Effect of Interface Passivation on the Performance of the Cu2ZnSnS4 Solar Cells Authors: Adèle Debono, Filipe Martinho, Jørgen Schou, Stela Canulescu Published: 2024 ABSTRACT This study examines the effect of ultrathin aluminum oxide (Al2O3) passivation layer on the performance of the kesterite Cu2ZnSnS4 (CZTS) solar cells. The Al2O3 layer was applied at the back CZTS/Mo interface using atomic layer deposition (ALD). Our findings indicate that the interface passivation with Al2O3 can significantly enhance the adhesion of CZTS to Mo, preventing delamination during annealing. An optimal Al2O3 thickness of 1–2.5 nm yielded a more than 30% increase in efficiency, primarily through enhancements of the open-circuit voltage (Voc) and short-circuit current (Isc), which resulted from reduced interface recombination and improved charge collection within the CZTS bulk. Furthermore, the Al2O3 layer modified MoS2 formation at the CZTS/Mo back interface, reducing series resistance (Rs). Nevertheless, a thicker Al2O3 layer led to a sharp decrease in efficiency due to increased series resistance and hindered the extraction of holes. Our work reveals that precise control of the Al2O3 layer thickness is essential to optimize CZTS solar cell performance by balancing the benefits of passivation with its potential drawbacks. Introduction Kesterite Cu2ZnSnS4 (CZTS) is a promising semiconductor for thin-film photovoltaics due to its cost-effectiveness, abundance of constituent elements, and nontoxic nature. CZTS has a direct band gap ranging from 1.48 to 1.63 eV depending on the Sn/Cu ratio.1 In addition, it exhibits a high absorption coefficient (>104 cm–1), allowing the deposition of ultrathin absorber layers and facilitating the manufacture of flexible solar cells.2 CZTS can be monolithically integrated with Si solar cells for tandem applications,3−5 either as it is or with Ag alloying to induce bandgap widening.6 CdS is commonly used as an n-type layer in kesterite CZTS and other thin-film technologies, but its toxicity poses challenges for large-scale industrialization. To replace CdS, nontoxic alternatives like Zn(O,S) are promising candidates. These materials can be deposited using methods such as atomic layer deposition, as demonstrated by previous work.7 CZTS solar cell performance faces limitations, particularly a significant open-circuit voltage (Voc) deficit, partly attributed to interface recombination losses.3 Addressing this issue is critical for advancing CZTS solar cell efficiency and viability in photovoltaic applications. Passivation, a long-standing technique in the solar cell industry, involves adding layers to protect materials from environmental corrosion or degradation. This method is used in high-performance silicon solar cells, such as passivated emitter and rear cells (PERC), which have achieved efficiencies of up to 26%.8 Passivation has been shown to improve the performance of solar cell devices by increasing the Voc.9 The back interface passivation consists of adding a passivation layer between the CZTS and the Mo back contact. Several materials have been investigated to passivate back contact, such as TiN,10,11 ZnO,10 Ag.12 The effects of the passivation layer on the device’s performance are complex and include: (i) reduction in the rear surface recombination velocity,13 (ii) control of MoS2 formation, which can be detrimental if too thick,14 (iii) improved homogeneity and compactness for the CZTS absorber, resulting in a reduction of voids,14 consequently enhancing Na diffusion in the absorber,14 (iv) reduction in the formation of secondary phases like SnS and ZnS.13 Given that high-performance CZTS solar cells are predominantly affected by interface recombination, passivation is a potential solution to mitigate this loss, as demonstrated for Si and CIGS cells.9,13,14 In this work, we investigate the impact of the nanometer-thick passivation layer of Al2O3, deposited via ALD at the “back interface” between the CZTS and Mo back contact. Experimental Section CZTS Solar Cell Fabrication The CZTS precursors were deposited by cosputtering of Cu, ZnS, and SnS targets on SLG/Mo substrates, as described in our previous work.15 The precursors were annealed at 585 °C for 15 min in a graphite box with additional S powder. A 50 nm-thick CdS was deposited by chemical bath deposition (CBD) using cadmium sulfate (CdSO) as a cadmium source, thiourea (CHNS) as a sulfur source, ammonium hydroxide (NHOH) as the complexing agent, and ammonium chloride (NHCl) as a buffer solution.16 A 50 nm intrinsic ZnO (i-ZnO) was deposited by sputtering an Al:ZnO target in Ar and O2 at 7.3 × 10–3 mbar using a 20 W RF power. Finally, the 350 nm-thick Al-doped ZnO (AZO) was deposited by sputtering of the same target in Ar at 6.7 × 10–3 mbar, 20 W RF power. Ultrathin Al2O3 Layer Deposition The Al2O3 layer was deposited by ALD at a temperature of 200 °C and pressure of 10 mbar. A cycle comprises four steps: trimethyl aluminum (TMA) deposition during 0.1 s with an N2 flow of 150 sccm, a purge for 3 s, water deposition during 0.1 s with an N2 flow of 200 sccm, and a purge for 4 s. The thickness of the deposited Al2O3 layers was measured using ellipsometry on silicon (Si) wafers. The results are presented in Table 1. It should be noted that the deposition rates determined via ellipsometry are slightly higher, approximately 0.11 nm per ALD cycle, compared to the internal calibration rate, which is around 0.097 nm per ALD cycle. Additionally, the thickness of Al2O3 might vary slightly when deposited on Mo compared to Si. Due to these variations, the thickness of the deposited layers is expressed in terms of the number of ALD cycles rather than in nanometers (nm). Table 1Thickness of the Al2O3 Layer at the Rear InterfaceALD cycles071115162126thickness (nm)00.881.201.662.452.322.93ALD rate (nm/cycle) 0.1260.1090.1110.1530.1100.113 Thickness of the Al2O3 Layer at the Rear Interface When Al2O3 is used at the back contact, a certain modification of the original Al2O3 layer could be expected due to the high-temperature sulfurization process (>500 °C) used in synthesizing the CZTS absorber layer. For example, a partial damage of the continuous Al2O3 layer or a modification of its composition and electrical transport properties could occur. Therefore, a wide range of ALD cycles were tested. Material and Device Characterization Scanning electron microscopy (SEM) images were acquired using a Zeiss Merlin microscope equipped with a high-efficiency secondary electron detector (HE-SE2). The SEM was operated at an acceleration voltage of 5 kV. Energy-dispersive X-ray spectroscopy (EDX) was performed using a Quanta Field Emission Gun (QFEG) at an acceleration voltage of 15 kV and working distance of 10 mm. The EQE measurements were performed on a QEXL setup with a monochromatic light chopped at a frequency of 109.9 Hz. Raman spectra were obtained using a Renishaw Raman spectrometer with 10 accumulations of 10 s each and a 50× magnification objective (numerical aperture 0.75) at room temperature. Three laser wavelengths were employed, namely 325, 532, and 785 nm. Results and Discussion Back-Contact Optimization in Kesterite CZTS Solar Cells Previous studies have determined that the optimal thickness for Al2O3 in nanopattern design is approximately 2 to 3 nm.9,14 The nanopatterned configuration has been shown to enhance the contact between the back contact and the CZTS layer. In cases without nanopatterning, the optimal Al2O3 thickness is expected to be lower. Therefore, the Al2O3 thicknesses investigated in this work ranged from 7 cycles (0.8 nm on Si) to 26 cycles (2.93 nm on Si), as given in Table 1. Figure 1 shows SEM cross-sectional images of CZTS solar cells without (a) and with an Al2O3 layer at the back interface (b, c). The images reveal that the interface between Mo and CZTS is significantly enhanced upon adding the Al2O3 layer. The adhesion of the absorber layer improves markedly with the incorporation of the Al2O3 layer. Conversely, the CZTS layer delaminates due to a poor CZTS/Mo junction interface. Notably, most reference devices without the Al2O3 layer experienced delamination either prior to or during CdS deposition. In contrast, all devices with Al2O3 at the back contact maintained structural integrity throughout the fabrication process. Statistics on the integrity of the devices are provided in Supporting Information Note 1. Lastly, no significant differences are observed in the grain morphology of the absorbers, despite previous work showing that Na diffusion from the Mo back contact into the absorber can be reduced by increasing the thickness of the Al2O3 layer.7 CZTS solar cells without and with Al2O3 layer at the back interface. Cross-section SEM images for a reference cell (a) and CZTS cells with 14 (b) and 26 ALD cycles Al2O3 layer (c). The red square marks the interface between Mo/MoS2 and CZTS. Impact of the Al2O3 Layer on the Cell Performance When Al2O3 is employed at the back contact, its structural and chemical properties may be modified by the high-temperature sulfurization process (>500 °C) used during CZTS synthesis. This thermal exposure can alter the Al2O3 layer, potentially affecting its composition and transport properties. To address these potential changes, a broader range of ALD cycles was tested for the rear interface to evaluate the impact of varying thicknesses on device performance. Next, we investigated the impact of Al2O3 thickness on the IV parameters of solar cells, specifically the short circuit current (Jsc), the open circuit voltage (Voc), and the fill factor (FF). Figure 2 presents boxplots of CZTS cells without Al2O3 layer (0 ALD cycles) and various Al2O3 layer thicknesses, up to 26 ALD cycles, excluding delaminated devices. Considering the median values, the optimal thickness is 11 ALD cycles, leading to a 20% improvement in efficiency (η = 2.72%) compared to the reference device (η = 2.16%). An optimal range is observed between 7 and 15 ALD cycles, corresponding to an Al2O3 thickness of 0.9–1.7 nm. The plateau in performance indicates that the process is tolerant to minor variations in thickness. However, beyond 16 ALD cycles, the efficiency decreases and falls below the reference devices. This suggests that an excessively thick layer increases series resistance and/or hinders hole collection by the back contact. Moreover, a significant decrease in the FF is observed for thicker alumina layers. A notable reduction in the shunt resistances (Rsh) of cells with 0, 11, and 26 ALD cycles was determined. The median Rsh value drops from 677 Ω cm2 for the reference cell to 100 Ω cm2 for the cell with 11 ALD cycles and further decreases to 37 Ω cm2 for the cell with 26 ALD cycles. Hence, the reduction in Rsh is responsible for the decrease in FF. CZTS solar cell performance upon back-contact passivation using Al2O3. Boxplots of (a) power conversion efficiency, (b) open circuit voltage (Voc), (c) short circuit current (Jsc), and (d) the fill factor (FF) for CZTS cells with and without Al2O3 layer (ALD cycles up 26). The blue dots represent each measurement included in the boxplot to the left. The green boxplot corresponds to the reference cells, while the cyan and blue boxplots represent two different series of CZTS cells. Each box spans from the first to the third quartile, with the extremities indicating the minimum and maximum values. The horizontal line within each box represents the median, the red square marks the mean, and the crosses denote outliers. Two types of cells solar cells were selected for further studies: a reference cell and a CZTS device produced from the same precursor with a 14 ALD cycle Al2O3 layer. The IV parameters of CZTS cells before and after back-contact passivation are shown in Figure 3 and Table 2. It includes box plots of efficiency, Jsc, Voc and FF, series resistance (Rs), shunt resistance (Rsh), ideality factor (A), and diode current (J0) determined from the method described in literature.17 The IV curves are shown in the Supporting Information Note 2. The most striking differences between the reference and passivated devices are the Voc and Jsc. The effect of passivation is evident in the increase of Voc, which reaches its maximum at 15 ALD cycles, likely attributed to reduced recombination.14Jsc follows the same trend as Voc, with a significant increase for the passivated device. The Jsc calculated from the EQE data (see also Supporting Information Note 3) increases from 12.75 to 13.48 mA/cm2 for 0 to 14 ALD Al2O3 cycles, respectively, which is consistent with the IV curve values (from 13.66 to 15.30 mA/cm2). However, Rsh is reduced, possibly because of impurities introduced during the ALD process. The diode saturation current density (J0) remains relatively high, and the ideality factor A exceeds 2 for all devices. An ideality factor greater than 2 is relatively common in kesterite devices.18 The large value of the ideality factor in CZTS devices indicates that charge carrier recombination in the space-charge region remains significant. IV parameters of CZTS reference device and the CZTS device with a 14 ALD cycle Al2O3 layer. Boxplot of IV parameters extracted for a reference cell (0 ALD cycles in green) and with 14 ALD cycles (blue). Table 2Comparison between Reference and Device with 14 ALD Cyclesnameη (%)Voc (mV)Jsc (mA/cm2)FF (%)Rs (Ω cm2)Rsh (Ω cm2)A (no unit)J0 (mA/cm2)no Al2O33.8246813.6660.031.315342.145.1 × 10–314 ALD cycles5.5658215.3063.790.364542.214 × 10–4relative change45.5%24.4%12.0%6.3% CZTS record1911.0173121.7469.272.5810001.446.8 × 10–11 Comparison between Reference and Device with 14 ALD Cycles The external quantum efficiency (EQE) spectra shown in Figure 4 illustrate the contribution of different wavelengths to the short-circuit current. For the best-performing cell, the maximum EQE increases from 64.20% for the reference cell to 69.15% for the alumina- passivated cell. There is a notable increase in the EQE of the device with a passivated surface in the long-wavelength region associated with CZTS absorption, indicating increased bulk collection efficiency. Figure 4b shows the EQE of the best cell (blue curve), internal quantum efficiency (IQE) (red curve), and 1 R (reflectance) (yellow curve). The device has a reflectance of about 10%. After accounting for optical losses, the IQE peaks at 75%, corresponding to an increase of 1 mA/cm2 short-circuit current. The band gap values extracted from the EQE spectra of over 84 cells, summarized in Table 3, reveal a mean band gap of 1.53 eV, consistent with theoretical values (1.48–1.63 eV). Quantum efficiency of CZTS solar cells. External quantum efficiency (EQE) of the reference device (a) and EQE, internal quantum efficiency (IQE), and 1 R (reflectance) of a solar cell with 14 ALD cycles (b). Table 3Band Gap Values of CZTS (Mean, Minimum, and Maximum Values) Extracted from Over 84 Solar Cells referencecell with 14 ALD cyclesmeanmaxminband gap (Eg)1.59 eV1.59 eV1.53 eV1.58 eV1.45 eV Band Gap Values of CZTS (Mean, Minimum, and Maximum Values) Extracted from Over 84 Solar Cells MoS2/Mo Back Interface CZTS was exfoliated from the substrate, and both surfaces were analyzed via SEM and Raman spectroscopy. Figure 5 (a–c) shows the top-view SEM images of the CZTS reference sample (0 ALD cycle) and with different Al2O3 thicknesses corresponding to 14 and 26 ALD cycles (CZTS side). The reference sample, with no ALD cycles, shows many holes, indicating poor adhesion. In contrast, the sample with 14 ALD cycles exhibits a much more compact structure, suggesting improved coverage and adhesion. On the other hand, the 26 ALD cycle sample shows no visible holes, indicating significantly enhanced adhesion and complete, uniform coverage of the CZTS layer. Figure 5 (d–f) shows the morphology of the substrate side (MoS2 side). The reference sample has small grains covering the surface, while the 26 ALD cycles sample has a typical Mo-sputtered film structure with an ultrathin layer of MoS2 on top. The 14-cycle sample appears to represent an intermediate morphology between the two. A schematic of the exfoliation mechanism is shown in the Supporting Information Note 4. SEM images of CZTS rear side and Mo back contact. The top-view SEM images of the exfoliated structures highlight the CZTS rear side (a–c) and MoS2 side (d–f). The number of ALD cycles corresponding to the Al2O3 thickness is indicated in each figure. Raman measurements taken on the surfaces of the films and substrate reveal distinct fingerprint features. Two dominant peaks associated with CZTS are observed at 286 and 339 cm–1, aligning with previous studies4 (Figure 6a). Notably, both surfaces have the in-plane E2g and out-of-plane A1g phonon modes of MoS2, confirming the presence of a MoS2 layer on both surfaces. Raman intensity maps (A1g peak) for different ALD cycles are shown in Figure 6c, d to provide a more detailed study of the surfaces. Raman spectra of CZTS rear side and substrate. Raman spectra of the CZTS back side (a) and MoS2 side (b) without an Al2O3 layer and with 14 and 26 ALD cycles. (c) and (d) show Raman maps of the A1g peak intensity for both sides. The lower row in (d) indicates the A1g peak position. The numbers above each figure in (c,d) denote the number of ALD cycles. All maps cover an area of 40 μm × 40 μm2. On the CZTS side, the A1g peak intensity increases with the number of ALD cycles peaking at 21 and then dropping to the background level. This trend suggests that the Al2O3 layer enhances the adhesion of CZTS to the back contact, resulting in more MoS2 on the exfoliated part. In contrast, for a thicker Al2O3 layer, the formation of MoS2 is significantly reduced. Conversely, Raman intensity maps on the MoS2 side reveal that the intensity of the Raman peak decreases to the background level above 21 ALD cycles, which can be interpreted as a reduced layer thickness. Furthermore, the position of the A1g mode, shown in Figure 6d, varies from 407 to 409 cm–1, indicating the presence of bimultilayer MoS2, which is in good agreement with earlier reports.20,21 For 21 ALD cycles and above, the intensity is too low for proper peak fitting; hence, it has been disregarded. The position of the A1g peak varies from 409 cm–1 for 0 ALD cycles to 407 cm–1 for 16 ALD cycles, indicating a decrease in thickness from bulk down to a few layers MoS2, consistent with the change in peak intensity. It is also worth mentioning that the MoS2 layer is likely vertically aligned onto the substrate, in contrast to the horizontal van der Waals structures obtained in a conventional chemical vapor deposition process.20 From these Raman mapping series, it can be concluded that Al2O3 effectively reduces the formation of a MoS2 layer between the Mo back contact and CZTS. The data indicate a clear correlation that as the thickness of the Al2O3 layer increases, the thickness of the MoS2 layer decreases. These findings agree with previous research by Liu et al.14 Lastly, Raman spectra at the excitation wavelength of 325 nm reveal the presence of ZnS at the back contact, regardless of the thickness of the Al2O3 layer (see Supporting Information Note 5). Conclusion This study highlights the critical role of the Al2O3 layer at the back interface of CZTS solar cells in enhancing device performance and structural integrity. The Al2O3 passivation layer significantly improved the interface between Mo and CZTS, enhancing adhesion and preventing delamination. An optimal thickness of the Al2O3 layer of approximately 1.7 nm was identified, resulting in over a 30% improvement in efficiency compared to reference devices without Al2O3. The power conversion efficiency of the cells remained similar for Al2O3 thicknesses between 0.9 and 1.7 nm, suggesting tolerance to minor thickness variations. The Al2O3 layer not only improved the interface between Mo and CZTS but also enhanced the adhesion of the CZTS absorber, which mitigated delamination issues observed in reference devices. The best-performing solar cell with a 1.5 nm Al2O3 layer exhibited enhanced Voc and Jsc, indicating reduced interface recombination and enhanced p–n junction quality. The overall efficiency increased from 3.8% in the reference devices to 5.5% in the optimally passivated devices. In summary, incorporating an Al2O3 layer at the back contact of CZTS solar cells significantly enhances device performance by improving interface quality, adhesion, and efficiency. Optimal thickness management is crucial to balance the benefits against potential drawbacks, such as increased series resistance and reduced shunt resistance. Future research should focus on refining the ALD process to enhance CZTS solar cell efficiency further.