Abstract Analysis of CdTe solar cells in relation to materials issues

Analysis of CdTe solar cells in relation to materials issuesM.Burgelman *,J.Verschraegen,S.Degrave,P.NolletUniversity of Gent,Electronics and Information Systems (ELIS),Pietersnieuwstraat 41,9000Gent,BelgiumAvailable online 15December 2004AbstractBy now,extensive experimental research is available on thin film solar cells based on CdTe and on CIGS,and their electrical and optical behaviour is characterised by a multitude of diverse characterisation techniques.At the same time,numerical simulation programmes have matured and are available to the research community to assist in interpreting these measurements consistently.Once multiple measurements are (more or less)quantitatively described,the numerical simulation can be used to explore the effect of a variation of materials parameter (e.g.the presence or absence of a property,or variation in a range of values)to the final solar cell characteristics.Examples of such analysis for CdTe solar cells are shown.In CdTe cells,much research has been devoted to the activation treatment of the absorber,and to the technology of the back contact.Analysis of ample measurements has evidenced the crucial role of the profile of the (effective)doping density through the device.It will be illustrated how this relative simple (but hardly mastered)materials property has a far reaching influence to the cell characteristics such as roll-over and cross-over of I –V curves,also in dependence on illumination and voltage,conventional and apparent quantum efficiency,and finally fill factor and efficiency.D 2004Elsevier B.V .All rights reserved.Keywords:Thin film solar cells;CdTe;CIGS1.IntroductionThin film solar cells based on CdTe and on CIGS are steadily approaching towards a mature stage of knowledge,understanding,fabrication and exploitation.By now,exten-sive experimental research on CdTe and CIGS materials and cells is available,and the electrical and the optical behaviour of these cells is characterised by a multitude of diverse characterisation techniques:examples are found in Refs.[1,2]for CdTe,in Refs.[3–5]for Cu(In,Ga)Se 2and in Ref.[6]for wide gap CuInS 2and Cu(In,Ga)(S,Se)2materials.At the same time,numerical simulation programmes have matured and are available to the research community to assist in interpreting these measurements consistently.A review on this was presented in Ref.[7].Modern polycrystalline cells are complicated structures,and the effects of some particular phenomenon,mechanism or materials parameter are often beyond intuition or simplerules of a thumb.Numerical simulation can then be used to provide insight,to interpret measurements and to assess the potential merits of a cell structure.Indeed,once multiple measurements are (more or less)quantitatively described,the numerical simulation can be used to explore the effect of a variation of materials parameter (e.g.the presence or absence of a property,or variation in a range of values)to the final solar cell characteristics.In this article,examples of such analysis are shown for CdTe thin film solar cells.In CdTe cells,much research has been devoted to two features,particular to this type of cell:the first is the treatment of the CdTe absorber in a chlorine containing environment,usually referred to as the CdCl 2treatment or the activation treatment (e.g.Ref.[8]),and the second is the technology of the back contact (e.g.Ref.[9]).Analysis of ample measurements has evidenced the crucial role of the profile of the (effective)doping density through the device.It will be illustrated here that this relative simple (but hardly mastered)materials property exerts a strong influence to the cell characteristics such as roll-over and cross-over of I –V curves,also in dependence on illumination and voltage,0040-6090/$-see front matter D 2004Elsevier B.V .All rights reserved.doi:10.1016/j.tsf.2004.11.011*Corresponding author.Tel.:+3292643381;fax:+3292643594.E-mail address:Marc.Burgleman@elis.ugent.be (M.Burgelman).Thin Solid Films 480–481(2005)392–398/locate/tsfconventional and apparent quantum efficiency,and finally fill factor and efficiency.Also in CIGS cells,materials related issues have been studied with numerical analysis.As this is not the emphasis of this article,we refer the interested reader further to the literature,e.g.:for an overview,Ref.[10],for the influence of grading of the materials composition(band gap grading) [11,12],for the influence of the details of the CIGS electronic surface structure on the I–V curves[13],for the influence of the CIGS defect structure in the bulk and at the surface[14,15].2.ExperimentalOur experimental work on CdTe cells was within the European CADBACK research project.The cells under discussion here were fabricated by Antec(Kelkheim(D)and Arnstadt(D)).The technology and the accomplishments of this company are well documented(e.g.Refs.[16,17]).The cells used in this work had had their activation treatment either in an air ambient or in vacuum,see Ref.[18].3.Some materials related issues in CdTe solar cells3.1.The CdTe contact barrierThere are fundamental materials reasons why finding a stable,ohmic contact to the CdTe layer in a CdTe–CdS thin film solar cell constitutes a scientific and technological challenge.The first is that CdTe is a p type material with a rather large bandgap of E g=1.45eV.According to the ideal (or na R ve)metal-semiconductor theory,the contact barrier U b for holes(measured between the hole Fermi level valence band edge at the contact)is simplyU b¼vþE gÀU mð1ÞEven with a quite common value of4.3to4.5eV for the CdTe electron affinity v,this would require a contact metal with a high work function U m exceeding5.5eV,which cannot be reached if one excludes noble metals for economic reasons.Even if the contact barrier were only partially determined by this ideal model,and partially by Fermi level pinning by contact interface states,the materials properties of CdTe do not favour a large choice of acceptable contact metals.The contact properties can be improved if the CdTe layer is heavily doped.However, obtaining a doping density exceeding1016cmÀ3in CdTe also is a problem,since CdTe is prone to self-compensa-tion.This is related to the rather high value of the band gap,to the introduction of chlorine by the activation treatment,and to many possible interactions and complex formations between impurities and native defects in CdTe [19,20].Thus,making a good,stable contact to CdTe is an art in itself,and it comprises several steps:surface pre-treatment(e.g.etching),control of diffusion of extrinsic dopants from the contact or buffer layer to the CdTe layer, the application of buffer layers to prevent unwanted diffusion or reactions.Copper containing contacts initially perform well(copper forms a shallow acceptor in CdTe), but their long time performance can be a problem[21],as the copper under circumstances diffuses through the CdTe layer and accumulates at the CdS junction and in the CdS layer.Stable contact structures containing no intentional copper were proposed instead[20].In any case,it is clear that the materials and processes used for the contact formation can affect the doping and impurity profile throughout the cell.The presence of a contact barrier affects the I–V characteristics of a CdTe thin film solar cell in a character-istic way:the I–V curves show a kink in the forward biased quadrant,which is commonly called roll-over of the I–V curves[22].The main effects on the cell performance are a decrease in fill factor FF when U b exceeds about0.45eV, and the associated efficiency loss[22,23].However,several secondary effects are reported:(i)A minority carrier (electron)current can be present at the CdTe back contact, giving rise to cross-over of the I–V curves[23];in special cases,all I–V curves of a set of experiments can cross over in one single point[24].These effects are favoured by red light illumination,by thin cells(b5A m)[25]and by a large apparent electron diffusion length,which can be caused by 2-D effects[24].(ii)When the CdTe doping at the back contact is sufficiently high,the hole current through the contact is described by drift and diffusion through the space charge layer(SCL);this effect weakens the aspect of the roll-over,and eventually the room temperature I–V curve looks as if there were only a weak barrier,or even no barrier at all[26].(iii)Also,high doping at the contact makes the apparent barrier dependent on bias voltage and illumination [26].Though these effects render the determination of the barrier from room temperature I–V measurements doubtful, they generally influence the fill factor and the efficiency in a positive way.These effects illustrate the crucial role of the doping profile in the cell.3.2.The CdTe activation treatment(or CdCl2treatment)The treatment of CdTe and other II–VI materials with chlorine containing species has been studied a long time since.In the late1980s,this treatment,with CdCl2as the chlorine source,was applied to CdTe-based solar cells [8,27],and was quickly adopted by the research community as an indispensable and somehow magic recipe of the fabrication of CdTe solar cells[28].The effects of this treatment have been investigated intensely,see,e.g.Ref.[29].We briefly summarise:grain growth occurs when the initial grains are small(41A m)but not when the initial grains are large(N1A m;this is the case for CdTe pro-duced with Close Spaced Sublimation(CSS));the internalM.Burgelman et al./Thin Solid Films480–481(2005)392–398393crystallographic structure of the grains improves in all cases (disappearance of sub-grain structure);the p type doping is established (type conversion)or improved;the minority carrier lifetime s n is improved;the density of deep electronic states in the bulk or at the interface can be re-duced,but other deep states can be introduced [30,31];intermixing between CdS and CdTe at the interface can be promoted.It is clear that these effects more or less thoroughly modify the set of input parameters to be used in numerical simulation of CdCl 2treated solar cells.Especially the doping and recombination properties of the cell are affected by the activation treatment.Hence,a detailed character-isation of the shallow and deep states is needed:doping type,concentration profile,energy levels or distribution,capture cross sections.4.Numerical simulations of CdTe solar cells:results and discussion4.1.The baseline parameter set for numerical simulations To illustrate the influence of the doping parameters (density profile,deep states,...)on the characteristics of a CdTe solar cell,we will start from a set of b baseline parameters Q ,and then vary a few selected parameters,while all other parameters retain their baseline value.Baseline parameters for a b generic Q CdTe cell were proposed in Ref.[32].Here,however,we want to start from a parameter set which consistently describes two specific CdTe cell series made by Antec:one,which had its activation treatment in air,and one with its activation treatment in vacuum.The cell technology and the measurements were presented in Ref.[18];it was concluded that the air-activated cells are more robust to variations in the contact technology than the vacuum-activated cells.These measurements were analysed with the simulation programme SCAPS [33],and a single set of parameters (for each cell)was found [34]which allows to simulate adequately the I –V (light and dark)curves and the I sc –V oc curves,the C –V curves,and the C –f curves,all in a range of temperatures,and the effects of (apparent)quantum efficiency at varying voltage bias.We will present elsewhere in detail how such a consistent parameter set was obtained [35];here we will just take it as a starting point for our simulations.We stress that the only deep states that were taken into account are those determined by us with DLTS [36,37].The main features of these baseline parameter sets are given in Fig.1for the air-activated cells and in Fig.2for the vacuum cells.The most relevant differences between the two are:the contact barrier U b (0.42vs.0.55eV for air/vacuum-activated cells),the CdTe thickness (5vs.8.7A m for air/vacuum-activated cells)and the doping profile (both have a rather high doping at the CdS junction,but air-activated cells have a substantially higher doping in the bulk and at the CdTe contact).The slight differences in the energy levels of the deep states between vacuum and air-activated cells (compare Fig.1to Fig.2)stem from our DLTS measurements [37].Fig. 1.Schematic representation of the model parameters for an air-activated CdTe solar cell.Top:band diagram with deep levels;the energies are in eV above the valence band.Bottom:profile of the shallow doping density.The width of the CdS layer (70nm)has been expanded by a factor of 10for bettervisibility.Fig.2.Schematic representation of the model parameters for a vacuum-activated CdTe solar cell.Top:band diagram with deep levels;the energies are in eV above the valence band.Bottom:profile of the shallow doping density.The width of the CdS layer (70nm)has been expanded by a factor of 10for better visibility.M.Burgelman et al./Thin Solid Films 480–481(2005)392–3983944.2.Results of numerical modellingIn Fig.3,the simulated I –V curves of vacuum and air-activated cells are shown when the contact barrier U b is varied.The effects illustrate what is commonly known.When U b increases,the I –V curves start to roll-over in the forward biased quadrant from about U b z 0.4eV at room temperature,then FF decreases,and finally an S-shaped I –V curve in the active quadrant develops.The associated efficiency loss is shown in Fig.4.The lower efficiency for the vacuum-activated cells is due to the overall lower doping density,as we will show.When the doping density at the CdTe contact is varied,the other parameters having their baseline value for each cell,the I –V curves of Fig.5are obtained.The resemblance with Fig.3,where U b wasvaried,is striking.This illustrates that the effects of contact doping can be mistakenly interpreted as an effect of contact barrier.In particular,a high doping density at the CdTe contact can effectively mask the effects of a contact barrier.This result has been presented before (e.g.Ref.[26]),but it is often overlooked.The influence of the shallow and deep doping profiles is further illustrated in Fig.6where the density of deep acceptors at 0.72eV above the valence band is varied,and in Fig.7where the shallow acceptor concentration in the bulk of the CdTe layer of vacuum-activated cells is varied.An increased defect density causes increased recombination inside and outside the spacechargeFig.3.Variation of the I –V characteristics with the energy barrier U b at the CdTe back contact.The other parameters have their baseline values.Top:a vacuum-activated cell;U b varies from 0.42to 0.60eV (step 0.02eV).Bottom:an air-activated cell;U b varies from 0.40to 0.52eV (step 0.02eV).Fig.4.Efficiency of a vacuum (dashed line)and an air-activated (solid line)cell as a function of the contact energy barrier U b .The baseline values for both areindicated.Fig.5.Variation of the I –V characteristics of a vacuum-activated cell (top)and an air-activated cell (bottom)with the doping density N Acontact at the CdTe back contact.The other parameters have their baseline values (thus U b =0.42eV (air)and 0.55eV (vacuum)).Top:a vacuum-activated cell;the N Acontact values are 1013,3Â1013,1014,...,1017cm À3.Bottom:an air-activated cell;the N Acontact values are 1014,3Â1014,1015,...,1017cm À3.Fig.6.Variation of the I –V characteristics of an air-activated cell with the deep acceptor density at junction (level at E V +0.72eV).The values of N Adeep are:1014,1015,1016and 1017cm À3.The other parameters have their baseline values.M.Burgelman et al./Thin Solid Films 480–481(2005)392–398395layer (SCL),and this deteriorates first V oc and then also J sc ,as can be seen in Fig.6.A low CdTe bulk doping density deteriorates FF ,as is illustrated in Fig.7for vacuum-activated cells.This has been ascribed to a too low electric field in the SCL [38].The influence of the densities of some of the components of the shallow doping and the deep levels is summarised in Fig.8for air-activated cells and in Fig.9for vacuum-activated cells.In both cases,a high shallow doping at the contact N Acontact is favourable.If N Acontact could be increased beyond 1016cm À3,an efficiency gain of 0.5%(absolute)can still be expected.Also the bulk doping should be as high as possible,though the gain is only marginal for N Abulk N 5Â1015cm À3(Fig.9).The deep acceptors at i E V +0.72eV only contribute to the doping at reverse bias voltage,but they enhance the recombination in the SCL in the bias active voltage bias range (0b V b V oc ):therefore,their concentration should be limited to about 1015cm À3(Fig.8).Other components of deep levels were investigated in literature.It was,e.g.pointed out that deep donors close to the back contact cause a loss of V oc due to compensation of the shallow acceptor doping in CdTe close to the back contact [25].This is consistent with our resultsof Fig.5,though a loss of FF seems to a more dominant effect for cells described by our parameter set.The efficiency does not depend monotonously on the doping density N Ajunction near the CdS/CdTe junction,as several effects are simultaneously present.At high N Ajunction ,V oc is high because of the higher built-in voltage,but the long wavelength current response decreases because of the small SCL width;at low N Ajunction ,the current collection is poor because of the electric field in the SCL is too weak;J sc is maximal for intermediate N Ajunction i 2Â1015cm À3,the maximum being broader for air-activated cells;for inter-mediate N Ajunction i 5to 8Â1015cm À3,V oc is decreased because the roll-over of the I –V curves already starts in the active quadrant.This complicated behaviour makes the CdTe doping near the junction the most critical parameter to adjust.Especially for vacuum-activated cells,this parameter should be in a rather narrow range.The simulation results obtained here maybe can explain the sometimes contradictory effects reported by introducing copper as a dopant from the back contact into the CdTe layer.It definitely diffuses easily completely through the CdTe layer,and accumulates at the CdTe/CdS junction and in the CdS layer [21].We showed here however that the resulting cell behaviour is very sensitive to variations in the doping profile near the junction.Hence,technologies which rely on establishing a well-tuned profile of copper related acceptors may bear a risk in terms of producibility and stability.The profile of shallow doping density in our cells is high–low–high,when going from the back contact to the junction.Such a profile has been invoked before to explain a set of measurements on cells of other fabrication [39].The variations in such a high–low–high profile have been invoked to explain the behaviour of CdTe cells upon ageing [40].Other profiles of the shallow doping have also been reported,e.g.a doping density increasing monotonously from the junction to the contact [25].The simulations shown here show that all components of the shallow and the deep doping profile have their influence on the final I –V characteristics of a thin CdTe solarcell.Fig.7.Variation of the I –V characteristics of a vacuum-activated cell with the shallow doping density N A bulk in the CdTe bulk.The N A bulk values are 1014,3Â1014,1015,...,1017cm À3.The other parameters have their baselinevalues.Fig.8.Efficiency of an air-activated cell as a function of the density of the shallow doping at the contact and at the CdS junction,and of the deep acceptor density at junction (level at E V +0.72eV).The other parameters have their baselinevalues.Fig.9.Efficiency of a vacuum-activated cell as a function of the density of the shallow CdTe doping,at the contact,in the bulk,and at the CdS junction.The other parameters have their baseline values.M.Burgelman et al./Thin Solid Films 480–481(2005)392–398396It is however evident that,whilst numerical simulation can lead to the conclusion that the doping profile in CdTe is a critical parameter,and give results about minimal or maximal values or optimal ranges,it cannot give any hints as how to obtain such a desired doping profile.Also the doping(shallow and deep)in the CdS layer plays its role.Its effects on the apparent quantum efficiency under short wavelength illumination(k b500nm or h m N2.5 eV)and under a forward bias voltage,recently have been studied intensely[41–44]and reviewed[45].We do not take up this subject in this paper,but refer to the literature.5.ConclusionsThe position dependence(profile)of the shallow doping in the CdTe layer of CdTe/CdS thin film solar cells seems a trivial materials property,but it is hardly technologically mastered.We showed here that it has a profound effect on all aspects of the I–V curves,and hence on the performance of the solar cell.This holds even more for the positional and energetic profile of the deep impurities in the material.We derived a set of baseline parameters which consistently describe an extended set of measurements on air and vacuum-activated CdTe cells;this is a laborious task in itself.We studied the influence of a selection of parameters related to the shallow and deep doping.Some parameters should be as high as possible:this is the case for the shallow doping in the bulk and at the contact.Others,like the density of deep acceptor states at E V+0.72eV,or the contact barrier U b should be minimised.We derived practical maximal(or minimal)values for these parameters,above (or below)which further improvement is marginal.The shallow doping density close to the CdS/CdTe junction influences multiple aspects of the I–V curves,and it should be within a tolerance window.The cells which had their CdCl2treatment in air generally meet these constraints better than vacuum-activated cells,especially what concerns the shallow doping near the contact.This makes air-activated cells more robust against variations in contact technology,which could produce variations in contact barrier height.AcknowledgementsThis work is supported by the Research Fund of the University of Gent(BOF-GOA).References[1]K.Durose,P.R.Edwards,D.P.Halliday,Journal of Crystal Growth197(1999)733.[2]K.Durose,D.Boyle,A.Abken,C.Ottley,P.Nollet,S.Degrave,M.Burgelman,R.Wendt,J.Beier,D.Bonnet,Physica Status Solidi.A, Applied Research229(2002)1055.[3]O.Lundberg,M.Bodeg3rd,J.Malmstr f m,L.Stolt,Progress inPhotovoltaics11(2003)77.[4]M.Turcu,U.Rau,Thin Solid Films431–432(2003)158.[5]R.Klenk,Thin Solid Films387(2001)135.[6]S.Siebentritt,Thin Solid Films403–404(2002)1.[7]M.Burgelman,J.Verschraegen,S.Degrave,P.Nollet,Progress inPhotovoltaics12(2004)143.[8]B.McCandles,L.Moulton,R.Birkmire,Progress in Photovoltaics5(1997)249.[9]D.B7tzner,A.Romeo,M.Terheggen,M.D f beli,H.Zogg,A.Tiwari,Thin Solid Films451–452(2004)536.[10]R.Klenk,Thin Solid Films387(2001)135.[11]T.Dullweber,G.Hanna,U.Rau,H.-W.Schock,Solar EnergyMaterials and Solar Cells67(2001)145.[12]C.H.Huang,Sheng S.Li,T.J.Anderson,Proceedings of the29nd IEEE Photovoltaic Specialists Conference(New Orleans), 2002,p.748.[13]U.Rau,M.Schmidt,Thin Solid Films387(2001)141.[14]I.Eisgruber,J.Granata,J.Sites,J.Hou,J.Kessler,Solar EnergyMaterials and Solar Cells53(1998)367.[15]A.Niemegeers,M.Burgelman,R.Herberholz,U.Rau,D.Hariskos,H.-W.Schock,Progress in Photovoltaics6(1998)407.[16]D.Bonnet,H.Richter,K.H.J7ger,Proceedings of the13th EuropeanPhotovoltaic Solar Energy Conference(Nice),1995,p.1456. [17]D.Bonnet,Proceedings of the2nd World Conference on PhotovoltaicSolar Energy Conversion(Wien),1998,p.397.[18]P.Nollet,M.Burgelman,S.Degrave,J.Beier,Proceedings of the29th IEEE Photovoltaic Specialists Conference(New Orleans), 2002,p.704.[19]Y.Marfaing,Thin Solid Films387(2001)123.[20]D.Halliday,M.Potter,D.Boyle,K.Durose,Materials ResearchSociety Symposia Proceedings668(2001)H1.8.1.[21]K.Dobson,I.Visoly-Fisher,G.Hodes,D.Cahen,Advanced Materials13(2001)149.[22]G.Stollwerck,J.Sites,Proceedings of the13th European PhotovoltaicSolar Energy Conference(Nice),1995,p.1995.[23]A.Niemegeers,M.Burgelman,Journal of Applied Physics81(1997)2881.[24]M.Burgelman,P.Nollet,S.Degrave,J.Beier,Proceedings of the28thIEEE Photovoltaic Specialists Conference(Anchorage),2000,p.551.[25]T.McMahon, A.Fahrenbruch,Proceedings of the28th IEEEPhotovoltaic Specialists Conference(Anchorage),2000,p.559. [26]J.Beier,M.K f ntges,P.Nollet,S.Degrave,M.Burgelman,MaterialsResearch Society Symposia Proceedings668(2001)H9.5.1. [27]B.McCandles,L.Moulton,R.Birkmire,Progress in Photovoltaics5(1997)249.[28]B.Ba Y ol,International Journal of Solar Energy12(1992)25.[29]K.Krishna,V.Dutta,Journal of Applied Physics96(2004)3962.[30]S.Ringel,A.Smith,M.McDougal,A.Rohatgi,Journal of AppliedPhysics70(1991)881.[31]A.Niemegeers,M.Burgelman,H.Richter,D.Bonnet,Proceedings ofthe14th European Photovoltaic Solar Energy Conference(Barcelona), 1997,p.2079.[32]M.Gloeckler, A.Fahrenbruch,J.Sites,Proceedings of the3rdWorld Conference on Photovoltaic Energy Conversion(Osaka), 2003,p.491.[33]M.Burgelman,P.Nollet,S.Degrave,Thin Solid Films361–362(2000)527.[34]Peter Nollet,PhD thesis,University of Gent/ELIS(2004).[35]P.Nollet,M.Burgelman,to be presented at the19th EuropeanPhotovoltaic Solar Energy Conference(Paris,June2004).[36]J.Versluys,P.Clauws,P.Nollet,S.Degrave,M.Burgelman,ThinSolid Films431–432(2003)148.[37]J.Versluys,P.Clauws,P.Nollet,S.Degrave,M.Burgelman,ThinSolid Films451–452(2004)434.[38]M.Burgelman,P.Nollet,S.Degrave,Applied Physics.A69(1999)149.M.Burgelman et al./Thin Solid Films480–481(2005)392–398397[39]V.Singh,D.Linam,D.Dils,J.McClure,G.Lush,Solar EnergyMaterials and Solar Cells63(2000)445.[40]S.Degrave,P.Nollet,G.Stojanoska,M.Burgelman,K.Durose,Pro-ceedings of the17th European Photovoltaic Conference(M q nchen), 2001,p.1058.[41]G.Agostinelli, D.L.B7tzner,M.Burgelman,Proceedings of the29th IEEE Photovoltaic Specialists Conference(New Orleans),2002, p.744.[42]P.Nollet,M.K f ntges,M.Burgelman,S.Degrave,R.Reineke-Koch,Thin Solid Films431–432(2003)414.[43]G.Agostinelli,D.B7tzner,M.Burgelman,Thin Solid Films431–432(2003)407.[44]G.Agostinelli, E.Dunlop, D.B7tzner, A.Tiwari,P.Nollet,M.Burgelman,M.K f ntges,3rd World Conference of Photovoltaic Energy Conversion(Osaka),2003,p.356.[45]M.Gloeckler,J.Sites,Journal of Applied Physics95(2004)4438.M.Burgelman et al./Thin Solid Films480–481(2005)392–398 398。