proceedings of the fourth annual conference on effects of

NATIONAL AERONAUTI CS AND SPACE ADMINISTRATION

Technical Memorandum 33-491

Proceedings of the Fourth Annual Conferenceon Effects of Lithium Doping

on Silicon Solar Cells

Held at the Jet Propulsion LaboratoryPasadena, California

April 29, 1971

Edited by

P. A. Berman

(N72- 10052

thruN72- 10065Unclas

t08095

(NAS A-CR-122850) PROCEEDINGS OF THE FOURTHANNUAL CONFERENCE CN EFFECTS OF LITHIUMDOPING ON SILICON SOLAR CELLS P.A. Berman(Jet Propulsion Lab.) 15 Sep. 1971 118 pCSCL 10B G3/03

J_ _ _

JET PROPULSION LABORATORY

CALIFORNIA INSTITUTE OF TECHNOLOGY

PASADENA, CALIFORNIA

September 15, 1971

Reproduced by

NATIONAL TECHNICALINFORMATION SERVICE

Springfield, Va. 22151. j

N72-10052

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION

Technical Memorandum 33-491

Proceedings of the Fourth Annual Conferenceon Effects of Lithium Doping

on Silicon Solar Cells

Held at the Jet Propulsion LaboratoryPasadena, California

April 29, 1971

Edited by

P. A. Berman

JET PROPULSION LABORATORY

CALIFORNIA INSTITUTE OF TECHNOLOGY

PASADENA, CALIFORNIA

September 15, 1971

REPRODUCED BYNATIONAL TECHNICALINFORMATION SERVICE

U.S. DEPARTMENT OF COMMERCESPRINGFIELD, VA. 22161

Prepared Under Contract No. NAS 7-100National Aeronautics and Space Administration

Page intentionally left blank

PREFACE

The purpose of this conference was to provide a forum for an in-depthreview and discussion of the results of investigations being carried out by var-ious organizations under NASA/JPL sponsorship as part of the Solar Cell Re-search and Development Program. Participating organizations included cellmanufacturers and university and industrial research laboratories. Becauseof the relevance of this program to activities outside JPL, members of theaerospace industry involved in the space effort were invited to attend the con-ference in addition to representatives of NASA, JPL, and JPL contractororganizations.

JPL Technical Memorandum 33-491 Preceding page blank| iii

LIST OF ATTENDEES

ANSPAUGH, B.Jet Propulsion Laboratory, Pasadena, Calif.

ARNDT, R.COMSAT, Washington, D. C.

BRANDHORST, H. W.Lewis Research Center, Cleveland, Ohio.

BRIGGS, D. C.Philco Ford Corp., Palo Alto, Calif.

BRIGLIO, A.Jet Propulsion Laboratory, Pasadena, Calif.

BRUCKER, G.RCA Astro Electronics Division of RCA,Princeton, N. J.

CARTER, J. R.TRW Systems, Redondo Beach, Calif.

COHN, E. M.NASA Headquarters, Washington, D. C.

COLEMAN, W. J.North American Rockwell, Downey, Calif.

CURTIS, O.Northrup Corporate Laboratories, Hawthorne,Calif.

DeANGELIS, H.Air Force Cambridge Research Laboratories,Bedford, Mass.

DOWNING, R. G.TRW Systems, Redondo Beach, Calif.

DRESSELHAUS, M.Massachusetts Institute of Technology,Department of Electrical Engineering,Cambridge, Mass.

FAITH, T.RCA Astro Electronics Division of RCA,Princeton, N. J.

GIVEN, R.Lockheed Missiles and Space Co. , San Jose,Calif.

GOLDSMITH, J.Jet Propulsion Laboratory, Pasadena, Calif.

HARNE, W. E.The Boeing Co. , Seattle, Wash.

ILES, P. A.Centralab Semiconductor Division of GlobeUnion, Inc. , El Monte, Calif.

JOHNSON, E. S.University of Southern California, LosAngeles, Calif.

JOSLYN, D.Aerospace Corp., Los Angeles, Calif.

KIRKPATRICK, A.Ion Physics Corp., Burlington, Mass.

KRULL, W.Lockheed Georgia Co. , Marietta, Ga.

LEADON, R. E.Gulf Radiation Technology Division of GulfEnergy and Environmental Systems Co.San Diego, Calif.

LUFT, W.TRW Systems, Redondo Beach, Calif.

NABER, J. A.Gulf Radiation Technology Division of GulfEnergy and Environmental Systems Co.San Diego, Calif.

PATTERSON, R.Jet Propulsion Laboratory, Pasadena, Calif.

PASSENHEIM, B.Gulf Radiation Technology Division of GulfEnergy and Environmental Systems Co.San Diego, Calif.

PAYNE, P.Heliotek Division of Textron, Inc. , Sylmar,Calif.

PRICE, W. E.Jet Propulsion Laboratory, Pasadena, Calif.

RALPH, E. L.Heliotek Division of Textron, Inc. , Sylmar,Calif.

RAUSCHENBACH, H. S.TRW Systems

REYNARD, D. L.Philco- Ford Corp. , Palo Alto, Calif.

JPL Technical Memorandum 33-491iv

SARGENT, G. A.University of Kentucky, Department ofMetallurgical Engineering and MaterialsScience, Lexington, Ky.

SPRY, R.Wright Patterson Air Force Base, Ohio

SROUR, J.Northrup Corporate Laboratories, Hawthorne,Calif.

STANNARD, J. E.Naval Research Laboratory, Washington,D. C.

STATLER, B. L.Naval Research Laboratory, Washington,D. C.

STOFEL, E.Aerospace Corp., Los Angeles, Calif.

TSANG, J. C.Massachusetts Institute of Technology,Cambridge, Mass.

WEINGART, J.Jet Propulsion Laboratory, Pasadena, Calif.

JPL Technical Memorandum 33-491 v

FOREWORD

The results of the investigations carried out during the past year repre-sent the most significant achievements attained thus far with respect to thelithium-doped solar cell program. Lithium-doped solar cells fabricated fromoxygen-lean and oxygen- rich silicon have been obtained with average initialefficiencies of 11.9% at air mass zero and 28 C, as compared to state-of-the-art N/P cells fabricated from 10 0-cm silicon with average efficiencies of11. 3% under similar conditions. Improvements in cell-processing techniqueshave made possible the fabrication of large-area lithium-doped cells. Ex-cellent progress has been made in quantitative predictions of post-irradiationlithium-doped cell characteristics as a function of cell design by means ofcapacitance-voltage measurements, and this information has been used toachieve further improvements in cell design. Specifically, analysis of irra-diated lithium-doped cells has shown that the recovery characteristics can bevery well predicted by the lithium density gradient near the junction and thatvery good cell-to-cell reproducibility of lithium density gradient can be ob-tained with reduced lithium distribution time schedules. Low-flux real-timeirradiation tests have shown, for the first time, that lithium-doped solar cellscan exhibit power levels higher than those state-of-the-art N/P cells in spaceenvironments. This is very encouraging, since the cells studied in theselatter experiments were, of necessity, cells fabricated about one year ago.Since that time, significant improvements in lithium-doped solar cell designhave been achieved, and it is expected that more recent cells would present aneven greater advantage than observed here. Also, tests recently conductedwith 28-MeV electron irradiation indicate a very marked advantage of lithium-doped cells over state-of-the-art N/P cells for radiation which results incluster defects.

The advancements made during this year with respect to lithium-dopedcell development have clearly shown the power of the interdisciplinary ap-proach which was adopted throughout the life of the program. Problems suchas cell instability, low efficiency, poor process control, unpredictability ofrecovery characteristics, variations in recovery rate, cell size limitations,etc. , that at one time appeared to be all but insoluble, appear to be rapidlyfalling by the wayside. These problems could not have been solved withoutthe involvement of the multi-expert team represented by the organizationsparticipating at this conference. Each organization isolated the problems withrespect to its own area of expertise, and the resultant body of information wascoordinated and distributed to the other organizations. It is strongly felt thatsuch an approach is applicable to many other programs (e. g., application ofspace technology to the economic generation of terrestrial power by means ofsolar energy conversion) and should prove to be of immense value in the ma-jority of cases.

ACKNOW LEDGMENTS

The conference chairman would like to acknowledge the support of NASAand JPL with respect to the investigations reported here. He would also liketo express his particular appreciation to all the organizations and personnelwho have diligently and conscientiously performed the work described at thisconference, and who have contributed to this program with significant personalinvolvement.

JPL Technical Memorandum 33-491vi

CONTENTS

SOME MAJOR RESULTS OF THE FOURTH ANNUALCONFERENCE ON THE EFFECTS OF LITHIUM DOPINGON SILICON SOLAR CELLS ..............................P. A. Berman

LITHIUM DOPED SOLAR CELLS FOR SPACE USE ..............P. Payne and E. L. Ralph

DEVELOPMENT AND FABRICATION OF LITHIUM-DOPEDSOLAR CELLS ......................................P. A. Iles

AIR FORCE/ION PHYSICS HARDENED LITHIUM-DOPED SOLARCELL DEVELOPMENT. ...............................A. Kirkpatrick, et al.

DEPENDENCE OF DEFECT INTRODUCTION ON TEMPERATUREAND RESISTIVITY AND SOME LONG TERM ANNEALINGEFFECTS ..........................................G. J. Brucker

DENSITY AND FLUENCE DEPENDENCE OF LITHIUM CELLDAMAGE AND RECOVERY CHARACTERISTICS ................T. J. Faith Jr.

DEGRADATION AND RECOVERY MECHANISMS IN LITHIUMDOPED SOLAR CELLS ................................R. G. Downing and J. R. Carter Jr.

RADIATION TOLERANCE OF ALUMINUM-DOPED SILICON ........O. L. Curtis, Jr., and J. R. Srour

EXPERIMENTAL AND COMPUTER STUDIES OF THE RADIATIONEFFECTS IN SILICON SOLAR CELLS .................R. E. Leadon, J. A. Naber, and B. C. Passenheim

HALL EFFECT STUDY OF ELECTRON IRRADIATED Si(Li) ........J. Stannard

THE OBSERVATION OF STRUCTURAL DEFECTS IN NEUTRONIRRADIATED LITHIUM-DOPED SILICON SOLAR CELLS ..........G. A. Sargent

RADIATION DAMAGE AND ANNEALING OF LITHIUM-DOPEDSILICON SOLAR CELLS ................................R. L. Statler

LOW FLUX IRRADIATION OF LITHIUM-DOPED SOLAR CELLS .....D. L. Reynard(Paper presented at conference but not available for Proceedings)

LOW FLUX IRRADIATION OF LITHIUM-DOPED SOLAR CELLS.William Krull(Paper presented at conference but not available for Proceedings)

RADIATION DAMAGE ANNEALING KINETICS .................M. S. Dresselhaus

JPL Technical Memorandum 33-491

4

13 vJ

31

35

41

53

65

71

81

85

91

99 < /

99 ~

9"101

vii

SOME MAJOR RESULTS OF THE FOURTH ANNUAL CONFERENCE ON EFFECTS

SOME MAJOR RESULTS OF THE FOURTH ANNUAL CONFERENCE ON EFFECTSOF LITHIUM DOPING ON SILICON SOLAR CELLS

P. A. BermanConference Chairman

I. INTRODUCTION

The conference was held at the Jet PropulsionLaboratory in Pasadena, Calif. , on April 29, 1971.Attendees were members of the photovoltaic com-munity interested in the radiation hardening ofsolar cells and personnel involved in research onlithium-doped silicon cells sponsored by NASA/JPL. A large body of information was presented,and a summary of all the significant points is notfeasible. Some of the points that will be of interestto the majority of readers are summarized.

II. SUMMARY OF RESULTS

Lithium-doped silicon solar cells having di-mensions as large as 1Z cm 2 are now possible,due to significantly improved boron-diffusion tech-niques, which stress the cells far less than tech-niques used previously. A large increase was ob-served in the short-circuit current measured intungsten (long wavelength) light for cells that werefabricated using the improved diffusion techniquesas compared with previous cells, indicating a pre-servation of minority carrier diffusion length inthe base region of the former cells. Sintering (atabout 600°C) of the contacts of lithium-doped cellsfabricated from Lopex silicon resulted in large in-creases in maximum power (of between 1 and4mW), mostly due to an open-circuit voltage im-provement, over non-sintered cells. Efficienciesas high as 12. 8% were observed, with the averageefficiency being about 11.9 %. The sample issmall, but the efficiency is significantly higherthan the average efficiency of 10 2-cm N/P cells(state-of-the-art) which is about 11. 3 to 11. 5%.

Measurements of the lithium surface concen-tration in silicon as a function of lithium diffusiontime indicated the occurrence of a peak in surfaceconcentration. The occurrence of lower lithiumconcentrations for longer diffusion times is indi-cative of diffusion from a limited source. Thepeak occurred more rapidly and was of greatermagnitude as the diffusion temperature wasincreased.

The uniformity of solar cells fabricated withthe addition of a redistribution cycle was more dif-ficult to control than that of cells fabricated usinga single-cycle lithium distribution. Also, longerlithium distribution cycles gave rise to larger

cell-to-cell variations in lithium concentrationgradient and open-circuit voltage.

The use of ion implantation has provided addi-tional flexibility in lithium solar cell design. Ionimplantation was used to implant an N+ region atthe back face of the cell and to implant lithiumthrough the N+ layer. In this way, the amount oflithium introduced prior to lithium distribution(which is still accomplished by diffusion at 300 to400°C) can be precisely controlled, and the ratioof electrically active lithium after distribution tothe amount of lithium initially present in the cellcan be determined. The presence of the N + regionat the back surface was found to inhibit loss oflithium after lithium distribution by a factor of tenover the loss observed in samples without the N+region. Very good post-irradiation recovery hasbeen obtained on ion-implanted cells fabricated inthis manner, with recovered powers being greaterthan that of state-of-the-art N/P cells.

In highly lithium-doped float zone (low oxygencontent) silicon the carrier removal reached equi-librium shortly after irradiation. In lightlylithium-doped float-zone silicon and highly lithium-doped crucible silicon, carrier removal reachedequilibrium approximately one to two years afterirradiation. A close pair vacancy-interstitialmodel applies quite well to lithium-doped crucible-grown silicon, but not as well for lithium-dopedfloat-zone silicon.

For a given fluence, the lithium-doped solarcell short-circuit current appears to be a linearfunction of the logarithm of the lithium densitygradient as measured with the capacitance-voltage technique and a linear function of the log-arithm of minority carrier diffusion length.

The damage coefficient of lithium-containingcells immediately after irradiation appears to bea function of the fluence and exhibits a square rootdependence on the lithium density gradient. Timeto half-recovery is linearly dependent on the log-arithm of the lithium density gradient as is the de-pendence of the recovered open-circuit voltage.

Long lithium-diffusion times resulted in largevariations in lithium density gradients (differencesas large as a factor of 24) while shorter diffusiontimes yielded very small variations of lithium


density gradients (within a factor of 2). This indi-cates that one can tightly control post-irradiationrecovery rates by proper control of the lithiumdistribution schedules.

Investigation of cells with various amounts oflithium coverage on the back prior to distribution(50, 80, and 100% coverage) indicated that the re-covery of the 80 and 100% coverage cells weresimilar to one another while the 50%-coveragecells recovered more slowly and less completely.The rate and extent of recovery was directly re-lated to the lithium concentration near the junc-tion. These results indicate that careful controlof the area coverage of the lithium prior to lithiumintroduction is not critical to the cell recoverycharacteristics.

It was postulated that the fact that the time tohalf-recovery appears to increase with increasingirradiation fluence might actually be a result ofannealing that occurs during the irradiation (higherfluences generally require longer irradiationtimes), so that the damage measured after bom-bardment does not represent the actual total dam-age, and the apparent half-recovery time is longerthan the actual half- recovery time. If the time tohalf-recovery does not increase with increasingfluence, this would simplify the description of theannealing kinetics.

It was observed that carrier removal ratesincrease with increasing lithium concentration andhence with distance from the P-N junction. Thatis, the amount of lithium which reacts with aradiation-induced defect during the recovery stageis proportional to the amount of lithium availablefor interaction with the defect, indicating that thedefect can nucleate precipitation of large numbersof lithium atoms if they are present in the vicinityof the defect. For example, the change in lithiumconcentration at 5 mrn from the junction was ob-served to be five times as great as the A-center(oxygen-vacancy defect) concentration at thatdepth. Furthermore, in lithium-containing cellsfabricated from float-zone silicon, the carrier re-moval rate was found to decrease with increasingfluence, again indicating a defect-lithium inter-action dependent upon the number of availablelithium atoms per defect. This indicates that themodel used in the past, which simply assumedthat two lithium donors were removed for eachannealed damage center, must be modified.

Annealing of 90% of the radiation damage wasobserved in lithium-doped silicon after isothermaland isochronal anneals of samples irradiated withfission neutrons and with high energy (30 MeV)electrons. Samples with high lithium concentra-tion exhibited degradation coefficients immediatelyafter electron irradiation which were greater thanthose of samples that did not contain lithium. Thepost- irradiation characteristics lend themselvesto interpretation by a two-defect model, one defectcontrolling the minority carrier lifetime T at lowtemperatures and the other controlling T at highertemperatures. The annealing appears to be afirst-order process, characterized by an activa-tion energy and an atomic frequency factor.

From electron spin resonance measurementsit was inferred that lithium inhibited the produc-tion of oxygen-vacancy radiation defect centers

and that the centers annealed at 350 K rather thanin the 600 K range observed for non-lithium-treated samples. In contrast, the production ofphosphorus-vacancy radiation defects did not ap-pear to be influenced by the presence of lithium.The radiation-induced divacancy, as observed bymeans of infrared spectroscopy, appeared to havean introduction rate in lithium-doped silicon simi-lar to that observed in non-lithium-doped silicon,but in the lithium-doped silicon annealing of thedivacancy proceeded much more rapidly. As theabsorption band associated with the divacancy de-creased through annealing, two new absorptionbands appeared, indicating the formation of a newdefect which is probably influenced by the asso-ciation with lithium.

Hall measurements performed at 4 K allowedseparate determination of donor and acceptor con-centration in silicon. Electron-irradiated lithium-doped silicon exhibited a defect level located 0. 14eV below the conduction band. It is postulatedthat this defect may be a vacancy-oxygen defect(A-center) associated with another impurity suchas an additional oxygen or carbon atom. It wasinferred that lithium interacted much morestrongly with the oxygen-vacancy defect than withthe phosphorus-vacancy defect and that as manyas four lithium atoms may be involved in the an-nealing of radiation-induced defect centers. An-nealing was observed to result in the disappear-ance of acceptor levels in irradiated lithium-containing samples but not in irradiated samplescontaining phosphorus but no lithium. Also, thecarrier concentration of irradiated lithium-containing samples was observed to decrease atroom temperature while the carrier concentrationof irradiated non-lithium-containing samples re-mained constant.

Electron microscopy of neutron irradiatedlithium and non-lithium-containing solar cells wasperformed using a surface replication technique.The density and diameters of the radiation-induceddisordered regions were obtained as a function oflithium doping and neutron fluence. It was foundthat the number of disordered regions observed bythis technique increased with increasing fluenceand increasing lithium density. The diameter ofthe disordered regions, however, were found todecrease with increasing fluence and with increas-ing lithium concentration. At a specific fluence,the total volume of the disordered regions appearsto be relatively constant with respect to lithiumdensity, increasing lithium densities resulting inmore numerous, but smaller-diameter, disorderedregions. The defect density of the lithium-containing cells did not appear to change withannealing temperatures up to 1200°C, while thedefect density of the non-lithium-containingsamples did change. Electron transmissionmicroscopy of the lithium- and non-lithium-containing cells indicated evidence of precipitateformation in both cell types.

Low flux irradiations obtained using a Co 6 0

source indicated a 4% superiority in maximumpower of lithium-doped cells fabricated fromcrucible-grown silicon over N/P (non-lithiumcontaining) 10 Q-cm silicon cells after a cumula-tive dose of approximately 5 X 1014 equivalentl-MeV electrons/cm 2 (100 days exposure) at atemperature of 60°C. Similar lithium-containing


cells irradiated at a temperature of 30°C were in-ferior to N/P cells at this fluence level. Similargroups of cells were also irradiated by means of ·a Van de Graaf generator to a fluence of 5 X 101-MeV electrons/cm2 and allowed to anneal at 30and 60°C. The same percentage superiority ofthe lithium-containing cells over the N/P cellswas observed for the 60°C anneal as was observedfor the cells exposed to the Co 0 irradiation atthis temperature. This indicates that the use ofaccelerated irradiation tests are valid for deter-mining the relative radiation resistance of lithium-doped cells with respect to N/P cells.

Two low flux experiments have also beencarried out using a Sr 9 0 source for a time periodof about six months (total fluence equivalent toabout three years in synchronous orbit). Thelithium-containing cells in these experiments rep-resented the best cells obtainable about one yearago, but not the best cells presently being fabri-cated. For the cells maintained at temperaturesof 50°C and above,lithium-doped cells fabricatedfrom crucible-grown silicon exhibited maximumpowers which were relatively constant over thelife of the test and about 8% higher than non-lithium-containing N/P cells. In one of the ex-periments the cells were post-irradiation annealedat the test temperature with no significant changesin power output being observed. This indicatedthat the cells were annealed as well as they couldbe during the test; that is, maximum annealing ofradiation-induced damage was occurring during thetest.

There appeared to be no additional radiationprotection afforded by 1-mil integral shields forthe low flux environment studied. One of the

'experiments indicated that the maximum power ofunilluminated cells was consistently about 4%higher than that of illuminated cells, while theother experiment was inconclusive with respect tothis phenomenon. If the results of the former ex-periment are accurate, this would indicate that theresults of most solar cell irradiation tests (i. e.,not only lithium-doped cell tests), which are per-formed with the cells in an unloaded unilluminatedcondition, may significantly underestimate theamount of damage to the solar cell maximumpower.

Lithium-doped solar cells fabricated from lowoxygen content silicon appear to have optimum an-nealing characteristics at a temperature of about20°C, while lithium-doped solar cells fabricatedfrom crucible-grown (oxygen- rich) silicon appearto be superior for temperatures of 50°C and above,for the conditions studied in these low flux experi-ments.

Irradiation of lithium-doped cells fabricatedfrom crucible-grown silicon by 28-MeV electronsto fluences of 5 X 1014 and 5 X 1015 e/cm2 ex-hibited recovered short-circuit currents 100%higher than similarly irradiated state-of-the-artN/P cells. This is consistent with previous in-vestigations using neutron irradiation, since inboth cases considerable cluster damage occurs.All results obtained thus far strongly indicate thatlithium is extremely efficient in reducing the detri-mental effects of radiation-induced defect clusterson the cell electrical characteristics. This meansthat for environments consisting of high energyelectrons, protons, or neutrons, lithium-dopedsolar cells should be vastly superior to state-of-the-art N/P solar cells.


0

N

LITHIUM- DOPED SOLAR CELLS FOR SPACE USE

P. Payne and E. L. RalphHeliotek Division of Textron, Inc.

Sylmar, Calif.

I. INTRODUCTION

This paper summarizes the lithium-dopedsolar cell research and development carried outby Heliotek over the past nine months under JPLContract 952547, Part II. The work performedcan be separated into two basic areas: (1) experi-mental work aimed at the improvement of cellprocesses and efficiencies, and (2) demonstrationof these improvements by the fabrication and sta-tistical analyses of quantities of various lithium-doped solar cell types.

Over the past several years the lithium-dopedsolar cell has progressed from a laboratory curi-osity, because of its unusual radiation damage re-covery characteristics, to a high-efficiencyradiation-resistant solar cell. Although signifi-cant improvements are still being made and abetter understanding of the device has beenachieved, it is a good time to stop and evaluate thestate-of-the-art. Some important factors are:

(1) Lithium-doped solar cells have been madethat have exhibited higher power outputsthan conventional N/P solar cells afterirradiation to a fluence of >1015 1-MeVelectrons per cm 2 .

(2) Lithium-doped cells have recently beenmade with very high efficiencies that aresubstantially higher than N/P cells.

(3) The size of lithium-doped cells have beenincreased from 1 X 2 cm to 2 X 2 cm andeven 2 X 6 cm as the result of the elimina-tion of the boron diffusion stresses.

These factors indicate that the lithium-doped solarcell basically meets or exceeds the electrical andradiation resistance characteristics of conven-tional N/P solar cells and improvements with lith-ium cells are still being made. The logical stepto be made next is the pilot production of thesecells, so that questions of cost, reproducibility,reliability, and space worthiness can be answered.

This paper summarizes the work that hasbeen done in the past nine months in respect tolithium cell power output, cell thickness, largearea cells, integrity of cell contacts and types ofexperimental cells prepared for radiation tests.

II. SUMMARY OF PREVIOUS DEVELOPMENTEFFORTS

The lithium-doped P/N solar cell has gonethrough distinct development stages over the pastfour years, with each stage showing a definiteimprovement in initial and post-irradiation re-covered efficiencies.

During the initial work (1966 to 1967) float-zone silicon was used. Lithium diffusion timesranged from 5 to 120 min, and diffusion tempera-tures ranged from 350 to 500 C. The concentra-tion gradients in the silicon slice were very steep,causing lithium cell instability, as exhibited byshelf-life degradation and redegradation after post-irradiation recovery. In order to reduce this gra-dient, redistribution cycles (removal of the ex-cess lithium from the cell surface and furtherheating of the cell to move the lithium deeper intothe slice) ranging from 30 to 120 min were used.Redistribution cycles produced higher output cells


with little change in the recovery characteristics.Table 1 shows typical air mass zero (AMO) elec-trical output values for cells diffused 5 and 90minand the change in output after 60- and 120-minredistributions.

Discovery in 1967 of the recovery capabilitiesof lithium-doped solar cells fabricated from cru-cible grown (C. G. ) silicon led to extensive investi-gation in this area. The higher oxygen concentra-tion C. G. lithium cells exhibited retarded roomtemperature post-irradiation recovery; however,recovery rates equivalent to those obtained withlow oxygen concentration lithium cells were ob-served when the cell temperature was raised to60 or 100°C. Cell stability, both shelf-life andafter radiation recovery, which was poor for somefloat-zone lithium cells, was not a problem withC. G. lithium-doped solar cells. In June 1969 theoutput of C. G. lithium cells was typically about2 mW higher (28. 3 versus 26. 5 mW) than similarlydoped float-zone lithium cells.

The next stage of development was the inves-tigation of lower lithium diffusion temperatures.Eight-hour diffusions at 325°C were performedand typical C. G. lithium cell efficiencies were ap-proximately 2 mW higher (30. 0 versus 28. 3 mW)than C. G. lithium cells that were diffused 90 minand redistributed 60 min at 425°C. The outputs ofthese cells after being irradiated with 1-MeV elec-trons to a level of 3 X 1015 e/cm2 were 11 to 20%higher than the 19. 5 mW average output of 10£-cm N/P cells.

These improvements in lithium cell charac-teristics were made independently of other pro-cesses in the fabrication procedure. They wereprimarily due to changes in lithium diffusionparameters. In order to further improve cell out-put, develop procedures which could be used inproduction and show capability of meeting basicrequirements for space quality cells, investigationin areas such as the boron diffusion, method oflithium application, and contact quality has beendone.

III. BORON DIFFUSION INVESTIGATION

The objective of the boron diffusion investiga-tion has been to develop a boron diffusion processwhich could be used to diffuse large quantities(>100 blanks per diffusion) of stress-free high-efficiency cells. BC1 3 has been the diffusionsource most commonly used in fabricating lithium-doped solar cells. This diffusion source was op-timized for fabrication of P/N cells (432-lam thick1 X 2 cm C. G. silicon blanks) early in the 1960's.This process had a very serious limitation in thatit could not be used for fabricating thin cells orcells of a 2 X 2 cm size or larger without a sig-nificant amount of bowing due to stresses intro-duced during the diffusion. Therefore, two meth-ods for reducing the stresses generated duringboron diffusion were investigated: (1) a modifica-tion of the standard BC1 3 diffusion, involving re-duction of the boron deposition time, and (2) useof BC13 with 02.

The BC13 diffusion, as it had been used over

the years, consisted of an 8-min warmup time, an8-min boron deposition time and a 10-min diffusiontime. During the boron deposition the BC13 reactswith the silicon; the silicon is etched and a heavy

deposit of boron silicides (Ref. 1) is left on thecell surface. This etch reaction and buildup ofboron silicides had previously been shown to beinfluenced by cell position on the boat during dif-fusion, with the cells first exposed to the BC13flow showing the greatest amount of silicon etched,the highest frequency of bowing, and the lowestoutput. In an experiment to determine whetherthe stresses could be reduced by reduction of theboron deposition time, the boron deposition timewas varied from 8 min down to 2 min. Both lappedand etched blanks were included in the experiment.It was found that the etched blanks (350 [am thick)were more susceptible to bowing than the lappedblanks (also 350 lam thick), i. e. , etched cells(1 X 2 cm, 350 am thick) were bowed after usingan 8-min boron deposition time, whereas lappedcells were not. Reduction of the boron deposi-tion time to 2 to 5 min reduced this heavy boronlayer and produced unbowed cells. By using a2-min boron deposition time 150 [lm thick 2 X 6cm blanks were then successfully diffused with nobowing. Typically, 8-min diffusions have producedbowed 2 X 2 and 2 X 6 cm cells (250 [lm thick)with the radius of curvature being as small as 8. 5cm. The absence of bowing of the large area cellsachieved with a 2-min boron deposition time indi-cated that the stresses were reduced considerably.

The other process modification investigatedfor reduced stresses was BC1 3 with 02. Just aswith the previously described diffusion, an 8-minwarmup time was used. The 02 was introducedinto the BC13 gas just before the gas flow enteredthe diffusion tube. The BC13 reacted with the OZto produce B 2 0 3 , which deposited on the cells asa glass layer. This glass layer and the subsequentdiffusion did not bow the cells.

Since lowstress cells were produced withboth processes discussed above, the two pro-cesses were compared to determine which resultedin the best cell electrical output. Two groups of100 cells were fabricated using the two differentprocesses. The silicon blanks used were 300 to380 lam thick with a resistivity of 0. 2 to 1. 2 2-cm.These were P/N cells with no lithium present.Etched rather than lapped blanks were used since,unlike the standard BC1 3 diffusion, neither of thesediffusion processes etch sufficient silicon to re-move the crystal surface damage present on alapped blank. As shown in Fig. 1, the AMO out-put of cells diffused in BC13 (no 02) with a 2-mindeposition time, averages 3 to 4 mW higher thanthe output of the cells diffused in BC1

3with 02.

The lower output of the cells diffused with 02 wasprimarily due to lower open-circuit voltage andhigher series resistance. The average open-circuit voltage of the cells diffused with 02 was590 mV, whereas the open-circuit voltage of cellsdiffused without 02 was 615 mV. The series re-sistance of cells from each group was measured.The series resistance of the cells diffused with 02was typically around 0. 8 2, whereas for the cellsdiffused with no 02 the series resistance wasaround 0. 3 Q. Table 2 summarizes the electricaldata on these two groups of cells.

At the present time, although neither of thesetwo diffusions discussed can be used for diffusionof one hundred cells or more per diffusion, theBCl 3 diffusion with a short deposition time andwithout 02 produces the best combination of lowstress and high efficiency.

JPL Technical Memorandum 33-491 5

IV. EFFECT OF BORON DIFFUSION PARAM-ETERS ON IMPROVEMENTS IN LITHIUMCELL OUTPUT

Once reduced stresses were obtained on P/Ncells by reducing the boron deposition time, theeffects of this process change on lithium cell out-put were investigated. Experimental lithium dif-fusions were performed using 20 Q-cm Lopexsilicon material with both 2- and 8-min boron de-position times. Both groups were placed in thesame lithium diffusion. The electrical character-istics of these cells are summarized in Table 3.The short-circuit current of the lithium cells sub-jected to the 2-min boron deposition averaged 11mA higher when measured in a 100 mW/cm 2 tung-sten light source and 6 mA higher when measuredin a solar simulator (AMO) than lithium cells sub-jected to an 8-min boron deposition time. Fig-ures 2 and 3 show the short-circuit current distri-bution obtained in the two light sources for the twogroups of cells (20 cells per group). These dis-tributions show that the Z-min boron depositionresults not only in higher, but more uniform,short-circuit currents.

In addition to the higher short-circuit currentsobtained for lithium cells subjected to a 2-minboron diffusion, open-circuit voltages (measuredat 25°C) as high as 615 mV were obtained; thehighest open-circuit voltage observed for cellsdiffused with an 8-min boron deposition was 595mV. The higher short-circuit currents and open-circuit voltages led to Lopex lithium cell AMOoutputs ranging from 27. 6 to 31. 7 mW, rather than23. 6 to 28. 8 mW as was obtained when an 8-minboron deposition was used.

Sintering of the contacted cell was investigatedto determine its effect upon the curve factor,whichfor some cells was below 0. 70. Sintering con-sisted of a heat treatment at 605°C for 6 min in anH 2 atmosphere. In most cases the sintering im-proved the curve factor slightly; in addition theAMO short-circuit current of lithium cells sub-jected to a 2-min boron deposition time increased3 to 5 mA with sintering. Typically improvementsof 10 to 25 mV in open-circuit voltage were alsoobserved. These improvements in curve factor,open-circuit voltage, and short-circuit currentresulted in AMO outputs 1 to 4 mW higher than theunsintered cell output. Figure 4 shows I-V curvesof a typical cell before and after sintering. Theopen-circuit voltage increased by 25 mW (600 to625 mV); the short-circuit current increased by3 mA (71. 5 to 74. 5 mA); the curve factor increasedfrom 0.706 to 0. 721; and the output increased by3. 3 mW (30. 3 to 33. 6 mW).

In fabricating Lopex lithium cells for Lot 11the following lithium diffusion schedules wereused: (1) 3 h at 340°C, (2) 7 h at 340'C, (3) 3 hat 360°C, and (4) 7 h at 360'C. Figure 5 showsthe distributions in maximum power (after sinter-ing) for the different groups of Lopex lithium cells.The various lithium diffusion parameters producedsimilar AMO outputs with the outputs for all thegroups ranging from 29 to 34 mW. This is equi-valent to an AMO efficiency range of 10. 9 to 12. 8%.The median efficiency is 11. 9%. These efficien-cies are the highest obtained on lithium cells thusfar and are comparable to high-quality conventionalN/P cells.

In Fig. 6 the Lopex cells which make up eachgroup in Fig. 5 are treated as a single group andthe AMO output is compared to a typical AMO out-put distribution of conventional 10 Q-cm N/P cells.The lithium cell output is 3 to 10% higher than theN/P cell output. Even if, for some reason, thesintering step is eliminated from the process, thelithium cell outputs are in the same range as the10 0-cm N/P cells. Figure 7 compares I-V char-acteristic curves of a typical 10 Q-cm N/P cell,and a typical lithium cell from this group of 105lithium cells. These curves show that the highlithium-doped cell output is due to short-circuitcurrents as high as 10 Q-cm N/P cell currentsand higher open-circuit voltages.

Not only has the BC13 diffusion with a 2-minboron deposition resulted in improved lithium celloutput, but it has also made it possible to fabri-cate 2 X 2 and 2 X 6 cm lithium cells. Figure 8shows various sized lithium cells which have beenfabricated. Sample cells, 2 X 2 cm and 2 X 6cm, which were fabricated from 20 Q-cm cruciblegrown silicon and lithium diffused 8 h at 325°Cexhibited efficiencies ranging from 10. 3 to 11. 3%.

V. CONTACT QUALITY

In order to evaluate Ti-Ag contact integrity,lithium-doped P/N cells were subjected to humidityand peel testing.

The humidity test consisted of exposing lith-ium cells with soldered contacts to 30 days of 90%relative humidity at 45°C. The cells were mea-sured at 25°C in 100 mW/cm2 tungsten lightsource before and after thirty days of humidity ex-posure. The cells exhibited 2. 5 to 3. 5% short-circuit current degradation, 0 to 2% current deg-radation at 450 mV, and no open-circuit voltagedegradation. Peel tests were performed on thesesame cells after humidity exposure. The testconsisted of soldering wires to both front and backcontacts and pulling at a 90-deg angle to the cellsurface until contact failure. The ten samples allexhibited peel strengths of 500 to 1000 g for bothfront and back contacts.

A tape peel test is also used to test contactintegrity on unsoldered cells. Scotch brand ad-hesive no. 810 is applied to both front and backsurfaces of the cells. If the contacts are weak,they will peel off with the tape. This test is usedon all lithium cells fabricated and the failure levelis less than 5%.

VI. CONCLUSIONS

Lithium-doped P/N solar cells meet the basicrequirements of solar cells for space use. Theefficiencies of the Lopex lithium cells which rangefrom 10. 5 to 12. 8% are equal to or better thanC. G. lithium cell and 10 0-cm N/P cell efficien-cies. Reduction in the stresses introduced duringboron diffusion has eliminated cell size restric-tions and both 2 X 2 and 2 X 6 cm lithium-dopedcells have been fabricated. The Ti-Ag contactswhich are tested 100% with a tape peel test, andpull tested and humidity tested on a sample basisare comparable to the Ti-Ag contacts on N/Pcells.


ACKNOWLEDGEMENT

The authors wish to thank P. Berman of JPLfor his technical coordination, and JPL for sup-port of the program under Contract No. 952547,Part II.

1. Powell, C. F., Campbell, I. E., and Gonser,B. W., Vapor Plating, p. 107. John Wiley &Sons, New York, 1955.


References

7

Table 1. Output of lithium-doped P/N solar cells

Diffusiontemperature, ° C

Diffusiontime, min

Redistributiontime, min

Pmax 'mW

425 5 25. 8 to 27. 3

425 5 60 29.2 to 29.4

425 5 120 30.4 to 30. 6

425 90 - 25.5 to 27. 0

425 90 60 2 6 .7 to 2 7 .6

425 90 120 2 8 .3 to 2 9.3

Table 2. Electrical characteristics of 1 Q-cm P/N cells (no lithium)

BC13 (no 02) BC1 3 (with 0 )Description 2-min deposition 2

range (median) range (median)

AMO output, mW 28 to 33 25 to 28(30. 5) (26. 9)

V , mV 600 to 625 570 to 600oc (615) (590)

Series resistance, 0.2 to 0. 5 0.4 to 1.0__ (0.3) (0.8)

Table 3. Electrical characteristics of lithium-doped P/N cells2-min versus 8-min boron deposition (20 cells per group)

Boron depositionrange (median)

Parameter

2 min 8 min

I , mAa 6 3. 5 to 6 6 . 0 48. 0 to 59. 0sc (65.0) (51.0)

I (AMO), mA 69. 0 to 71. 0 60.0 to 66. 0(70. 4) (613. 7)

(Voc)b mV 571 to 612 552 to 595.- (595) (578)

Output (AMO), mW 27. 6 to 31. 7 23. 6 to 28. 8(29.5) (26. 5)

a.

Measured in a tungsten light source at an intensity of 100 mW/cm .

bMeasured at 25°C.


C

30 :0

28 _ _4

0.01 0.05 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99 99.8 99.99"OR MORE" CUMULATIVE FREQUENCY, %

Fig. 1. Maximum" power distributions for 1 -cm P/N cells diffused with two different boron diffusiontechniques (measured at 25°C in solar simulator at 140 mW/cm2 intensity)

H.

66 kgE- ,;I±i ' J -: 1i

I

4.

1 .

Ii

t

X _

iZlP

4 -1 T'l' tq

99.99501ET E-tht, i 4 I. i I t t TH E- it 4 #t

0.01 0.05 0.2 0.5 1 2 90 95 98 99 99.85 10 20 30 40 50 60 70 80


z

E

urU

y

0'A

52

"OR MORE" CUMULATIVE FREQUENCY, %

Fig. 2. Comparison of short-circuit current of lithium cells as a function of boron deposition time(measured at 25°C in a tungsten light source at 100 mW/cm2 intensity)

i _-- 4

-4.

_ _

ti:

.t

E,I

LL1 1-it

9

68Z

- 66

UI-

0

0.01 0.05 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99 99.8 99.99


Fig. 3. Comparison of short-circuit current of lithium cells as a function of boron deposition time(measured at 25°C in solar simulator at 140 mW/cm Z intensity)

80 0 0 0 0 0

70

60- BEFORE SINTERINO -30.3 mW

AFTER SINTERING33.6 mW

50

VOLTAGE, v

Fig. 4. I-V characteristics curves of a typical lithium cell before and after sintering(measured at 25°C in solar simulator at 140 mW/cm 2 intensity)

10g1 JP Technical Memran

10 JPL Technical Memorandum 33-491

E

30

E 28

3h 340°C 22 CELLS26

7h 340°C 19 CELLS

3h 360QC 35 CELLS24

7h 360°C 29 CELLS

22 -_0.01 0.05 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99 99.8 99.99


Fig. 5. Maximum power distributions of Lopex lithium fabricated for Lot 11 (measured at 25°C in solarsimulator at 140 mW/cm 2 intensity)

72

0.01 0.05 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99


99.8 99.99

64 UE

N

x

60 Ž

o

56 0

52

48

44

Fig. 6. Comparison of sintered and unsintered Lopex lithium cell output distributions to a typical10 Q-cm N/P cell output distribution (measured at 25'C in solar simulator at140 mW/cm2 intensity)


-U

EU

-qx

o

E

E

:E

32

30

28

26

24

22

I

11

50

u E

O 40

P. 30

10

0.1

1 X 2 cm LITHIUM CELL 31.7 mW

2 x 2 cm 1 Oil-cm N/P CELL 59.2 mW

0.2 0.3 0.4

VOLTAGE, v

0.5 0.6

20

60 3

- 40

0.7

Fig. 7. Typical 10 f2-cm N / P cel l v e r s u s typical Lopex l i thium cel l (measured at 25°C in so la r s imula to r at 140 m W / c m ' intensity)

Fig. 8. Var ious s ize l i thium cel ls fabr ica ted by Heliotek

12 J P L Technica l Memorandum 33-491

0'

DEVELOPMENT AND FABRICATION OF LITHIUM-DOPEDSOLAR CELLS

P. A. IlesCentralab Semiconductor

Globe-Union Inc. , E1 Monte, Calif.

I. INTRODUCTION

This paper outlines the progress made in thepast year in the understanding and performance oflithium-doped solar cells.

II. SURVEY OF 1970 PRESENTATION

In the corresponding presentation a year ago(Ref. 1), the following achievements were reported:

(1) Improved boron diffusion methods, espe-cially the use of reduced tack-on borontrichloride cycles, were developed.These methods had led to improved out-put from cells, especially when oxygen-lean silicon was used. Thus, the fullcapability of the several forms of siliconcrystal growth was available for exploringa wide range of possible recovery timesand degrees of cell stability.

(2) The lithium diffusion cycles used werepredominantly with a single time cycleand with diffusion temperatures below425°C, extending as low as 325°C.

(3) The cell output resulting from the com-bination of these boron and lithium diffu-sion methods was higher, as shown inFigs. 1 and 2, where the cumulativepower distributions for JPL shipmentsC-11 and C-12 are shown.

(4) The lithium distribution throughout thecells was better understood by combiningresistance probing, which gave the donor

concentration in the bulk of the cell withcapacitance-voltage analysis. Thisanalysis explored the donor concentrationvery close to the PN junction. The dis-tributions resulting from five differentlithium diffusion schedules are given inFig. 3 (bulk distribution) and Fig. 4(distribution near the PN junction).

III. SURVEY OF WORK SINCE 1970

The work reported here has extended the re-sults summarized in the foregoing in the followingareas:

A. Cell Fabrication Sequence

The main fabrication steps required for cellfabrication (silicon growth, cutting, surface prep-aration, boron diffusion, lithium diffusion, con-tact application, coating application) have consid-erable interactions. Also the order of thefabrication processes can affect cell performance,often to a greater extent than anticipated. Toillustrate the effects of the order, if contacts areapplied to the front (boron diffused) P+ surfacebefore lithium diffusion, fairly severe heatingcycles can be used to sinter the contacts. How-ever, both these front contacts (and the antire-flective coating, if present) can be attacked bylithium during its application and the diffusioncycle unless the lithium source is sufficientlydiluted.

On the other hand, if the lithium is diffusedfirst, there is no attack of the front contact andcoating, and in addition, the P+ layer can be


protected from possible interaction with the lith-ium. However, the range of sintering cyclesavailable is now restricted to those which do notcause appreciable disturbance of the lithium con-centration in the cell. Evaluation work reportedby RCA (Ref. 2) showed good correlation betweencell recovery time after irradiation, and the lith-ium concentration gradient near the PN junction.Thus, it is particularly important that the sinter-ing cycles do not change this lithium gradientsignificantly.

In other words, the cell performance beforeirradiation may be relatively insensitive to thedifferences in lithium distribution. In addition,tests using thin slices (approximately 7 mils)showed that the redistribution cycle could lead tosevere cell degradation as a result of excess lith-ium present in the PN junction depletion region,although the average lithium level was low through-out the N region.

D. Lithium Distribution

Work in.the past year has shown that it is pos-sible to apply contacts and coating after lithiumdiffusion to obtain good electrical output and satis-factory contact adhesion by sintering for shorttimes at temperatures less than the lithium dif-fusion temperature.

The effect of more severe sinter cycles on thelithium distribution is also being investigatedquantitively.

B. Lithium Application

The paint-on lithium method provided well-controlled groups of cells (e. g. , Figs. 1 and 2)despite apparent disadvantages in lack of uniform-ity. However, for consistency in larger scalefabrication of cells, vacuum evaporation of lith-ium metal was reevaluated. This reevaluationwas prompted by several changes of cell design,all reducing some of the previous difficultiesfound with lithium evaporation. These favorabledesign changes included the need for less lithiumin the cells than in previous years, and the use oflower lithium diffusion temperatures. Also thealtered sequence described in the foregoing Sub-section A, with contacts and coating applied last,allowed protection of the front cell surface duringthe lithium diffusion. Many of the cell groups pre-pared using lithium evaporation have shown satis-factory close distribution of cell parameters andlithium concentrations,

Some of the evaporated lithium groups hadwide spread in all parameters. Systematic testsshowed where the evaporation method wastechnique-dependent, and allowed greater controlto be imposed. In analysis of these wide-rangegroups, the methods developed to test lithiumdistribution proved to be effective over the wholerange.

C. Lithium Diffusion Cycles

This year's work has concentrated on therange of temperatures from 375 down to 325°C,with various times from 2 up to 8 h at thesetemperatures.

Two-stage diffusion cycles, e. g. , those com-bining a tack-on followed by a redistribution cycle,are now considered to be inherently more difficultto control. First, the lithium concentration mustbe controlled on application; later the amount ofthe initial source left after tack-on must also beclosely controlled. Also, these two-stage cyclesoften involve longer times at a given temperaturethan a single cycle, with more chance of variationsoccurring. At the lower lithium levels now beingused, it is possible for the wide spread in lithiumcharacteristics during the two cycles to be maskedby the general improvement in cell characteristics.

More information and insight has been addedto the way in which the lithium is distributed andthe resultant effects on cell properties.

1. Surface concentration (Cs). The buildupof the lithium concentration at the back surfacewhere the lithium was applied, was measured as afunction of time at various temperatures.

Figure 5 shows the measured values. Quali-tatively, the results are as expected, with a grad-ual buildup of C s for longer times, to a maximumvalue - the rate of increase and the maximumvalue being higher for higher temperatures.

Two other features were noted. First, thebuildup was slower than expected, often requiringaround one hour to reach the maximum C s value.Second, after the saturation region, C s decreasedfor longer diffusion times. The points shown onFig. 5 for times greater than 2 h are those calcu-lated for the C13 cells, discussed in Section Elfollowing.

2. Concentration near the PN junction.Theoretically, both the lithium concentration andits gradient at any plane in the cell, including theimportant region near the PN junction at the frontof the cell, are both directly proportional to C s .However, as Figs. 3 and 4 showed the lithiumconcentration gradient near the junction must bemuch increased over the theoretical value in orderto give concentrations which agree with those cal-culated from C-V measurements. Gradients upto seven times the theoretical value have beenmeasured. The cause of this increase and theassociated lower concentration at the junction arebelieved to be the perturbation of the theoreticallithium distribution by the electric field in thejunction depletion region, which assists the flowof lithium into the P+ layer.

The changes in lithium distribution were fol-lowed by measurements on samples diffused at agiven temperature for several different times.Figure 6 shows the results of such a test. Thelithium buildup at the PN junction was followed bycapacitance measurements. For the three shortertimes, no measurable increase at the junction wasobserved in keeping with the theoretical curvesshown. The two longer times are seen to giveincreases in lithium near the junction. Thechanges needed in the theoretical distribution areshown by the dotted lines. Good consistencybetween measured and theoretical lithium distri-butions was obtained if the dotted line changesoccurred over a distance around 20 JLm.

3. Changes in concentration with diffusiontime. The trend noted in the foregoing SubsectionD(1) for the C

svalue at a given temperature to

decrease steadily with diffusion time gave


increased spread in C s values. This, in turn, ledto increased spread toward lower values of lithiumconcentrations and gradient near the junction.Although the cause of these changes was firstsought in effects near the junction, the C s mea-surements showed that the likely cause was thedecrease in C s resulting from depletion of thelithium source.

This depletion can be reduced by controllingthe amount of lithium deposited or by modificationof the diffusion procedure to minimize losses. Inthe following Subsection EZ the cell parametersresulting from wide variation in C s are comparedto the tighter grouping obtained when the lithiumsource depletion was reduced.

E. JPL Shipment C13

This shipment consisted of ten groups, eachwith thirty cells. The lithium diffusion schedulesused were between those used for Cll and they areshown in Table 1. At each diffusion temperature,the lithium was applied to batches of slices and thediffusion was performed on a split boat, allowingwithdrawal of the shorter time group first, leavingthe longer time group in the furnace. The differ-ences in cell properties for both short and longtime diffusions showed more correlation with theseparate batches, showing that the main variablewas in the paint-on lithium application. Thefabrication sequence applied front contacts andcoating before lithium diffusion.

In general, the cells showed good control andgood I-V performance. The controlling variablewas the diffusion time. Figure 7 shows the averagevalues of Pmax, Voc, Isc, and capacitance for thethirty cell groups inC13. Pmax varies over about4%. Voc and Isc show slightly larger ranges (upto 5. 5%) but their out-of-phase variations led toreduced spread in Pma,,. Also Voc and capaci-tance vary in phase, the spread in capacitancebeing up to 30%.

Figures 8 and 9 show in more detail some ofthe parameters for the C13 groups. In these fig-ures, average values and the spread in the groupsfor Pmax, Voc, capacitance, and donor concen-tration gradient near the junction are plotted as afunction of increasing diffusion time. The Pmaxvalues are satisfactorily high and tightly grouped.The other parameters all show a trend towardlower values and larger spreads at the longer time.These two figures show that the I-V characteristicscan have small spread, but may be associated withlarger spread in the lithium concentration or itsgradient.

Figure 10 plots the cumulative percentage ofPmax, averaged for all 300 cells in C13 with theupper and lower thirty cell groups shown also.For comparison, the distribution for C12 is given.This figure shows that the general level of celloutput was good.

Figure 11, for ten cells in C13G and C13Hshows good correlation between Voc and the lith-ium concentration and concentration gradient nearthe junction, calculated from capacitance measure-ments. For C13G, the shorter time diffusion, thegrouping is tighter and both Voc and the lithiumparameters are higher. Even for cells with lesslithium, the Voc values are generally high, and

this figure shows that greater resolution of theunderlying correlations is now feasible.

Figure 12 plotted for 100 cells, ten in eachC13 group, shows that Voc is approximately pro-portional to the logarithm of the lithium concen-tration. The estimated slope of the curve isapproximately 0. 026 V-1, a value expected if thePN junction is operating at current levels whereit approaches diode diffusion theory operation.The total spread in Voc is around 5%, whereas thetotal spread in concentration is 400%.

Table 2 lists the values of surface concentra-tion calculated for each of the C13 groups. TheseC s values were obtained by combining the lithiumconcentration and concentration gradient near thePN junction (estimated from C-V methods). As-suming the gradient occurred for a distance of20 pim from the junction, a value of lithium con-centration was obtained at this 20-[im plane. Thenthe theoretical distribution was extrapolated fromthis plane to the back surface to give the Cs value.The groups in Table 2 are plotted in increasingorder of diffusion time, and show the steady fall-off of C s.

In summary, C13 provided 300 cells of goodoutput, but with sufficient differences between thegroups to show that diffusion time was a moreimportant controlling parameter than had beensuspected earlier. To reduce the effects of sourcedepletion at the longer times, a later shipment(C15) was planned to include both paint-on andevaporated sources, to check that the sourcedepletion was reduced for the more concentratedsource.

1. JPL shipment C14. The intention in C14was to compare oxygen-rich (C. G. ) silicon andoxygen-lean (Lopex) silicon in two of the lowertemperature single time cycles. Also lowerresistivity Lopex silicon (15 l2-cm) was used tocheck if the five-fold increase in background con-centration was of advantage for higher fluenceirradiations. The fabrication sequence for C14applied lithium before any contacts or coatingswere present. Table 3 gives details of the siliconand lithium cycles used.

Figure 13 gives the average values and thespread of Pmax, Voc and capacitance for the C14groups. The general level of Pmax (and Voc) wasgood. For a given lithium cycle, the C. G. siliconcells had higher Voc (and Isc) than the Lopex cells.However, this higher Voc was accompanied by thelower capacitance (and thus lower lithium concen-tration near the junction) usually found for C. G.silicon. This lower capacitance can explain thehigher Isc values obtained for C. G. silicon, butdoes not explain the Voc differences. In this case,the differences in silicon grown by the two methodshave more effect on I-V characteristics than thedifferences obtained in lithium concentration.

2. JPL shipment C15. The purpose of C15was to compare three different silicon ingots usingeither paint-on or evaporated lithium sources foreach group and a diffusion schedule similar to thoseused in C13. A schedule of 350°C for 240 min waschosen; and in addition, a group of C. G. cells wasfabricated using the evaporated source and 350°Cfor 480 min to check if the lithium depletion wasreduced for the longer diffusion time.


Table 4 shows details of the C15 groups.

Figure 14 plots the average values and thespread for the various C15 groups. Again, theoverall level of Pmax is good and, as seen in Cl and C14, the Pmax values for C. G. silicon areslightly higher than those for Lopex cells. Thereis no significant difference in these preirradiationmeasurements between the medium and high resis-tivity Lopex groups. Also, as for C14, the lithiumconcentration near the PN junction, as measuredby capacitance, was higher for the Lopex ingots.

The overall spread is not markedly lower forthe lithium evaporation groups, possibly becausethe diffusion time is not long. However, thespread in Pmax and especially in capacitance forthe 8-h diffusion group C15G was much lower thanobtained for the longer time C13 groups. Thisshows that lithium depletion has been minimized bythe more concentrated source.

During fabrication of the evaporated lithiumgroups, some batches showed signs of severe lith-ium depletion. These signs included both seriousreduction in Voc (with accompanying higher Isc)

and low capacitance values often near the levelfound for cells made with no lithium. Resistanceprobe measurements showed that the lithium con-centration was much reduced for these batches.These tests showed that lithium evaporation wastechnique dependent. Using improved techniquesmainly in reducing the interaction of lithium andsilicon before the slices were placed in the dif-fusion furnace, gave increased lithium concen-trations with very closely grouped sets of param-eters. Figure 15 compares the spread of threecell parameters, Voc, capacitance and Pmax forthree different conditions of lithium application -namely paint-on, lithium evaporation with varyinginteraction, and well-controlled lithium evapora-tion. The silicon used for these three tests wasthe same (Lopex, 75 Q-cm) as was the lithium dif-fusion schedule at 3500C for 240 min. The paint-on source again shows tight grouping of Pmax andVoc with a wider spread in capacitance. The lesscontrolled evaporation group had lower values forPmax, Voc, and capacitance with wide spreads.The well-controlled evaporation group had highvalues with small spread for all parameters.

F. More Complex Structures

The present design of lithium cell with suitablychosen silicon and lithium diffusion conditions cangive cells with good performance for many possiblemissions. Any further complexity added to thecell design must be justified by the more severeconditions imposed by particular missions.

The study of two of the possible complexitiesneeded to maintain cell stability following highfluence irradiation has been continued this year.

1. Use of additional N + layer at the backsurface. This layer, using donors other than lith-ium, has been used particularly with oxygen-leansilicon to reduce the back contact resistance.However, as will be noted in the following Sub-section G1, present contact resistances are satis-factorily low. More tests are needed to check thethreshold fluence at which carrier removal bylithium depletion (occurring during the recovery

phase following irradiation) increases the contactresistance appreciably.

2. Use of additional N + layer near the frontsurface. Past results have shown that this layercan increase the stability of the PN junction whenlithium carriers are removed following highfluences. Again tests are needed to find thethreshold fluence which makes this N+ layer neces-sary. The threshold fluence will be higher as thestarting donor concentration is increased. Thuscomparison of the C15 Lopex groups will be ofinte rest.

The other possible use of the front surface N+layer is to control the final distribution of lithiumnear the PN junction after the lithium diffusioncycle. Tests in progress will show if any advan-tageous distributions can be obtained.

3. Front surface introduction of lithium. Inthe past year, tests continued on the possiblemerits of introducing lithium through the frontsurface of both P/N and N/P cells. The cellsmade to date with this method did not show anyadvantages over the conventional design cells.

G. Other Topics

1. Contacts. For the levels of lithium usedin the past year, Ti-Ag contacts have given curvefill factors (CFF) values between 0. 70 and 0. 76.The sequence used for shipment C13 (front con-tacts applied before lithium) generally gave CFFaround 0. 72. The later shipments C14 and C15,where both contacts were applied after lithiumdiffusion, had CFF around 0. 74 with minimumlithium depletion at the back surface, CFF > 0. 75have been obtained. This lessens the need for anextra N + layer at the back surface to reduce con-tact resistance.

Evaluation of various contact sintering cyclesis in progress. The cycles will be compared foradhesion of the contacts, the effects on the I-Vcharacteristics of the lithium cells, and the effectson lithium distribution within the cells.

2. Improved antireflective coatings. Im-proved coatings, using double layers of dielectricshave given slight increase (2%) in the output oflithium cells. The increase was less than thatobtained on N/P cells because present lithium cellsalready have a form of double-layer coating.

3. Maximum output versus lithium concen-tration. A search was made for the highest Pmaxvalues obtained for different lithium concentrationsnear the PN junction. The curve joining the pairsof corresponding points for all the cells measuredis given in Fig. 16. For lithium concentrationsaround 5 X 1014 cm-3 , Pmax = 33 mW is possiblefor 2-cm 2 cells. 31 mW was obtained for 1015lithium atoms cm-3 for C. G. silicon, and for >1015lithium atoms cm-3 for Lopex silicon. This figureshows the current feasibility of reaching an opti-mum trade off between Pmax and recoverycharacteristics.

IV. SUMMARY

This year has shown further improvement inthe performance and understanding of lithium cells.


They have proved to be superior to N/P cells forseveral possible missions, particularly when ener-getic heavy particles (protons or neutons) are asignificant fraction of the bombarding particles.The high output and repeatability are obtainablefrom both oxygen- rich and oxygen-lean silicon.The precision of analysis of cell behavior has beenimproved.

The analytical techniques developed will alloweven further control of the lithium distribution,aiding in closer grouping of both pre- and post-irradiation performance.

The fabrication sequence alterations have ledto higher cell output, better appearance, and in-creased contact strength.

The present methods of fabrication and anal-ysis are well suited to beginning the evaluation ofpossible pilot-line scaling up.

There is still a need for a figure of merit toallow assessment of the radiation capability of thecells. Before lithium cells are used in largenumbers, they must pass a relatively simple testto prove that all the cells on the array will degradeand recover in a similar fashion. At present, thetest suggested by RCA (Ref. 2), wherein the lith-ium concentration gradient is measured, provides

the best figure of merit, but further tests arenecessary.

As shown in Fig. 15, there is still some roomfor improvement in the cell output for a given li-thium concentration. It will be of interest to seeif other papers on irradiation effects substantiatethe optimism that the cell fabricators have re-garding lithium cells.

REFERENCES

1. Iles, P. A. , "Progress Report onLithium-Diffused Silicon Solar Cells" in Proceed-ings of the Third Annual Conference on Effects ofLithium Doping on Silicon Solar Cells, Jet Pro-pulsion Laboratory, Pasadena, Calif., April 1970.

2. Faith, T. J., Corra, J. P., and Holmes-Siedle, A. G., "Room Temperature Stability andPerformance of Lithium-Containing Solar Cells -An Evaluation," in the Conference Record of theEighth IEEE Photovoltaic Specialists Conference,Seattle, Wash. , August 1970.

ACKNOWLEDGMENT

We would like to thank NASA and JPL for theirsupport under Contract No. 952546 and Mr. PaulBerman for his technical support.


Table 1. Lithium diffusion schedules for C11 and C13

temperature, ° CCell group

325

325

325

330

330

340

340

350

350

360

360

370

370

375

Cl1A

CllB

Cl C

C13A

C13B

C13C

C13D

C13E

C13F

C13G

C13H

C13I

C13J

Cl1D

480

480

480

180

420

180

420

300

300

180

420

240

360

180

Table 2. Surface concentration of lithium for C13 groups

Cell group Temperature, °C Time, min Calculated Cs, cm

17C13A 330 180 1.3 X 10

C13C 340 180 1.7 X 1017

C13G 360 180 1.6 X 10

C13I 370 240 0.95 x 1017

C13E 350 300 1.2 X1017

C13F 350 300 1.2 x 10

C13J 370 360 0.75 X 1017

C13B 330 420 0.70 x 1017

C13D 340 420 0.50 X 10

C13H 360 420 0.65 x 1017


Diffusion t Diffusion time, min

18

Table 3. Details of JPL shipment C14

Cell group No. of cells Silicon used Lithium diffusion

C14A 30 Lopex- 15Q1-cm Paint on at 375°C for 180 min

C14B 30 C. G. - 30 Q-cm Paint on at 375°C for 180 min

C14C 30 Lopex- 15Q-cm Paint on at 350°C for 300 min

C14D 30 C. G. - 30Q-cm Paint on at 350°C for 300 min

Table 4. Details of JPL shipment C15

Cell group No. of cells Silicon used Lithium diffusion

C15A 30 Lopex- 15 Q-cm Paint on at 350°C for 240 min

C15B 30 Lopex- 15 2-cm Evaporated at 350°C for 240 min

C15C 30 Lopex-75 1Q-cm Paint on at 350°C for 240 min

C15D 30 Lopex-- 75 2- cm Evaporated at 350°C for 240 min

C15E 30 C.G. -30 2-cm Paint on at 350 C for 240 min

C15F 30 C.G. -- 30 2-cm Evaporated at 350°C for 240 min

C15G 30 C.G. -30 2-cm Evaporated at 350°C for 480 min


E

2 29

E

o 28

270

26

25

24 I 12 5 10 20 30 40 50 60 70 80 90 95 98 99 99.8 99.99

PERCENT WITH POWER GREATER THAN P4 5 0

Fig. 1. Cumulative Pmax distribution for JPL shipment C11

33 I I I I I I I I I I I I

E C.G. SILICONLi 325°C - 480 min) 32

E31

30 .

29 I I I I I I I I I I I I I I I1 2 5 10 20 30 40 50 60 70 80 90 95 98 99 99.8 99.99

PERCENT WITH POWER GREATER THAN P4 5 0

Fig. 2. Cumulative Pmax distribution for JPL shipment C12


1018

1017 / 400-90.. _375-180 400-90 400-90

420-90

Ei / < 350-300 350-300U

zo 375-180

325-480325-480

0 2 4 6 8 10 12

0z0

1015

BACKGROUND CONCENTRATION

10DI STANCE FROM FRONT, mi0 2 4 6 8 10 12

DISTANCE FROM FRONT, mils

Fig. 3. Donor concentration profiles (in the bulk) forvarious lithium diffusion schedules


E

z

uO

zUZ

Oc

20 30 40 50

DISTANCE FROM P-N JUNCTION,/m

Fig. 4. Donor concentration profiles (near the P/N junction)for various lithium diffusion schedules


4

107

8

6

C') II lr ~325E 4

0o

Z 2Uz0U

10 16Z 10'

8

6

4

2

1015

0 60 120 180 240 300 360 42

LITHIUM DIFFUSION TIME, min

Fig. 5. Build-up of surface concentration of lithium for increasingtime at various temperatures


0

23

U 202

015z 10

zO 6

:D4 - DONOR CONCENTRATIONI FROM CAPACITANCEMEASUREMENTS

2

1410

6

4 K P-N JUNCTION

2

10I I I I16 14 12 10 8 6 4 2 0

DISTANCE FROM BACK, mils

Fig. 6. Changes in lithium concentration profile and the valuesmeasured at the P/N junction as the diffusion time at330°C was increased

JPL Technical Memorandum 33-491Z4

I I

32-

I I I

I I I- - -

RANGE,

4

3

5.5

C13 GROUP

Fig. 7. Average values for Pma,, Voc, Isc, and capacitancefor the ten C13 groups

C13 GROUP

E

Sx-Eo

32

30

28

630

> 610

>° 590

570

17

t

uZ

u

Uz

U

U

15

11

9

K I~~~T1 < I II - I III I I I I

I I I I II~~~ I

I I I I I I

0 60 120 180 240 300 360 420 480


Fig. 8. Average values and spread for Pmax, Voc, and capacitance for theC13 groups, plotted as a function of diffusion time


E

30

I I I I I I I

I I I I

oIu1olU

> 600

>0 590>o

E

z

U

U

74l

1 1 1 F I I

68 I I- I

IT

25

C13 GROUP

AC G I E F J BD H

I I r

II

I

60 120 180 240 300


IL360 420 480

Fig. 9. Average values and spread of donor concentration gradient at theP/N junction, for the ten C13 groups, plotted as a function ofdiffusion time

33 I I I I I I I I I I I

32C12 C.G. AT 325°C FOR 8 h

31

E

Ea- 29 LIMITS FOR SEPARATE

GROUPS IN C13 \GROUPS IN C13 AVERAGE FOR ALL

GROUPS IN C13

28

27

26 I I I I I I I I I I I I I I I I1 2 5 10 20 30 40 50 60 70 80 90 95 98 99 99.8 99.9 99.99

PERCENT WITH POWER GREATER THAN P

Fig. 10. Cumulative Pmax distribution for total and extreme groups in C13(C12 shown for comparison)


10

8

6

4

2

0

5'E

x

zo

uz-z

z0

uzou

zou:

I II 'i I

Z6

20 I I

XX16

E X XE x x

X X X

12 O

;~ x~~~~~~x8-

4 0

0 o o ° o I

17 X

X

15 X

U-

z13 0

0_ O

0 C13H AT 360°C FOR 7 h

O O0 09 I I I

570 580 590 600 610 620

OPEN-CIRCUIT VOLTAGE, mV

Fig. 11. Voc plotted versus capacitance andlithium concentration gradient nearthe junction for two 10-cell groupsin C13


620 IN C13

(10 CELLS IN EACH OF ENVELOPE OF MEASURED VALUES10 GROUPS)

5 600

I -0IO

580

560

540 2 4 1 1

1013 2 4 6 1014 2 4 6 1015

LITHIUM CONCENTRATION AT P-N JUNCTION, cm-3

Fig. 12. Vo plotted against logarithm of the donor concentration for 100 cells, 10 in each C13 groups

GROUP

SILICON

DIFFUSION TEMPERATURE

E

E

610

E590

570

26

Uz

U

U

22

18

14

C14A C14B

LOPEX C.G.

375°C 375°C

C14C

LOPEX

3500 C

C14D

C.G.

350°C

I I I I I I

I I I I I I

0 I I T I I

0 60 120 180 240 300 360 420


Fig. 13. Average values and spread of Pmax, Voc, andfor C14 cell groups

capacitance


GROUP

SILICON

LITHIUM DIFFUSION

CI5A C15B C15C C15D C15E CI5F

Lx(15) Lx(15) Lx(75) Lx(75) CG(30) CG(30)

350°C FOR 240 min -

LITHIUM SOURCE PO EV PO EV PO EV

Fig. 14. Average values and spread for Pmax, Voc, and capacitancefor C15 cell groups


x 28

26E26

24

610

E- 590

570

24

t 20Uz< 16

< 12u

8

I I I I I I I

I I I I I I I

.I I I I I IC15G

CG(30)

350'C480 minEV

29

2

_JU

0

z

560 580 600 8 10 12 14 16 18 20 22 24 26 28 30 32

VOC' mV CAPACITANCE, nF Pmox' mW

Fig. 15. Histogram plots of Voc, capacitance, and Pmax for three different conditions of lithiumapplication in two C15 groups

36

AMO, 140 mW/cm 2

CELL 2 cm

- AT 28°C

32 -

30 -

N UPPER LIMIT FOR ALL TESTS

I -

U T

UPPER LIMIT FOR C13 / """""-~"'" '''''

0 1 2 3 4 5 6 7 8 9

LITHIUM CONCENTRATION NEAR JUNCTION, X 1014 cm-

3

Fig. 16. Highest observed pairs of values for Pmax and lithium concentration


I I I I I I I I I I I I I I I

LOPEX SILICON (75 n -cm)

20 _ DIFFUSED AT 350 0 C FOR 4 h

5 C-15C Li (PAINT ON)

0

15 C-1SD Li EVAPORATION (POOR)

10

15 h C-15D Li EVAPORATION (GOOD)

0

O n n n11 r n ni 0

34

E

0:0~Z

-TDXO~

10

1

30

AIR FORCE/ION PHYSICS HARDENEDLITHIUM DOPED SOLAR CELL DEVELOPMENT

A. Kirkpatrick, F. Bartels, C. Carnes,J. Ho, and D. Smith

Ion Physics CorporationBurlington, Massachusetts

I. INTRODUCTION

Ion Physics Corporation (IPC) is working underAir Force Contract F33615-70-C-1491 to developa lithium-doped solar cell hardened against bothspace radiation and nuclear weapons burst environ-ments. Performance objectives for the hardenedcell are sufficiently severe that they are not metby any of the lithium cells which have been reportedon to date. IPC is developing a cell structure de-signed to meet these program goals. The first12 months of the 24-month program have involveddevelopment of elements of the cell structure. Thecompleted structure is now to be assembled andoptimized.

II. THE IPC CELL STRUCTURE

Because the performance characteristics ofthe lithium-doped solar cell are determined by theconcentration profile of lithium throughout the cell,IPC is attempting to develop a structure allowingindependent control over profile shape in the base,concentration in the base and profile in the vicinityof the junction.

The IPC cell is illustrated in Fig. 1. The cellis 10 mils thick and has 5- 1 lm aluminum contacts.Base material is 20 f-cm phosphorus-doped sili-con with <100> orientation. The cell is a P+ N+NN+ structure in which the N+ phosphorus-dopedregion completely encloses the base N region.Purpose of the N+ region at the front below the P+layer is to reduce diffusion of lithium into the P+layer and to prevent widening of the space chargeregion due to lithium depletion after irradiation.

The N+ layer at the cell back prevents outdiffusionof lithium and insures low resistance ohmiccontact.

The cell is fabricated in the followingsequence:

(1) The silicon wafer is diffused at 1000°Cwith PH

3to produce the back N+ layer.

(2) The front P+ layer is produced by dif-fusion at 950°C with BZH6 .

(3) The cell is sized using an orientationdependent etch which leaves <111> planesas the cell edges.

(4) Phosphorus is ion implanted from thefront to produce the buried N+ layerwhich also extends down the cell edgesurfaces.

(5) Lithium is introduced through the backsurface by ion implantation and is dis-tributed through the cell using conventionalelevated temperature distribution cycles.

(6) Aluminum contacts and antireflectioncoating are applied.

A. Lithium Implantation

Implantation of the lithium is fundamental tothe IPC cell design. In the implantation process ahigh energy beam of the desired ion is formed,analyzed and rastered over the surface to be


implanted. The number of ions introduced ismeasured directly by integration of total chargecollected at the substrate. Typically IPC used250 keV Li 7 which penetrates approximately 1 FLminto the silicon.

Implantation of the lithium offers many im-portant advantages over other lithium introductiontechniques:

(1) The amount of lithium introduced isprecisely controlled and is uniform andreproducible.

(2) The amount of lithium introduced isindependent of the distributiontemperature.

(3) No surface treatment is required afterimplantation as there is no surfaceresidue, surface etching or pitting.

Following implantation, the lithium is distrib-uted using distribution cycles approximatelyequivalent to those employed with diffused lithiumcells. During distribution a large portion of theimplanted lithium is lost from the back surface ofthe cell. The N+ barrier layer at the back surfaceinto which the implant is made reduces this loss bymore than an order of magnitude. Shown in Fig. 2are experimental lithium profiles after distributionat high temperature of cells with and without theN+ barrier. Because IPC is able to "count" everylithium ion introduced into the cell, it is possibleto determine from each profile a "utilizationfraction," f, which is the fraction of initially im-planted lithium remaining after distribution.Typically for implants at 250 keV into the N+barrier with distribution times of the order ofhours at 450°C or below this fraction rangesbetween 0. 02 and 0. 03.

Desired cell base lithium concentrationsaveraging slightly above 1016 cm-3 are obtained byimplanting 1 X 1016 lithium ions cm- 2 . This levelcan be produced in about 2 min per cell from theIPC research implantation facility. As seen inFig. 3, the lithium concentration throughout thebase can be varied almost linearly with implantfluence.

The N+ layer at the back surface into whichthe lithium implantation is made has an additionalimportant function. Ion implantation into a crystallattice causes structural damage which for mostapplications is then annealed at elevated tempera-tures typically in the range of 700°C and above.When lithium is implanted into silicon, the latticedamage which occurs is not completely annealedat temperatures below 450°C. Short 5-min annealsat 700°C are sufficient but could cause reproduci-bility problems with the lithium profile. If theimplantation is made into 20 l-cm silicon withoutan N+ surface layer and adequate anneal is notprovided, the resulting solar cell has high seriesresistance of several ohms. However, when theimplant is made into the heavily doped N+ layer,

it is found that no anneal is required to insuregood ohmic contact and low series resistance.

B. The Implanted Phosphorus Front Barrier

In order to reduce lithium diffusion into the P+region and to prevent widening of the junctionspace-charge region due to lithium depletion afterirradiation, IPC will implant a thin N phosphoruslayer immediately below the front P+ layer. With-out this N+ barrier it is possible to control shapeand concentration of the lithium profile through thebase but desired control over the lithium profilein the junction vicinity is dependent upon use of thebarrier. This barrier is the only component ofthe planned structure on which development hasnot yet been successfully completed.

Desired concentration profiles near the junctionare illustrated in Fig. 4. Phosphorus is implantedat 200 to 300 keV to penetrate approximately0. 15 Bm below the junction and to produce a peakconcentration of approximately 101 6 cm-3. Thephosphorus N+ region below the P+ surface layeris to act as a barrier to reduce loss of lithiuminto the P+ layer sink during distribution and con-sequently to provide control of the lithium level.Peak concentration in the N+ layer is selected tobe approximately equal to the lithium level in thevicinity of the junction and should serve as abarrier to lithium diffusion until concentrationreaches this level. If the phosphorus concentra-tion is made much higher than the required lithiumlevel, the N+ region will retard flow of holes tothe junction and will reduce cell performance.

Investigation of the front barrier is proceed-ing. Results to date have been inconclusive. Witha few exceptions, cells having the front barrierhave shown losses in Voc of 30 to 100 mV relativeto otherwise identical cells without the barrier.Some cells have shown evidence of high junctionvicinity lithium concentration and low concentra-tion gradient, but these results have not beenconsistent or reproducible.

III. PRESENT STATUS

A first lot of 110 crucible-grown silicon cellshas been fabricated without the front N + barrier.Lithium implant fluence was 1 X 1016 cm-2 at250 keV followed by 450°C 120-min distribution.Layer stripping showed lithium utilization fractionof 0.028 with a profile peak of 1. 3 X 1016 cm-3 .Median maximum power of these 10-mil thickaluminum contacted cells was 58 mW.

Introduction of lithium by ion implantationeliminates reproducibility and surface problemdeficiencies of other introduction techniques.Implantation has been demonstrated to make pos-sible a degree of control over the cell lithiumcontent which has not previously been available,Front barrier development remains to be com-pleted. Successful development of the barrier willmake available the freedom to select optimumlithium concentration throughout the cell, includingin the vicinity of the junction.


BORON DIFFUSEDP+ LAYER

<100> / IMPLANTED PHOSPHORUSt ~ / N+ LAYER

<111> /10-mil LITHIUM-DOPED N REGION \

PHOSPHORUS DIFFUSEDN+ LAYER

Fig. 1. Ion Physics Corp. lithium-doped cell

5 x 1022

1022

E

z

2

I-

Zz0U

0z00

! :: : 1 . r : !: : : t__ _ :-: : : !

...._ :_ :_:WITH N+ LAYER f f=0.043, .. i__ .'-

~__-- ... I t i .t-7 _

: · 77 -- : i. . :.

- .'I _ WITHOUT N+ LAYER[ i4f 0.002_ -....

:-_ :......l-"' 1 .....' ? .. ' :- -

_-- __=7 -

_ v .- -:.--, --- - - - -_

SILICON <100> C.G.

_ -i - LITHIUM IMPLANT 1 x 1016 cm 2I - 300 keV

DISTRIBUTION: 700°C 5 min

0 2 4 6 8 10 12

DEPTH FROM CELL BACK, mils

Fig. 2. Effect of back N + layer on lithium profile


Alm

33

0.25/m

F- 0.4/m

5x1022 ! i i

: r: ::=-' :- .-i.::-:-- i:l. -- -t _- - !

__ 17 _, 7 ____ 1 ,__ i7Ii

_ 1 ---I-_ i---- ~ l $'1' 2IMPLANT FLUENCE 1 x 10 cm

io22 -- if 0.026 -;10

_ _:: - -- - ' : -__

o ILT------ - F c --E---

2 -, i f = 0.021

u .. .t ' _z --- - 7- 7 _- !7

l:i' I . !N .. ...

2 -DISTRIBUTION 425°C/200 mn 1021

DEPTH FROM CELL BACK, is. : :

Fig. 3. Effect of implant :::fluence on profile:

-_ _. --- -- T -- ITT!-" 1 - ,

........ i ..... _-' ,- . . . ...... , i--..... .102 0 I' ' i - ' - .- '-': ....

1022

Z _ . DIFFUSED BORON P+

UzO IMPLANTED PHOSPHORUS N

+

/0 16~/LITHIUM DISTRIBUTIONol6 - /

PHOSPHORUS BACKGROUND10~ - -

1014 _ ____ _ I I _0 1 2 3

MICRONS FROM FRONT SURFACE

Fig. 4. Planned concentration profiles


•1

DEPENDENCE OF DEFECT INTRODUCTION ON TEMPERATUREAND RESISTIVITY AND SOME LONG-TERM ANNEALING EFFECTS

G. J. BruckerAstro-Electronics Division of RCA

Princeton, New Jersey

I. INTRODUCTION

The effort reported here represents data whichwas obtained after the termination (June 1970) ofour contractual investigation of lithium propertiesin bulk-silicon samples before and after irradia-tion. The objective of this phase of the study wasto obtain the analytical information required tocharacterize the interactions of lithium withradiation-induced defects in silicon, and hopefullyto develop a model of the damage and recoverymechanisms in irradiated-lithium-containing solarcells. The approach to the objectives was basedon making measurements of the Hall coefficientand resistivity of samples irradiated by 1-MeVelectrons. Experiments on bulk samples includedHall coefficient and resistivity measurementstaken as a function of: (1) bombardment tempera-ture, (2) resistivity, (3) fluence, (4) oxygen con-centration, and (5) annealing time at temperaturesfrom 300 to 373 K.

II. RESULTS

A. Carrier Removal

Previous work of several investigators (Refs.1, 2) demonstrated that the defect introductionrate in phosphorus-containing silicon decreases asthe bombardment temperature is decreased andalso as the dopant concentration is increased at afixed bombardment temperature. One of theobjectives of our study was to determine the de-pendence of carrier-removal rate in irradiatedsilicon on bombardment temperature and lithiumconcentration. To achieve this objective, severalHall bars fabricated from 1500 Q- cmand 5000 Q2 -cm

float-zone refined silicon and from 30 to 50 Q-cmquartz-crucible silicon were diffused with lithiumto concentrations from 2 X 1020 to 3 X 1023 Li/M 3

and irradiated at bombardment temperatures from79 to 200 K. The rates of carrier-removal weredetermined after each bombardment. These ratesmeasured in float zone (FZ) silicon after annealingto a temperature of 200 K versus the reciprocal ofthe bombardment temperature are shown in Fig. 1.Lithium concentrations of the samples used in themeasurements are shown as a parameter. Thecarrier-removal rates were normalized at a bom-bardment temperature of 200 K, and the ordinateof the curves is labeled as the probability of defectformation. The curves shift along the temperatureaxis toward lower temperatures as the lithiumdensity is decreased over three orders of magni-tude. These concentrations of lithium correspondto resistivities from 0. 03 to 20 Q-cm. This shiftof the temperature dependence portion of thecurves is in qualitative agreement with the pre-diction of the interstitial-vacancy close-pairmodel (Refs. 1,-2) of radiation damage in siliconat low temperatures. It was previously reported(Ref. 3) that the saturated value of carrier-removal rate measured at a bombardment temper-ature of 200 K appeared to increase with lithiumconcentration. This conclusion is in agreementwith the results obtained on the samples containingconcentrations to 3 X 1023 Li/m 3 .

A similar set of curves is shown in Fig. 2 forthe samples fabricated from quartz-crucible sili-con with lithium concentrations from 2 X 1021 to3 X 1023 Li/m 3 . The saturated value of carrier-removal rate obtained for the samples dopedwith 3 X 1023 Li/mn 3 is equal to 10 m- 1 which is


the same value obtained for the samples dopedwith 2 X 1022 Li/m3.

B. Annealing Properties-High LithiumConcentrations

Following the completion of the bombardments,the float-zone samples doped to 3 X 1023 Li/m 3

were annealed at temperatures ranging from 250 to300 K. Hall and resistivity measurements wereobtained on these samples as a function of timeand measurement temperature. The time to halfrecovery, tR of the mobility (TM = 79 K) wasdetermined from these annealing cycles, and thereciprocal of this quantity is shown in Fig. 3. Theerror in the measurement is large at the highertemperatures since the sample recovered to itsoriginal value of mobility in less than a minute at300 K. Nevertheless, the activation energy deter-mined from the slope of the curve is 0. 68 eV whichis close to the activation energy (0. 66 eV) for lith-ium diffusion in oxygen-lean silicon measured byPell (Ref. 4) and also reported (Ref. 5) in mea-surements made on solar cells.

C. Long-Term Annealing Effects

Results reported (Ref. 6) on the long-termannealing effects in lithium-containing quartz-crucible and float-zone samples with low lithiumconcentrations showed a disturbing property,namely, the continuous loss of carrier density(measured at low and high temperature) as afunction of time in irradiated samples annealingat room temperature. These samples were re-cently remeasured, and the carrier densities mea-sured as a function of time are shown in Fig. 4 forthree samples. Two float-zone samples weredoped to a low and a high concentration, respec-tively, and the quartz-crucible sample containeda high concentration of lithium. It appears thatthe runaway loss of carriers has finally ceased tooccur in the quartz-crucible (>2 yr) and low-lithium-density sample of float-zone silicon. Thus,the precipitation-like process which produces acharge-neutral defect complex and which was pre-viously postulated (Ref. 6) as one of the annealingmechanisms appears to equilibrate. The behaviorof the other sample of float-zone silicon is typicalof the more heavily doped samples, namely, aninitial dissociation of defects with an increase ofcarrier density, and then they reach equilibriumover the long-term period.

III. DISCUSSION OF RESULTS

The irradiation temperature dependence of theproduction rate of impurity complexes in siliconhas been explained (Refs. 1, 2) by a model basedon the formation of close-spaced interstitial-vacancy pairs by electron bombardment. Thismodel yields a probability PC, of vacancy libera-tion from the interstitial to form vacancy-impuritydefects described by

PC = 1 + g exp · ·F · (1)C [F KT

where g is the ratio of the number of ways thestate can be occupied to the number of ways the

state can be unoccupied, EF(n0, T) is the Fermilevel for electrons which is a function of initialcarrier concentration n o and temperature T, andEM is the energy level of the metastable interstitial-vacancy pair. Equation (1) was used in an attemptto fit the data of Figs. 1 and 2. The preliminaryresults indicate that there is not a unique pair ofvalues for the parameters g and EM which willsatisfactorily fit the data obtained on the float-zonesamples over the three orders of magnitude rangein lithium concentration. The best value of Eappears to be =0. 09 eV, but several values of gare required to fit the entire range of concentra-tions. In addition, the values of g are small (aslow as 10- 3 ) and thus not physically reasonable.This model does not appear to be the completedescription of defect production mechanisms inlithium-doped float-zone silicon over this range ofconcentrations.

The attempts at curve fitting were more suc-cessful with the quartz crucible samples. In thiscase, EM = 0. 07 eV and g = 0. 1 gave a reason-able fit over the two orders of magnitude in lithiumconcentration of these samples. This higher valueof g is more physically reasonable, and the lowervalue of EM is in agreement with the value used byprevious investigators (Ref. 1) in fitting their dataon phosphorus-doped silicon. Our investigationshave shown (Ref. 4) that the oxygen-vacancy defectis the dominating carrier-removal center in thelithium-containing silicon. Thus, it is reasonableto expect that these samples would behave simi-larly to the phosphorous-doped (5 X 1020 P/M3 ,p = 10 Q2-cm) quartz-crucible silicon as our dataindicates.

The mobility recovery shown in Fig. 3 demon-strated again in agreement with the detailed mea-surements obtained on solar cells (Ref. 5) that thediffusion of free lithium is definitely related to therecovery process. This conclusion applies to theshort-term recovery period that is within the firstfew minutes or hours following bombardment.

The long-term annealing effects shown inFig. 4 are important, since they indicated that thecarrier density changes will finally equilibrate inlightly doped float-zone silicon or in the moreheavily doped quartz-crucible silicon after a suf-ficiently long annealing time. This behavior isobserved only under the conditions of these irradi-ations, namely the damage level was not largecompared to the lithium concentrations. Theauthor and other investigators (Ref. 7) have ob-served that after large fluences, the loss of car-riers does not equilibrate, and the samples willreturn to the starting resistivity of the siliconbefore diffusion with lithium. This applies to bothfloat-zone and quartz-crucible silicon.

IV. CONCLUSIONS

The simple close-pair vacancy-interstitialmodel does not completely describe the dependenceof defect introduction on temperature and resistiv-ity in lithium containing float-zone silicon. Betteragreement of the model with experimental resultsis obtained in samples of lithium-containing quartz-crucible silicon. Further evidence of the relation-ship between diffusion of free lithium and recoveryof damage was obtained. This was deduced fromthe activation energy for recovery of 0. 68 eV mea-sured in lithium-containing float-zone silicon.


Equilibrium and stability of the carrier density ispossible after long-term annealing at 300K inirradiated quartz-crucible (>2 yr) or lightly doped(2 to 5 X 1020 Li/m 3 ) float-zone silicon (<1 yr).

REFERENCES

1. Stein, H. J., and Vook, F. L., Phys. Rev.,Vol. 163, p. 790, 1967; also in RadiationEffects in Semiconductors, pp. 99 and 115.Edited by F. L. Vook. Plenum PublishingCorp., New York, 1968.

2. Gregory, B. L., and Barnes, E. E., inRadiation Effects in Semiconductors, p. 124.Edited by F. L. Vook. Plenum PublishingCorp., New York, 1968.

3. Brucker, G. J. , et al. , Third QuarterlyReport, RCA, under JPL Contract No. 952555April 10, 1970.

4. Pell, E. M. , Phys. Rev., Vol. 119, p. 1222,1960; see also J. Appl. Phys. Vol. 32, p. 61961.

5. Brucker, G. J., et al., Fourth QuarterlyReport, RCA, under JPL Contract No. 952555July 10, 1970.

6. Brucker, G. J., IEEE Trans. Nucl. Div.,Vol. NS-17, p. 144, 1970.

7. Goldstein, B., Final Contract Report, RCAunder Contract No. F19628-68-C-0133, July 1968.


3 5 7 9 11

103 /T B,K-

- 2x1 20 Li/M3

\ 5x 1020 Li/M3

\ 3 x 102 1 Li/M3

2 ), 1022 Li/M3

\ 3 x 1023 LU/M3

13 15

Fig. 1. Probability of defect formation versusreciprocal bombardment temperaturefor electron irradiated float-zone siliconmeasured at 79 K after annealing to200 K. Initial lithium concentration isthe curve parameter

100

ZO

02

OU

wLI-

bU

>.-

0

coea-

10 - 1

10-2

2 x102 1 L/JM3

5 6 7 8 9 10 11 12 13 14

103 /TB, K-

1

Fig. 2. Probability of defect formation versusreciprocal bombardment temperaturefor electron irradiated quartz-cruciblesilicon, measured at 79 K after anneal-ing at 200 K. Initial lithium concentra-tion is the curve parameter


100

10-1 FLOAT-ZONESILICON

ZO

O

UU-

0

a-op!

I I I I I l

38

FLOAT-ZONE SILICON

10-2

q ' - \ EA = .68 eV

10- 3

10- 4 i i t I I 3.0 3.2 3.4 3.6 3.8 4.0 4.2

103 /TA, K- 1

Fig. 3. Reciprocal half recovery time of themobility in electron irradiated float-zone silicon versus the reciprocalannealing temperature. Measure-ments of mobility at 79 K. Activationenergy for the recovery process isindicated. Initial lithium concentra-tion 3 X 1023 Li/M 3

C-,_

LU4

UC3-c.,

0C

1 .7

1.5

c-E

1.3wL

1.1 o04

OC

0.9

0.7

100 101

TIME AFTER IRRADIATION, days

Fig. 4. Carrier density versus time after electron irradiation for samples of float-zone andquartz-crucible silicon, measured at 300 K. Initial lithium concentration isindicated for each sample


10- 1

39

J

Density and Fluence Dependence of Lithium Cell Damageand Recovery Characteristics

T. J. Faith, Jr.RCA Astro Electronics Division

Princeton, N. J.

ABSTRACT

Experimental results on lithium-containingsolar cells point toward the lithium donor densitygradient dNL/dw as being the crucial parameterin the prediction of cell behavior after irradiationby electrons. Recovery measurements on a largenumber (180) of oxygen-rich and oxygen-leanlithium cells fabricated by all three manufacturershave confirmed that cell recovery speed is directlyproportional to the value of the lithium gradienrt forelectron fluences ranging from 3 X 1013 e/cm ~ to3 X 1015 e/cm 2 . An approximate relationshipbetween the time to half recovery, 0, (at roomtemperature) the lithium gradient, dNL/dw, andthe l-MeV electron fluence 1' was derived foroxygen-rich cells:

8 dNL/dw = 2.7 X 1012- 0.57 days/cm4

which holds for the entire range of fluences tested.For oxygen-lean (Centralab and Heliotek) cells therelation

e dNL/dw = 8.5 X 108 0. 61 days/cm4

holds up to 1 X 1015 e/cm 2 , above which a morerapid increase with fluence occurs, probably dueto the greater significance of lithium depletionduring irradiation. A similar relationship holdsfor oxygen-lean Texas Instruments cells, but witha proportionality constant approximately ten timesthat of the cells from the other manufacturers.


Seventy oxygen-rich (C13) cells with initialperformance significantly above a group of com-mercial 10 R-cm n/p cells were recently irradi-ated to fluences ranging from 3 X 1013 to 3 X 1015e/cm2 . Through pre-irradiation capacitancemeasurements (giving dNL/dw) and pre- and post-irradiation short-circuit current and diffusionlength measurements, a relationship was derivedbetween the diffusion length damage constantbefore recovery, KL(O), and dNL/dw and O.

KL(O) = 5. 3 X 101

8 (dNL/dw)/ (1 - 0. 063 log1

0 D)

Gradient measurements have also been cor-related with lithium diffusion schedules. Resultshave shown that long diffusion times (>_5 h) witha paint-on source result in large cell-to-cellvariations in gradient, probably due to a loss ofthe lithium source with time. They also indicatethat this problem can be overcome either byshort diffusion times or by use of an evaporatedlithium source.

I. INTRODUCTION

The principal performance parameters char-acterizing lithium cell behavior in a radiation en-vironment are: (1) the initial (pre-irradiation)performance levels, (2) the rate of cell degrada-tion or damage constant, (3) cell recovery rateversus cell temperature, and (4) final photovoltaicperformance after recovery from a given fluence.Under the present contract, experiments on solarcells irradiated by l-MeV electrons have beendesigned with the aim of eventually relating the

Preceding page blank[ 41

above performance parameters and the manufac-turers' fabrication parameters to a set of physicalparameters obtainable through non-destructivemeasurements on unirradiated cells. This wouldprovide the cell manufacturers with: (1) the opti-mum set of fabrication parameters for a given cellapplication, and (2) a set of quality-control teststhat can be integrated into the production line.

In earlier work on this contract the dependenceof recovery rate on cell temperature was obtainedfor both oxygen-rich and oxygen-lean lithium cellsthrough measurements of activation energy fordiffusion-length recovery (Ref. 1). For a givencell temperature, the recovery rate of short-circuit current in the 1014 e/cm 2 fluence rangewas shown to vary linearly with the lithium densitygradient (Ref. 2) for both oxygen-rich and oxygen-lean cells. In recent work: (1) recovery rateversus lithium gradient measurements were ex-tended to cover the fluence range from 3 X 1013 to3 X 1015 e/cm 2 ; (2) a relationship betweendiffusion-length damage constant KL, lithium gra-dient dNL/dw, and 1-MeV electron fluence 4 wereobtained through measurements on seventy irradi-ated cells of lot C13, and (3) pre-irradiation capac-itance measurements on one hundred C13 cellsestablished relationship between fabrication param-eters (temperature and time of lithium diffusion)and the degree of control of the lithium densitygradient, dNL/dw, near the junction.

II. INITIAL CELL PERFORMANCE

A shipment of one hundred high performancelithium cells fabricated from quartz-cruciblesilicon (lot C13) were recently received from JPL.They consisted of ten groups (C13A to C13J), tencells per group, with varying lithium diffusiontemperatures and diffusion times. Lithium densitygradients were obtained from reverse-bias capaci-tance measurements (Ref. 3) on all of the C13cells. Some of the individual groups showed largecell-to-cell variations in gradient. This findingand its relationship to the individual cell groupswill be discussed later in the paper; in this sectionthe cells are divided into three batches accordingto lithium gradient, without distinctions betweengroups. The three photovoltaic parameters, maxi-mum power, Pmax; short-circuit current, I sc;and open-circuit voltage, Voc, are plotted inFigs. 1, 2, and 3, respectively. For these figuresthe cells were divided into batch 1, consisting of20 cells with gradient ranging from 0. 65 X 1018 to3. 3 X 1018 cm-4 ; batch 2, with 40 cells rangingfrom 3. 3 X 1018 to 9.0 X 1018 cm- 4 ; and batch 3,with 40 cells ranging from 9. 0 X 1018 to 1. 6X 1019 cm-4 . 1

The data points appear along the abscissa atthe average value of dNL/dw for the batch and thefive ordinates represent (from top to bottom) thehighest value, the values exceeded by 20, 50, and80% of the cells, and the lowest value for the batch.Equivalent values for a batch of 20 commercial10 Q-cm N/P cells are shown on the left in eachfigure. The general trends show the power andshort-circuit current to remain approximately

constant over the first two batches, then drop withbatch 3, while the open-circuit voltage increasesmonotonically with gradient. The maximum powerof all lithium batches exceeds that of the n/pbatch, most of this advantage being due to thehigher Voc which is due, in turn, to the heavierbase doping in the lithium cells.

III. DAMAGE CONSTANT

Seventy of the C13 cells were irradiated with1 MeV electrons; seventeen cells to a fluence of3 X 1013 e/cmz, nine to 1 X 1014 e/cm2 , twentyto 3 X 10 1 4 e/cm 2 , eight to 1 X 1015 e/cm2 , andsixteen to 3 X 1015 e/cm 2 . The cells were chosenso that seven cells from each group were irradi-ated, and so that the cells irradiated to 3 X 1013e/cm 2 , 3 x 1014 e/cm 2 and 3 X 1015 e/cm2

covered the widest possible range of lithium gra-dients. Room-temperature2photovoltaic charac-teristics under 140 mW/cm tungsten illuminationwere taken on all cells immediately after irradi-ation. The cells were then stored at 80°C torecover.

A number of interesting results were obtainedfrom the short-circuit current readings madeimmediately after irradiation (before recovery).This current is plotted in Fig. 4 versus lithiumdonor density gradient, dNL/dw, for the seventyirradiated cells. Within the scatter of the data,the points at each fluence fit well along a straightline, all five lines having approximately the sameslope. A best least-squares fit to the equation

Isc = IA - B logl0 (10-18 dNL/dw) (1)

was calculated for each fluence.

The values of B were found to be 8. 4, 8. 7,8. 4, 8. 5 and 8. 9 for 4 = 3 X 10 1 3 , 1 X 1014,3 X 1014, 1 X 10 1 5 , and 3 X 1015 e/crm2, re-spectively. These equations give a specific rela-tion between the amount of initial damage (beforerecovery) and the lithium gradient in crucible-grown cells. It is advantageous to describe thisdamage in terms of a more standard quantity,namely the diffusion-length damage constant be-fore recovery, KL(O), given by

1 1KL(O)D = 2- 2

L(O) L0

(2)

where Lo and L(O) are the diffusion lengths inthe base region of the cell before irradiation andimmediately after irradiation, respectively. Toobtain KL(O), the relationship between Isc and Lmust be known; Fig. 5 gives such a plot for C13cells. The data for this figure was generatedfrom simultaneous short-circuit-current and dif-fusion length2 measurements (Ref. 4) made on theC13 cells, 30 of which were unirradiated, the other

1 Batch 1 has the smaller number of cells because of the comparatively small fraction of cells in the lowgradient range.

2 Obtained from measurements using band-gap light that was calibrated with the electron-voltaic methodusing 15 different lithium and N/P solar cells.


70 being at various stages of recovery after irradi-ation. The best fit to this data is

Isc = 34.4 log 1 0 L (3)

For fluences above 3 X 1013 e/cm 2 , 1/L(O)2 >>1/L6 in Eq. (2) and the latter term can be droppedwith less than 10 percent error. This approxima-tion, combining Eqs. (1), (2) and (3), and usingB = 8. 6 in Eq. (1), gives

0- (+IA/ /17.2) dNLKL(O) = A (4)~L 4) V dw

which is valid for all of the fluences employedexcept D = 3 X 1013 e/cm2 . For this lowest fluenceKL(O) was calculated taking Lo into account. Theresult, which is shown in Fig. 6, was

KL(O) = 8.5 X 10 19 /dNL/dw (5)

Equations (4) and (5), together with the appropriatevalues of IA listed in Fig. 4, give the fluence de-pendence of KL(O), which is plotted in Fig. 7 fordNL/dw = 101 cm-4. Figure 7 shows a logarith-mic dependence on fluence described (for dNL/dw= 1018 cm -

4 ) by

KL(O) = 5.3 X 10-9(1 - 0.063 log10 D) (6)

Inserting the lithium gradient dependence, theexpression for KL as a function of 'p and dNL/dwis given by

KL(O) =

5. 3 X 10-18 (dNL/dw)l/z (1 - 0. 063 log1 0 D) (7)

The applicability of this relationship to othercrucible cells was tested using the post-irradiationdata on thirteen previously irradiated cells fromlot H3A. These H3A cells covered a wide range ofgradients: 3 X 1017 < dNL/dw s 1. 3 X 1019 cm-4 ;they were irradiated to a fluence of 3 X 1014 e/crn2 .Figure 8 shows KL(O) plotted against dNL/dw forthese cells. The square-root dependence on gra-dient is evident, the best least-squares fit beingobtained with the relationship

KL(O) = 4.4 X 10 -1

9 (dNL/dw)l/2 (8)

3 Lots T3, T4, T5, T6, T7, T9, T10, H5, H7, H1A,C8F, C1OC, C1OF, and CllC were represented.

4 Trademark of Texas Instruments Corp.

This is within 10% of the value of 4. 8 X 10 - 1 9

(dNL/dw)l/ 2 obtained for the C13 cells, as can beseen in Fig. 7, where the H3A data point is showntogether with the C13 data.

The C13 cells are now recovering at 80°C.At completion of short-circuit current recovery,the effective damage constant, KL(R), after cellrecovery can be computed from

1 1KL(R), = 2 2

L(R) L00

(9)

where L(R) is the diffusion length after recovery.There is a large uncertainty in KL(R) for lowfluences since I(R) - Io. Results at 3 X 1013 e/cm2

show a very large scatter, KL(R) ranging from0.5 X l0-1 to 2.0 X 10-10 e-1. The results at1 X 10-14 e/cm2 are more coherent and show aslight increase with lithium gradient, KL(R

{

ranging from 1.0 X 10- 10 to 1. 6 X 10-l e- forgradients ranging from 4 X 1018 to 1. 6 X 1019cm 4.

IV. RECOVERY CHARACTERISTICS

Prior to the tests on the C13 cells, a largenumber of previously unirradiated oxygen-leancells from past shipments were gathered andirradiated to fluences ranging from 3 X 1013 e/cm 2

to 3 X 1015 e/cm 2 . Cells from Texas Instruments,Heliotek and Centralab3 of both Lopex4 and float-zone silicon were represented. They included awide range of diffusion schedules and initial per-formance levels. After irradiation, short-circuitcurrent was measured as a function of time on allof the cells. The purpose of the experiment wasto test the validity of the previously observed(Ref. 2) linear relationship between recoveryspeed and lithium density gradient for a largebatch of cells covering the widest possible rangeof lithium gradients and a wide range of fluences.

A typical short-circuit current versus timecurve during recovery is shown in Fig. 9. Thetime to half recovery, 0, defined by

I(R) - I() _ 0. 5I(R) - I(O) -

(10)

where I(R) is the short-circuit current at peakrecovery, provides the most well-defined indexof (inverse) recovery rate. For the cell in Fig. 9,O = 55 min (or 0. 038 days). The values of 0 forthe Centralab and Heliotek cells are plottedagainst lithium density gradient in Fig. 10 forfour fluences ranging from 3 X 1013 to 3 X 1015e/cm 2 . Included are all the cells tested fromthese manufacturers except those of lots C4 andC5 and those with lithium gradients greater than1020 cm -

4 . The points on these logarithmic plots

H5A, H7A, H (NASA-furnished in 1967), C4, C5,


fit remarkably well along straight lines with minusone slope, confirming the linear relationship be-tween recovery speed and lithium gradient. The6 dNL/dw products (averaged over the cells ateach fluence) are: 1. 7 X 1017 days/cm4 for3 X 1013 e/cm 2 ; 3.4 X 1017 days/cm4 for 1 X 1014e/cm 2 ; 7. 2 X 1017 days/cm4 for 3 X 1014 e/cm2 ,and 6. 8 X 10 1 8 days/cm4 for 3 X 10 1 5 e/cm2 .There are two cell lots that do not follow thesecurves, lots C4 and C5, which recover at a fasterrate than the curves predict. However, these twolots had already been identified as mavericks inprevious work (Ref. 2), having been shown to suf-fer open-circuit voltage instability due to a de-crease in lithium gradient with time. In addition,cells with gradients greater than 1020 cm-4

(C1OC and C10OF cells) recovered more slowly atlow fluence than predicted by the curves. Afeasible explanation for this is that the gradient inthese cells is not a good index of the averagelithium density near the junction. The capacitancemeasurements on these high gradient cells giveevidence of this in the form of a leveling off of thelithium density; i. e., a decrease in gradient, atdistances of less than 1 pam from the junction.

Curves of 0 versus dNL/dw for oxygen-richcrucible cells are shown in the upper portion ofFig. 10. The data are drawn from previous re-sults on cells that recovered at room temperature,60 0 C, or 80°C. (In the latter two cases equiva-lent recovery time at room temperature was cal-culated using the activation energy previouslyobtained (Ref. 1) for crucible cell recovery. ) At1 X 1014 e/cm2 and 3 X 1014 e/cm 2 the separationbetween the oxygen-lean (FZ and L) curve and thecrucible curve is z700, which is approximatelythe ratio of the room-temperature lithium diffusionconstant in oxygen-lean silicon to that in oxygen-rich silicon (Ref. 5). At the highest fluence,3 X 1015 e/cm2 , the separation is only =250.This suggests that lithium is lost more rapidly indefect formation in oxygen-lean cells than inoxygen-rich cells, supporting previous carrierremoval observations (Ref. 1) in bulk-samplemeasurements.

A puzzling anomaly was observed in the caseof the oxygen-lean T cells. While the constancyof the 0 dNL/dw product was satisfied by thesecells at each fluence as shown in Fig. 11, theproducts were approximately an order of magni-tude higher than the equivalent products for theCentralab and Heliotek cells. This can be seenby comparing the products in Fig. 11 with thoseof Fig. 10. This discrepancy is not understood atpresent. One of the main differences between themanufacturers is that the TI cells have used anevaporated lithium source whereas all of theoxygen-lean C and H cells tested to date have useda paint-on source. It would seem likely that adifference in silicon type would cause differencesin recovery rate. This would be particularlyfeasible for Lopex versus float-zone recoverysince the oxygen-content is generally higher inLopex silicon. However, as was shown in Fig. 10,the same recovery rate applies for both Lopex andfloat-zone C and H cells. Dislocation counts arenow being made on some of the cells in an effortto confirm the silicon types employed.

The approximate linear dependence of re-covery speed on lithium gradient at all the fluencestested enables prediction of the recovery speed of

any lithium cell with lithium gradient between 1017and 1020 cm - 4 in this fluence range. This isillustrated in Fig. 12, which gives plots of the 0dNL/dw products of all the cells tested versus1-MeV electron fluence.

Below 1015 e/cm 2 this product increasesgradually with fluence in all types of cells. Inoxygen-lean cells there is a more pronounceddependence on fluence above 1015 e/cm2. In theoxygen-rich cells the relation

o dNL/dw 2. 7 X 1012 0. 57 days /cm 4

(11)

provides a good approximation of cell recoveryspeed over the entire range of fluences.

V. LITHIUM DIFFUSION SCHEDULE ANDDENSITY CONTROL

From the above and previous results, it isevident that many of the characteristics of lithiumcells under electron irradiation can be predictedthrough knowledge of the lithium density gradient,dNL/dw. Consequently, it is desirable to findrelationships between this parameter and thefabrication parameters of the cell manufacturers.This has been accomplished for cell lots C13 (100cells) and H3A (15 cells). Lot C13 consists of10 groups of 10 cells, comprising nine differentlithium diffusion schedules with diffusion tempera-tures ranging from 330°C to 370°C and diffusiontimes from 3 to 7 h. The lithium source was alithium-in-oil suspension painted on the back sur-face of the cell. Figure 13 gives the distributionof lithium gradients measured for the cells ineach of the 10 cell groups. There are five sepa-rate ordinates, each running from 1 to 10. Thevalue of the ordinate at a given lithium gradientindicates the number of cells of that particulargroup with a lithium gradient greater than thevalue of the abscissa. A pair of cell groups shareseach ordinate, since two groups were diffused ateach lithium diffusion temperature. Each groupis identified by a letter followed by a number inparentheses which gives the lithium diffusion timein hours.

Several important factors are brought out inFig. 13: (1) for a given diffusion temperature,the shorter diffusion time gives a narrower gra-dient distribution; i. e., better gradient control;(2) the gradient distribution for the shorter diffu-sion time is always situated near the upper limitof that for the longer diffusion time; (3) at thehighest diffusion temperature, even the short timediffusion shows a rather broad distribution; and(4) the average gradient (for short diffusion times)increases with diffusion temperature. Items (1),(2), and (3) indicate that as the diffusion time in-creases the lithium reservoir is somehow lost tothe cell, either through lithium depletion or throughinterruption of the lithium-silicon interface.Therefore, for lithium introduction by the paint-ontechnique, the shortest practical diffusion timeshould be used.

A previously received cell lot, H3A, consistedof 15 quartz-crucible cells diffused with lithiumfor 8 h at 325°C. Ten of the cells used a paint-onsource; the other five, an evaporated lithium


source. The cell distribution versus lithium gra-dient is shown in Fig. 14. The cells using thepaint-on source have a very broad distributionsimilar to those of the C13 cells using long diffu-sion times. The cells using an evaporated source,however, have a narrow distribution at the highend of the gradient scale. This supports the hypoth-esis of loss of the lithium source in the paint-oncells and also indicates that an evaporated sourcemay provide the solution to this problem.

VI. CONCLUSIONS

Previous work has shown the lithium densitygradient, dNL/dw, obtained from non-destructivecapacitance measurements, to be a convenient anduseful way to characterize the lithium density in alithium cell by a single parameter. The presentwork has investigated the relationship between thisphysical parameter and the performance and re-covery parameters of the lithium cell. Resultsshow that knowledge of the lithium gradient enablesthe prediction of recovery speed for both oxygen-rich and oxygen-lean cells within a factor ofapproximately 2 for 1 MeV-electron fluences from3 X 1013 to 3 X 1015 e/cm 2 .

A relationship between the diffusion-lengthdamage constant immediately after irradiation(before recovery) KL(O), the lithium gradient,dNL/dw, and the electron fluence, 4, has beenobtained for C13 crucible cells:

recovering, and this will be investigated uponcompletion of recovery.

These relations make it possible, in principle,to predict cell behavior in an electron environmentusing a simple non-destructive capacitance mea-surement. It is evident that a similar approachto predicting cell behavior under heavy particleirradiation should also be examined.

Gradient measurements have also been corre-lated with lithium diffusion schedules. Resultshave shown that long diffusion time (>_5 h) with apaint-on source result in large cell-to-cell varia-tions in gradient, probably due to a loss of thelithium source with time. The results also indi-cate that this problem can be overcome either byshort diffusion times or by use of an evaporatedlithium source.

REFERENCES

1. Brucker, G., et al., Fourth QuarterlyReport, RCA under JPL Contract No. 952555,July 10, 1970.

2. Faith, T., Brucker, G., and Holmes-Siedle,A., Sixth Quarterly Report, JPL ContractNo. 952555. Prepared by RCA and issuedJan. 10, 1971; see also Faith, T., Corra,J. P., and Holmes-Siedle, A., ConferenceRecord of the Eighth Photovoltaic Spec. Conf.,IEEE Catalog No. 70C 32 ED, p. 247, 1970.

KL(O) = 5.3 X 10 - 1 8(dNL/dw)l/2 (1 - 0.063 1 og

1 0, )

A check on previous post-irradiation data showsthat this relationship also holds for H3A cellsirradiated to 3 X 1014 e/cm 2 . One additionalrelationship, that between the damage constantafter recovery KL(R), dNL/dw, and 4, is requiredto complete the description of cell dynamicsunder electron irradiation. The C13 cells are now

3. Hilibrand, J., and Gold, R. D., RCA Rev.Vol. 21, p. 245, 1960.

4. Rosenzweig, W., Bell Sys. Tech. J.,Vol. 41, p. 1573, 1960.

5. Pell, E. M., J. Appl. Phys. Vol. 31, p. 291,1960; Vol. 32, p. 6, 1961; and Phys. Rev.Vol. 119, p. 1222, 1960.


PERCENT CELLSWITH POWER >P

0%

20%

C13 CELLS140 mW/cm2

TUNGSTENILLUMINATION

*BATCH 1 (20 CELLS)

0 BATCH 2 (40 CELLS)

0 BATCH 3 (40 CELLS)

1017 1018 1019 1020

LITHIUM DONOR DENSITY GRADIENT, dNL/dw (cm-4 )

Fig. 1. Initial maximum power distributions for C13 quartz-crucible lithium cells versus lithium donor densitygradients with comparisons of 10 0-cm commercialN/P cells

1118 111910THIUM 10DNOR DENSITY GRADIENT, dNdw (-4)

LITHIUM DONOR DENSITY GRADIENT, dNL/dw (cm- 4 )

10 2 0

Fig. 2. Initial short-circuit current distributions for C13 quartz-crucible lithium cells versus lithium donor densitygradients with comparisons of 10 Q-cm commercialN/P cells


34

33 h

E32

w:

31 j 10 c mn n/p (20 CELLS) 50%2

X 30

<

< 29I-

28

27

- 0%

- 20%

- 50%

-- 80%

- 100%

80%0

100%

BATCH 3

| BATCH I - -

BATCH 2 V -

E

LiZuJ

0rE,

:rLOI

1017

46

5o

r.- -1 I

PERCENT CELLS WITHOPEN-CIRCUIT VOLTAGE >Vnr

H

O _

C13 CELLS140 mW/cm 2

TUNGSTEN ILLUMINATION

BATCH 1 (20 CELLS)

O BATCH 2 (40 CELLS)

O BATCH 3 (40 CELLS)

10 l2cm n/p (20 CELLS)--0%

BATCH 1 BATCH 2 BATCH 3

loi8 101910THIUM 18 1DENSITY GRADIENT N

LITHIUM DONOR DENSITY GRADIENT, dNL/dW (cm-

4 )

Fig. 3. Initial open-circuit voltage distribution for C13 quartz-crucible lithium cells versus lithium donor densitygradients with comparisons of 10 i-cm commercialN/P cells

I 63EU 60

1- 55

0

a 50

rLU

LU

r3m 40

35

Cl 30

28

1, 1018 1019

LITHIUM DONOR DENSITY GRADIENT, dNL/dw (cm-

4 )1020

Fig. 4. Short-circuit current immediately after irradiationversus lithium density gradient for seventy C13cells irradiated by 1-MeV electrons to fluencesranging from 3 X 1013 e/cm2 to 3 X 1015 e/cm 2


E 620EU0> 610

i(9

) 600

- 590

U

9 580

z

0.O 570

- 560Z

5501 C

O _

--20%-- 50%

-- 80%--100%

017 1020

117

n

47

70-

E 7

65

zZ *Zwa 60

(C13 CELLS)H

0

55 ,- · iSC = 34.4 LOG10 L

V)

10 20 40 o60 0 100

DIFFUSION LENGTH, L (m)60 80 100

Fig. 5. Short-circuit current versus dif-fusion length for 100 crucibleC13 cells (30 cells unirradiated,70 cells at various stages of re-covery after irradiation)

z 10

0 8

w -Y

Z I

-I 2

0

U..O

0

01

zI , 10- 8

oLU

O

E-

o 7I 110-1

200

200 0 1017 10 1 8 1019

LITHIUM DONOR DENSITY GRADIENT, dNL/dw (cm - 4 )

Fig. 6. Diffusion-length damage constant im-mediately after irradiation (beforerecovery) versus lithium gradient forseventeen C13 cells irradiated to afluence of 3 X 1013 e/cm 2

1014 1015

1-MeV ELECTRON FLUENCE, 7 (e/cm 2 )

Fig. 7. Diffusion-length damage constant im-mediately after irradiation (at dNL/dw =1018 cm-4 ) versus 1-MeV electronfluence for C13 and H3A cells


C13 CELLS

(0= 3x 1013 e/cm 2

KL = 8 -19(dN L/dw)'

102 0

48

lo - lo I I I III I I I I I I I i17 1018 1019 020

LITHIUM DONOR DENSITY GRADIENT dNL/dw (cm- 4 )

Fig. 8. Diffusion-length damage constant im-mediately after irradiation versuslithium gradient for H3A cells irrad-iated to a fluence of 3 X 1014 e/cm 2

1020

-J

0

1019

101.1

_o 10 '2 11 7x1(3.4 o 1 7)

Z (1.7 x 101 7)

3 18101 ec 2

D °6 3 10145e/c, 2 0

0 32 1015 e5 c 2

1I -2 1010>. o2 05 /m

TIME T

65

E:60

Z 55Urrc:

D 50

() 45Cr

F- 40r3

IX3 3510 - 1 100 101 102 103

TIME AFTER IRRADIATION, T (MINUTES)

Fig. 9. Typical short-circuit current recoverycurve illustrating the index of cellrecovery rate 0 used in subsequentillustrations

100 101 1o2

rO HALF RECOVERY, 0 (DAYS)

Fig. 10. Time to half recovery at room temperature versuslithium gradient for oxygen-rich cells and forCentralab and Heliotek oxygen-lean cells irradiatedto fluences ranging from 3 X 1013 to 3 X 1015 e/cm 2


z 108

h-EI z

"r

w

J

o

- I I I] I I I I i I i j

KL = 4.4 x 10-19 dN/dw)Y

H3A CELLS 0= 3 x 101 4 e/cm2

F,

U.

5

103

49

C 1020

E

-o

-Jz

zLU 1019

1:

ULzLU

- 10180

0

1017

10-1 100 101

TIME TO HALF RECOVERY, 0 (DAYS)

Fig. 11. Time to half recovery at room tem-perature versus lithium gradient foroxygen-lean Texas Instruments cellsirradiated to fluences ranging from3 X 1013 to 3 X 10 1 5 e/cm 2

E

-

Lu 1015

2

LL

0cc

-JwL

- I III

-dNL/dw = 8.5 x 1080 0.61

(O I x 1015 e/cm2 )

, OdNL/dw = 2.7 x 1012

I! /a/ _

.· o/O 0/ FLOAT-ZONEAND

/ LOPEX CELLS,

/ O T CELLS

yo/ . C AND H CELLS

I . . II I . . I . I I n 13 7i0 TM 1018 1019 1020 102 1

TIME TO HALF RECOVERY X LITHIUM GRADIENT, 0 dN (DAYS/cm 4 )dw

Fig. 12. Product of time to half recovery and lithium gradient,0 dNL/dw plotted versus 1-MeV electron fluence foroxygen-rich and oxygen-lean lithium cells


CRUCIBLE CELLSe

oq. a o

10 -2

' ''I I I I I I I ' I I ';

1022

50

T = 370'

T = 360'

T = 350'

T = 340'

I I IOC \1(4)

Jt6)

OC H(7)

OC F() )

OC

A 10

, . V~~ I

T = 330° C

1018

LITHIUM DONOR DENSITY GRADIENT, dNL/dw (cm- 4 )

Fig. 13. Distribution of cells versus lithium gradient for the tendifferent cell groups within Lot C13. Numbers inparentheses indicate duration of lithium diffusionin hours; T = diffusion temperature

1 .0

I-oos

n

j 0.6

- Z

o AZHO Z 0.4

,- 0wL. 0.2

0 L

1017 1018 1019 . 1020

Fig. 14. Distribution of cells versus lithium gradient for H3Acells using two different lithium sources, paint-onand evaporated


10

Io

0o

-1-

I-'EiB:

uJ zLIZ

z

w -Z(..

loT

10-

otI 0'F

0lo1017

3 CELLS., GROUPS A TO JCELLS PER GROUPARTZ CRUCIBLE

1020

C12

1019

51

IVI

DEGRADATION AND RECOVERY MECHANISMS INLITHIUM-DOPED SOLAR CELLS*

R. G. Downing and J. R. Carter, Jr.TRW Systems Group

Redondo Beach, California

I. INTRODUCTION

Several large groups of lithium-doped solarcells were received from JPL for radiation hard-ness evaluation. These cells included some of themost advanced devices produced to date. Theevaluations consisted of the determination of thelithium concentrations in the devices and the effectof irradiations with 3 X 1014 and 3 X 1015 e/cm 2

with l-MeV electrons. In addition some studieswere performed using 28-MeV electrons. Capac-itance measurements were used to systematicallystudy the changes in lithium concentration whichoccur during irradiation and recovery.

II. LITHIUM-DOPED SOLAR CELL EVALUATION

During the past year, lithium-doped solarcells from Centralab and Heliotek have been irra-diated with 1-MeV electrons and their recoverycharacteristics have been studied. Several dif-ferent processing techniques were represented inthese cells, including different diffusion gases andvarying percentage of lithium coverage of the rearsurface. Data for cell groups CllA through CllD,H1A, HZA, H4A, H5A1, H5AZ and H5A3, arelistedin Table 1. All of the cells received radiationexposure to 1-MeV electrons. Tungsten I-V char-acteristics and capacitance versus voltage mea-surements were then obtained as a function of timeat either room temperature or 60°C. The generalradiation damage and recovery characteristics ofeach cell group are summarized in Tables I and 2.The recovered levels given in the tables are the

peak of the recovery curve and do not take intoaccount any redegradation that may have occurred.In general, it can be observed that the higher lith-ium concentrations result in lower initial charac-teristics, higher recovered levels, and more rapidannealing rates, whereas with lower lithium con-centrations, higher initial and slower recoveryrates exist. It should be noted that most cellgroups tested were superior to the contemporaryN/P cells in recovered level. The initial short-circuit current of many of the cells studied wasinferior to the contemporary N/P cells.

A. Centralab Cells

The Centralab cells submitted for evaluationwere all fabricated from quartz-crucible-grownsilicon, with the exception of group CllC. Thevariable of boron dopant gas was investigated bydiffusing the P-type front surface with borontrichloride in the case of group C11A and borontribromide in the case of group CGlB. The short-circuit current values of cells received in groupC1lB were rather low (53 mA) compared to thatgroup CllA. This difference is not believed to berelated to the use of boron tribromide. It is knownthat when Texas Instruments manufactured solarcells, their P-type diffusions were made withboron tribromide. A second important differencebetween the cells of these two groups is concen-tration of lithium found at the junction. Althoughthe cells of both CllA and CllB groups were lith-ium diffused in the same manner (480 min at325°C), the data in Table 2 indicates that group

*This work was performed for the Jet Propulsion Laboratory, California Institute of Technology, spon-sored by the National Aeronautics andSpace Administration under Contract No. NAS 7-100.


Preceding page blank

CllA contains twice the lithium concentration ofgroup CllB. This difference is not considered tobe related to use of boron tribromide. The effectof this lower lithium concentration can be seen inthe 60°C recovery data in Table 2. The cells ofgroup Cl lA are nearly fully recovered 200 h afterirradiation; however, those of CllB appear to beonly half recovered at a comparable recoverytime. This slower recovery is probably due tothe lower lithium concentration. The recoveredIsc values of the CllA group cells are 38 mA afterirradiation of 3 X 1015 e/cm 2 . The value is sig-nificantly better than the comparable data for con-ventional N/P cells. Recovery in the CllB grouphas not progressed sufficiently to determine finalrecovered Isc value.

The cells of group ClD were lithium diffusedat the slightly higher temperature of 375°C for180 min. This diffusion schedule resulted in ahigher concentration of lithium at the junction.These cells have about 5 X 1014 lithium atoms/cm 3 . The CllA cells diffused at 325°C for480minhad only 3 X 1014 Li/cm 3 . It is also noted thatslope of the log capacitance versus log voltage is-0. 26. This is the lowest value found in the cur-rent group of cell under evaluation. The high lith-ium concentrations in group CllD are reflected ina slightly more rapid recovery after irradiation.The initial Isc values of the cells in groups CllAand CllD are comparable with those of conven-tional N/P solar cells. The cells of groups Cl lA,CllB, and CllD are interesting in that they havethe lowest lithium concentrations of any quartz-crucible-silicon cell evaluated by TRW. The rel-atively poor performance of cells in group C lBindicate that for this type of cell, the concentrationof lithium at the junction should be kept above2 X 101 4 /cm3 .

The remaining Centralab group, C11C, wasfabricated from Lopex (low oxygen, low dislocation)silicon. In general, the performance of the groupwas very good. The cells were lithium diffused at325°C for 480 min. The resulting lithium concen-tration in this group was 1.5 X 1014 /cm3 . Theresults of the 3 X 1014e/cm2 and 3 X 101 5e/cm2

radiations are shown in Table 1. In both cases,the Isc recovered to values greater than those ofcomparably irradiated N/P solar cells, as shownby the dashed lines on these graphs. Although therecovery kinetics are relatively slow in these cells,due to the lower lithium concentration, the irra-diated cell performance is among the best receivedto date.

The cells from the C13 series, manufacturedby Centralab, represent a matrix of lithium dif-fusion temperatures and times. The matrix isdesigned to investigate the optimum lithium dif-fusion and the reproducibility of the process. Allthe cells were fabricated from quartz cruciblegrown silicon with resistivities between 25 and40 Q-cm. The diffusion matrix is shownin Table 2.Also shown in the table are the mean donor con-centrations at the junction of the cells of eachgroup in the matrix. The voltage-capacitancerelationship or range of relationships is alsoshown for each group. The donor concentrationswere determined by means of capacitance. Sincethe phosphorus concentration is approximately1.5 X 1014 atoms/cm3 , the lithium concentrationswere estimated by subtracting the phosphorus con-centration from the donor concentrations. Since

the phosphorus concentration may go as high as2 X 10 1 4 atoms/cm3 (i. e. , 25 0-cm), somegroups may contain cells with very low lithiumconcentrations. Specifically these are the cellswhich received diffusions of 6 or 7 h. It can alsobe observed that both higher temperatures andlonger diffusion times result in greater variationsin lithium concentration.

It can be seen from the data in Table 2 that,in the time span studied, the lithium concentrationsat the junction are decreasing with time. Such adecreasing lithium concentration is inconsistentwith diffusion from an infinite source. Since thedonor concentration originally was equal to thephosphorus concentration, it was necessary forthe donor concentration to rise from the originalto a maximum before declining. This decrease inlithium concentration with diffusion time is char-acteristic of diffusion from a starved source. Asthe lithium source is exhausted, the surface con-centration of the cells will decrease and concen-trations at the junction will decrease. In the past,manufacturers of lithium solar cells have inter-rupted the lithium diffusion and removed the lith-ium source to produce starved source diffusion.This second diffusion has been referred to as aredistribution. The data in Table 2 indicate thatredistribution is not necessary. It is not clearwhat causes the exhaustion of the lithium source.

Three cells of each C13 series group wereirradiated with 3 X 1015e/cm2 (1 MeV) and wereallowed to recover at 60°C. The most pronouncedchanges during recovery occur in the short-circuitcurrent. The open-circuit voltages of nearly allcells in the C13 series were about 0.6 V. After a3 X 101 5 e/cm 2 irradiation, the Voc was reducedto about 0. 45 V in all cases. Very little recoveryis subsequently observed in this parameter. Thechanges in I c for cells of the C13 series areshown in Tagle 2. Several general observationscan be made about these cells. With a few excep-tions, the 3 X 1015e/cm2 irradiation reduced theIsc to 20 to 25 mA. In most cases, the time neces-sary for the Isc to reach the half recovery pointwas 100 to 200 h at 60'C. This time is somewhatlonger than what was previously found in similarcells. Except for cells with low lithium concen-tration, most cells ultimately recover to Iscvalues of 35 to 40 mA. The most important factorin the extent of the recovery is the lithium concen-tration at the junction rather than the diffusionschedule. As previously noted, the shorter low-temperature diffusions produce the more consis-tent results. The donor concentration of each cellirradiated is noted in Table 2.

B. Heliotek Cells

All Heliotek cells received for evaluation dur-ing the past quarter were fabricated from eitherfloating zone or Lopex silicon and, therefore, hadlower oxygen concentrations. There are two dif-ferent experimental variables represented in theseHeliotek cells. Two groups (H1A and H4A) werediffused at lower temperatures: 325°C lithiumdiffusion for 480 min. Group H2A was lithium diff-used at 425°C for 90 min with a 120-min re-distribution cycle. This latter diffusion schedulehas been used extensively in the past and can beregarded as control. The capacitance measure-ments results from the H1A group, shown inTable 1, indicate that very little or no lithium


reached the junction. For this reason the irra-diation recovery results shown in Table 1 are verypoor. Although some recovery is observed after3 X 1014 e/cm2, the higher fluence of 3 X 1015e/cm2 exhausts the lithium and no recovery isobserved. These results are in direct conflictwith those for group H4A which had an identicalhistory. The H4A cells had approximately5 X 1014 lithium atoms/cm3 at the junction, andexhibited satisfactory recovery as shown inTable 1. The recovered Isc values of the H4Acell would probably have been higher if the before-irradiation Isc values had been higher than 46 mA.This condition is not necessarily a result of thelithium diffusion, as other cells with similar lith-ium concentrations have initial Isc values inexcess of 60 mA. Despite this difficulty, the dataindicated that cells of group H4A recover to Iscvalues of 40 mA after a fluence of 3 X 1015e/cm 2 .This is considerably higher than a comparableirradiated N/P solar cell.

The irradiation recovery results for the cellsof group H2A are shown in Table 1. As mentionedpreviously, this lithium diffusion schedule has pre-viously been used many times to produce superiorlithium cells. The results in Table 1 confirm thatsuch cells exhibit excellent Isc values when re-covered from an irradiation. The results in thecase of the 3 X 1015e/cm2 fluence are particularlyinteresting in that the recovered Isc reached avalue of 44 mA. The fact that these cells werefabricated from Lopex silicon as opposed to float-zone silicon is not considered significant.

The remaining groups of Heliotek cells repre-sent a series of experiments to determine theeffectiveness of area coverage during the appli-cation of lithium diffusion source material to theback of the cell. Groups H5A1, H5A2, and H5A3,respectively, received 100, 80, and 50% back sur-face area coverage. The results of this experi-ment are very interesting for comparative analysis.The first point of interest is the measured lithiumconcentrations at the junctions of the variousgroups as seen in Table 1. The cells with 100%coverage (H5A1) have approximately 6 X 1014 lith-ium atoms/cm 2 at the junction. The groups whichreceived less coverage (H5A2, H5A3) had roughlyhalf the above lithium concentration. The resultsindicate quite clearly that decreased area coveragereduces the concentration of lithium at the junction.The relationship does not appear to be linear,since the cells with 80% coverage (H5A2) have lith-ium concentrations nearly as low as those with50% (H5A3). It can be concluded that incompletearea coverage with the lithium source materialsignificantly reduces the lithium concentration atthe junction. It is also of interest to compare thecells of these groups to cells of other groups. Thecells of group H2ZA were made with the same mate-rial and diffusion schedule, but presumably no con-trol on area coverage. The data in Table 1 indi-cate the H2A cells had much lower lithium concen-trations than any of the cells in H5 groups. Itmust be concluded that there are other unknownfactors which extend strong influences on the con-centration of lithium reaching the junction. Onepossible factor could be the chemical activity ofthe lithium in the source material.

The effects of various lithium source areacoverages on radiation response can be seen inTable 1. The initial Isc values of these cells are

all relatively low. The values average approx-imately 51 mA. This parameter influences radi-ation recovery behavior, because the maximumrecovered parameters can only approach and notexceed their initial values. Despite this problem,the cells of group H5A1 (100% coverage) recoveredto a maximum Isc of 50 mA after an irradiation of3 X 10 1 4 e/cm and 39 mA after 3 X 1015e/cm2 . Inboth cases these values are above those of simi-larly irradiated conventional N/P solar cells.The H5A1 cells, irradiated with 3 X 1014 e/cm 2 ,show some redegradation of Isc after the maxi-mum was reached. The radiation recovery ofcells of group H5A2 (80% coverage) was not dras-tically altered by the reduced coverage. The cellsof group H5A2 which were irradiated with 3 X 1014e/cm3 recovered to Isc values of 45 mA. Therecovery probably would have exceeded the abovevalue if the initial Isc value had been greater than46 mA. The 3 X 1015e/cm2 irradiation of H5A2cell allowed the Isc recovery to 39 mA after irrad-iation. This value is equal to that achieved in thegroup having 100% coverage (H5A1). This resultis difficult to explain, considering the lower lith-ium concentration and the studies of D. L. Kendallat Texas Instruments which indicated very littlelateral spreading of lithium during diffusion. Theresults for cells of H5A3 (50% coverage) showcomparable performance after 3 X 10 1 4 e/cm2

(Table 1). In the case of the higher electron flu-ence on the H5A3 cell, the recovered Isc valueswere significantly reduced. It appears that incom-plete coverage with diffusion source material doesnot reduce recovery behavior, except in extremecases.

III. ELECTRON ENERGY DEPENDENCE OFLITHIUM-DOPED CELL RECOVERY

Most of the studies of radiation behavior oflithium-doped cells have been done with a 1-MeVelectron environment. With the exception of a fewstudies with reactor neutrons and one early studywith protons, there is a complete lack of much ofthe data needed to accurately predict behavior inspace. One area in which more data are desirableis response of these cells to electrons with ener-gies greater than l-MeV. Through the courtesyof Dr. J. A. Naber of Gulf Radiation Technology,several lithium-doped cells were irradiated with28-MeV electrons.

Because of the necessary delays between theirradiation at the Gulf facilities and analysis atTRW, cells made from quartz-crucible-grownsilicon were selected because of their slow recov-ery rate at room temperature. This allowed thecells to be mailed to TRW before any post-irradiation recovery occurred. The cells used inthis experiment were from the CllD series,manufactured by Centralab. These cells are con-sidered to be typical of good lithium-doped cellsand adequate 1-MeV electron data had previouslybeen obtained. Cells were irradiated with 3 X 1014and 3 X 10 1 5 e/cm2 . The Isc recovered at 60°C to46 and 31 mA, respectively, for the two fluences.The degraded and recovered Isc values as a func-tion electron fluence are shown in Fig. 1. Thepreviously accumulated 1-MeV data are also shownand typical 1- and 28-MeV data for a 10 S2-cm N/Psolar cell are also shown as dashed lines. Thedotted line indicates the Isc value (38 mA) at whichthe critical fluence has been arbitrarily defined.It has been observed that all solar cells are


'7 VQ55

degraded by penetrating radiation at a rate ofabout 6. 5 mA/cm 2 under tungsten illumination foreach decade of fluence added. It is apparent thatthe initially degraded Isc values of the lithium-doped cells decrease at a much smaller rate. Inthis case, the lower rate is probably due to anneal-ing of defects which occurs during the irradiation.After the 600C recovery, the Isc of the lithium-doped cells decline 13 mA per decade of fluence.The important point in Fig. 1 is that, while at1-MeV the lithium cells are slightly superior tothe conventional N/P cell, at 28-MeV the lithiumcell will withstand over ten times the fluencebefore being degraded to the same Isc value of asimilarly irradiated conventional cell.

IV. LITHIUM-DOPED SOLAR CELL REMOVALRATE STUDIES

During the past quarter, extensive capacitancestudies were made on Heliotek and Centralab cellspreviously evaluated under 1-MeV electron irradi-ation. By using techniques previously discussedin reports of this series, the donor concentrationin the N-type base can be determined as a functionof distance into the base. These studies can bemade before irradiation, after irradiation, andafter recovery. In this manner the changes incarrier concentration occurring during irradiationand recovery can be determined. These data canbe of use in the construction of physical models ofthe damage and recovery processes and to provideinformation for the design of improved solar cells.The basic equation used in the analyses is

N SVCd qE

where

Nd = donor concentration

V = voltage

C = capacitance

q = electronic charge

E = dielectric constant

The factor S is related to the exponent of thek = VCn relationship for the cell.

A capacitance study was done on Centralabcells from the CI lC group. These cells werefabricated from Lopex silicon and showed excel-lent initial and recovery characteristics. Thephosphorus concentration of the silicon was ap-proximately 6 X 1013/cm3 . The before-irradiationdata indicate a very much lower donor concentra-tion compared to other cells which have exhibitedgood recovery. Cell CllC-7 and CllC-3 wereirradiated with 3 X 1014 and 3 X 1015 e/cm2 elec-trons, respectively. After recovery from anirradiation of 3 X 10 1 5 e/cm 2 , the donor concen-tration is approaching the phosphorus concentration.This indicates near exhaustion of lithium donors.The removal rates were calculated for these cellsand are plotted in Fig. 2 as a function of barrierwidth. The removal rates are of interest in thiscase because of the two different electron fluencesused. Although there was some difference in lith-ium concentration between the two cells, the dif-ferences in removal can be considered to be largely

due to differences in fluence. One might expectthat the removal rates would be relatively inde-pendent of electron fluence, if the behavior isbased upon discrete solid-state reactions in whichone atomistic species reacts stoichiometricallywith another. The fact that during satisfactoryrecoveries the removal rate has been observed tovary between 1. 5 to 4 times that observed duringirradiation, indicates that no discrete quantity oflithium appears to react with the radiation prod-ucts. The data in Fig. 2 indicate that in C11Ccells, the removal rates during irradiation andrecovery from a 3 X 10 1 4 e/cm2 fluence are sig-nificantly greater than those resulting from a3 X 1015e/cm 2 fluence. Not only are the removalrates lower with the higher fluence, but the ratioof the rates during recovery and irradiation de-crease from 4 to 2 when the fluence is increasedfrom 3 X 1014 to 3 X 1015e/cm2. It is obviousthat the models which we have proposed in the pastmust be modified to account for the widely varyingremoval rates which we have observed in cellsmade from low-oxygen silicon.

Our analysis of lithium distributions also in-cluded two cells made from silicon grown fromquartz crucibles. Our previous studies haveshown that the very small removal rate observedduring irradiation of this type cell can be explainedby the production Si-A centers (oxygen-vacancypairs) with an introduction rate of about 0. 2 cm - 1 .The two cells (CllA-13, CllD-3) had previouslyexhibited excellent recovery when irradiated with3 X 10 1 5 e/cm2 . The result indicated a very lowremoval rate during irradiation of both cells.This is because the concentrations of Si-A centersproduced are barely detectable in these donor con-centrations. Much larger changes in donor con-centration occur during the recovery phase. Tofurther study possible relationships, the removalrate data are replotted in Fig. 3 as a function oflithium concentration present at the particularbarrier width position after irradiation. In addi-tion similar data from a cell (H14-4921) analyzedduring our previous years work are added to Fig. 3.

It can be seen from the data of the three cellsshown in Fig. 3 that the removal rate during recov-ery is directly proportional to the concentration oflithium present at that point in the cell. Since ourprevious work has shown that Si-A centers areintroduced uniformly throughout the active N-typearea of the cell, it must be concluded that theamount of lithium which reacts with the radiationproduced defects is independent of the concentra-tion of Si-A centers. The earlier proposed modelsinvolved the reaction of one or possibly two lith-ium donor atoms with a Si-A center. It is clearthat this model will require considerable modifi-cation. It is possible that precipitation ratherthan ion pairing may be the basis of the recoveryreaction. It has already been proposed that theO-V vacancy pair is the nucleation site for theprecipitation of lithium in silicon and germanium.In such a process the quantity of lithium donorsreacting during recovery would be independent ofthe number of radiation defects, and the ratio oflithium donors reacting during recovery to thenumber of radiation defects could vary considerably.

Using previously described capacitance versusvoltage methods, the donor concentrations at vari-ous distances into the N-type base were determined


before and after irradiation and during the recov-ery process. The results of such a study of cellC13D-1 are shown in Fig. 4. After 720 h the re-covery process, as indicated by the Isc is nearlycomplete and the changes in donor concentrationhave diminished. To allow a more systematicstudy of the lithium donor changes, the data forspecific barrier widths were reduced to lithiumdonor concentration and normalized to that presentimmediately after an irradiation with 3 X 1015e/cm 2 . These data are shown as a function oftime after irradiation, at three different points inthe cell, in Fig. 5.

The mathematical form of the change in lith-ium concentration with time could indicate infor-mation regarding the nature of the recovery pro-cess. The data in Fig. 5 are plotted in semilogform to detect any relationship between the loga-rithum of the lithium concentration and the timeelapsed. Such relationships may indicate first-order chemical kinetics or some aspects of theprecipitation of a second phase. The data in Fig. 5indicate that the lithium concentration decreasesin a nearly straight line form until half of the lith-ium donors have reacted. The rate of reactionappears to slow to zero as the lithium concentra-tion approaches about 25% of its original value.

It was established earlier that an irradiationof 3 X 10 1 5 e/cm2 produced a uniform concentra-tion of defects (probably Si-A centers) of 6 X 10 1 4

cm-3 in a lithium-doped cell made from quartz-crucible silicon. Our work has indicated that suchan irradiation will cause at least 70% of lithiumpresent to react with the defects. Using cellC13D-1 as an example, this means that near thezero bias position of the N-type side of the spacecharge region (2 pLm) about 6 X 101 4 cm- 3 lithiumdonors react during the recovery. This amountsto a one-to-one relation between defects and react-ing lithium donors. At a distance of 3 B±m deeperin the N-type region, about 3 X 10 1 5 cm-3 lithiumdonors react during recovery. At this point, the

concentration of lithium donors reacting duringthe recovery period is five times that of the radia-tion defects detected. It would be highly desirableto extend these measurements to deeper areas ofthe N-type base. After the 3 X 1015e/cm 2 irradia-tion of cell C13D-1, the measured diffusion lengthof the cell was 4.3 pLm. The data in Fig. 4 indi-cate that immediately after the irradiation, thecapacitance-voltage measurement investigates thelithium concentration of the entire active portionof the N-type base. During the recovery phase,the measured diffusion length increased to 17. 5 pm.The distance is currently beyond the limits of thecapacitance technique. If the original concentra-tion of donors found before irradiation is extrap-olated to 17.5 Clm, about 2 X 10 1 6cm- 3 lithiumdonors are present. If 75% of these react duringrecovery phase, 50 lithium donor ion cores reactfor each radiation defect detected. Such behavioris clearly beyond explanation by models previouslyproposed. A model involving the nucleation oflithium precipitation by the radiation defectsappears to be the only one consistent with observedresults.

V. CONCLUSIONS

Several groups of lithium-doped solar cellshave been evaluated under 1-MeV electron irradi-ation. Many of these groups indicated superiorelectrical output after irradiation and recovery ascompared to similarly irradiated N/P solar cells.The superior cells are those with lithium concen-trations of 2 to 5 X 1014 atoms/cm3 at the junction.An irradiation of lithium-doped cells with 28-MeVelectrons indicated a tenfold advantage of lithium-doped cell over N/P cells. Studies of the changesin the lithium concentration during recovery haveshown the amounts of lithium reacting is highlynonlinear in regard to the electron fluence andvaries greatly with distance from the junction.The results indicate that precipitation of lithiumon radiation defects may be the cause of recoveryrather than ion pairing.


Table 1. Float-zone silicon cell recovery characteristics

Cell Diffusion NLi, Average C 1-MeV electron Initial Damaged Recoveredgroup schedule -3 versus fluence, level Is

c , level Isc, level Is c ,( ° C/min/min) cm slope e/cm2 mA mA mA

CllC 325/480/0 1. 1 X 1014 -0. 37 3 X 1014 60.3 34 54

H1A 325/480/0 0 -0.47 3 X 10 1 4 58 33 45

H2A 425/90/120 0.4 X 101 4 -0.32 3 10 1 4 51 31 48

H4A 325/480/0 4.9 X 1014 -0. 35 3 X 1014 47 27 46

H5A1 425/90/120 4.2 X 10 -0.34 3 X 10 52.5 27 50

H5A2 425/90/120 2.8 X 1014 -0.34 3 X 1014 46.5 27 45.5

H5A3 425/90/120 2.3 X 101

4 -0.36 3 X 1014 52.0 27 47

CllC 325/480/0 1.9 x 1014 -0.34 3 X 10 58.5 24.3 39

H1A 325/480/0 0 -0.47 3 X 10 1 5 56 24 25

H2A 425/90/120 2.7 X 1014 -0.30 3 X 1015 56 22 44

H4A 325/480/0 5.6 X 10 1 4 -0.33 3 X 1015 46 19.5 40

H5A1 425/90/120 8.5 X 104 -0.32 3 X 10 1 5 47.5 17.2 39

H5A2 425/90/120 4.4 X 1014 -0.35 3 X 1015 51 19 39

H5A3 425/90/120 3.8 X 10 -0.36 3 X 1015 53 18 33

co

4

L(

0rco

0

w

\o

Table 2. Crucible lithium solar cell recovery characteristics, 60°C recovery

Cellgroup

CllA

C11B

C11D

C13A

C13B

C13C

C13D

C13E

C13F

C13G

C13H

C13I

C13J

Diffus ionschedule,

°C/h

325/8

325/8

375/3

330/4

330/6

340/3

340/7

350/5

350/5

360/3

360/7

370/4

370/6

-3NLi' cm

3.1X

1.5 X

5.1 X

2.5 X

1014

1014

1014

1.6 X 1014

2.3 X 1014

2.1 x 1014

3.1X1014

4.8 X 1014

6. 5 X 1014

3.7 X 1014

5.8 X 10 4

3.0 X 10 4

AverageC

versusV slope

-0. 28

-0. 32

-0. 26

-0. 29

-0. 29

-0. 28

-0. 29

-0.28

-0. 28

-0. 26

-0.31

-0. 29

-0. 31

1 -MeVElectronfluence,e/cm 2

3 X

3X

3 x

3 X

3X

3x

3X

3x

3x

3x

3 X

3 X

3 X

3 X

3 X

3 x

3 X

3X

3X

3x

3X

3 X

15

1015

1015

1015

1014

1015

10 T1i0 1 4

1015

1014

10 1 5

10 1 4

105

15

104

1015

1014

1015

1014

1015

114

3 X 10 1 4

Initiallevel Isc,

mA

64. 2

53.3

61.5

64. 5

64. 0

61.8

61.8

64.0

62.5

63.8

65.0

64.2

59.0

61.5

59.8

57.0

61.5

63.5

61.5

61.0

58.0

60.5

62.8

Damagedlevel Isc,

mA

22.0

24.9

20. 7

22.5

33.8

22.7

35. 7

21.4

30. 0

22. 5

35.8

24. 5

30.1

22.0

29.8

19.5

29. 5

26. 0

30. 0

23.0

30.5

24. 0

37. 8

Recoveredlevel Isc,

mA

>37

>27

>36

39.0

53.0

37.0

51.8

39.4

50.2

37.8

50.5

35.0

47.2

36.0

49.0

36.5

48.0

35.0

49.0

35.0

47.8

34.0

47.0


z

UuU

I-

m0

0 'l 10 ' 10o1 1I

ELECTRON FLUENCE (e/cm 2)

Fig. 1. Short-circuit current versus electron fluence

1016


0.

0.7

0.6

0.5

IE

U nU.

0.2

0.1

BARRIER WIDTH (p)

Fig. 2. Removal rates of cells of CllC group


0.5

0.4

o 0.3

0z

1 0.2

E

20

ut

o

0

co0

z

I

z

z

U0

e-,

20

LITHIUM DONOR CONCENTRATION (1015 ATOM/ cm3 )

Fig. 3. Removal rate versus lithium donor concentration


5 x 1015

4x 10''

24 HRS

o 3x101 5

oS/ 72 HRSz

"

0V 720

0 2x101 5

z

1 x 1015

PHOS. CONC.

0 1 2 3 4 5 6

WB (rm)

Fig. 4. Donor concentration during recovery, C13D-1 cell


3.5 pm

0.2

2 pm

0.0 200 400 600 80CTIME (HRS)

Fig. 5. Change in lithium donor concentration during recovery, C13D-1 cell


O0AS

RADIATION TOLERANCE OF ALUMINUM-DOPED SILICON*

O. L. Curtis, Jr., and J. R. SrourNorthrop Corporate Laboratories

Hawthorne, California

ABSTRACT

Data are presented that indicate that heat-treated Czochralski-grown aluminum-doped siliconthat undergoes an appreciable resistivity increase(5a factor of two) during heating at -450°C is sig-nificantly more resistant to both neutron andgamma irradiation than boron-doped material (oraluminum-doped material that either has not beenheat-treated or does not experience an appreciableresistivity change during such a treatment). Datainterpretation is, however, quite difficult due tothe occurrence of severe trapping effects in heat-treated material. Studies of radiation effects onaluminum-doped silicon solar cells are plannedfor the near future to resolve these uncertaintiesin the bulk material results.

I. INTRODUCTION

The development of methods for increasingthe radiation tolerance of silicon could result insignificant additions to the technology employed infabricating semiconductor devices that are to beused in a radiation environment. Lithium-dopedsilicon, for example, exhibits radiation resistance(Refs. 1, 2) and is currently being studied in detailfor possible solar-cell applications by variousworkers. The quite interesting properties exhib-ited by lithium-doped material lead one to speculate

as to whether there might be other dopants whichwould create desirable radiation-tolerance char-acteristics. (As a recent example, it has beenreported (Ref. 3) that copper-doped N/P siliconsolar cells appear to be more radiation tolerantthan conventional cells.)

During recent years, there have been occa-sional reports by different researchers within thetechnical community that aluminum-doped siliconis more resistant to radiation than is materialdoped with other acceptors. For example,Mandelkorn, et al., (Ref. 4) reported that aluminum-doped silicon solar cells were more radiation-tolerant than boron-doped units (1-MeV electronand 10-MeV proton bombardment). Additionally,we have reported (Ref. 5) that neutron-irradiatedaluminum-doped bulk silicon exhibits a lifetimedamage constantl significantly larger (i. e., moreradiation-resistant) than that for boron- andgallium-doped material. Attempts by otherworkers to reproduce results obtained onaluminum-doped material apparently were not successful.Our previous finding for neutron-irradiated bulkmaterial (Ref. 5) was reproduced (Ref. 6), but itwas not entirely clear at that time what conditionswere necessary in order to prepare radiation-tolerant aluminum-doped silicon. In this paper,we present results of a study of the radiationresistance of aluminum-doped silicon in which it

*Work supported in part by the National Aeronautics and Space Administration through the Jet PropulsionLaboratory, California Institute of Technology, and in part by the Defense Atomic Support Agencythrough Harry Diamond Laboratories.

1 Defined as the amount of radiation required to reduce the lifetime of initially perfect material to 1 [is.That is, damage constant = K = t[(1/T) - (1/To)]-1 , where 4 is the fluence, and To and T are the pre-and post-irradiation lifetimes, respectively.


is shown that heat-treated aluminum-dopedCzochralski-grown material can be prepared thatappears to exhibit significantly improved toleranceto both neutron and gamma irradiation when com-pared to boron-doped silicon.

II. BACKGROUND

A chronological review of the results of ourprevious studies on aluminum-doped silicon ispresented first for the purpose of placing our mostrecent results in the proper perspective. Figure 1shows the results of a systematic study (Ref. 5) ofdamage constants for neutron-irradiated P-typesilicon of various resistivities. The increase indamage constant with increasing resistivity isreadily explained in terms of the Hall-Shockley-Read model, but the particular parameters shownshould not be considered significant. The sampledesignation system used in Fig. 1 is as follows:first letter corresponds to manufacturer (K =Knapic, D = Dow, M = Merck, T = TexasInstruments); second letter corresponds to growthprocess (C = Czochralski, V = vacuum float zone,L = Lopex); third letter(s) corresponds to dopant(gallium, boron, and aluminum). Only onealuminum-doped specimen was examined, but itwas found that the damage constant was signifi-cantly higher than for boron- and gallium-dopedmaterial of the same resistivity. However, sincethe pre-irradiation lifetime for this sample wasvery low, and since a significant amount of trap-ping was observed in the photoconductivity decay,this result was considered tentative at best.

Following our initial finding, a more detailedinvestigation of aluminum-doped material was con-ducted (Ref. 7). Samples prepared from aGzochralski-grown silicon ingot exhibited an ex-tremely short pre-irradiation lifetime, so it wasdecided to anneal the specimens in an attempt toincrease the lifetime. Anneals at 350 and 400°Capparently increased the lifetime but also changedthe resistivity. The samples were neutron-irradiated along with boron-doped control samples,and damage constant results of this earlier studyare shown in Fig. 2 (circles only). It was foundthat the aluminum-doped material was consider-ably more radiation-resistant that the boron-dopedspecimens. However, trapping was still presentto such an extent that the data was not consideredto be highly accurate.

In an attempt to remedy the trapping problem,which was presumably associated with the presenceof a large oxygen concentration, we repeated theabove experiment with aluminum-doped float-zonematerial. These samples had long lifetimes andtrapping effects were small. However, the floatzone specimens were found to be just as radiation-sensitive as boron-doped silicon. Subsequentexperiments with better Czochralski-grown mate-rial were likewise unsuccessful. This led us topostulate that the radiation resistance observedearlier was associated with the heat-treatmentgiven to the specimens. Additional work indicatedthat heat-treated samples were not radiation resis-tant unless the annealing resulted in appreciablecarrier concentration changes, which is the topicof the present paper. Before presenting out mostrecent findings, a brief review of pertinent litera-ture on the effects of heat-treating on silicon ispresented. ,

Resistivity changes in undoped heat-treatedpulled-crystal silicon were first reported 16 yearsago (Ref. 7). Such changes were explained quali-tatively on the basis of the formation of silicon-oxygen compounds, one of which (SiO4 ) acts as adonor below -500°C (Ref. 8). A detailed study ofthe effects of acceptors and oxygen on heat-treatedsilicon was performed by Fuller, Doleiden, andWolfstirn (Ref. 9). In this study, results forgallium-doped material were found to be quitesimilar to those obtained for boron doping, butaluminum-doped material exhibited a somewhatdifferent behavior. Before discussing the dif-ferences, some general features of the results ofFuller, Doleiden, and Wolfstirn are summarized.Heating oxygen-containing acceptor-doped siliconfor long periods at the proper temperature (typi-cally in the range of 400 to 500°C) results in theformation of donor sites with accompanying car-rier concentration changes. The amount of changedepends primarily on four parameters: (1) ac-ceptor concentration; (2) oxygen concentration;(3) heat-treatment temperature; (4) amount oftime the specimen is heat-treated at a particulartemperature. The reactions that produce theobserved carrier concentration changes were notdefinitely determined by Fuller, Doleiden, andWolfstirn, but the data supported the postulationof the formation of primarily three distinct com-pounds: (1) SiO4 donor sites; (2) neutral oxygen-acceptor sites; (3) acceptor-oxygen sites that actas donors. Considerably larger carrier concen-tration changes could be obtained in aluminum-doped material than in comparable boron-dopedspecimens. In fact, particular aluminum-dopedsamples were actually converted to N-type; thiswas not possible in the boron case. It is postu-lated that the reason for this difference may bethat acceptor-oxygen donor sites are formed morereadily in the aluminum case than for borondoping.

III. EXPERIMENTAL RESULTS

A considerable amount of research onaluminum-doped silicon was performed, theprimary conclusion of which was the following:heat-treated Czochralski-grown aluminum-dopedmaterial that undergoes an appreciable resistivityincrease (5a factor of two) is apparently muchmore radiation tolerant than boron-doped materialor aluminum-doped material that either has notbeen heat-treated or does not experience an ap-preciable resistivity change during such a treat-ment. It should be noted that float-zone aluminum-doped silicon is excluded as a radiation-tolerantmaterial by this conclusion because the resistivitydoes not change appreciably during heat-treatment.

Much of the work leading up to the conclusionstated above involved studying the variation ofresistivity during heat-treatments of -450°C,employing the results of Fuller, Doleiden, andWolfstirn as a guide. Resistivity was monitoredat various times during anneals using four-pointprobe techniques. Figure 3 shows the resistivityvariation with time at 450°C for a low-resistivityspecimen. After 315 h, the resistivity had in-creased by a factor of -1.7. In a number of-1 Q-cm specimens examined, the resistivitytypically doubled after 24 h at 450°C.


Following heat treatment, pre-irradiationlifetime measurements were performed and thenspecimens were gamma-irradiated at a Co 6 0

source (dose rate -1.2 X 105 R/h). Other speci-mens were subjected to neutron irradiation in aTRIGA reactor. Following irradiation, damageconstants were determined. Figure 4 shows typi-cal results for gamma-irradiated specimens. Itis seen that aluminum-doped material that ex-perienced a 50 to 100% resistivity change uponheat treatment was significantly more radiationtolerant than both the boron-doped control samplesand the aluminum-doped samples which did notundergo an appreciable resistivity change. Fig-ure 2 shows results for neutron-irradiatedaluminum- and boron-doped specimens (trianglesonly). It is seen that the earlier results (circles)were qualitatively reproduced, once again indicat-ing tolerance to neutron irradiation.

A carrier removal experiment was performedon neutron-irradiated heat-treated aluminum-doped silicon. The study involved examination ofboth pre- and post-irradiation resistivity profilesfor both boron- and aluminum-doped samples. Theneutron dose employed was 1.65 X 1014 nvt(>10 keV). No significant differences between thetwo types of samples were observed; i. e. , the per-cent resistivity increase in aluminum-doped mater-ial due to carrier removal was similar to thatobserved in boron-doped material. This resultsuggests that heat-treated aluminum-doped siliconmay be advantageous from the standpoint of life-time degradation but may offer no improvement ofthe carrier removal problem.

IV. DISCUSSION

The interesting question is raised as towhether heat-treated boron-doped material, whichexperiences an appreciable resistivity change,would also be more radiation tolerant. Fuller,et al., (Ref. 9) indicate that significant resistivitychanges are possible for oxygen-rich boron-dopedspecimens. However, they also found that donorsproduced in heat-treated boron-doped materialdisappear above -800°C, as compared to -1100°Cin aluminum-doped material. Hence, heat-treatedboron-doped material would seem to be less suit-able for applications requiring high-temperaturediffusions (>800°C).

It is difficult to specify at present what mech-anism is responsible for the observed decrease inradiation sensitivity for aluminum-doped material.The effect of the radiation-induced defects on car-rier lifetime is presumably diminished by an inter-action between these defects and one or more ofthe three types of compounds (mentioned above)thought to be formed during heat treatment. TheSiO4 donors can most likely be ruled out becausethey are presumably present in considerablequantity in heat-treated material which has notexperienced a significant resistivity change. Be-cause of the indicated correlation between resistiv-ity change and decreased radiation sensitivity, it'is tempting to speculate that the stable aluminum-oxygen donor sites are involved in reducing theeffectiveness of the radiation-induced defects.

There is a problem regarding the interpre-tation of our data which should be emphasized.Damage constant measurements for bulk materialhave all been based on minority carrier lifetimesdetermined using photoconductivity decay tech-niques (Ref. 10). Because of severe trappingthat occurs in the heat-treated aluminum-dopedmaterial, analysis of transient decays has beenquite difficult at best. Our feeling is that theradiation-tolerance results are still qualitativelyaccurate. However, final proof will not comeuntil diffusion length measurements are made onheat-treated aluminum-doped material beforeand after irradiation. We are currently workingtoward this goal. Solar cells are presently beingfabricated from aluminum-doped material, andradiation testing will follow. If the cells areradiation hard, as our bulk studies have predicted,a considerable effort would undoubtedly have tobe expended to optimize the fabrication techniquein terms of maximizing both the pre-irradiationconversion efficiency and the radiation toleranceof aluminum-doped solar cells.

REFERENCES

1. Vavilov, V. S., in Radiation Damage inSemiconductors, p. 45. Academic Press,New York, 1964.

2. Wysocki, J. J., IEEE Trans. Nucl. Sci.Vol. 13, p. 168, December 1966.

3. Usami, A. , Solid-State Elect., Vol. 13,p. 1202, 1970.

4. Mandelkorn, J., et al., J. Appl. Phys.,Vol. 35, p. 2258, 1964.

5. Curtis, O. L. Jr., IEEE Trans. Nucl. Sci.,Vol. 13, p. 33, December 1966.

6. Curtis, O. L., Jr., Paper 7G. 1 in Recordof the 1969 IEEE International Convention,New York, March 24-27, 1969.

7. Fuller, C. S., et al., Acta Met. Vol. 3, 97,1955.

8. Kaiser, W., Frisch, H. L., and Reiss, H.,Phys. Rev. Vol. 112, p. 1546, 1958.

9. Fuller, C. S., Doleiden, F. H., andWolfstirn, K., J. Phys. Chem. Solids,Vol.13, p. 187, 1960.

10. Curtis, O. L., Jr., and Wickenhiser, R. C.,Proc. IEEE,Vol. 53, p. 1224, 1965.


oKC Ga

* KCBo KVB* DVBo MVB* TCAIa TCGa

- TLB

* TCB

0 /

I. Ir III II .I- 45

n

I I1 I I I I I I - I IT I I I I

a I I I 0 ~ ·I IIo~~~~~~~

_ f _ I _ 11 __ __ __

I_1 zel. -I. .. I .I. .. . 4 I ~~~~~0, a . n.. a

0.1 0.2 .4 0.6 0. 1 2 4 6RESISTIVITY ({-cm)

I 10

/

40 60 81

Fig, 1. Damage constant versus resistivity for neutron-irradiated P-type silicon. Damage constantis expressed in nvt (>10 keV) required to reduce the lifetime of an initially perfect sample to1 Fs. The theoretical curve corresponds to energy level parameters determined fromtemperature dependence measurements.


X10o

i7

6

5

/Io

u1

r-

c

0-

I-C)z0

wo=E

1r

I.

A

68

10'1

8

6

4

2

I-

8z

6w

4

2

io10L llQ50.6 0.8 2 4 6 8 10 20

RESISTIVITY (2-cm)

Fig. 2. Comparison of damage constants for neutron-irradiated Czochralski-grown aluminum- andboron-doped silicon. The aluminum-dopedspecimens were heat-treated.

).-I.-

I--

(1

Ir

0 40 80 120 160 200 240 280TIME (h)

320

Fig. 3. Variation of resistivity with time at -450 °Cfor a low-resistivity aluminum-dopedsilicon sample


Present- I Ref. 6 Data

0 Al-doped -

- 0 B-doped A - .

* o

- - - - I

is

I I I 0---

I _ _ _ _ 1 1 1 >I_

UD I I rI I - 7SAMPLE TCAI 0.252

T n450 °Co

o/oo~0

I I

0. , I0

2I I I I I II

69

2 U /l-gupa " 'u I2v ,

o 2 n very slight

4 4Fig. 4. Com

LUz 2O

107 - - -

101 2 4 6 8 100 2 4 6 8 101 2 4 6 102RESISTIVITY (_Q-cm)

Fig. 4. Comparison of damage constants for gamma-irradiatedCzochralski-grown aluminum- and boron-dopedsilicon. The aluminum-doped specimens wereheat-treated.


(A

EXPERIMENTAL AND COMPUTER STUDIES OF THE RADIATIONEFFECTS IN SILICON SOLAR CELLS*

R. E. Leadon, J. A. Naber, and B. C. PassenheimGulf Radiation Technology

San Diego, Calif.

I. INTRODUCTION

This paper is presented in two parts. Thefirst section will be a summary of selected experi-mental results obtained on lithium-diffused bulksilicon. Particular emphasis was placed on theradiation-induced degradation and thermal annealof minority-carrier in bulk silicon because solarcell output is related to the minority-carrier life-time. The temperature dependence of theminority-carrier lifetime indicates the density andenergy levels of the recombination centers, pro-viding clues to their identity. Electron spin res-onance and infrared absorption techniques wereused to investigate the introduction and anneal ofthree specific radiation-induced defects, whichare thought to contribute to the recombinationprocess.

The results reported herein are a continuationand summary of work previously reported (Refs.1-8).

The second part of the paper will discuss themerits of a computer code developed by GulfRadiation Technology which calculates the current-voltage (I-V) output of a computer simulated P-Njunction. With this code, the user stipulates theinitial properties and environment of a cell. Thecell's properties and environment may then bevaried in a manner which might reflect the antic-

*Work performed for the Jet Propulsion Laboratory,Administration under Contract NAS 7- 100.

ipated radiation exposure and mission environ-ment, and the current-voltage output is recal-culated. Thus, a variety of cell types and experi-mental conditions can be rapidly evaluated. Thiscode can also be utilized by cell manufacturers toevaluate the merit of different cell types, dopingprofiles, and cell parameters. It can be utilizedby those engaged in solar cell testing programs toevaluate a greater variety of cell types, radiationhistories, and ambient conditions than is feasibleby experimental techniques alone. Systems engi-neers can use the code to predict the radiationvulnerability of solar cells in varied environments.

II. EXPERIMENTAL PROGRAM

A. Samples and Radiation

The experimental program has been devotedto an investigation of bulk lithium-diffused silicon.Electron spin resonance and infrared absorptionwere used to study certain specific defects, andminority-carrier lifetime measurements wereused to study radiation-induced recombinationcenters. Samples were lithium-diffused by eithera lithium-oil paint-on technique or by a lithium-tinbath technique. Samples with lithium donor den-sities from as low as 2 X 1020 Li/m3 (2 X 1014Li/cm 3 ) to as high as 4 X 1023 Li/m3 (4 X 1017Li/cm3 ) were investigated. Resistivity profileswere used to monitor samples for diffusion uni-formity. Donor density was occasionally verified

sponsored by the National Aeronautics and Space


by Hall effect measurements. Samples were dif-fused from vacuum float-zone (FZ) and quartz-crucible (QC) phosphorus-doped silicon, rangingin initial room-temperature resistivity from 102Q-m (104 Q-cm) to 0. 005 Q-m (0. 5 2-cm). WhileQC silicon has approximately a hundred times theoxygen concentration of FZ silicon, its dislocationdensity is much less.

Samples were damaged with 30-MeV electronsat temperatures ranging from 115 to 300 K or byfission neutrons (E > 10 keV) at 273 to 300 K. Thedamage was thermally annealed at temperaturesup to 673 K.

B. Study of Three Specific Defects

The introduction of three specific defects by30-MeV electron irradiation and their subsequentthermal anneal were investigated using electronspin resonance or infrared absorption techniques.Specifically, the oxygen-vacancy (Si-B1), diva-cancy (Si-G7), and phosphorus-vacancy (Si-G8)centers have been studied. The results of thesestudies have been previously reported elsewhere(Refs. 7, 8) but may be summarized as follows.

The oxygen-vacancy and phosphorus-vacancycenters were studied by ESR techniques. In QCsilicon, the Si-Bl center introduction rate is thesame as in nondiffused silicon (-15 m- 1 = 0. 15cm- 1) as long as the lithium density is much lessthan the oxygen density. When the lithium and oxy-gen densities are comparable, and B1 center in-troduction rate is significantly reduced (-2. 5 m- 1 0. 025 cm- 1). This can be attributed to the lithium-oxygen pairing, reducing the number of oxygenatoms available to form oxygen-vacancy (Si-Bl)centers, or to competition between oxygen and thepositively charged lithium or lithium oxide donorsfor the vacancies. The B1 center in lithium-diffused silicon is found to anneal below 400K in-stead of near 600K, as in nondiffused silicon.The LiO+ density and conductivity decreased asthe Si-Bl center annealed.

Electron spin resonance was also used to studythe phosphorus-vacancy (Si-G8) center. Inphosphorus-doped lithium-diffused silicon(1022 P/m 3 = 1016 P/cm 3 , -1022 Li/m3 - 1016Li/cm3 ), 30-MeV electron irradiations of 1021e/m2 (1017 e/cm 2 ) below 150K are found to pro-duce Si-G8 (phosphorus-vacancy) defects at a ratecomparable to nondiffused silicon. Production andannealing studies on the Si-G8 center at these den-sities and fluence levels indicate that the presenceof lithium had little or no effect on the creation orannealing of that center. It is possible that atthese high fluences, relative to the initial lithiumdensity, the active lithium has been depleted.

The divacancy (Si-G7) was observed by moni-toring the introduction and anneal of the 1. 8-ylmoptical absorption band. In this investigation, itwas found that for 30-MeV electron irradiationsat less than 150K to fluences of 102 1 e/m 2 (1017e/cm 2 ), the introduction rate of the Si-G7 center(divacancy) is comparable in electron-irradiatedlithium-diffused (5 X 1022 Li/m 3 = 5 X 1016 Li/cm 3 ) and nondiffused silicon. The divacancy an-neals at or below 300 K in diffused silicon com-pared with -325 to 575 K in nondiffused silicon.As the 1. 8-Bim divacancy band disappears, new

bands near 1. 4 and 1. 65 uLm appear, and theseanneal near 600 K. These bands have been ob-served by other investigators (Ref. 9) and attri-buted to the divacancy plus one or two lithiumatoms.

C. Minority-Carrier Lifetime Measurements

The introduction and anneal of recombinationcenters were determined from minority-carrierlifetime measurements. Minority-carrier lifetimewas determined from measurements of the decayof photoconductivity or by steady-state techniques.These measurements are described in detail else-where (Refs. 3, 8). The present results were atinjection levels (An/no) of less than 1%. Inter-pretation of the results is based on Shockley-Readrecombination theory (Ref. 10). Capture cross-section temperature dependence as estimated byLax (Ref. 11) or Leadon (Ref. 12) was assumed.The temperature dependence of minority-carrierlifetime was measured before and after irradiationand again after thermal anneal. Figure 1 showsminority-carrier lifetime versus inverse tem-perature for a lithium-diffused quartz-crucibleand a lithium-diffused float-zone sample. Theminority-carrier lifetime temperature dependenceof the QC silicon can be attributed to recombina-tion centers further than about 0. 3 eV from eitherband edge, but the lifetime of FZ silicon has amore severe temperature dependence and a dis-continuity near 1000/T = 4. 5 K- 1. This can beattributed to recombination centers about 0. 17 eVbelow the conduction band, which were not ob-served in QC silicon, in addition to centers fur-ther than 0. 35 eV from either band edge. Thelines drawn through the data represent the lowinjection-level lifetimes calculated by Shockley-Read theory, assuming the temperature-dependentcapture cross sections of Lax. The presence ofa recombination center near Ec - 0. 17 eV in FZsilicon and its absence in QC silicon suggests thatthis center is not the oxygen-vacancy (Si-B1) cen-ter, which is known to have an energy level in thisrange.

The minority-carrier lifetime degradationrate is defined in terms of lifetime T and electronfluence 4 to be

K = A(1)

compared with phosphorus-doped silicon, and wasfound to increase with increasing lithium donordensity. The degradation constant was found to beessentially independent of oxygen content. Thisshows that the 30-MeV electron-induced defectseither contain lithium or are affected in their pro-duction by lithium. The minority-carrier lifetimedegradation constant was found to vary with tem-perature in a way which could be attributed to thetemperature dependence of the capture crosssection.

Both isothermal and isochronal annealing ex-periments have been performed on a variety oflithium-diffused samples. The unannealed fractionof annealable defects for isothermal annealing isgiven by


1/T - 1/Tf

1/T 0 - l1/Tf

where TO is the lifetime after irradiation and be-fore anneal, Tf is the fully annealed lifetime, andT

tis the lifetime at the annealing temperature after

time t. Linearity of the isothermal annealing dataimplies first-order annealing kinetics (Ref. 13)in which the number of recombination centers isgiven by

N = NO

exp (-Rt)

where N o is the initial number of annealable defectsand R is the rate constant, which is given by

R = v exp (-E/kT)

where E is the activation energy and v is the effec-tive frequency factor. In every case, our iso-thermal anneals indicated first-order annealingkinetics were appropriate.

For isochronal anneals, the unannealed frac-tion of annealable defects is given by

1/TT - 1/Tf

/TO

- 1/Tf

where rT is the lifetime observed at 300K afteranneal at temperature T, and T

Oand Tf are the

pre- and post-anneal lifetimes.

Analysis of isochronal and isothermal anneals,using first-order kinetics, yields the activationenergy and frequency factor of the rate constant.Figure 2 shows the results of five separate iso-chronal annealing experiments on lithium-diffusedQC silicon. A single first-order annealing stageat 380 ±10K is readily apparent. Similar resultswere obtained on fission-neutron-irradiatedsilicon.

Table 1 summarizes our experimental findingsfor the annealing of radiation-induced recombina-tion centers in lithium-diffused silicon exposed to30-MeV electron and fission neutrons (E > 10 keV).The activation energies compare with the 0. 65+ 0. 05 eV energy for lithium diffusion in oxygen-lean FZ silicon, and with 1. 07 ± 0. 05 eV energy oflithium-oxygen dissociation and diffusion in oxygen-rich QC silicon (Refs. 14-16). This data supportsthe supposition that radiation-induced damage, in-cluding the clustered damage attributed to fissionneutrons, is annealed by the diffusion of lithium tothe defect.

In addition, the effective frequency factors aremuch less than the atomic frequency factor of about1013 s- 1 and scale with the lithium donor density.This is consistent with a process involving thelong-range migration of one annealing species toanother (Ref. 13), which we take to be the migra-tion of lithium to the damage site.

A more detailed discussion of our experimen-tal findings and theoretical interpretations is pre-sented elsewhere (Ref. 8), and a considerable bodyof additional data, provided by us and other inves-tigators using other techniques, is available onbulk silicon. The following section of this paperdescribes one means of applying this body of datato the problem of predicting the response of asolar cell to various radiation and environmentalconditions.

III. CODE FOR PREDICTING PERFORMANCEOF SOLAR CELLS

The purpose of this presentation is to de-scribe a computer program that can be used forpredicting the output performance of solar cells.This code was developed under Defense AtomicSupport Agency contract and applied to solar cellsunder the JPL contract. Briefly, this code, whichis fully operational at the present time, solves thecomplete equations for a one-dimensional P-Njunction. The code starts with the current andcontinuity equations for electrons and holes,Poisson's equation for the electric field inside thecell, and appropriate equations for the generationand recombination of excess carriers. Arbitraryone-dimensional space profiles of doping density,mobilities, recombination lifetimes, and ioniza-tion densities are inputs to the code. The partialdifferential equations are converted to finite dif-ference approximations, which are then solved byiteration. In principle, there are no adjustableparameters in the code. If one had accurate valuesfor all of the material properties, the code wouldpredict the absolute magnitudes of the current andvoltage for selected values of load resistance.Thus, the accuracy of the results is primarilydetermined by how well one can measure or esti-mate the important physical parameters of thecell.

This code could be useful both in the design ofnew devices and in the analysis of radiation effectsto solar cells. For example, if a cell manufac-turer is designing a new type of cell, perhaps withan unusual doping profile, he could estimate theperformance of the new device relative to a con-ventional cell and optimize its parameters beforehe goes through the expense of constructing it.Similarly, the radiation-effects physicist coulduse such a code to help him analyze his solar celldata in terms of kinds and densities of defectscreated by the radiation and to extrapolate the re-sults to other radiation environments.

The three basic equations of the code are theone-dimensional continuity equations for holes (p)in the valence band and electrons (n) in the con-duction band, and Poisson's equation for the elec-tric field (E).

ap(x, t)=at = g(x,t)- R(x,t)- ax

an (x,t)an(x,t) t) - R(x,t)- x

aE(x, t) = [p(x,t) - n(x, t) + AN(x)]Ox

(1)

(2)

(3)


In the above equations, the dimension x is normalto the face of the cell and g(x, t) is the time- andspace-dependent ionization rate of electron-holepairs by the external radiation. This quantity iscalculated ahead of time, using the spectral dis-tribution of the incident light and the attenuationcoefficients for the particular material, and isthen an input into the code. The quantity q is themagnitude of the electronic charge, K is the di-electric constant in MKS units, AN(x) is the netdoping of the semiconductor (positive for n-type,negative for p-type), and Jp and Jn are the particlecurrent densities given by

Jp = pE- D a2 (4)p = np pax

J = -n)iE - D ax (5)

In Eqs. (4) and (5), the first terms are the drifcomponents of the currents with mobilities ip a}in, while the second terms give the diffusion clrents with diffusion coefficients Dp and D n .

The term R(x, t) is the rate of recombinatiof excess carriers via recombination centers.There are several different options available touser, depending on the sophistication he wishesuse in describing the recombination process.many situations, it is adequate to use a ShockleRead-type recombination rate for a single recobination level:

2np- n.

1

T (p + PF ) + rO (n + nF)n P

where ni

is the intrinsic carrier density, PF ann F are the values of p and n when the Fermi le-coincides with the level of the recombination ceter that is being considered, and TOn and r O athe recombination lifetimes of electrons in hpa'p-type material and of holes in heavily n-typematerial, respectively.

This recombination equation does have thelimitation that it only simulates the simultaneouannihilation of a free electron and a free hole.does not consider the trapping of carriers on defect centers and, therefore, carrier removal.Since the latter phenomena can be important whthe details of the trapping kinetics affect the carier lifetimes or removal rates, the code alsothe capability of simulating both single-level andouble-level traps. This could be important, fexample, if the radiation produces a high-densiregion of defect centers which significantly de-pletes the majority carriers and creates a p-njunction effect.

For a single-level trap with density N1

andtwo possible charge states, negative (N 1 ) andneutral (N?), the (-R) term in Eq. (1) would bereplaced by

-Rp = cp(pN - N4PF)

74

tandur-

on

o a

and the (-R) term in Eq. (2) would be replaced by

-R° = n(nNOl N nF)n n I I (8)

The quantity ap is the product of the thermal ve-locity of a hole in the valance band and the crosssection a- for capture of a hole by a negativelycharged Sefect, and ao is a similar term for cap-ture of an electron by a neutral defect. In additionto modifying Eqs. (1) and (2), the density of thenegatively charged center (Ni) must be includedin the summation of charges in Eq. (3). Further-more, additional equations are required for therate of change of N1 and N?. For constant totalN 1 ,

dN_ dN 0

1 1 oR-dt dt n p

Similar equations are available in the code for adouble-level defect with three charge states, N2,No, and NJ, but, for brevity, they are not pre-sented here.

s to To obtain a solution to the above equations,In they are converted to finite difference form forey- some prescribed mesh distribution. The distri-)m- bution of the meshes can be chosen arbitrarily by

the user to most judiciously utilize the permissiblenumber of meshes. On the Univac 1108, corestorage limits the number of meshes to about 230,which is an adequate number to simulate a three-

(6) layer planar transistor, and is certainly enoughfor a two-region solar cell.

Since this code was originally developed tostudy time-dependent processes in semiconductor

nd devices, the finite difference equations are solvedvel by iteration for finite time steps. Under condi-en- tions of steady illumination and circuit load, theare code quickly goes to the steady-state solution.vily The resulting steady voltage and current form one

point on the I-V curve. Additional points are ob-tained by changing the external load resistanceand again allowing the problem to run to equilib-rium.

Is

It Since Eqs. (1), (2), and (3) are time-e- dependent partial differential equations, one must

specify boundary conditions at the two ends of theLen device and self-consistent initial distributions on

Lr- n, p, and E. The self-consistent initial distri-has butions could have been specified in any number of-d ways. To simplify the task of the user, a newor problem is started from a condition of zero elec-ity tric field and charge neutrality everywhere. This

is usually a physically unrealistic, but mathe-matically correct, starting point. The code thendoes the work of proceeding to the correct solu-tion. For a subsequent run, with the same device,

i it is usually possible and economical to restartfrom any previous cycle.

Two types of boundary conditions are avail-able. The first, called "bulk, " forces the car-rier densities at the two ends of the device to have

(7) zero slopes. This is the condition that wouldapply if the p-n junction were far from the contacts


(9)

and the densities have a chance to approach theirbulk values far from the junction (hence the name"bulk"). This boundary condition is suitable whenthe details of the contact are not important in theproblem.

On the other hand, when the details of thecontact between the semiconductor and the metalare important, the contact potential between thetwo materials can be simulated in the second typeof boundary condition. A large surface recom-bination velocity can be simulated by defining anarrow region at the surface with a very short re-combination lifetime. Of course, an additionalrequirement is that the current through the ex-terior circuit must equal the current through thedevice, including the displacement current in time-dependent problems. Detailed descriptions of thestarting and operating procedures for the code canbe found in Refs. 17 and 18.

Since this code solves the complete equationsfor the solar cell, the answers that it providesshould be as good, within the accuracy of the finitedifference method of solution, as the input datasupplied to the code. For comparative designstudies, the theoretical device characteristics canbe specified exactly and the effect of varying oneor more parameters individually can be evaluated.For the analysis of radiation effects in an actualdevice, the problem is somewhat different becausethe parameters for the undamaged test device,such as doping profile and lifetimes, are usuallynot accurately known. For the best results, it isalways desirable to avail oneself of all possiblenondestructive tests to determine the deviceparameters, such as capacitance-voltage mea-surements and resistivity for the doping densityprofile and diffusion lengths for the lifetimes.

However, in the final analysis, it may benecessary to adjust these parameters somewhatby trial and error to obtain acceptable agreementwith the preirradiated test data. Once these char-acteristics are determined for the undamaged cell,the change in the bulk parameters due to radiationdamage can then be estimated and included in thecode. If the test data on radiation effects in thebulk material are sufficiently good so that themodel for the radiation damage and the change perunit dose can be correctly determined, the pre-dicted results for the irradiated cell should be inacceptable agreement with the post- irradiation testdata. If there is significant disagreement, it mustmean that the model for the radiation damage orthe magnitude of the radiation effects is differentfrom that estimated. In this way, the code can beused as a tool for analyzing and interpreting radia-tion effects test data. After agreement has beenobtained between the experimental and calculatedresults for a number of test conditions, the codecan then be used to extend the test results to othertest or operating conditions.

As an example of the results from this code,Fig. 3 shows predicted I-V curves for a particularsolar cell for several fluences of 300-keV protonsimpinging on the illuminated surface. The param-eters for the undamaged cell were adjusted to giveagreement with the performance curve supplied bythe manufacturer. The proton damage was simu-lated by a distribution of recombination centers,concentrated mainly in the region -of the maximum

range of the protons. If adequate test data for theproton fluences were available, the validity of themodel for the proton damage could be tested bycomparing experiment to prediction.

REFERENCES

1. Naber, J. A., Horiye, H., and van Lint,V. A. J., "Radiation Effects on Silicon,"Final Report GA-8668, Gulf General Atomic,Inc., Aug. 20, 1968.

2. Naber, J. A., Horiye, H., and Berger,R. A. , "Production and Annealing of Defectsin Lithium- Diffused Silicon After 30-MeVElectron Irradiation at 300°K," Proceedingsof the Conference of Effects of Lithium Dopingon Silicon Solar Cells, held at the Jet Pro-pulsion Laboratory, May 9, 1969, TechnicalMemorandum 33-435. Jet Propulsion Lab-oratory, Pasadena, Calif., Aug. 15, 1969.

3. Naber, J. A., Passenheim, B. C., andBerger, R. A. , "Study of Radiation Effectsin Silicon Solar Cells," Report No. GA-9909,Annual Report for Contract No. NAS 7- 100,Gulf General Atomic, Inc., Jan. 23, 1970.

4. Naber, J. A., Passenheim, B. C., andHoriye, H. , "Introduction and Annealing ofDamage in Lithium-Diffused Silicon," Pro-ceedings of the Third Annual Conference onEffects of Lithium Doping on Silicon SolarCells, held at the Jet Propulsion LaboratoryApril 27 and 28, 1970, Technical Memorandum33-467. Jet Propulsion Laboratory, Pasadena,Calif. , April 1, 1971.

5. Passenheim, B. C., and Naber, J. A.,Radiation Effects, Vol. 2, p. 229, 1970.

6. Passenheim, B. C., Naber, J. A., andBerger, R. A. , "Production and Annealingof Defects in Lithium-Diffused Silicon AfterIrradiation With 30-MeV Electrons and FissionNeutrons at 300'K," Conference Record of the8th IEEE Photovoltaic Specialists Conference,Seattle, Wash., Aug. 4-6, 1970, IEEE CatalogNo. 70C-32 ED, pp. 260-266.

7. Naber, J. A., Horiye, H., and Passenheim,B. C., "Lithium-an Impurity of Interest inRadiation Effects in Silicon," to be publishedin Radiation Effects (1971).

8. Passenheim, B. C., et al., "Study of Radia-tion Effects in Silicon Solar Cells;" ReportNo. Gulf-RT- 10482, Annual Report for Con-tract No. NAS 7- 100, Gulf Radiation Tech-nology, Jan. 20, 1971.

9. Young, R. C., et al., J. Appl. Phys., Vol.40, p. 271, 1968.

10. Shockley, W., and Read, W. T., Jr., Phys.Rev., Vol. 87, p. 835, 1952.

11. Lax, M., Phys. Rev., Vol. 119, p. 1502,1960.

12. Leadon, R., and Naber, J. A., J. Appl.Phys., Vol. 40, p. 2633, 1969.


13. Damask, A. C., and Dienes, G. J., PointDefects in Metals, Gordon and Breach,Science Publishers, New York, 1963.

14. Pell, E. M., J. Appl. Phys., Vol. 31,p. 291, 1960.

15. Fuller, C. S., and Severiens, J. C., Phys.Rev., Vol. 96, p. 21, 1960.

16. Maita, J. P., J. Phys. Chem. Solids,Vol. 4, p. 68, 1958.

17. Leadon, R., and Vaughn, M., "Short-PulsedRadiation Effects on Dynamic ElectronicComponents," Final Report on ContractDASA01-68-C-0123, DASA Report No. 2358Defense Atomic Support Agency; also GulfGeneral Atomic Report GA-9397, June 5,1969.

18. Leadon, R. E., et al., "Radiation Effectson Dynamic Electronic Components," FinalReport on Contract DAOSA01- _69-C-0113,DASA Report No. 2546 Defense AtomicSupport Agency; also Gulf RadiationTechnology Report GA-9775, July 1970.


Table 1. Summary of the results of isochronal and isothermal annealing experiments on electron- andneutron-irradiated lithium-diffused float-zone and quartz-crucible silicon

Sample type Float zone Quartz crucible

Lithium donor 2. 5 X 1022 4. 5 X 10O

2 X 10 2. 5 X 1021 4. 5 X 1020 2. 5 X 10

density, Li/m- 3

Fluence 1017 10 6 to X 1014X .5 X 1016 1016 4 X 1014

(particles/m ) to 8 X 1016

Radiation 30-MeV 30-MeV Fission 30-MeV. 30-MeV Fission

electrons electrons neutrons electrons electrons neutrons

Activation 0. 85 +0. 10 0.60 +0. 10 0.67 +0. 05 1.0 ±0. 2 0.75 m0. 10 1.2 m0. 6

energy, eV

Frequency factor, 10 10 Unknown -107 108 to 1010 107 to 108 Unknown-1

sec


200 6200 i 2 ° After 2,7 x 10 elm at 300'K

FZ SILICON

4.5 x 1020 Li/m 3

100 · Preirradiation

80 -8 \ . After 9 x 1015 e/m3

at 300°K-0 -- Shockley-Read Theory

060

40 O

U_

1 0 20 ·

2 I I I3.0 4.0 5.0 6.0 7.0

(IOOO/T) (K- )

Fig. 1. Inverse temperature dependence of minority-carrier lifetime of lithium-diffused float-zone and quartz-crucible silicon beforeand after irradiation with 30-MeV electronsat room temperature


(D

gC LITHIUMI1.0 0 DENSITY FLUENCE

o 0 (Li/m3 ) (e/m2 )

0 2.5x102 1 7.8x101 6

O Ao 2.5x102 1 2.40x1 1 6DA 21 16

' 0.8 V 0 2.5x102 1 8.4x101

6

c \ ° V I.OxlO 2 2 6 .0x1016

-jz 0.6z

< 0.4L_

= 0.2

0

300 320 340 360 380 400 420 440

ANNEAL TEMPERATURE (°K)

Fig. 2. Unannealed fraction of annealable defects after 5-min isochronal anneals at indicatedtemperatures after 30-MeV irradiation of lithium-diffused N-type QC siliconirradiated at 300 K

30

25 10'~ p/cm '

20

!0 20

I I I I I I a 1, I l I0.1 0.2 0.3 0.4 0.5 o.6

VOLTAGE (volt)

Fig. 3. Current-voltage curves of a graded-junction cell degraded by 300-keV protons tovarious fluencesvarious fluences


HALL EFFECT STUDY OF ELECTRON IRRADIATED Si (Li)*

J. StannardNaval Research Laboratory

Washington, D.C.

I. INTRODUCTION

Measurement of the Hall coefficient as a func-tion of temperature between 15 and 300K allowsthe separate determination of donor and acceptorconcentrations in silicon. In this experiment theseconcentrations were monitored in samples whichwere irradiated at 240K by l-MeV electrons andthen thermally annealed at 300 K. Presumably inboth lithium- and phosphorus-doped silicon onewould expect irradiation to increase the acceptorconcentration and decrease the donor concentra-tion due to the formation of vacancy donor pairs.Annealing should not change either concentrationin phosphorus-doped silicon and should cause bothconcentrations to decrease in silicon doped withmoderate amounts of lithium. In addition to this,in both lithium- and phosphorus-doped silicon theconcentration and energy level of the A center canbe measured at each point in the sample's history.This experiment was limited to float-zoned silicondoped with less than 6 X 1014 donors/cm 3 .

II. THE A CENTER

Before going into the details of this experimentsome evidence will be given concerning the anoma-lous activation energy previously reported for theA center. Measurements made on a variety ofsamples indicated a level at Ec -0.14 eV, insteadof at Ec -19 eV as expected for the A center(Ref. 1). All of these samples were cut from oneof two boules made by Dow Corning. A sample ofSemi-element's pulled silicon was irradiated and

the A center energy measured to be 0. 19 eV.Another sample of 14Q2-cm float-zoned Si(P) madeby Dow Corning was then lightly irradiated andthe production rate and energy of the A centermeasured. Then the same sample was irradiatedmuch more heavily so that enough deep centerswould be created to depopulate the level at 0. 14 eVand indicate the presence of any deeper levels.The results of this experiment were as follows:

Concen- ProductionDose, Energy,e/cm2 tration, eV rate

e/cm Z cm_3 eVcm

2 X 1014 2. 5X 1013 0.14 0.13

1.3X 1015 1.5 X1013 0.19 0.01

This data resolves the problem. In this boulecenters at 0. 14 eV and 0. 19 eV are both beingformed as a result of irradiation. However, thecenter at 0. 14 eV is being formed with a muchhigher production rate than the usual A center at0. 19 eV. When the sample was heavily irradiatedthe Hall effect ceased to count the level at 0. 14 eVand also began to indicate the presence of a levelat Ec -0. 19 eV. The level at Ec -0. 14 eV may berelated to the level at Ec -0. 13 eV seen by Steinin pulled silicon (Ref. 2). If so, the only unusualproperty of this Dow Corning material is that theformation rates of these two levels are reversed

*Work performed for the Jet Propulsion Laboratory (Contract WO-8056) and supported by the NationalAeronautics and Space Administration under Contract NAS7-100.


in their normal order of importance. Since oxygendoes form a number of different complexes in sili-con, it is most likely that the 0. 14 eV center isthe usual A center but with perhaps an additionaloxygen present in a nearby interstitial site. Thepossibility that another impurity such as carbonmay be involved cannot be discounted, however.

III. LITHIUM AND THE A CENTER

The experiment to be described was designedto study the interaction of lithium with the A cen-ter. In order to determine what effects were dueto the presence of lithium it was necessary to usea sample and a control cut from the same bouleand having similar resistivities. The presence ofboth phosphorus and lithium in the sample reducedsomewhat the variety of information obtainable,i.e., it was not possible to demonstrate that theLiV complex is formed during irradiation. Aseparate experiment would be necessary to investi-gate this question.

Two samples were prepared from the sameboule of float-zoned 14 2-cm Dow Corning siliconcontaining 3. 5 X 1014 donors/cm3 . Into onesample an additional 3.4 X 1014 lithium donors/cm 3 were introduced from a lithium in oil suspen-sion using a tack-on and redistribution of 5/60/425°C. Measurements of the Hall effect versustemperature were made; (1) before irradiation,(2) after irradiation at 240 Kby 1.7 X 1014 1 MeVe/cm2 , (3) after 2 h, (4) and after 17 h at 300 K.The electrical contacts degraded due to thermalcycling after 17 h thereby terminating the experi-ment. Irradiation was performed at 240 Kso thatlithium could be considered immobile duringirradiation. As a result the effects of irradiationcould be separated from those due to lithiummobility. Donor and acceptor concentrations wereobtained by least squares analysis of the datausing Fermi statistics for silicon containing twokinds of donors and two kinds of acceptors. Thesecalculations assumed the values of excited statesplitting and ground state degeneracy obtainedfrom optical experiments reported in the litera-ture (Refs. 3-5).

Figure 1 shows the behavior of room tempera-ture carrier concentration in the two samples asa function of annealing time and is the behaviorexpected. There was no measurable annealing inthe phosphorus-doped sample and a decreasingcarrier concentration in the sample containinglithium.

Figure 2 shows the change in donor concen-tration in these samples caused by annealing at300 K. The most apparent difference between thetwo samples for annealing times longer than 2 h isthat the donor concentration decreases in thelithium-doped sample and is constant in thephosphorus-doped sample. This change in thelithium-doped sample was clearly caused by radi-ation damage, since such samples do not show achange in resistivity with time prior to irradia-tion. From the data shown here it is not possibleto show that the LiV complex forms during irradi-ation. From the pre-irradiation and post-irradiation points it is clear that there is moredonor loss during irradiation in the sample con-taining both lithium and phosphorus than in thesample doped only with phosphorus. Unfortunately,several samples doped only with phosphorus

showed a range of values for carrier removalrate. An experiment similar to this one, butusing a sample doped only with lithium, woulddemonstrate the formation of LiV.

Comparison of the data before annealing andafter 2 h of annealing indicates a decrease indonor concentration caused by bringing the sampleup to room temperature from 250 K. This was notexpected in the phosphorus-doped sample andrepresents a short-term thermal annealing in thedirection of increased damage. Other samplescontaining phosphorus were irradiated at 250 Kbut were held at 250 Kfor 2 h before making anymeasurements. This was found to decrease themagnitude of the annealing stage, indicating thatannealing occurs at 250 K as well as 300K. Instudies of unannealed Si(Li), data should be takenafter substantial annealing at 250 K, which willremove this stage without allowing the lithium tobe mobile. This short-term annealing is evenmore striking in Fig. 3, which shows the acceptorconcentration.

The total measured acceptor concentrationand A center concentration are represented by Atand 0. 14, respectively. For the phosphorus-doped sample the value indicated for the E-centerconcentration is the measured decrease in donorconcentration caused by irradiation. Inspectionof this data indicates that this annealing may becaused by thermal release of trapped vacancieswhich then form additional damage complexes.Such vacancies would have to be much moretightly bound than the vacancy-interstitial close-spaced pair, as the observed annealing tempera-tures are very high (Ref. 2). This behavior mayalso be a characteristic peculiar to these boules'of silicon. As was expected, the behavior ofacceptor concentration with continued annealingwas very different in the two samples. Whereasannealing causes acceptor concentration to de-crease in the lithium-doped sample, it has noeffect in the phosphorus-doped sample. It isevident from this figure that A centers interactmuch more strongly with lithium than does the Ecenter. Comparing the changes in total acceptorconcentration with those in A center concentrationit is apparent that almost 80% of those centersthat are neutralized within 17 h were A centers.Since there were probably as many E centers inthe lithium-doped sample as there were in thecontrol sample, it is clear the A centers com-pletely monopolized the neutralization processdespite the fact there were more E centers thanA centers.

By comparing the changes in donor and ac-ceptor concentrations seen in Figs. 2 and 3 anidea canbe obtained as to how many lithium ions arerequired to neutralize the A centers. The con-centration changes between 2 and 17 h are:

ALi = -4.5 X 1013 cm 3

AAt

= -1.3X 1013t

A0.14 = -1.0X 1013

Since the total acceptor concentration (At) in-cludes the A center concentration (0. 14) it isapparent that 80% of the annealed centers wereA centers. Comparing this change to the much

JPL Technical Memorandum 33-4918Z

larger loss in lithium concentration it seems thatas many as four lithium atoms may be requiredto neutralize this A center.

REFERENCES

1. Sonder, E., and Templeton, L., J. Appl.Phys., Vol. 31, p. 1279, 1970.

2. Stein, H., and Vook, F., Phys. Rev., Vol. 163,p. 790, 1967.

3. Aggarwal, R., et al., Phys. Rev., Vol. 138,p. A882, 1965.

4. Long, D., and Myers, J., Phys. Rev.,Vol. 115, p. 1107, 1959.

5. Aggarwal, R., and Ramdas, A., Phys. Rev.,Vol. 137, p. A602, 1965.


I I I I I I [ I

20 30 40 50 60 70

TIME, h

80

Fig. 1. Annealing of room-temperature carrierconcentration, irradiated at 240 K by1.7 X 1014 1-MeV e/cm 2

1

o 0

x

E

Z 0O

U

0U

0

u

uO

.0

.8

.6

6

Io

x

5

0zU

z 40

o

z20

3z 43_

2

Li+P

P _~~~~~~~~~~0

I I I I I I I I

0 4 8 12 16

TIME, h

Fig. 2. Room temperature annealing of donorconcentration in lithium andphosphorus-doped silicon

.4 . 0.14iv~ ~ ~

0ci, ' ,, 1

O0 p1 I I I I I0 1 { 1 1 }

0 4 8 12 16 20 0 4 8 12 16 20

TIME, h

Fig. 3. Room temperature annealing ofacceptor concentration in lithiumand phosphorus-doped silicon


J I I I I I I

Si(Li) 6 n- cm FZ

A I A I I I I I I

n 6.0

8 5.0

' 4.0

x

zO

F 3.0

u 2.5

U 2.

1.5

I I I I I I I

Si(P) 14 n,- cm FZ

.. I ^1 I I I 1

S~~~~~~ _ 0

0 10

I I I I I

Si (P)

At-

q 16E

I I I I I

Si(Li)

At

'U

- __O_

)I

0

I

84

THE OBSERVATION OF STRUCTURAL DEFECTS IN NEUTRON-IRRADIATED LITHIUM-DOPED SILICON SOLAR CELLS

G. A. SargentDepartment of Metallurgical Engineering and Materials Science

University of KentuckyLexington, Kentucky

ABSTRACT

Electron microscopy has been used as a tech-nique to observe the distribution and morphologyof lattice defects introduced into lithium-dopedsilicon solar cells by neutron irradiation. Uponetching the surface of the solar cells after irradi-ation, crater-like defects are observed that arethought to be associated with the space chargeregion around vacancy clusters. The crater defectdensity was found to increase with increasingirradiation dose and increasing lithium content;however, the defect size was found to decreasewith increasing dose and lithium.

Thermal annealing experiments showed thatthe crater defects were stable in the temperaturerange 300 to 1200K in all of the lithium-dopedsamples. Some annealing of the crater defectswas observed to occur in the undoped cells whichwere irradiated at the lowest doses.

I. INTRODUCTION

It is now generally accepted that a localizedcluster of lattice defects may be produced by arecoil from a single collision between an energeticneutron and a lattice ion (Refs. 1-5). The energyof the recoil is dissipated by creating lattice dis-order. Subsequent rapid quenching of the latticeshould freeze a large concentration of latticedefects into the neighborhood.

Until recently, the behavior of irradiatedsemiconductors has been interpreted on the basisof isolated Frenkel defects in terms of the model

of James and Lark-Horovitz (Ref. 6). However,Gossick (Ref. 7) and Crawford and Cleland (Ref. 8)have proposed a model of disordered regions thatis more applicable to neutron-irradiated semi-conductors and predicts the existence of regions ofhighly localized damage. Their model assumesthat a region of lattice disorder is produced uponirradiation which may contain a high concentrationof defects such as vacancies. Surrounding the dis-ordered region is a potential well that arisesbecause the position of the Fermi level relative tothe energy band is different within the disorderedregion compared to that outside. Crawford andCleland have estimated the size of the defect re-gion to be of the order of 15 to 20 nm. For P-typesilicon the dimensions of the space-change regionsurrounding the defect region due to the potentialwell is predicted to be about 200 to 250 nm indiameter.

The measurement of electrical properties byClosser (Ref. 9) and Stein (Refs. 10-12) have sub-sequently provided direct evidence for the exis-tence of damage regions as predicted by the theo-retical models of Gossick and Crawford andCleland.

Until recently little work has been carried outto determine the exact structural nature of thelattice disorder created by irradiation damage.X-ray diffraction (Ref. 13) techniques and directobservation of thin foils by electron microscopy(Refs. 14-16) met with limited success in thisrespect. However, an alternative method forobserving defects in semiconductors using surfacereplication electron microscopy has been


perfected by Bertolotti and his coworkers(Refs. 17-20).

Bertolotti found that upon etching the surfaceof neutron-irradiated silicon samples craters wereproduced, the dimensions of which were found tocompare with the dimensions of the space-chargeregions as predicted by the theory of Gossick(Ref. 7) and Crawford and Cleland (Ref. 8). Withinthe craters a small well-defined region could alsobe observed, the size of which compared well withthe theoretical estimates for the size of the defectclusters.

The main objective of the work to be describedin this present paper was to make use of the tech-nique developed by Bertolotti to determine theeffects of neutron irradiation on the structuralcharacteristics of undoped and lithium-doped sili-con solar cells. The overall objective was toobtain a better understanding of the morphology ofthe defects produced by irradiation damage andtheir interaction with dopants such as lithium.Such knowledge could lead to the development ofsolar cells which are more radiation resistant.

II. EXPERIMENTAL TECHNIQUE

The work was carried out on commercialundoped and lithium-doped silicon solar cells pre-pared by Heliotek. The cells were produced fromfloat-zone melted single crystals of phosphorus-doped N-type silicon. Boron was diffused into theslice to give a junction depth of 0.5 pLm. Lithiumwas diffused in using the paint-on technique toproduce three types of cells: 10 1 5 , 1016 and 10 1 7

lithium atoms/cm3 . In addition, a fourth type ofcell was produced with no lithium.

Samples of the undoped and lithium-dopedcells were irradiated with monoenergetic 14. 7-eVneutrons produced by a Cockroft-Walton generator.The irradiation was carried out at room tempera-ture, and the dose was controlled by varying thedistance of the samples from the target.

After irradiation the surface of the solar cellswas prepared for replication by mechanicallygrinding and polishing and then by etching withCP4A etchant. A carbon replica of the surfacewas obtained by evaporation. The replica wasshadowed with chromium and was observed in anelectron microscope at 75 kV.

To study the recovery of the irradiation dam-age a number of samples were annealed at temper-atures of 293, 593, 700, 900 and 1200K for 10 minin a vacuum furnace (10-6 torr).

III. RESULTS AND DISCUSSION

Figure 1 shows a photograph obtained in theelectron microscope of a surface replica takenfrom the etched surface of an unirradiated samplewhich contained no lithium. This photographshows a finely etched uniform structure withoutany additional significant features. In contrast,Fig. 2 shows a surface replica from a samplewhich contained no lithium but was irradiated with1011 neutrons/cm2 . This is a typical example ofthe appearance of the etched surfaces of thesamples after neutron irradiation. The areashows a finely etched background structure onwhich many crater-like depressions can beobserved.

86

According to the Gossick theory (Ref. 7),- thedimensions of the craters revealed by the abovetechniques actually correspond to the size of thespace-charge region which surrounds a cluster oflattice defects formed during the irradiation.Therefore, the density of the craters should relateto the total defect volume produced at a givenirradiation dose.

The average density and diameter of the cra-ters were measured as a function of dose andlithium content. Values of the average craterdefect density and diameter as functions of irradi-ation dose and lithium content are presented inFigs. 3 and 4, respectively. It can be seen thatthe average defect density increases with increas-ing irradiation dose and increasing lithium content.On the other hand, however, it can be seen fromFig. 4 that the average defect diameter decreaseswith increasing dose and lithium content. It wasfound from these results that, for a given irradi-ation dose, the total defect volume is essentiallyconstant. It appears that the presence of lithiumprovides more nucleation sites for the defectclusters to form, and as a result the averagedefect diameter is smaller. Under comparableirradiation conditions, 14. 7-MeV neutrons at adose of 10 1 2 /cmZ, the present results are in goodagreement with those reported by Bertolotti, et al.,(Ref. 19). Therefore, the present results supportthe theories of Gossick (Ref. 7) and Crawford andCleland (Ref. 8).

From the annealing experiments it was foundthat no change could be detected in either theaverage defect density or the average size overthe temperature range 300 to 1200K, with theexception of those samples which were undopedand irradiated at the two lowest doses, i.e., 1010and 1011 n/cm2 . These results are shown inFigs. 5 and 6 where the average defect densityand size are plotted, respectively, as a functionof annealing temperature. In these samplesannealing was found to be significant at a tempera-ture of about 800 K and was reflected in both anincrease in defect density and a decrease in defectdiameter.

It would appear from these results, therefore,that the defect structure which is formed duringthe irradiation at room temperature of either theundoped or doped material, at the highest doses,represents the most stable defect structure. Thelarge defect clusters which are formed at the lowdoses in the undoped material have a tendency tocollapse upon annealing to form smaller defects ata higher density.

The results of the annealing experimentscarried out in the present work appear to be signif-icantly different from those observed previouslyfrom recovery of electrical properties in neutronirradiated silicon, byStein (Ref. 21) and Passenheimand Naber (Ref. 22). However, it is not possibleat the present time to make direct correlation withthese results as the irradiation conditions andannealing times were significantly different.

It is not apparent from the present resultsexactly what role lithium plays in the nucleationand stabilization of the irradiation-induced defects.At the high temperatures used in the annealingexperiments it is unlikely that lithium would re-main in the bulk lattice. There is some evidence(Refs. 23, 24), however, that at the temperature


at which the irradiation is carried out theirradiation-induced defects are trapped at pre-cipitated metallic lithium and form stable clusters.The subsequent annealing of the defect cluster,once it has reached a stable configuration or criti-cal size, could be quite independent of the presenceof lithium. Hence, the presence or mobility oflithium at the higher annealing temperature wouldnot necessarily influence the annealing kinetics ofthe defect clusters.

REFERENCES

1. Seitz, F., Disc. Faraday Soc., Vol. 5,p. 271, 1949.

2. Seitz, F., and Koehler, J. S., in SolidState Physics. Edited by F. Seitz and D.Turnbull, Academic Press Inc., Vol. 2,1956.

3. Siegel, S., Phys. Rev., Vol. 75, p. 1823,1949.

4. Billington, D. S., Nucleonics, Vol. 14,p. 54, 1956.

5. Brooks, H., Am. Rev. Nuclear Sci., Vol. 6,p. 215, 1956.

6. James, H. M., and Lark-Horovitz, K.,Z. Physick Chem., Vol. 198, p. 107, 1951.

7. Gossick, B. R., J. Appl. Phys., Vol. 30,p. 1214, 1959.

8. Crawford, J. H., and Cleland, J. W., J.Appl. Phys., Vol. 30, p. 1204, 1959.

9. Closser, W. H., J. Appl. Phys., Vol. 31,p. 1693, 1960.

10. Stein, H. J., J. Appl. Phys., Vol. 31,p. 1309, 1966.

11. Stein, H. J., Phys. Rev., Vol. 163, p. 801,1967.


13. Fujita, F. E., and Gonser, U., J. Phys.Soc. Japan, Vol. 13, p. 1968, 1958.

14. Parsons, J. R., Balluffi, R. W., andKoehler, J. S., Appl. Phys. Letters, Vol. 1,p. 57, 1962.

15. Hemment, P. L. F., and Gunnersen, E. M.,J. Appl. Phys., Vol. 37, p. 2912, 1966.

16. Pankratz, J. M., Sprague, J. A., andRudee, M. L., J. Appl. Phys., Vol. 39,p. 101, 1968.

17. Bertolotti, M., et al., Nuovo Cimento,Vol. 29, p. 1200, 1963.

18. Bertolotti, M., et al., J. Appl. Phys.,Vol. 36, p. 3506, 1965.

19. Bertolotti, M., et al., J. Appl. Phys.,Vol. 38, p. 2645, 1967.

20. Bertolotti, M., Radiation Effects in Semi-conductors, p. 311, Plenum Press, 1968.


22. Passenheim, B. C., and Naber, J. A.,Radiation Effects, Vol. 2, p. 229, 1970.

23. Sargent, G. A., Research Study of DamageProduced in Silicon Semiconductors byNeutron Irradiation, Technical ReportUKY 31-70-Met 13, University of Kentucky,1970.

24. Berger, R. A., et al., Radiation Effects inSilicon Solar Cells, Quarterly Report byGulf General Atomic to Jet PropulsionLaboratory, October 1969.


TjK"'* *r

Fig. 1. Surface rep l ica of an undoped and un i r r ad i a t ed so l a r ce l l

Fig . 2. Surface r ep l i ca of an undoped so la r ce l l i r r a d i a t e d with lO*! n e u t r o n s / c m ^

J P L Technica l Memorandum 33-491

109

E

4)a.

C

0

I10d

*

Li DopantAtoms/cm 3

01015016

i'07

10" 10'2Neutron Dose (Neutrons per cm" )

0

A

10'3

Fig. 3. Average defect density as a functionof irradiation dose and lithiumdopant

I0

-10xE

4)

E

a

4)C,

Neutron Dose (Neutrons per cm2 )

Fig. 4. Average defect diameter as a functionof neutron dose and lithium dopant


loe

1c4IC Li O1 Atomsm/c , Dose 10" n/cm2

Eo

Li 0, Dose 10 '°n/cm 2

.d)

500 1000

Temperature ° K

Fig. 5. Effect of thermal annealing ondefect density

15

Li O Dose 10 n/cm

IoOxE

Co

E0r

- 5 - Li O4D ~ Dose 10'3 n/cm2

Li I': Atoms//cm Dose nI m2

Li 10'7 Atoms/tm 3 Dose I 13 n/cm2

500 1000

Annealing Temperature OK

Fig. 6. Effect of thermal annealing ondefect diameter


Li IOrrAtoms/cmr, Dose 10' n/cm'

90

RADIATION DAMAGE AND ANNEALING OF LITHIUM-DOPEDSILICON SOLAR CELLS*

R. L. StatlerNaval Research Laboratory

Washington, D. C.

II. EXPERIMENTAL

The origination of the lithium-doped radiation-hardened silicon solar cell in 1966 (Ref. 1) hasbeen followed by an intensive effort to study thespecifics of the radiation damage and recoveryprocesses. As one step in the direction towardoptimizing the chemical and physical parametersof the cell for maximum radiation hardness, it isessential to determine the performance of thissolar cell in various radiation environments. Theamount and rate of this recovery process havebeen shown to depend on the oxygen impurity in thesilicon, the type of irradiating particle, thefluence, and the temperature of the cell followingthe irradiation. In addition, low-flux rate irra-diation as contrasted with Van de Graaff bombard-ment, have provided valuable information on thebehavior of these cells in a simulated space en-vironment. A part of this paper will discuss therecent results of a continuing real-time-rate irra-diation (Ref. 2) which was begun in September1969. Since that time, better lithium-doped cellshave been made, and placed into the study. Weshall also discuss the results of 1-MeV electronand 4-MeV proton Van de Graaff irradiation andannealing of a solar cell made from crucible-grown silicon which has been lithium-diffused for8 h at 325°C.

Three modes of radiation were used to studydamage and recovery in the lithium solar cells.They comprised 1. 2-MeV gamma photons fromthe Naval Research Laboratory (NRL) Cobalt 60source, 1-MeV electrons from a Van de Graaff,and 4-MeV protons from a Van de Graaff. Thedamage caused by the 1.2-MeV gamma photoncomes about from energized electrons which areproduced in the chamber walls, in the solar cellholder, and within the solar cell by means ofCompton interactions of the gamma ray with elec-trons in the material (Ref. 3). These electronshave a spectrum of energies ranging upward to0. 8 MeV. These energetic electrons create latticevacancies in the silicon, followed by the formationof defects and recombination centers similar tothose occurring in I-MeV electron irradiatedsilicon. A first approximation for equivalency ofdamage in solar cells from Cobalt 60 gammas, ascompared with electrons, can be made by deter-mining the number of gamma photons that willproduce the same number of lattice displacementsas a 1-MeV electron. If values (Ref. 4) are usedfor the total number of displaced silicon atomsper unit of incident flux of 10-2 for l-MeV gammaphotons and 4.6 for 1-MeV electrons, the equiv-alent electron dose corresponding to 1 rad (Si),

*This work was performed for the Jet Propulsion- Laboratory, sponsored by the National Aeronauticsand Space Administration under Contract No. NAS7-100.


I. INTRODUCTION

91

which is 2. 22 X 107 photons/cm2 , is 4.35 X 106e/cm 2 . This equivalency factor is applicable onlywhen the gamma environment is one of electronicequilibrium for the irradiated sample. In thiscase, this condition is essentially satisfied.

The flux values, in units of particles/cm 2 -s,varied widely. In the case of gamma flux, thevalue was 5 X 10 6 e/cm 2 -s, later increased to2.7 X 107 e/cm Z -s. For the electron Van deGraaff irradiations the flux was 5 X 1011 e/cm 2 -s,and for the proton Van de Graaff, 3 X 109particles/cm 2 -s.

The solar cells discussed here are five types.There are four groups of Heliotek lithium-dopedP/N cells, one of Centralab lithium-doped P/N ccells, and a group of Centralab 10-l-cm N/Pflight-quality solar cells. Table 1 shows the ex-perimental matrix for the gamma ray portion ofthis study.

The experimental apparatus for the gammairradiations consisted of three stainless-steelcylindrical cans about 7. 6 cm in diameter, 24 cmlong, with a 0. 51-cm-thick wall. The solar cellswere held in contact with temperature-controlledbrass plates by means of spring clips. Each cellwas loaded with a 10-Q resistor, with pressurecontact made through the spring clips and brassplate.

Illumination was provided during irradiationby means of five 12-V automobile lamps in eachcan. Replacement of lamps was required aboutevery 3 mon because of radiation darkening of theglass bulbs. Then cans were evacuated after seal-ing, and back-filled with 0.7 N/cm2 (1 psi) ofargon to provide for thermal conduction of the heatfrom the lamps to the chamber walls. (Undervacuum, the bulbs failed in a few hours. )

The cell temperatures were controlled bymeans of a combination of electrical strip heatersand water-carrying copper tubing fastened to therear of the solar cell mounting plates. One canwas held at 30 0 C, and two cans at 60°C with avariation of L1°C. For measurements of the I-Vcurves, the solar cells were removed from thecans and placed under a Spectrosun X-Z5L solarsimulator calibrated for 140 mW/cm2 air masszero. Solar cell temperature was 25°C for allmeasurements at the simulator. During the timesthe cells were out of the irradiation chambers,they were held at dry-ice temperature to preventannealing, except for the actual measurementtime.

The irradiations at the NRL 2-MeV electronVan de Graaff were carried out on solar cells atroom temperature, with forced-air cooling on thesample. The fluence obtained was 5 X 1014l-MeV e/cm 2 . All of the solar simulator mea-surements were performed with cell temperaturesof 25°C. The annealing of the solar cells afterirradiation was done in forced-draft laboratoryovens at 60 and 30°C with solar simulator mea-surements being performed at intervals through-out the recovery period.

The 4-MeV proton bombardment was per-formed at the NRL 5-MeV proton Van de Graaff toa fluence of 2. 2 X 1011 p/m2 , with the two solar

cells in a vacuum of 6. 7 X 10-3 N/m2 during theirradiation. As in the case of the electron-irradiated cells, these samples were measured atthe solar simulator at 25°C at intervals through-out their 60°C annealing period.

III. RESULTS

The results of Cobalt 60 gamma irradiationsover a period of 18 mon will be discussed. Duringthis time the solar cells were removed from thesource 12 times for measurements. A totalgamma exposure of 1.4 X 108 R was received bythe cells, equivalent to a 1-MeV electron fluenceof 6. 1 X 1014 e/cm 2 . It should be mentioned thatthe dose rate was increased from 5 X 106 e/cm2 -sto 2. 7 X 107 e/cm2 -s after a fluence of 2 X 1014.The trend of the data show that the results werenot affected by this increase. Figure 1 is a plotof the power output of the cells as they are irra-diated at 60°C in a continuous gamma ray environ-ment. The power is measured with a cell tem-perature of 25°C at one sun of illumination. Itcan be seen from Table 2 that the four types oflithium cells were much lower in initial efficiencythan the N/P 10 Q-cm cell which was used forcomparison. The results of the 30 0 C irradiationshowed slightly more damage than these at highfluences, although up to 1 X 1014 e/cm 2 there wasgenerally little difference between the 30 and 60°Cdata (Ref. 2). The groups H2 and H5 were desig-nated generally to be low lithium content. In fact,junction capacity measurements on H5 indicatedno lithium concentration near the junction. Thepoor radiation hardness attests to this finding.On the other hand, the more heavily doped H6 andH9 were also so low in efficiency that they werenever competitive to the HZ and N/P cells in thecourse of the experiment.

A quite different picture of the performanceof a lithium cell is given in Fig. 2. The maximumefficiency shown here for the CllA cell is 11.2%compared with 10.4% for the N/P cell. However,the rapid decrease in power brings the lithium cellperformance exactly to the level of the N/P cellat 60°C irradiation temperature. The results ofthe 30°C irradiation show that the Cl1A cell out-put is definitely lower. One might assume thatthe increased mobility of lithium at the highertemperature causes a more rapid annealing ratefor the damage centers which are being produced.The most important point of this figure is that thepower level of the C11A cell ultimately stays abovethe N/P cell at fluences beyond 4 X 1014 e/cm2 ,with an improvement of about 5%. In terms oflong exposures in an electron environment, itappears that the lithium cell will offer an advantageover the conventional N/P cell.

The next part of the study was to investigatethe annealing rate of rapidly damaged solar cellsas a function of temperature and cell type. TheC lA lithium- doped cell was chosen because of itsgood characteristics, and was compared with thesame type of N/P cell used throughout this work.Figure 3 depicts the post-irradiation annealing ofthe power lost through 1-MeV electron irradiationin C11A and N/P solar cells. The N/P cell showsa slight recovery immediately after the irradia-tion, then quickly saturating and even slightly de-clining. The CllA cell shows additional damageoccurring within the first few hours following the


irradiation. The 30 0 C annealing does not regainany of the power lost during the irradiation. How-ever, annealing at 60°C produces a large recoveryof the lost power. This result can be seen morequantitatively in Fig. 4 where the percentage ofrecovered power is shown. Percentage recoveryis defined as

P -Pa r X 100

P0 - Pr

where P0 is the original power, Pr is measuredafter the irradiation, and Pa after annealing. InFig. 4 one observes that the lithium cell hasrecovered 42% of its power after Z00 h at 60°C,while the N/P cell has recovered only 100%o. Inactual power level, the lithium cell is about 5%greater at 200 h (Fig. 3).

All of the previous results, for 60'C gammaradiation and 60°C annealing after electron radia-tion are summarized in Fig. 5. The maximumannealing time for this data is 100 days, whereasthe time to gamma irradiate cells to this dose isabout 200 days. With this fact in mind, we ob-serve that more damage is produced during theslow irradiation with simultaneous annealing thanis left in the cells when they are rapidly damaged,then allowed to anneal for an equivalent length oftime. Some of the Heliotek lithium cells used inthe first part of the gamma experiment were re-moved after a fluence of 2 X 1014 e/cm 2 andplaced in a 60°C annealing temperature with andwithout illumination. No recovery of these cellswas observed; indeed, many cells degraded some-what further. Thus it is unlikely that the gamma-irradiated C11A cells would show recovery if theradiation were removed. It is interesting to notethat in both experimental situations, the lithium-doped solar cells demonstrate a power margin ofabout 5% over the N/P cell.

The final portion of the experiment consistedof a 4-MeV proton bombardment of two CllA cellsto a fluence of 2.2 X 1011 p/cm2 . The cells wereallowed to anneal in an oven at 60°C for 800 h, withI-V measurements made at intervals throughout.This data is summarized in Fig. 6, along with theelectron-damage annealing data from Fig. 3. Inboth electron and proton Van de Graaff bombard-ment, the P/N lithium cells are degraded muchmore than the N/P cell. This difference betweenradiation hardness in P-type and N-type siliconsolar cells has been established for many years(Refs. 5, 6). During the annealing, the lithiumcell recovers to a higher power level than theN/P cell after 800 h at 60°C. Carter and Downing(Ref. 7) report like behavior in similar solar cellsirradiated to 3 X 1015 l-MeV e/cm 2 and annealedat 100°C. In their case, the lithium cell powerrecovered to the value for the N/P cell in 2 h. Sincerecovery rate depends on irradiation fluence andannealing temperature (Ref. 8), these two resultsare notnecessarily conflicting. The recoveryfromprotondamagefollows a different behavior. Thereis no additional damage found during the first fewhours after irradiation; recovery begins at onceand proceeds more rapidly and to a higher degreethan for electron damage. The percentage recov-ery at 800 h is 38% for electron damage and 68%for proton damage. The proton data shown for the

N/P cell is a typical value extrapolated from aprevious experiment (Ref. 6). No annealing datawas available for the N/P cell.

IV.. SUMMARY

Very conclusive evidence has been presentedthat a lithium-diffused crucible-grown siliconsolar cell can be made with better efficiency thanthe flight-quality N/P 10 Q-cm solar cell. Whenthis lithium cell is exposed to a continuous radia-tion environment at 60°C (electron spectrum fromgamma rays) it has a higher power output than theN/P cell after a fluence of 4 X 1014 e/cm Z

(equivalent 1 MeV).

A comparison of annealing of proton- andelectron-damage in this lithium cell reveals adecidedly faster rate of recovery and higher levelof recoverable power from the proton effects.This fact strongly suggests that the lithium cell ina low-rate continuous proton flux would have evengreater superiority over the N/P cell than it hasin the case of electrons. Therefore, the lithiumcell shows a good potential for many space mis-sions where the proton flux is a significant fractionof the radiation field to be encountered. This isthe case for many specified earth-orbit trajector-ies as well as interplanetary missions.

The need for additional proton radiation stud-ies is obvious, in view of the superior performanceof the lithium cells. Such work will be carried oututilizing various other proton energies of interest.

REFERENCES

1. Wysocki, J. J., "Lithium-Doped RadiationResistant Silicon Solar Cells, " IEEE Trans.NuclSci., NS-13, p. 168, 1966.

2. Statler, R. L., "Real-Time Irradiation ofLithium-Doped Solar Cells, " Proceedings ofthe Third Annual Conference on Effects ofLithium-Doping on Silicon Solar Cells, Tech-nical Memorandum 33-467, pp. 81-84,April 1971.

3. Price, B. T., and Horton, C. C., RadiationShielding, Pergamon Press, New York,pp. 24-30, 90-92, 1957.

4. Transient-Radiation Effects on ElectronicsHandbook, DASA 1420, edited by R. K.Thatcher, Battelle Memorial Institute,Columbus, Ohio, August 1967.

5. Statler, R. L., "Electron BombardmentDamage in Silicon Solar Cells," NRL Report6091, Naval Research Laboratory,Washington, D. C., October 1964.

6. Statler, R. L., "Radiation Damage in SiliconSolar Cells From 4. 6-MeV Proton Bombard-ment," NRL Report 6333, Naval Research Lab-oratory, Washington, D. C. , November 1965.

7. Carter, J. R., and Downing, R. G., "Be-havior of Lithium in Irradiated Solar Cells,"Colloque International Sur Les CellulesSolaires, Toulouse, France, July 1970.

8. Berman, P. A., "Effects of Lithium Doping onthe Behavior of Silicon and Silicon Solar Cells, "Technical Report 32-1514, Jet Propulsion Lab-oratory, Pasadena, Calif., February 1971.


Type

Li P/Ncrucible

Li P/Ncrucible

Li P/NFZ

Li P/NFZ

Li P/Ncrucible

N/Pcrucible10 Q2-cm

Table 1. Co 6

Experiment sample matrix

Cell group

IH-2

H-6

H-5

H-9

Cl lA

Centralab

Quantity of samples

Irradiated ControlLi diffusionparameter s

Illuminated30°C

5

5

5

5

2

5

All cells are illuminated with tungsten light and are individually loaded with a 10-Q2 resistordeveloping a load voltage of 0. 21 to 0. 24 V. The cells are removed from the chambers formeasuring the electrical performance at 25°C under a solar simulator.

90 mnin, 425°C-60 min, 425°C

90 min, 450°C-60 min, 450°C

90 min, 350°C-60 min, 350°C

90 min, 425°C-60 min, 425°C

480 min, 325°C

NA

Illum inated60°C

Table 2. Photovoltaic parameters of experimental cells

Cell group Type IS, mA Pax mW Efficiency, %lo

H-2 Li CG 64. 5 27. 8 9.9

H-6 Li CG 60. 0 24.7 8.8

H-5 Li FZ 70. 0 27.4 9.8

H-9 Li FZ 61. 0 25. 1 9.0

ClIA Li CG 71. 6 31. 0 11. 1

N/P 10 Q-cm 71.5 29.2 10.4

3

3

3

3

Illuminated60°C

5

5

5

5

2

5


A

! 0

I 1 I I

CA

O

A

0 0

0

0

A

A

O o

o

O

00

o

I I I

N-ON-P, 10 OHM-CM

H2, Li DOPED CG

H6, Li DOPED CG

H5, Li DOPED FZ

H9, Li DOPED F2

I I IoV 2 4 6 8 1014 2 4 6 8 o015

FLUENCE ( E/CM 2 - i MEV EQUIVALENT)

Fig. 1. Power loss in lithium-doped P-on-N solar cell1 compared toan N-on-P 10 Q-cm solar cell irradiated by CoO° gamma fluxat 60°C

0

A

A

0

A

A

0

A

0A

AA

A

A

32

30

28

I-

-J_j

o

0Q..

26

24

22

20

I I I I IOCOBALT 60 GAMMA IRRADIATION

-O

18 L/

I I I

o N-ON-P, 10 OHM-CM, Ti =60 0 C

O CilA, L; DOPED CG, Ti

= 6 00 C

A CIlA, Li DOPED CG, Ti

= 300 C

I I I I I I I I

oV 2 4 6 8 1014 2 4 6 8

FLUENCE (E/CM2

- MEV EQUIVALENT)

Fig. 2. Power loss in lithium-doped P-on-N solar cells and 10 Q-cmN-on-P! solar cells irradiated at 30 and 60 °C by Co60 gammaflux


1015

95

28

26

24

3

3 22

-r

20

018

18

1R6

30r

DT r I

o

Q-U

32

30O

A PRE-RAD

Io

28-

I I I I I I I I

VAN DE GRAAFF IRRADIATION

5 x10 4 1 MEV E/CM 2

O N-ON-P, 10 OHM-CM, 600C

o CIIA, Li DOPED CG, 60°C

L CllA, Li DOPED CG, 300C

ANNEAL

ANNEAL

ANNEAL

26

24-

22-

_ /oPOST-RAD

20-\o

a - aA

I I I I I I I I I I100 1000

16 I IIIt0

TIME (HOURS) AFTER IRRADIATION

Fig. 3. Post-radiation annealing of power lost through l-MeV electronirradiation in lithium-doped P-on-N solar cells and N-on-P10 f-cm solar cells

50 -

40

I I I II I IF II i I I ,I

VAN DE GRAAFF IRRADIATION

5 x04 1 MEV E/CM 2

- ........ -0.....o N-ON-P, 10 OHM-CM, 60-C ANNEAL

-- 0 CIIA, Li DOPED, CG, 60 0 C ANNEAL

A .CIIA, Li DOPED CG, 300 C ANNEAL

_ /--~h

I II I I, _L

-10 _

10 100 1000

I I I I

10,000


Fig. 4. Percentage of power recovered through post-radiation annealingin solar cells irradiated by 1-MeV electrons


V,F-

I-

-J3:

w

0a-

I I I I

10,000

60

30

10

0

zw0

wa.

W

00wa

a:w3.0a-

I

'18

I I 1 I

.. -1 -

20C

I

96

I ,I I I,

I I Iw lI I

3 N-ON-P, 10 OHM-CM _

O CliA, LI DOPED CG AFTE60C ANNEALAFTER 5xIO 4 E/CM 2

COBALT 60 GAMMA IRRADIATED

I r I I4 6 8 1014 2

I I4 5 6 8 1015

FLUENCE ( E/CM - I MEV EQUIVALENT)

Fig. 5. Comparison of power recovery by annealing following a t-MeVelectron radiation, and the power lost in a real-time Co Ogamma radiation for lithium-doped P-on-N solar cells andN-on-P 10 0-cm solar cells. The annealing time extendsto 2300 h, whereas the time to gamma irradiate to 5 X 1014is 200 days


32

30

- IRE-RAD

-b

28-

26-

24-

I--_jJ

30O_ 22

D

t- POST

RAD

-- 020

2

97

32 I

-PRERADRAD30 5x4 MEV E/CM 2 2.2 x110 4MEV P/CM 2

3 0 Q N-ON-P, 10 OHM-CM A TYPICAL N-ON-P, 10 OHM-CM

0 CIA, Li DOPED CG 0 CliA, Li DOPED CG

600 C ANNEALING TEMPERATURE

26 -

_J

22 I

l 20 I

18

0 110 10O 1000 10,000


Fig. 6. Comparison of power recovery following a l -MeV electronradiation and a 4-MeV proton radiation for lithium-dopedP-on-N solar cells


LOW FLUX IRRADIATION OF LITHIUM-DOPED SOLAR CELLS

D. L. ReynardPhilco-Ford Corporation

Palo Alto, Calif.

LOW FLUX IRRADIATION OF LITHIUM-DOPED SOLAR CELLS

W. KrullLockheed-Georgia Company

Marietta, Ga.

The above papers were not available at time of publication.


5

RADIATION DAMAGE ANNEALING KINETICS

M. S. DresselhausDepartment of Electrical Engineering,

Center for Materials Science and Engineering,Massachusetts Institute of Technology, Cambridge, Mass.

I. INTRODUCTION

Our primary task is to understand how thepresence of lithium assists in the recovery processof irradiated silicon solar cells. Although lithiumis more effective in accelerating the recovery ratefrom neutron and proton bombardment, lithium hasalso proved useful in the recovery process fromelectron irradiation. Since neutron irradiation isnot especially important for typical NASA spacemissions involving solar cells, and, furthermore,since it is relatively easy to provide protectionagainst proton damage, our efforts have beenmainly directed toward the study of recovery fromelectron irradiation.

II. BACKGROUND

To see how to tackle the problem let us startat the beginning and consider the solar cell as ap-n junction where for our case, the base regionis the n-region, as is illustrated in Fig. 1. Lightincident on this base region produces electron-holepairs, with the holes capable of surmounting thepotential barrier at the p-n junction and reachingthe collecting electrode. We are accustomed tothinking of typical carrier diffusion lengths incrucible-grown quartz as L - 1.2 X 10- 4 m so thatcollection can result from electron-hole pairsformed far from the junction. It is this long dif-fusion length that encourages us to think that thekinetics in bulk silicon might be the crucial factorin determining solar cell behavior. We shall nowshow why, on one hand, the consideration of bulkeffects might by themselves lead to erroneousconclusions and, on the other hand, why it is


nevertheless important to consider the kineticsof bulk silicon.

Any kinetics modeling based on a first-principles approach requires the observation of:(1) the defects that are formed during irradiation,(2) how these defects are annealed out in the re-covery process, and (3) how these defects affectthe diffusion length, minority carrier lifetime,carrier removal rate, etc. In dealing with thisproblem, we must classify the defects into variouscategories, and since each category tends to havea rather different cross section for minority car-rier removal, different concentration ranges couldsignificantly affect the minority carrier diffusionlength.

In the first category we have the shallow donorand acceptor levels. These defects do not repre-sent a major perturbation to the periodic potentialcharacteristic of the host silicon lattice and con-sequently offer just a small cross-section forminority carrier removal. Typical concentrationsof such defects have little effect on the minoritycarrier transport problem. On the other hand,the deep defect levels represent appreciable per-turbations to the periodic potential and offer largecross-sections for a number of carrier removalprocesses. For such defects (e. g. , gold), a con-centration as low as -10 1 9 /rn3 (which is merely atrace concentration) can appreciably affect theminority carrier diffusion length (Ref. 1). Somedeep defect levels act like traps and are impor-tant in minority carrier removal processes assoc-iated with transient phenomena. But in the steadystate, once a trap is filled, it will remain filled.

Preceding page blank 101

On the other hand, a single recombination centercan continuously remove minority carriers bycombining them with majority carriers in either aradiative or a non-radiative recombinationprocess.If the process is radiative, it is, in principle,observable by radiation recombination measure-ments; on the other hand, non-radiative processesare very difficult to study experimentally.

The types of defects most prevalent in the baseregion before irradiation are: (1) the shallowdonor defects which are present in concentrationsas large as 10 2 1 /m 3 , and (2) deep impurity defectswhich are present in trace concentrations. Beforeirradiation, it is these trace quantities of deepimpurity recombination centers which control theminority carrier diffusion length. The types ofdefects introduced by irradiation with electronsin the energy range of -1 MeV are predominantlyin the deep defect level classification, since theradiation-induced defects (such as the vacancy)represent a major perturbation to the periodiclattice of the host crystal. For fluences typicalof the space environment of a synchronous orbitmission of 5-yr duration, the radiation-induceddeep defect concentrations will generally exceedthe native deep-defect concentrations. Uponannealing, the radiation-induced defects tend toform complexes which are also associated withdeep level states but generally represent less ofa perturbation to the periodic potential of the hostcrystal. In both the irradiation and recoveryprocesses, majority carriers are removed whenthey become bound to the various defect centers.This carrier removal results in a degradation ofthe junction.

To consider the kinetics problem from firstprinciples, we must identify the primary radiation-induced defects and determine their effect on theminority carrier lifetime. We must then identifythe centers that result upon annealing of the pri-mary centers and determine the effect of these newcenters on the minority carrier lifetime. In thisconnection it would be desirable to carry out mea-surements of both the minority carrier lifetimeand the energy level spectrum of the defects. Suchjoint measurements should be made on samplesthat are well-characterized with regard to bothmaterial parameters and radiation environment.

III. EXPERIMENTAL TECHNIQUES

A number of experimental techniques areavailable for the study of defect levels. The infra-red absorption technique is one of the most usefultechniques for establishing the energy levels assoc-iated with defects. We must remember here thatthere are many different species of defects presentin the base region of an irradiated solar cell. Fur-thermore, each defect can be found in any one of anumber of energy states as shown in Fig. 2. Thisfigure shows the energy levels of the divacancy,one of the common defects occurring in irradiatedfloat-zone silicon (Ref. 2). If the photon energy ofthe infrared radiation is just sufficient to exciteelectrons from the singly charged (-1) occupiedlevel, 0. 4 eV below the Fermi level, to the con-duction band, there is an absorption of power fromthe incident beam. An experiment closely relatedto the infrared absorption is the photoconductivityexperiment whereby we look at the current of thephotoexcited electrons rather than at the powerabsorbed from the incident beam. These experi-

ments are not actually equivalent insofar as thefinal state in the photoconductivity experimentmust be a band state, while a bound state is anacceptable final state in the infrared absorptionprocess. At any rate, both types of experimentshave, in fact, been carried out under our programand some results of the photoconductivity studiesby Corelli are shown in Fig. 3. In interpretingsuch curves, the maxima in the photocurrent areidentified with the resonant absorption from or toa defect level. These measurements by them-selves do not identify the nature of the defectcenters; to make such identifications it is neces-sary to study concentration dependences, temper-ature dependences, effects of stress or possiblymagnetic fields. To study the effect of thesecenters on solar cell characteristics, measure-ments of the minority carrier lifetimes should bemade on the same samples and correlated withthe optical data.

There are, however, some more fundamentaldifficulties with these data: in order to observethese defects optically, it was necessary to irra-diate the samples with fluences which create defectdensities far in excess of the lithium concentration.These fluences are larger than are of practicalinterest for space missions. Since the benefits oflithium donors are realized primarily when thelithium concentration exceeds the damage centerconcentration (Ref. 3), these experiments are onlyof limited utility for our kinetic modeling program.Furthermore, this difficulty would not be resolvedby increasing the lithium concentration, since thelithium concentrations used in this work werealready as high as might be of interest to possiblesolar cell applications. In order for optical tech-niques to be useful to us, the sensitivity of thetechnique must be increased by perhaps two ordersof magnitude. To accomplish this, a modulationtechnique could be exploited to advantage.

To study defect centers in bulk material somerather elegant recombination luminescence studieshave been carried out by the University of Illinoisgroup. In recombination luminescence, one mea-sures the spectrum of the radiation emitted when,for example, an electron associated with a partic-ular defect level combines with trapped hole orwith hole in the valence band. An example of sucha spectrum is shown in Fig. 4. Because recombi-nation luminescence in silicon is not an efficientprocess, these experiments have also been plaguedwith the same difficulties as the optical experi-ments insofar as it has been necessary to employhigh irradiation fluences in order to observe re-combination luminescence signals. Recently,increased sensitivity has been achieved in siliconthrough use of a laser source to achieve the initialexcitation (Ref. 4). Perhaps such techniques couldalso be employed in the luminescence recombi-nation work in our program. In the luminescencestudies made under our program, considerable suc-cess has been achieved in the identification of thenature of the observed defects. For example, thestructure at 0.97 eV is identified with the divacancy0.19 eVbelow the conductionband and the structure at0.79 eV is identifiedwith a two- oxygen defect complex.To make such studies valuable to a kinetic model-ing effort, it will be necessary to correlate theluminescence studies with minority carrier life-time measurements in order to understand the roleof such defects in the solar cell degradation uponirradiation and subsequent recovery.


The difficulty with the bulk measurementsmade under conditions of intense irradiation canalso be understood in terms of Fig. 5, where thedependence of the minority carrier diffusion lengthL is plotted as a function of fluence for 1-MeVelectron irradiation on standard n/p solar cells,made from crucible grown silicon (Ref. 5). WhileL is -1. 20 X 10-4 m for the pre-irradiated cells,we see a rapid decrease in L with increasingfluence so that for a fluence -10 2 0 /m 2 the dif-fusion length is already below 10- 5 m. This fig-ure tells us that if we are considering fluences inexcess of -102 0 /m 2 , the solar cell current gener-ation in the base region will become less impor-tant compared with the behavior in the vicinity ofthe junction. Since the impurity concentrationsand defect densities tend to be rather differentnear the junction as compared with the bulk be-havior, the kind of data on bulk silicon that isof primary interest for kinetic modeling is thatfor low fluence levels corresponding to typicalspace environments. On the other hand, weshould not conclude from this figure that mea-surements on bulk silicon are irrelevant. Weshall soon see why it is important to considerboth the data on the devices and data on the bulkmate rial.

Another powerful technique that has beenapplied to the study of radiation-induced defectsand their subsequent annealing in the presenceof lithium is the electron spin resonance (ESR)technique. This technique is applicable to thestudy of defects in a singly charged state; forthese states, the ESR technique provides a micro-scopic and detailed probe. With this technique,it has been possible to identify resonances asso-ciated with the oxygen-vacancy, the divacancy,the phosphorus-vacancy, the lithium-oxygen, aswell as many other centers. The interpretationof these experiments indicates that for oxygen-lean silicon, the presence of lithium favors theformation of lithium-vacancy complexes ratherthan phosphorus-vacancy complexes and that inthe annealing process, lithium can help toneutralize phosphorus in the phorphorus-vacancycomplexes that happen to be present (Ref. 6).For the oxygen-rich silicon, the presence oflithium tends to favor formation of LiOV centersrather than the OV center itself. The fluencesnecessary for observation of ESR signals arestill rather large (like 1020 to 1021 electrons/m 2 )

but somewhat smaller than are required for theoptical measurements.

We have, however, encountered several dif-ficulties in utilizing some of the ESR data: (1) Themeasurements employ 30-MeV rather than 1-MeVelectrons, the lower energy matching more closelyother work in the program as well as the environ-mental conditions in space. Since the productionand annealing rates for a given defect are stronglyenergy dependent, it would be desirable to haveESR data on similar samples as a function ofelectron irradiation energy. (2) The lithium con-centration tends to be higher than the range ofinterest for solar cell applications and yet the lith-ium concentration is often less than the density ofirradiation defects. Thus, the parameters do notreally fall into the desired range for solar cellmodeling. (3) Not all centers can be studied bythe ESR method because they have the wrongcharge state or are unoccupied.

From the point of view of cell modeling, our%nost useful data have come from less microscopican'd perhaps less elegant techniques, but neverthe-less capable of yielding information on the defectproduction and annealing in the parameter rangeof interest to solar cells and the type of spaceenvironment where the cells will be utilized. Suchtechniques are the bulk Hall and resistivity mea-surements, minority carrier lifetimes in bulksilicon and in solar cell devices, and capacitancemeasurements. We will have more to say aboutrelating such data to the kinetic modeling later.From the temperature dependence of the Halleffect and from analysis of the temperature depen-dence of the minority carrier lifetime data, it isalso possible to deduce defect levels. Such tech-niques have been widely exploited for this purposein our program by the RCA, TRW, and Gulf-Atomic groups. If we now consider all the defectlevels that have been reported in the literature todate we get the typical representation shown inFig. 6. This diagram shows that a large numberof defects have been identified and that a givendefect can exist in various energy states. Notonly is the number of defect levels overwhelminglylarge, but the quantity of reliable and necessaryinformation on each level is discouragingly small.From this situation, we must conclude that it isnot feasible at this time to construct a first-principles kinetic model for the annealing ofelectron radiation damage in lithium-diffusedsilicon solar cells.

IV. KINETICS MODELING

We have, therefore, looked toward a phenom-enological approach whereby a parametric modelis developed to predict lifetime damage constantsand carrier removal rates relevant to the oper-ation of the solar cell. In dealing with the kineticsproblem, we divide it into two parts: (1) a deter-mination of the minority carrier lifetime as afunction of the material and environmental param-eters, and (2) the deduction of solar cell perfor-mance from the minority carrier lifetime, andfrom doping profiles in the n- and p-material andof the diffused lithium in the solar cell. Thesecond part of the problem has been dealt with bythe Exotech and Gulf-Atomic groups. For thisreason, we have mainly concerned ourselves withthe determination of minority carrier lifetimes.(We have recently learned that the Gulf-Atomicgroup has also been successful in modeling minor-ity carrier lifetimes under a variety of materialand environmental parameters).

Almost all lifetime data is analyzed in termsof the Shockley-Read-Hall theory for recombi-nation processes. According to this theory, thelifetime for a single level defect is given by theexpres sion,

/n0+ n1 + An\T = T + + n +

PO0 P0 + anT (po + l An\n0 n0 + P0 + An)

where the lifetimes in heavily p-type and n-typematerial are given by

1TPO NRvth(T) crp(T)


and

1n 0 NRVth(T) n ( T )

In these expressions

NR

= recombination center density

vth = thermal velocity

n¢ = cross sections for recombination ofp, n holes or electrons

nOP p= thermal equilibrium carrierconcentration

An = Ap = concentration of excesscarriers

T = temperature

nl = N e[(EREC)/kT ]

1 = N e[(EVER)/kT]Niv

where Nc and Nv are integrals over the density ofstates for the conduction and valence bands,respectively, and

N = JP (E) e[-(EEc)/kTdE]

Nv

= / Pv ( E) e[-(EV-

E)/kTdE]

In order to interpret lifetime versus temperaturedata, an explicit temperature dependence of thecross sections and of the thermal velocity vth mustbe assumed. The temperature dependence of -r iscomplicated but has been treated theoretically(Ref. 8). Although there exist divergent opinionson the temperature dependence of a-, there isgeneral agreement that the temperature dependenceof vth goes as T1/2 . Most interpretations of life-time data consider (r(T) -T-1/2 so that ap- and-n0 are temperature independent. If temperature

dependences for Vth and o-(T) are introduced, thenit is possible to fit the experimental lifetime mea-surements as a function of temperature and therebyto deduce the energy of the defect level ER.

We are at present trying to understand how touse the Shockley-Hall-Read theory to find thiseffective defect level in an unambiguous way. Itis our hope that the minority carrier lifetime iscontrolled by one or two dominant defects, e. g.,the divacancy in oxygen-lean silicon and theoxygen-vacancy complex in oxygen-rich silicon.If this is the case, then we might expect the effec-tive defect level ER to be correlated with an impor-tant defect level that has been identified by moredirect measurements on bulk material, e. g.,optical, luminescence, ESR, temperature depen-dence of the Hall effect, etc. In this context, itwould be most desirable to make measurements onthe same samples by different techniques.

The application of the Shockley-Read-Hallapproach to the minority carrier lifetime problemis itself beset with complications. To illustratethese complications, Fig. 7 displays the variousenergy levels identified with radiation defects pro-duced by 1-MeV electrons in n-type silicon. Inall cases, the levels were obtained from Shockley-Read-Hall analysis of the temperature dependenceof lifetime data. The two most important reasonsfor the variety of effective levels in this diagramare: (1) differences in material parameters in then-type silicon, and (2) differences in the use of theShockley-Read-Hall theory. Differences in thematerial parameters are difficult to handle be-cause it has not been customary to give a completecharacterization in the literature of the materialand environmental parameters. This lack of in-formation has made it exceedingly difficult tocompare the work of different groups. This lackof sample characterization may also be responsiblefor the tendency on the part of some workers toignore past work and to report their findings withlittle or no attempt to correlate their results withprevious measurements. Because of the collabo-rative aspect of the JPL program, we have aunique opportunity to correlate lifetime measure-ments in a more significant way. Differences inthe use of the Shockley-Read-Hall theory ariselargely through differences in the assumptions forthe temperature dependence of the cross sections.As can be seen in Fig. 7, the upper energy levelis relatively less sensitive than is the lower levelto these two classes of differences.

As a first stab at the annealing problem, wehave decided to use the Shockley-Read-Hall ap-proach and to consider the recombination centerdensity NR to be time dependent. The determi-nation of NR then is governed by a rate equation,such as has been developed by Fang and others(Ref. 9).

Although we have given some thought to radi-ation annealing in the presence of lithium, we arestill developing techniques for handling the kineticsproblem for non-lithium-diffused irradiated sili-con. We are trying to understand such questionsas the presence of multiple competing defects anddefects with more than one important recombina-tion level. The introduction of lithium donorssuggests a need for such a generalization to multi-level Shockley-Read-Hall models.

Because of the complexity of the annealingprocess itself, it will be necessary to developsome physical insight into the problem in order toknow what type of approximations can be made.The first problematical fact about lithium is itsrather low solubility in silicon: -10 2 0 /m 3 . Yet,it is customary for us to unconcernedly charac-terize our materials with Li concentrations as highas 10 2 3 /m 3 . Electron microscopy studies such asthose shown in Fig. 8 show a precipitation of lith-ium metal in oxygen-lean silicon having nominallya 10 2 3 /m 3 lithium concentration. Substantive evi-dence for such precipitation comes from the identi-fication of body-centered cubic electron diffractionpatterns with the lattice constant of metallic lith-ium. We have not yet understood how this propertycan be incorporated into the kinetic model.

A second property that must be considered isthe fact that the lithium profile in a solar cell


exhibits a linear concentration gradient across thejunction, as well as concentration gradients farinto the base region, as is seen in Fig. 9. Sincethe minority carrier diffusion lengths for irradi-ated material become comparable to a few junctionwidths, this lithium gradient is important in anal-ysis of lifetime data. The presence of lithiumgradients introduces electric fields. Since theminority carrier drift to the collecting electrodesis assisted by these electric fields, the interpre-tation of the minority carrier lifetime measure-ment in solar cell devices becomes more compli-cated. The effective minority carrier diffusionlengths now become an average over the lithiumconcentration profile and thus we cannot now writesimply L = /-DT.

One promising technique for studying theminority carrier diffusion lengths in just such asituation is the spectral response curve. Forincident light with photon energies in excess ofthe indirect gap, absorption will occur. As thephoton energy is increased above the band gap, theabsorption will increase and the effective lengthfor the penetration of the light will decrease, ascan be seen in Fig. 10, showing the frequencydependence of the optical absorption coefficient insilicon. The presence of impurities and radiationdefects has no important effect on such a curve,which is characteristic of the valence and con-duction band states. Therefore, we can use thewavelength of the light to control the dimensionsof the active region of the solar cell, i. e., min-ority carrier generation will proceed only up tothe penetration depth of the light.

In the spectral response curve itself shownin Fig. 11, we measure the short-circuit currentas a function of the wavelength (or photon energy)of the light. In the long wavelength region, wherethe absorption coefficient is small and the lightpenetrates far into the base of the cell, the spectralresponse is dominated by the base region minoritycarrier lifetime or diffusion length. At shorterwavelengths where the optical penetration depth issomewhat larger than the junction depth, the effectof concentration gradients near the junctionbecome important. At yet shorter wavelengthswhere the optical penetration depth is confined tothe diffused layer, the spectral response will besensitive to the minority carrier lifetime in thediffused layer and to the front surface recombi-nation velocity. Our main interest in this work isto study the spectral response curves for opticalpenetration depths traversing the base region ofthe cell. It should be noted that the spectral res-ponse curve rises for photon energies above theindirect band gap even though the optical penetra-tion length decreases; this is because the opticalabsorption is increasing with increasing photonenergy. For sufficiently high photon energies theeffect of the decreased penetration depth begins todominate over the effect of increasing opticalabsorption.

Spectral response studies have been made innon-lithium solar cells covering a variety ofenvironmental parameters, though very little hasso far been done on lithium-diffused solar cells.The spectral response of a solar cell with uniformdoping in the base region and in the front diffusedlayer can be analyzed relatively simply. In addi-tion, exponential carrier concentration gradients

and their resulting electric fields can also beincluded in the analysis.

The spectral response method is complemen-tary to the short-circuit current determination ofthe damage constant where one measures someeffective diffusion length characteristic of theentire base region of the cell. On the other hand,in the analysis of spectral response measurementsit will be necessary to include the effect of: (1)lithium concentration gradients, and (2) the elec-tric field resulting from such gradients. Thisanalysis will involve a fairly complicated com-puter program. We aim to correlate thesespectral response measurements with diffusionlength measurements made on the same samples.

A spectral response apparatus has been builtand tests are being made on a preliminary batchof cells. Here we measure the wavelength depen-dence of the collection efficiency, Q = Isc/qNphwhere Isc is the short-circuit current produced bythe incident photon flux Nph and q is the electroniccharge.

A block diagram of the appartus is shown inFig. 12. Here the monochromator is a double-pass Perkin-Elmer prism monochromator with aresolution of better than 1 cm X 10-8, and thelight source is a flat filament tungsten lamp. Theoutput of the monochromator is chopped andimaged uniformly over the active area of the cellby a toroidal section mirror. In these measure-ments, the cell is loaded to short-circuit con-ditions. In the first mode of operation, the mea-surements will be made first using only theillumination of the monochromator. In the secondmode of operation, spectral response will beprobed by the low intensity light from the mono-chromator while the cell is simultaneously illu-minated by the solar simulator (a powerful tungstensource). The simulated sunlight will allow obser-vation of injection level dependence in the recom-bination mechanisms. In the second mode ofoperation, high power output will be obtained fromthe solar cell and, consequently only a small loadwill be needed to keep the voltage low. The acsignal from the appropriately loaded cell isamplified, synchronously rectified, and recorded.

As a first step in our experimental program,we will develop a reliable technique for spectralresponse measurements using non-irradiated(i e., undamaged) standard n/p solar cells.

The next step will involve the extensions ofthe spectral response measurements to non-irradiated lithium solar cells. In this work, wewill determine values of the same set of charac-teristic parameters for the lithium-doped cells asfor the non-lithium-doped cells.

As a third step, we will irradiate both stand-ard cells and the lithium diffused cells with I-MeVelectrons and measure the spectral response atseveral fluence levels. This procedure is impor-tant for the identification of the particular param-eters which are being degraded by the irradiationand in ultimately determining the damage con-stants for minority carrier lifetime degradation.If the kinetic modeling of the lifetime damagemechanism is to be successful, it is imperativethat the cell degradation be attributed to the


appropriate degraded parameters. If, for example,the electric field changes in a lithium-doped celldue to radiation damage, then there will be con-sequent changes in the short-circuit current whichof course must not be attributed to changes in life-time. Such misinterpretation of the cell degrada-tion could lead to a seriously distorted model ofthe damage mechanism. It is this kind of separa-tion of the radiation damage effects which makesthe spectral response method a valuable tool forstudying the complexities of the annealing kineticsin lithium-diffused solar cells.

REFERENCES

1. Tkachev, V. D., Plotnikov, A. F., andVavilov, V. S., Fiz. Tverd. Tela, Vol. 5,p. 3188, 1963; [Soviet Phvs. -Solid State,Vol. 5, p. 2333, 1964].

2. Cheng, L. J., et al., Phys. Rev., Vol. 152,p. 761, 1966.

3. Carter, J. R., Jr., J. Phys. Chem. Solids,Vol. 31, p. 2405, 1970.

4. Cherlow, J., (private communication).

5. Rosenzweig, W., Bell System Tech. J.,Vol. 42, p. 1573, 1962.

6. Passenheim, B. C., Naber, J. A., andBerger, R. A. , "Production and Annealingof Defects in Lithium-Diffused Silicon AfterIrradiation With 30-MeV Electrons andNeutrons at 300'K, " paper presented at theEighth IEEE Photovoltaic Specialists Con-ference, Seattle, Wash., August 1970.

106

7. Shockley, W., and Read, W. T., Jr., Phys.Rev.,Vol. 87, p. 835, 1952.

8. Lax, M., Phys. Rev.,Vol. 119, p. 1502,1960

9. Fang, P. H., J. Appl. Phys., Vol. 41,p. 3453, 1970.

10. Corelli, J. C., Final Report on JPL Contract952456, Rensselaer Polytechnic Institute,Troy, N. Y., October 1970.

11. Compton, W. D., Semiannual Progress.Report on JPL Contract 952383, Universityof Illinois, July 1970.

12. Glaenzer, R. H., and Wolf, C. J., J. App.Phys., Vol. 36, p. 2197, 1965.

13. Baicker, J. A., Phys. Rev., Vol. 129,p. 1174, 1963.

14. Wertheim, G. K., Phys. Rev., Vol. 10,p. 1086, 1958.

15. Carter, J. R., Downing, R. G., andFlicker, H., TRW final Technical Report4161-6023-R000, May 25, 1966.

16. Sargent, G. A., Final Report on JPLContract 952561, University of Kentucky,July 1970.

17. Iles, P., Proceedings of the Third AnnualConference on Effects of Lithium Doping onSilicon Cells, Technical Memorandum 33-467,pp. 3-6, Jet Propulsion Laboratory,Pasadena, Calif., April 1, 1971.


SIMPLE BAND PICTURE (Si)

t n I\ELECTRONS 066

1.12 eV IA''2 .9

ENERGY I

5 - 0 \) 2 I

(n0

"- 5--6 1 -3

4Lxlo m-0 --- xl 0.642 2 300

Fig. 1. The P/N junction as a photovoltaic device in silicon o-- I.541 1100I

-- 085 Before'- 5-- Anneal

02- Z IO IU 10 I t

0000 o.5°FC -5 o JCONDUCTION BAND / 0.20 085

I 1 /0.18 eV /2 Q

0.4 eV >1 I--_ _ _ _ _ _ _ _ E F H

LIJ I Unirradia______ ? _ 0.22 1 Materia

______ 02 DIVACANCY 2

0.25 eV

f 5///// VALENCE BAND /// //

Fig. 2. Energy levels of thedivacancy defect insilicon ENERGY (eV)

Fig. 3. Relative photoconductivityas a function of photonenergy (Ref. 10).


zO05-r

z

-

PHOTON ENERGY, eV

Fig. 4. Luminescence spectrum of n-type pulled silicon,phosphorous doped, irradiated with 1021electrons/m 2 at 2. 5 MeV. Here the in-direct energy gap is 1. 164 eV (Ref. 11)

o

z

zO

Lo

100

10 t

1015 1ol6 1018

ELECTRON FLUX (m- 2 )

Fig. 5. Minority carrier diffusionlength as a function offluence for 1-MeV electronirradiation on standardN/P solar cells made fromcrucible-grown silicon(Ref. 5)


]0]7

108

//// /////

-1

-2

0-1

Li+?

V+O

V+V

V+PV+V

Li+

Aus1.10 < Eg< 1.165

(Ec - 0.085)

(Ec - 0.21)

(Ec - 0.4)

(Ec - 0.54)

+

+l

V+OV+V

Fe' (Ev + 0.55)

A9i

Avu

-1 V+OLi

+1 V+V

+1 V+O

(Ev + 0.35)

(Ev + 0.18)

E , VALENCEv / /// // / / / / / / / / / / / / / / BAND

Fig. 6. Energy levels of various defect levels in silicon given in unitsof electron volts


CONDUCTIONBAND

109

1 .1 eV

1 0

0.5

- TYPE P - DOPED 1 - M e V ELECTRONS

REF. 13

REF. 10

REF. 11

REF.10

REF.11

REF. 12

Fig . 7. Ene rgy levels of rad ia t ion defects produced by 1-MeV e l ec t rons in n-type si l icon as obtained from ana lys i s of Shockley-Read-Hal l l i fet ime data

*tf tf ?*' tf> & \* \S

Fig. 8. E lec t ron m i c r o s c o p y s tudies of un i r r ad i a t ed f loat-zone si l icon: l i thium a t o m s / m , and showing l i thium prec ip i ta t ion (Ref. 16)

(a) no l i th ium, (b) with 10 23

110 J P L Technica l Memorandum 33-491

4250 C FOR 90 min

102 3 k

E

O

zuzu

Zoc)

1022

1021

LITHIUM PROFILES AND DIFFUSION SCHEDULES(P. ILES, NASA TM 33-467, IAPRIL 1971)

3250 C FOR 480 min

10 20 30 40

DISTANCE FROM PN JUNCTION, 10- 6 m

Fig. 9. Lithium profiles and diffusion schedules (Ref. 17)


104 [

103

0.4 0.6 0.8 1.0

X(10- 6 m)

Fig. 10. Optical absorption coefficient ofsilicon versus wavelength

0.4 0.6 0.8 1.0

WAVELENGTH (10-

6 m)

Fig. 11. A spectral response curve for a silicon solar cell exposed to atungsten light source


10

z

nu

I-

V

U

UWI-0m:Vw

:5W

8

6

2

05

112

Fig. 12. Block diagram for spectral response measurements

JPL Technical Memorandum 33-491NASA - JPL- Col., L.A., Colif.

113

proceedings of the fourth annual conference on effects of

Documents