a-Si H-based heterojunction solar cells

Downloa Amorphous silicon has been widely investigated as a viable material for making inexpensive and efficient solar cells. In particular the p-i-n structure a-Si:H-based solar cells have attracted a great deal of attention. Over the past few years p-i-n heterojunction structure cells with a wide- band-gap emitter layer have emerged as the structure of great promise in solar cell applications on account of improved conversion efficiency h.1–4 It is imperative to understand precisely what factors are responsible for this improvement in h. This can best be achieved with the help of a detailed first principles computer model capable of simulating carrier transport, recombination, and trapping as a function of posi- tion in the device. It is the aim of the first part of this report to understand the factors leading to this improvement in h. It has been suggested that the p/i interface states in- duced by the heterojunction limit the performance of such cells, and that a significant improvement in h can be achieved by introducing a p/i graded interfacial layer.5–8 The role played by a graded band-gap p/i buffer layer introduced between the emitter and the absorber layers has been inves- tigated in detail in this report. The effect of a graded p/i interfacial layer and its role in improving h has already been studied in some detail by the Tokyo modeling group5,7,8 and to some extent by Hou et al.9 and Suntharalingam et al.10 In the former set of calculations5,7,8 the deep dangling bond ~DB! states have been represented by two delta functions, while Hou et al.9 assumed two rectangular distributions in states. Neither of these represent the true DOS picture in these materials. In Ref. 10 as well as in the present calcula- tions the more realistic Gaussian-type distributions have been utilized to represent the deep DB states. Also in all the reports in the existing literature a neutral ~no band bending! contact to the transparent conducting oxide ~TCO! surface has been assumed whenever wide-band-gap emitter-layer p-i-n solar cells have been considered. This, as the present report goes on to reveal, seems unlikely for the TCO mate- rials developed to date. Finally the mobility band gap Em and the optical band gap Eopt were assumed to be the same even for i-a-Si:H in Refs. 5, 7, and 8, which contradicts recent experimental findings.11 In the present report we have taken proper account of the difference between Em and Eopt for a-Si:H. Other aspects studied in this report are the influence of the p-layer thickness on solar cell terminal characteristics and the problem of apportionment of the p/i band-gap dis- continuity between the conduction and valence bands. The aim of the former study is to predict theoretically the opti- mum thickness for the p layer when it has a wider band gap than the intrinsic absorber. The latter aspect has been ex- plored experimentally by a number of authors;12–14 however, the results of these measurements have not been very con- clusive. The only theoretical investigation in this respect deals with heterojunction cells that do not have a p/i buffer layer.15 Also the drawbacks of that investigation15 are: 7339J. Appl. Phys. 79 (9), 1 May 1996 0021-8979/96/79(9)/7339/9/$10.00 © 1996 American Institute of Physics A computer analysis of the effect of a on the performance of a-Si:H-based he Parsathi Chatterjee Energy Research Unit, Indian Association for the Cultivatio ~Received 14 February 1995; accepted for publication A first principles computer model for simulating the pe cells has been applied to study the effect on solar cell band gap larger than that of the intrinsic absorber. conducting oxide/p-layer contact barrier height fb0 i band gap of the emitter layer Em(p) is to depress the o cell efficiency h, although the short-circuit current Jsc the decrease of Voc , FF, and h at constant fb0 is that a the gradient in the electron affinity at the p/i interface collapse of the field over the intrinsic absorber layer. C in these structures is practically independent of this p influence on Voc , FF, and h. In fact, the observed impro cell performance when a wider-band-gap p-a-SiC:H l assuming that fb0 in this case is larger than that for a cell. A study of the p/i interface states induced by the Jsc , FF, and the efficiency of such cells. A reduction in a graded band-gap buffer layer at this junction has bee by more than 25%. We further find that cells with conversion efficiency when this layer has a wider ban show that when a buffer layer is introduced at the p/ p-layer/i-layer band-gap discontinuity does not have performance. © 1996 American Institute of Physics. I. INTRODUCTION ded¬12¬Feb¬2011¬to¬221.192.238.11.¬Redistribution¬subject¬to¬AIP wide-band-gap emitter layer terojunction solar cells n of Science, Calcutta 700 032, India 8 January 1996! rformance of amorphous-silicon-based solar performance of using an emitter layer with We surprisingly find that if the transparent s held constant, the effect of increasing the pen-circuit voltage Voc , fill factor ~FF!, and increases as expected. The main reason for s Em(p) expands, the field corresponding to increases also, leading more and more to a onsidering the effect of fb0, we find that Jsc arameter. However, fb0 exerts considerable vement in a-Si:H-based single junction solar ayer is introduced can only be explained by p-i-n homojunction structure a-Si:H solar heterojunction reveals that these mainly limit the number of these interface states by using n found to enhance the conversion efficiency p-layer thickness ;80 Å have the highest d gap than the intrinsic absorber. Finally, we i heterojunction, the energy location of the a crucial impact on heterojunction device @S0021-8979~96!06008-5# the density-of-states ~DOS! model to simulate these DB ¬license¬or¬copyright;¬see¬http://jap.aip.org/about/rights_and_permissions Downloa ~i! the assumption of a constant distribution of DB states instead of Gaussian distributions; ~ii! the assumption of neutral contacts to the TCO sur- face; and ~iii! taking Em equal to Eopt even for i-a-Si:H. Nowadays it is common practice to use a p/i buffer layer to reduce the detrimental effect of band-gap mismatch; hence, the present investigation explores the effect of different ap- portionments of the p/i band-gap discontinuity on the per- formance of heterojunction a-Si:H-based solar cells that em- ploy a graded band-gap p/i buffer layer. This article is organized as follows: Section II describes the simulation model and the mathematical implications of a wide-band-gap p layer. Section III and its subsections ex- plore the impact of using a wide-band-gap p layer on solar cell performance. The concluding Sec. IV summarizes the salient features of the report. II. SIMULATION MODEL Our first principles computer model16,17 solves the Pois- son’s equation and the two carrier continuity equations under nonequilibrium steady-state conditions. Very general bound- ary conditions are used. They require the front fb0 and back fbL contact barrier heights and the recombination speeds of these contacts to be specified. In the present calculations the latter are taken equal to 107 cm s21 for both holes and elec- trons. The band diagram under short-circuit conditions of a typical wide-band-gap p-layer solar cell, with the band-gap discontinuity equally divided on the conduction- and the valence-band sides, is shown in Fig. 1. For the heterojunction structures, the usual current den- sity equations ~e.g., Ref. 16! are expressed in the more gen- eral form, FIG. 1. Energy band diagram of a p/i heterojunction solar cell with a graded p/i buffer layer. The front fb0 and back fbL contact barrier heights as well as the p and n regions are indicated. Ec , Ev , EFn, and EFp repre- sent, respectively, the conduction-band edge, valence-band edge, and the quasi-Fermi levels for electrons and holes. 7340 J. Appl. Phys., Vol. 79, No. 9, 1 May 1996 ded¬12¬Feb¬2011¬to¬221.192.238.11.¬Redistribution¬subject¬to¬AI Jn5qnmn“EFn, Jp5qpmp“EFp. ~1! Here Jn (Jp) are the electron ~hole! current density, n (p) the free electron ~hole! density, mn (mp) the band micro- scopic electron ~hole! mobility, EFn (EFp) the quasi-Fermi level for electrons ~holes!, and q the electronic charge. The equation for Jn can be expressed as Jn5qnmn“~EV.L.2k!1qDn“n2qnDn“~ ln Nc!, ~2! where EV.L. is the vacuum level, k is the electron affinity, Dn the electron diffusion constant, and Nc the effective DOS in the conduction band. Here term 1 is the drift due to the ‘‘effective field’’ on the electrons, which in a heterojunction includes, besides the usual electrostatic field, also a contri- bution due to the gradient in the electron affinity. In other words term 1 includes the gradient of the conduction-band edge Ec . Term 2 is the usual diffusion term due to the carrier concentration gradient, while the last term determines the diffusion due to the gradient ~if any! in the number of avail- able states per cm3 (Nc) in the conduction band. Likewise, the effective field causing the drift of holes in a heterojunc- tion structure can be shown to be equal to the gradient of the valence-band edge Ev . The gap state model16,17 used in these calculations con- sists of both the donorlike and acceptorlike tail states, as well as Gaussian distribution functions to simulate the deep dan- gling bond states. The mobility gaps of the i and n layers have been assumed to be 0.2 eV larger than the respective optical gaps in the light of recent experimental evidence.11 III. RESULTS AND DISCUSSION We now investigate the effect of a wide-band-gap emit- ter layer on the performance of a-Si:H-based heterojunction solar cells. The material and device parameters of the ‘‘stan- dard’’ heterojunction cell ~named PBFSD! are given in Table I. This cell is called the standard cell since in the following sensitivity studies, when the p-layer energy gap or thickness, or the front contact barrier height is changed, all the other parameters are held at their values given in Table I. Now in Refs. 5, 7–10, a neutral ~no band bending! TCO/p-a-SiC:H barrier height (fb05fb0n ) has been assumed. Calculations for this cell reveal that fb0n is ;1.6 eV. It has been proved both theoretically16,18,19 and experimentally20 that fb0 for TCO/p-a-Si:H is rather low ~fb0,1.15 eV! and that such contacts show considerable Schottky barriers. The mobility gap of p-a-Si:H is 1.9 eV,11 while that to be assumed for p-a-SiC:H in the following calculations is 2.02 eV. Since fb0 depends on the difference in the electron affinities of the TCO and the p-doped semiconductor, it is unlikely that the widening of the band gap of the latter by only 0.12 eV would increase fb0 from 1.15 to 1.6 eV. We have therefore assumed a lower fb0. Also the p-layer thickness in this study has been taken equal to 80 Å. We intentionally did not create a p layer that would electrostatically shield the i layer from the front contact, since this study concerns solar cells where the typi- cal p layer thickness is ;100 Å. The particular value of 80 Å for the p-layer thickness has been assumed since, as is shown subsequently in Sec. III D, the optimum p-layer thickness for a wide Em(p) solar cell is around 80 Å. In all Parsathi Chatterjee P¬license¬or¬copyright;¬see¬http://jap.aip.org/about/rights_and_permissions the maximum variation of the potential energy of an electron emitter-layer cells without a p/i buffer layer. The reason for TABLE I. Principal input parameters for the standard p-i-n heterojunction structure solar cell containing a graded band-gap p/i buffer layer. General parameters Device length ~ Front contact b Electron mobili Parameters for the n Thickness ~Å! 250 Mobility band g 1.9 Optical band ga 1.7 Activation ener 0.25 Effective DOS 231020 Effective DOS 231020 Donor tail char 50 Acceptor tail characteristic energy ~meV! 60 55–35 27 30 23 18 17 15 18 Downloa across the device in thermodynamic equilibrium!. Also shown for comparison are the values of these quantities when fb051.15 eV ~cell named PB115!, the fb0 value found to give good agreement16 with experiments in standard a-Si:H p-i-n homojunction structure cells; and when the TCO/p contact is neutral ~fb05fb0n 51.62 eV—cell named PB162!. Also given are the performance parameters for the cell PINSD, which is a homojunction cell with output char- acteristics close to the cell fabricated by Hamakawa21 and simulated in Ref. 16. this assumption is as follows: In order to explain the rather low blue response observed in such cells, the net trapped positive charge density pT must not be high in the very de- graded p/i interface region. Otherwise the electric field in the p/i region would be high, thereby reducing interface re- combination and bringing up the blue response. Two possible ways of reducing pT , without bringing down the number of p/i interface states ~which are known to be large in the ab- sence of any p/i buffer layer due to the band-gap mismatch! are TABLE II. fb0, Vbi , Jsc , Voc , FF, h, and the TCO/p front contact loss R front under the maximum power output condition for a-Si:H-based solar cells of different structures simulated with the model of Ref. 16. Solar cell structure Structure name fb0 ~eV! Vbi ~V! J sc ~mA cm22! Voc ~V! FF h ~%! R front ~%! p-i-n homojunctiona PINSD 1.15 1.10 10.93 0.786 0.667 5.72 21.3 p-i-n heterojunction with buffer PB115 1.15 1.14 14.05 0.716 0.687 6.91 10.4 Same structure as above PBFSD 1.29 1.22 14.10 0.855 0.726 8.76 8.9 Same structure as above PB162 1.62 1.35 14.63 1.007 0.761 11.90 4.3 p-i-n heterojunction without buffer PH129 1.29 1.09 12.39 0.824 0.612 6.25 6.4 aFrom Ref. 16. 7341J. Appl. Phys., Vol. 79, No. 9, 1 May 1996 Parsathi Chatterjee the calculations in this section, 100 mW of AM1.5 light is assumed to be incident on the device through the TCO/p contact. Reflection from the back contact has been ignored, since the aim of this report is to understand the impact of increasing the emitter layer band gap, rather than design any state-of-the-art cell structure. The output parameters of the standard cell, PBFSD, are given in Table II, together with fb0, the built-in potential Vbi , and the TCO/p front surface loss R front under the maxi- mum power output condition ~Vbi for this study is defined as Donor midgap density ~Gaussian—cm ! 2310 Acceptor midgap density ~Gaussian—cm23! 231018 Donor Gaussian peak position ~eV!a 0.96 Acceptor Gaussian peak position ~eV!b 0.56 Standard deviation ~Gaussians—eV! 0.15 Charged capture cross sections ~Gaussians—cm2! 10214 Neutral capture cross sections ~Gaussians—cm2! 10215 Charged capture cross sections ~tails—cm2! 10215 Neutral capture cross sections ~tails—cm2! 10217 aMeasured from the valence-band edge. bMeasured from the conduction-band edge. ded¬12¬Feb¬2011¬to¬221.192.238.11.¬Redistribution¬subject¬to¬AIP In the discussion to follow we also consider cases when the p/i buffer layer is absent. The cell under this condition ~named PH129! has identical parameters in the p , i , and n regions as in Table I, except that the i-layer thickness here is 5000 Å. However, in the first 25 Å of the i-layer adjacent to the p region, the density of DB states has been taken equal to 331019 cm23. Also the charged capture cross sections for the donorlike DB states snd here had to be taken one order of magnitude higher than the rest of the device, to match the low blue response generally observed in wide-band-gap 2310 2310 4.5310 231017 231015 4.531018 0.95–0.91 0.9 0.9 0.55–0.51 0.5 0.5 0.15 0.15 0.15 10214 10214 10214 10215 10215 10215 10215 10215 10215 10217 10217 10217 : mm! 0.533 Temperature ~K! 300 arrier height ~eV! 1.29 Back contact barrier height ~eV! 0.25 ty ~cm2 V21 s21! 25 Hole mobility ~cm2 V21 s21! 6 different layers p p/i i 80 150 4850 ap ~eV! 2.02 2.0–1.92 1.9 p ~eV! 2.02 2.0–1.92 1.7 gy ~eV! 0.4 ••• 0.9 in the valence band ~cm23! 231020 231020 231020 in the conduction band ~cm23! 231020 231020 231020 acteristic energy ~meV! 90 80–55 46 ¬license¬or¬copyright;¬see¬http://jap.aip.org/about/rights_and_permissions 8 Downloa ~a! to assume a high charged capture cross section for the donorlike DB states snd in the p/i region, so that these states mainly provide channels for recombination, or ~b! to assume9 that a certain portion of the boron atoms from the p layer diffuses into the p/i interface region, to produce active boron doping. The latter effect reduces the net effective positive charge density in this region. However, even in cells deposited in a separate chamber mercury sensitized photo-chemical-vapor- deposition ~photo-CVD! system7 the blue response has been shown to be low in the absence of a p/i buffer layer. As shown by Kim et al.7 in their analysis, via secondary-ion- mass spectroscopy, of boron atom profiles in cells fabricated by photo-CVD, the slope of the boron profile at the p/i in- terface is very sharp. Hence, the low blue response of cells without a p/i buffer layer cannot be explained by assuming an active boron doping density of 331018 cm23 in the p/i interface region as done by Hou et al.9 In view of this we have assumed a higher snd in this region as compared to the rest of the device. The solar cell output parameters and R front of this cell are also given in Table II. The front contact loss for the p-i-n homojunction device is seen from Table II to be much larger than those for the wide-band-gap emitter layer heterojunction structures. In such a homojunction structure a large fraction of the shorter wavelengths is absorbed in the p layer. Also a considerable number of electrons back-diffusing from the i layer are able to enter this region. It has been shown by computer analysis16,19 that for such structures when the front contact barrier height is less than the corresponding neutral contact value, the band bending in the part of the p layer adjacent to the contact acts as a sink that attracts photogenerated elec- trons to the front contact. Widening of the band gap of the p layer decreases absorption of incident photons in this region. At the same time the probability of back-diffusing electrons produced in the intrinsic absorber entering the p layer dimin- ishes both on account of the larger repelling field on the electrons at the p/i contact and absorption in the p/i inter- face layer; thus, R front is lower. In fact, our observation is that while the low short-circuit current Jsc and low blue re- sponse in the homojunction p-i-n structure cell ~PINSD! are mainly due to the large R front , it is the recombination through the p/i interface states that limits these quantities in an abrupt p/i heterojunction structure cell ~PH129!, provided the i layer quality is good. When a p/i buffer layer is present, recombination at this interface decreases. A larger fraction of the back-diffusing electrons are therefore able to reach the TCO/p contact where they recombine ~PB115 and PBFSD!. R front decreases with increasing fb0, until for fb05fb0 n ~cell PB162!, R front is the lowest. A. Sensitivity to the p-layer mobility gap We have studied the effect on solar cell performance of increasing the band gap of the emitter layer with fb0 held constant at 1.29 eV. The range of variation of the p-layer mobility gap is from 2.0 to 2.2 eV. The activation energy of the p layer has been held constant as the mobility gap is increased. Although for such a thin and low defect density p layer, decoupling of the front contact barrier height fb0 and 7342 J. Appl. Phys., Vol. 79, No. 9, 1 May 1996 ded¬12¬Feb¬2011¬to¬221.192.238.11.¬Redistribution¬subject¬to¬AIP the p-layer mobility gap Em(p) is not possible in practice, this was done in order to obtain an insight into the actual role played by each parameter in influencing the solar cell termi- nal characteristics. The band-gap discontinuity has been as- sumed to be equally distributed between the valence