Study and optimization of alternative MBE‐deposited metallic precursors for highly efficient kesterite CZTSe:Ge solar cells

Nowadays, most of the best efficiencies of Cu2ZnSn(S,Se)4 (CZTSSe) solar cells are obtained from absorber layers fabricated using sequential processes, including the deposition of metallic stack precursors, typically by sputtering, and followed by reactive annealing under chalcogen atmosphere. The sputtering technique is widely known for the easy growth of metallic layers, although the deposition rates, growth morphology and nucleation, or the roughness can sometimes be an issue leading to inhomogeneities in the final layers. Nevertheless, MBE (molecular beam epitaxy) technique could have some advantages to obtain high‐quality metallic layers, with accurate control of the growth due to ultra‐high vacuum conditions and high purity. In this work, we study the use of advanced MBE systems to grow metallic stack precursors, alternatively to sputtering or thermal evaporation techniques, to obtain high‐quality CZTSe:Ge absorbers. Due to differences in the nature of each type of precursor, thermal annealing optimizations are presented by modifying some critical selenization parameters, such as the temperature or the selenium amount in order to obtain well‐crystallized absorbers. Detailed morphological, compositional, and structural characterizations show relevant features of each precursor, mainly related to the formation of MoSe2 at the back interface, and Se and Sn composition after selenization in different conditions. Regarding the solar cell devices, main efficiency limitations come from VOC and FF, which could be tentatively related to a noncontrolled selenization; different precursor reactivity, porosity, or composition; and different alkali diffusion during the reactive annealing. Finally, in the first optimization, a 9.2% efficiency device has been achieved with promising perspectives for future improvements.


| INTRODUCTION
Cu 2 ZnSn(S,Se) 4 (CZTSSe) semiconductor materials have attracted considerable attention in the last years, being one of the most promising thin film photovoltaic absorbers, primarily through the use of earth abundant elements and low toxicity. [1][2][3] This makes kesterites CZTSSe an interesting mid-to long-term alternative to the widely known CuIn 1-x Ga x Se 2 (CIGSe), thus allowing reducing the use of scarce elements like In and Ga. 4 Apart from that, kesterite has several advantageous properties to be a very suitable material for photovoltaic applications: kesterite has p-type conductivity naturally due to intrinsic point defects; it is direct band gap semiconductor with a high absorption coefficient (~10 4 cm −1 ) 5 ; its band gap can be easily tuned with the ratio S/Se, from 1.0 eV, for the pure selenium Cu 2 ZnSnSe 4 (CZTSe) compound, to 1.5 eV, for the pure sulfur Cu 2 ZnSnS 4 (CZTS) 6,7 ; and it is highly compatible with CIGS technology, sharing several processing steps and techniques. Furthermore, the fact that kesterite absorbers can be synthesized with a large variety of techniques is another advantage to consider, especially for future industrial perspectives. [8][9][10][11][12][13][14][15][16][17][18] Regarding the deposition techniques, these are usually classified as vacuum (mostly physical vapor deposition [PVD]-based) and nonvacuum techniques.
Historically, solar cells prepared by chemical-based routes have demonstrated better performances than the obtained by physical-based ones. Nevertheless, in the last few years, as illustrated in Figure 1, there has been a remarkable improvement of devices fabricated through PVD-based approaches like sputtering and coevaporation, which is of great importance from the industrial point of view. In terms of scalability, in general, these vacuum-based pro-   annealing profile consisted in a two-step process; first, 30 minutes at 400°C and 1 mbar (with Ar flux), and second, 15 minutes at different temperatures from 500°C to 575°C and 1000 mbar total Ar pressure.
After that, the samples were naturally cooled down until room temperature. Further details can be found elsewhere. 24,37 In order to complete the solar cells, a chemical etching with (NH 4 ) 2 S was performed to prepare the absorber surface before the growth of the CdS buffer layer (50 nm) by chemical bath deposition

| RESULTS AND DISCUSSION
To compare the metallic precursors, Cu/Sn/Cu/Zn stacks were deposited either by sputtering or MBE system. Figure  The optimization of the thermal annealing processes is a key factor when the properties of the precursors are modified due to different deposition systems. The porosity, the compactness or the roughness are some of the characteristics that can be affected by the deposition technique, and these have to be considered to adjust the reactive annealing processes, since it can affect the final layer morphology, the degree of molybdenum selenization, the presence of undesired secondary phases, etc.
Therefore, different annealing conditions were tested for both types of metallic precursors, by modifying two of the most relevant selenization parameters: (a) the second step annealing temperature (ie, crystallization temperature) and (b) the selenium quantity, which will impact on the selenium partial pressure during the process. Figure 3 shows the SEM analysis of the samples fabricated using different annealing conditions. Here, the effect of the annealing temperature is assessed (from 500°C to 575°C) for each type of precursor, keeping the Se amount constant (100 mg). As one could expect, the increasing crystallization temperature leads to a gradual increase of the grain size for both types of precursors. Nevertheless, larger grains are systematically obtained for the sputtered precursors (see Figure 3 A), and additionally the degree of selenization of the molybdenum back contact is much lower in this case. This likely indicates a higher porosity of the MBE-deposited precursors, allowing a fast diffusion of the Se vapors towards the back contact, which could be controlled to some extent by the Se vapor pressure as will be shown below. A reasonable explanation for the different grain growth could be the different nature of the MBE and sputtered precursors; indeed, compared to the former, sputtered films are typically more compact with reasonably higher compressive stress, which is known to act as a driven force for the grain growth during the annealing treatment. 39,40 As previously commented, the Se amount introduced during the selenization can be directly related to the Se partial pressure during the process. In Figure 4, a better crystallization with bigger grains is Further compositional analysis corroborates some of the previous observations. As can be seen in Figure 5, the Se content in the absorbers from sputtered precursors remains practically constant, regardless of the temperature increase or the Se quantity used in the reactive annealing, while it gradually increases for the MBEdeposited samples as both temperature and Se quantity are increased.
Thus, this confirms the overselenization previously shown by SEM On the other hand, looking at some characteristic compositional ratios (see Figure 5), intriguingly they show that Sn becomes additionally incorporated with the increasing annealing temperature and Se amount, in particular for the MBE-deposited samples. Cu/Sn and Zn/Sn ratios are significantly reduced as both parameters are increased. As it widely known, Sn tends to be adjusted by means of SnSe 2 presence during the reactive annealing, and probably due to the higher reactivity of the MBE-evaporated precursors, the control of this element becomes even more critical than for the sputtering.
To further investigate the effect of this compositional variations on the structural properties, the crystalline quality or the presence of undesired secondary phases, Raman spectroscopy analysis were performed on the different samples (see Figure S1 in the Supporting Information). Interestingly, rather small differences were observed in terms of crystalline quality, defects, or secondary phases (those detectable with green excitation wavelength, 532 nm). Thus, suggesting a very similar material, structurally speaking.
To eventually see the impact on the device performance, all these layers were made into solar cells and measured under a solar simulator. Figure 6    Further optimizations of the precursor composition, finally led to a conversion efficiency of 9.2% (total area, 0.522 cm 2 ) with MgF 2 antireflective coating for MBE-deposited precursors. Figure 9 shows the champion J-V curve and EQE for samples from MBE-evaporated layers, compared with the best sputtering-based cell. This champion sputtering-based solar cell was fabricated in a different batch, using the optimized baseline process, achieving an efficiency greater than 11%, and is shown for comparison. As can be seen, relatively high current densities are achieved, exceeding 39 mA/cm 2 , which are at the same level of the best cells reported so far for the pure selenide CZTSe compound. 21,25 Nonetheless, V OC and especially FF values remain still lower compared with these record devices. As discussed earlier, this might be explained by a less controlled selenization and a nonoptimal alkali content, which can have a great impact on these solar cell parameters.
In the same vein, Figure 9A shows the clear difference between the two champion solar cells, the one synthesized from MBEevaporated precursors and the one from sputtered precursors, and the V OC is the main responsible, leading to slightly higher voltage deficits for the MBE system (about 340 mV vs 325 mV as determined by using the Schockley-Queisser limit). Additionally, Figure 9B shows  ORCID Sergio Giraldo https://orcid.org/0000-0003-4881-5041