Solar Energy Materials & Solar Cells 61 (2000) 203}211 Letters Metal/CdTe/CdS/Cd Zn S/TCO/glass: A new 1~x x CdTe thin "lm solar cell structure Isaiah O. Oladeji!, Lee Chow!,*, Christos S. Ferekides", Vijay Viswanathan", Zhiyong Zhao" !Department of Physics, University of Central Florida, Orlando, FL 32816, USA "Department of Electrical Engineering, University of South Florida, Tampa, FL 33620, USA Received 12 August 1999; accepted 13 August 1999 Abstract The potential of CdTe/CdS/Cd Zn S structure as an alternative to CdTe/CdS structure in 1~x x photovoltaic application has been demonstrated. The unoptimized solar cell structure grown on transparent conducting oxide coated soda lime glass of 3 mm thickness with no antire#ection coating yielded a 10% e$ciency. This e$ciency is the highest ever recorded in any Cd Zn S "lm containing CdTe solar cells. ( 2000 Elsevier Science B.V. All rights reserved. 1~x x Keywords: CdTe; Solar cell; CdZnS 1. Introduction The average solar radiation [1] reaching the earth surface, AM1.5, is very rich in photons with energies ranging from about 1.4 eV}3 eV. In traditional CdTe solar cells CdS is popularly used as a window material and as an n-type heterojunction partner to p-CdTe. Since the bandgaps of CdTe and CdS are 1.5 and 2.4 eV, respectively, it means that only photons with energies within this range reach CdTe layer where they contribute to the cell's photocurrent. The absorption coe$cients of II}VI compounds in the spectrum window are high [2], about 104}105 cm~1. Thus, the carrier generation by higher energy photons in the n-CdS layer takes place at the surface far away * Corresponding author. Tel.: 001-4078-232-325; fax: 001-4078-235-112. E-mail address: le@physics.ucf.edu (L. Chow) 0927-0248/00/$ - see front matter ( 2000 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 7 - 0 2 4 8 ( 9 9 ) 0 0 1 1 4 - 2 204 I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 Fig. 1. Band edge energies of ZnS, CdS, and CdTe compounds [6]. Where E is the band gap, E conduc' # tion band edge, and E valence band edge. 7 from the depletion region where the generated carriers can be collected. As a result, these carriers are often lost to surface recombination current. This current increases the dark current and consequently reduces the useful current delivered by the cell to a load. In these CdS/CdTe cells, reducing the thickness of CdS layer [3] to less than 1000 As is routinely done to reduce the absorption and minimize the surface recombination current. At this thickness they appear to be low yield due to increase in the shunting probability. Below 700 As , there is a general degradation [3] in the cell performance owing to considerable decrease in shunt resistance. It should have, therefore, been more appropriate to use a thick layer of ZnS "lm with 3.7 eV bandgap as a window, but there are several drawbacks associated with this idea. First, ZnS is highly resistive and hard to dope [4] and could signi"cantly increase the cell's series resistance. Second, the lattice mismatch between ZnS and CdTe is about 16%, making it a poorer heterojunction partner to CdTe compare to CdS. Depositing CdTe directly on a thick Cd Zn S "lm will not help either, because detrimental property, 1~x x especially lattice constant [5], of this "lm lies between those of CdS and ZnS. The arguments above then make Cd Zn S/CdS structure a suitable window structure 1~x x that can maintain the traditional CdS/CdTe contact and improve the short wavelength spectral response of the cell at the same time, without compromising the transport properties, the series and the shunt resistances of the cell. Let us look at the feasibility of CdTe/CdS/Cd Zn S structure for solar cell 1~x x application from the band structure standpoint. According to Lehmann [6] the relative position of conduction and valence band edges of ZnS, CdS, and CdTe are as shown in Fig. 1. Based on the relative positions of the band edges of these materials, an e$cient cell is only possible with ZnS heavily n-type doped, CdS lightly doped, and I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 205 Fig. 2. Ideal band diagrams of Cd Zn S/CdS/CdTe system (a) before and (b) after junction formation. 1~x x Where qs is the electron a$nity in the n-type semiconductor and qs that in the p-type semiconductor, / 1 E the Fermi level, x the CdS/CdTe metallurgical junction, = original CdS layer thickness, x and x are F + ' / 1 depletion layer edges in the n and p-type regions, respectively. a normal p-type doped CdTe "lms. But the di$culty in doping ZnS makes Cd Zn S a suitable alternative. As a result the latter shall be used as the doped 1~x x form of ZnS. The conduction and valence band edges of this material lie between [5] those of ZnS and CdS depending on the Zn or Cd content. After doping, the band structure of the "lms before and after the junction formation is shown in Fig. 2(a) and (b), respectively. In Fig. 2(b), the high electron a$nity (s ) of CdS compared to that of / CdTe (s ) helps eliminate the energy spike at the CdS/CdTe junction. At the 1 Cd Zn S/CdS interface we also note that there is no energy spike. This is because 1~x x these materials are highly soluble [6] into each other, and the graded junctions between them do exist in which band edges change over certain distance with a "nite slope corresponding to electric "eld strength. For a thin CdS "lm, the slope covers the whole of this intermediate layer. This, coupled with the high conductivity of doped Cd Zn S layer ensure [7] that the depletion region extends through out the 1~x x original thickness, = , of CdS layer and up to x in the CdTe layer. Under AM1.5 ' 1 illumination, if Cd Zn S is transparent to this spectrum, the carriers generated by 1~x x the higher energies photons down to 2.4 eV in the CdS layer, now well removed from the cell surface, are no longer lost to surface recombination. Instead, they are separated and collected by the assistance of the "eld present in this region. The consequence therefore is a strong spectral response of the solar cell in almost all 206 I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 AM1.5 spectrum resulting in higher e$ciency CdTe solar cell. The objective of this study is to demonstrate that the traditional &Metal/CdTe/CdS/TCO/glass' structure of CdTe solar cells can be changed to &Metal/CdTe/CdS/Cd Zn S/TCO/glass' which 1~x x has the potential of attaining the highest e$ciency ever recorded in CdTe solar cells. 2. Experimental CBD and CSS were used to grow the window and the absorber layers of Cd Zn S/CdS/CdTe solar cell structures, respectively. The Cd Zn S layer was 1~x x 1~x x formed from the processed CBD grown ZnS/CdS multilayer. The CBD growth of CdS thin "lms was carried out in a 803C, 140 ml stirred solution consisting of 18.4 M) deionized water, 0.05 g cadmium acetate dihydrate (Cd(CH COO) ) H O), 0.1 g 3 2 2 ammonium acetate (NH CH COO), 6.5 ml ammonium hydroxide (NH OH) of 4 3 4 normality 14.3, and 0.03 g thiourea (SC(NH ) ). The CBD growth of ZnS thin "lms, 22 on the other hand, was carried out in a similar bath at 503C where the reagents are 0.4 g zinc sulfate heptahydrate (ZnSO ) 7H O), 0.4 g ammonium sulfate 4 2 (NH ) SO ), 0.11 g thiourea (SC(NH ) ), 6 ml ammonium hydroxide (NH OH) of 42 4 22 4 normality 14.3, 10 ml of 100% hydrazine monohydrate (N H ) H O), and 0.5 g of 2 4 2 nitrilotriacetic acid dissolved in 1.45 ml KOH solution. For the CdTe "lms deposition closed space sublimation (CSS) reactor was used. Here, the 99.999% pulverized CdTe powder source at 6503C in 15}30 torr He/O mixture ambient was sublimated onto 2 a substrate at 6003C which was 0.2 cm away from the source. Our primary substrate was a cleaned Libbey Owens Ford TEC 8 transparent conducting oxide coated soda lime glass of 3 mm thickness. The superstrate for the deposition of CdTe were of two types: (1) In the "rst type, a 0.07 lm thick ZnS "lm was "rst deposited on TCO/glass substrate. This was then dipped in 1% CdCl methanol solution for about 30 s to 2 improve the conductivity, then dried with an infrared lamp, and then rinsed in the deionized water. This was followed by an additional deposition of 0.05 lm CdS thin "lm to complete the superstrate fabrication. The solar cells fabricated from these substrates are called type 1. (2) In the second type, a 0.03 lm thick CdS "lm sandwiched between two 0.04 lm thick ZnS "lms were "rst deposited by CBD on TCO/glass substrate. The glass multilayers substrate was then given the usual CdCl treatment described above. 2 The sample was then annealed in CdCl /Ar ambient at 4003C for 15 min. The 2 substrate fabrication was completed by the deposition of an additional 0.035 lm thick CdS "lm. The solar cells fabricated from these ones are called type 2. We need to note that the in-situ di!usion of Cd into ZnS layer in the CdS growth bath is quite signi"cant. This together with the low temperature annealing ensured the formation of Cd Zn S/CdS layer as reported elsewhere [8]. 1~x x Prior to the deposition of CdTe in each case the CdS/Cd Zn S/TCO/glass 1~x x structure was further annealed in situ in the CSS reactor chamber in H ambient for 2 I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 207 about 15 min. For the purpose of comparison a CdS/CdTe structure was also fabricated. In all cases after CdTe deposition, the structure was treated with CdCl , annealed 2 in He/O ambient at 4003C, and etched in 0.5% Br methanol solution to form Te 2 2 rich surface. The structures were then masked; Cu doped graphite contact applied and processed, and the cells isolated. Indium was soldered on TCO around each device to improve current collection. To determine the extent of Zn di!usion SIMS analysis were performed on types 1 and 2 cells from which the contacts were stripped. SIMS data were taken using a CAMECA IMS-3f with O` primary beam and detection of secondary positive ions. 2 The impact energy was 5.5 keV. Typical sputtering conditions were 150 nA O` into 2 a 200 lm]200 lm crater. Ions were detected from 60 lm diameter region at the center of the crater. Illuminated I}< measurements were performed on the cells using xenon-arc lamp simulator with NREL-con"rmed devices. The J}< curves were used to extract < , 0# series, and shunt resistance. External quantum e$ciency (QE) measurements were performed using zero-light bias with total intergrated current set to measure J of 4# each device. 3. Results and discussions The solar cell data extracted from the illuminated J}< characteristics of Fig. 3 are summarized in Table 1. From Table 1 and Fig. 3, we observe that the J of the type 4# 1 cell is the highest and that of CdS/CdTe is the least. The explanation of these observations is straightforward by taking a closer look at the spectral response curves, Fig. 4, of the cells. In the following order CdS/CdTe, type 1, and type 2 Cd Zn S/CdS/CdTe solar cells have 80% or slightly higher QE in the 1~x 1~x 600}850 nm, 525}850 nm, and 510}850 nm wavelength regions. Below these regions, at 500 nm, say, the QE is about 62% for CdS/CdTe cell, 65% for type 1 cell, and 75% for type 2 cell; and at 400 nm, the QE are 38%, 40%, and 38% in that same order. We note that cells with Cd Zn S layer (Types 1 and 2 cells) generally have better short 1~x x wavelength response, as expected, and may explain why they have higher J . How4# ever, for these same type of cells, between 550 and 850 nm the QE of type 2 cell is slightly lower than that of type 1, but from 450 to 550 nm the QE of type 2 cell is higher than that of type 1. This slight response reversal may explain why the J of 4# types 1 and 2 cells di!er by only 0.1 mA/cm2 in favor of type 1. The reduced J and 4# QE between 550 and 850 nm of type 2 cell could be as a result of some recombination at the interface due possibly to stress induced defects. The low < of type 2 cell, and 0# type 1 cell for that matter, seems to con"rm our speculation, since these defects are reported to have stronger impact [9,10] on this parameter. To "nd a possible source of these interface defects that reduce the < of type 1 and 0# type 2 cells, we consider the SIMS analysis, Fig. 5, of these cells. Zn seems to have di!used through to the CdS/CdTe interface and more into CdTe layer in the type 2 cell than in the type 1 cell. It is known that lattice mismatch of Cd Zn S/CdTe is 1~x x 208 I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 Fig. 3. J}< characteristics of thin "lm CdTe solar cells. Table 1 Cell data for CdS/CdTe and type 1 and 2 Cd Zn S/CdS/CdTe solar cells 1~x x Cell type < (mV) 0# J (mA/cm2) 4# R () cm2) 4 R () cm2) 4) FF g (%) CdS/CdTe 1 2 804 739 727 21.3 21.9 21.8 4.5 5.0 10.3 2150 889 1054 0.67 0.62 0.54 11.4 10.0 8.6 higher than that of CdS/CdTe. This mismatch increases with increase in Zn content in Cd Zn S. The < of 625 mV that was previously reported [11] for 1~x x 0# Cd Zn S/CdTe and 630 mV for ZnS/CdTe solar cells [10] suggested that the 1~x x defects due to this lattice incompatibility have a direct impact on this parameter. This study gives direct evidence that the lattice mismatch is indeed the culprit. For, our type 2 cells with Zn about the interface, in amount probably lower than the previous case [10,11], the < is 727 mV. Whereas, the < of type 1 cell with low Zn content is 0# 0# I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 209 Fig. 4. Spectral response curves of CdTe thin "lm solar cells. 739 mV, and that CdS/CdTe cell with no Zn except those non-intentionally introduced in the structure during processing and the growth of the window layer from the CBD bath has 804 mV as its < . Since it is the relative displacement of the 0# quasi-Fermi energy levels at the interface that determines the solar cell < , it is 0# therefore logical to say that the lattice mismatch of the materials in this region does pin these levels restricting their motion, and hence the < of the cell. The shunt 0# resistance, the FF, and the e$ciency are also a!ected by these defects as indicated by their values in the Table 1. Nevertheless, the 10% e$ciency of our best Cd Zn S/CdS/CdTe solar cell is much higher than [10] 2% of ZnS/CdTe and [11] 1~x x 7% of Cd Zn S/CdTe solar cells. Our design approach then represents the right 1~x x approach if the present record [9] 16% e$ciency of CdTe solar cell is to be raised further. 4. Conclusion The thorough understanding of CBD growth mechanisms of ZnS and CdS and the improvement of their processing in the course of our previous studies [8] has enabled us to propose and fabricate a new CdTe solar cell structure. The trend in the spectral response of our CdS/CdTe solar cell is same as that of CdS/CdTe that has produced 16% e$ciency, the highest ever recorded [9] in CdTe solar cells. The 11.4% e$ciency recorded in this study has to do with low quality soda lime glass. This glass has 3 mm thickness translating into low transmission. Plus, its absorption edge starts around 500 nm where AM1.5 is quite rich [3] in photons. It 210 I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 Fig. 5. SIMS depth pro"les of (a) type 1 and (b) type 2 CSS-CdTe/CBD-CdS/Cd Zn S solar cells. 1~x x also contains sodium that has the tendency of di!using into the active "lm layers where they create defects or defect complexes that act as recombination centers. However, the spectral responses in the short wavelength region of AM1.5 of our Cd Zn S/CdS/CdTe solar cells are better than CdS/CdTe cells. Thus if the 1~x x I.O. Oladeji et al. / Solar Energy Materials & Solar Cells 61 (2000) 203}211 211 thickness of Cd Zn S and CdS "lms and the post-growth treatment can be 1~x x optimized to minimize the Zn di!usion to the interface with a borosilicate glass used as the substrate, a record e$ciency could be achieved. Nevertheless, in this study, using soda lime glass as the substrate with no anti-re#ection coating we have produce a 10% e$cient Cd Zn S layer containing CdTe solar cell, the highest ever reported 1~x x for this type of cell. Acknowledgements We would like to acknowledge Mr. Fred Stevie of Lucent Technologies and all our colleagues at the Material Characterization Facility of the University of Central Florida for the SIMS data and other discussions. References [1] H.J. Moller, Semiconductor for Solar Cells, Artech House, Boston, 1993. [2] B. Ray, II}VI Compounds, Pergamon Press, Oxford, 1969. [3] C. Ferekides, J. Britt, Y. Ma, L. Killiam, Proceedings of the Twenty Third IEEE photovoltaic Specialists Conference, 1993, p. 389. [4] P. Capper, in: P. Capper (Ed.), Narrow-gap II}VI Compounds for Optoelectronic and Electromagnetic Applications, Chapman and Hall, New York, 1997, p. 211. [5] K.L. Chopra, S.R. Das, Thin Film Solar Cell, Plenum Press, New York, 1983. [6] W. Lehmann, in: S. Shinoya, H. Kobayashi (Eds.), Proceedings of the Fourth International Workshop on Electroluminescence, Vol. 38, Springer-Verlag, Berlin, 1989, p. 371. [7] J.J. Liou, Advanced Semiconductor Device Physics and Modelling, Artech House, Boston, 1994, p. 105. [8] I.O. Oladeji, Ph.D. Dissertation, University of Central Florida, 1999. [9] H. Ohyama, T. Aramoto, S. Kumazawa, H. Higuchi, T. Arita, S. Shibutani, T. Nishio, J. Nakajima, M. Tsuji, A. Hanafusa, T. Hibino, K. Omura, M. Murozono, Proceedings of the Twenty sixth IEEE photovoltaic Specialists Conference, 1997, p. 343. [10] D. M. Oman, Ph.D. Dissertation, University of South Florida, 1995. [11] T.L. Chu, S.S. Chu, J. Britt, C. Ferekides, C.Q. Wu, Proceedings of the Twenty Second IEEE photovoltaic Specialists Conference, 1991, p. 1136.