Back Contact Engineering to Improve CZTSSe Solar Cell Performance by Inserting MoO₃ Sacrificial Nanolayers

Abstract

Earth-abundant Cu₂ZnSn(S,Se)₄ (CZTSSe) is a promising nontoxic alternative compound for commercially available Cu(In,Ga)(S,Se)₂ thin-film solar cells. In this study, a MoO₃ nanolayer was applied as a sacrificial layer to optimize the quality of the interface between the CZTSSe and Mo back contact. MoO₃ nanolayers can greatly improve CZTSSe grain growth and suppress the formation of some harmful secondary phases, especially the undesirable MoS(e)₂. In terms of device performance, the series resistance was reduced from 1.83 to 1.54 Ω·cm², and the fill factor was significantly enhanced from 42.67% to 52.12%. Additionally, MoO₃ nanolayers improved CZTSSe absorber quality by lowering the defect energy levels from 228 to 148 meV. Furthermore, first-principles calculations demonstrate that the partial sulfoselenized MoO₃ nanolayers may function as the (p-type) hole-selective contacts at Mo/CZTSSe interfaces, leading to an overall improvement in device performance. Lastly, a CZTSSe solar cell with about 26% improvement (compared with reference cells) in power conversion efficiency was achieved by inserting 5 nm MoO₃ sacrificial layers.

1. Introduction

Thin-film chalcogenide-based solar cells such as CdTe [1] and Cu(In,Ga)(S,Se)₂ (CIGSSe) [2,3,4] are commercially available and demonstrate over 20% power conversion efficiency (PCE). However, the toxicity of cadmium (Cd), and the scarcity of indium (In) and tellurium (Te) are major concerns for their applications [5,6]. Meanwhile, Earth-abundant and nontoxic CZTSSe, which shows a tunable direct bandgap of 1.1–1.5 eV by varying the S/Se ratio and high absorption coefficient of over 10⁴ cm⁻¹ [7,8,9], is treated as a promising alternative for CdTe and CIGSSe [10,11]. However, the open-circuit voltage (V_OC) deficit and low fill factor (FF) critically obstruct PCE in CZTSSe solar cells, with the highest efficiency reaching 12.6% [12], which is significantly lower than the established chalcogenide absorbers [4]. On the basis of the existing configuration of CIGSSe solar cells, molybdenum (Mo) is also used as the back contact for CZTSSe. Nevertheless, Mo induces the detrimental decomposition of CZTSSe and the formation of performance-harmful binary secondary phases, such as Cu_xS(e)_y, Sn_xS(e)_y and Zn_xS(e)_y, leading to some voids near the back sides, and MoS(e)₂ [13]. Although MoS(e)₂ improves the adhesion and quasiohmic contact if it is thin enough [14,15,16], thick MoS(e)₂ is more likely to be formed due to a required high sulfur (S) and/or selenium (Se) pressure [17] during the formation of a CZTSSe absorber. Moreover, CZTS-based absorbers are generally chalcogen-poor, and S/Se vacancies are classified as deep-level defects [18] leading to high series resistance (Rs), thus degrading the overall solar cell performance, especially the V_OC deficit and low FF [17].

In order to suppress thick MoS(e)₂ formation, several different intermediate layers were introduced between the CZTSSe absorber and Mo back contact, such as TiN [17,19], TiB₂ [20], carbon [21], and ZnO [22]. In this regard, introducing a very thin MoO_x at the interface between CZTSSe and Mo back contact can be advantageous because it may modify the surface energy of the Mo substrate and thus further enhance CZTSSe grain growth. From a band alignment point of view, oxidized MoS₂ (i.e., MoS_2-xO_x) layers with p-type conductivity are also preferred [23,24,25,26,27]. N-type MoS₂ layers with low work function are not suitable for the hole-transporting contact with p-type CZTSSe because it may form another p–n junction. In solar cell devices with nonohmic contacts, the V_OC is limited by the work function difference between the top and bottom electrodes instead of the quasi-Fermi level splitting of the absorbers [10,28]. Therefore, MoO_x can be a promising sacrificial material because not only the formation of a conductive p-type MoS(e)_2-xO_x layer is beneficial, but also the voids and unwanted secondary phases can be inhibited by avoiding direct reaction between CZTSSe and Mo.

In this work, we utilized a thermally evaporated ultrathin MoO₃ nanolayer as a sacrificial layer to optimize the back contact interface between CZTSSe and Mo. Cross-sectional transmission electron macroscopy (TEM) combined with energy-dispersive X-ray spectroscopy (EDS) results confirmed that the sacrificial nanolayers increase grain growth (from ~800 to ~1000 nm) and also greatly suppress the formation of MoS(e)₂ layer (from ~350 to ~100 nm) and other harmful secondary phases. As a result, R_s was reduced from 1.83 to 1.54 Ω·cm², and FF was significantly enhanced from 42.67% to 52.12%. Additionally, a MoO₃ nanolayer improved the CZTSSe absorber quality by lowering the defect energy levels from 228 to 148 meV, which was determined by admittance spectroscopy (AS). To fundamentally investigate the role of MoO₃ nanolayers at the back contact of CZTSSe solar cells, a first-principles study was conducted using the Vienna Ab initio Simulation Package (VASP) [29,30,31]. Calculations reveal that the partial sulfoselenized MoO₃ nanolayers may be treated as a hole-selective contact between Mo/CZTSSe interfaces, and is beneficial for the device performance. Lastly, we demonstrate that introducing an ultrathin MoO₃ sacrificial nanolayer remarkably improved the CZTSSe absorber’s quality, and optimized the back contact interface between CZTSSe and Mo, leading to the enhancement of overall solar cell performance.

2. Experimental Details

Multimetallic stacked Sn/Zn/Cu (CZT) precursor film of 600 nm thick was deposited onto a 1 μm Mo-coated (with/without ultrathin MoO_x nanolayers) soda-lime glass (SLG) substrate using the RF magnetron sputtering process. To synthesize CZTSSe absorbers, the CZT precursor films were annealed in a semisealed graphite box (7 cm³ in volume) with Se pellets at a temperature of 550 °C for 7 mins in 20% H₂S (99.95%) diluted with Ar (99.99%) at atmospheric pressure (Cu/Sn + Zn ~0.9 and Zn/Sn ~1.14 measured by X-ray fluorescence (XRF), XRF-1800, Shimadzu Scientific Instrument). The p–n heterojunction was formed by depositing 40 nm CdS buffer layer on top of CZTSSe film using chemical bath deposition method. A 300 nm DC-sputtered ITO/40 nm ZnO contact was then directly deposited onto CdS/CZTSSe films. Lastly, a 300 nm Ag top metal contact and 100 nm MgF₂ antireflection coating were deposited on top of all devices by thermal evaporation. The device area (~0.105 cm²) was defined by shadow masks and mechanical scribing, resulting in a cell effective area of ~0.095 cm² [10,11]. The current density–voltage (J–V) characteristics of the solar cell devices were measured using a Keithley 2400 source meter under the illumination of an AM 1.5G Solar Simulator (Newport) with light intensity of 100 mW/cm², which was calibrated with an NREL-traceable Si reference cell. The spectral response and the external quantum efficiency (EQE) of cells were obtained using an ENLI Technology model EQE-D-3011 system. Temperature-dependent (120–300 K) capacitance–frequency (C–F) measurements were conducted by an Agilent E4980A LCR meter in the frequency range of 10²–10⁶ Hz at 0 V with an AC amplitude of 30 mV under dark conditions [32].

3. Results and Discussion

To confirm the successful deposition of an ultrathin MoO_x film on the Mo-coated SLGs, X-ray photoelectron spectroscopy (XPS) measurements were performed, as shown in Figure 1. For the pristine Mo-coated SLGs (red line), the peaks of Mo₃d_5/2 and Mo₃d_3/2 were located at 228.3 and 231.5 eV, respectively, with a 3.2 eV energy difference, consistent with the literature values [33], indicating that there was almost no native metal oxide on the pristine Mo-coated SLGs. Once some MoO_x had been thermally evaporated on the Mo-SLG substrates, except for the residual small peaks located at 228.3 eV (Mo₃d_5/2 for pristine Mo), these XPS peaks were shifted to 232.9 eV (Mo₃d_5/2 for MoO₃) and 236.1 eV (Mo₃d_3/2 for MoO₃), respectively, with 3.2 eV energy difference, demonstrating that 5 and 10 nm ultrathin MoO₃ (not MoO₂) nanolayers were successfully deposited on the Mo substrates. Moreover, the ultrathin MoO₃ layers can be prepared by the reactive sputtering of 1 μm thick Mo on SLGs, which is compatible with the subsequent fabrication for CIGSSe solar cells [34].

The 600 nm thick Sn/Zn/Cu (CZT) multimetallic stacked precursors were sputtering-deposited on our Mo-SLGs with or without MoO₃ layers [11,35]. The X-ray diffraction (XRD) of metal stacked precursors on Mo-SLGs (in Figure 2) showed two characteristic peaks of well-intermixed metal alloys at 30.26°, corresponding to bronze (Cu₆Sn₅ JCPDS no. 45–1488), and 43.26°, belonging to brass (Cu₅Zn₈ JCPDS no. 25–1288). Along with these two phases, the characteristic peaks of Sn were also found at 30.71°, 32.10°, 43.95°, and 44.97° (JCPDS No. 650296). The well-intermixed precursors are beneficial to the CZTSSe growth during sulfoselenization [11,35].

Raman spectroscopy with 473 nm wavelength excitation (around ~550 nm sampling depth) was conducted to examine the quality of CZTSSe films, as displayed in Figure 3. The signatures at 176, 207, and 239 cm⁻¹ correspond to the CZTSe vibration modes, and the peak at 329 cm⁻¹ corresponds to the CZTS vibration mode [11,35,36]. No peak shifts or secondary phases were observed between the pristine and 5 or 10 nm MoO₃-modified CZTSSe films. The progressive peak sharpening suggests the better quality of CZTSSe films grown on MoO₃ modified substrates compared with the pristine one. The Raman spectra can also be utilized to determine the sulfur and selenium (S/Se) ratio in the CZTSSe by taking the ratio of the peak shift for the sulfide based of the pure CZTS (338 cm⁻¹) and selenide based on the pure CZTSe (196 cm⁻¹) [9]. By calculating the ratio of the peak shift for sulfide-based (329 cm⁻¹) and selenide-based (207 cm⁻¹), an [S]/[S] + [Se] value of ~0.45 was obtained, leading to the S/Se ratio of 45:55.

To investigate the MoO₃ effects on solar cell performances, the light J–V was measured on all the devices, as shown in Figure 4. For statistical study, there were 10 device cells for each condition. The average device performance and standard deviations (±) for different device conditions are summarized in

Table 1. Solar cell parameters of CZTSSe devices with different thickness of MoO₃.

	PCE (%)	JSC (mA/cm2)	VOC (mV)	FF (%)	RS (Ω·cm2)	RSH (Ω·cm2)
Pristine	5.21 ± 0.66	25.86 ± 2.26	450 ± 15	42.67 ± 1.5	1.83 ± 0.44	133 ± 45
5 nm MoO3	6.80 ± 0.85	27.33 ± 2.20	473 ± 5	52.12 ± 3.0	1.54 ± 0.23	153 ± 68
10 nm MoO3	6.49 ± 0.55	26.30 ± 1.57	474 ± 7	50.69 ± 1.5	1.77 ± 0.26	252 ± 63
Best cell (5 nm MoO3)	7.78	28.84	480	55.83	1.68	180

Statistics from 10 cells for each condition.

. All cells with MoO₃ nanolayers exhibited higher short circuit current (J_SC), V_OC, and FF compared to the reference cells. The significant improvements in Voc from 450 to 473 mV, J_SC from 25.86 to 27.33 mA/cm², and FF from 42.67% to 52.12% for devices inserted with a 5 nm MoO₃ layer were attributed to the increased shunt resistance (R_SH) from 145 to 151 Ω·cm² and the reduced R_S from 1.83 to 1.54 Ω·cm², which may have resulted from the improvement in back contact quality. Lastly, a CZTSSe solar cell with 7.78% PCE was obtained by inserting 5 nm MoO₃ layers, which was an improvement of about 26% compared with 5.76% PCE of the pristine one.

To inspect the CZTSSe absorber quality near the Mo back contacts, cross-sectional TEM and EDS mapping was executed, as presented in Figure 5 and Figure 6, respectively. MoS(e)₂ thickness was reduced from ~350 to ~100 nm, and CZTSSe grain sizes were enlarged by inserting 5 nm MoO₃ layers (Figure 5). In the pristine absorber, the void formation near the back contacts may have arisen from the vaporization of highly volatile secondary phases (i.e., Sn_xS and Sn_xSe_y) during the focused ion beam (FIB) fabrication of TEM samples. A large number of voids from secondary phases forming at the back interface limited the free-carrier transportation, leading to higher R_S and thereby lower J_SC [21]. EDS mapping (Figure 6) also directly illustrates the formation of unwanted secondary phases, such as Cu_xS_y, Zn_xS_y, Sn_xS_y, and Sn_xSe_y, near the untreated Mo back contacts, which further proves that the CZTSSe/Mo interface was chemically unstable [19]. However, the inserted MoO₃ nanolayers could not be identified by the oxygen/molybdenum elemental mapping images in Figure 6b. MoO₃ nanolayers may have been fully sulfoselenized to p-type MoS(e)_2-xO_x during a high-temperature synthesis process. Therefore, the increased V_OC and FF, and reduced R_S in the modified devices may have been due to the larger grains of the CZTSSe absorber (i.e., fewer grain boundary recombination centers), the suppression of some harmful secondary phases, and the thinner MoS(e)₂ by inserting MoO₃ sacrificial nanolayers.

To understand why the device performance was not monotonically improved as MoO₃ thickness increased, cross-sectional TEM analysis was conducted, as shown in Figure 7. Inserting 10 nm MoO₃ increased the Rs, and reduced the J_SC and FF, thus decreasing the overall PCE (Table 1). The TEM results (Figure 7) show that the orientation of MoS(e)₂ can be vertically or horizontally aligned with the Mo substrate, which can provide different contact resistances. On the basis of the direction of the photocarrier separation (i.e., built-in electric field direction), the vertically aligned MoS(e)₂ was more conductive than the horizontally aligned one is. In fact, MoS(e)₂ is a layered material with a weak van der Waals bond between each layer; thus, its interlayer (i.e., horizontally aligned) conductivity is two orders of magnitude lower than its intralayer (i.e., vertically aligned) conductivity (Figure 7) [37,38,39]. The vertically aligned MoS(e)₂ came from the pristine Mo contact that reacted with S/Se vapor during the annealing process, while the horizontally aligned MoS(e)₂ came from the MoO₃ layers that lost their oxygen during the sulfoselenization process due to the high vapor pressure of S/Se [40]. Therefore, it is preferable that the MoO₃ layers be transformed into horizontally aligned MoS(e)₂ (or MoS(e)_2−xO_x) because it can hinder S/Se diffusion during annealing process [40]. Hence, undesirable MoS(e)₂ thickness can be minimized. This shows that the higher Rs of 1.77 Ω·cm² for 10 nm MoO₃ devices compared with 1.54 Ω·cm² for 5 nm MoO₃ devices was due to a thicker MoO₃ that transformed into a more horizontally aligned MoS(e)₂. As a result, J_SC was reduced from 27.33 to 26.30 mA/cm² by inserting the thicker MoO_3, as shown in Table 1.

To further study the effect of the MoO₃ sacrificial layer on defect states in CZTSSe, AS measurement was performed [32]. Figure 8 reveals the C–F profiling for CZTSSe devices with MoO₃ thickness of 0, 5, and 10 nm. Using the model in the work of Kimerling [41,42], capacitance at a high frequency represents the response of the free carrier density, while capacitance at a low frequency represents the response of the sum of free carriers and deep traps. The Arrhenius plot (Figure 9) was used to determine the activation energy of deep trap states with various MoO₃ thickness levels. For each AS spectrum, the inflection point frequency (or step frequency ${\omega }_0$) is determined by applying angular frequency at the maximum of the $\omega$dC/d$\omega$ plot. The inflection point frequency of each AS spectrum can be used to construct the Arrhenius plot by fitting the spectra according to the following equation:

$${\omega }_0=2\pi {\nu }_0{T}^{2}exp\left(\frac{E_a}{kT}\right)$$

where ${\omega }_0$ is the step frequency, $E_a$ is the energetic depth of the defect relative to the corresponding band edge, ${\nu }_0$ is the pre-exponential factor comprising temperature-independent parts such as defect capture cross-sections for holes σ_p, thermal velocity ${\nu }_{th}$, and the effective density of states in the valence band N_v.

Activation energy E_a ascertained with the Arrhenius plot was approximately the energy difference between the defect level and the valence band edge. Assuming that their energy levels were within a small range, the E_a value deduced from the Arrhenius plot could also represent the average value of activation energies of a band of defects in the band gap. The activation energies of CZTSSe devices with MoO₃ thicknesses of 0, 5, and 10 nm were 228, 148, and 199 meV, respectively (Figure 9). Both results support the argument that, besides suppressing MoS(e)₂ and secondary formation at the back interface, introducing MoO₃ as a sacrificial nanolayer may help in improving CZTSSe absorber quality by not only increasing the grain growth, but also lowering the activation energy of defect states.

To fundamentally investigate the role of MoO₃ nanolayers at the back contacts of CZTSSe solar cells, first-principles calculations were conducted through VASP [29,30,31] to examine the electronic properties of S-doped MoO₃. The projector-augmented wave (PAW) method was adopted, and Perdew–Burke–Ernzerhof (PBE) exchange correlation functionals were used to account for electron–electron interactions in our system [43]. Monkhorst–Pack K-point grids of 2 × 2 × 2 and 400 eV energy cutoff for plane wave basis set were used for geometric optimization. The geometric structures were entirely relaxed when the total energy was converged to 10⁻⁶ eV in the 3 × 2 MoO₃ supercell model. The density-of-states (DOS) calculation of S-doped MoO₃ was performed using the same parameters except for the 6 × 6 × 6 Monkhorst–Pack K-point grids. Due to sulfidation, sulfur appeared in MoO₂, and this could be regarded as a sulfur substitution for oxygen in which three different sites substitutive defects could be formed, denoted as S_O1, S_O2, and S_O3. S_O1 is single oxygen coordinated, while S_O2 and S_O3 are double and triple oxygen coordinated, respectively, as shown in Figure 10.

Figure 11 reveals the DOS of MoO₃ with S_O defects, which indicates that there were additional defective states (left-hand side near 0 eV) with slightly higher energy than that of the valence band maximum (VBM). There was no significant difference between sulfurization and selenization for MoO₃. The small difference was that the additional defective states of selenization were a little serious compared with the sulfurization one. The calculated DOS was almost the same for these three different defect sites, which means that the DOS was not sensitive to the chemical environment of S_O, and only one is shown. These defective S_O states above the VBM, which contribute to the p-type carriers (holes) within the partial sulfoselenized MoO₃ nanolayers, may be treated as hole-selective contacts between Mo/CZTSSe interfaces, further improving the overall PCE, as demonstrated experimentally.

4. Conclusions

While CZTSSe is a very promising alternative for CdTe and CIGSSe for thin-film solar cells, its V_OC deficit and low FF critically obstruct the PCE. The formation of many harmful secondary phases at the back interface between CZTSSe and Mo leads to high series resistance (Rs), thus degrading overall solar cell performance. In this study, a MoO₃ nanolayer was applied as a sacrificial layer to optimize the back contact interface between CZTSSe and Mo. Cross-sectional TEM and EDS results show that the MoO₃ sacrificial nanolayers not only improved CZTSSe grain growth, but also successfully suppressed the formation of an undesirable MoS(e)₂ layer (from ~350 to ~100 nm) and other harmful secondary phases. From the device point of view, the Rs was reduced from 1.83 to 1.54 Ω·cm² and the FF was significantly increased from 42.67% to 52.12%, resulting in an increase in Voc by 23 mV. Importantly, MoO₃ nanolayers also improved CZTSSe absorber quality by lowering the defect energy levels from 228 to 148 meV. In addition to the experimental results, the first-principles calculations revealed that the partial sulfoselenized MoO₃ nanolayers may be treated as the (p-type) hole-selective contacts between Mo/CZTSSe interfaces, contributing to an overall improvement in device performance. Lastly, a CZTSSe solar cell with 7.78% PCE was obtained by inserting 5 nm MoO₃ nanolayers, which was an improvement of about 26% compared with the 5.76% PCE of the pristine one.