Hydrazine-Free Solution-Deposited CuIn(S,Se) 2 Solar Cells by Spray Deposition of Metal Chalcogenides

: Solution processing of semiconductors, such as CuInSe 2 and its alloys (CIGS), can signi ﬁ cantly reduce the manufacturing costs of thin ﬁ lm solar cells. Despite the recent success of solution deposition approaches for CIGS, toxic reagents such as hydrazine are usually involved, which introduce health and safety concerns. Here, we present a simple and safer methodology for the preparation of high-quality CuIn(S, Se) 2 absorbers from metal sul ﬁ de solutions in a diamine/dithiol mixture. The solutions are sprayed in air, using a chromatography atomizer, followed by a postdeposition selenization step. Two di ﬀ erent selenization methods are explored resulting in power conversion e ﬃ ciencies of up to 8%.

T he use of vacuum-based deposition techniques for the preparation of chalcogenide absorber layers has resulted in highly efficient Cu(In,Ga)Se 2 (CIGS) photovoltaic devices. 1 Vacuum technologies, however, have a high capital cost, which is a major obstacle in the commercialization of CIGS. As a consequence, recent research on CIGS solar cells has been focused on developing solution-based approaches, where the deposition occurs under atmospheric conditions. These solution-based techniques are attractive, because the avoidance of high-vacuum processing equipment makes the process less expensive. Also, solution processes are more easily implemented in large area applications, as well as being better-suited to flexible substrates. 2 To date, the most successful approach in terms of solar cell performance was developed by Todorov et al., who demonstrated a 15.2% efficient device. 3 Their methodology involves the dissolution of metal chalcogenides (Cu 2 S and In 2 Se 3 ) and Ga with excess chalcogen (S, Se) in hydrazine and the clean decomposition of the formed precursors. As opposed to other precursors which have been used for solution-processed CIGS deposition, 4 metal chalcogenides are free from impurities, such as of halogens, carbon, or oxygen. These impurities can be easily incorporated into the thin film and as a consequence reduce device performance. However, metal chalcogenides are not easily dissolved in common solvents because of their strong covalent bonds, with hydrazine being the only solvent that can dissolve metal chalcogenides via the "dimensional reduction" process 5 in the presence of excess chalcogen. Although hydrazine is an excellent solvent for this purpose, it is difficult to implement in large scale production because of its high toxicity and explosive nature. As a consequence, recent research on low-cost deposition techniques for CIGS solar cells has been focused on eliminating the use of hydrazine and substituting it with a safer solvent.
Recently, Webber et al. reported that nine V 2 VI 3 chalcogenides (where V = As, Sb, Bi; and VI = S, Se, Te) can be dissolved at ambient conditions in a solvent mixture comprised of 1,2-ethanedithiol and 1,2-ethylenediamine. 6 Our group has also reported that the same solvent mixture can be used for the dissolution of Cu 2 S, Cu 2 Se, In 2 S 3 , In 2 Se 3 and that single phase CuIn(S, Se) 2 films can be obtained with good crystal growth after the selenization treatment of a spin-coated thin film. 7 Duan et al. have shown that CdS and SnSe have a high solubility in the same solvent combination and SnS, CdSe, ZnSe, and MoS 2 can be dissolved at lower concentrations. 8 Here our work is extended further by improving our precursor deposition and selenization methodology and by doing so, CuIn(S,Se) 2 (CISS) thin film absorber layers with high crystalline quality, suitable for solar cell applications, have been obtained.
The deposition of the precursor solution involved spraycoating instead of spin-coating. The spin-coated samples from our previous work 7 were highly porous, which hindered the formation of highly crystallized and dense films. However, denser films are formed when spray deposition is used, because a thinner layer of the material is deposited at each deposition pass, when compared to the spin coating method. This minimizes the volume contraction and the formation of cracks that normally occur during thermal decomposition of the molecular complexes into an inorganic phase. 9 The as-sprayed material is therefore much denser than the spin-coated material, which facilitates crystallization during the selenization step. Apart from the higher crystalline quality of the material, the use of spray deposition methods requires a smaller amount of starting material for the same film thickness, compared with spin-coating. This enables the thickness of the film to be built up easily and quickly. Spray deposition is a high-throughput and scalable technology and is very attractive from a manufacturing process viewpoint. The spray pyrolysis technique has been successfully implemented in literature for the preparation of CIGS from metal salts and thiourea. 10 The drastic improvement in the film quality compared to our previous work 7 was mostly obtained by substitution of spin coating with spray coating technique (denser as-deposited films). However, dilution of the precursor solution and the use of graphite box during selenization (see the Supporting Information) showed additional improvement in the material quality.
The thin films are deposited in ambient atmospheric conditions within a fume hood using a glass chromatography atomizer. The solutions are prepared targeting a copper poor film stoichiometry (see the Supporting Information). The films are sprayed on a molybdenum coated glass substrate, placed on a preheated hot plate. An intermediate drying step is performed in-between the spraying of each layer. Finally, the postdeposition selenization treatment is performed in a tube furnace or a Rapid Thermal Processing (RTP) oven with selenium pellets. RTP was explored as an alternative selenization method in an effort to reduce the thermal treatment time and to improve the crystallization of the film and the device performance. By using RTP selenization, a power conversion efficiency (PCE) of 8.0% was obtained without an antireflection coating or grids, for a cell with an area of 0.24 cm 2 .
SEM and XRD were used to evaluate the surface morphology and the crystal structure of the absorber. Figure 1a shows an SEM image of an absorber surface, which was selenized in a tube furnace and deposited from a mixture of Cu 2 S and In 2 S 3 , targeting a copper poor ratio (Cu/In = 0.8) of the precursor. EDS analysis shows that the Cu/In ratio remains approximately  . This is indicative of the decreasing dspacing, as would be expected with increasing sulfur substitution in the crystal lattice. 11 This shift indicates that only a small amount of sulfur is present in the film and that sulfur is mostly displaced by selenium during the selenization step. The broad peaks at 2θ ≈ 32 and 56°(noted with circles) correspond to a MoSe 2 layer, most likely located between the Mo and the absorber. Such an interfacial region creates an ohmic contact between the CIGS/CIS and Mo layer and it also improves the adhesion of the absorber to the Mo back contact. 12 The XRD spectrum of the as-deposited sample ( Figure S6) consists of two low-intensity peaks at 2θ ≈ 28 and 46.7°, which correspond either to the (112) and (220) peaks of CuInS 2 (JCPDS 15−0681), or to a Cu x S phase. In combination with the EDS analysis ( Figure S3), these results suggest that some CuInS 2 is formed during spraying, with selenium displacing sulfur during selenization. Figure 2a shows the current density−voltage (J−V) characteristics of a device based on the tube furnace selenization approach, measured under dark and simulated AM1.5G illumination. These results are for the highest performing cell of the sample area (2.5 × 2.5 cm). The configuration of the device was: AZO/iZnO/CdS/CISS/Mo/ glass. A PCE of 7.3% was obtained, with an open circuit voltage (V oc ) of 432 mV, a short circuit current density (J sc ) of 31.1 mA cm −2 and a fill factor (FF) of 54%. The V oc and J sc of this device is comparable to that of a hydrazine-processed CuIn(S,Se) 2 solar cell with low sulfur content (n = 10.7%, V oc = 440 mV, J sc = 35.9 mA cm −2 , FF = 67.5%). 13 In comparison to the hydrazine-based device, our cell is limited by the low FF, which is mainly attributed to a high series resistance of 3.5 Ω cm 2 . This high series resistance can be related to excessive formation of MoSe 2 12 and the incomplete crystallization of the absorber throughout its thickness. It is expected that further improvement of the selenization step, the transparent conductive oxide (TCO) deposition, as well as the evaporation of current collecting grids will increase the FF of the device significantly. In parallel with these improvements, the formation of MoSe 2 will be carefully optimized. Figure 2b shows the external quantum efficiency (EQE) measurement of the same device. The observed decrease in EQE at wavelengths below 550 nm corresponds to the absorption of the CdS buffer layer. The EQE data shows good collection efficiency from 550 to 700 nm, but the collection is reduced gradually at longer wavelengths. The drop in EQE at wavelengths >700 nm indicates that there is high recombination toward the back of the absorber. This suggests that the device is not fully recrystallized throughout the thickness and that the selenization step should be improved further. The band gap (E g ) was estimated from the EQE curve, by plotting [E.ln(1-EQE)] 2 versus E 14 (shown in the inset of Figure 2b), where E g was extracted from the band edge by extrapolating to the horizontal axis intercept. The estimated bandgap of 1.06 eV corresponds to CuIn(S, Se) 2 with low sulfur content. On the basis of previous work that associates the S/Se ratio with the band gap (assuming a linear relationship between the bandgap values of CuInSe 2 and CuInS 2 ), 11,15 the estimated band gap from the QE data corresponds to a sulfur content of 4 at %. This result suggests that sulfur is mostly displaced by selenium during selenization, in agreement with the XRD data and EDS analysis.
The selenization process needs to be optimized for the complete crystallization of the absorber to further improve performance. A longer selenization time could potentially improve the crystallization, but it may also increase further the series resistance by excessive MoSe 2 formation. A different procedure was therefore used for a CISS sample deposited with the same conditions. An alternative selenization approach using an RTP oven was explored, which enables a significantly higher heating ramp rate compared to the tube furnace. Apart from being attractive as a low cost technique since it has a shorter processing time and lower thermal budget, use of RTP improves the material quality of chalcopyrites. Specifically, it was reported that the use of RTP resulted in improved crystal growth as detrimental phases can be avoided by passing intermediate formation temperatures rapidly. 16 Also, a different TCO layer combination was used compared to the previous device. The configuration of the RTP processed device was: IZO/ZTO/CdS/CISS/Mo/glass (see the Supporting Information). It should be noted that these two devices were completed

ACS Applied Materials & Interfaces
Letter in different laboratories; therefore different configurations were employed according to the respective lab standard process (between Loughborough and NREL). Although a direct comparison between devices is difficult to make, because of the different TCOs and buffer layers used, we are highlighting the merits of each selenization process in terms of crystal growth and MoSe 2 formation. Figure 3a shows the SEM image of the absorber top surface of a sister sample, selenized in the RTP oven. Information on the instrumentation, temperature profiles, and composition of the sample is given in the Supporting Information. The SEM image indicates that the films are homogeneous and crack-free, with grains of up to 1 μm size. Figure 3b shows the XRD spectrum of the sample, which verifies the presence of a singlephase chalcopyrite material. The MoSe 2 peaks are evident, as in the case of the sample selenized in the tube furnace. The intensities of the MoSe 2 (100) and Mo (110) diffractions can be related to the thickness of the MoSe 2 layer. 12 Here, the intensity ratio of the Mo/MoSe 2 peaks (at 2θ ≈ 40 and 32°) is higher for the RTP-processed sample (6.1 compared to 3.5). This suggests that the MoSe 2 layer is thinner for the RTPprocessed sample. Raman was also performed in combination with XRD to verify the absence of secondary phases ( Figure  S7). A peak at ∼175 cm −1 is identified as the A1 peak of chalcopyrite CuInSe 2 and a double peak is identified as the B2 or E mode. 17 Figure 4a shows the dark and light J−V curves for a device, with the absorber selenized in the RTP oven. The highest performing cell of the sample area (2.5 × 2.5 cm) has a PCE, V oc , J sc , and FF of 8.0%, 452 mV, 31.4 mA cm −2 , and 56%, respectively. In comparison to the 7.3% cell, the RTP-selenized device has a slightly improved V oc and a lower series resistance, which results in a better FF. The lower series resistance could be attributed to the shorter annealing time, which results in a thinner MoSe 2 layer formation at the interface. This is in agreement with the XRD spectra and the cross sectional SEM images of the two devices, which are shown in Figure S8. The cross-sectional images show a similar crystallization of the absorber for the two cells with ∼600 and 500 nm thick crystalline layer for the tube furnace and RTP-processed cell, respectively. This might suggest that the small difference in the series resistance is mainly attributed to the different MoSe 2 thickness, which is indicated in the images. Figure 4b shows the EQE measurement of the RTP processed device. In comparison with the previous device, there is a decrease in collection below 550 nm, good collection efficiency from about 550 to 700 nm and a reduced collection at longer wavelengths. This indicates that the absorber is not fully recrystallized throughout its thickness, which is also shown in the cross sectional SEM images of Figure S8. In comparison to the EQE of Figure 2b, this cell has a slightly better collection at longer wavelengths. This could be attributed to a slight improvement in the absorber quality or the use of a different TCO layer combination.
The estimation of the band gap (E g ) is shown in the inset of Figure 4b and equals to 1.03 eV, which corresponds to pure CuInSe 2 . 11,18 This verifies that sulfur is mostly displaced by selenium during the selenization process, within the accuracy of this calculation.
Overall, the EQE data shows that both devices have good collection but room for improvement remains. Specifically, further optimization of the selenization process should result in the complete crystallization of the absorber, which will improve the EQE at longer wavelengths and consequently the J sc of the device. Apart from the selenization step, there is room for improvement in terms of the deposition and drying steps, in order to eliminate small voids (up to 200 nm) in the asdeposited film.
In summary, we have developed an atmospheric solutionbased process for the deposition of CuIn(S, Se) 2 absorbers, using a safer solvent combination than hydrazine, combined with metal chalcogenide precursors. Using this approach, we have fabricated a 8.0% CuIn(S, Se) 2 solar cell, with the performance of this device limited by high series resistance and incomplete recrystallization throughout the absorber. On the basis of these promising results, there is scope for further improvement of this new approach. Specifically, the increase in the band gap, by incorporating excess gallium or sulfur in the process, would result in an increase of the V oc and an improvement in conversion efficiency. Also, no sodium or antimony dopants have been used intentionally in this process. These dopants have been shown to improve the electronic

ACS Applied Materials & Interfaces
Letter properties 19 and enhance the grain growth of CIGS 20 respectively.
It is anticipated that further optimization of the device processing will result in higher device performance. The proposed approach using safer solvents than hydrazine has the potential for large scale manufacturing thin film CIGS devices. The deposition procedure is performed in air and avoids complicated precursor synthesis. It is a simple and easily controlled process. The use of a safer solvent mixture and the simplicity of this method have the potential to lead to an industrially scalable solution deposition methodology for CIGS.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10