kWh/m2 electricity

210 kWh/m2

552 kWh/m2

524 kWh/m2

762 kWh/m2

936 kWh/m2

Dirt penalty

–5% = >200 kWh/m2 –5% = >525 kWh/m2

–7% = >488 kWh/m2

–9% = >695 kWh/m2

–9% = >852 kWh/m2

Annual revenue at 10¢/kWh

$20/m2

$52.5/m2

$48.8/m2

$69.5/m2

$82/m2

Cell cost per m2 module

$50/m2

$300/m2

$100/m2

$100/m2

$137/m2

Module materialb

$50/m2

$60/m2

$67/m2

$67/m2

$67/m2

Optics

0

0

$57/m2

$65/m2

$65/m2

Module cost

$100/m2 ($1.11/W)

$224/m2 ($1.20/W)

$232/m2 ($0.92/W)

$360/m2 ($1.89/W)

$75/m (1-axis)

$150/m (2-axis rigid)

$150/m2 (2-axis rigid)

$25/m2

$25/m2

$25/m2

$40/m2

$40/m2

$160/m2

$460/m2

$324/m2

$422/m2

$459/m2

$35/m

Installation Total module area cost

2

2

2

$269/m2 ($0.85/W)

$75/m (1-axis)

Array support structure

2

Module system costc

$1.78/W

$2.42/W

$1.74/W

$1.67/W

$1.44/W

System with inverter at $0.3/Wd

$27/m2 $2.08/W

$57/m2 $2.72/W

$56/m2 $2.04/W

$76/m2 $1.97/W

$96/m2 $1.74/W

System payback time at $0.1/kWhd

9.4 years

9.8 years

7.8 years

7.2 years

6.8 years

a

Las Vegas. Glass, frame, Wire, J-box, etc. Without inverter cost. An explanation of the assumed peak power rating is required. For the two planar cases, this is straightforward. The efficiencies listed should be interpreted as future projected module efficiency at operating temperature. In this case, the Wp rating assumes 1 kW/m2 global illumination. So, the module peak W power ratings are 90 W/m2 for the thin-film case and 190 W/m2 for the single-crystal silicon case. For the HCPV cases, the efficiencies listed should be interpreted as future projected cell efficiency at operating temperature. However, the Wp rating now assumes 850 W/m2 of direct illumination (DNI) and 85% optical efficiency. So, for the 35% HCPV column, the module peak W power rating is 253 W/m2 and for the 44% column, the module peak W power rating is 318 W/m2. For the LCPV case, there is no standard. However, since the 3-sun sees DNI plus one-third of diffuse, the logical standard to be consistent with the other two would be to assume 90% of global illumination or 900 W/m2. In the table, a 3-sun module rating of 220 × 0.9 × 0.94 = 186 W/m2 is assumed. d Including inverter cost. From above, given that the inverter size for the system is set by the system peak power rating, one can calculate inverter cost per square meter. For the thin-film column, this is $27 per square meter; for the Si column, this is $57 per square meter; for 35%, this is $76 per square meter; for 44%, this is $96 per square meter; for LCPV, this is $56 per square meter. b c

61

62

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

2015) projected system costs for thin-film PV systems, single-crystal silicon flatplate systems, a JX Crystals 3-sun LCPV evolutionary system, an HCPV system using today’s 35% multijunction cells, and a JX Crystals HCPV Dual-Focus Cassegrainian system using a 44% efficient cell combination. These systems will be described in more technical detail in Chapters 12–17. Each column lists projected module outputs in annual kilowatt hour per square meter for Las Vegas, and each column summarizes projected system-level area-related costs in dollar per square meter. These costs include not just module cost but also support structure and installation costs. Given annual projected energy outputs and total area-related costs, it is then possible to calculate a simple system payback time given an assumed energy value of 10 ¢ per kilowatt hour. The good news is that all five columns for all PV technologies promise a simple payback time of less than 10 years. These systems should last for over 20 years. The first column is for the thin-film PV case. This case suffers from a low cell efficiency of 9%. Module costs are assumed to be those stated by First Solar and other thin-film PV companies of only $1.11 per watt. However, even given these low module level costs, because of the low cell efficiencies and lower-energy production per square meter, it takes a long time to pay off the cost of the array support structures and system installation. The projected simple payback time is 9.4 years. The second column is for the conventional single-crystal silicon planar module case. Here, the good news is the 19% cell efficiency, which then leads to 2.5 times more annual energy production per square meter of module area compared to the thin-film case. Simple one-axis tracking is one of the benefits here, giving more kilowatt hour per kilowatt installed. Trackers, unfortunately, are not affordable for the thin-film PV case. However, the cell cost is much higher than the thin-film case. Note that the cell cost assumed here is a future projection of $1.58 per watt. Given a higher-energy output, the field-related area costs are paid off faster than for the thin-film PV case, and the simple payback time works out to be 9.8 years. The third column is for an evolutionary LCPV case as will be described in Chapter 12. This case assumes the same 1-sun single-crystal cells and module manufacturing process as for the second column case. The three changes are the cut-up of the 1-sun single-crystal cells into thirds, the addition of low-cost linear mirrors, and the addition of a thin aluminum sheet heat spreader at the back of the laminated standard module. This approach has already been demonstrated in the field. The immediate benefit here is low risk, straightforward manufacturing, and a reduction of the cell cost by a factor of 3. Additional benefits are the ability to utilize more than just direct sunlight on less than perfect sunny days and higher optical throughput. The same simple one-axis trackers can be used as for the planar silicon module case. The reduction of the cell cost relative to column 2 with no additional penalties leads to a lower simple payback time of 7.8 years. The fourth column is for the HCPV case using today’s 35% multijunction cells. Here, of course, the 35% cell efficiency is spectacular. However, there are some negatives. One negative is that these HCPV systems only see direct sunlight

THE IMPORTANCE OF SINGLE CRYSTALS

63

(DNI) and not global illumination. While DNI can be as high as 90% of global illumination on the best blue-sky day, on an annual average even in Las Vegas or in Phoenix, it is only 78% of global illumination. A typical optical throughput of 85% for HCPV is also a negative. These two negatives make a 0.85 × 0.78 × 35% = 23% silicon planar cell competitive. However, the hoped-for good news for the HCPV approach is a lower module cost because of the cheaper optical materials relative to the high-efficiency single-crystal solar cells. If the low-cost modules ($0.92 per watt assumed here) and if the low-cost precision tracking ($150 per square meter assumed here) can be demonstrated in high-volume production, then this system will beat the planar system in sunny locations with a simple payback time of 7.2 years. However, this approach is more risky and will take longer to build up the manufacturing infrastructure than for the LCPV case. In the still longer term, column 5 presents an even higher-efficiency HCPV approach as will be described in Chapter 15. This approach uses a recently announced improved triple-junction monolithic cell (39% efficient) at the focus in the center of a Cassegrain optical configuration. A second infrared cell is also used behind a secondary mirror with a dichroic coating allowing the infrared to get to the second cell. The second GaSb infrared cell adds an addition 5% to give a cell combination efficiency of 44%. The GaSb cell, which is a diffused junction cell [16], is relatively inexpensive to make. Even with the cost of this second cell, its added efficiency leverages down all of the system cost to give an even more attractive payback time of 6.8 years. This approach, while all of the separate components have already been demonstrated, will require a lot of time and money to bring into high-volume production.

3.9

THE IMPORTANCE OF SINGLE CRYSTALS

From the above discussion of system-level cost trades, why have CPV systems not received more attention and why have thin film systems received so much attention? One answer is that searching for a 20% efficient low-cost thin-film solar cell is a very attractive dream. However, in this chapter, we have talked about electrons as waves and semiconductors as crystals to convey the message that this dream is not well founded on scientific principles. In fact, in graduate school solid-state physics classes, the bandgap in semiconductors is rigorously derived based on the assumption of the perfect periodic single-crystal lattice. However, the importance of single crystals to semiconductor devices is not generally conveyed in a simple understandable way. It is certainly not knowledge available to funding sources or the financial community. Figure 3.14 is an attempt to rectify this situation by making an analogy between an electron traveling in a solid and a car traveling through a forest. Organizing the atoms in single crystals is like removing the trees to make a road through a forest. Atoms out of place or atomic impurities are obstacles for the electron just like trees are obstacles for a car. Collisions with these obstacles force the electron (or the car) to lose energy. Efficiency is dramatically reduced.

64

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Figure 3.14. Single-crystal versus thin-film solar cells. Organizing the atoms in single crystals is like removing the trees to make a road through a forest. Atoms out of place or atomic impurities are obstacles for the electron just like trees are obstacles for a car. Collisions with these obstacles force the electron (or the car) to lose energy. If you were a car driving through the national forest, or an electron passing through a solar cell, which path would you rather take?

In any case, after 25 years of effort on thin-film solar cells, their module efficiencies are still low and they have not penetrated the mainstream electric power marketplace. Concentrator solar cells have not entered the marketplace yet in high volume either. There are several reasons for this, but it is not for lack of performance. The technology for solar concentrators is well founded on established scientific and engineering principles. One of the problems for concentrators is that a larger investment is required for hardware like lenses and trackers as well as for new solar cell manufacturing facilities. It is time for a serious top-level funding commitment. The technology is ready.

REFERENCES

65

ABBREVIATIONS AM1.5—1 and ½ Air-Mass (see Chapter 19) CPV—concentration photovoltaic DNI—direct normal incident sunlight E—electric field Eg—bandgap energy GaAs—gallium arsenide GaSb—gallium antimonide Group III–V—compounds consisting of group III and group V elements from the periodic table (LED semiconductor compounds) HCPV—high-concentration photovoltaic LCPV—low-concentration photovoltaic LED—light-emitting diode P/N junction—positive/negative junction PV—photovoltaic or solar cell QE—quantum efficiency (the number of electrons collected per incident photon) λ—wavelength

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]

A. Einstein. On the quantum theory of radiation. Phyikalische Zeitschrift 18, 121 (1917). L. M. Fraas, J. Avery, J. Gee, et al. Over 35% efficient GaAs/GaSb stacked concentrator cell assemblies for terrestrial applications. In 21st IEEE PV Specialist Conference, p. 190. Kissimmee, Florida, 1990. IEEE, New York (1990). R. P. Feynman, R. B. Leighton, and M. Sands. The Feynman Lectures on Physics, Volume III—Quantum Mechanics. Reading, MA, Addison Wesley (1965). N. Bohr. Radiation Spectra and the Hydrogen Atom. Philos. Mag. 25, 10 (1913). N. Bohr. The Theory of Spectra and Atomic Constitution. Fys. Tidsskr. 19, 153 (1921); 9, 1 (1922). R. P. Feynman, R. B. Leighton, and M. Sands. The Feynman Lectures on physics. In The Hydrogen Atom and the Periodic Table, Vol. 3, Chapter 19, Reading, MA, Addison Wesley (1965). A. Sommerfeld. Z. Phys. 47, 1 (1928). F. Bloch. Z. Phys. 52, 555 (1928). A. H. Wilson. Proc. R. Soc. A 133(458), 134, 277 (1931). J. M. Ziman. Principles of the theory of solids. In Electron States, pp. 72–74. Cambridge and London, Cambridge University Press (1964). C. Kittel. Introduction to Solid State Physics. 3rd Edition. New York, John Wiley & Sons (1967). S. M. Sze. Physics of Semiconductor Devices. New York, Wiley-Interscience (1969). L. M. Fraas and R. C. Knechtli. Design of high efficiency monolithic stacked multijunction solar cells. In 13th IEEE Photovoltaic Specialist Conference, Conference Record (A79-40881 17–44), Washington, D.C., June 5–8, pp. 886–891. Institute of Electrical and Electronics Engineers, Inc., New York (1978).

66 [14] [15]

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS R. R. King, D. C. Law, K. M. Edmondson, et al. 40% efficient metamorphic GaInP/ GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90 (18), p. 3516 (2007). L. M. Fraas, H. X. Huang, and J. E. Avery. Low cost high power GaSb photovoltaic cells. Thermophotovoltaic generation of electricity. In 3rd NREL Conference, T. Coutts, C. Allman, J. Benner, eds, p. 33. Woodbury, N.Y., AIP (1997).

4 SOLAR CELL DEVICE PHYSICS LARRY PARTAIN Varian Medical Systems

4.1 DEVELOPMENT OF QUANTUM MECHANICS AND SOLID-STATE ELECTRONICS Solar cells are likely the ultimate quantum mechanical and solid-state electronic devices. Although their behavior (photovoltaics) was observed in the early 1830s, their optimization and rapid advancement awaited the development of both these fields starting at the turn of the twentieth century. Many of these discoveries are summarized by the set of equations immediately below. The person most associated with each discovery is shown in parentheses along with the year of the discovery. Asterisks identify those advances that were awarded Nobel Prizes in physics [1, 2]. (Planck, 1900)*

E N = Nh f

(4.1)

(Einstein, 1905)*

E = hf

(4.2)

(Compton,1923)*

p = hf c

(4.3)

(De Broglie, 1923)

λ=h p

(4.4)

f ( ε ) = 1 [1 + exp {( ε − ε F kT )}]

(4.5)

(Fermi and Dirac, 1926) (Schrodinger, 1928)* (Shockley et al., 1947)*

[ − (h

2

8π m ) ∇ + V ( r )] FK ( r ) = EK FK ( r ) 2

2

ε*Fe − ε*Fh = qV and J n = qnμ n dε*Fe /dx.

(4.6) (4.7)

Equation 4.1 explained the shape of the black body light emission spectrum from hot objects including the sun. The energy at any emitted light frequency f consists Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

67

68

SOLAR CELL DEVICE PHYSICS

of N discrete (quantized) energy packets each with a small energy content proportional to f with a proportionality constant (Planck’s constant) value of h. Equation 4.2 more definitively established that the wave phenomena of light is actually made up of discrete photons each with a quantized energy value hf that explained the experimentally observed photoelectric effect of electron emission from metal surfaces in a vacuum illuminated with monochromatic light. Equation 4.3 established that each such photon possesses a quantized momentum value, p, which accurately explains the trajectories of photon “collisions” with electrons. Strangely enough, Equation 4.4 indicates that rest mass particles like electrons also behave like waves with wavelengths λ that are inversely proportional to their momentum value, p. This is one description of the wave–particle duality. In 1913, Neil Bohr’s [3] Nobel Prize-winning work (when combined with the De Broglie result of Eq. 4.4) explained the discrete electron energy levels of hydrogen atoms have just the right sizes to accommodate integer multiples of each electron’s wavelength, λ. Thus, the hydrogen atom has allowed electronic energy states with unoccupied energy gaps between each allowed state or energy “band.” It is reasonable to expect that when atoms are brought together to form solids, their atomic discrete levels are perturbed and broadened into allowed electronic energy state bands. Equation 4.5 describes the probability that any of a material’s allowed electronic states is actually occupied by an electron, in terms of an energy variable expression containing Boltzmann’s constant k and referenced to a special energy value εF (the Fermi energy). The latter has ultimately been found to be the electron’s potential energy value [4, 5]. 4.2 FUNDAMENTALS OF SOLAR CELL OPEN-CIRCUIT VOLTAGE The density of available quantum states to “free” electrons in a semiconductor can be approximated by solving Schrodinger’s wave equation for the case of a cube of semiconductor material where the states’ potential energy, V, is zero everywhere inside the cube and is infinite everywhere outside this cube [6]. All of the allowed solutions to the wave equation are sine waves whose wave numbers K (i.e., in the x direction Φ(x) ∼ sin(Kx)) are integer multiples of π/a where a is the dimension of one side of the cube equal to the atomic spacing of the semiconductor. If one then integrates such quantized K values in three-dimensional space, using a differential volume that is normalized by the volume (π3/a3) occupied by one individual quantum state and applying the Fermi–Dirac expression for the probability that any available state is actually occupied, then one obtains the well-known expressions for the volume density of electrons n in the conduction band (and similarly the volume density of holes p in the valence band) with the respective conduction and valence energy band edges specified by εC and εV [7, 8]: n = N c e −(εC − ε F ) kT

(4.8)

p = N v e −( ε F − ε V ) kT .

(4.9)

FUNDAMENTALS OF SOLAR CELL OPEN-CIRCUIT VOLTAGE

69

Here, NC = 2Mc(2πmdekT/h2)3/2 and Nv = 2(2πmdhkT/h2)3/2. mde and mdh are the density-of-states effective masses for “free” electrons and holes, respectively, and Mc is the number of equivalent minima at the conduction band edge [8, 9]. Each hole denotes the absence of an electron in the valence band. Note that the product of n × p gives np = N C N V exp[− ( ε C − ε V ) kT = N C N V exp ( − ε G kT ) = ni2 ,

(4.10)

which only varies with the bandgap and temperature regardless of the doping levels to make the semiconductor either n or p type or alternatively to characterize its intrinsic or “i” properties if undoped. Then ni is the electron and hole density in such “i” materials. While the density states derivation is based on the periodic atomic spacing a, any dependence on this a disappears in the total mathematical calculation process. The results, at least to the first order, appear to accurately describe the properties of even amorphous semiconductor materials with distributions of atomic spacings and not just a well-defined and single constant a as long as there is an approximately defined bandgap. Equations 4.8 and 4.9 are equilibrium expressions for materials in the dark. With light exposure, both conduction band electron and valence band hole concentrations increase above their equilibrium values. Absorbed photons with enough energy “pump” valence band electrons into the conduction band, leaving behind the valence band holes. The standard assumption is that Equations 4.8 and 4.9 still apply but with two different quasi-Fermi energy values, ε*Fe for the electrons and ε*Fh for the holes instead of the single-equilibrium Fermi energy εF0. For the latter, subscript “0” is added to denote its equilibrium position. These expressions clearly show that for n to increase, ε*Fe must move away from εF0 toward the conduction band edge εC. Similarly, ε*Fh must also move away from ε*F0 for p to increase but toward the opposite valence band edge εV. Since quasi-Fermi energies correspond to potential energies, the light-exposed semiconductor populations of electrons and holes actually possess measurably different potential energy values that, according to the first Shockely expression in Equation 4.7, is a measureable voltage, V. The magnitude of this voltage is expressed as V = [ Δε LFe + Δε LFh ] q ,

(4.11)

where Δε LFe = ε*Fe − ε F0 and Δε LFh = ε*F0 − ε Fh with the superscript “L” added to denote the nonequilibrium light exposure. It is this splitting of quasi-Fermi and potential energy values of electrons and holes that is the first fundamental step in the quantum mechanical conversion of the microscopic photon energy (typically expressed in electron volts per photon) into a macroscopic DC voltage, V, which can be physically measured and used to do work. The challenge is to include properly configured semiconductor materials into a device that allows such a macroscopic physical V measurement with a standard voltmeter.

70

SOLAR CELL DEVICE PHYSICS

A simple explanation of the underlying physics was presented in the First Edition (pp. 2–4 [5]). It details how concentration gradients (normalized by the total concentration value) produce forces per unit charge that are just as real as the forces that electric fields exert on such charges. Moving quantized particles through such concentration gradient force fields involves work that changes potential energies. This derivation is not repeated here.

4.3

SHOCKLEY DIODE MODEL OF SOLAR CELLS

In 1947, Shockley et al. [10] produced the Nobel Prize-winning discovery and demonstration of solid-state electronic transistors fabricated from single-crystal semiconductors with energy bandgap properties. Shockley’s breakthrough devices incorporated characteristics well summarized by the two expressions of Equation 4.7 (pp. 305–310, 463 [9]) that he used extensively in his work. The second Equation 4.7 expression also clearly implies quasi-Fermi level connections to potential energy because the gradient of the former provides a force that translates into velocity (and then to a current density, J) with a proportionality constant that is the electron mobility μn, where n is the density of conduction electrons in the semiconductor material. The energy gradient and resulting velocity prescribe how much of the starting potential energy and thus the efficiency is lost in establishing a DC terminal voltage, V, and a current flow as detailed below. Interesting enough, Shockley never clearly identified quasi-Fermi levels as measures of potential energy. However, there had long been hints of such a relationship, not the least being the Einstein relation that the ratio of diffusion coefficient to mobility is a known and calculable constant [8, 9]. In their bipolar forms, transistors consist of two back-to-back Shockley diodes. All of these quantum mechanics and solid-state electronic discoveries led directly to the detailed description of the Shockley diode solar cell behaviors and their controlling parameters needed for reproducible progress in this field. These preceding advances received a significant share of the Nobel Prizes in physics in the first half of the twentieth century, and their exploitation in the second half of the twentieth century, along with one more Noble Prize-winning discovery, has resulted in major increases in solar cell performance from ∼1% efficiency level in ∼1900 to over 40% currently. One wonders what the next century of Nobel Prizewinning discoveries may provide. In 1954, Chapin et al. [11] reported a 6% efficiency for an abrupt p/n junction diode single-crystal silicon solar cell conversion device. Seven years later, the derivation and treatment of the classic Shockley model was published by Shockley and Queisser [12] for the current–voltage properties and the efficiencies of this particular configuration. The corresponding geometry and nonequilibrium energy band structure for this type of solar cell exposed to light is shown schematically in Figure 4.1. This contact metallization partially covers its emitter to allow light entrance. A macroscopic and measureable voltage, V, is clearly indicated as being equal to

SHOCKLEY DIODE MODEL OF SOLAR CELLS Emitter

71

Base Depletion region

Light n

p tE tE'

Grid metal

tB xE

tB'

xB

Back metal x

Metal fermi level V ~ 0.6 Voc

0

xn

xP

εc

Loss δεn

Light ε ε

qV

* Fe

q(V-JR)

Dark ε

* Fh

XW

Light Loss

* Fe

ε

Dark

Metal fermi level

* Fh

δεp

εv

Figure 4.1. The Shockley diode configuration (top) and its energy band diagram (bottom) in the light.

the different majority carrier Fermi levels at opposite sides of the solar cell. A significant difference in Figure 4.1 from most earlier models is its clear identification of the Fermi energy in the metal contacts and their alignment with the majority carrier Fermi levels in the p/n junction solar cell. The assumption here as that the metal contacts do not significantly respond to the incident light so that their nonequilibrium behavior is well described just by their single-equilibrium Fermi energy values. At best, the band diagrams in Shockley’s 1950 book (fig. 12-3, p. 310 [9]) only hint at such a diagram, and his classic model publication contains no energy band diagrams [12]. However, it is metal conductors that deliver electric power to loads, and any potential energy conversion in a solar cell can only be utilized if such potential energy differences are efficiency transferred to the metal contacts. Hovel’s often quoted text only weakly relates voltage to differing Fermi levels and with no connection to metal Fermi energy levels (fig. 3, p. 9 [13]). Even though Sze’s (fig. 4, p. 794 [8]) textbook does relate the open-circuit voltage to the Fermi level split, there is no connection made to Fermi levels in the metal contacts. It is such latter levels and their alignments (or misalignments) with conduction and valence band edges that form the basis for projected significant increases (or losses) in solar cell efficiencies that is the major point of this chapter.

72

SOLAR CELL DEVICE PHYSICS

The quasi-Fermi level splitting in Figure 4.1 between ε*Fe and ε*Fh gives the highest potential energy difference available from the solar cell conversion process as characterized by setting the open-circuit voltage Voc to the voltage V of Equation 4.11. For low injection conditions, where the concentrations of the light-generated carriers ΔnL = n – n0 and ΔpL = p – p0 are much greater than that of the minority carriers but are much less than that of the majority carrier concentrations, the opencircuit expression simplifies to Voc = ε G q + ( kT q ) ln [ Δ nL po ( N C N V )]

(4.12)

for the p-type side of the solar cell and to Voc = ε G q + ( kT q ) ln [ Δ pL no ( N C N V )]

(4.13)

for the n-type side of the solar cell. For high injection where ΔnL and ΔpL become larger than the majority carrier concentration, these expressions respectively become 12 Voc = ε G q + ( 2 kT q ) ln ⎡⎣ Δ nL ( N C N V ) ⎤⎦ and

(4.14)

12 Voc = ε G q + ( 2 kT q ) ln ⎡⎣ Δ pL ( N C N V ) ⎤⎦ .

(4.15)

Only significant voltages are generated if ΔnL or ΔpL >> no or po, respectively (low injection), or is greater than both (high injection). In abrupt p/n junctions with significant doping levels and low injection, this means that voltage is only really generated by minority carriers. Note the twice higher temperature coefficient of the second term in the Voc expressions for high injection versus low injection. Also note that Voc approaches the εG/q bandgap value (in all four expressions) when the temperature T approaches absolute zero [12] or when the light-generated carrier concentrations approach 1018–1019 cm−3 concentrations (depending on the density-of-states effective mass values) where the arguments in the ln functions approach values of one. It is this latter case of high light-generated carrier concentrations giving bandgap Voc values that is another major point of this chapter for significantly increasing solar cell efficiencies. Given the absorption of a photon that produces an electron–hole pair, this specific optimization minimizes the loss of energy below the bandgap. This contrasts to minimizing the losses above bandgap in the multiple bandgap devices that have recently provided efficiencies exceeding 40% (Perharz and Bett, Chapter 14; [14]). The fundamental Shockley transistor assumption implies that any minority carrier that enters the depletion region exits the other side with relatively little change in the potential energy it had on entrance. To the degree that this assumption is accurate, it provides an efficient, low-loss process for transforming minority carriers into majority ones. This conversion process from minority to majority

SHOCKLEY DIODE MODEL OF SOLAR CELLS

73

carriers is the second major step of the abrupt p/n junction solar cell energy conversion process. It is instructive to note here the most efficient step theoretically possible for converting an absorbed photon’s energy in electron volts into a DC voltage, V. This occurs when all absorbed photons’ energies hf (i.e., monochromatic photons all of a single frequency f) just exceed the bandgap energy εG, and the open-circuit voltage Voc equals the bandgap value εG/q. This would only be possible if the quasi-Fermi levels ε*Fe and ε*Fh split all the way to the band edges εC and εV. It would be a perfect quantum mechanical conversion step with no direct loss. When the energy of an absorbed photon is greater than the bandgap, its energy excess greater than the bandgap creates “hot” electrons (and/or holes) whose energies are rapidly thermalized to steady-state values near the band edge by inelastic collisions with other charged carriers and/or with semiconductor lattice vibratrions (phonons). The latter quickly convert such extra energy into heat that no longer can contribute to steady-state carrier potential energy but only to the semiconductor’s temperature increase. The third major step in the energy conversion process is the extraction of the majority carriers’ light added energy from the semiconductor and the deposition of this energy into the metal contacts whose average electron potential energy values are also specified by their respective metal Fermi energy levels indicated by the arrows in Figure 4.1. Because of semiconductor band structure, mobile electrons can only occupy electronic states whose total energy is at or above the conduction band edge εC. They thus enter the metal with offset energy δεn as a kinetic energy (hot electron) that is again quickly thermalized by scattering down to the metal’s Fermi energy level. Since only the potential energy is available at the external metal contacts, the typical way to reduce this δεn loss is to raise the electron quasi-Fermi level by moderate to heavy doping at the semiconductor’s boundary. An analogous δεp loss process also occurs for majority carrier holes at the p-type interface with its metal contact. This is likewise typically reduced by moderate to heavy doping at the semiconductor’s boundary. For the highest conversion efficiencies, both these δεn and δεp need to be simultaneously reduced to as close to zero values as possible. The second expression of Equation 4.7 indicates that a loss of potential energy is required to transport charge from one part of a device to another as described by a current density, J. The latter is a fundamental and unavoidable potential energy loss even if the conversion step to open-circuit voltage is perfect and loss free. Despite the lack of band diagram clarity, it is the one-dimensional Shockley and Queisser [12] equations, as extended by Hovel [13], and only slightly extended more in the First Edition (chapter 1 [5]), that have formed the basis for most of the successful analysis and performance improvements that have eventually and recently led to cell efficiencies exceeding 40% ([14]; Perharz and Bett, Chapter 14). These well-known derivations will not be repeated here but will only be summarized by the resulting expressions that identify key parameters. In the Shockley abrupt p/n junction model of Figure 4.1, an electrostatically charged depletion region penetrates a depth, xE, into the emitter of thickness tE and

74

SOLAR CELL DEVICE PHYSICS

also penetrates a depth, xB, into the base of thickness tB. Outside the depletion region, there is no net charge, so any potential energy change, for the mobile electrons and holes there, cannot be due to an accumulation of net electrostatic charge. The majority carrier transport is simply described by the linear Ohm’s law, current–voltage relationships, specified by a single series resistance. The more important physics is contained in the nonlinear expressions (Eqs. 4.8 and 4.9) and relationships that describe the minority carrier transport, where the effects of concentration gradients are dominant. Combining the transport equations with the one-dimensional continuity equations and the standard boundary conditions, and the assumption of current dominated by minority carrier diffusion current at each edge of the depletion region [12, 13], one obtains the standard Shockley diode equations for current density J versus voltage V in an abrupt n/p junction solar cell as J = J 0 {exp [ q (V − JR ) kT ] − 1} − J L ,

(4.16)

V = ( kT q ) ln [( J J 0 ) + ( J L J 0 ) + 1] + JR,

(4.17)

or solving for V,

where Jo = (Joh + Joe) and J oh = q ( ni2 N D ) vdh f h and J oe = q ( ni2 N A ) vde f e and the light-generated current JL is ∞

J L = ∫0 qF ( λ ) [1 − r ( λ )] QE ( λ ) dλ.

(4.18)

Here, fh and fe are hyperbolic functions of the surface or interface recombination velocities, the diffusion velocities, the diffusion lengths, and the undepleted n and p layer thicknesses specified in the First Edition [5]. The R = RA is the specific series resistance, where R is the actual series resistance in ohms describing majority carrier transport and A is the cross-sectional area of the device. The ND and NA are the respective donor and acceptor doping concentrations of the emitter and base. The F(λ) is the photon flux density of light of wavelength λ, r(λ) is the light reflectivity of the emitter surface at λ, and QE(λ) is the quantum efficiency of the solar cell at the indicated wavelength. The open-circuit voltage Voc comes from setting J = 0 in Equation 4.17 to give Voc = ( kT q ) ln [( J L J 0 ) + 1] ,

(4.19)

which is equivalent to Equations 4.12 and 4.13’s low injection expressions applied to the boundaries of the depletion region. The short-circuit current density Jsc (where V = 0) is given from Equation 4.16 by the transcendental expression J sc − J 0 [ exp ( − qRJ sc kT ) − 1] = − J L .

(4.20)

SHOCKLEY DIODE MODEL OF SOLAR CELLS

75

If R is known, the light J–V characteristics are uniquely determined when any two of the three coefficients JL(∼ –Jsc), Jo, or Voc are known. Equation 4.19 gives the third when the other two are specified. The power density is the JV product (=J{[KT/q] In[(J/Jo) + (J/JL) + 1] + JR}) from Equation 4.17 that is maximum when the d(JV)/dJ = 0, at a maximum power current density Jm and voltage Vm. This Jm is negative, with a magnitude slightly less than JL (as shown in Fig. 4.2), and it is specified by the transcendental expression ln [( J L + J m + J 0 ) J 0 ] = − J m ( J L + J m + J 0 ) − 2qRJ m kT .

(4.21)

Substituting this Jm back into Equation 4.17 then gives Vm. The efficiency of the cell η is given by 2

1

0

Voc

J (A/cm2)

–1

–2 JL –3

R = 0.3 Ω-cm2 0.22

–4 0.0091

Jsc

0.15

–5

–6 –1

–0.5

0

0

0.06

0.5

1

1.5

2

V (V)

Figure 4.2. The light and dark J–V characteristics calculated with Equation 4.16 for Jo = 9.23 (10−20) A/cm2, JL = 5.17 A/cm2. The open-circuit voltage Voc and short-circuit current density Jsc are shown as a function of the specific series resistant R in ohm-square centimeter.

76

SOLAR CELL DEVICE PHYSICS

η = − Vm J m A Pin = − Voc J sc FF ( Pin A) = − [( ε G q ) VF][ J scmax JF] FF ( Pin A) ,

(4.22)

where Pin is the optical input power to the solar cell and A is the cell’s crosssectional area. The negative signs come from Jm and Jsc being negative. The fill factor FF is conveniently defined as the ratio FF = Vm J m Voc J sc = ( J m J L )(Vm Voc )( J L J sc ) ,

(4.23)

which specifies the “squareness” of the light J–V characteristics (see Fig. 4.2), and it has a maximum value of one and accounts for the unavoidable potential energy loss in transporting the charge to the solar cell boundaries for maximum power and the unavoidable “dark” current losses of forward biased diodes as well as losses due to high series resistance values. The (qVoc/εG) = VF loss fraction is the loss when the Voc is less than the bandgap. The loss fraction JF = J sc J scmax describes when all the incident photons with energies greater than the bandgap εG do not contribute to the Jsc (i.e., the external quantum efficiency is kT/q [12] and 0 for Voc 95% reduction in dark current in the champion cell with a fill factor of 85% of figure 1.2 of the First Edition [5]). This indicates that theoretical maximum efficiencies are in the 85–90% range that falls to the ∼80–85% levels after downward adjustments for realistic material parameter values versus “ideal” ones (see [5]).

4.10

CONCLUSIONS

Fifty years of quantum mechanics theory development followed by 50 years of its application to solid-state electronic device problems have led to solar cell efficiency performance increasing from ∼1% to over 40% currently. Solar cell theory

CONCLUSIONS

103

continues to predict substantially higher efficiencies when high excited carrier concentrations are considered. As existence proofs, LEDs have achieved excited carrier concentrations at the high levels (that penetrate band edges at ∼1019 cm−3) where greatly improved performance is projected for solar cells. These high carrier concentrations correspond to diode operation where the quasi-Fermi levels split to the bandgap edges. They also correspond to terminal voltages that approximately equal the bandgaps of the photoactive regions. In retrospect, it is intuitive that such operation corresponds to the highest possible conversion efficiency. However, no current champion solar cells have come close to demonstrating open-circuit voltages at bandgap levels. The associated deficit is directly described by the voltage fraction VF shown in Figure 4.5 and identifies a promising area for substantial future device performance improvements. Prior modeling has projected 50% terrestrial efficiencies for two-junction cells at 1000X sunlight concentration that rise to 72% for 36 tandem junctions, ignoring the losses in the interconnect tunnel junctions [8, 54]. Bandgap open-circuit voltage values should assist with the challenges of exceeding 50% efficiency levels. The last 15 years has provided wide technical and market experience with solar cell configurations that depart dramatically from the traditional Shockley diode abrupt p/n junction geometry with one-dimensional symmetry. These include the a-Si, the point contact, the IBC, and the HIT solar cells. All of these except the point contact one are currently commercially successful products. And all but the first use crystalline silicon as the primary photo active component. All use textured surfaces to enhance light absorbance and trapping, and all use high doping of the regions next to the metal contacts to assure good ohmic contact formation and Fermi level alignments with the nearest band edges. Except for the HIT cell, none have applied the Nobel Prize-winning concept of a quantum well heterojunction barrier to shield excited carriers in a photoactive region from the high recombination properties of ohmic contacts. Combining the lessons learned from analyzing their behavior with the insights gained from LEDs, it appears that abrupt p/n junctions are not required for the highest efficiency performance. Indeed, under low injection limits, one is fundamentally blocked from achieving such Fermi levels bandgap splitting performance. So far, the non-abrupt configuration has not been applied to the multijunction III–V devices that currently provide the 40% efficiency levels of performance. They achieved their improvements mainly by reducing the losses of absorbed photon energy that lie above the bandgap of the absorbing material. The opportunities for improvement in improved voltage fraction VF arise from reducing the photon energy losses below the bandgap. The latter involves two components. One is the quasi-Fermi level bandgap splitting from high excited carrier concentrations. Direct bandgap materials are likely preferred due to their comparatively higher absorption coefficients to provide high light absorption in thinner layers and their absence of significant non-radiative recombination losses. The other is using heavily doped semiconductor regions next to the metal contacts to align the metal Fermi level to the associated band edge in the solar cell. The latter becomes particularly effective when implemented with the quantum well

104

SOLAR CELL DEVICE PHYSICS

heterojunction barriers. In addition, suggested options are recommended improvements to the crystalline silicon configurations for higher performance, mainly by implementing the quantum well and heterojunction barrier configuration pioneered with HIT solar cells. Summarizing the rationale for p-i-n quantum well concentrator solar cells, if the light-generated carrier concentrations can be made to equal or to exceed the standard doped semiconductor concentration levels, there is no longer any need for doping. Once generated, carriers have no memory of coming from either thermal ionization of dopants or from optical absorption of photons and thus reduce the resistive effects by equivalent amounts. However, elimination of doping can greatly increase the excited carrier lifetimes and the associated lightgenerated carrier concentrations, particularly when all cell boundaries are passivated or have heterojunction barriers shielding excited carriers from ohmic contacts’ high surface recombination. The highest possible efficiency conversion occurs by the absorption of photons with energies just above the quantum well bandgap that produce carriers that exit the solar cell with a voltage equal to the well bandgap value. Such voltage generation in the metal contact requires the metal Fermi levels to align with the well band edges, which only occurs at high doping levels and when the high light-generated carrier concentration splits the quasi-Fermi levels all the way to the band edges. Such highly doped semiconductor layers are typically optically dead and should thus be kept as thin as possible to reduce their light absorption losses. Carrier transport parameters fundamentally reduce the power delivered out of the cell by the fill factor FF ratio. However, much higher conversion efficiencies should be made possible than any current champion devices exhibit, as open-circuit voltages are made to approach quantum well bandgap values.

ABBREVIATIONS A—solar cell junction area a—atomic spacing of a semiconductor Al—aluminum AlGaAs—aluminum gallium arsenide AM1.5D—1.5 air mass direct component of sunlight a-Si—amorphous silicon ASTM—American Society of Testing and Materials c—speed of light in free space CdS—cadium sulfide CdTe—cadmium telluride CIGS—copper indium gallium diselenide Cu—copper CuxS—copper sulfide DC—direct current

ABBREVIATIONS

105

EN—black body radiation energy of N photons of frequency f E(x)—electric field as a function of position x exp—exponential f—frequency of a light photons fe—hyperbolic function describing effects of minority carrier electron surface recombination FF—fill factor fh—hyperbolic function describing effects of minority carrier hole surface recombination FSF—front surface field F(λ)—light flux intensity of wavelength λ f(ε)—probability that a quantum state of energy ε is occupied by an electron GaAs—gallium arsenide Ge—germanium h—Planck’s constant HIT—heterojunction with intrinsic thin layer i—intrinsic I—current IBC—interdigitated back contact ITO—indium tin oxide J—current density of electrons and holes J0—Shockely diode reverse saturation current density of electrons and holes J0e—electron contribution to the Shockely diode reverse saturation current density J0h—hole contribution to the Shockely diode reverse saturation current density J L1X —one sunlight-generated current density J Lc—concentrator light-generated current density JF—current density factor of J sc J scmax JL—light-generated current density from electrons and holes Jm—maximum power current density Jn—current density of electrons Jsc—short-circuit current density J scmax—light-generated current density when external quantum efficiency equals one JTH—threshold current density in SCLI devices k—Boltzmann’s constant K—quantum wave number from Schrodinger’s equation solutions L—width of the intrinsic or low-doped region ln—natural logarithm Mc—number of equivalent minima at the conduction band edge mde—electron density-of-states effective mass mdh—hole density-of-states effective mass md/mo—density-of-states effective mass ratio n—density of electrons n0—equilibrium electron carrier density ni—intrinsic carrier density nα—electron concentration boundary condition at the forward bias electron injecting contact

106

SOLAR CELL DEVICE PHYSICS

nβ—electron concentration boundary condition opposite the forward-bias electron injecting contact N—number of photons of frequency f emitted from a black body NA—hole (or acceptor) doping density NC—effective density of conduction band states ND—electron (or donor) doping density Nv—effective density of valence band states p—density of holes p—momentum of a photon p0—equilibrium hole carrier density Pin1 X —unconcentrated one sun optical input power to the solar cell Pinc —concentrated optical input power to the solar cell PERL—passive emitter, rear locally diffused p-i-n—p-type-intrinsic-n-type structure Pin—optical input power to the solar cell p/n—abrupt junction from p-type to n-type structure Pt.—point ptio—hole concentration in ith trap at equilibrium in a SCLI device QE(λ)—quantum efficiency of the solar cell at wavelength λ q—charge of an electron R—solar cell series resistance R—specific solar cell series resistance equal to series resistance R times junction area A r(λ)—the light reflectivity of the emitter surface at wavelength λ SCLI—space charge-limited current Si—silicon SiO—silicon oxide T—absolute temperature tB—solar cell base width t B′ —undepleted solar cell base width TCO—transparent conducting oxide tE—solar cell emitter width t E′ —undepleted solar cell emitter width TiO2—titanium dioxide UV—ultraviolet vde—diffusion velocity of electrons vdh—diffusion velocity of holes VF—voltage factor of qVoc/εG Vm—maximum power voltage Voc—open-circuit voltage VTFL—trapped-filled limit voltage in SCLI devices X—sunlight concentration ratio xB—base depletion region width xE—emitter depletion region width V(r)—potential energy function value at a space vector position r

REFERENCES

107

Wn—width of n contact region of IBC solar cell Wp—width of p contact region of IBC solar cell xn—x position of emitter depletion region edge xp—x position of base depletion region edge xw—x position of base boundary next to its metal contact Zn—zinc ΔnL—light-generated change in electron concentration ΔpL—light-generated change in hole concentration ε—dielectric constant εC—energy level of conduction band edge εF—Fermi energy Δε LFe—light-induced change in electron quasi-Fermi energy Δε LFh —light-induced change in hole quasi-Fermi energy δεn—electron energy loss moving from conduction band edge to metal Fermi level δεp—hole energy loss moving from valence band edge to metal Fermi level η—efficiency ηC—concentrator efficiency ε∗Fe—quasi-Fermi energy of electrons εF0—equilibrium Fermi energy ε*Fh—quasi-Fermi energy holes εG—bandgap energy εV—energy level of valence band edge λ—wavelength of a photon of frequency f ρo(x, Φ)—charge distribution as a function of position x in a SCLI device exposed to a light intensity Φ at zero bias voltage ρI(x, Φ, V)—the change in charge distribution as a function of position x in a SCLI device exposed to a light intensity Φ due to the application of a charge injecting voltage V μn—electron mobility π—the universal mathematical constant “pi” ΦK(r)—electronic wave function of quantum number K is a vector position in space r Φ—light intensity in a SCLI device ∇—mathematical operation “del”

REFERENCES [1] [2] [3] [4]

N. Herbert. Quantum Reality. New York, Doubleday (1987). Nobel Prizes. Available at http://nobelprize.org/. Accessed April 3, 2009 (2009). Bohr Model (1913). Wikipedia available at http://en.wikipedia.org/wiki/Bohr_ model. Accessed May 2, 2010. L. Partain and J. Sheldon. Quasi-Fermi level interpretation as potential energy. Lawrence Livermore Laboratory Report No. UCRL-83787, Lawrence Livermore National Laboratory, Livermore, CA, p. 16 (1980).

108 [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

[30] [31] [32] [33] [34] [35] [36]

SOLAR CELL DEVICE PHYSICS L. Partain, ed. Solar Cells and Their Applications. New York, Wiley (1995). D. Neamen. Semiconductor Physics and Devices, 3rd edition, pp. 25, 83–114, 327, 458. Boston, McGraw-Hill (2003). C. Kittel. Thermal Physics. New York, Wiley (1969). S. Sze. Physics of Semiconductor Devices, 2nd edition. New York, Wiley (1981). W. Shockley. Electrons and Holes in Semiconductors. Huntington, NY, Krieger (1976). J. Bardeen, Phys. Rev. 71, 717 (1947), and W. H. Brattain and W. Shockley, Phys. Rev. 72, 345. D. Chapin, C. Fuller, and G. Pearson. J. Appl. Phys. 25, 676–677 (1954). W. Shockley and J. Queisser. J. Appl. Phys. 32, 510–519 (1961). J. Hovel. Solar Cells, Semiconductors and Semimetals, Vol. 11. New York, Academic (1975). J. F. Geisz, D. J. Friedman, J. S. Ward, A. Duda, W. J. Olavarria, T. E. Moriarty, J. T. Kiehl, M. J. Romero, A. G. Norman, and K. M. Jones. Appl. Phys. Lett. 93, 123505 (2008). M. Green, K. Emery, D. King, Y. Hishikawa, and W. Warta. Prog. Photovoltaics Res. Appl. 15, 35–40 (2007). C. Kim and R. Schwartz. IEEE Trans. Electron Devices 16, 657–663 (1969). R. Schwartz and M. Lammert. Int. Electron Devices Mtg. 21, 350–352 (1975). R. Schwartz and M. Lammert. IEEE Trans. Electron Devices 24, 337–342 (1977). D. Carlson and D. Wronski. Appl. Phys. Lett. 28, 671–673 (1976). T. Ashley and C. Elliot. Semicond. Sci. Technol. 6, C99–C105 (1991). A. Prokopyev and S. Mesheryakov. Solid State Electron. 43, 1747–1753 (1999). O. Nichiporuk, A. Kaminski, M. Lemiti, A. Fave, and V. Skryshevski. Solar Energy Mater. Solar Cells 86, 517–526 (2005). A. Al-Omar. Solid State Electron. 50, 1656–1666 (2006). Z. Alferof, V. Andreev, and V. Rumyantsev. III-V solar cells and concentrator arrays. In High-Efficient Low-Cost Photovoltaics, V. Petrova-Koch, R. Hezel, and A. Goetzberger, eds., pp. 101–133. Berlin, Springer (2009). A. Rai, R. Bhattacharya, and Y. Park. Thin Solid Films 114, 379–398 (1984). L. Partain. J. Appl. Phys. 62, 4003–4005 (1987). L. Partain and D. Liu. Appl. Phys. Lett. 54, 928–930 (1989). N. Xu. The physical origin of prebreakdown electron “pin-holes.” In High Voltage Vacuum Insulation, R. Latham, ed., pp. 124–129. London, Academic (1995). N. Kaminar, D. Liu, H. MacMillan, L. Partain, M. Ladle Ristow, G. Virshup, and J. Gee. Concentrator efficiencies of 29.2% for a GaAs cell and 24.8% for a mounted cell lens assembly. In 20th IEEE Photovoltaic Specialists Conference, Las Vegas, NV, pp. 766–768. New York, IEEE (1988). D. King and B. Hansen. Validation of PDML and PTEL Primary Reference Cell Calibrations, Sandia National Laboratories Internal Memo (1991). M. Green, K. Emery, Y. Hishikawa, and W. Warta. Prog. Photovoltaics Res. Appl. 17, 85–94 (2009). H. Kroemer. Rev. Modern Phys. 73, 783–793 (2001). C. Hamaker. J. Appl. Phys. 58, 2344–2351 (1985). M. Klausmeier-Brown. Status, prospects and economics of terrestrial, single junction GaAs concentrator cells. In Solar Cells and Their Applications, L. Partain, ed., pp. 125–139. New York, Wiley (1995). Z. Fang and D. Look. J. Electron. Mater. 22, 1361–1363 (1993). V. Andreev, V. Khvostikov, V. Larionov, V. Rumyantantsev, E. Paleva, M. Shvarts, and C. Algora. 5800 suns AlGaAs/GaAs concentrator solar cells. In Proc. International

REFERENCES

[37] [38] [39] [40] [41] [42] [43] [44] [45]

[46] [47] [48] [49] [50]

[51] [52] [53] [54] [55] [56] [57]

[58] [59] [60]

109

Photovoltaic Science and Engineering Conference, pp. 147–148. Sapporo, Japan (1999). R. Swanson. Solar Cells 17, 85–118 (1986). R. Swanson. Device physics for back-side contact solar cells. 33rd IEEE Photovoltaic Specialists Conference, San Diego, CA, May 11–16, 2008. New York, IEEE (2008). W. Mulligan, M. Cudzinovic, T. Pass, D. Smith, N. Kaminar, K. McIntosh, and R. Swanson. SunPower Corporation, assignee, Solar cell and method of manufacture. U.S. Patent 7,339,110 B1, March 4 (2008). E. Van Kerschaver and G. Beaucarne. Prog. Photovoltaics Res. Appl. 14, 107–123 (2006). N. H. Fletcher. Proc. IRE 43, 551–559 (1955). M. Green. Solar Energy 74, 181–192 (2003). Y. Tsunomura, Y. Yoshimine, M. Taguchi, T. Baba, T. Kinoshita, H. Kanno, H. Sakata, E. Maruyama, and M. Tanaka. Twenty-two percent efficienct HIT solar cell, Solar Energy Materials & Solar Cells, 93, 670–673 (2009). M. Taguchi, H. Sakata, Y. Yoshimine, E. Maruyama, A. Terakawa, M. Tanaka, and S. Kiyama. An approach for the higher efficiency in the HIT solar cells. In 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp. 866–871. U. Das, S. Bowden, M. Burrows, M. Lu, and R. Birkmire. Effect of surface passivation on Si heterojunction and interdigitated back contact cells. 17th Workshop on Crystalline Silicon Solar Cells & Modules: Materials and Processes, Vail, CO, August 5–8, 2007. D. Reynolds, G. Leies, L. Antes, and R. Marburger. Phys. Rev. 96, 533–534 (1954). L. Partain. J. Appl. Phys. 54, 5218–5226 (1983). R. Moorthy, L. Partain, D. Okubo, and D. Henderson. Space-charge-limited current and capacitance in CuxS/CdS solar cells. In IEEE International Electron Devices Meeting, Washington, DC, December 1979, pp. 442–445. New York, IEEE (1979). L. Partain, G. Armantrout, J. Leong, and P. Warter. J. Electron. Mater. 9, 467–484 (1980). A. Catalano, R. Arya, C. Fortman, J. Morris, J. Newton, and J. O’Dowd. High performance, graded bandgap a-Si : H solar cells. In 19th IEEE Photovoltaic Specialists Conference, New Orleans, LA, April 5, 1987, pp. 1506–1507. New York, IEEE, (1987). L. Partain. J. Appl. Phys. 61, 5458–5466 (1987). M. Lampert and P. Mark. Current Injection in Solids. New York, Academic (1970). L. Partain. J. Appl. Phys. 63, 1762–1767 (1988). C. Henry. J. Appl. Phys. 51, 4494–4500 (1980). R. Nelson and R. Sobers. J. Appl. Phys. 49, 6103–6108 (1978). J. Baur, F. Baumann, M. Peter, M. Engl, U. Zehnder, J. Off, V. Kuemmler, M. Kirsch, J. Strauss, R. Wirth, K. Streubel, and B. Hahn. Phys. Status Solidi C 52, S905–S908 (2009). A. Laubsch, M. Sabathil, M. Strassburg, W. Bergbauel, M. Peter, H. Laubauer, N. Linder, B. Hahn, and K. Streubel. Improving the high current efficiency of LEDs. Available at http://spie.org/documents/Newsroom/Imported/1476/1476_5485_0_ 2009-04-09.pdf. Accessed June 6, 2009 (2009). N. Holonak. Am. J. Phys. 68, 864–866 (2000). A. Erchak, D. Ripin, S. Fan, P. Rakich, J. Joannopoulos, E. Ippen, G. Petrich, and L. Kolodziejski. Appl. Phys. Lett. 78, 563–565 (2001). G. Zhang, Z. Yang, Y. Tong, Z. Qin, X. Hu, Z. Chen, X. Ding, M. Lu, Z. Li, T. Yu, L. Zhang, Z. Gan, Y. Zhao, and C. Yang. Opt. Mater. 23, 183–186 (2003).

PART II TERRESTRIAL SOLAR CELL ELECTRICITY

5 CRYSTALLINE SILICON SOLAR CELLS AND MODULES LEONID RUBIN Day4 Energy Inc.

5.1

INTRODUCTION

During the last decade, the PV industry has demonstrated an impressive 30–50% annual growth rate and, as a result, in 2008, the total production capacity has surpassed 4 GW with cumulative annual revenues reaching $20 billion. There are several reasons for this success. Initially, in Japan and later in Germany, government-supported programs for the PV industry development played extremely important roles for attracting capital investments for the PV industry. Government subsidy programs helped bridge the gap in cost between PV-generated electricity and conventional electrical grid prices, thus creating a viable marketplace for PV product manufacturers. There were also a number of other factors that have contributed to the rapid growth of the PV industry. The sudden decline in the microelectronic industry due to the Internet bubble and the availability of excess silicon feedstock production capacity provided silicon feedstock for the PV industry. The growing concern about global climate change due to greenhouse gas emission combined with oil and gas price rises suggested that the development of alternative energy sources should become an important part of government policy. Substantial progress was also made in PV cell efficiency, and PV cells were being successfully mass produced as well. One may say that it was a unique “once-in-a-life-time” opportunity for the PV industry and so, the PV companies took full advantage of it. Nevertheless, the electric energy produced with solar cells continues to be noncompetitive with traditional nonrenewable sources. The PV industry is still strongly dependent on subsidies, and PV provides a negligible contribution to the Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

113

114

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

overall energy generation market. It was expected that the economy of scale would eventually result in cost reduction sufficient to make the PV industry cost competitive with conventional sources of electrical power. There is no doubt that in recent years, some cost reduction associated with PV cells and module mass production has been achieved. But this progress has not been sufficient so far to make the PV industry cost competitive. One of the key roadblocks to further cost reduction is in the overall lack of technological innovation in PV cell and PV module manufacturing. Ironically, the very subsidy system that resulted in PV product demand explosion is also partially responsible for the lack of technological innovation during this same period. With demand exceeding supply by a wide margin, manufacturers have focused their efforts and capital on expanding production capacity, economies of scale, and profit. Crystalline silicon technology development has been largely “on the back burner.” The fact is that the majority of PV cell manufacturers are utilizing identical production technologies. Without manufacturing technology improvements, the difference between the efficiencies of mass-produced and advanced PV cells has been increasing in the last decade and is now pronounced. Even less diversity is found between PV module producers. They continue to use 35-year-old tabbing and stringing technology. In this chapter, we will present a brief overview about the design and performance of conventional mass-produced crystalline silicon PV cells and the main factors limiting their light-to-electric energy conversion efficiency. We will describe novel technologies that can lead to PV cell efficiency improvement. We will then describe some advanced PV cells with outstanding efficiency. The current status of PV module fabrication technologies and the possibilities for their improvement will be described. Finally, the optimization of PV cell and module designs to maximizing annual kilowatt hour generation capacity for PV power generation systems will be addressed.

5.2

INDUSTRIAL CRYSTALLINE SILICON PV CELL

Since the PV industry does not require the same silicon material purity as for microelectronics, it was initially possible to utilize the waste from microelectronic silicon production. This is referred to as solar-grade silicon for initially monocrystal and later for mc silicon wafer production. The most widely used technology for making monocrystal silicon is the CZ. For CZ silicon crystal growth, a silicon monocrystal seed is dipped into a crucible of molten high-purity silicon and is withdrawn slowly pulling a cylindrical monocrystal. This silicon crystal is then sliced into thin CZ wafers that are used for CZ PV cell production. Production of mc silicon typically involves a casting process in which molten silicon of slightly lower grade, referred to as solar-grade silicon, is directly cast into a typically squared mold where it is slowly crystallized into 3D ingots that are then sliced into square mc wafers. The temperature cooling gradient profile is one of the main factors that determine the impurity distribution divided between the

INDUSTRIAL CRYSTALLINE SILICON PV CELL

115

silicon microcrystal and the boundaries between them. Segregation of impurities into the grain boundaries is referred to as gettering and influences the quality of the mc wafers. It is known that the PV cells produced from CZ silicon typically have higher efficiency than the PV cells from mc material. At the same time, the throughput from the CZ process is about two times lower compared with the mc process, and the mc process consumes less energy and material with a lower cost. Therefore, the price of mc silicon wafers is lower for PV cell manufacturers, but there is a compromise leading to lower PV cell conversion efficiency. Recently, there has been substantial progress in both single-crystal and mc process technology leading to higher cell efficiencies for both. Currently, the market share of mc PV cells is slightly higher compared with CZ. Due to growing demand from the PV industry, a new solar-grade crystalline silicon production capacity has been established based not only on conventional solar grade but also on some new metallurgical silicon purification technologies. However, while the silicon material availability has increased to meet the PV industry demand, unfortunately, the PV cell efficiency produced from metallurgical silicon has been reduced. It is now evident that production of high-efficiency and low-cost CZ or mc PV cells must be based on sufficiently good quality initial feedstock silicon material. Conventional crystalline silicon PV cells are generally produced from p-type monocrystal or mc semiconductor wafers of 150- to 300-μm thickness. The front side of industrial PV cells is typically doped with phosphorus, thus creating an n-doped area of 0.1- to 2-μm thickness and a resistivity typically of 50–65 Ω per square. It is also possible to produce PV cells from monocrystalline n-type crystalline silicon with a boron-doped p-type front surface. These n-type PV cells do not suffer from light-induced efficiency degradation as often happens with monocrystalline p-type PV cells. The junction between the n-doped surface area and the p-doped bulk silicon creates a charge separation region with a strong dipole electric field. The surface area is referred to as the emitter and the bulk region is referred to as the base. When a PV cell is illuminated by light, photons produce electron–hole pairs and the dipole electric field provides for a separation of these charges. This displacement of free charges results in a voltage difference between the p and n regions of the PV cell. The emitter technical characteristics such as thickness, doping profile, and doped element concentration in the PV cell surface region are extremely important for PV cell efficiency. Conventional PV cells are typically equipped with current collecting metal contacts on the front and rear sides provided for conducting electric current when the p and n regions are connected through an external load. Figure 5.1 shows a drawing of a typical square mc silicon cell showing the front-side collection metallization pattern comprising screen printed fingers and two bus bars that collect electric current from the PV cell. The conversion efficiency of solar energy into electric energy is considered to be the main PV cell characteristic. Under illumination, PV cells generate a maximum voltage value referred to as the open-circuit voltage or Voc when the

116

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

70

Bus bar

Bus bar

Fingers

Figure 5.1. Top view drawing of a typical 156-mm2 MC silicon cell.

external circuit is open. If the external circuit is shorted, then the light-generated electric current reaches its maximum value referred to as the short-circuit current or Isc. The dependence of the current I versus the voltage V for different values of the external load is known as the I/V curve where the values of Isc and Voc are the intersection points with the current and voltage axis, respectively. Unfortunately, the maximum value of generated electric power is not the product of Isc × Voc. This maximum is never achieved due to inevitable power losses. The real value of generated power may be evaluated from the experimental I/V curve containing the so-called maximum power point referred to as Pmpp, at which I, V, and P reach their maximum values Pmpp = Impp × Vmpp. The ratio between Immp × Vmmp/ Isc × Voc represents PV cell general power losses and is referred to as the fill factor (FF). The PV cell energy conversion efficiency η is expressed as H = Voc × Isc × FF E , with FF being the fill factor and E being the light energy in watt per PV cell area.

5.3

EFFICIENCY LIMITATIONS

Although the theoretical efficiency of a crystalline silicon PV cell approaches 29% and the world record for the best silicon PV cell is 24.3%, the average efficiencies

EFFICIENCY LIMITATIONS

117

of a typical industrial monocrystalline or mc PV cells are 17% and 16%, respectively. There are several causes limiting PV cell efficiency. First of all, there are limitations based on the fundamental properties of silicon semiconductors. Photons with energies less than 1.12 eV are lost due to the silicon semiconductor band gap, and photons with energies exceeding 1.12 eV loose energy via dissipation into heat. The maximum value of the PV cell open-circuit voltage Voc is substantially lower than the silicon semiconductor band gap because it is defined by the quasi-Fermi level separation. The typical Voc values for conventional monocrystalline and mc PV cells are about 625 and 610 mV, respectively. Even high-efficiency PV cells demonstrate a Voc value not higher than 722 mV.

5.3.1

Optical Losses

The theoretical Isc maximum value for a crystalline silicon PV cell may reach 42 mA/cm2 under AM 1.5 sun irradiation. In reality, it is hard to reach this value because the Isc value of a PV cell is not determined simply by the incident solar energy intensity but instead, it depends on the fraction that is absorbed by the PV cell and converted without losses into electric energy. The main problem with crystalline silicon semiconductor material is that it absorbs light very poorly due to its high refraction index of about 3.9 and corresponding high light reflection of about 40%. The most efficient way to solve this problem is to utilize light trapping. Modern wet chemical etching and laser processing technologies allow for arranging for different types of high-efficiency textured structures like inverse pyramids or honeycomb-type structures on monocrystalline and mc PV cells. Light trapping textures can decrease light reflection to below 10%. An additional decrease in refection is achieved after antireflective dielectric coating. Sample AR coatings can consist of SiNx with a refraction index of up to 2.2. They are applied on top of the initially textured surface of the PV cell. This process can decrease the light reflection further to less then 4%. It is worth noting that most advanced technologies for achieving extremely efficient light trapping surfaces are kept as proprietary know-how. An additional optical loss is associated with the partial transmission of light in the IR spectrum. This effect is most pronounced when the PV cell thickness is less than 250 μm. Therefore, there is a need to utilize backside mirror reflective coatings, thus preventing efficiency decline due to insufficient absorption of light. The area that is occupied by the metal current collection contacts on the front side of the PV cell is referred to as the shading area. It impacts the optical losses by preventing solar radiation from reaching the surface of the PV cell and thereby generating electric current. This shading area typically occupies up to 7–10% of a PV cell’s available front surface, thus decreasing the PV cell efficiency accordingly. Later in this chapter, we will describe several promising current collecting technologies that have been developed for decreasing front-side shading.

118

5.3.2

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

Recombination Losses

The photon-generated electrical charges have to diffuse to the charge separation area and further to the current collecting contacts located on the front and rear sides of a PV cell. In general, this diffusion process encounters two main types of power losses. Charge recombination losses can occur in the semiconductor bulk and on the PV cell front and rear surfaces. Recombination in the bulk strongly depends on semiconductor impurities and crystal dislocation concentrations. These defects are responsible for energy states that function as efficient trap and recombination centers. Free electrons captured at these centers are passing to the valence band dissipating energy as heat. This type of recombination is typical for mc and monocrystalline PV cells and strongly depends on materials purity. In the case of mc material, bulk recombination may be reduced by impurity gettering to the microcrystal boundaries during casting crystallization and/or by bulk passivation, for example, with hydrogen during the SiNx AR coating application. It is obvious that PV cells produced from metallurgical silicon are particularly exposed to this type of recombination, thus demonstrating substantially lower efficiency. The recombination on the PV cell surface depends on the density of defects on the surface due to silicon crystal edge breakdown, as well as the presence of the metal current collecting contacts, and dopant concentration on the surface. Several technological methods have been developed and introduced into mass production to minimize surface recombination, which may be referred to as surface passivation. Charge recombination on the front and rear sides of a PV cell may be substantially reduced by passivation with a thin layer of a dielectric material, such as SiO2, SiNx, or SiC, by employing industrially available technologies and equipment [1, 2]. Recombination on a p-type PV cell rear surface may also be decreased by doping, for example, with boron or aluminum, thus creating a p+ layer or BSF.

5.3.3

Resistivity Losses

Conventional high-quality silicon monocrystalline PV cells of 156-mm2 area with optimized optical loss minimization typically generate Isc values of up to 8.5 A. Keeping in mind that the Voc value remains practically at the relatively low level of 625 mV, it is a challenge to collect this current with minimum power losses (I2 × R) because the PV cell resistivity starts to have an extremely high impact on the PV cell efficiency. Overall, the PV cell resistivity includes the following main components: (1) the series resistance (R) of the current collecting pattern on the front and rear sides of the PV cell, (2) the parallel or shunt PV cell resistance (Rsh), and (3) the contact resistance between the front and back-side metal patterns with the emitter (Rem) and bulk silicon (Rbl), respectively. The typical technology for making front- and back-side metal contacts is based on a conventional screen printing process in which an electrically conductive paste, containing silver and/or aluminum powder particles, is printed through a

EFFICIENCY LIMITATIONS

119

screen onto the front and back surfaces of a PV cell. The front-side screen is typically configured to produce a plurality of thin parallel line contacts referred to as “fingers” connected typically to two or three wider lines referred to as “bus bars.” The fingers collect the electrical current from the front side of the PV cell and transfer it to the bus bars. Metal ribbons are typically soldered to the bus bars to conduct electric current to the neighboring PV cell and further to the electrical circuit. Typically, the width and the height of each finger are approximately between 110 and 120 μm and between 10 and 20 μm, respectively. This corresponds to a height-to-width aspect ratio ranging between 0.1 and 0.15. It is evident that if the ratio between finger width and height is improved by making them narrower and thicker, finger conductivity will be improved along with decreased shading. It has been demonstrated that the utilization of hot melt pastes with modern heated screen printing equipment allows for printing fingers with 1 ms that allows minority carriers to diffuse from the illuminated front surface through the entire wafer thickness to arrive at the junction and current collecting contacts of both polarities at the rear side. Front-side recombination has been additionally decreased by introducing n+-doped and SiO2 passivating layers on the front side. The back side of the IBC PV cell is processed in a very special manner. Initially, “interdigitated” n+ and p+ parallel narrow nonoverlapping strips between the PV cell’s opposite edges are produced by sequential diffusion processes. An efficient electrical insulation is built between these strips to guarantee high shunt resistance. The entire back-side surface of this PV cell is covered by a SiO2 layer providing efficient rear-side passivation. Contacting holes through this back-side SiO2 layer are produced in precision alignment with corresponding n+ and p+ strips. Metallic contacting narrow fingers are further printed in a precision-aligned manner along corresponding contacting holes, thus providing electric contact through the holes with the underlying n+ and p+ strips. Two terminal bus bars are screen printed on the opposite edges of the IBC PV cell’s rear side in such a way that one of them is connected to all n+ fingers and the other to the p+ fingers. These terminal bus bars are used for PV cell testing and interconnection in series by means of special tabbing during PV module production. These PV cells demonstrate high current (>40 mV/cm2) not only due to efficient light trapping and low shading but also due to better external quantum efficiency. The blue response is improved due to a highly n+-doped front diffusion and emitter localization on the back side, thus eliminating the front-side dead layer associated with a conventional PV cell structure with emitters localized on the front side. A long-wavelength response in the near IR spectral region has been improved as a result of a perfect rear-side passivation provided by the SiO2, which covers the entire rear-side area except for the contacting holes. Optimal front- and rear-side processing provides for an optical thickness for the PV cell about six times its actual thickness. Due to perfect frontand rear-side passivation, a Voc of about 700 mV has been achieved. An optimized screen printing process allows the FF to reach a value of >80%. In contrast to conventional crystalline silicon cells that demonstrate a Voc value decline of

126

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

2.2 mV/°C, the SunPower cell’s Voc dependence on temperature is relatively lower at about 1.87 mV/°C, thus decreasing its temperature dependence for power losses to 0.38%/°C compared to 0.5%/°C typical for conventional PV cells. This differentiation certainly will play an important role in securing annual kilowatt hour power generation. Although SunPower asserts that the production cost of these cells is low, it is evident that the cost should be substantially higher when compared with conventional PV cell fabrication due to the more expensive silicon material and the technology complexity. Several PV companies have made attempts to produce PV cells employing back contact concepts, but none of them have succeeded so far in introducing their designs into mass production mainly due to production technology complexity and high cost. The fact is that practically all crystalline silicon PV cell manufacturing companies are using screen printing technology for current collection, although it is evident that the limited conductivity of screen printed metallic patterns represents a real bottleneck for further PV cell efficiency improvement and cost reduction. In reality, the PV cell is just a small electric generator. Imagine what would happen if some gas electric generating plant were renovated and then its production capacity is doubled, and it is then placed back in service while the corresponding transmission line is kept on the same level? Isn’t it evident that a large portion of its additional electric energy will dissipate into heat due to parasitic losses in the transmission line? The same happens when conventional screen printing technology fails to collect electric power without parasitic losses from more efficient PV cells. Therefore, there is an urgent request from the PV industry for a more efficient current collecting technology.

5.5.6

Day4 PV Cell

In 2003, Day4 Energy Inc. developed a novel concept for a PV cell with improved current collection hereinafter referred to as D4 technology or D4 electrode or D4 PV cell [18, 19]. This concept is based on a proprietary current collecting electrode (Fig. 5.3) composed of copper wires coated with low melting alloys that are spaced apart and imbedded into an optically transparent adhesive layer that in turn is firmly

Film

Adhesive Wire Alloy

Figure 5.3. D4 wire-tape electrode concept.

EXAMPLES OF NOVEL PV CELLS

127 Bus bar

Wires

Wires

PV cell

Fingers

Figure 5.4. The placement of the D4 wire-tape electrode on a PV cell.

adhered to the supporting transparent polymeric film. Since the thickness of the adhesive layer is lower than the thickness of the wire it secures, then at least part of the wire protrudes above it. One may see (Fig. 5.4) that the electrode wires are soldered to the copper bus bar of about 5- to 10-mm width and 50- to 200-μm thickness. The bus bar is typically positioned outside the PV cell perimeter, thus not occupying its front surface, thereby preventing shading. The D4 electrode is placed on top of the front side of the PV cell in such a way that its wires are oriented in a transverse direction in respect to the cell’s screen printed fingers, thus allowing each wire to be placed on top of each and every grid finger. The entire composite structure is exposed to the vacuum lamination process comprising sequential vacuuming, heating, and pressing. During this process, the air is pumped out, the alloy melts, and under pressure, the electrode wires become firmly soldered to the screen printed fingers. At the same time, the electrode’s adhesive layer softens and under pressure firmly fixes the polymeric film onto the PV cell front surface. The D4 electrode secures a very low ohmic contact between the electrode wires and the screen printed aluminum on the rear side of a PV cell due to the alloy properties and strong mechanical compression during the vacuum lamination. Due to the large numbers of soldering points between the electrode wires and the screen printed fingers, this electrode secures a substantially lower risk of electrical layout failure when compared with the conventional tab spot soldering using the conventional screen printed bus bars. It is known that resistive power losses are proportional to the square of the current flow distance. Figure 5.5 shows a typical screen printed cell with screen

128

CRYSTALLINE SILICON SOLAR CELLS AND MODULES Bus bar Wires

4

70

Bus bar

Bus bar

Fingers

Figure 5.5. Comparison between the conventional and D4 PV cell.

printed bus bars on the left in contrast to a D4 cell with a wire-tape electrode on the right. Referring to Figure 5.5 (left), the typical length that the electric current must flow through the screen printed finger before reaching the screen printed bus bars is not less than 35 or 20 mm in the case of a 6-in.2 conventional PV cell that utilizes two or three screen printed bus bars. In contrast, in the case of the D4 PV cell shown in Figure 5.5 (right), with the same size PV cell with the same grid pattern, the D4 electrode allows replacing two or three screen printed bus bars with 40 D4 electrode wires, thus reducing the current flow distance along the screen printed grids to only 2 mm. One may see that these wires are soldered to the bus bar provided for series interconnection with the following PV cell. Therefore, the resistive losses for the case of the conventional current collection through bus bar ribbons are at least 20 times higher when compared with the D4 current collecting technology. Employment of the D4 technology provides several additional advantages. First, it allows eliminating conventional screen printed bus bars from the front side, thus increasing the PV cell’s Rsh value and removing silver pads from the back side of PV cells, resulting in lowered recombination losses due to improved front- and rear-side passivation. Removal of silver/aluminum pads on the rear side allows eliminating one screen printing step and one drying step. It also decreases the silver paste consumption by 40% along with lowering production cost with no compromise in the cell efficiency. In fact, the D4 technology improves the PV cell efficiency by >0.1% absolute mainly due to higher Voc and FF values. In addition, it opens the possibility to introduce into mass production PV cells with substantially narrower screen printed fingers. It has been demonstrated that after industrial conventional multicrystalline (MC) PV cells are equipped with D4 electrodes, FF values of not less than 77% with screen printed fingers of 70-μm width and less than 6-micron height provide an improved cell efficiency up to 0.5% absolute due

PV MODULE

129

to lower shading and better front and rear-side passivation. Recent results demonstrated that production costs of the PV cells with narrow fingers may be substantially reduced due to >250% lower silver paste consumption relative to conventional PV cells with two bus bars. It should be appreciated that D4 technology opens a wide range of possibilities to introduce into mass production novel low cost and more efficient PV cells based on several promising technological concepts such as extremely narrow current collection metallic patterns produced either by screen printing or electroless plating, or aerosol, or ink-jet noncontacting direct printing, or laser-fired contacts that have been already developed but failed to find commercial application so far. There is realistic possibility to employ D4 technology for electric power collection and interconnection of the back contact PV cells, thus decreasing their production cost and simplifying PV module production.

5.6

PV MODULE

In contrast to the extensive progress in PV cell technology, the development of the production technology for PV modules with crystalline silicon PV cells has remained virtually unchanged for more than 30 years.

5.6.1

Conventional PV Module Production Technology

Since a PV cell is actually is a source of low-voltage ≈0.6 V with a DC electric current of about 34 mA/cm2 in sunlight, there is a need to interconnect a large number of PV cells in series in order to achieve a necessary and useful value for the DC voltage that in turn may be converted into AC by means of an inverter. Neighbored PV cells typically are interconnected in series by means of tinned copper tabs that are spot soldered onto the front side of the bus bars of a first PV cell and then onto the silver pads on the rear side of the next sequential PV cell. A certain number of PV cells interconnected in series are known as a PV string. Strings in their turn are interconnected in series by means of tinned copper bussing, thus producing a PV module layout. PV modules with series-interconnected PV cells perform optimally only when all the series-interconnected PV cells are illuminated with an approximately similar light intensity. However, if even one PV cell within the PV module layout is shaded while all other cells are illuminated, the entire PV module is adversely affected, resulting in a substantial decrease in power output from the PV module. In addition to temporary power loss, the module may be permanently damaged as a result of cell shading because when a PV cell is shaded, it starts to act as a large resistor rather than as a power generator. In this situation, the other PV cells in the PV string expose the shaded cell to a reverse voltage that drives electric current through this large resistor. This process may result either in the breakdown of a shaded PV cell or in heating it to such a temperature that it can destroy even the

130

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

entire PV module if a very high temperature persists. In order to eliminate the risk of a PV module damage in the event of shading, practically all PV modules employ BPD connected across each of the PV string and/or an entire module depending on specific PV module design and the quality of the input PV cells. The number of PV cells in a single PV string depends on the PV cell quality, namely, the ability to withstand back-voltage breakdown that each PV cell may be exposed to if one of them within the PV string is shaded. For example, for PV cells of good quality that can withstand back-voltage breakdown of 14 V and given that each PV cell generates a Vmax = 0.56 V, then the number of PV cells in one string should not exceed 24. Given that PV cells produced from metallurgical silicon typically have lower quality and back-voltage breakdown for these cells is not higher than 7 V, using them in PV strings comprising more than 12 cells is not recommended. Although employment of the BPD allows protecting the PV cells and PV strings against damage, it also causes substantial power losses of a PV module because the shading of just one PV cell results in an entire PV string switch-off. According to the industry estimate, almost 30% of kilowatt hour annual generation may be lost in field condition due to different sources of shade. Therefore, there is a need to optimize a PV module lay out in order to secure not only sufficient shading protection of PV cells but also to minimize annual kilowatt hour losses of PV systems as well. Since PV modules are generally expected to operate outdoors for typically 25 years without degradation, their construction must withstand various weather and environmental conditions. The front side of a typical PV module construction involves the use of a transparent sheet of low-iron tempered glass. The PV cell strings are sandwiched between sheets of polymeric encapsulant material, such as ethylene vinyl acetate, or thermal plastic material, such as polyvinyl butyral. An array of PV cells is placed onto the polymeric encapsulant material in such a way that the front sides of the cells face the transparent glass sheet. The back side of the array is covered with an additional layer of encapsulant material and a backsheet layer of weather protecting material, such as Tedlar® by DuPont, or a glass sheet. The additional layer of encapsulant material and the back-sheet layer typically have openings to provide for terminal electrical conductors to be passed from the PV cell strings through the back-side encapsulant layer and back sheet of weather protecting material to the outer surface for connecting with the electrical load through a junction box. For a PV module having an array of two PV strings, typically four conductors are arranged to pass through the openings so that they are all in proximity with each other so they can be terminated in a junction box mounted on the back-sheet layer. The glass, encapsulant layers, cells, and backsheet layer are typically vacuum laminated at about 14–160°C to build a free-of-air bubbles, strong bound structure that protects the PV cells from moisture penetration from the front- and back sides and also from the edges. The electrical interconnections of the PV strings and connections to BPDs are made in the junction box. The junction box is sealed on the back side of the PV module and is equipped with electric cables for neighbored PV module interconnection.

PV MODULE

131

An aluminum frame extends around the perimeter of the PV module and protects it against damage, provides mechanical strength against wind and snow loads, and facilitates mounting of the module to a support. At the same time, it is possible to employ PV modules without the aluminum frame if external supporting structures are sufficient to provide necessary mechanical strength. The fabrication of PV modules as described above with conventional technology is quite complicated and expensive. Layout of the PV module before lamination requires a separate step of “bussing” in which PV strings are electrically interconnected typically by means of tinned copper busses that increase the area that is occupied by the PV module, thus decreasing its conversion efficiency. Due to differences in thermal expansion coefficients between the copper and silicon and glass, there is a certain risk that the soldered spots between conventional tabs and front-side bus bars and rear-side silver pads may break causing PV module irreversible damage under inevitable changes of ambient temperature. There is an additional risk especially associated with utilization of thin PV cells. Exposing cells to spot soldering may result in PV cell breakage due to local heating and pressure. Once PV cells are interconnected in series by means of conventional technology, the series resistance of the produced PV module eventually exceeds the sum of all PV cell resistances due to the additional impact of soldering points, tabs, and bussing resulting in PV module FF value decline and corresponding efficiency losses when compared to the efficiency of the PV cells utilized in the module’s fabrication. For example, when PV cells with an average FF value of ≈76% are used to produce a 48-cell PV module, the module FF value typically will be lower than 73%. In other words, excessive series resistance associated with conventional PV module production technology is responsible for about 4% power loss when compared to the power that the input PV cells were capable of generating before being interconnected in the PV module. These insufficiencies become even more pronounced when modern highefficiency PV cells are utilized for PV module production. This is a clear demonstration of a growing conflict of interest between PV cell and PV module producers. In fact, PV cell producers are motivated to build PV cells with increased power output combined with decreased production cost. The easiest way to achieve this goal is to make PV cells thinner, thus securing economy of silicon material and to increase PV cell size, thus increasing the production capacity of existing manufacturing equipment with minimal additional investments. During the last 5 years, the size of PV cells has been increased from 100 × 100 cm2 to 125 × 125 cm2 and further to 156 × 156 cm2. Even a larger cell of 210 × 210 cm2 area has been developed and tested. PV module producers have been forced to buy these cells for an increased price based on dollar per Wp, but in return, they do not have sufficient benefit because thinner cells become more fragile and experience more pronounced bowing with corresponding higher-yield losses due to higher risk of breakage. To minimize this risk, there is a need for additional investment for more sophisticated production equipment for tab soldering and general handling. At the same time, conventional PV module fabrication technology does not allow increasing PV cell

132

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

power output at PV module level if the module series resistance is not decreased accordingly. Another challenge that PV module producers have to deal with is the necessity to achieve certain voltage values that a number of interconnected PV modules must generate per occupied area in order to secure the inverter’s most efficient operation. It is evident that although increased size of more efficient PV cells results in higher power output at the PV cell level, it eventually has a negative effect on the PV module performance due to the decreased voltage value since the number of PV cells per occupied area is diminished. Yet another challenge that PV module producers have experienced is the necessity to redesign the PV module layout to allow for an increasing number of PV strings now composed of a lower number of PV cells as required for cells made with metallurgical-grade silicon. Another example of a growing conflict of interest between PV cell and PV module producers is the employment of a PV cell with a selective emitter that demonstrates a higher efficiency in the blue spectral region. This advantage allows PV cell manufacturers to sell their cells at a higher price. At the same time, PV module producers do not observe the same efficiency gain on the PV module level not only because the glass and encapsulant materials substantially cut off the blue light but also due to the inevitable accumulation of dirt on the front side of PV modules. These general considerations reflect an urgent market request for developing novel, simpler, flexible, and cost-efficient technologies that are capable of utilizing more efficient PV cells in PV module production without loosing generated power.

5.6.2

Day4™ Technology for PV Module Production

One such promising technology has been recently developed at Day4 Energy Inc. [20, 21]. It was further demonstrated [22] that PV cells may be interconnected in series by means of electrically connecting the front-side D4 electrode of the first PV cell with the back-side D4 electrode of the sequential PV cell via the tinned copper bus bar. Since the thickness of this bus bar may be ≤100 μm, it securing extremely low series resistance of this interconnection. This approach allowed replacing conventional tabbing and stringing technology along with simplifying the PV module layout. New technology provides an elegant solution to produce U-type PV strings without conventional bussing by just turning each of two sequential PV cells with front and back electrodes by 90° before placing it on top of the rear-side electrode of the previous PV cell. It further allows eliminating conventional string interconnection by means of bussing, thus not only simplifying and optimizing PV module layout but also decreasing its production cost due to material and production step reduction. At the same time, PV module efficiency is increased due to its lower series resistance and economy of space occupied by conventional bussing. It is worth to notice that flexibility of the Day4 technology allows designing and producing PV modules with an increased number of PV strings, thus improving protection against shading and preserving annual generated kilowatt hour energy.

CONCLUSION

133

D4 PV module technology and specialized production equipment were introduced into mass production in 2006 after being UL and TUV certified. To date, about 40 MW of these PV modules have been installed on European, North American, and other markets and have received positive references from customers confirming excellent and reliable performance of installation. According to neutral party test results, installations with these PV modules demonstrate higher values of annually generated kilowatt hour in comparison with neighboring installation employing PV modules from competitive producers due to sufficiently higher shunt resistance of Day4 PV cells without bus bars, resulting in higher conversion efficiency at low light intensities. Depending on the quality of the input PV cells, the average power output of a mass-produced D4 PV module containing 48 MC PV cells is not less than 175 W. The efficiency of the best produced D4 module so far is 15%, while the average is typically >14% relative to the30 years of numerous numbers of PV systems’ reliable performance in different geographical locations. That is why the 25 years warranty on the crystalline silicon PV modules is based on a real solid ground. Therefore, we believe that the crystalline silicon PV industry will continue to be attractive for the long-term investments and to be the core basis for industrial solar electric energy generation. The material presented above illustrates the high potential for cost reduction and efficiency improvement practically at all steps of the crystalline silicon PV cell and module production that may be sufficient to make PV electric energy cost competitive with conventional nonrenewable sources. The challenge is how to convert this potential into sufficient cost reduction of kilowatt hour generated in-field conditions. It might sound strange but the current financial crisis in fact has created a moment of truth that is forcing the PV industry either to change its business model or to die. During the last 8 months, there has been an amazing price reduction for the PV cells (>40%) and PV modules (>50%). Keeping in mind that last year the Si-based PV cell contributed about 70% of the PV module cost and now it is lowered to 55%, there is sufficient potential for further PV module and kilowatt hour cost reduction. This trend must be combined with a modified PV industry business model that must change its current focus from dollar per Wp toward dollar per kilowatt hour and ROI. The selling policy must also be changed from the current 25-year guarantee for PV module Wp performance toward the 25-year guarantee of a certain kilowatt hour annual generation. We believe that the success and survival of the PV industry depends on the implementation of this ambitious but feasible program.

ABBREVIATIONS AC—alternating current AR—antireflective coating BPD—bypass diode BSF—back surface field CAD—computer-assisted drawing CZ—Czochralski process D4—Day4 Energy DC—direct current FF—fill factor Gen—generation HIT—heterojunction with intrinsic thin layer I—current IBC—interdigitated back contact Impp—current at maximum power point

REFERENCES

135

IR—infrared Isc—short-circuit current ISE—Institute for Solar Energy max—maximum mc—multicrystalline n—negatively or donor-doped semiconductor NREL—National Renewable Energy Laboratory p—positively or acceptor-doped semiconductor PERC—passivated emitter and rear cell PERL—passivated emitter, rear locally diffused Pmmp—power at maximum power point PV—photovoltaic R—resistance Rbl—bulk resistance Rem—emitter resistance ROI—return on investment Rs—series resistance Rsh—shunt resistance SiC—silicon carbide SiNx—silicon nitride antireflective coating SiO2—silicon dioxide TUV—Technischer Überwachungs-Verein (Technical Inspection Association) UL—Underwriters Laboratories V—voltage Vmmp—voltage at maximum power point Voc—open-circuit voltage Wp—watt peak η—efficiency 3D—three dimensional

REFERENCES [1]

[2] [3] [4] [5]

S. W. Glunz, A. Grohe, M. Hermle, M. Hofmann, S. Janz, T. Roth, O. Schultz, M. Vetter, I. Martin, R. Ferré, S. Bermejo, W. Wolke, W. Warta, R. Preu, and G. Willeke. Comparison of different dielectric passivation layers for application in industrially feasible high-efficiency crystalline solar cells. Presented at the 20th European Solar Conference and Exhibition, June 6–10, 2005, Barcelona (2005). S. W. Glunz. High-efficiency crystalline silicon solar cells. Article ID 97370. Review article. In Advances in Optoelectronics, doi:10.1155/2007/97370 (2007). A. Brenner and E. Riddell. Nickel plating on steel by chemical reduction. Journal of Research of the National Bureau of Standards, 37, 31–34 (1946). A. Brenner. Electroless plating comes of age. Metal Finishing 37, 61–68 (1954). M. V. Sullivan and J. H. Eigler. Electroless nickel plating for making ohmic contacts. Journal of the Electrochemical Society 104, 226–230 (1957).

136 [6] [7] [8]

[9] [10] [11]

[12]

[13] [14] [15] [16] [17] [18] [19]

[20]

[21]

[22]

CRYSTALLINE SILICON SOLAR CELLS AND MODULES L. F. Durkee. Method of plating by means of light. U.S. Patent No. 4 144 139, Solarex Corporation, Rockville, MD (1979). L. A. Grenon. Electroplating method. U.S. Patent No. 4 251 327, Motorola, Schaumburg, IL (1981). S. W. Glunz, J. Knobloch, D. Biro, and W. Wettling. Optimized high-efficiency silicon solar cell with Jsc = 42 mA/cm2 and η = 23.3%. In Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 392–395 (1997). K. F. Teng and R. W. Vest. Application of ink jet technology on photovoltaic metallization. IEEE Electron Device Letters 9, 591–593 (1988). T. Rivkin, C. Curtis, A. Miedaner, J. Perkins, J. Alleman, and D. Ginley. Direct write processing of photovoltaic cells. In Proceedings of the 29th IEEE Photovoltaic Specialists Conference, New Orleans, LA, pp. 1326–1329 (2002). M. F. Kaydanova, A. M. van Herst, A. Miedaner, C. Curtis, J. Alleman, M. S. Dabney, E. Garnett, S. Shaheen, L. Smith, R. Collins, J. I. Hanoka, A. M. Gabor, and D. S. Ginley. Direct write contacts for solar cell. In Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp.1305–1308 (2005). M. Hoerteis, P. L. Richter, and S. W. Glunz. Improved front side metallization by aerosol jet printing of hotmelt inks. Presented on the 23rd European Photovoltaic Solar Energy Conference and Exhibition, September 1–5, 2008, Valencia, Spain (2008). H. J. Gould. Method of manufacturing a back contact for semiconductor die. U.S. Patent 5,451,544 (1993). R. Preu, E. Schneiderlöchner, S. Glunz, and R. Loedeman. Method of producing a semiconductor-metal contact through a dielectric layer. U.S. Patent 6,982,218 (2001). M. A. Green and S. R. Wenham. Australian Patent No. 5703309, Australia (1984). M. Green, J. Zhao, A. Wang, and S. R. Wenham. Very high efficient silicon solar cell- science and technology. IEEE Transactions on Electron Devices 46(10), 1940– 1947 (1999). R. Swanson. Method of fabricating back surface point contact solar cells. U.S. Patent No. 4,927,770 (1990). L. Rubin and G. Rubin. Electrode for photovoltaic cells, photovoltaic cell and photovoltaic module. EP Patent 1 547 158 B1, 2002, August 29 (2007). A. Schneider, L. Rubin, A. Osipov, A. Smirnov, and P. Antipov. A new approach in solar cell module interconnection technique resulting in 5-10% higher PV module power output. Presented at the IEEE 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, May 8–12 (2006). A. Schneider, L. Rubin, and G. Rubin. Solar cell efficiency improvement by new metallization techniques—The Day4™ electrode concept. Presented at the IEEE 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, May 8–12 (2006). A. Schneider, L. Rubin, and G. Rubin. The Day4 electrode—A new metallization approach towards higher solar cell and module efficiencies. Presented at 21st European Photovoltaic Solar Energy Conference, September 4–8, 2006, Dresden, Germany (2006). L. Rubin and V. Nebusov. Photovoltaic module with edge access to PV strings, interconnection method, apparatus, and system. PCT/CA/2007/002301 (2007).

6 THIN-FILM SOLAR CELLS AND MODULES ROBERT BIRKMIRE Institute of Energy Conversion, University of Delaware

6.1

INTRODUCTION

Thin-film solar cell technologies originated in the 1960s with Cu2S/CdS, which was the first (flexible) thin-film solar cell and the first to achieve 10% efficiency in 1981 [1]. From 1970 through the 1980s, CuInSe2 [2], CdTe [3], and a-Si [4] became the solar cell materials of interest with all three achieving ∼10% efficiency. The mantra through this period was thin films will be in large-scale production and will cost ∼$1 per watt next year, which continued through the 1990s. In the 1990s, however, performance levels of the thin-film solar cells increased with CuInGaSe2 >19% [5], CdTe >16% [6], and a-Si >10% (stabilized) [7]. This, coupled with the expansion of c-Si technologies in the mid-1990s, has led to significant efforts to commercialize thin film technologies. At the turn of the century, PV began its rapid expansion with production increasing about 35%/year with most of the capacity in the c-Si module production. In 2008, worldwide solar cell production was ∼6.85 GW up from 3.44 GW in 2007 [8]. In the past few years, thin film manufacturing has undergone rapid expansion lead primarily by a-Si and CdTe technologies, and in 2008, thin film production was ∼0.89 GW up about 120% [8]. The emergence of the thin film technologies has been, in part, driven by the shortage of Si feedstock, which limited new production of c-Si and mc-Si modules. First Solar emerged as the largest thin-film PV manufacturer in the world and has expanded its manufacturing capacity of CdTe to 1 GW in early 2009. It is the first thin film company to demonstrate the “promise of thin-film PV” that high-throughput processing and large-scale facilities will significantly reduce the cost of PV modules and has achieved the lowest manufacturing cost per watt in the industry of less than $1 per Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

137

138

THIN-FILM SOLAR CELLS AND MODULES

TABLE 6.1. Best Verified Solar Cell Efficiencies for Thin-Film Solar Cells along with c-Si for Comparison under Standard Test Conditions Classification

Eff. (%)

Area (cm2)

VOC (V)

JSC (mA/cm2)

FF (%)

Test Center (Reference)

CdTe

16.5

1.03

0.845

25.9

75.5

NREL [6]

CuInGaSe2 CuInGaSe2

19.4

0.99

0.716

33.7

80.3

NREL [5]

20.0

0.419

0.692

35.7

81.0

NREL [5]

a-Si

9.5

1.07

0.859

17.5

63.0

NREL [7]

nc-Si

10.1

1.2

0.539

24.4

76.6

JQA [10]

a-Si/a-Si/a-SiGe

12.1

0.27

2.297

69.7

NREL [11]

Si (crystalline)

25.0

4.0

0.705

42.7

82.8

Sandia [12]

Si (multicrystalline)

20.3

1.002

0.664

37.7

78.9

NREL [13]

7.56

For a complete listing of the best cell efficiencies, see Green et al. [14].

watt [9]. a-Si manufacturing is rapidly increasing with Applied Materials, Oerlikon Corp, and Sharp leading the way. The production of CuInSe2-based modules has also increased substantially through efforts of Global Solar and Wurth Solar along with the entrance of Showa Shell and Honda into the market. This has spurred an infusion of venture capital for a large number of start-up companies in the United States and throughout the world in CuInSe2-, CdTe- and a-Si-based technologies as well as sparked the interest of several larger companies. The market share of thin-film PV modules is expected to increase substantially in the coming years. The performance of commercially available thin-film modules is less than crystalline Si (c-Si) modules that typically have efficiencies ranging from 12% to 19%. CuInSe2-based modules have the highest efficiency of thin-film modules in the range of 9–12% with CdTe modules from 9% to 11% and a-Si typically in the 5–8% range. The thin film technologies are at a similar development stage as c-Si was in the 1980s, and the next challenge is to increase module performance at the manufacturing level. A summary of the best cell and module performance measured at certified test laboratories is given in Tables 6.1 and 6.2 for a-Si-, CdTe-, and CuInSe2-based devices along with both single-crystal and multicrystal Si as reference points. It is interesting to note that both a-Si and CdTe, the most commercially advanced technology, have not recently increased the best performance value, mostly likely since the focus has been on increasing manufacturing capacity. In Figure 6.1, the best efficiency thin-film solar cell device is compared to the maximum obtainable single junction efficiency along with the best cell and module efficiency for single-crystal Si as a reference point. Single-crystalline Si is within about 85% of the efficiency limit, while the best CuInGaSe2 cells are ∼70% of the efficiency limit and can be directly compared

INTRODUCTION

139

TABLE 6.2. Best Verified Module Efficiencies for Thin Film along with c-Si for Comparison under Standard Test Conditions Classification

Eff. (%)

Area (cm2)

VOC (V)

ISC (A)

FF (%)

Test Center (Reference)

CdTe

10.9

4874

26.21

3.24

62.3

NREL [15]

31.2

2.18

68.9

NREL [16]

3.285

66.0

NREL [17]

CuInGa(Se,S)2

13.5

3459

a-Si/a-SiGe/a-SiGe

10.4

905

Si (large area)

20.1

16,300

66.1

6.35

78.7

Sandia [18]

Si (multicrystalline)

15.5

1017

14.6

1.37

78.6

Sandia [19]

4.353

For a complete listing of best module efficiencies, see Green et al. [14].

35 30

Cell Module Black-body limit

Efficiency (%)

c-Si 25 20

CuInGaSe2

AM0 CdTe

15 AM1.5 a-Si

10 5 0.5

1.0

1.5 Bandgap (eV)

2.0

2.5

Figure 6.1. Comparison of maximum achievable efficiency for single-junction solar cells to best-reported solar cell for different technologies. The best module performance is also included.

to the best multicrystalline Si cells since they have about the same bandgap and are polycrystalline in nature (not shown on graph). CdTe is about 55% of the efficiency limit, which indicates that advances in the laboratory are needed for the technology to reach its full potential. It is difficult to evaluate a-Si PV technology on this basis since the best devices are multijunction structures. The differential between the best cell and module can be used as a crude measure of the readiness of the technology for manufacturing. The efficiency of the CdTe and CuInGaSe2 modules is about 50–60% of the best cell, while multijunction a-Si and c-Si modules are about 80% of the best cell and mc-Si is 75%.

140

THIN-FILM SOLAR CELLS AND MODULES

The selection of PV modules will be based on annualized power output and levelized cost of energy for the specific location of the array and type of application. Figure 6.2 compares the annualized output for c-Si, mc-Si, a-Si, and CdTe at different geographic locations in the United States for residential rooftop arrays mounted at a fixed tilt for the local latitude. The graph was generated from the SAM [20], developed by DOE/NREL. Table 6.3 lists the module manufacturers and efficiencies used from SAM. The array size is ∼2.5 kW for all technologies and the SAM database is used for the modules. It is important to note that the lower-efficiency a-Si (5.7%) and CdTe (7.7%) arrays perform nearly as well as c-Si (19.3%) and better than mc-Si (13.3%) as the result of a better temperature coefficient. However, there is a large variation in temperature coefficient data for PV modules, and the effects of illu-

1500

1000

500

CdTe

0 AZ

mc-Si HI

MT Location

c-Si PA

pe

a-Si

Ty

Annual Output (kWh/kW)

2000

WA

Figure 6.2. Comparison of the annualized power output for deployment on a ∼2.5-kW roof residential rooftop array at different geographic locations for thin-film and Si modules. TABLE 6.3. Modules Used to Generate the Annualized Power from the Solar Advisor Model Database Module

Type

Eff. (%)

SunPower

c-Si

19.3

Kyocera

mc-Si

13.3

Uni-Solar

a-Si

5.7

First Solar

CdTe

7.7

THIN-FILM SOLAR CELLS AND MODULES

141

mination intensity on the module performance are not taken into account in the calculations. Thus, an accurate comparison requires a more detailed model that is validated by real data. Also, the lower-efficiency arrays require significantly more area, which increases the balance of system cost. The “figure of merit” will thus be the levelized cost of energy.

6.2

THIN-FILM SOLAR CELLS AND MODULES

The motivation for thin-film solar cells was, and still is, the potential for highspeed/high-throughput manufacturing and minimum material requirements to reduce cost. To meet this goal, large-scale manufacturing facilities are needed to obtain economies of scale, but until recently, the large up-front investment required inhibited the full-scale commercialization of thin film technologies. Thin-film modules all have several common components: (1) substrate; (2) a TCO and/or grid; (3) metal contacts; and in many cases, (4) a monolithic integration scheme. The substrate is either glass or a flexible web that is either plastic or metal foil, and the device configuration is either a superstrate or substrate design as shown in Figure 6.3. The superstrate is always glass for the thin film technologies and is coated with a TCO such as ITO, SnO2, or ZnO. Both CuInSe2- and a-Si-based modules are currently commercially being manufactured using glass and/or flexible substrates, while CdTe is always on a glass substrate. (There are several start-up companies exploring CdTe on a metal substrate.) For the substrate device configuration, glass is used for CuInSe2-based modules where Mo is used as the back contact, and flexible metal foils or polyimide substrates (web) are used for both a-Si- and CuInSe2-based modules. The attraction of the flexible substrate is the use

Figure 6.3. The two basic cell configurations for thin-film solar cells.

142

THIN-FILM SOLAR CELLS AND MODULES

of continuous roll-to-roll processing for manufacturing and was first proposed and demonstrated for Cu2S solar cells on a flexible Cu web [21]. Energy Conversion Devices and Iowa Thin Film Technologies adapted this for a-Si in the 1980s. In the 1990s, CuInSe2-based solar cells were developed on a polyimide and stainless substrate, and recently, several start-up companies are pursuing this approach on both polyimide and metal webs. A major advantage of thin film technologies is that large glass substrates or continuous webs can be coated as compared to c-Si, where individual wafer cells are fabricated and physically connected into modules using a “tab and stringing” process. For the glass and insulating webs, monolithic integration schemes are used for both superstrate and substrate modules to series connect individual cell segments, eliminating the need to assemble individual cells into the modules. Figure 6.4 illustrates the monolithic integration structure where the details of the processing steps depend on the device configuration and substrate/web. The cell segments and interconnections are defined by a series of laser and/ or mechanical scribes at different processing stages for glass substrates and are typically done as the last processing step for flexible insulating webs. For conductive webs used for both a-Si and CuInSe2, individual cells are prepared by cutting the web and applying a metal grid structure to provide the front contact to the cell and then are connected in series in a manner similar to assembling a c-Si module.

6.3

POLYCRYSTALLINE THIN FILM

Thin-film CdTe and CuInSe2 and related alloys are polycrystalline in nature with grain sizes of approximately 1–5 μm. Both CdTe- and CuInSe2-based solar cells

Figure 6.4. Typical monolithic integration scheme for thin-film modules.

POLYCRYSTALLINE THIN FILM

143

are heterojunction devices where CdTe is configured in a superstrate structure and CuInSe2 in a substrate structure. Both devices operate in a similar manner and have nearly 100% internal quantum efficiency and thus are generally not limited by Jsc. Approaches to improve device performance have primarily focused on Voc and FF, but the approach is different for each materials system. The primary mechanism limiting Voc, in both devices, is generally recombination in the space charge region, which is consistent with Shockley–Read–Hall recombination at defects within the bandgap [22, 23]. 6.3.1

CdTe Solar Cells and Modules

CdTe solar cell technology is leading the way in the development of thin-film PV in terms of cost and high-speed/high-throughput manufacturing. CdTe has a bandgap of 1.45 eV, which is a near perfect match to the AM 1.5 terrestrial spectrum for maximum solar cell efficiency (see Fig. 6.1). It is a direct bandgap semiconductor with an absorption coefficient >105 cm−1 at 700 nm and therefore only requires about a 1-μm thick film to absorb most of the incident light. All highefficiency CdTe solar cells have been made using a superstrate structure where the final processing steps, used to control the electronic properties and formation of a back ohmic contact, are carried out on a free CdTe surface. Efforts to use a substrate configuration where the CdTe is deposited on a conductive opaque substrate have not been successful, and thus CdTe solar cells are not compatible with a flexible substrate. Small area laboratory devices fabricated on a borosilicate glass substrate have achieved an efficiency of 1.3 eV, independent of the alloying materials, are less suitable for solar cells [41]. All high-efficiency CuInSe2-based devices are made using a substrate device structure where the substrate can be glass or a flexible plastic or metal foil. Efforts to use a superstrate configuration where the CuInSe2-based material is deposited on a transparent conducting substrate have not been successful [42]. The highest-efficiency solar cell, 20%, was made using a CuInGaSe2 film deposited by multisource evaporation [5], and the highest-efficiency module, 13.5%, was made using a CuInGa(Se,S)2 film grown by the reaction process [13]. CuInSe2-based solar cells are p/n heterojunction devices where a wide bandgap, n-type window layer such as CdS is used as the heterojunction partner with the p-type CuInSe2 base material. A schematic of the device structure is shown in Figure 6.7. Mo is used for the back ohmic contact and is deposited by sputtering. CuInSe2based thin films are generally deposited by the multisource elemental thermal evaporation or reaction of a precursor such as sputtered metals in a Se and/or S atmosphere. The remarkable property of CuInSe2 materials is the tolerance to

TABLE 6.4. CuInSe2 Alloy Materials Material

Eg (eV)

CuInSe2

1.0

CuGaSe2

1.68

CuAlSe2

2.72

CuInS2

1.53

CuGaS2

2.53

CuAlS2

3.5

CuInTe2

1.1

CuGaTe2

1.23

POLYCRYSTALLINE THIN FILM

147

Figure 6.7. Standard CuInSe2-based solar cell structure highlighting the individual layer in the device.

composition variations. The Cu concentration in the film can vary by 2–3% without affecting the device performance provided that the Cu-to-group III ratio is less than one. This is attributed to the broad single-phase region of CuInSe2 at the growth temperatures and may explain why, when forming a wider bandgap alloy, materials are less tolerant to composition. CuInSe2 may be the only I-III-V chalcopyrite material that has the broad single-phase regime. Another critical issue in the growth of CuInSe2-based materials is the incorporation of Na into the film that modifies the growth habit of the film and appears to reside at the grain boundaries and improves the properties of the solar cell device [43]. The soda lime glass substrate provides a source of Na, which diffuses from the glass thru the Mo contact into the growing CuInSe2-based film. However, for manufacturing, a diffusion barrier is, generally, used on the glass and the Na is controllably incorporated into the film. Two primary approaches have been used to deposit CuInSe2-based materials. Multisource evaporation has been used to deposit the CuInSe2-based thin films used for the highest-efficiency solar cells. Figure 6.8 is a schematic of the deposition system. The incident flux and substrate temperature are varied during the film growth to control the through-film composition, grain structure, and bandgap. Typically, for a CuInGaSe2 film, the through-film composition of Ga and In is controlled by varying the incident flux during film growth, creating a gradient in the bandgap through the film that depends on the Ga to (Ga + In) ratio. Variation in the Cu flux during growth can affect the structure of the film, but the Cu diffuses rapidly through the film, resulting in a uniform distribution of Cu. The best laboratory solar cell was fabricated using a three-stage process where the initial growth is just GaIn-Se followed by Cu-Se and then Ga-In-Se [5]. However, solar cells with efficiencies around 17% have been grown by simpler uniform growth and two-stage processes, which are more compatible with manufacturing [44].

148

THIN-FILM SOLAR CELLS AND MODULES Substrate

Se sparger tubes

θ Se Cu Se Ga Se

In

Se

Z Cu, Ga, In sources

Y X

Figure 6.8. Schematic of an elemental source thermal evaporation in-line system for depositing CuInGaSe2 films on a flexible web.

Figure 6.9. System for reacting precursor materials to form CuInGaSe2 films.

The second approach to growing CuInSe2-based materials is the reaction precursor films in a Se2 and/or S2 atmosphere, generally H2Se and/or H2S, and is referred to as selenization or surfurization. Figure 6.9 is a schematic of the reaction system. This approach was initially commercialized by ARCO Solar, Siemens Solar Industries, and Shell Solar Industries [45] and is the focus of many companies since the reaction process is simpler to translate to manufacturing. The simplest example of the reaction process is to sputter deposit Cu-In-Ga layers and to react these precursors in H2Se gas at atmospheric pressure. Typically, the reaction temperature is limited to less than 500°C since the H2Se will react with the Mo contact at higher temperature. Additionally, as a result of the reaction kinetics, the Ga resides near the back of the film, and thus the benefits of incorporating Ga to widen the bandgap and to increase Voc are not realized. Sulfur has been incorporated at the front surface where the junction is formed to increase the Voc, and a small area

POLYCRYSTALLINE THIN FILM

149

device of about 16% has been reported by Siemens [46]. Several companies are developing precursor processes that eliminate the metal sputtering process based on inks consisting of nanoparticles containing Cu-In-Ga-Se or electrodeposited metals. The n-type window layer for many CuInSe2-based devices is CdS and has been used for the highest-efficiency laboratory cells as well in commercial modules. The CdS is deposited by a wet chemical process, referred to as chemical bath deposition (CBD), and conformal CdS films can be deposited with a thickness of about 50 nm to minimize optical losses in the CdS layer. There has been a significant effort to replace the CdS with a non-Cd-containing window layer to make the modules more environmentally friendly. Some of the materials that have been evaluated are Zn(S,O,OH) [47], ZnSe [48], ZnIn2Se4 [49], In2S3 [50], and Inx(OH,S)y [51], which were deposited by a variety of techniques such as CBD, sputtering, MOCVD, and ALE. However, the device performance is, in general, not as good as devices using CdS. As with CdTe devices, a high-resistive TCO is deposited on the CdS, which is usually ZnO and has a resistivity from 1 to 100 Ω-cm with a thickness of ∼50 nm. Since there is a significant probability of pinholes in CdS, the highresistive TCO eliminates the possibility the conductive TCO from contacting the CuInSe2-based absorber, which can result in secondary diodes. Finally, the conductive TCO is deposited, which is usually doped ZnO or ITO having a resistivity of 10−3 to10−4 Ω-cm. Manufacturing of CuInSe2-based modules is at the early stages of commercialization and is being led by Showa Shell, Honda, Wurth Solar, and Global Solar, which have production capacity in excess of 15 MW along with numerous start-up companies throughout the world. Both multisource evaporation and reaction of precursors in a Se2 and/or S2 atmosphere are being used for the deposition of the CuInSe2-based absorber on glass, polymer, and metal foil where roll-to-roll processing is being developed using polymer and metal foil webs. Module efficiencies of close to 12% have been reported, and cell efficiencies on the steel substrate are about 10%. Recently, Ascent Solar reported monolithic integrated modules with efficiencies greater than 9.5%, verified by NREL using a roll-to-roll processing on a polyimide web. CuInSe2-based modules have the potential to achieve 15% at the commercial level based on laboratory results but lack fundamental knowledge to control properties, and the process has limited progress. The challenge with elemental multisource evaporation is the high source temperatures, required to obtain high film growth rates; typically, the Cu source operates at ∼1600°C, coupled with a Se environment [52]. Design and control of the system, particularly when depositing on a continuous web, requires in situ process control and diagnostics [53], which are being developed by several companies. For the reaction of precursors, the reaction time to form the CuInSe2-based films is limited by the reaction/diffusion rates. For a flexible web, the challenge is to react the precursors on a moving web where uniform delivery of the gas and reaction time become important issues. For modules fabricated on a glass substrate, a batch process is used where a large

150

THIN-FILM SOLAR CELLS AND MODULES

number of plates are reacted in a furnace. Developing a robust high-throughput manufacturing process needs to be demonstrated along with low cost comparable to that demonstrated for CdTe modules to accelerate the CuInSe2-based PV technologies.

6.4

HYDROGENATED a-Si, a-Si ALLOYS, AND nc-Si

Thin-film a-Si solar cells/modules have been commercially available starting with SANYO’s first commercial production facility in 1980. Today, a-Si-based modules are available from a number of companies and typically have efficiencies from 5 to 8%. The challenge for a-Si-based materials has been to improve the stabilized efficiency, which is less than their initial efficiency due to light-induced degradation, the Staebler–Wronski effect [54], which is a fundamental property of the an a-Si material. The basic device is a p-i-n structure where the n- and p-layers are 10–20 nm thick and the intrinsic i-layer is from 100 to 200 nm, depending on the device design. Engineering approaches to minimize the light-induced degradation have focused on multijunction device structures that minimize the thickness of the intrinsic layer in order to reduce the light-induced degradation. a-Si can be characterized as a direct bandgap material; however, the conduction and valence band edges are not well defined as in crystalline material as a result of band tail states. To compare the bandgap of different a-Si materials, the optical absorption coefficient as a function of energy is analyzed according to Tauc, where an optical or Tauc bandgap is rigorously defined [55]. The optical bandgap for a-Si varies from 1.7 to 1.8 eV depending on the H2 concentration in the film and has an absorption coefficient >105 m−1 at 550 nm. The bandgap can be modified from 1 to >2 eV by alloying with Ge to narrow the bandgap (a-SiGex) or with C (a-SiCx), O2 (a-SiOx), or N2 (a-SiNx) to widen the bandgap. The a-Si can be doped p-type with boron and n-type with phosphine and is used as a thin-film transistor in the display industry. The amorphous nature of the film can be transitioned to a nanocrystalline structure by controlling growth conditions and has been exploited for p-type window layers, μc-Si, to enhance optical transmission and for i-layers, nc-Si, to increase the absorption above 850 nm for use as a bottom cell in a tandem device structure [56]. The design of thin-film a-Si-based solar cells/modules is the most flexible of any of the thin film technologies. Both superstrate and substrate devices have been made on glass, foil, and plastic substrates, all based on a p-i-n structure as shown in Figure 6.10. To achieve the best performance, the device is illuminated through the p-layer, but the order of the deposition depends on the specific device design. The n-layer is deposited first for the substrate configuration, and the p-layer is deposited first for a superstrate structure. The contact to the n-layer is generally ZnO with a metal reflector, independent of the device structure, to facilitate light trapping in the device. The contact to the p-layer is a TCO where both SnO2 and ZnO have been used for modules.

HYDROGENATED a-Si, a-Si ALLOYS, AND nc-Si

151

Figure 6.10. Device structures used for a-Si-based solar cells.

Most a-Si, μc-Si, and nc-Si films used for commercial modules are grown by a PECVD process using RF (13.56 MHz) or VHF (40–100 MHz) excitation where SiH4 is the primary precursor gas and the substrate temperature is between 150 and 250°C. The growth rate for “device quality” a-Si films is from 1 to 5 A/s for the RF PECVD and is over 20 A/s for VHF PECVD [57]. However, VHF PECVD has the advantage of improved stabilized performance at the higher growth rates compared to RF PECVD, but designing the VHF electrode for uniform deposition over large areas is more challenging. Other growth processes have been used including microwave PECVD, photo- and catalytic (hot wire) CVD, and sputtering. Performance, especially stabilized, decreases as the growth rate increases regardless of the deposition method. The utilization of the SiH4 is, generally, less than 10% depending on the design equipment. Figure 6.11 shows a schematic of an RF–PECVD system. The intrinsic a-Si layer is grown using a gas mixture, SiH4 and H2, typically referred to as hydrogen dilution (ratio SiH4/H2) to control the bandgap and crystalline fraction in the film. Device quality nc-Si films have a crystalline fraction that is typically around 60% and consists of 10- to 20-nm crystallites that are encased by an a-Si:H “tissue.” The optical properties of the nc-Si:H are a mix of a-Si:H and c-Si, and thus the thickness of the intrinsic nc-Si:H layer needs to be about 1–2 μm to absorb a significant fraction of the long wavelength light. The conductivity of the a-Si is controlled by adding a dopant gas to the SiH4/H2 gas mixture, either diborane (B2H6), borntrifloride (BF3), or trimethylboron (B(CH3)3), for doped p-typing or phosphine (PH3) for n-type doping. The structure of a p-type a-Si window layer, which does not contribute to the current in the cell due to poor

152

THIN-FILM SOLAR CELLS AND MODULES Gas inlet

Substrate RF Electrode Substrate Transport RF

Plasma

Pump

Burn box

Figure 6.11. RF–PECVD system for growing a-Si films.

carrier properties, is grown using high H2 dilution to grow a μc-Si film that improves the short wavelength transmission and has a higher crystalline fraction than the nc-Si i-layer. Commercialization of a-Si-based module technology is undergoing a rapid expansion with several companies selling “turnkey” systems for single and multijunction modules based on RF or VHF PECVD. Module dimensions are increasing with Applied Materials supplying equipment to manufacture a 2.5 × 2.3 m module on a glass substrate. The Prometheus Institute projections for the thin film manufacturing capacity in 2012 is 9.6 GW with a-Si leading the way with a capacity of ∼5.2 GW. These projections are very conservative since 2008 thin-film capacity has already exceeded the projections in the report [58]. a-Si modules face several challenges to remain competitive with other thin film technologies and with c-Si wafer-based modules. The module efficiencies are low, and it is difficult to project efficiencies beyond 10%. However, a-Si-based modules have the lowest temperature coefficient of any PV technology, and thus the annualized power output can be comparable to or higher than modules with higher-efficiency ratings (see Fig. 6.2). For example, the annualized power output of a-Si modules exceeds that of c-Si by 5–15% in a large number of studies [59]. Two critical issues to improve the viability of a-Si technology are (1) improvement in stabilized performance, which is being addressed by going to a nc-S/a-Si tandem structure and (2) increased growth rate to reduce capital costs, which is particularly important for the nc-Si bottom cell where the i-layer is 1.5–2.0 μm. However, there appears to be a trade-off of performance and processing speed. a-Si technology also has several advantages: (1) flexibility in device design and structure; (2) pro-

ABBREVIATIONS

153

cessing is done at low temperature, less than 200°C; (3) material availability is not an issue even though SiH4 utilization is low; and (4) the flat panel display industry underpins the technology and scientific base.

6.5

OUTLOOK FOR THIN-FILM MODULES

The production of thin-film PV modules will continue to expand and probably, the manufacturing capacity will exceed c-Si in the next 5–10 years, primarily replacing mc-Si module market share. CuInSe2-based manufacturing challenges will be overcome and modules will reach or exceed 15% on glass and possibly flexible substrates utilizing high-speed processing technology to drive the manufacturing cost well below $1 per watt. New companies will enter the CdTe module market and module efficiencies will approach 13% with manufacturing cost less than $0.65 per watt. There will be a competition between the polycrystalline and a-Si technologies, which will be driven by the application and will levelize cost of power. As the production capacity expands, material availability of In and Te may become a critical issue, which would benefit the a-Si technology and be a driver for the organic PV technologies. The future of PV will consist of a mix of c-Si and thin film technologies where cost, application, location, and aesthetics will determine the choice of the PV technology. Thin film technologies will be the modules of choice for residential roof application and commercial build integrated applications. Medium-scale power- and utility-scale PV arrays will use both polycrystalline and high-efficiency c-Si modules. However, in hot arid locations, very high-efficiency concentrator PV arrays will also be used for utility-scale applications.

ABBREVIATIONS ALE—atomic layer epitaxy a-Si—amorphous silicon C—carbon CBD—chemical bath deposition Cd—cadmium CdS—cadmium sulfide Cd2SnO4—cadmium stannate CdTe—cadmium telluride Cl—chlorine c-Si—crystalline silicon CSS—close-spaced sublimation Cu—copper CuInGaSe2—copper indium gallium diselenide Cu2S—copper sulfide CuInSe2—copper indium diselenide

154

THIN-FILM SOLAR CELLS AND MODULES

DOE—Department of Energy Eff.—efficiency FF—fill factor Ga—gallium Ge—germanium H—hydrogen Hg—mercury I—undoped or intrinsic semiconductor III—elements in column 3 of periodic table In—indium ITO—In2O3:Sn Jsc—short-circuit current MBE—molecular beam epitaxy mc-Si—multicrystalline silicon Mo—molybdenum MOCVD—metal-organic chemical vapor deposition N—nitrogen n—negatively doped semiconductor Na—sodium nc-Si—nanocrystalline silicon NREL—National Renewable Energy Laboratory O—oxygen P—phosphorous p—positively doped semiconductor Pd—palladium PECVD—plasma-enhanced chemical vapor deposition PV—photovoltaics or solar cells PVD—physical vapor deposition RF—radio frequency S—sulfur SAM—Solar Advisor Model Se—selenium Si—silicon SiH4—silane Sn—tin SnO2—tin oxide TCO—transparent contact oxide Te—tellurium V—elements in column 5 of periodic table VHF—very high frequency Voc—open-circuit voltage VT—vapor transport Zn—zinc ZnO—zinc oxide μc-Si—microcrystalline silicon

REFERENCES

155

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24]

R. B. Hall, R. W. Birkmire, J. E. Phillips, and J. D. Meakin. Appl. Phys. Lett. 38, 925 (1981). R. A. Mickelsen and W. S. Chen. In Conf. Rec. 15th IEEE Photovoltaic Specialist Conf., pp. 800–803 (1981). Y.-S. Tyan and E. A. Perez-Albuerne. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 794–800 (1982). A. Catalano, R. V. D’Aiello, J. Dresner, B. Faughnan, A. Firester, J. Kane, H. Schade, Z. E. Smith, G. Swartz, and A. Triano. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 1421–1422 (1982). I. Repins, M. A. Contreras, B. Egaas, C. DeHart, J. Scharf, C. L. Perkins, B. To, and R. Noufi. Prog. Photovoltaics 16(3), 235–239 (2008). X. Wu, J. C. Keane, R. G. Dhere, C. DeHart, A. Duda, T. A. Gessert, S. Asher, D. H. Levi, and P. Sheldon. In Proceedings of the 17th European Photovoltaic Solar Energy Conference, October 22–26, 2001, Munich, pp. 995–1000 (2001). J. Meier, J. Sitznagel, U. Kroll, C. Bucher, S. Fay, T. Moriarty, and A. Shah. Thin Solid Films 451–452, 518–524 (2004). SolarBuzz Annual World Solar PV Industry Report. Available at http://www.solarbuzz.com/Marketbuzz2009-intro.htm (2009). First Solar Quarterly Report for 2009: www.firstsolar.com. K. Yamamoto, A. Nakajima, Y. Tawada, M. Yoshimi, Y. Okamoto, and S. Igari. In Proc. of the 2nd World Conference on Photovoltaic Energy Conversion, pp. 1284– 1290 (1998). J. Yang, A. Banerjee, S. Sugiyama, and S. Guha. In Conf. Rec. 26th IEEE Photovoltaic Specialist Conf., pp. 563–568 (1997). J. Zhao, A. Wang, M. A. Green, and F. Ferrazza. Appl. Phys. Lett. 73, 1991–1993 (1998). O. Schultz, S. W. Glunz, and G. P. Willeke. Prog. Photovoltaics 12, 553–558 (2004). A. M. Green, K. Emery, Y. Hishikawa, and W. Warta. Prog. Photovoltaics 17, 85–94 (2009). D. Cunningham, K. Davies, L. Grammond, E. Mopas, N. O’Connor, M. Rubcich, M. Sadeghi, D. Skinner, and T. Trumbly. In Conf. Rec. 28th IEEE Photovoltaic Specialist Conf., pp. 13–18 (2000). Y. Tanaka, N. Akema, T. Morishita, D. Okumura, and K. Kushiya. In Proc. 17th European Photovoltaic Solar Energy Conference, pp. 989–994 (2001). J. Yang, A. Banerjee, T. Glatfelter, K. Hoffman, X. Xu, and S. Guha. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 380–386 (1994). D. Rose, O. Koehler, N. Kaminar, B. Mulligan, and D. King. In Proc. 4th World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 2018– 2023 (2006). D. L. King, W. K. Schubert, and T. D. Hund. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 1660–1662 (1994). Available at http://www.nrel.gov/analysis/sam. A. M. Barnett, B. N. Baron, J. V. Masi, and T. W. F. Russell. Method for the continuous manufacture of thin film solar cells. U.S. Patent 4,318,938, March 9, 1982. S. Hegedus and B. McCandless. Sol. Energy Mater. Sol. Cells 88, 75–95 (2005). S. Hegedus and W Shafarman. Prog. Photovoltaics 12, 155 (2004). J. Jordan and S. Albright. Photovoltaic cell with thin CdS layer. U.S. Patent 5,279,678 (1994).

156

THIN-FILM SOLAR CELLS AND MODULES

[25]

T. Takamoto, T. Agui, H. Kurita, and M. Ohmori. Sol. Energy Mater. Sol. Cells 49, 219–225 (1997). B. McCandless and R. Birkmire. In Proceedings of the 28th IEEE PVSC, ed., pp. 491–494 (2000). B. McCandless. Unpublished work. X. Wu et al. In Conf. Rec. 28th IEEE Photovoltaic Specialist Conf., pp. 470–474 (2000). B. McCandless, L. Moulton, and R. Birkmire. Prog. Photovoltaics 5, 249–260 (1997). F. J. Kruger. Electrochem. Soc. 125, 2028–2034 (1978). Y. Tyan and E. Perez-Albuerne. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 794–800 (1982). R. Powell, G. Dorer, N. Reiter, H. McMaster, S. Cox, and T. Kahle. Apparatus and method for depositing a material on a substrate. U.S. Patent 5,945,163 (1999). B. McCandless, I. Youm, and R. Birkmire. Prog. Photovoltaics 7, 21–30 (1999). R. Wendt, A. Fischer, D. Grecu, and A. J. Compaan. Appl. Phys. 84, 2920–2925 (1998). R. Sudharsanan and A. Rohatgi. Sol. Cells 31, 143–150 (1991). J. Jordan. Photovoltaic cell and method. International Patent Application WO93/14524 (1993). S. Ikegami. In Technical Digest of the 3rd Photovoltaic Science and Engineering Conference, pp. 677–682 (1987). G. M. Hanket, B. E. McCandless, W. A. Buchanan, S. Fields, and R. W. Birkmire. J. Vac. Sci. Technol. A, 24(5), 1695 (2006). B. McCandless and J. Sites. In Handbook of Photovoltaic Science and Engineering, A. Luque and S. Hegedus, eds., pp. 633–644. John Wiley & Sons (2003). R. Birkmire. In Conf. Rec. 33rd IEEE Photovoltaic Specialist Conf., paper 370 (2008). W. Shafarman, R. Klenk, and B. McCandless. In Conf. Rec. 25th IEEE Photovoltaic Specialist Conf., pp. 763–768 (1996). T. Yoshida and R. W. Birkmire. In Proc. 11th European Photovoltaic Solar Energy Conference, pp. 811–816 (1992). M. Bodegard, K. Granath, A. Rockett, and L. Stolt. Sol. Energy Mater. Sol. Cells 58, 199–208 (1999). W. Shafarman and J. Zhu. Thin Solid Films 361–362, 473–477 (2000). R. Gay, M. Dietrich, C. Fredric, C. Jensen, and K. Knapp. In Proc. 12th European Photovoltaic Solar Energy Conference, pp. 935–940 (1994). D. Tarrant and J. Ermer. In Conf. Rec. 23rd IEEE Photovoltaic Specialist Conf., pp. 372–375 (1993). A. Ennaoui, U. Blieske, and M. Lux-Steiner. Prog. Photovoltaics 6, 447–451 (1998). Y. Othake, K. Kushiya, A. Yamada, and M. Konagai. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 218–221 (1994). Y. Othake, T. Okamoto, A. Yamada, M. Konagai, and K. Saito. Sol. Energy Mater. Sol. Cells 49, 269–275 (1997). J. Guillemoles, B. Canava, E. Yousfi, P. Cowache, A. Galtayries, T. Asikainen, M. Powalla, D. Hariskos, H. Schock, and D. Lincot. Jpn. J. Appl. Phys. 40, 6065–6068 (2001). D. Hariskos, R. Herberholz, M. Ruckh, U. Ruhle, R. Schaffler, and H. W. Schock. In Proc. 13th European Photovoltaic Solar Energy Conference, pp. 1995–1998 (1995). S. T. Junker, R. W. Birkmire, and F. J. Doyle III. AIChE J. 51(3), 878 (2005).

[26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52]

REFERENCES [53] [54] [55] [56] [57] [58]

[59]

157

K. Mukati, B. A. Ogunnaike, E. Eser, S. Fields, and R. Birkmire. Ind. Eng. Chem. Res. 48, 5975–5991 (2009). D. Staebler and C. Wronski. Appl. Phys. Lett. 31, 292 (1977). J. Tauc. In Optical Properties of Solids, F. Abeles, ed., pp. 277–313. Amsterdam, North-Holland (1972). A. Shah, J. Meier, P. Torres, U. Kroll, D. Fischer, and N. Beck. In Conf. Rec. 26th IEEE Photovoltaic Specialist Conf., pp. 569–574 (1997). H. Takatsuka, M. Noda, Y. Yonekura, Y. Takeuchi, and Y. Tamauchi. Solar Energy 77, 951–960 (2004). S. Grama and T. Bradford. Thin-Film PV 2.0: Market Outlook Through 2012: Executive Summary, the Prometheus Institute. Available at http://www.greentechmedia.com/GreentechMedia/Report/Thin-FilmPV20MarketOutlookThrough2012. html. S. Hegedus. Thin film solar modules: The low cost, high throughput and versatile alternative to Si wafers. Prog. Photovoltaics 14, 393 (2006).

7 TERRESTRIAL MODULE FABRICATION AND ASSEMBLY TECHNOLOGIES CHRISTOPHER BUNNER Spire Corporation

7.1

INTRODUCTION

The global desire to produce energy from various renewable energy options increases as political, environmental, and economic landscapes transform. As expected, the solar industry benefits from this renewed attention where governments recognize and support the need for renewable energy options through grants, low tariffs, and deferred taxes to diminish the consumption of fossil fuels. Unfortunately, environmental pressure remains tied to the cost of ownership. The regional desires for renewable solutions directly correlate with governmental financial incentives. Fortunately, the solar industry has attracted notice from complementing industries as their primary customers stagnate. This attention provides new ideas and initiatives that benefit the solar module manufacturing process. Overall, the goal of terrestrial module manufacturing remains the same: to consistently produce the least expensive and most reliable module. While the basic module construction and manufacturing processes have changed little, the methods and equipment available have matured. This chapter examines the available improvements to the module fabrication process through the evolution of materials and equipment.

7.2

MATERIALS

While the solar industry continues to grow and mature, the necessary materials to construct a silicon-based photovoltaic (PV) module continue to encompass over Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

159

160

TERRESTRIAL MODULE FABRICATION Materials, 0.413, 73.1%

Labor, 0.032, 5.6% Op. & Maint., 0.020, 3.5%

Depreciation, 0.043, 7.7%

OH & G&A, 0.057, 10.1%

Figure 7.1. A breakdown is shown for the module cost elements for 220-W crystalline silicon modules produced by an automated production line at 50 MW/year excluding the cell costs. The module cost elements, excluding the cell cost, in dollar per watt, account for a total cost = $0.565 per watt.

60% of the overall module costs. This driving cost creates opportunities for material improvements and cost reduction pressure. The increasing number of module manufacturers has simultaneously consumed available material resources and have challenged material suppliers to increase their capacities. Spire Solar for many years has been doing comprehensive and exhaustive cost analysis studies for module manufacturing costs. Figure 7.1 shows a breakdown for the module cost elements for 220-W crystalline silicon modules produced by an automated production line at 50 MW/year. The module cost elements, excluding the cell cost, in dollar per watt, account for a total cost = $0.565 per watt. Silicon prices directly affect the growth of the solar industry. The silicon wafer remains the highest-priced component within the module. Silicon manufacturers have been challenged over the past decade to predict the consumption needs of the volatile semiconductor market as well as the rapidly growing solar industry. For years, the solar industry has consumed any excess higher grade and the lower grades of available silicon feedstock not desirable to the semiconductor industry. However, as capacities and innovation grow, the solar industry’s appetite for highquality silicon increases. As expected, this increasing demand, coupled with stationary supplies, leads to increasing silicon feedstock prices and higher wafer, cell, and module costs. Increasing silicon production capacities require a large commitment of capital and equipment. Over the last several years, the success of the solar industry has convinced silicon manufacturers to increase capacities. This increased capacity is just beginning to yield lower costs. Cell prices have begun to trend downward. A major driver in this cost reduction, however, has been the fabrication of thinner silicon wafers into solar cells. Where 10 years ago the “standard” solar cell thickness was 250–300 μm, today’s solar cells typically are fabricated from 180- to 200-micron wafers.

BASIC MODULE ASSEMBLY

161

These thinner cells have presented special needs for the manufacturing process to achieve the required yields necessary to take advantage of the thinner material. These special requirements include low-impact cell handling, small stacks if cells are coin stacked, more stringent heating/cooling requirements, and a myriad of machine design features to maintain manufacturing process yield. Over the last decade, traditional silicon solar cell efficiency has increased from around 12% to 17%. While the individual PV cells become more efficient and evolve into larger formats, their versatility remains. Cells may be arranged in various series and parallel configurations to meet system requirements. Three years ago, a typical module produced 70 W through the interconnection of 36 125-mm cells in series. Today, the standard module design is 220 W using 60 156-mm cells in series. A larger cell size yields various benefits. First, by increasing cell area, cell power is increased. While each cell produces about 0.6 V, regardless of size, as surface area increases, cell current increases. Second, as the cell efficiency is improved, fewer cells are required to produce the same module power. Additionally, some cell manufacturers have taken additional paths to improving cell efficiency. Reducing or eliminating the top surface contacts by use of a rear-contact cell structure has gained more absorption area. Furthermore, the use of larger and more powerful cells reduces cell handling, increases solar cell production capacities, and reduces packaging and shipping costs. Cost reductions achieved in the production of cells directly relate to reduced module costs. While these wafers and cells will require additional handling and processing requirements, the efficiency improvements will likely benefit other solar applications. Cell testing and classification is an important process step for both the cell and module fabrication processes. While cell testing quantifies cell performance, it also enables cell sorting into performance bins, which allows module fabricators to maximize power output. Essentially, the lowest-performing cell within a series string will limit the output current and the subsequent power produced by the solar module. This limiting effect also extends to the installed system. Thus, system designers expect narrow module classes to reduce loses from module mismatching (Fig. 7.2).

7.3

BASIC MODULE ASSEMBLY

While alternative materials are available for flexible solar products, low iron tempered glass remains the standard load-bearing component for today’s silicon-based solar module. The glass must be fully cleaned to ensure adequate adhesion to the subsequent module layers. After the glass is permitted to dry, the first encapsulant layer is added. EVA continues to be the primary encapsulating plastic used in PV modules. Today’s faster formulations have cut lamination cycle times in half. Additionally, EVA has demonstrated improved reliability, and UV stabilizers have been added

162

TERRESTRIAL MODULE FABRICATION

Figure 7.2. Picture of cell testing machine. In addition to machine throughput improvements now with over 1000 cells/h, today’s cell test machines must be able to handle everthinning cell thicknesses.

to protect the laminate package against the sun’s spectrum and have prevented the degradation of EVA. Other encapsulant materials have continued to play an active role in the production process. Furthermore, some thin-film processes may be incompatible with the EVA process and have returned to using PVB—a material used in the 1970s before EVA became the prominent encapsulant. Next, solar cells are interconnected into series strings. While some module manufacturers use manual labor to interconnect solar cells, most new entrants to module building are utilizing available automated tabbing and stringing equipment. There are various soldering approaches, including conduction (hot bar), convection (hot air), and radiation (infrared lamps). The standardization of cell contact printing has enabled equipment manufacturers to standardize alignment and soldering equipment. Roughly 5 years ago, four stringing machines were required to support a 12-MW line. Today, the standard throughput exceeds 500 cells/h. Coupled with larger solar cells, one stringing machine can connect soldered strings to support a 12-MW production line (Fig. 7.3). The connection of cell strings (busing) remains similar. The strings are connected using a thicker and/or wider bus wire (ribbon) that brings the module current and voltage through an egress hole or slit in the back sheet ultimately terminated in a junction box. After the cell circuit is completed, an optional fiberglass layer is added to ensure cell ribbon isolation from the rear layer of the module. Additionally, this layer aids air evacuation and minimizes cell string movement during the lamination process.

BASIC MODULE ASSEMBLY

163

Figure 7.3. Photo of automated tabbing and stringing machine. Backsheet

Fiberglass

EVA

Cells

Glass

Figure 7.4. Picture of laminate layers.

A second sheet of encapsulant and the back-sheet layers are added to encapsulate the power-producing section of the module. While module designs may vary based on system requirements, all current designs require multiple power leads to egress the rear layer for string output monitoring. While PVF is still the dominant back-sheet material, increasing module production has stressed the available supply. This has encouraged additional material suppliers to explore other material options. These newer materials are beginning to supplement PVF shortages. Additionally, glass back sheets remain a viable option for the rear surface, especially in BIPV and in thin-film modules. Modern production equipment, such as laminators, must exhibit the flexibility to handle these various module constructions (Fig. 7.4). The layered package is placed into a lamination machine. The cycle time will depend on the curing time required for the encapsulant layers. Typically, the lamination cycle is divided into to two phases. First, the air is evacuated from the heated

164

TERRESTRIAL MODULE FABRICATION

processing chamber while the encapsulant layers melt. During the second phase, pressure is applied onto the laminate package to evacuate any remaining air. The total lamination time is approximately 10–20 min, depending on materials. After the laminate is permitted to cool, excess encapsulant and back-sheet material is trimmed; module frames, a junction box, and cables are added. Within the junction box, bypass diodes are incorporated to sense reverse current. A reverse current condition is possible if one or more cells are shaded or covered during daylight hours. Essentially, the cell would not pass the module current; the cell would overheat and potentially break the module glass. The installed bypass diodes protect the module until the shading is resolved. The final testing of modules can be a labor-intensive exercise. The module must be presented to a sun simulator. The module’s cables must be connected. Next, the module is transferred to a required high-voltage tester, where the module cables and frame are connected. The test results are used to rate the module and to monitor module line performance. Ultimately, after combining all of these module materials, the resulting module is expected to last in excess of 25 years.

7.4

EQUIPMENT OPTIONS

While module manufacturers endeavor to produce the best or the cheapest module to remain competitive, the equipment manufacturers must evolve to effectively handle new materials and larger module formats. The function of the equipment is to minimize handling and stress on the solar cells whenever possible. Some module manufacturers chose to establish or move factories to economically challenged areas, to take advantage of lower labor rates. Alternatively, more automation may be incorporated to reduce labor costs. The line concept has evolved to include powered conveyors and robotics. The addition of conveyors is a natural evolution as the solar industry matures. Conveyors are available to transport the glass through the factory as it transforms into a laminate and ultimately into a solar module. By utilizing conveyors, automation may be integrated at the various process steps. Essentially, by limiting operator contact, the resulting module can be produced faster and more consistently. Robots are now engaged throughout the module assembly process. The introduction of robotics was assisted by experience from other industries. While operators feed material into the work cells, robots perform the assembly steps and advance the module through the processes. For example, an automated 50-MW module assembly line utilizes robotics to load the module glass, to complete soldered cell string layup, and to install bus ribbon prior to the lamination process (Fig. 7.5). The fundamentals of busing and laminate assembly remain similar; however, the implementation of automation has allowed module manufacturers to manage these process steps more efficiently. For example, the utilization of robots transforms the soldering and assembly processes into a timed process step eliminating the dependence on an operator’s level of experience (Figs. 7.6 and 7.7).

EQUIPMENT OPTIONS

165

Figure 7.5. Photo of glass loading. A glass-loading robot eliminates labor from the beginning of the assembly process.

Figure 7.6. After cell soldering, the robot places completed strings onto the prepared glass.

Some module equipment is developed to reduce handling and others are developed to increase throughput. Larger lamination machines address both goals. Equipped with conveyors to automatically load and unload laminates, one laminator can now laminate four 220-W laminates per cycle. Coupled with the fastest cure EVA, just two laminators are necessary for a 50-MW fully automated module line (Fig. 7.8).

166

TERRESTRIAL MODULE FABRICATION

Figure 7.7. Robots with specially equipped soldering arms complete the module circuit.

Figure 7.8. Photo of larger area laminator. Larger area laminators integrated with conveyor systems provide higher throughput with minimal to no handling.

Additionally, robots perform the edge trimming, framing, and junction box placement process steps as the laminate travels along conveyors to the sun simulator (Figs. 7.9–7.11). Larger modules are very important to reducing system costs. While modules are purchased by module power (watts), the overall system costs are lower due to fewer junction boxes, cables, and frames. Additionally, fewer modules per installation reduce the number of DC connections and simplify system mounting requirements. This has led to the development of special equipment designed to handle the increased module sizes. Today, automation can eliminate manual lifting,

EQUIPMENT OPTIONS

167

Figure 7.9. Module frames are affixed in a robotic work cell.

Figure 7.10. Photos of framing and junction box application.

loading, and connecting from the operation. Automation can also apply the power label after testing. New ideas and initiatives that benefit the solar module manufacturing process will emerge from the increasing attention enjoyed throughout the renewable energy industries. Technology improvements will likely compliment other solar segments. For example, alternative and more efficient cell designs may be incorporated into concentrator and BIPV applications.

168

TERRESTRIAL MODULE FABRICATION

Figure 7.11. The completed module is conveyed over the sun simulator for electrical performance test results. Next, the module is transported through the high-voltage tester. Lastly, the performance and safety labels are affixed to the rear side of the module.

Figure 7.12. Picture of building integrated PV (BIPV) project at Denali National Park Visitor Center in Alaska.

Through additional governmental incentives, it becomes more popular to integrate solar products into new and existing structures. Attractive BIPV projects capitalize on the versatility of silicon cells, encapsulants, and other module materials (Fig. 7.12).

7.5

FUTURE

A combination of improved cell efficiency, reduced material costs, and governmental support shall increase the amount of electricity generated from silicon-

ABBREVIATIONS

169

based modules in the coming years. Ultimately, the renewed attention and global push for renewable energy solutions will invigorate the solar industry. The collaboration of ideas and people from other industries will further challenge today’s best practices and increase productivity while lowering system costs. While this growth is beneficial, in order for the industry to evolve, lower material costs are needed. The availability and quality of bus wire for cell-to-cell and circuit connections, encapsulating plastics, and composite back-sheet materials directly affect module manufacturing. While the basic module construction and manufacturing processes have changed little, the methods and equipment available have matured. ABBREVIATIONS BIPV—building integrated photovoltaics DC—direct current EVA—ethylene vinyl acetate PVF—polyvinyl fluoride PVB—polyvinyl butyral UV—ultraviolet

8 CHINESE SOLAR CELL STATUS WANG SICHENG Energy Research Institute, National Development and Reform Commission

8.1

INTRODUCTION (BY THE EDITOR)

Driven by the pressure of global warming, rising oil prices, and a deteriorating natural environment, clean renewable energy is receiving greater attention worldwide. China’s central government has stated its support for the development of a sustainable energy system that maximizes energy efficiency and the use of renewable energy sources. A key aspect of that initiative is the application of PV technology to convert light into electricity. The Chinese PV manufacturing industry has grown dramatically in recent years due to a strong demand from overseas markets. China’s production of solar cells and modules has grown at an average annual rate of 49.5% since 2002. By 2008, the production of solar cells reached over 2 GW, which was 33% of the global production. China now contributes a large part of the worldwide solar production and is now the largest producer of solar cells in the world. Intelligent and supportive Chinese central government policy has contributed to this rapid growth in the Chinese solar cell industry. This chapter provides an overview of the coordinated efforts in China to develop a strong domestic PV industry. This chapter was written by Wang Sicheng, who has contributed to the formation of China’s government policy. It was translated into English by Huang Han Xinag at JX Crystals Inc., and editorial comments are by Lewis M. Fraas. 8.2

CHINA’S SOLAR ENERGY RESOURCES AND DATA

8.2.1

Solar Irradiation Resources in China

PV power generation largely depends on the solar energy resources of the location. Therefore, it is necessary to know the distribution of the solar energy resources in Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

171

172

CHINESE SOLAR CELL STATUS

II I III

IV

Figure 8.1. Geographic distribution of the solar irradiation of China. The territory can be divided into four resource levels shown. TABLE 8.1. Solar Irradiation Resources in China on a Horizontal Surface Level of Irradiation

Region Number

Annual Irradiation Max (MJ/m2)

Annual Irradiation Min (kWh/m2)

Daily Irradiation Average (kWh/m2)

Very high

I

≥6300

≥1750

≥4.8

High

II

5040–6300

1400–1750

3.8–4.8

Average

III

3780–5040

1050–1400

2.9–3.8

Low

IV

40%).

during the course of the day and season and is usually not equal to the reference spectrum. Due to the resulting current limitations, it is expected that concentrator systems with triple-junction cells are more sensitive to changes in the solar spectrum than CPV systems using single-junction cells. The influence of the spectral variations on the electrical parameters of a triplejunction solar cell can be investigated in the laboratory by using a multi-light source simulator [34, 35]. Here, the electrical parameters of the triple-junction solar cell are measured, while the incident spectrum is changed systematically. Thus, on the cell level, the influence of the spectral changes is well investigated. Under varying sun spectra, the power of triple-junction solar cells can be expected to vary by about 20%. Unfortunately, the influence of the solar spectrum on the energy production of concentrator systems equipped with triple-junction cells is not well investigated at present. Long-term field experience is still lacking due to the rather young triple-junction technology and the lack of information on the changing solar spectrum combined with concentrator optics. However, some investigations of Fresnel lens concentrators equipped with triple-junction cells have been already performed and show the spectral impact in the morning and evening hours [36–38]. At these high air masses, spectral distribution of the sunlight is relatively “redrich.” As a result, the top cell current is significantly lower than the current of the other subcells and the efficiency of the concentrator system decreases.

RECENT EXAMPLES AND DEVELOPMENTS

327

Obviously, the annual energy production of concentrator systems using triplejunction solar cells is more affected by the solar spectrum than systems with singlejunction cells. In a number of theoretical studies, the annual energy production of multijunction solar cells with respect to the solar spectrum has been investigated [39, 40]. The basic conclusion from these theoretical investigations is that despite the penalty of current limitation, triple-junction solar cells outperform other solar cell concepts. This is not surprising when considering that air masses