Abstract

We report the fabrication of superstrate-structured solar cells by the deposition of Cu-In-S (CIS) films on under air by spray pyrolysis. The cells had an open-circuit voltage of 0.551 V, a photocurrent density of 9.5 mA/cm2, a fill factor of 0.45, and a conversion efficiency of 2.14%. However, transmission electron microscopy/energy dispersive X-ray (TEM-EDX) analysis revealed significant differences between the atomic ratio of the setting material in the spray-deposition solution and the elements in the layer. Moreover, TEM-EDX measurements suggested strong segregation of the Cu-In-S elements in the spray-pyrolysis-deposited layer. The degree of segregation depended on the substrate ( , , or ), although Cu3In5S9 nanoparticles were segregated in the sulfur layer.

1. Introduction

Compound solar cells with high-conversion efficiencies (20.3%) have been fabricated by vacuum methods [1], which are slow and require costly equipment. Nonvacuum processes for solar cell materials, such as Cu(In1-xGax)Se2, CuInSe2, and CuInS2 (CIS), have been investigated with the aim of developing fast and inexpensive fabrication methods. Printing [2, 3], spray-pyrolysis-deposition (SPD) [4, 5], selenization of printed metal oxides with H2S or H2Se [6], and electrochemical deposition [7, 8] have been explored as alternative methods. Less toxic materials are required because nonvacuum methods are carried out in air; therefore CuInS2 absorber materials are more suitable than CuInSe2 or Cu(In,Ga)Se2. This is particularly important for SPD methods because aerosols are formed from the compounds. In this work, the structures of spray-pyrolysis-deposited Cu-In-S layers, which have 2.14% conversion efficiency in superstrate-structured solar cells, were studied by transmission electron microscopy-energy dispersive X-ray (TEM-EDX) analysis. It was confirmed that SPD formed a segregated structure in the Cu-In-S layers and the Cu-In-S structure differed between substrates.

2. Experimental

Cu-In-S films were deposited on substrates at 300°C in air, by SPD. The substrates were glass, , and . Figure 1 shows the fabrication processes for the Cu-In-S layer and the resulting Cu-In-S structures. The glass substrates were coated with a TiO2 layer ( = 100 nm) by SPD at 400°C. The solution used for depositing the TiO2 compact layer was a mixture of titanium acetylacetonate (TAA) and ethanol (9 : 1 v/v). TAA was prepared by slowly injecting acetylacetone (99.5%, Kanto Chemical Co. Inc., Tokyo, Japan) into titanium tetraisopropoxide (97%, Kanto Chemical Co. Inc.) at a mole ratio of 2 : 1. The In2S3 layers ( = 100 nm) were used as the buffer layers and deposited using a precursor solution of InCl3 (10 mM) and thiourea (20 mM) in water (10 mL) at 300°C. The Cu-In-S spray deposition was performed using an aqueous solution of CuCl2·2H2O (30 mM; 97.5%, Kanto Chemical Co.), InCl3 (30 mM; 98%, Kishida Chemical Co., Ltd, Japan) and thiourea (150 mM; Tokyo Chemical Industry Co., Ltd, Tokyo, Japan) giving a Cu/In/S ratio of 1 : 1 : 5. The Cu-In-S absorber layer was deposited by spraying the Cu-In-S SPD solution (40 mL) on to each substrate at 300°C at a speed of 2 mL min−1. The Cu-In-S layers were scratched from the substrates and powdered, dispersed in ethanol, and attached to Mo grids for TEM observation. A TEM-EDX system (JEOL JEM2100) was used to characterize the Cu-In-S materials. XPS analysis conducted previously showed an oxidation state just on the Cu-In-S surface [5]. Thus, the powdered Cu-In-S samples for TEM observation show the inside of the SPD layer.

SEM cross-section images showed that the surface is relatively flat [5] with a roughness factor of only 1. However, there were several pinholes due to dust and aggregates. Moreover, observation by atomic scale view indicated a very rough surface because it is not a single crystal layer, hence the superstrate-type structure of the solar cells has been fabricated.

The back contact electrode formed from an Au layer deposited by vacuum evaporation produced superstrate-structured solar cells . The samples used to obtain the photo I-V curve were 5 mm × 5 mm. The photovoltaic measurements employed an AM 1.5 solar simulator equipped with a xenon lamp (YSS-100A, Yamashita Denso, Japan). The power of the simulated light was calibrated to 100 mW cm−2 by using a reference Si photodiode (Bunkou Keiki, Japan). The I-V curves were obtained by applying an external bias to the cell and measuring the generated photocurrent with a DC voltage current source (6240, ADCMT, Japan). It is very important to obtain information regarding the electrical properties of samples on various substrates for the preparation of Cu-In-S solar cells. However, since the resistance of the Cu-In-S layer was very high, the electrical resistance and the Hall effects could not be measured.

3. Results and Discussion

Figure 2 shows the photo and dark I-V curves for the superstrate-structured solar cell fabricated by SPD. The crossing of the curves suggests double-diode effects in the CIS solar cells. The open-circuit photovoltage, the short-circuit photocurrent density, the fill factor, and the conversion efficiency were 0.521 V, 9.68 mA cm−2, 0.423, and 2.14%, respectively, and were improvements compared with our previous report [5], obtained by optimizing the SPD conditions. The conversion efficiency is far removed from the best Cu-based solar cells (20.1%) [1]; however, if the efficiency of SPD-processed solar cells can be enhanced by over 10% in the near future, this would be a very significant improvement due to the high-speed and nonvacuum processing methods used in the fabrication of cost-effective solar cells. Thus, further investigation is required to improve the conversion efficiency. Here, we report changes in the CIS layer as observed by XRD and TEM-EDX.

Figure 3 shows the XRD patterns for the Cu-In-S layers on the different substrates that produced the , , and structures. The peak at 26.4° was attributed to SnO2 (26.6° by JCPDS 88–0287) in FTO (Figure 3(a)); therefore, an SnO2 peak was not observed in the XRD pattern of Cu-In-S on the glass substrate (Figure 3(b)). The peak at 27.6° in Figure 3(a) was assigned to In2S3 (27.4° by JCPDS 73–1366) and TiO2 (27.7° by JCPDS 88–1173) and was not observed in the (Figure 3(c)) and (Figure 3(d)) XRD patterns. Hence, the In2S3 layer had diffused into the Cu-In-S layer, confirmed by EPMA mapping analysis of the cross-sectional image [5]. The TiO2 peak was buried below the Cu-In-S layer (Figure 3(c)) because of the very thin TiO2 layer ( = 200 nm) and made evident by the small XRD signal.

Although the positions of the peaks were very close to the CuInS2 peak (27.9° by JCPDS: 85–1575), the Cu-In-S XRD peaks (27.8–28.4°) varied significantly among substrates (Figures 3(b)–3(d)) and the peak shifts suggested variations in the types of Cu-In-S crystals. Due to the large ratio of sulfur in the Cu-In-S layer (Cu-poor), which will be addressed in the TEM-EDX discussion to follow, assignment of the peaks from 27.8° to 28.4° in Figures 3(b)–3(d) would be CuIn11S17 (27.7° by JCPDS: 34–0797), CuIn5S8 (27.7° by JCPDS: 72–0956), and CuIn5S9 (28.4° by JCPDS: 49–1383). Therefore, it was thought that the peak shifts may be due to variations in the Cu-In-S crystals, which are projecting the compositions of the Cu-In-S.

The elemental analysis of Cu-In-S by TEM-EDX is shown in Table 1 and the locations analyzed are shown in Figures 4, 5, and 6. The Cu-In-S samples for TEM-EDX analysis were taken from the layers deposited on the different substrates: glass, glass/TiO2, and glass/TiO2/In2S3. It was confirmed that the Cu-In-S elements were segregated in the SPD layer. The components of the Cu-In-S materials were CuInS2, Cu3In5S9, Cu2In4S7, CuIn3S5, and CuIn5S8 [9] with atomic ratios calculated as 25.00 : 25.00 : 50.00, 17.65 : 29.41 : 52.94, 15.38 : 30.77 : 53.85, 11.11 : 33.33 : 55.56, and 7.14 : 35.71 : 57.14, respectively. Therefore, the Cu-poor and In-rich portions of the segregated Cu-In-S layer were close to the composition ratio of Cu3In5S9.

Figure 4 shows the TEM images of Cu-In-S deposited on a glass substrate. It was confirmed that, although the spray-pyrolysis solution for the Cu-In-S layer contained a significant amount of copper and indium, the layer on the glass substrate was mainly pure sulfur (Figure 4(a), Table 1). The sulfur particles were several hundred nanometers in size. The atomic lattice was not visible in the TEM images because sulfur is much lighter than the heavy metal elements. However, the electron scattering patterns showed a single crystal (inset, Figure 4(a)). A small number of nanoparticles around 10 nm in diameter were observed (Figure 4(b)) and revealed the atomic lattice because they contained metallic elements. The electron sputtering pattern (inset, Figure 4(b)) suggested the existence of a polycrystalline structure. EDX analysis indicated that the Cu-In-S elemental ratio was close to Cu3In5S9. Hence, a considerable amount of the copper metal source (CuCl2) did not end up in the Cu-In-S layer and was dispersed somewhere else during SPD. The structure of the Cu-In-S layer (Figure 1(a)) shows the segregated Cu3In5S9 particles in the sulfur layer.

Figure 5 shows the TEM images of Cu-In-S deposited on a glass/TiO2 substrate. In contrast to the glass substrate (Figure 4), most of the Cu-In-S layer consisted of Cu3In5S9 nanocrystals, 10 nm in diameter (Figure 5(a)). The electron diffraction pattern also suggested it had a nanocrystalline structure. Again the heavy metal elements meant that the nanocrystal atomic lattice could be observed. Although the nanocrystals contained copper, the Cu/In ratio was only 0.54 : 1, which was half that of the setting ratio (1 : 1) in the spray-deposition solution. The CuCl2 in the Cu-In-S solution was again dispersed and not deposited in the Cu-In-S layer. In the minor part of Cu-In-S layer, a small amount of copper was observed (Figure 5(b)). However, when the substrate was changed from glass to glass/TiO2, the amount of metal ([Cu + In]/[Cu + In + S]) in the sulfur rich part increased from 0.06% (Figure 4(a)) to 0.44% (Figure 5(b)). The electron scattering suggested a single crystal (inset, Figure 5(b)), although an atomic lattice was not observed, which may be because sulfur is a light element. Therefore, it was concluded that the substrate ( or ) significantly affected the composition and structure of the spray-pyrolysis-deposited CuInSe2 layer. The structure of Cu-In-S layer deposited on the glass/TiO2 substrate by spray pyrolysis is shown in Figure 1(b).

Figure 6 shows the TEM images of Cu-In-Se deposited on a glass/TiO2/In2S3 substrate. The atomic ratio showed that, for Cu-In-Se, the layer mainly consisted of Cu3In5S9 (Figure 6(a)), although a small amount of Cu3In1S80 (Figure 6(b)) was present. Pure sulfur (Figures 4(a) and 5(b)) was not observed because of the In2S3 buffer layer and interdiffusion to the upper Cu-In-S layer. Cu3In5S9 was present as nanocrystals, around 10 nm in diameter (Figure 6(a)), and the electron diffraction pattern suggested a polycrystalline structure (inset, Figure 6(a)). The Cu3In1S80 particles were around several hundred nanometers in size (Figure 6(b)) and the atomic lattice was not visible because of the light sulfur atoms, although the electron diffraction pattern suggests a single-crystalline structure (inset, Figure 6(b)). Therefore, Cu-In-S was segregated as Cu3In5S9 and Cu3In1S80. The structure of the Cu-In-S layer deposited on the glass/TiO2/In2S3 substrate by spray pyrolysis is shown in Figure 1(c).

The ratios of Cu-In-S, such as Cu3In5S9 and Cu3In1S80, simply show integral multiplication of the TEM-XRD results of Table 1. From a scientific point of view, the crystal structures should be revealed by XRD analysis. However, the results from normal XRD were obscure and GIXRD may be a more suitable method. Nevertheless, although GIXRD may have provided better results, the Cu-In-S was a mixture of small crystals and amorphous phases of different Cu-In-S compositions, making crystal identification in this work very complicated. It would be desirable to improve the analytical methods in the future.

4. Conclusions

Compound solar cells based on Cu-In-S (CIS) were prepared by SPD under air. Although the Cu-In-S layer had a segregated structure, the solar cells had a 9.68 mA cm−2 short-circuit photocurrent density and 2.14% conversion efficiency. The segregated structures differed among substrates. The resulting Cu-In-S layers were copper poor due to the loss of CuCl2 during SPD. Therefore, the next step in improving the conversion efficiency is to control the amount of copper in the Cu-In-S layer. Since the significance of the contribution of a Cu-poor CuInS2 and CuInSe2 solar cells has been discussed in the publications [10, 11], we believe that a Cu-poor CuInS2 layer, and hence the presence of Cu3In5S9, may contribute to the photovoltaic effects.

Samples prepared for this kind of work may be solid mixtures of CuInS2, CuIn3S5, and so on and should be clearly identified. Each major part of the sample was expressed in the TEM figures in this paper. It is very difficult to obtain the exact components of such segregated materials by other methods, such as XPS, Auger, SIMS, Raman, XRD, and EPMA, because the resolutions in these methods reach the submicron range and give mixed information about major and minor parts of the materials. Advancements in the analytical tools may reveal further details in the future.

Acknowledgment

This work was funded in part by the Innovative Solar Cells Project (NEDO, Japan).