Efficient CuInS2 solar cells from a rapid thermal process (RTP)

*Corresponding author.Tel.:#49-30-8062-2584,fax:#49-30-8062-2931.E-mail address:siemer @hmi.de (K.Siemer).Solar Energy Materials &Solar Cells 67(2001)159}166E $cient CuInS solar cells from a rapid thermal process (RTP)Kai Siemer *,Jo Klaer,Ilka Luck,Ju rgen Bruns,Reiner Klenk,Dieter Bra unigHahn-Meitner-Institut Berlin,Dept.SE3,Glienicker Strasse 100,D-141009Berlin,GermanyAbstractCuInS -based solar cells are prepared by a rapid thermal process (RTP).We use a sequential preparation with metallic layers of Cu and In rapidly heated in elemental sulfur vapor.Absorber layers from this process show good crystallinity,as seen from XRD and SEM.For further analysis of the defect chemistry,photoluminescence and admittance spectroscopy measurements are carried out.Solar cells prepared from these RTP absorbers have reached 11.4%total area e $ciency (A "0.5cm ),which is to our knowledge the best CuInS -based solar cell so far. 2001Elsevier Science B.V.All rights reserved.Keywords:CuInS;Solar cells;Rapid thermal process 1.IntroductionCuInS is a promising absorber material for the development of thin "lm solar cells.We have recently presented a sequential process with conventional heating resulting in 11.1%total area e $ciency for a glass/Mo/CuInS /CdS/ZnO cell structure [1].For potential commercial applications short processing times and low thermal budgets are of great importance in order to meet high throughput and low cost.Additionally,it has been reported for selenide chalcopyrites that the use of rapid thermal processes improves crystal growth as detrimental phases can be avoided by passing intermedi-ate temperature ranges rapidly [2].In this work we use a rapid thermal process for the preparation of CuInS absorbers.Both absorbers and corresponding solar cells are characterized with respect to their structural and electronic properties.0927-0248/01/$-see front matter 2001Elsevier Science B.V.All rights reserved.PII:S 0927-0248(00)00276-2Fig.1.Temperature course of a typical sulfurization by RTP.Temperatures given are nominal,sample temperature is up to 803C higher.2.Experimental2.1.Sample preparationCu and In are deposited subsequently as a stacked elemental layer by DC magnet-ron sputtering onto Mo-coated #oat-glass substrates.All samples are prepared under Cu excess with atomic ratio [Cu]/[In]"1.8.The layer thickness is about d ! "550nm and d ' "650nm,respectively.For the formation of CuInS ,a rapid thermal process is performed in a customized halogen lamp furnace with a quartz tube as reaction chamber.The samples are rapidly heated by radiation from the front to about 6003C and kept at this stage for a few minutes.Sulfur is added in elemental form prior to processing.Both the samples and the sulfur are placed onto a quartz sample holder.The furnace is evacuated but no pumping is done during the heating cycle.The pressure in the chamber is thus only determined by the vapor pressure of sulfur.The temperature and the time of sulfurization as well as the amount of sulfur added to the reaction are varied.The process temperature is determined using a chromel/alumel thermocouple which is inserted into a small quartz tube close to the center of the furnace.For comparison,the temperature was measured with a Pt100RTD directly attached to the sample.It shows that the sample temperature is about 803C higher than the temperature measured with the thermocouple.Fig.1shows the temperature course of the sulfurization for a typical process.To reduce the amount of sulfur to be added to the reaction,the size of the reaction volume was decreased by the use of a petri dish covering samples and sulfur for a number of experiments.This reduces the reaction volume from about 10 cm to about 170cm .The amount of sulfur per volume is kept comparable.The large reaction volume however is partly bordered by cold #anges.This might reduce the vapor pressure of sulfur during the process due to condensation.In contrast,in the small reaction volume,the vapor pressure is probably constant or even increasing.It can,however,not be assumed that this small chamber is sealed with respect to the rest of the reaction tube.160K.Siemer et al./Solar Energy Materials &Solar Cells 67(2001)159}166K.Siemer et al./Solar Energy Materials&Solar Cells67(2001)159}166161 For completion of the solar cell,segregation of CuS due to Cu excess has to be removed by a cyanide etch.Subsequently,a CdS bu!er and a two-layer ZnO window are deposited.The cell is contacted with a Ni}Al grid.Best cells are further improved by a MgF anti-re#ection coating.2.2.CharacterizationThe crystallinity and morphology of absorber layers are studied by X-ray di!rac-tion(XRD)and scanning electron microscopy(SEM),respectively.The defect chemistry is examined by means of photoluminescence(PL).These measurements are carried out on absorbers at10K with a diode laser( "668nm)as excitation source giving a power density of about100W/cm .Additionally,admit-tance spectroscopy measurements are carried out on completed cells to get informa-tion about the distribution of deep defects in the band gap of the absorber.These measurements are performed in the temperature range between40and360K at frequencies of100Hz to1MHz.Performance of the solar cells under AM1.5global illumination with100mW/cm as well as quantum e$ciency are measured.3.Results3.1.Structural characterizationFig.2shows a SEM picture of a CuInS absorber layer from RTP after cyanide etch taken under453angle.The grain size is about1}2 m which is comparable to layers from a conventional heating process[3].The cross section shows a closed layer above the back contact though the layer is very rough.Thus,the existence of pin holes in the absorber layer cannot be excluded from these observations.The adhesion of the absorber layer to the molybdenum back contact seems to be poor.However,this might be a consequence of the sample preparation for SEM including breakage of the substrate.Fig.3shows the X-ray di!ractogram of a layer comparable to the one seen in Fig.2. All peaks can be easily identi"ed with CuInS roquesite phase.For large angles,the splitting of Cu K lines can be observed which is a consequence of the small linewidth of the peaks and thus an indication for good crystal quality.Tetragonal splitting can be seen for(004)/(200),(204)/(220),and(116)/(312)peaks which indicates a well ordered sublattice of Cu and In atoms of the roquesite phase.The relative intensity of the observed peaks is in good accordance with JCPDS standard spectra.Therefore,we can conclude that these layers have no preferred orientation.3.2.Defect distributionTo study the defect structure of the absorber layers,samples are examined by means of photoluminescence.Fig.4shows PL of di!erent CuInS layers after cyanide etch.Fig.2.SEM picture of a RTP-CuInSabsorber layer after cyanide etch (453tilted).Fig.3.X-ray di !ractogram of a RTP-CuInS absorber layer after cyanide etch.Numbers indicate roquesite re #exes except where indicated.The spectra usually show an excitonic emission at 1.53eV and a broad band emission at 1.2eV which has not yet been clearly identi "ed.It possibly results from structural defects.Additionally,in some spectra shoulders at 1.50and 1.45eV can be seen.The latter one can be ascribed to a donor acceptor recombination <1}<! [4].The relative intensity of the peak at 1.2eV with respect to the excitonic emission correlates with process parameters as shown in Fig.4.With increasing process162K.Siemer et al./Solar Energy Materials &Solar Cells 67(2001)159}166Fig.4.PL-spectra of RTP-CuInSabsorber layers after cyanide etch prepared at di !erent temperatures.temperature the relative intensity of the broad band emission decreases which can be explained by an increasing crystal quality.The intensity of the excitonic emission increases and/or (structural)defects decrease in density with increasing process tem-perature.This is in good accordance with observations from XRD where the linewidth of roquesite peaks decreases with increasing process temperature.The same correla-tion of PL with process parameters has been observed for increasing slope of the heating-up ramp and }to a minor extent }for increasing processing time (not shown here).For further examination of electronic defects,admittance spectrocopy measure-ments were carried out for solar cells.A single defect level at 140meV is found which we attribute to interface defects [5](not shown here).So far,we have not observed any signi "cant correlation of defect distribution deduced from admittance spectra with parameters of the RTP process nor with any electrical parameters of RTP-CuInS-based solar cells.3.3.Solar cell performanceFig.5a shows the quantum e $ciency of two CuInS solar cells prepared in di !erent reaction volumes.To obtain the smaller volume both the samples and the sulfur was covered by a petri dish (see above).At about 450nm the absorption in CdS can be seen.Between 500and 800nm the quantum e $ciency is at a constant level of about 80%showing a constant collection e $ciency independent of penetration depth.Quantum e $ciency rapidly decreases at the band edge of the absorber.Therefore,it is assumed that the di !usion length is no limitation for the collection e $ciency.The spectral dependence is almost the same for both cells;however,the overall level is signi "cantly higher for the sample from the small reaction volume resulting in higher e $ciency.Fig.5b shows the current-voltage measurement of the same solar cells as shown in Fig.5a.As expected from the quantum e $ciency results,the short circuitK.Siemer et al./Solar Energy Materials &Solar Cells 67(2001)159}166163Fig.5.(a)Quantum e $ciency of two RTP-CuInS solar cells prepared in di !erent reaction volume sizes.(b)j }<measurement of the same cells under AM 1.5globalillumination.Fig.6.j }<-measurement of the best RTP-CuInS solar cell as independently con "rmed by Fhg-ISE,Freiburg,Germany.Electrical parameters are given as inset.current density is signi "cantly higher for samples prepared in the small reaction volume.Furthermore,the "ll factor has increased from about 65%to 70%.This results in an increased e $ciency of more than 11%compared to about 10%for the sample from the larger reaction volume.From this result,we can conclude that the sulfur vapor pressure during the preparation plays an important role on the device performance.The large reaction volume of 10 cm is partly bordered by cold #anges.Sulfur probably condenses at these #anges during the process.When the small reaction volume is used,in contrast a constant or even increasing vapor pressure of sulfur can be assumed to be present during the process.Fig.6shows the result of our best cell as independently con "rmed by Fhg-ISE,Freiburg.This cell was prepared at 5003C nominal for 3min,heating-up ramp was164K.Siemer et al./Solar Energy Materials &Solar Cells 67(2001)159}166Fig.7.Open-circuit voltages and "ll factors as a function of process parameters.Best cell of each sample is shown.(a)Di !erent amount of sulfur added to the reaction (¹+5803C,t "3min.,ramp 10K/s,reaction volume 170cm ).(b)Di !erent process temperature.Temperatures given are nominal,at 5003C about 80K must be added for sample temperature (reaction volume 10 cm ).10K/s.The small reaction volume of about 170cm was used with 50mg of sulfur added.This cell has an open-circuit voltage of < "729.4mV,FF "71.7%"ll factor,and short-circuit current density of j "21.83mA/cm resulting in "11.4$0.4%total area e $ciency for A "0.511cm .3.4.In y uence of process parametersTo examine the in #uence of the amount of sulfur added to the reaction we have taken j }<-curves under AM 1.5global illumination at 100mW/cm .Fig.7a shows the open-circuit voltage and the "ll factor as a function of the amount of sulfur.The reaction volume for this experiment was about 170cm .Sample temperature was about 5803C,kept for 3min.The heating-up ramp was run with a constant slope of 10K/s.The values for the best cell on each sample are shown.Obviously,there is a minimum amount of sulfur necessary to get reasonable results.For the lowest amount of sulfur,no diode behavior is obtained.With a minimum of 50mg in this case,the open-circuit voltage exceeds 700mV and the "ll factor is in the 70%range.With increasing amount of sulfur the open-circuit voltage slightly improves while the "ll factor decreases.For 1000mg sulfur both values decrease signi "cantly.For 100to 500mg sulfur cells have 11%e $ciency or above.The best value for < )FF and the highest e $ciency is obtained with 100mg sulfur.The characterization of these samples in terms of structural properties is on the way.Fig.7b shows a typical example of the impact of process temperature on solar cell performance.The large reaction volume of about 10 cm was used with 3000mg of sulfur added,which proved to be the best condition for the large volume.The slope of the ramp was again 10K/s.The top temperature was varied to keep the product K.Siemer et al./Solar Energy Materials &Solar Cells 67(2001)159}166165166K.Siemer et al./Solar Energy Materials&Solar Cells67(2001)159}166¹(3C)t(s)constant to partly discriminate between the in#uence of thermic power and the in#uence of temperature.At3003C all cells were ohmic.Between4003C and6003C fair results were obtained with a maximum at5003C nominal.Additionally,homogen-eity is by far the best at5003C with all cells on the sample in the10%e$ciency range.4.ConclusionWe have presented a RTP process for the preparation of CuInS absorber layers. These layers are of pure roquesite phase with no preferred orientation observed from XRD.SEM pictures show rough but closed layers with grain sizes of1}2 m. Photoluminescence spectra show excitonic as well as deep broad band emission.The relative intensity of these two emissions correlates with process parameters such as temperature and ramp slope.Admittance spectra show a maximum in defect distribu-tion about140meV which we attribute to interface states.The vapor pressure of sulfur and its constancy during the process has an important in#uence on device quality as the e$ciency of solar cells was signi"cantly improved by the use of a smaller reaction volume.Solar cells made from RTP-CuInS layers have reached11.4%e$ciency which is to our knowledge the best value for CuInS based cells so far.AcknowledgementsFor sample preparation,the help of E.Mu ller,C.Kelch,T.Mu nchenberg,M. Kirsch,and P.Szimkowiak is gratefully acknowledged.The authors wish to thank K. Diesner for XRD measurements.Part of this work has been funded by the Joule III Programme of the European Commission(Contract JOR3-CT98-0297).References[1]J.Klaer,J.Bruns,R.Henninger,K.Siemer,R.Klenk,K.Ellmer,D.Bra unig,Semicond.Sci.Technol.13(1998)1456}1458.[2]V.Probst,F.Karg,J.Rimmasch,W.Riedl,W.Stetter,H.Harms,O.Eibl,Materials Research SocieySymposium Proceedings,Pittsburgh,PA,Vol.426,1996,pp.165}176.[3]J.Klaer,J.Bruns,R.Henninger,K.To pper,R.Klenk,K.Ellmer,D.Bra unig,A tolerant two-stepprocess for e$cient CuInSsolar cells,In:J.Schmid et al.(Eds.),Second World Conference Photovol-taic Solar Energy Conversion,Vienna1998,537}540.[4]K.To pper,J.Bruns,R.Scheer,M.Weber,A.Weidinger,D.Bra unig,Photoluminescence of CuInSthin"lms and solar cells modi"ed by postdeposition treatments,Appl.Phys.Lett.71(4)(1997)482}484.[5]J.Kneisel,Admittanzspektroskopie an CuInS-Du nnschichtsolarzellen,Diploma Thesis,Technical University Berlin,Germany,1999。