chapter x energy conversion with photovoltaics

Chapter X

ENERGY CONVERSION WITH

PHOTOVOLTAICS

Chapter X.— ENERGY CONVERSION WITH PHOTOVOLTAICS

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .393

Photovoltaic cells. . . . . . . . . . . . . . . . . . . . . ...395Cells Designed for Use in Concentrated

Sunlight . . . . . . . . . . . . . . . . . . . . . . . . . .. .400Optimizing Conventional Silicon Cells

for Use in Concentrated Sunlight .. ...400Interdigitated Back-Contact Cells . . . . . . .400Gaiiium-Arsenide Cells. . ..............401The Vertical Multifunction Cell . ........401Horizontal Multifunction Cells. . . . .. ....402Thermophotovoltaic Cell . . . . . . . .Multicolor Cells . . . . . .................403

Concentrator Systems . . . . . . . . . . . . . . . . . . . . 403Photovoltaic Cogeneration .. ..............406The Credibility of the Cost Goals . . . . . . . . . . . . 407

Markets . . . . . . . . . . . . . . . . . . . . . . . . . . .. ...408An Analysis of Manufacturing Costs. .. ....412

Silicon. . . . . . . . . . . . . . . . . . . . . . . . ......412Thin Films . . . . . . . ....................420

MaterialsAvailability and Toxicity. ... .......421Silicon . . . . . . . . . . ......................422Cadmium . . . . . . . . . . . . . . . . . . . . . . . . . . . . .422Gallium Arsenide. . .....................423

Performance of Cellsin Real Environments. ..423PassiveCooling. . ......................424ActiveCooling. . . ......................424

Photochemical EnergyConverslon. . ........425

LIST OF TABLESTableNo. Pagel. a) Key Milestones for National Photovoltaic

Conversion Program; and b) Goals forConcentratorSystemsUsingSilicon Cells .394

2. PhotovoltaicCell Efficiencies . . . . . . . . ....3983. Temperature Dependence of Photovoltaic

Devices . . . . . . . . . ......................407

4. Market Forecasts for Photovoltaic Devicesat Different Prices. . ....................410Years To Achieve Indicated Price Reductionas a Function of Learning Curve andAssumed Growth in Demand . . . . . . .......412The Distribution of Costs in the Manufactureof Silicon Photovoltaic Devices UsingContemporaryTechnology. . .............414

7. The MajorSt6psin Manufacturing a SiliconPhotovoltaic Device . . . . . . . ...:..... ....419

8. EffectiveConvective Heat-TransferCoefficients [ke] for a Variety of CoolingSystems . . . . . . . . ......................424

LIST OF FIGURESFigure

PageArray of Silicon Photovoltaic CellsProviding PowertoPumpWater . ........393A Typical Photovoltaic Device . . . . . . . . . . .396An lnterdigitated BackContact SolarCelland CrossSection. . ...................401The Vertical MultijunctionCell . .........401Thermophotovoltaic Converter . .........403Experimental Concentrating PhotovoltaicArray in Operation. The Array EmploysFresnel Lenses and Silicon PhotovoltaicCells . . . . . . . . . . . .....................404Breakeven Costs for ConcentratingPhotovoltaicCollectorsCompared WithFlat-Plate Devices. .,..... . ................405Historical Learning Curves From theSemiconductorIndustry. . ..............409PhotovoltaicArray Prices for Several MarketProjections. . . . . ......................411Commercial Silicon Photovoltaic Cells ...413Silicon Ribbon Being Grown by the ’’Edge-Defined Film-FedGrowth” Process .. ....417Thin-Film Cadmium Sulfide-CopperSulfideSolarCel ls . . . . . . . . . . . . . . . . . . . . . . . . . . .420Photochemical Energy Conversion. .. ....425

Chapter X

Energy Conversion With Photovoltaics

INTRODUCTION

Photovo taic cells (often called “solar cells”) are semiconductor deviceswhich are capable of converting sunlight directly into electricity. Generatingelectricity in this way has obvious advantages. The photovoltaic cells have nomoving parts, and they do not require high-temperature or high-pressurefluids. Since they are extremely reliable, quiet, safe, and easy to operate, theyare ideally suited to onsite applications. Photovoltaic energy production isunlimited by scarce materials, since cells can be manufactured from silicon,one of the Earth’s most abundant elements.

There is no doubt as to the technical feasibility of generating electricityfrom photovoltaic devices. The units have been used for many years in thespace program and to provide power in remote terrestrial sites. The array ofsilicon cells shown in figure X-1 provides power to pump water in a remotelocation in Arizona.

The most immediate barrier faced by all photovoltaic systems is the pres-ent high cost of the devices. Photovoltaic arrays could be purchased in largequantities in late 1977 for about $11 per peak watt of output. * Electricity fromsystems using such cells costs $1.50 to $2.00/kWh.

Figure X-1 .—Array of Silicon Photovoltaic CellsProviding Power to Pump Water

.

SOURCE Solar Technology International

Development work to reduce the cost ofphotovoltaic energy can be divided intothree general categories:

1. Reducing the cost of manufacturing thesingle-crystal silicon celIs which arenow on the market;

2. Developing techniques for mass pro-

3

ducing and increasing the performanceof cells made from thin films of mate-rials such as CdS/Cu2S or amorphoussilicon, and

Developing high-efficiency cells whichcan be installed at the focus of magni-fying optical systems,

It is technically possible to use any ofthese approaches to reduce costs to $1 to $2

* Photovoltaic prices throughout this paper refer tothe sel I Ing price of arrays of cel Is encapsulated to pro-tect them from the weather, f o b the manufacturingfacility In 1975 dollars The peak output of a cellrefers to the output which would be obtained If thecell were exposed to the Sun at the zenith on a clearday.

393

394 ● Solar Technology to Today’s Energy Needs

per peak watt (electricity costing $0.10 t o$0.40/kWh) during the next 3 to 5 years. Theachievement of costs below $1 to $2 perwatt will require a considerable amount ofengineering development work. Progress inany of a number of current research pro-grams would give us considerably more con-fidence about the prospects for achievingsubstantialIy lower costs.

A set of goals for reducing the cost ofsilicon photovoltaic devices was establishedsomewhat arbitrarily during the crash Proj-ect Independence studies conducted in1973. The Department of Energy (DOE) be-lieves that, with some relatively minor ad-justments, these goals are achievable and isusing them for planning purposes. The pres-ent goals are: $1 to $2 per watt by 1980-82,$0. SO/watt by 1986, and $0.10 to $0.30/wattby 1990. Current goals are shown in table

X-1. The lower cost goals appear to be op-timistic but not impossible.

The DOE price goals assume that the ar-rays will last approximately 20 years. (Pres-ently, terrestrial arrays are guaranteed for 1year.) This seems to be an attainable objec-tive, although more data is required on thedegradation rate for arrays exposed to theenvironment for long periods of time. Sili-con cells apparently fail only when the ma-terial used for encapsulation cracks orleaks. Structural failures and corrosion haveoccurred in improperly encapsulated de-vices, ’ and most clear plastics darken withprolonged exposure to sunlight, cuttingdown the light reaching the cells. Glass oracrylic encapsulation should, however, beable to prevent these problems.

‘A Johnson (M ITRE Corporation), private com-munication, 1977.

Table X-1

a) Key Milestones for National Photovoltaic Conversion Program——

Array price in 1975, Production rate,dollars per peak megawattspeak watt per year

1. End of FY 1977 . . . . . . . . . . 112. End of FY 1978 . . . . . . . . . . 73. End of FY 1982 . . . . . . . . . . 1-2 204. End of FY 1986 . . . . . . . . . . 0.50 5005. End of FY 1990 . . . . . . . . . . 0.1-0.3 50,000

SOURCE Photovoltaic program program summary January 1978, U S Department of Energy, Division of Solar Technology

b) Goals for Concentrator Systems Using Silicon Cells

Technical Commercialfeasibility* equipment* ●

1975 dollars Silicon 1975 dollars Siliconper peak watt cell per peak watt celI

(entire system) efficiency (entire system) efficiency— —— —1. End of FY 78. . . 2 16% 15 13.5%2. End of FY 80. . . 1 180/0 2.75 16%3. End of FY 82. , . 0.50 200/0 1.60 180/0

“ Laboratory proof of concept + reasonable estimate of cost of commercial system based on concept

“* F o.b. price of production technology.

SOURCE Annual Operation Plan, Systems Deflnltlon Prolect (FY 1978), Sandla Laboratory

Ch. X Energy Conversion With Photovoltaics ● 3 9 5

PHOTOVOLTAIC CELLS

All photovoltaic installations will consistof small, individual generating units — thephotovoltaic cell. Individual cells will prob-ably range in size from a few millimeters toa meter in linear dimension, A surprisingvariety of such cells is available or is in ad-vanced development since it has only beenin the past 5 years that serious attention hasbeen given to designing cells for use in any-thing other than spacecraft. Devices areavaiIable with a large range of efficienciesand voltages. Some cells are designed towithstand high solar intensities and highoperating temperatures, while others are de-signed to minimize manufacturing costs.The variety should not give the mistaken im-pression that this is an area where largeamounts of private or public research fund-ing have been directed; indeed, many of theprojects described here are being conductedby small research laboratories.

The photovoltaic equivalent of powerengineering is something of an anomaly inthe generating industry since it involvesmanipulations in the miniature and silentworld of semiconductor physics instead ofsteam tables, gears, and turbine blades. Thefollowing discussion provides only a verybrief excursion through this complex field.More complete discussions of the topicscovered can be found in several recentp u b l i c a t i o n s . 2 3 4 5 All that can be done hereis to outline some of the major effects in-fluencing cell cost and performance.

The energy in light is transferred to elec-trons in a semiconductor material when alight photon collides with an atom in thematerial with enough energy to dislodge anelectron from a fixed position in the mater-

2t-i ). Hovel, Semiconductors and Semimeta/s,Volume 11, Solar Cells, Academic Press, New York,1975

‘C E Bacus, (ed ), Solar Cc//s, I E EE Press, New York,1976

4 A Rothwarf and K W Boer, frog, in Solid StateChem., 10(2), p 77 (1975)

‘K W Boer and A Rothwarf, Ann. Rev. Mat. Sci. (6),p 303 (1976)

ial (i. e., from the valence band), giving ite n o u g h e n e r g y t o m o v e f r e e l y i n t h ematerial (i. e., into the conduction band).

A vacant electron position or “hole” isleft behind at the site of this collision; such“holes” can move if a neighboring electronleaves its site to fill the former hole site. Acurrent is created if these pairs of electronsand holes (which act as positive charges) areseparated by an intrinsic voltage in the cellmaterial,

Creating and controlling this intrinsicvoltage is the trick which has made semicon-ductor electronics possible. The most com-mon technique for producing such a voltageis to create an abrupt discontinuity in theconductivity of the cell material (typicallysilicon in contemporary solid-state compo-nents) by adding small amounts of im-purities or “dopants” to the pure material.This is called a “homojunction” cell. A typi-cal homojunction device is shown in figurex-2.

An intrinsic voltage can also be createdby joining two dissimilar semiconductor ma-terials (such as CdS and CU2S), creating a“heterojunction,” or by joining a semicon-ductor to a metal (e.g., amorphous silicon topaladium) creat ing a “ S c h o t t k y ” b a r r i e r

junction.

A fundamental limit on the performanceof all of these devices results from the factthat (1) light photons lacking the energy re-quired to lift electrons from the valence tothe conduct ion bands ( the “band gap”energy) cannot contribute to photovoltaiccurrent, and (2) the energy given to electronswhich exceeds the minimum excitat ionthreshold cannot be recovered as usefulelectrical current. Most of the unrecoveredphoton energy is dissipated by heating thecell.

The bulk of the solar energy reaching theEarth’s surface falls in the visible spectrum,where photon energies vary from 1.8 eV(deep red) to 3 eV (violet). In silicon, only


\Rear Metal Contact

about 1.1 eV is required to produce aphotovoltaic electron, and in GaAs about1.4 eV. Choosing a material with a higherenergy threshold results in capturing a largerfraction of the energy in higher energyphotons but losing a larger fraction of lowerenergy photons. The theoretical efficiencypeaks at about 1.5 eV, but the theoretical ef-ficiency remains within 80 percent of thismaximum for materials with band gapenergies between 1 and 2.2 eV.6 7

‘M. Wolf, Energy Conversion, 11, 63(1 971 ).‘Joseph J. Loferski, “Principles of Photovoltaic

Solar Energy Conversion,” 2Sth Artnua/ Proceedings,Power Sources Conference, May 1972.

E l e c t r o n s a c t u a l l y l e a v e c e l l s w i t henergies below the excitation voltage be-cause of losses attributable to internal re-sistance and other effects, not all of whichare understood. a (An electron leaves a typi-cal silicon cell with a useful energy of about0.5 eV.)

The same kinds of fundamental limits ap-ply to photochemical reactions in which alight photon with energy above some fixed

‘M. Wolf, University of Pennsylvania, “Recent im-provements and Investigations of Silicon Solar CellEfficiency, ” U . S . / U . S . S . R . joint Solar E n e r g yWorkshop, Ashkhabad, U. S. S. R., September 1977.

Ch. X Energy Conversion With Photovoltaics ● 397

excitation threshold is able to produce achemical reaction or a structural changewhich can be assigned a fixed energy. T h etheoretical limit to the performance of sev-eral types of cell designs is shown in tablex-2.

The performance of real cells (also shownin table X-2) falIs below the theoretical max-imum for a number of reasons. One obviousproblem is reflection of light from the cellsurface (which can be reduced with specialcoating and texturing) and reflection fromthe electrical contacts on the front surfaceof the cell (which can be reduced with care-ful contact design). ’ Losses also result fromthe fact that the photo-generated electronsand holes, which fail to reach the region inthe cell where they can be separated by theintrinsic voltage, cannot contr ibute touseful currents.

Photo-generated charges can be lostbecause of imperfections in the cell crystalstructure, defects caused by impurities, sur-face effects, and other types of imperfec-tions. Losses are minimized if a perfect crys-tal of a very pure semiconductor material isused, but producing such a crystal can be ex-tremely expensive. Manufacturing costs canprobably be greatly reduced if cells con-sisting of a number of smalI crystal “grains”can be made to operate with acceptable ef-ficiencies. The size of the grains which canbe tolerated depends on the light absorbingproperties of the cell material. If absorptionis high, photovoltaic electron-hole pairs willbe created close to the cell junction wherethe voltages exist; relatively smalI grainsizes can be tolerated since the chargesneed drift only a short distance before beingsorted by the field.

Silicon is a relatively poor absorber oflight and, as a result, cells must be 50 to 200microns thick to capture an acceptable frac-t ion of the incident l ight . This placesrather rigorous standards on the sizes of

‘J Llndmayer and J F All ison, Cornsat Tech. # 3 ,1973, p 1

‘OJ R Hauser and P M Dunbar, IEEE Transactionson E/ectron Devices, E D-24 (4), 1977, p 305

crystal grains which can be tolerated, and al Icommercial sit icon celIs are now manufac-tured from single crystals of silicon. It isbelieved that if polycrystalline silicon is tobe used, individual crystal grains must be atleast 100 microns on a side if efficiencies ashigh as 10 percent are to be achieved. 11 12 Itis important that the grains be oriented withthe grain boundaries perpendicular to thecelI junction so that charge carriers canreach the junction without crossing a grainboundary, A number of research projectsare underway to develop inexpensive tech-niques for growing such polycrystallinem a t e r i a l s 13 14 15 and for minimizing the im-pact of the grain boundaries. ” Efficienciesas high as 6.7 percent have been reportedfor vapor-deposited polycrystalline siliconcelIs with grains about 20 to 30 microns on aside’ 7, and a proprietary process capable ofproducing grains nearly a millimeter on aside reportedly can be used to produce cellswith efficiencies as high as 14 percent,Work is underway to improve crystal grow-ing techniques and to enlarge grains withlasers and electron beams.

Perhaps the most intriguing recent devel-opment is the discovery that an amorphoussilicon-hydrogen “alloy” can be used to con-

‘)A R Kirkpatrick, et al., Proceedings of the ERDASemiannual Photovoltaic Advanced Materials Pro-gram Review Meeting, Washington, D C , Mar 22-23,1977

‘lA, Baghdad[, et al , ERDA Semiannual Photo-volta ic Advanced Materla Is Program Review Meeting

‘ ‘T. L. Chu, et al., IEEE Transaction on E lect ronDevices, ED-24 (4), 1977, pp. 442445.

14H. Fischer and W. Pschunder, (A EG-Telefunken),“Low-Cost Solar Cells Based on Large-Area U nconven-tional Silicon, ” IEEE Transactions on E/ectron Devices,ED-24(4), 1977, pp. 438-441

‘5L. D. Crossman, Research and Development of

Low-Cost Processes for Integrated Solar Arrays, (Thirdquarterly progress report, ERDA EC (11-1 }2721,Energy Research and Development Admlnistratlon,Washington, DC., 1975), p. 38

“T H Distefano and J. J Cuomo, 1 BM, “Enhance-ment of Carrier Lifetime in Polycrystalllne Silicon, ”Proceedings, National Workshop on Low-Cost Poly-crystal I ine Silicon Solar Cel Is, Dallas, Tex , May 18-19,1976.

“T. L Chu, et al., op. cit pp 442-445‘“H` Fischer and W Pschunder, op clt , p 438


Table X-2.— Photovoltaic Cell Efficiencies—

1.

2.

3.

4.

5.

6.

7.

8.

9.

Perform-ance

of com-mercial

cells

10-15——

2-3

——

—

——

—

12.5

——————

Reference

(1 ,2)(4,5)(6,7)

(8,9)

(lo)(9)

(11 )

(12)(1 ,13)

(3)

(14,15,16)

(17)(18)(19)(20)(21 )(22)

● Techniques for reporting efficiencies differ Wherever Possible, efficiencies were chosen which assume air mass 1 and Include losses due toreflection and contact shading

H. J. Hovel, Semiconductors and Semimetals, Vol. 2:So/ar Cc//s (Academic Press, New York, 1975).

J. Lindmayer and C. Y. Wrigley, “Development of a 20Percent Efficient Solar Cell,” (NSF-43090, NationalScience Foundation, Washington, D. C., 1975),

J. V. DeBow, paper presented as part of the Pro-ceedings of the Energy Research and DevelopmentAdministration Semiannual Photovoltaic AdvancedMaterials Program Review Meeting, Washington,D. C., Mar. 22-23, 1977.1.1. Chu, S. S. Chu, K. Y. Duh, H. 1. Yoo, /EEE Transac-tions on Electron Devices, ED-24 (No. 4), 442 (1977).H. Fischer and W. Pschunder (AEG-Teiefunken). “Low-Cost Solar Celis Based on Large-Area Unconven-tional Siiicon, ” IEEE Transactions on ElectronDevices. pp. 438-441.D. E. Carlson, IEEE Transactions on Electron Devices,ED-24 (No. 4) 449 (1977).D. E. Carlson and C. R. Wronski, App/. Phys. Lett. 28,671 (1976).K. W. Boer and A. Rothwarf, Annu. Rev. Mater. Sci. 6,303 (1976).A. M. Barnett, J. D. Meakim, M. A. Rothwarf, “Progressin the Development of High Efficiency Thin Film Cad-

10.

11.

12.

13.

14.

15.

16.

17.

mium Sulfide Solar Ceils,” (ERDA E (49-18)-2538,Energy Research and Development Administration,Washington, D.C. 1977).G. A. Samara, paper presented as part of the Pro-ceedings of the ERDA Semiannual Photovoltaic Ad-v a n c e d Materiais P r o g r a m R e v i e w M e e t i n g ,Washington, D. C., Mar. 22-23, 1977.J. 1. Shay, S. Wagner, H. M. Kasper, App/. Phys. Lett.27,89 (1975).L. J. Kazmerski, S. R. White, G. K. Morgan, ibid., 29,268 (1976).R. J. Stirn (Jet Propulsion Lab). “High Efficiency Thin-film GaAs Soiar Cells”, 730-9 First Interim Report forDOE/NASA, Dec 1977.

1. S. Napoli, G. A. Swartz, S. G. Liu, M. Klein, D. Fair-banks, D. Tames, RCA Rev. 38,76 (1977).

H. J. Hovel (IBM), “The Use of Novei Materiais and De-vices for Sunlight Concentrating Systems,” Pro-ceedings DOE Photovoltaic Concentrator SystemsWorkshop, Tempe, ARiz., May 1977, pp. 35-54.

J. Gibbons (Stanford University), personal communi-cation, October 1977.M. D. Lambert and R. J. Schwarz, IEEE Transactionson Electron Devices, ED-24 (No. 4), 337 (1977).


Table X-2 references—continued

18.

19.

20.

R. M. Swanson and R. N, Bracewell, “ S i l i c o nPhotovoltaic Cells in Thermophotovoltaic Conver-sion” (EPRI ER-478, Electric Power Research in-stitute, Palo Alto, Cal if., February 1977).

J. Harris, et. a/. (Rockwell International). “High Effi-ciency AIGaAs/GaAs Concentrator Solar Cells,” TheC o n f e r e n c e Record of the T h i r t e e n t h /EEEPho fo vo/fa ic Specia Iis ts Conference- 1978.Washington, D. C., June 1978.

L. W. James and R. L. Moon (Varian Associates).“GaAs Concentrator Solar Cells” The ConferenceRecord of the Eleventh IEEE Photovoltaic SpecialistsConference, Scottsdale, Ariz., May 1975, pp. 402-408.

struct photovoltaic cells with useful effi-ciencies. Efficiencies of 5.5 percent havebeen measured” and 15 percent efficiencymay be possible. 20 The hydrogen apparentlyattaches to “dangling” silicon bonds, mini-mizing the losses that wouId otherwiseresult at these sites. Acceptable perform-ance is possible, despite the large number ofremaining defects, because the amorphousmaterial is an extremely good absorber oflight; test cells are typically 1 micron or lessin thickness. 21 The properties of this com-

plex material are not welI understood.

GaAs or CdS/Cu2S are also much betterabsorbers of Iight than crystalline silicon, socells made from these materials can be thin-ner and tolerate smaller crystal grains thanwas possible with the crystalIine siIicon.Commercial CdS/Cu2S cells will probably be6 to 30 microns thick, 22 23 and the crystalstructure produced with a relatively simplespray or vapor deposit process is largeenough to prevent grain structure from sig-nificantly affecting celI performance.

The primary drawback of most of the“thin film” celIs is their low efficiencies.Research is proceeding rapidly in a number

“D E Carlson, RCA Laboratories, “Amorphous Sili-con Solar Cells, ” IEEE Transactions on ElectronDevices, ED-24(4), 1977, pp 449-452

20 and C R Wronskl, App Phys, Lett. 28, 1976, p671

2 ‘A R Moore, E/ectron and Ho/e Drift Mobr’/ity inAmosphous Si/icon (to be published)

22 Boer, op clt , p 3192]) F jordan, Photon Power, Inc., El Paso, Tex ,

private communication, February 10, 1977

21.

22.

J. J. Loferski (Brown University), “Tandem Photovol-taic Solar Cells and Increased Solar Energy Conver-sion Efficiency,” The Conference Record of theTwe/fth /EEE Photovo/taic S p e c i a l i s t s Con/er-ence-1976, Baton Rouge, La., November 1976, pp.957-961.H. J. Hovel (IBM). “Novel Materials and Devices forSunlight Concentrating Systems,” /BM Journa/ ofResearch and Development. Vol. 22, No. 2, Mar 1978,T. T. Rule, et. a/. (Arizona State University). “TheTesting of Specially Designed Silicon Solar CellsU n d e r H i g h S u n l i g h t I l l u m i n a t i o n , ” Twe/fthIEEE- 1976. Pp. 744-750.

of areas, however, and a number of thin filmcells may be able to achieve efficienciesgreater than 10 percent.

Older designs of cadmium cells degradedand fai led relat ively rapidly, but accel-erated lifetime tests on modern designs ap-pear to indicate that cel ls hermetical lysealed in glass with proper electrical loadingcould have a useful life of decades.24 25 26

Research is also underway on a number ofother materials which may be used to manu-facture inexpensive thin film cells. Experi-mental cells have been constructed whichsubstitute iridium phosphide or copper in-idium selenide for the copper sulfide in theC d S / C u2S heterojunction, and efficienciesgreater than 10 percent have been demon-strated in single-crystal laboratory cellsmade with these materials. 2728

It is also possible to convert light energydirectly into electricity by exposing elec-trodes immersed in chemicals to sunlight. ifthe materials are properly chosen, a currentcan be produced without any net chemicalchange in the materials used. Conversion ef-

L

“J . Besson, et al , “Evaluation of CdS Solar Cells asFuture Contender for Large Scale E Iectriclty Produc-t ion, ” The C o n f e r e n c e Record of the E/eventh /EEEPlrotovo/taic Specialists Conference–1 975, Scotts-dale, Ariz , May 6-8, 1975 (Institute of E Iectrlcal andElectronics Engineers), pp 468-475

z 5E w c reenwich and F A Shlrland ( W e s t i n g h o u s e

Research Labs), “Accelerated Life Performance Char-acteristics of Thin-Film CU25 Solar Cel/s-CdS, ” froceedings, E Iectrochem Ica I Society Solar Energy Sym-posium, Washington, D C , May 2-7, 1976

“j F jordan, P h o t o n Power,lnc , p r i v a t e c o m -munication, Apr 5, 1976

27S Wagner, et al , Appl. Phys. Lett., 26, 229, 1975“j Shay, et al , App/. Phys. Lett,, 27, 89,1975

400 . Solar Technology to Today’s Energy Needs

ficiencies greater than 5 percent have beenreported with polycrystalline CdSe-basedp h o t o e l e c t r o c h e m i c a l c e l l s .29 30 31 32 33 34

CELLS DESIGNED FOR USE IN

CONCENTRATED SUNLIGHT

High efficiencies are important for cellsused in concentrated sunlight, since in-creased cell performance means a reductionin the area which must be covered with themagnifying optical equipment, which domi-nates the system’s cost.

Modified silicon cells, designed to per-form in concentrated sunlight are not in-herently more expensive than ordinary cells,but are always likely to cost more per unit ofcell area since production rates will belower and since more care will be taken intheir manufacture. However, since the cellscover only a fraction of the receiving area,much more can be spent on any individualcell. Cells such as the thermophotovoltaicdevice and the GaXA l1-XAs/GaAs cells may beconsiderably more expensive per unit areathan the silicon devices, but this cost maynot be significant since the devices can beused with concentrations of 500 to 1,000 ormore and can be much more eff icient. A v a -riety of approaches are being investigatedfor achieving high-efficiency performance inconcentrated sunlight.

Optimizing Conventional Silicon Cells forUse in Concentrated Sunlight

The current from a photovoltaic cell in-creases almost Iinearly with increasing sun-light intensity and the voltage increasesslightly faster than the logarithm of the in-

jgMark ‘j WrightOn, MIT, private communication,

Mar 26, 1977.30A. B, Ellis, et al., ). Amer. Chern. Soc., 98, 1635,

6418, 6855(1 976)“G. Hodes, et al , Nature, 216, 403,1976,‘2B. Miller and A. Heller, Nature, 262, 680, 1976.“D. L. Laser and A. ). Bard, j. E/ectrochem, Soc.,

123, 1027,1976.“K. F Hardee and A j Bard, ), E/ectrochem. Soc.,

724, 215,1977

tensity. These effects would lead to an in-crease in overall cell efficiency, except thatthe increased current densities in the celllead to increased resistive losses and othereffects. The design of standard silicon cellscan be optimized for operating in intensesunlight by carefully designing the wiresused to draw current from the cells, optimiz-ing the resistivity of the cell material, chang-ing the thickness of the cell junction, andotherwise taking pains in celI manufacture(e.g., better antireflective coatings and sur-face texturing, higher quality silicon, pre-cisely designed gridlines, etc.). Efficienciesas high as 17.9 have been reported for sili-con celIs operating at 1000 C in sunlight con-centrated 200 times. 35

Several ingenious techniques have beensuggested for improving performance of siIi-con devices used in intense sunlight withnovel designs.

Interdigitated Back-Contact Cells

An “interdigitated back-contact” cell ex-poses an unobstructed wafer of intrinsic(i.e., very pure) silicon crystal directly to thesunlight. The junctions that produce the cellvoltages, and that are attached to electricalleads, are entirely on the back of the cell.(The name comes from the shape of the pos-itive and negative electrical contacts on theback side of the cell. See figure X-3.) An effi-ciency of 15 percent concentration, withratios up to about 280, has been reported fora preliminary version of this cell; it is be-lieved that straightforward design improve-ments will result in cells which are at least20-percent efficient. ”

15L s NapOil, et al,, RCA, “ H i g h L e v e l concentra-

tion of Sunlight on Slllcon Solar Cel Is, ” RCA Review,Vol 38, No 1, March 1977, p 91

16M D Larnmert and R. j Schwartz, pU rdue Un iver-

sity, “The Interdlgitated Back Contact Solar Cell: ASilicon Solar Cell for Use in Concentrated Sunlight, ”/EEE Transactions on E/ectron Devices, ED-24(4), 1977,pp 337-341,

Ch. X Energy Conversion With Photovoltaics 401

Figure X-3.— An lnterdigitated Back ContactSolar Cell and Cross Section

I Hlgh-lifetlme bulk region I

SOURCELammert M D and R J Schwartz “The Interdlgltated Back ContactSolar Cell A Sll; con Solar Cell for Use (n Concentrated Sunllghl ”IEEE Transacf/ons ori E/ecfron Dev/ces, VOI ED-24 (1977) p 337

Gallium-Arsenide Cells

Gal l ium arsen ide has a h igher theoret ica l

photovoltaic eff iciency than s i l i c o n b e c a u s ei t s exc i ta t ion thresho ld is bet ter matched to

the energy in the Sun’s spectrum.

CaAs cells can be used to achieve high ef-ficiencies in intense radiation, particularly ifthey are covered with a layer of GaXA l1-XA swhich has the effect of reducing surface andcontact losses. Efficiencies as high as 24.5percent have been measured for suchdevices operating in sunlight concentrated180 times, 38

In addition, both theory and experimentshow that the efficiency of GaAs cells isreduced less by high temperature than is theefficiency of siIicon celIs.

“H Hovel, op clt , p 195“J Harris, et al , Rockwell International, “High Effi-

ciency AI GaAs/G aAs Concentrator Solar Ce[ Is, ” TheConference Record of the Thirteenth /EEE Photovoltaic Specialists Conference-1978, Washington,D C , June 1978

The Vertical Multifunction Cell

The edge-illuminated, vertical, multifunc-tion cell shown in figure X-4 consists of astack of silicon homojunction devices quali-tatively similar to standard cells, illumi-nated in such a way that the light enters thecells parallel to the junctions. Several recentcalculations have indicated that these de-vices have the potential of achieving effi-ciencies as high as 30 percent, 39 althoughvery little work has been done on designingand testing optimized cells. The maximummeasured efficiency to date is 9.6 percentfo r a dev ice w i thou t an an t i re f l ec t i vecoating. 40

Figure X-4.—The Vertical Multifunction Cell(edge illuminated)

I incident sunlight

current

‘9H.J Hovel, IBM, “Novel Materials and Devicesfor Sunllght Concentrating Systems, ” IBM journal ofResearch and Development, 22(2), March 1978,

40T T Rule, et al , Arizona State University, “TheTesting of Specially Designed Silicon Solar CellsUnder High Sunlight Illumination, ” The ConferenceRecord of the Twelfth /EEE Photovo/taic SpecialistsConferenc&1976, Baton Rouge, La., November 1976,pp 744-750.

402 ● Solar Technology to Today’s Energy N e e d s

The high potential efficiency results fromseveral features of the device:41 42 43

●

●

●

●

Since the vertical multifunction devicesare connected in series, they producehigher voltages and lower currents thanother concentrator cells with the samepower output. The low currents reducethe resistance losses. Series connec-tions also mean, however, that consid-erable care must be taken to ensurethat all of the cell elements are illumi-nated since an unilluminated unit willact Iike a large series resistance.

Like the interdigitated cell discussedearlier, the vertical multifunction devices do not require contacts on the sur-face (the aluminum connections cancover less than 1 percent of the surfacearea)” and thus front surface reflec-tions are minimized.

The multifunction device should beable to make more efficient use of lightwith relatively long and short wave-lengths.

The multifunction devices should beable to perform better than conven-tional silicon cells at high tempera-tures, and its performance at high tem-peratures improves in high light con-centrations; the cells should, for exam-ple, be only half as sensitive to temper-ature at 1,000 x suns as they are in un-concentrated sunlight.45

41B, L. Sater and C. Goradia, “The High IntensitySolar Cell –Key to Low Cost Photovoltaic Power, ”NASA Technical Memorandum ,NASA TM X-71718,1975.

42C Goradia and B. L. Sater, “A First Order Theoryof the P-N-N Edge-1 Illuminated Silicon Solar Cell atVery High Injection Levels,” IEEE Transactions onElectron Devices, Vol. ED-24, 4, p. 342,1977,

“H. ), Hovel, IBM, “The Use of Novel Materials andDevices for Solar Concentrating Systems,” Pro-ceedings DOE Photovoltaic Concentrator SystemsWorkshop, Scottsdale, Ariz., May 24-26,1977.

4 46. L. Sater, “Current Technology Status of theEdge-illuminated Vertical Multifunction (VM)) SolarCell, ” November 1977 (unpublished),

45 Goradia and Sater, op. cit., p. 350.

The extra manufacturing steps required tofabricate these devices will make themsomewhat more expensive than conven-tional silicon cells, but this difficulty wouldbe offset if the high efficiencies are realized.

Horizontal Multifunction Cells

There have also been proposals for usinghorizontal multifunction devices using sili-con or GaAs. It may be possible to design acell array capable of producing relativelyhigh voltages on a single chip, thereby re-ducing the cost of interconnecting devicesand reducing series resistance in connec-tions. The inherent efficiencies of thesedevices are approximately the same as con-ventional cells, but it may be easier to usethe approach to develop practical cellswhich can more nearly approximate thepotential of the materials.46

Thermophotovoltaic Cell

The “thermophotovoltaic” cells shown infigure X-5 may be able to achieve efficien-cies as high as 30 to 50 percent by making anend-run around the fundamental limits oncell performance discussed earlier. This isaccomplished by shifting the spectrum oflight reaching the cell to a range where mostof the photons are close to the minimum ex-citation threshold for silicon cells. The Sun’senergy is used to heat a thermal mass to1,800 ‘C (the effective black body tempera-ture of the Sun is about 5,7000 C). A largefraction of the surface area of this massradiates energy to a silicon photovoltaicdevice. (Reradiation to the environment canoccur only through the small aperture wherethe sunlight enters. ) A highly reflective sur-face behind the photovoltaic cell reflectsunabsorbed photons back to the radiatingmass and their energy is thus preserved inthe system .47

“H. J. Hovel, IBM, “The Use of Novel Materials andDevices for Solar Concentrating Systems,” op. cit., p.38

“R. M. Swanson and R. N. Bracewell, S i l iconPhotovo/taic Cells in Thermophotovo/taic Conversion,EPRI ER-478, February 1977.


Figure X-5.— Thermophotovoltaic Converter

CONCENTRATED SUNLIGHTFROM PARABOLOID

ThermophotovoltalcConverter

SOURCESwanson, R M and R N Bracewell (Stanford Umverslty) “Sll(conPhotovoltalc Cells In Therm ophotovoltalc Conversion” EPRI ER.478 ~1977)

Multicolor Cells

Another approach to achieving high cellefficiencies involves the use of a number ofdifferent materials to form cells optimallydesigned for different colors of light. Thereare two basic approaches: (1) the use ofselective mirrors to separate colors anddirect them to different cells, and (2) the useof a vertical stack of photovoltaic cells ar-ranged so that upper layers absorb only highenergy (i.e. short wavelength) photons, al-

lowing the remaining photons to reach lowercell levels,48 Little experimental work hasbeen done on these devices, but theoreticalanalysis has predicted that the light-filterdevices could achieve efficiencies as high as46 percent if two separate cells are used and52 percent if three cells are used. ” MuIti-Iayer cells using Ge, Si, GaAs, and othermaterials with two or more layers may alsobe able to achieve efficiencies above 40 per-cent. 50 An ingenious scheme for separatingcolors has been recently proposed whichuses a series of dyes capable of absorbingsunlight and reradiating the energy in a nar-row frequency band matched to the bandgap of each of a series of cell junctions. s’ 52A considerable amount of developmentwork will be required to design practicaldevices, however, and the ultimate cost offabricating the devices cannot be forecastwith any confidence.

48H. J Novel, IBM, “The Use of Novel Materials andDevices for Solar Concentrating Systems, ” op cit., p

3849N. S Alvi, et al., Arizona State University, “The

Potential for Increasing the Efficiency of PhotovoltalcSystems by Using Multiple Cell Concepts,” TwelfthIEEE-7976, OP cit., pp. 948-956

50] ] Loferski, Brown University, “Tandem Photo-voltalc Solar Cells and Increased Solar E nergy Conver-sion Efficiency, ” Twe/fth /EEE-1976, o p clt , p p957-961

“ A Goetzberger a n d W Breubel, App/. Phys., 14,1977, p 123.

52 The technique was originally suggested for use inconnection with scintil Iation counters in R L Carwln,“The Design of Liquid Scintlllatlon Cells, ” Rev. Sci, in-st. 23, 755 (1 952); and R L Garwin, “The CO I Iectlon ofLight From Scintillation Counters, ” Rev. SCI. Inst. 37,1010 (1 960).

CONCENTRATOR SYSTEMS

The contr ibut ion of photovolta ic cel lcosts to the overall cost of an installed pho-tovoltaic system can be greatly reduced ifan optical system is used to concentratesunlight on the cell, even though cells de-signed for use in concentrators may costmore per unit area of cell surface than flat-plate cells. If such systems are used, prob-lems of reducing cell fabrication costs are

replaced with problems of mechanical en-gineering. In most cases, the energy requiredto manufacture a concentrator array ismany times lower than the energy requiredto manufacture a flat-plate cell array with asimiIar area.

The variety of concentrating collectorswhich can be used with photovoltaic de-


vices is reviewed in chapter XII. PhotovoI-taic devices can be attached to most typesof tracking systems with minimal modifica-tions to the basic collector design. (A two-axis tracking system using Fresnel lenses tofocus light on silicon cells is shown in figureX-6. ) Attaching a photovoltaic device to aconcentrating collector can, however, pre-sent some unique design problems:

● One-axis tracking collectors are unableto illuminate the entire receiver surfaceduring most of the day. Thus, only apart of the receiver can actually becovered by cel ls. In general , i t isnecessary to connect photovoltaic de-vices in a receiver in series to achieveacceptable system voltages. Nonillumi-nated photovoltaic devices have high

Figure X-6.—Experimental Concentrating Photovoltaic Array in Operation.The Array Employs Fresnel Lenses and Silicon Photovoltaic Cells

SOURCE Sandia Laboratories, Albuquerque, N Mex., 1976.

Ch. X Energy Conversion With Photovoltaics . 405

resistance and the output of a string ofcells connected in series would begreatly reduced if one element of thestring were shaded.

● It is desirable to maintain a relativelyuniform illumination on most cells tomaximize cell performance. This is dif-ficult to accomplish with most collec-tor designs. It is possible to design cellscapable of performing with acceptableefficiencies in the i Illumination patternsof specific CoIIector designs. This canbe done, for example, by modifying thepattern of electric contacts on the frontsurface of the cell, but the market forsuch specialized cells would necessari-ly be limited.

It is d i f f i cu l t to compare the a t t rac -tiveness of concentrating and nonconcen-trating photovoltaic devices because of thelarge number of variables involved; the onlycompletely satisfactory way of making sucha comparison is to conduct a complete Iife-cycle cost analysis of competing systemsoperating in realistic environments (such asthose reported elsewhere in this study). Thefollowing formula, however, can be used toobtain a crude estimate of the cost of a con-centrating collector which would be com-petitive with a flat-plate device (assumingthat no credit can be given for the thermalenergy which can be produced from the con-centration systems):

where the variables are defined as follows:

c c = allowed cost of the concentratin g

col lector, excluding the solar cel ls($/m 2)

C fP = cost of fIat plate cell array ($/kW)n fP = the efficiency of the f tat-plate arrayr e = ratio of the efficiency of the concen-

trator photovoltaic system (includin g

losses in the collector optics) to the effi-ciency of the flat-plate array (includingpacking factors)

rs = ratio of solar energy reaching cells inthe tracking collector to the sunlight(direct and diffuse) reaching the flat-plate collector

r C = ratio of the cost of the concentrating

cell ($/kW in one sun) to the cost of afIat-plate cell array

C i = cost of installing the flat-plate collec-tors ($/m2)

r l = ratio of cost of installing a concen-trator to the cost of installing a flat-plate device

C o=annual cost of maintaining the flat-plate system (cleaning etc.) in $/m2 o fcolIector

rO = the ratio of the cost of maintainingthe concentrating collector to the costof maintaining a flat-plate system (perm 2)

k l= the effective cost of capitalcm ––the cost of f lat -plate support ing

structures ($/m2)C,= concentration ratio of collector

This formula has been used to construct thecurves shown in figure x-7.

Figure X-7.— Breakeven Costs for ConcentratingPhotovoltaic Collectors Compared With Flat. Plate

Devices

0.10 0.20 0.30 0.40

EFFICIENCY OF CELLS USED IN THECONCENTRATOR

(not including optical losses)


Assuming that concentrating systems costso percent more than flat-plate arrays to in-stall and three times as much to operate onan annual basis , concentrator systemswould be competi t ive with $500/kW—15percent efficient flat-plate arrays if the con-centrating collector costs about $70/m2 a n dconcentrating cells are 20-percent efficient.The concentrator could cost about $180/m2

if concentrating cells are 40-percent effi-cient. (This example assumes that the op-tical efficiency of the concentrators is 80percent and the concentration ratio is atleast 10 times the ratio between concen-trator cell costs and the cost of fIat-plate ar-rays. ) The allowed cost of concentrators willbe considerably larger if there is a useful ap-plication for the thermal energy availablefrom an active cooling system.

PHOTOVOLTAIC

The analysis thus far has considered onlythe electric output of collector systems, butthe attractiveness of photovoltaic devicescan be increased significantly if effectiveuse can be made of the thermal energy car-ried away by water pumped over the backsurfaces of collecting celIs. Such systemsare the photovoltaic analogs of cogenera-tion devices, and an analysis of the op-portunities presented by such devices ismuch the same as those conducted for con-ventional systems. I n both cases, the net ef-ficiency of systems can only be understoodby examining the combined demands forelectricity and thermal energy in each pro-posed situation. There are clearly manyuseful applications for the thermal energywith temperatures in the 500 to 100‘C rangewhich can easily be extracted from mostphotovoltaic cogeneration devices. About23 to 28 percent of the primary energy con-sumed in the United States is used at tem-peratures below 108 oC .53

53 Battelle Columbus Laboratories, Survey of the Ap-plications of Solar Thermal Energy Systems to /n-dustrial Process Heat.

While the cost of mounting cells on track-ing collectors is clearly a major concern,there is also cause for concern about thecost of mounting flat-plate arrays. If arrayprices fall below $300 to $500/kW, the costof supporting and installing the arrays couldbegin to exceed the cost of the arraysthemselves.

General Electric has proposed a designfor a photovoltaic shingle which may beable to substitute for a building roofingmaterial. If costs reach $100 to $300/kW, itmay also be economical to use photovoltaicsheets as a part of a wall surface. I n mostlocations in the United States, the output ofa collector mounted vertically on a south-,east-, or west-facing walI is 40 to 60 percentlower than the output of a collector fixed atan optimum orientation.

COG EN ERATION

With photovoitaic systems, the criticalquestion is whether the electric-generatingefficiency which is lost because of operatingthe cells at higher temperatures is compen-sated by the value of the thermal energyproduced. C e l l p e r f o r m a n c e d e g r a d e salmost Iinearly with temperature at tem-peratures of interest primarily because of adrop in the operating voltage of the cell (athigh temperatures, thermally excited elec-trons begin to dominate the electrical prop-erties of the semiconductor device). Thistemperature dependence is commonly ex-pressed in the following form:

where n (T) is the efficiency of a cell at tem-perature T (expressed in “C) and ß is the tem-perature coefficient. The temperature coef-ficient (ß) measured for a number of dif-ferent cells is illustrated in table X-3. It canbe shown in most cases that if a use for low-temperature thermal energy exists, it is pref-erable to accept these losses of efficiencyand use the thermal output from celIs direct-Iy rather than to maximize cell performance


Table X-3.—Temperature Dependence of Photovoltaic Devices

Cell Base Concentrate ion ß= temper-efficiency resistivity ratio ature coef -

Material at 28° C(% ) C r ficient

Silicon . . . . . . .,Silicon . . . . .Silicon ., . . . . . . . .Silicon . . . . . . . . .Silicon . . . . . . . . .GaAs . . . . . . . . . .GaAs . . . . . . . . . .GaAs . . . . . ., . . . .CdS ., . . . . . . . .

10.411.812.412.211.817.019.218.57.8

0.10.30.32.0

10.0

11

40111

1001000

1

0.00350.00350.00320.00400.00460.00220.00230.0021

0.004-0.005—

SOURCES Silicon cell data from Edward Burgess, Sandla Laboratories, private communication.GaAs data from L James, Varlan private communicationCdS data from John Meakln, Institute of Energy Conversion, Unlverslty of Delaware (private communlcat{on)

and attempt to use a photovoltaic-poweredheat pump to produce thermal energy,

GaAs and other high-efficiency cells areless affected by high temperature operationthan are silicon devices .54 A commercialsilicon cell operating with an efficiency of12 percent at 270 C has an efficiency of only8 percent if operated at 1000 oC 55, while aGaAs cell with an efficiency of 18.5 percentat 280 C can operate with 16-percent effi-

54H j Hovel, IBM, Semiconductors and Sernirneta/s,op cit , p 168

‘SE L Burgess and J C Fossum, Sand iaLaboratories, Albuquerque, “Performance of n + -pSillcon Solar Cells in Concentrated Sunlight, ” /EEETransactions on E/ectron Devices, E D-24(4), 1977, p.436

Sandia cellSandia cellSandia cellCommercial cellCommercial cellVarianVarianVarianUniv. of Delaware

ciency at 1000 C and about 12-percent effi-ciency at 2000 oC.56

It is possible to operate flat-plate collec-tors at elevated temperatures, but cogenera-tion will probably be easier to justify forconcentrating systems. Care must be takento cool concentrator cells even if waste heatis not employed. If photovoltaic cogenera-tion proves to be attractive, concentratorphotovoltaic systems may continue to beeconomically attractive even if the pricegoals for fIat-plate arrays are achieved.

56L. James, Varian Corp , private communication,1977,

THE CREDIBILITY OF THE COST GOALS

The most satisfactory technique for an-ticipating the future cost of photovottaic de-vices wouId be to anticipate future manu-facturing techniques and develop a precisecost estimate for each processing step; thisapproach is taken in the next section. Un-fortunately, the results of such analyses areinconclusive since many future manufactur-

ing processes capable of dramatic costreductions are based on techniques whichare now only laboratory procedures orwhich anticipate progress in research. It ispossible, however, to make some estimatesof potential cost reduction by examining thehistory of cost reductions achieved in sim-iIar types of manufacturing.

408 Solar Technology to Today’s Energy Needs

A c o m m o n s t a t i s t i c a l t e c h n i q u e f o ranalyzing the history of cost reductions iscalled the “learning-curve” technique. 57

This technique attempts to correlate theprice of a product with the cumulative pro-duction experience of the industry manufac-turing the product. In many cases, a roughcorrelation appears to exist, although therate at which prices decrease varies greatlyfrom one industry to another. The rate of“learning” is quantified by determining howmuch the price decreases when the cumula-tive production volume doubles: if a doubl-ing of accumulated volume results in a pricedecrease of 10 percent, the system is said tobe on a “90-percent learning-curve;” if adoubling of production volume results in aprice decrease of 30 percent, the system issaid to be on a “70-percent learn i rig-c urve.”

The most obvious place to look for ahistorical analogy for predicting the pricereductions possible in photovoltaic devicesis the semiconductor industry which pro-duces silicon devices using many of thetechniques now used to construct siliconphotovoltaic devices. The history of pricesin the transistor and silicon diode industriesis illustrated in figure X-8. The results are dif-ficult to interpret since the price reductionsclearly do not follow a simple linear learn-ing curve, but they do indicate that a learn-ing curve of 70 percent is not impossible.The analogy is far from perfect, of course,since miniaturizat ion techniques whichwere used effectively to reduce the price ofsemiconductor electronics cannot be usedin the manufacture of photovoltaic devices;the size of photovoltaic devices used in flat-plate arrays cannot be reduced since powerproduced by a unit area of photovoltaic sur-face is limited by the intensity of sunlight onthe Earth’s surface. The learning-curvemethod is probably valid for making crudeestimates of future cell prices since learningcurves of 70 to 90 percent have been ob-served for most products, even when minia-turization is not used in the manufacturingprocess.

“Boston Consulting Group, Perspectives on Ex-perience, 1971.

MARKETS

The learning curve cannot be used to esti-mate the rate at which pr ices wi l l fa l lwithout information about the size of themarket at each future price. Unlike mostother power sources, the demand for photo-voltaic devices exists at a large range ofprices, since the equipment can providepower in remote areas where conventionalalternatives are extremely expensive. Theunique features of the equipment may leadto the discovery of markets for energy whereno market now exists. The large elasticity ofmarkets, coupled with the fact that individ-ual installations can be very small, has al-lowed an evolutionary growth in sales and agradual reduction in cell prices,

The free-world market for cells at 1976prices was about 380 kW58 of which about280 kW were soId by U.S. manufacturers.U.S. Government purchases during thisperiod were about 108 kW, of which about50 kW were used in satellites.59 Major com-mercial markets have appeared in commun-ications equipment (68 kW), corrosion pro-tection for bridges, pipelines and like ap-plications (28 kW), and aids to navigation(20 kW).6O Sales during 1977 were expectedto be about twice 1976 levels.

Several market surveys for photovoltaicequipment have been completed during thepast several years, and some of these aresummarized in table X-4. The considerabledifferences in the forecasts reflect differingjudgments about the future cost of conven-t ional energy and other forms of solarenergy, about the rate at which an industrialinfrastructure capable of supporting Iarge-scale production can be established, andabout the potential costs of support equip-ment (storage, controls, etc. ) and instalIa-

58BDM Corporation, Characterization of the PresentWorldwide Photovoltaics Power Systems Market, May1977, p. 1-1.

S9 I ntertechnology Corporation, Photo vo/ta ;C power

Systems Market /identification and Analysis, Aug. 23,1977.

‘“lntertechnology Corporation, op. cit.


Figure X-8.— Boston Consulting Groups’ Learning Curve

SILICON TRANSISTORS

1954 1960 1963 1968

$100.00

0.1 1 10 100

Industry Total Accumulated Volume(million units)

1955

SILICON DIODES

1959 1965 1968

3 10 100 1,000 10,000

Industry Total Accumulated Experience(million units)


Table X-4.– Market Forecasts for Photovoltaic Devices at Different Prices[in megawatts of annual sales]

—Array prices in dollars per peak watt

Marketing study 10 3 1 0.5 0.1-0.3

1. BDM/FEADOD market . . . . . . 10Worldwide commer-cial market . . . . . . 1.5

2. Intertechnology Cor-poration . . . . . ., . . 0.5

3. Motorola . . . . . . . . . . . . . 1.5

4. Texas Instruments. . . . . 0.4

5. RCA. . . . . . . . . . . . . 0.8

6. Westinghouse . . . . —

75 100 — —

20 70 100 —

13 126 270 —

20-30 — — —

2.6 30 100 20,000

13 200 2,000 100,000— — 96,000

DOE planning, objectives . . 1.0 8 75 500 5,000

SOURCES ERDA brteflng on the “Photovoltalcs Program, ” Sept 9, 1977.

BDM Corporation, “Characterlstlcs of the Present Worldwide Photovoltalc Power Systems Market, ” May 1977, p1.1

Irrtertechnology Corporation, “Photovoltaic Power Systems Market Identlflcatlon and Analyses, ’ Aug 23, 1977

BDM Corporation, DOD Phofovo/tam Energy Conversion Systems Market /nventory and Ana/ys/s, SummaryVolume, p 17.

P D Maycock and G F Wakefield (Texas Instruments) “Business Analysis of Solar Photovoltatc Energy Conver-sion, ” The Conference Record of the E/evenfh /EEE P/rotovo/fam Specla//sfs Conference. 1975 Scottsdale, ArlzMay 1975, pp 252.255

The Motorola Corporation

tion. Some manufacturers, for example,have been skeptical about the high forecastsfor small remote applications since many ofthese applications require a considerableamount of expensive marketing and engi-neering.

The surveys seem to agree that a signifi-cant fraction of sales during the next fewyears will occur in developing countries.Photovoltaic equipment is ideally suited toplaces where no utility grid is available andwhere labor for installing the equipment isrelatively inexpensive. Consumers in thecapital cities of many developing nationsnow pay as much as $0.20 to $0.25/kWh forelectricity, and prices in more remote areasare often higher (if power is available atalI). 61 The modular nature of photovoltaic

“Telegrams received from A I.D from posts innumerous developing nations during July and August1977 [n response to a request for information aboutlocal utlilty rates

equipment has the additional advantage ofallowing functioning power sources to be in-stalled quickly and in sizes appropriate foreach application. Moreover, an investmentin the photovoltaic power source does notcommit a nation to finding a reliable sourceof fuel or to maintaining a highly trainedgroup of operators—two serious problemsfor developing countries,

Other possible areas where sales of photo-voltaic equipment may increase rapidly dur-ing the next few years include applicationsby the U.S. Department of Defense (large,cost-effective purchases appear possibleduring the next few years)62 and the armedforces of other nations; the agricultural sec-tor, for irrigation and other pumping ap-

‘2BDM Corporation, DOD Photovoltalc Energy Cor--version S ysterns Market /nventory and A na/ys IS, 5um-mary Volume, p. 17


placations; and the transportation industry,for highway markers and lighting. 63 64

If prices fall below about $0.50/watt, anexplosive growth in sales could occur sinceat this price photovoltaic equipment mightprovide electricity which is competitive withresidential and commercial electricity ratesin many parts of the United States. 65 66 B ythe time prices fall to $0.10 to $0.30/kW, thephotovoltaic electricity may be competitivewith electricity sold at buIk rates to large in-dustrial consumers. Estimating sales at theselevels is extremely speculative since gener-ating large amounts of power from photo-voltaic devices would require a fundamen-tal change in the ways in which the Nationnow supplies and consumes electric energy.Moreover, when array prices reach these lowlevels, the overall cost and attractiveness ofphotovoltaic systems are likely to be domi-nated by factors other than the cost of thecelIs themselves. Before turning to an anal-ysis of integrated systems, however, it willbe useful to examine the costs and capabili-ties of the assortment of photovoltaic de-vices which are or may be available in thenear future.

Each of the projections of markets can beused to predict the rate at which prices fallgiven an assumed rate of “learning.” Threeforecasts and three different learning curveshave been used to construct the forecastsshown in figure X-9. It can be seen that withthe RCA estimate of markets and a 70-per-cent learning curve, the technique predictsthat prices will reach $500/kW in 1986. Ifprices fall according to a 75-percent learn-ing curve, however, the price will not reach$500/kW until after 1990. Using the less op-timistic Texas Instruments (Tl) estimates ofmarkets, prices would not fall to $1,000/kWuntiI nearly 1990, even with the 70-percentlearning curve,

‘‘Intertechnology Corporation, op. cit.64BDM Corporation, Photovo/taic Market, op cit.“Westinghouse Corporation, Fina/ Report: Concep-

tual Design and Sys tems Ana/ysis of Photovo/taicPower Systems, Vo/ume /, Executive Summary, April1977, p 45

“See volume I I of this study

Figure X-9. -Photovoltaic Array Prices forSeveral Market Projections

NOTE: With the exceptfon Of the “DOD sales” curves, It has been assumedthat the Federal Government spends $129 ml Illon for photovoltalc systemsbetween 1978 and 1981, of which 40 percent IS actually used to purchasecell arrays (the remainder being spent for des Ign, i nstallatlon, storage, con-trol systems, and other supporting devices) The forecasts assume thatcumulatwe production of cel Is through 1975 was 500 kW and that averagearray prices at the end of 1975 were $15/Watt

The figure also illustrates the extreme sen-sitivity of the forecast of potential price re-ductions to the assumptions made about in-termediate markets. If the BDM Corporationforecasts of potential cost-effective militaryapplications are correct and the militar y

purchases devices for all cost-effective ap-plications, prices can fall to $500/kW by1986, even if a 75-percent learning curve i s

assumed.

If the cost of photovoltaic devices is to bereduced through an expansion of the mar-ket, the industry will have to grow extremely

rapidly for the cost goals to be met. For ex-ample, if prices follow a 70-percent learningcurve, the cost goals can be met if the pro-


duction rate of cells doubles each year. (Thetimes required to achieve different price re-ductions, given a learning curve and produc-tion doubling time, are summarized in tableX-5.) Such expansions are possible but theycould make it difficult for small companiesnow manufacturing cells to expand fastenough to meet demands.

[n spite of its limitations, the learning-curve technique can provide some usefulguidance by establishing whether the costgoals anticipated for photovoltaic devicesexceed any historic rates (the goals are op-timistic, but not impossible by this test), andthey can be used to explore the sensitivity ofprice reductions to different forecasts ofpotential markets.

AN ANALYSIS OFMANUFACTURING COSTS

Silicon

The vast majority of the photovoltaiccelIs now being sold are single-crystal siIiconcells in fIat arrays. (See figure X-1 O.) The bulkof Federal funding to reduce the cost ofcells is being directed to silicon technology.The Federal Low-Cost Silicon Solar Array(LSSA) project, managed by the Jet Propul-sion Laboratory (jPL) has made a carefulanalysis of the component costs of eachstep in the manufacturing process and is

Table X-5.— Years To Achieve Indicated Price Reduction as a Function ofLearning Curve and Assumed Growth in Demand

——Learning Number of times production rates

Price reduction curve double each year

(PO/ P) (percent) 4 2 1 1/2

70 3.0 5.5 10.0 18.030 75 3.6 6.6 12.3 22.6

80 4.4 8.4 15.7 29.4

70 3.4 6.2 11.4 20.950 75 4.0 7.5 14.1 26.1

80 5.0 9.5 18.0 34.0

70 3.8 7.2 13.4 24.7100 75 4.6 8.7 16.5 30.9

80 5.8 11.0 21.1 40.2

70 4.1 7.8 14.5 27.0150 75 5.0 9.4 17.9 33.7

80 6.2 11.9 22.9 43.8

Assumptions: Price in 1976 = PO = $15,000/kWea

Cumulative sales volume in 1976 = So = 500 kweb

Annual sales in 1976 = B = 250 (kWe)e

= to = 1976 (base date)

aJ w, yerk~.s, (.ARc(_J solar, lnC ) Pr!vatf? Communtd(onbJ Llndmeyer, (Solarex Corp ) Private Comrnunlcatlon.C A Cllfford, (Solarex Corp ) Private Communlcatlon. Oct 1976SOURCE” Prepared by OTA

`Ch. X Energy Conversion With Photovoltaics ● 413

Figure X-10.— Commercial Silicon Photovoltaic Cells

Sil icon Concentrator Photovoltaic Cells Commercial Sil icon Photovoltaic Cells

Manufacturer:1. Solarex.2. Arco Solar (Printed contacts)3. Motorola.4, Optical Coatings Laboratory, Inc.

(Space Cell Utilized In Sat elites)

systematicalIy examining techniques forreducing costs in four major areas.

●

●

●

●

Production of the pure silicon feed-stock,

Preparation of a thin sheet of silicon,

Fabrication of cells, and

Arrangement of the cells in a weather-proof array.

JPL objectives for cost reductions in eacharea are shown in table X-6. I n reachingthese goals, maintaining high cell efficiencywilI be criticalIy important since many costsin cell manufacture and in the instalIationof photovoltaic arrays are proportional tooveralI celI area and not to power.

SILICON PURIFICATION

The purified polycrystalline silicon usedas the raw mater ia l of commercial cel lmanufacture now costs about $65/kg. T h ejet Propulsion Laboratory estimated that, ifthe goal of $0.50/watt is to be reached, thesilicon material cost must be reduced to

Scale in Centimeters Photos John Furber, OTA

Manufacturer:1-4. Spectrolab.

5. Sandia Laboratories,6. Solarex.

about $1 O/kg and the amount of siIiconwasted in the manufacturing process con-siderably reduced. b’ Perhaps even more im-portantly, current techniques for manufac-turing silicon are extremely inefficient intheir use of energy; approximately 7,000kWh of energy is required to manufacture acell with a peak output of 1kW (assumingthat celIs are 100 microns thick and 82 per-cent of the silicon entering the manufactur-ing process is wasted). 68 This means that thedevice must operate in an average climatefor about 4 years before it produces as muchenergy as was consumed in manufacturingthe component silicon.

Several promising techniques for improv-ing the purification have been experimen-tally verified, and it should be possible to

‘7H Macomber, j PL, Proceed ings , ERDA Serni-a nnua / Solar Photo vo/ta ic Program Review Meeting,Silicon Technology Programs Branch, San Diego,Calif , January 1977, p 68.

“L P Hunt, Dow Corning Corp , “Total Energy UseIn the Production of S i I icon Solar Cel Is From RawMaterials to Finished Product, ” Twe/fth IEEE-1976, op.cit `, pp 347-352


Table X-6.—The Distribution of CoStS in the Manufacture of silicon Photovoltaic DevicesUsing Contemporary Technology [aii costs in $/peak W]

Ingot technologyNon ingot

technology

Representativedata for 1976 1978 1980 1982 1984 1986

Polysilicon . . . . . . . . . 2.01 1.16 .76 .28 .14 .06Crystal growth andcutting . . . . . 4.18 2.50 1.43 .67 .23 .10

Cell fabrication . . . . . . . 5.97 1.87 1.01 .58 .34 .19Encapsulationmaterials . . . . .78 .22 .12 .09 .07 .03

Module assembly andencapsulating . 3.99 1.25 .68 .38 .22 .12

Price . 16.96Goal . . . . . . . . . . 20.00 7,00 4.00 2.00 1.00 50

——‘Does not Include Inflation— prices In 1975 constant dollars

NOTE The detailed price goal allocations are for slllcon Ingot technology up through 1982 and for slllcon sheet Technology In 1984 dnd 1986 Theseal local Ions for use w It h In the LSSA Pro)ec I are subject to rev IS Ion ds better knowledge Is gal ned and s ho[) Id not be ( onstr ued aspredicted prices

I t Is ex pec ted t hal after 1982, Ingot and sheet lec h ndogy mod u Ies w I I I be cost corn pet I t I ve, w I t h the poss I btl I t y that the $ 50/W goal ~ an beachieved by e It her technology Technology developments and future product I on c OS! parameters w I I 1 be malor factors I n determ I n I nq t hc~ mostcost effective deslg ns

SOURCE Jet Propulslorl Laboratory Low Cost SIIIC on Solar Array Pro)e( t Recel~ed by OTA June 1977

develop cell arrays capable of producing allthe energy used in their manufacture in 3 to4 months. ” A significant amount of c h e m -ical engineering and process development isneeded, however, to demonstrate that theselaboratory experiments can be scaled-up bymany orders of magnitude to form the basisof a commercial faciltiy.

It is possible to produce photovoltaiccelIs with siIicon less pure than the “semi-conductor grade” material now used in cellmanufacture, but a careful analysis must bemade to determine whether the lower sili-con costs would compensate for the addi-tional system costs which would be incurredby the reduced cell efficiency which results.Manufacturing very high efficiency cells foruse in concentrators may even require sili-con which is of higher purity than the mate-riaI now used to produce most semicon-ductors.

Silicon costs probably represent the singlegreatest technical barrier to meeting JPL’scost goals for nonconcentrating arrays in theearly 1980’s. This is because construction of

b9Hunt, Ibid,

plants capable of manufacturing silicon inquantities large enough to achieve the re-quired cost reductions would need to beestablished very quickly— probably withinthe next year — and the investments requiredwill be large, compared to previous spend-ing in photovoltaic manufacturing New sili-con plants are Iikely to require more capitalinvestment per unit of cell production thanany other stage in the celI production proc-ess — between $20 miIIion and $40 miIIionfor a single plant. There is no incentive to in-vest in such equipment of this magnitudesolely for the purpose of selIing sit icon tothe semiconductor industry because mate-rial costs for these devices are already asmall part of the device cost. Silicon prices,therefore, are unlikely to fall by 1982 unlessthe Government takes some action.

Improved sawing techniques may be ableto cut silicon material requirements by two-thirds by producing thinner silicon cells andreducing the material lost as sawdust. 70 Im-

‘ (’A Kran, I HM, Proceeding, S yn?po~lun? on t h eM,] ter/a/s .$cience$ Aspect o f T h i n / //n? $y~ tern~ forSo/ar F rwrgy C o n \ wsion, Tuscon, Arlz , Mdy 20-22,1074


proving the techniques used to grow crystalscould also reduce losses. Development of aribbon or thin-film, crystal-growing processcould greatly reduce wastage. Developmentof an amorphous silicon cell with adequateperformance would dramaticalIy reduce sili-con requirements in cells since these cellswould probably require less than I percentof the silicon used in commercial cells, Sili-con requirements can also be greatly re-duced if concentrator devices are used.

FORMATION OF SILICON SHEETS

Growing pure silicon crystals and sawingthem into thin waters now represents about25 percent of the price of arrays. 7’ A numberof active programs exist for improving thebatch processes in which crystal ingots arecurrently being produced and sawed intowafers 72 techniques have also been de-signed for drawing single-crystal sheets orribbons directly from molten siIicon, but aconsiderable amount of engineering workmust be done before a commerciaI processif available. Development of an Ingot tech-nique adequate to meet the $1 to $2 perwatt cost goal appears to be assured, andimproved ingot techniques may even beadequate to meet the 1986 cost goal. Theproblems remainlng in this technology ap-pear to be largely ones of improving me-chanical designs, this is another area wherethe program could be accelerated by theGovernment Although crystal growingequipment IS relatively expensive, unsubsi-dlzed commercial Interest in this kind ofequipment In the next few years is Iikely tobe greater than commercial interest in ad-vanced silicon refinement processes. Re-search progress which makes it possible touse polycrystalIine or amorphous materialswouId substantialIy reduce the cost of thisstep In production

——.—— —

7‘“ Low-(’ost SI I IC on Sol at Array Pro]ect, ” Qudrter)y

Report-l, J Uly-september 1976, pp 1-4“jet propulylon Ldbordtory, Low-cost $//icon .$o/dr

Array F’ro/ect, Report-J, July- September 1976, pp4-57

7‘M~comber, ] P1 , op c It , p 71

Crystal Growing and S/icing

Ali of the silicon photovoltaic devicesnow sold are manufactured from waferssawn from single-crystal boules (2- to 3-inchdiameter cylinders of silicon). These wafersrepresented about 35 percent of the cost ofphotovoltaic arrays in 1976 (see table X-6)The crystals are commonly grown by dipping a seed crystal into a silicon melt in a

quartz crucible which is a few degreesabove silicon’s melting point and by slowlywithdrawing the growing cyrstal from themelt . The crucible and the crystal arecounter-rotated to grow a straight crystal ofuniform, circular cross-section The crystalis pulled from the crucible until most of themolten silicon has been withdrawn andremoved from the melt. The entire appa-ratus is then aIlowed to cool so that thecrystal can be taken out of the airtightchamber, As the remaining molten siliconsolidifies in the crucible, the crucible usual-ly breaks, adding about $50 to the cost ofthe boule This procedure is cal led the“Czochralskl ” (Cz) or “Teal-Little” method.The boule is sawed into thin wafers whichare sent to the next stage of the cell fabrica-tion process. Several commercial siliconcelIs manufactured using this technique areshown in figure X-10. A variety of new con-cepts have been proposed for reducing thecost of the crystal-growlng processes, someInvolve radicalIy new approaches where thintilms or ribbons are solifified directly fromthe molten silicon.

• I reproved sawing techniques might reduce the material lost in the sawin g

process; currently, nearly 50 percent ofthe crystal grown is lost as s i l i c o nsawdust. ” Techniques are being devel-oped which use muItiple saws wIth thinsaw blades, sawing wires, and other ad-vanced processes to decrease the mate-rial lost In sawing and increase thenumber of silicon wafers producedfrom a single crystal by producing thin-ner wafers. 74 75

“Texas Instruments, I nc , and L’drlan have F RDAcontrdcts to Improve sawl ng technlquej See R GForney, et al J PL, op clt

7’LSSA Quarter/v Report, op clt , pp 4-117, 4-115


The molten silicon in the quartz cruci-ble from which the crystal boule is with-drawn can be continuously replenished,leading to longer crystal draws and alower requirement for crucibles. (Thisprocess could also save a number ofprocessing steps.)

The silicon material lost in sawingcould be recycled for further use if it isnot contaminated in the cutting proc-ess.

Techniques can be used to increase thediameter of the crystals grown by usingthe standard Czochralski process. Crys-tals now in use are typically 3 inches indiameter, but crystals 10 inches indiameter have been grown in labora-t o r i e s . An experiment is now under-way to grow three 12-inch diametercrystals in one continuous heating cy-cle. 77 Choice of an optimum diameterwill depend on a detailed study of themanufacturing process. (Problems incutt ing the crystals increase withcrystal diameter. )

The rate of crystal growth can be in-creased if columnar grains can be toler-ated in cell wafers (research is requiredto determine the extent to which suchgrains can be tolerated or their effectsminimized).

Considerable energy and capital equip-ment could be saved if a process couldbe developed for growing a crystal in amold. The Crystal Systems Company iscurrently examining a concept in whicha seed crystal is placed in one end of aninsulated mold filled with molten sili-con. A temperature gradient is main-tained along the mold so that the crys-tal grows from the end with the seedcrystal to fiII the mold. 78 This techniqueis commonly used to grow metal crys-tals, but has not yet been successfully

“M. GoIdes (President of SunWind Ltd ), privatecommunlcatlon, Aug 16, 1976

“R G Forney, et al , j PL, op clt7 8R C Forney, et al , jPL, op cit.

adapted to the growth of semiconduc-tor-grade silicon crystals.

improvements in cu t t ing c i rcu la rwafers into the hexagons required forclose packing are possible through theuse of a Iaser-slicing technique beingdeveloped by Texas Instruments.

Table X-6 indicates that if current tech-niques are converted to mass production,silicon wafer blanks can be produced forabout $112/m2 ($0.80/watt if the cell is 14-percent efficient) if polycrystalline materialwere purchased for $35/kg. This could bereduced to approximately $78/m 2 if poly-crystal Iine material were purchased for$10/kg. A recent analysis conducted forDOE by Texas Instruments estimated that itwould be possible to produce “solar-grade”polycrystalline silicon f o r $ 1 0 / k g ( o r$214/kW at 14 percent efficiency),79 and forapproximately $5/kg if arrays are to bemanufactured for $1 ()()/k W. 80

Single Crystal Growth—Ribbons and Sheets

Research has been underway for at least15 years to develop a process by whichsingle-crystal silicon can be produced in theform of ribbons or sheets. Thin ribbonswould be appropriate for use in solar cellswithout the crystal-slicing steps required forboules. This would eliminate the costs in-volved in slicing the crystals and couldmake more efficient use of silicon. ThiswouId not only eliminate expensive oper-ating steps, but could reduce tr imminglosses when cutting round cells to hexagonalcells. The Department of Energy has beenfunding Mobil-Tyco, Solar Energy Corp.,IBM, Motorola, RCA, the Universi ty ofSouth Carolina, and Westinghouse to assessthe merits of a variety of processes for grow-ing single-crystal ribbons and sheets. 81

One technique for doing this is the “edge-defined, film-fed growth” process (EFG);

“ Ibid.‘“lblda’ E RDA, Semiannua/ Nationa/ Yo/ar Photovo/tage

Program Review Meeting, University of Maine, August1976


work began on this process about a decadeago. The EFG process utilizes a graphite dieon top of a crucible of molten silicon. A nar-row pool of molten silicon forms on the topof the die by capillary action. A thin ribbonof silicon is slowly withdrawn from thispool.82 Such a ribbon is shown in the growth

82B Chalmers, “ T h e Photovoltaic Ceneratlon o fE Iectrlcity,” Scientific American, November 1976, p34

process in figure X-11. Photovoltaic cellsmade from EFG ribbons have shown effi-ciencies as high as 10 percent .83 Intensivework is proceeding on the EFC process butseveral difficulties remain: 1 ) contain i nation

‘~K, Ravi, et al., 12th IEEE Photovoltaic SpecialistsConference, Baton Rouge, La., Nov. 15,1976

Figure X-1 1 .—Silicon Ribbon Being Grown by the “Edge-Defined Film. Fed Growth” Process

(Mobil Tyco Solar Energy Corp )

— —

478 Solar Technology to Today’s Energy Needs

of theresultssiIiconby the

ribbon by impurities from the diein ef f ic iencies lower than other

cells; 2) the graphite die is attackedmolten silicon; and 3) the process is

currently quite slow. Ribbons 2- to 21A-cmwide can now be grown at rates of 2 to 7 cmper minute. The objective is a growth rate ofapproximately 18 cm. per minute. a’ A recentanalysis by IBM indicated that a large-diameter Czochralski boule can grow siliconcrystal areas at a rate equivalent to 20 to 40simultaneously pulled ribbons .85

Other ribbon techniques under investiga-t ion include the IBM “Capi l lary Act ionShaping Technique” (CAST) which uses awetted die, and the RCA “inverted stepa-nov” technique which uses a nonwetted die.Both techniques share many of the prob-lems of the EFG approach. a’

Another process for manufacturing rib-bons is the web-dendrite crystal growthprocess. Unlike EFG, this process requires nodie, and the die contamination and die ero-sion problems are, therefore, avoided. Stripsof silicon 2 meters long, 0.15mm thick, and22mm wide have been grown at rates of 2 to3 cm per minute; the objective is a growthrate of 18 cm per minute.87 D i f f i c u l t i e swhich remain with this process include: 1)the fact that careful temperature control isnecessary, which probably precludes the si-multaneous growth of several ribbons fromone melt; and 2) relatively slow growth rates.

Polycrystalline Sheets

Work is also underway to develop low-cost techniques for producing sheets ofpolycrystalline silicon with crystal grainslarge enough in size and oriented properly

“lbld“ 5A Kran (1 BM), “Single-Sillcon Crystal Growth by

Czochralski and Ribbon Techniques–A Comparisonof C a pa bl I I t Ies, ” Proceedings, S yrnpos I urn on theMaterial Sciences Aspect of Thin Film Systems forSo/ar [nergy Conversion, Tuscon, Ariz , May 20-22,1974, Pp 422-430

“’’Low-Cost Silicon Solar Array Project, ” QuarterlyReport-1, july-September 1976, p 4-57

“lbld

to allow the production of cells with accept-able efficiencies. Some processes are yield-ing individual crystal grains 1 to 10 mm on aside, blurring the distinction between singleand polycrystalline devices. Perfection ofsuch processes could lead to more efficientuse of the silicon feedstock and replaces theexpensive task of growing a perfect crystalwith a casting, forming, or depositing proc-ess which may be much less expensive.Some of the methods proposed would cre-ate a thin layer of polycrystalline materialon a sol id backing, or substrate, thuseliminating the cost and material wasteassociated with slicing the silicon materialinto wafers. Concepts include:

1. Dipping a substrate into molten silicon,covering the substrate with a thinc o a t i n g o f polycrystalline siIicon.(Honeywell).

2. Continuously pulling a thin sheet of sili-

3

4

5

con from the surface of a pool ofmolten metal where silicon vapor is be-ing deposited. (General EIectric).

Depositing silicon from a silicon-halidegas directly onto a hot substrate. It maybe possible to combine the final stageof silicon purification (distillation) withthis deposition process (Rockwell inter-national and Southern Methodist Uni-versity).

Forming heated silicon material underpressure in rollers (University of Penn-sylvania).

Casting b l o c k s o f poiycrystalIinematerial and slicing the blocks (CrystalSystems, Salem, Mass.).

CELL FABRICATION

The process of converting silicon wafersinto an operating photovoltaic array with 1kW peak output involves a number of indi-vidual steps which now require many hoursof hand labor,88 adding about $6 to the price

‘8P D. Maycock and G F Wakefleld, Texas in-struments, “Business Analysls of Solar PhotovoltalcEnergy Conversion, ” Eleventh IEEE-7975, pp 252-255

-.—.


of a 1 watt cell. Processes include the crea-tion of the photovoltaic junction, the addi-tion of electrical contacts, and the applica-tion of anti-reflective coatings.89 Studies ofmass-production techniques conducted forJPL by Motorola, Texas Instruments, andRCA all indicate that this cost could bereduced to $0.30 to $0.90/watt90 usingknown mass-production apparatus, if plantscapable of producing about 5 to 50 MW an-nualIy were constructed. The costs asso-ciated with each step in the fabricatingprocess are shown in table X-7. Projectingfurther price reductions, however, requires afair amount of optimism. It can be seenfrom table X-4, however, that cell pricesmust fall to $1 to $2/watt before there willbe markets large enough to support severalcompeting fabricating plants of the size en-

visioned in the JPL studies. It is likely thatmanufacturers wilI show greater desire to in-vest in fabrication equipmentmore capital-intensive devices————

“’F RDA “ Photovol talcs Program “‘(’h! acomber, J []l., op clt , pp 71-72

than in therequired to

manufacture silicon wafers. This is becausesignificant cost reductions can b e a c -complished with plants much smaller than50 MW/year, and it is much less likely thatfabr icat ing plants would become obso-lete–even i f breakthroughs dramatical ly

reduce the cost of manufacturing siliconwafers.

The laborious and expensive processes formaking photovoltaic celIs are also extreme-ly inefficient in their use of energy. Siliconmaterial, for example, is melted when it ispurified, cooled, and shipped to the crystal-g r o w i ng facility where it must be meltedagain to grow a crystal , again cooled,heated at least once more during cell fabri-cation, and then cooled to ambient tem-peratures for the third time. No attempt ismade to recover any of the heat wastedwhen the cells are cooled. The efficiency ofthis process could clearly be improvedsignificantly to reduce both the cost of thecell and the time required for the cell to“pay-back” the energy used in its manufac-ture once it is installed.

Table X-7 . —The Major Steps in Manufacturing a Silicon Photovoltaic Device

—

Surface Junction Other Moduleprep format ion Metalization processes fabrication Totals

(alternate)—.— -.——

M o t o r o l a 0.282 0.080 0.082 0.042 0.385 0871R C A 0.003 Ion 0.044 0.101 0,045 0.114 0309

(Spin 0.081 )& PO CI3

to0.343

(Print 0.048)

Texas Instru -ments 0.035 Spin 0.048 Thik Film 0.050 0.093 Hermetic 0.272

(Ion O 092)0391

Nonhermetic 0135 to0.586

ASSumptionsI) cells would be made uslnq wafers Sliced from ingots

2) Based upon an extrapolation of present technology with reasonable development but wlthout technological breakthroughs

3) 50 MW/year productlon of a slngle deslgn.

4) 14 percent cell efficiency

‘ Does not Include cost of sllicln material nor wafers

SOURCE JPL L[, h CJst Slhr(,r) Sc~lar Array Prole t (r~t mred by OTA June 19771


MODULE ASSEMBLY AND ENCAPSULATION

Finally, the process of connecting cellstogether into arrays and encapsulating themto protect the cells from the weather mustbe reduced by a factor of 10 from currentcosts to about $0.12/watt .91 A variety oftechniques have been proposed and the costreduction seems feasible, but the exacttechnique which will be used is not yetclear.

Thin Films

The nonsi l icon thin-f i lm technologieswhich show the greatest potent ia l forreaching the cost goals early in the 1980’sare all based on the CdS/Cu2S heterojunc-tion cell. Much more work has been done onthe problem of reducing the cost of manu-facturing CdS/Cu 2S than on any other thinfilm.

Cadmium sulfide cells are produced com-mercially in the United States by SolarEnergy Systems (SES) of Newark, Del. Thesecells are produced in batch processes92 a n dhermetically sealed in glass. Their efficiencyis 3.2 percent with an array efficiency of 2.5percent. 93 SES is attempting to keep their ar-ray prices competitive with the market priceof silicon arrays. An eariy prototype cell isshown in figure X-12.

No technical barriers to scaling up thecurrent manufacturing procedures are fore-seen, but a significant amount of “processdevelopment” wil l be required to bringprices below $1,000/kW.94 Researchers atWestinghouse estimate that a factory basedon this process capable of manufacturingcells for $1 to $2/watt could be producingcommercial products within 2 years of adecision to initiate the project.95

“See table X-6‘2F A Shlrland, West inghouse Research Labs,

private communication, Mar. 26, 1976‘3S DiZlo (President, SE S) private comr:]unication,

Apr 7 1976.“ I b i d .95F. A Shirland, West inghouse Research Labs r

private communlcatlon, Mar. 26, 1976.

Figure X.12.—Thin-Film Cadmium Sulfide-CopperSulfide Solar Cells

PHOTO: Courtesy of the University of Delaware

Photon Power, Inc., of El Paso, Tex., isdeveloping a chemical spray process whichforms cadmium sulfide solar cell panels di-rectly on hot float-glass as it comes out ofthe glass factory.

Photon Power has completed a pilot plantin El Paso for testing the technique. If theprocess works as well in large- scale produc-tion as it has in laboratory tests, the facilitywill be expanded into a smalI manufacturingplant which, it is hoped, will be able to pro-duce arrays which can be sold for $2 to$5/watt by 1980.9’ Low costs are possiblebecause all of the processes in cell manufac-turing (cell growth, junction formation, ap-

9’G. Rodrick, (President, Photon Power, Inc.),private communication, March 1978


plying contacts, and encapsulation) involvespraying a series of chemical layers onto hotglass moving through the plant on a con-tinuous conveyor. Laboratory devices pro-duced with a similar process have yielded5.6-percent efficiencies, and cells with 8- tolo-percent efficiencies appear possible. ”Even higher efficiencies may result if theprom i sing resuIts of experiments mixing zincwith cadmium can be integrated into theprocess.

Photon Power is working with Libby-Owens-Ford (a part owner of the company)to design a process which can be attachedto a float-glass plant. A preliminary calcula-tion indicated that a plant costing about$140 million would be able to produce 2G We/ y e a r f o r $ 0 . 0 5 t o $ 0 . 1 5 / w a t t . 9 8The major technical challenge in increasingoutput will be finding a way to increase thespeed of the spray application process fromthe 2 cm per minute in the pilot plant byabout an order of magnitude to match therate at which glass is produced from a flat-glass facility.99 It will also be necessary to

“G A Samara (Sandia Laboratories, Albuquerque),Proceedings, ERDA Semiannual Photovoltaic Advanc-ed Ma reria /s Program Review A4eetlng, Wash I ngton,D C , Mar 22-23, 1977

“j F jordan, P h o t o n P o w e r , Inc , InternationalConference on Solar Electricity, Toulouse, France,April 1976

“j F Jordan, Photon Power , Inc , private com-munlcatlon, April 1976

determine how much of the cadmium notinitially captured on the glass can be recy-cled (this is important since cadmium is ex-pensive and toxic) and to determine howwell the hydrogen chloride, sulphur, andsulfurous acids can be recycled.

The French National petroleum company,Compagnie Francaise des Petroles, ownscontrolling interest in Photon Power. Libby-Owens-Ford and D. H. Baldwin Co., ownminority interests.

Patcentre International of Hartfordshire,England is working on a spray process verysimilar to Photon Power’s. They report S. S-percent efficiency for their laboratory de-vices, and hope to market a commercialproduct in 1980. ’00

A number of other cells, usually based onheterojunctions between an element ingroup III and group V of the periodic table,are being investigated as candidate thin filmcells, but few are beyond the stage of basiclaboratory research. The present perform-ance of some of these experimental devicesis shown in table X-2.

IOOG Kaplan, IEEE Spectrum, “Power/Energy Prog-ress Report, ” IEEE Spectrum, Vol 15, No 1, january1978, p 51

MATERIALS AVAILABILITY AND TOXICITY

Enthusiasm about cells based on materials other than silicon must be tempered tosome extent by uncertainties abou‘t thehealth hazards which they may present andabout the limits imposed on their produc-tion by U.S. and world supplies of compo-nent materials. Both cadmium and arsenic,used in GaAs cells, are toxic and while itmay be possible to reduce the hazards theypresent to manageable proportions andwhiIe both materials are already used exten-sively in commercial products and manufac-

turing, i t clearly wil l be necessary to ex-amine these hazards with some care beforerecommending an energy system whichwouId significantly increase the use of thesematerials near populated areas. Domesticsupplies of both cadmium and gallium willbe sufficient to supply annual productionrates in excess of several thousandmegawatts a year but production beyondthis level could tax domestic supplies and, inthe case of cadmium, known world reserveslimit the ultimate potential of CdS/Cu 2S


devices to about the level of current U.S.energy consumption.

SILICON

Silicon is nontoxic and in plentiful supply(about one atom in five in the Earth’s crust isa silicon atom), although current productionrates of purified silicon are not adequate tosupport a large solar industry. The manufac-ture of silicon devices with present tech-niques involves the use of a number ofhazardous chemicals (PH3, BCl3, H2S 2, HCI,H C N ) .101 Existing State and Federal lawsshould be sufficient to ensure that releasesof these materials into the air and water arekept to acceptable levels, although vigi-lance will be needed to ensure compliancewith these regulations. Meeting the stand-ards may add to the cost of the devices.

CADMIUM

Cadmium is a cumulative heavy-metalpoison with a half-life in the human body of10 to 25 years. Cadmium poisoning is be-lieved to lead to accelerated aging, in-creased risk of cancer, heart disease, lungdamage, bir th defects, and other prob-Iems. 102 Chronic exposure to airborne cad-mium can lead to emphysema and other res-piratory problems. 103 Cadmium, however, isused in many commercial and industrialproducts; it is used for corrosion protectionon bolts and screws, in paints, plastics, rub-ber tires, motor oil, fungicides, and sometypes of fertilizers.104 Increased use of cad-mium resulting from large-scale manufac-ture of CdS/Cu2S cells could result in in-

10’] G. Holmes, et al , “Environmental and Safetyimplications of Solar Technologies,” Proceedings,1977 Annual Meeting American Section of the /nterna-tiona/ Solar Energy Society, Orlando, Fla., june 1977,p. 28-7

‘O’N A Esmen, et al., University of Delaware, En-vironmental Aspects of Cadmium Sulfide Usage inSolar Energy Conversion– Part P– Toxicological andEnvi ronmenta l Health Considerations – A 13ib/iog-raphy, NSF/RANN, SE/C I 34872 /T R73/5, june 1, 1973,Section I I

‘OJHolmes, op. clt , p, 287.‘04 Esmen, op cit., section II & IV

creased release of cadmium from mining, re-fining, and manufacturing of the cell, and itwould increase the amount of cadmiumpresent in products located near populatedareas, thereby increasing the risk of ex-posure in the event of fires, accidents, or thedemolition of buildings.

More stringent standards may be neededboth in manufacturing processes involvingthe material and in permitting usage of thematerial in commercial products. At pres-ent, cadmium is not subject to any environ-mental controls, but under both the ToxicSubstance Control Act (P. L. 94-469) and theResource Conservation and Recovery Act(P. L. 94-580), EPA has jurisdiction over itsregulation which, if instituted, could includeboth solar and other household uses.

It is apparently possible to significantlyreduce the cadmium released in the manu-facturing process, and it appears that withproper encapsulation CdS cells can be usedin residential areas with minimal hazard. 10 5Exposure to the material would only occurduring fires, accidental breakage, or build-ing demolition. The acceptabiIity of such ex-posure can be judged better when meaning-ful standards have been developed.

U.S. production of cadmium in 1970could support the annual manufacture ofcell arrays with a peak output of 5,000 to20,000 MWe (assuming cells are lo-percentefficient). World production is about fivetimes greater than U.S. production. 106 T h ehigher figure applies to the spray process formanufacturing cells, which results in cellsabout 6 microns thick. Significant increasesin cadmium production may be achievable,however, if demand increases. IdentifiedU.S. reserves of cadmium are sufficient toproduce cells with an annual output equalto the current U.S. consumption of electrici-ty. Known world reserves are about fivetimes greater than U.S. reserves. ’07

‘05 Esmen, op. cit) ObH .j Hovel, IBM, S e m i c o n d u c t o r s a n d

Semimeta/s, op. cit., p. 219.107R A Heindl, Minera/ Facts and F’rob/ems, U.S.

Bureau of Mines, U S. G.P.O ,1970, p 515.

. . -— —


GALLIUM ARSENIDE

Undisassociated GaAs is harmful but ap-parently not highly toxic. A lethal dose foran average adult is about one-third kg (0.7lb). ’08 To ingest this much GaAs, a personwould have to eat the amount of CaAs inabout 20m2 (200 square feet) of fIat-plate ar-rays or the amount of GaAs in a field of con-centrating arrays covering about 3 acres,clearly a Herculean task. The coatings andprotective encapsulation placed on cellsshould be able to reduce any hazards asso-ciated with normal operation of GaAs de-vices to acceptable levels.

T h e m a j o r d a n g e r a r i s e s w h e n t h ematerial disassociates as a result of a fire orsome other accident, since many arseniccompounds are highly toxic and recognizedcarcinogens, 109 110 The AS 20 3 which wouldbe released in a fire could contaminate landand water near the fire site. Since the cellswouId most probably be used in connectionwith high concentrations of sunlight, caremust be taken to ensure that the materialsdo not vaporize and escape if a breakdownin the cell cooling system occurs. Theamounts of arsenic used in a concentrating

collector, however, wouId be extremely

1‘aBased on l-13j0 of 4,700 mg/kg body weight citedIn the Regtstry of Toxic Effects of C h e m i c a l Sub-s tances , 1975 Edltlon, U S Department of HEW,N IOSH

‘“’N I Sax, Dangeroqs P r o p e r t i e s of lndustrla/Materials, Fourth Edition, Van Nostrand Reinhold Co ,1975, pp 420-421

1‘“George L Waldbott, Hea/th E f f e c t s o f En-vlronmen?a/ Pollutants, C V. Mosby Co., 1973, pp.202-203

PERFORMANCE OF CELLSSince the output of cells varies as a func-

tion of temperature, it is necessary to com-pute cell temperatures in order to obtain anaccurate estimate of the output of cells in-stalled in various types of collectors. Thetemperature of the cells depends on the ef-fect ive heat-transfer coeff icient of thesystem used to provide cooling. In simpleflat-plate systems, the heat loss from the

small. A device with a concentration ratio of1,000 would, for example, use only about0.16 gm per m2 of collector area. This con-centration is 250 to 1,500 times smalIer thanthe concentration of A S 20 3 r e c o m m e n d e dby the U.S. Department of Agriculture forweed control. 111

Gallium supplies should not place a sig-nificant constraint on the use of GaAs de-vices. U.S. consumption of gallium in 1973was 8.5 metric tons per year, most of whichis impor ted . If this amount of GaAs wereused to produce 50-micron-thick cells usedin tracking collectors producing concentra-tion ratios of 1,000, the arrays would have apeak output of about 10 GWe at an efficien-cy of 20 percent. These arrays would havean average output of about 1 percent of theaverage output of all electric-generatingfacilities in the U.S. in 1977. 113 D o m e s t i cproduction of gallium could be increasedsubstantially if an attempt is made to ex-tract the gal l ium associated with coal ,aluminum, and zinc ores. A recent study in-dicated that the United States could pro-duce about 600 metric tons/year of galliumfrom these sources. ’ So the gallium re-source is ample. World resources-of galare about 1.1 x 105 metric tons.}‘5

‘ 1‘USDA Herblclde Manual, 1965‘“Y C M Yeh, j PL, private communication,

25, 19751 l ‘ H . j H o v e l , IBM, Semiconductors

Semimeta/s, op. cit , ~ 219

lium

Apr.

and

‘“R N. Anderson,” Available Supply of Gallium andArsenic, NASA-Langley, Contract Order No L42953A,1976

f‘‘H j Hovel, IBM Semiconductors and Semimeta/s,Op Clt , p 220

IN REAL ENVIRONMENTSfront (and possibly also the back) of the ar-rays provides adequate cooling. In concen-trating devices, it is usually necessary to pro-vide more sophisticated cooling systems.These systems can be passive [i. e., large,finned radiating surfaces attached to thecell), or active cooling can be provided bypumping liquids across the back of the cell.

—


PASSIVE COOLING

The heat-transfer capabilities of a varietyof passive heat exchangers have beenmeasured, although much more work mustbe done to obtain adequate data in thisarea. The results, shown in table X-8, are ex-pressed in terms of an “effective heat-transfer coefficient” ke where

k e=[rate of heat removal (kW/m2

of heat exchanger)]/(T – Ta) (x-3)

In this, ke varies with wind speed and thusdepends not only on the type of heat ex-changer attached to the cells, but also onthe air circulation possible in each installa-tion.

Table X-8.— Effective Convective Heat-TransferCoefficients [ke] for a Variety of Cooling

Systems

Device Wind velocity ke(kWjm’OC)

Air circulating passive-ly through metal panin back of cells . . . . 0 0.015

Same as above . . . . . . . . ‘ ‘average’ 0.025

Flat plate with finnedmetal heat-reject ionsurfaces . . . . . . . . ‘ ‘average’ 0.30

Active water coolingwith ‘ ‘simple flow’ . . — 1-3

Active water coolingwith “impingementcooling’ . . . . . . . . . — 3-1o

SOURCE Table prepared by OTA from data supplied by F TBartels, (Spectrolab, Inc ,) Mar 1976.

In order to obtain a relation between celltemperature T and insolation level I, anenergy balance is written

(x-4)

where a is the absorptivity of the encap-sulated cell and Ta is the ambient air tem-perature. The value of I is given by thefollowing relations for f tat-plates or concen-trating systems:

II

where:

I D = d i r e c t n o r m a l(kW/m2)

Ø i = angle between themal to the cell

solar intensity

Sun and the nor-

Equations (X-2, X-3, and X-4) can be com-bined to yield an expression for cell efficien-cy in terms of insolation level:

ACTIVE COOLING

If water is pumped through a heat ex-changer attached to the cells, the value ofK e can be made as high as 3 to 10 kW/m2 “C.With such high rates of heat removal, celltemperatures can bz maintained at 150

0 F(65 ‘C), even at concentration ratios near1,000. The removed thermal energy avail-able in the cooling fluid (QA) is given by

where I is defined as before. UL is the heatloss coefficient which differs with each col-lector design. This equation assumes theentire receiver area is covered with cells. T f

is the average temperature of the coolingfluid (a more sophisticated equation con-sidering cell coverage ratios less than unityand the difference in inlet fluid temperatureis developed in chapter VIII.

The cell temperature (T) is given by

and the cell efficiency (n) is given by


PHOTOCHEMICAL ENERGY CONVERSION

Photochemical energy conversion hasmany of the advantages of photovoltaicenergy conversion and, in addition, the ad-vantage that the chemicals may be storedfor later use. However, the technology is notas well developed as photovoltaics, andmuch research remains to be done. The nas-cent state of photochemical energy conver-sion is compounded by the fact that publicawareness and Government funding areboth low, and thus research has been for themost part confined to a few universities,

In 1972, Fujishima and Honda announcedthe discovery of a process by which lightenergy incident on an electrode suspendedin water could be used directly to convertthe water into hydrogen and oxygen withoutany noticeable degradation of the elec-t rode .116 The original Japanese work hasgenerated a considerable amount of interestand a number of other materials have beenexamined. 117 118

A system being studied by research teamsof the Allied Chemical Corporation andMassachusetts Institute of Technology is il-

‘“A Fu]ishlma andhK H o n d a , “ E l e c t r o c h e m i c a lPhotolysls of Water at a Semiconductor Electrode,”Nature 238, july 1972, pp 37-38

‘ 1‘K L Hardee and A J Bard, /, Electrochern Soc.,~~q, 215 (1977)

1‘EM Wrlghton and J. M. Bolts, “Characterlzatlonof Band Positions Photoelectrodes for the Photoelec-trolysls of Water, ” paper delivered to the AmericanChemical Society, San Francisco, Calif , September1976, and /, Phys, Chem.,/X), 2641, 1976.

‘‘qHydrogen Production From Solar Energy, SITREPEnergy R&D, Report No. 154, Navy Energy and NaturalResources R&D Off Ice, Feb 18,1977.

lustrated in figure X-13. In this system,hydrogen is produced at the plat inumcathode and oxygen is generated at ananode made from a semiconductor materialsuch as T1O3 or CaAs. Some work has alsobeen done on systems in which both elec-trodes are semiconductors. Other tech-niques for generating hydrogen have alsobeen demonstrated. Researchers at theCal i fornia Inst i tute of Technology havedemonstrated that hydrogen is releasedwhen light strikes a complex Rhondiumcompound. 20 Research at the University ofNorth Carolina at Chapel Hill has Ied to thedevelopment of a ruthenium compoundwhich can split water into hydrogen and ox-ygen when applied in a monolayer to a glasssurface and exposed to sunlight. ’z’ All ofthese processes now exhibit very low effi-ciencies, and long-term stability has notbeen demonstrated. Almost no work h a sbeen done to determine the potential costof such systems.

The photosynthet ic process used byplants to convert sunlight into storable fuelsources is clearly a photochemical systemwith great potential. Work is underway tobetter understand the basic chemistry of theprocess and, of course, there is increasing in-terest in the use of plant materials as a fuelsource. The use of biological materials as anenergy source is beyond the scope of thecurrent study.

1]OM Simon, “UNC Professor Patents Solar EnergyProcess,” C/rape/ Hill Week/y, june 1976.

‘ “ ’ ’Gray Discovers Rhodium C o m p o u n d a tCaltech, ” The California Tech, Oct. 7, IW7, pp. 1,4.

28-.LI. I2 O - 28

—


Figure X-13. — Photochemical Energy Conversion

Phoioelectrolysls

H 20

PlatIn um Photocatalyticelectrode

*Window

Light

f

SOURCEGoodenough, J B (Professor, Oxford Unlverslty, England) “The Opt Ions forUsing the Sun,” Techrrology Flev/ew (Ml T.) 79(l), OctoberlNovember 1976,pp 62.71

chapter x energy conversion with photovoltaics

Documents