Two types of back-contact structures with SiO_xN_y layers were applied: Mo/SiO_xN_y and Mo/SiO_xN_y/Mo. As shown in Fig 1, a standard Mo layer (≈ 550 nm) consisting of two sub-layers (≈ 275 nm for each) was deposited by Ar plasma (power density: 6.1 W/cm², pressure: 2.7×10⁻³ mbar) onto the 1 mm soda lime glass (SLG). SiO_xN_y layers (10, 25 and 40 nm) were deposited from Si sputter target by mixed Ar-N₂-O₂ plasma (power: 160 W, pressure: 2×10⁻³ mbar) onto the standard Mo layers. For Mo/SiO_xN_y/Mo back-contact structure, the top Mo layer (≈ 50 nm) on SiO_xN_y was deposited under the same conditions as for the standard Mo layers. For both back-contact structures, a standard procedure for the fabrication of solar cells in our lab was performed [8, 16, 17]. In this study, a dry cleaning with Ar plasma (power: 100 W, pressure: 5×10⁻³ mbar, duration: 90 s) was performed on the as-grown Mo/SiO_xN_y and Mo/SiO_xN_y/Mo back contacts. For the formation of CZTSe absorber, a metallic precursor with a structure of Zn/Cu-Sn/Zn was deposited onto the above-mentioned back contacts by DC-sputtering at room temperature, followed by the annealing with selenium (Se) pellets and tin (Sn) wires in a tube furnace at 530°C for 20 minutes. As buffer layers, cadmium sulfide (CdS) was deposited onto the as-grown CZTSe absorbers (≈ 1.2 μm) via chemical bath. Furthermore, i-ZnO and Al:ZnO layers were deposited by RF-sputtering as front contacts. Finally, every sample was divided by mechanical scribing into solar cells with an average area of ≈ 0.25 cm². In the following statement, the reference solar cell with a standard 550 nm Mo back contact is denoted by ‘‘M”. The solar cells grown on Mo/SiO_xN_y back contacts are denoted by ‘‘MS10”, ‘‘MS25” and ‘‘MS40”, for which the thicknesses of SiO_xN_y layers are 10, 25 and 40 nm, respectively. ‘‘MM” denotes the reference solar cell with a Mo/Mo back contact, in which a top layer of Mo (≈ 50 nm) is deposited on the standard Mo layer. ‘‘MS10M”, ‘‘MS25M” and ‘‘MS40M” denote solar cells grown on Mo/SiO_xN_y/Mo back contacts, in which the thicknesses of SiO_xN_y layers between the top and the standard Mo layers are 10, 25 and 40 nm, respectively.

A FEI Helios Nanolab 600i scanning electron microscope (SEM) was used for the characterization of cross-section morphology. A Keithley 2400 SMU was adopted for current-voltage (I-V) measurements of CZTSe solar cells under standard AM 1.5 illumination in a PET SS100AAA solar simulator. A Bentham PVE300 system was used for EQE measurements. Measurements of capacitance-frequency (C-f) and thermal admittance spectroscopy (TAS) were performed with a Solartron impedance analyzer SI-1260. The admittance spectra were recorded for a frequency range from 10 Hz to 1 MHz and a temperature range from 50 K to 330 K. The processes of heating and cooling were performed in a closed cycle Helium cryostat at a base pressure < 10⁻⁵ mbar. For precise measurements of the temperature, a thermal sensor was glued on top of a dummy cell placed next to the real sample.

Fig 2 shows SEM cross-section morphology of CZTSe samples grown on Mo and Mo/SiO_xN_y back contacts. The reference sample M, which has no SiO_xN_y diffusion-barrier layer, shows the formation of a ≈ 1 μm MoSe₂ interfacial layer. Samples with SiO_xN_y layers in various thicknesses (i.e., 10, 25 and 40 nm) show a significantly suppressed formation of MoSe₂ layers. The specific thicknesses of formed MoSe₂ layers lie in the range of 30 ~ 40 nm with no visible trend. This suggests that SiO_xN_y acts as an effective diffusion barrier for Se and prevents the strong reaction of Mo and Se to form MoSe₂ during a high temperature (≈ 530°C) annealing.

Performance of the CZTSe solar cells grown on Mo and Mo/SiO_xN_y back contacts are shown in Fig 3. Compared to the reference M, which is grown on pure Mo back contact, all the solar cells from Mo/SiO_xN_y back contacts show a deterioration in all parameters, i.e., open-circuit voltage (V_oc), short-circuit current density (J_sc), fill factor (FF) and energy conversion efficiency (η). The reasons for the observed deterioration of solar cell properties may include a possible formation of an extra potential barrier induced by SiO_xN_y layers. Furthermore, it cannot be excluded that a small amount of oxygen atoms from SiO_xN_y may diffuse into the CZTSe absorbers and lead to additional impurity states in the absorber, which could impact the defect landscape or influence the phase structure of kesterite material in relevant regions. Moreover, thicker SiO_xN_y diffusion barriers may further cause a harmful influence on the device performance by blocking sodium diffusion from SLG, which is generally considered to enhance the absorber quality in CIGS and kesterite solar cells [18–20]. However, based on our previous research, in which SiO_xN_y was investigated as barrier layers between SLG and Mo, a total blocking effect for sodium was only achieved with much thicker SiO_xN_y layers [21]. That means the blocking effect of SiO_xN_y layers for sodium have most probably only a very minor influence in the present case. As a consequence, the CZTSe solar cells grown on Mo/SiO_xN_y back contacts show overall poor performance, regardless of a possible positive effect expected from the reduced thicknesses of MoSe₂ layers.

Fig 4 shows the EQE of CZTSe solar cells grown on Mo and Mo/SiO_xN_y back contacts. The overall EQE drops for the samples with SiO_xN_y layers compared to that of the reference. Specifically, MS10 and MS25 show a slight drop while MS40 shows a strong one. This result matches the performance shown in the previous I-V measurements. In particular for the sample MS40, the strong drop in EQE and in the extracted short-circuit current (S1 Fig) suggest not only an increased series resistance, for which the EQE drop is wavelength independent, but also a potential barrier introduced by the thick SiO_xN_y layer and/or a deterioration in absorber quality.

Fig 5 shows results from C-f measurements of CZTSe solar cells grown on Mo and Mo/SiO_xN_y back contacts. According to literature about similar material-systems, inflection points related to shallow defects can be observed in capacitance measurements between 10 Hz and 100 kHz at temperatures between 50 K and 200 K [22–25]. In our case, samples M and MS10 do not show clear inflection points. Samples MS25 and MS40, which are grown on back contacts with thicker SiO_xN_y layers, show clear inflection points. This may suggest a higher defect-density for CZTSe solar cells grown on Mo/SiO_xN_y back contacts with the thicker SiO_xN_y layers. Specific distributions of defects are investigated by TAS measurements and shown in the following part.

Density of states (DOS) derived from TAS measurements on CZTSe solar cells M and MS40 are shown in Fig 6. The results of sample M, as shown in Fig 6A, reveal a small peak at ≈ 0.13 eV and a large peak at ≈ 0.18 eV. For sample MS40, as shown in Fig 6B, both peaks are broadened in comparison to the previous case and their maxima are shifted to ≈ 0.10 eV and ≈ 0.20 eV, respectively. According to literature based on Cu₂ZnSn(S,Se)₄, both peaks may be linked to bulk defects [22, 25]. The significant broadening of the deeper defect is visibly accompanied by an enhanced density of states at deeper levels, which could cause higher recombination rates and thus a deterioration of solar cell properties.

Cross-section morphology of CZTSe solar cells grown on Mo/Mo and Mo/SiO_xN_y/Mo back contacts is shown Fig 7. In comparison to the case of sample M, the thicknesses of interfacial MoSe₂ layers of MS10M, MS25M and MS40M are reduced to a range of 230 ~ 240 nm with no obvious trend for the thicknesses of the investigated SiO_xN_y layers. It indicates that only the top Mo layers (≈ 50 nm) in Mo/SiO_xN_y/Mo structures contribute to the formation of MoSe₂ layers during the annealing and the Mo layers (≈ 550 nm) underneath SiO_xN_y barriers remain intact. Surprisingly, for the reference sample MM, in which no barrier is applied, the thickness of interfacial MoSe₂ stays also in a similar range as for the samples grown on Mo/SiO_xN_y/Mo back contacts. This indicates a barrier-like behavior at the Mo/Mo interface, which could be related to a natural passivation due to the process-break in Mo fabrication or a blocking effect due to crystal discontinuity in this layered structure.

Fig 8 shows parameters of CZTSe solar cells grown on Mo/Mo and Mo/SiO_xN_y/Mo back contacts. For the sample MS10M, all the device parameters (V_oc, J_sc, FF and ŋ) stay in a similar or slightly improved range in comparison to the reference MM. In cases of thicker SiO_xN_y layers (MS25M and MS40M), the device performance deteriorates. However, the deterioration here may differ from the samples grown on Mo/SiO_xN_y back contacts. In cases of Mo/SiO_xN_y back contacts, the deterioration could possibly be a combined result from the negative influence of SiO_xN_y at the back interface and the deteriorated CZTSe absorber quality. For Mo/SiO_xN_y/Mo back contacts, a thin SiO_xN_y layer may have a positive effect due to its role as passivation layer in between the formed porous MoSe₂ and the intact bottom Mo, which could compensate the disadvantages such as causing an electrical barrier or an extra series resistance. As a result, for MS10M, without a noticeable drawback in FF, a similar V_oc and a small increase in J_sc lead to an improved ŋ at around 11% in comparison to the case of MM. And compared to M, MS10M gives a similar ŋ with a higher homogeneity in the morphology at the back interface. However, for the cases of MS25M and MS40M, in which the SiO_xN_y layers are thicker, the possible negative effects of SiO_xN_y on the back interface and the CZTSe absorber quality dominate and cause a significant drop in FF thus the ŋ of the whole devices.

Fig 9 shows EQE of CZTSe solar cells grown on Mo/Mo and Mo/SiO_xN_y/Mo back contacts. Compared to that of the previous samples grown on Mo/SiO_xN_y back contacts, the change in thicknesses of the added SiO_xN_y layers has a much smaller impact on the EQE, which shows no clear trend. This indicates that the negative effects of a thick SiO_xN_y layer are significantly reduced by applying a 50 nm Mo layer on top. EQE of all devices reach ≈ 90% at around 600 nm. The extracted values of J_sc remain constant at around 35 mA/cm² (S2 Fig).

Fig 10 shows C-f measurements of CZTSe solar cells grown on Mo/Mo and Mo/SiO_xN_y/Mo back contacts. Compared to that of the samples without the 50 nm Mo top layers, no clear inflection point is observed for all the samples grown on Mo/SiO_xN_y/Mo back contacts. This result is consistent with the above-discussed I-V and EQE measurements, suggesting that the additional Mo layers on top of the SiO_xN_y can suppress or compensate the negative impacts from SiO_xN_y layers. The presence of MoSe₂ may have beneficial effects on the overall interface quality due to a more favorable alignment of the work functions and hence the bands of the adjacent layers.

Fig 11 shows the DOS derived from TAS measurements of the CZTSe solar cells MM and MS40M. In comparison with Fig 6, the overall influence of SiO_xN_y on DOS is much smaller. In details, compared to the case of MM, the small peak at ≈ 0.12 eV is reduced and the large peak at ≈ 0.18 eV is slightly broader for MS40M. If the shallow levels at ≈ 0.12 eV is related to the interface defects, this result could indicate that SiO_xN_y is beneficial for suppressing interface defects. However, the slight broadening of the deep levels at ≈ 0.18 eV still may indicate a negative influence of SiO_xN_y on raising CZTSe bulk defects. As a result, the combined outcome from the changes at back interface and in absorber bulk is reflected in the device performance.