A. Cation disordering in CZTSSe solar cells

Experimental evidence on the facile formation of Cu–Zn antisites was provided by Schorr et al.^{Reference Schorr, Hoebler and Tovar29} using neutron powder diffraction because conventional X-ray diffraction cannot distinguish cations with similar atomic factors. They observed a kesterite structure of CZTS with Zn and Cu occupying the z = 1/4 and z = 3/4 in a disordered manner, leading to the formation of Cu_Zn and Zn_Cu antisites. Similar observations have been reported by Washio et al.,^{Reference Washio, Nozaki, Fukano, Motohiro, Jimbo and Katagiri30} Lafond et al.,^{Reference Lafond, Choubrac, Guillot-Deudon, Fertey, Evain and Jobic31} and Stone et al.^{Reference Stone, Christensen, Harvey, Teeter, Repins and Toney32} using the resonant X-ray diffraction technique, which measures diffraction intensity as a function of X-ray energy across an absorption edge. Since each element has different resonant energies and the scattering power of the resonant atom changes near the absorption edge, the resonant X-ray diffraction technique can effectively distinguish elements with similar atomic numbers (e.g., Cu and Zn).^{Reference Stone, Christensen, Harvey, Teeter, Repins and Toney32} Washio et al. prepared CZTS thin films with different compositions and observed the Cu substitution for Zn at the 2d site and Zn substitution for Cu at the 2c site at a photovoltaic compositional ratio (Cu/(Zn + Sn) = 0.8) using synchrotron radiation X-ray diffraction.^{Reference Washio, Nozaki, Fukano, Motohiro, Jimbo and Katagiri30} Stone et al. further validated the existence of high concentrations of Cu_Zn and Zn_Cu antisites on the 2c and 2d positions as well (Fig. 4).^{Reference Stone, Christensen, Harvey, Teeter, Repins and Toney32}

FIG. 4.

Structure for kesterite (a), disordered kesterite (b), and stannite (c). Reproduced with permission from Ref. Reference Lafond, Choubrac, Guillot-Deudon, Fertey, Evain and Jobic31.

To probe the cationic disorder in CZTSSe, spectroscopic studies, such as photoluminescence (PL)^{Reference Gokmen, Gunawan, Todorov and Mitzi17,Reference Gershon, Shin, Bojarczuk, Gokmen, Lu and Guha33,Reference Grossberg, Krustok, Raadik, Kauk-Kuusik and Raudoja34} and nuclear magnetic resonance (NMR),^{Reference Choubrac, Paris, Lafond, Guillot-Deudon, Rocquefelte and Jobic35,Reference Paris, Larramona, Bais, Bourdais, Lafond, Choné, Guillot-Deudon, Delatouche, Moisan and Dennler36} were also carried out. The results showed a rather broader width of the spectroscopic peak as evidence of the Cu–Zn disorder. Using ¹¹⁹Sn MAS NMR spectroscopy, Paris et al.^{Reference Paris, Larramona, Bais, Bourdais, Lafond, Choné, Guillot-Deudon, Delatouche, Moisan and Dennler36} demonstrated that A-type [V_Cu + Zn_Cu] defect complexes exist under Cu-poor and Zn-rich conditions, which is consistent with the previous theoretical studies.^{Reference Choubrac, Lafond, Guillot-Deudon, Moëlo and Jobic23,Reference Lafond, Choubrac, Guillot‐Deudon, Deniard and Jobic24} Such a high population of Cu–Zn antisites is expected to cause electrostatic potential fluctuation, leading to band tailing as reported by Gokmen et al.^{Reference Gokmen, Gunawan, Todorov and Mitzi17} At a low temperature, charge carriers become localized in a potential well due to insufficient thermal energy. Therefore, separation of carriers localized in the well makes the recombination rather difficult and results in a longer lifetime compared to that of CIGS. Moreover, high concentrations of defects give rise to a broader PL width, indicating that the recombination is facilitated by the tail states.^{Reference Gokmen, Gunawan, Todorov and Mitzi17,Reference Gershon, Shin, Bojarczuk, Gokmen, Lu and Guha33}

However, the abovementioned spectroscopic and structural studies on the Cu–Zn disorder area lacking in spatial resolution and therefore give information from a rather large volume of interests. To understand point defects and related defect clusters at the atomic scale, researchers have performed scanning transmission electron microscopy (STEM) studies. Mendis et al.^{Reference Mendis, Shannon, Goodman, Major, Claridge, Halliday and Durose37} used an aberration corrected STEM along with electron energy loss spectroscopy (EELS) to analyze three Cu/Zn atom columns between Sn atoms along the [001] c-axis. As shown in Figs. 5(a) and 5(b), atom columns were clearly rich in Zn (∼1020 eV) but poor in Cu (∼930 eV), indicating the presence of a high density of Zn_Cu antisite atoms. When scanning a region over a 10 × 10 nm² area, a strong Cu L-edge was visible in the EELS spectra with a Cu/Zn ratio of 0.4 [Fig. 5(b)]. Consequently, Zn_Cu antisites were observed with the nanoscale composition inhomogeneity of a lateral size between 1.5 and 5 nm. Aguiar et al.^{Reference Aguiar, Erkan, Pruzan, Nagaoka, Yoshino, Moutinho, Al‐Jassim and Scarpulla38} also used STEM to probe the partial Cu–Zn disorder and the resulting cation ratio fluctuations within CZTSSe. As can be seen in Figs. 5(c) and 5(d), STEM-based energy dispersive X-ray spectroscopy (EDS) revealed mixed colors of red and blue at some points (white dashed lines) while distinct color separation was observed in the other maps of different cation ratios ([Zn/Sn & Cu/Sn] and [Cu/Zn & Cu/Sn]). The results suggested the existence of partially disordered regions at the nanometer scale (∼20 nm) where Cu influences the presence of Zn and Sn. Using atomic-resolution TEM, Kattan et al.^{Reference Kattan, Griffiths, Cherns and Fermin39} and Mendis et al.^{Reference Mendis, McKenna, Gurieva, Rumsey and Schorr40} reported the antisite domain boundaries (ADBs) in pure sulfide and pure selenide, respectively. These boundaries are typically 5–20 nm apart and have displacements of 1/2[110] or 1/4[201] depending on the cationic substitution. In other words, there are Zn_Cu and Cu_Zn antisite defects on the 2c and 2d sites in the bottom half for 1/2[110] (001) ADBs, whereas Sn_Cu antisite defects are present on the 2c site in the top half for 1/4[201] (101) ADBs.^{Reference Kattan, Griffiths, Cherns and Fermin39}

FIG. 5.

(a) STEM high angle annular dark field image of [010]-oriented Cu₂ZnSnS₄.^{Reference Mendis, Shannon, Goodman, Major, Claridge, Halliday and Durose37} (b) EELS spectra for atom 1, atom 2, and atom 3.^{Reference Mendis, Shannon, Goodman, Major, Claridge, Halliday and Durose37} (c) STEM-based EDS image at a grain boundary of Cu₂ZnSnS₄.^{Reference Aguiar, Erkan, Pruzan, Nagaoka, Yoshino, Moutinho, Al‐Jassim and Scarpulla38} (d) STEM-based EDS image far away from a grain boundary of Cu₂ZnSnS₄.^{Reference Aguiar, Erkan, Pruzan, Nagaoka, Yoshino, Moutinho, Al‐Jassim and Scarpulla38} Reproduced with permission from Refs. Reference Mendis, Shannon, Goodman, Major, Claridge, Halliday and Durose37 and Reference Aguiar, Erkan, Pruzan, Nagaoka, Yoshino, Moutinho, Al‐Jassim and Scarpulla38.

B. Secondary phases in CZTSSe solar cells

A theoretical study by Chen et al.^{Reference Chen, Gong, Walsh and Wei18–Reference Xu, Chen, Huang, Xiang, Gong and Wei20} revealed that the stable range of chemical potentials of CZTSSe against secondary phases [Fig. 2(b)] is very narrow. Such a limit in the phase inhomogeneity originates from the easy formation of Zn(S,Se) or Cu₂Sn(S,Se)₃ because the chemical potential of Zn also lies in a narrow region.^{Reference Chen, Walsh, Gong and Wei19} Moreover, considering that the CZTSSe solar cells with high efficiencies have been fabricated in nonstoichiometry conditions, the formation of secondary phases is inevitable. The existence of secondary phases is known to be detrimental to device performance since they may enhance the recombination due to their inappropriate band gap and electrical properties. Secondary phases with a band gap smaller than that of the kesterite (Cu₂SnS₃, Cu₂S, and SnS in a pure sulfide and Cu₂SnSe₃ in a pure selenide) can limit the V _oc, whereas those with a larger band gap can hinder the charge transport.^{Reference Siebentritt25,Reference Kumar, Dubey, Adhikari, Venkatesan and Qiao41} Therefore, it is desirable to analyze the component uniformity and to detect the secondary phases in CZTSSe at the atomic level. However, detection of the secondary phases at the nanoscale level and their characterization require advanced techniques because they are structurally complex and seemingly similar to the CZTSSe matrix.

In this regard, the combination of atom probe tomography (APT) with ab initio DFT calculations is an effective strategy to study compositional inhomogeneity at the nanoscale. The largest strengths of APT as an analytical technique are three-dimensional compositional imaging in the nanoscale analysis of buried features and high analytical sensitivity. APT simultaneously achieves a resolution finer than 0.3 nm in all three directions of a three-dimensional image.^{Reference Kelly and Miller42} Schwarz et al.^{Reference Schwarz, Marques, Botti, Mousel, Redinger, Siebentritt, Cojocaru-Mirédin, Raabe and Choi43–Reference Schwarz, Cojocaru-Mirédin, Mousel, Redinger, Raabe and Choi45} extensively studied the secondary phases in CZTSSe using the laser-assisted APT analysis. They examined CZTSe samples before and after annealing. Three-dimensional elemental mapping revealed Cu enrichment and Sn depletion at grain boundaries before annealing. The annealing resulted in a strong enrichment of grain boundaries with Zn (nanometer sized ZnSe formation up to 30 at.%), strong Cu and Sn depletion, and slight Se depletion. Moreover, a substantial amount of Na (∼0.4 at.%) was segregated at the majority of grain boundaries as well as at some parts of the ZnSe/CZTSe interfaces, which was accompanied by co-segregation of K (∼0.15 at.%).^{Reference Schwarz, Cojocaru-Mirédin, Choi, Mousel, Redinger, Siebentritt and Raabe44} It is known that the valence band maximum (VBM) of CZTSe is determined by the Se–Cu p–d hybridization.^{Reference Jaffe and Zunger46,Reference Chen, Gong, Walsh and Wei47} Cu-enriched grain boundaries induce an upward shift of the VBM, which consequently increases the recombination. The Zn-enriched grain boundaries with Cu depletion, however, induce a downward VBM shift, which can act as a hole barrier and is beneficial for solar cell performance.^{Reference Persson and Zunger48,Reference Gloeckler, Sites and Metzger49}

The beneficial role of the downward VBM shift at the grain boundaries was also reported by Kim et al.^{Reference Kim, Choi, Choi, Gershon, Lee, Wang, Shin and Chung50} Using STEM, EDX, and EELS, they studied the electronic structure of internal interfaces in CZTSe. In CZTSSe solar cells, air annealing has been shown to enhance V _oc.^{Reference Kim, Choi, Choi, Gershon, Lee, Wang, Shin and Chung50,Reference Sardashti, Haight, Gokmen, Wang, Chang, Mitzi and Kummel51} The study by Kim et al. aimed to investigate the mechanism behind the improvement of the device performance with air annealing. They demonstrated the appearance of nanoscale layers with a thickness of a few nanometers along the grain boundaries after air annealing, where the Se sites are substituted by oxygen as shown in Figs. 6(a) and 6(b). The EELS measurement further confirmed the existence of oxygen in the lattice: The oxygen K-edges were clearly seen at the interface [Fig. 6(c)]. From the DFT calculations, they were able to clarify the effect of oxygen on the electronic band structure of CZTSe. Oxygen substitution leads to a downward shift of VBM by 0.49 eV compared with that of CZTSe and hinders the hole transport to the grain boundaries. Thus, a large energy barrier to the hole transport formed due to the oxidation and efficiently inhibited the charge carrier recombination at the grain boundaries.

FIG. 6.

(a) High resolution electron microscopic images with two different types of grain boundaries, (b) high-angle annular dark-field STEM images and EDX maps at grain boundaries, and (c) annular bright-field STEM image (top) and EELS spectra (bottom) in CZTSe after air annealing. Reproduced with permission from Ref. Reference Kim, Choi, Choi, Gershon, Lee, Wang, Shin and Chung50.

Furthermore, Schwarz et al. reported the formation of Cu₂Zn₅SnSe₈ and Cu₂Zn₆SnSe₉ phases in CZTSe.^{Reference Schwarz, Marques, Botti, Mousel, Redinger, Siebentritt, Cojocaru-Mirédin, Raabe and Choi43} The APT study was performed on two CZTSe samples with different compositional ratios: one grown under Cu-rich conditions and another under Cu-poor conditions. Three-dimensional elemental maps of Cu, Zn, and Sn are shown in Fig. 7(a) for two different regions (>200 nm below the surface and >500 nm above the Mo back contact) for Cu-rich CZTSe. They observed Cu₂Zn₅SnSe₈ and Cu₂Zn₆SnSe₉ phases as well as Cu_2−xSe and ZnSe precipitates in Cu-rich CZTSe samples, whereas only ZnSe and Cu₂Zn₅SnSe₈ phases were detected in Cu-poor CZTSe samples [Fig. 7(a)]. However, neither Cu₂Zn₅SnSe₈ nor Cu₂Zn₆SnSe₉ was detected using XRD, indicating that the two phases have structures similar to CZTSe and ZnSe or the volumes of fractions are very small in CZTSe. Based on DFT calculations, the authors suggested that a possible disorder among the cations (the presence of three different cations on the Zn sites) distorts the original symmetry and the resultant metastable phases can co-exist in the CZTSe [Figs. 7(b) and 7(c)]. DFT calculation further yielded the positive energy difference with respect to the decomposition into Cu₂ZnSnSe₄ and ZnSe.^{Reference Schwarz and GmbH52}

FIG. 7.

(a) Three-dimensional elemental maps of Cu (blue), Zn (orange), and Sn (green). Structure of (b) the Cu₂Zn₅SnSe₈ and (c) Cu₂Zn₆SnSe₉ phase calculated by DFT: Cu (blue), Zn (grey), Sn (rose), and Se (green).^{Reference Schwarz, Marques, Botti, Mousel, Redinger, Siebentritt, Cojocaru-Mirédin, Raabe and Choi43} (d) and (e) APT elemental maps of Zn (orange), Cu (blue), Sn (dark green), and Na (green) acquired from Cu₂ZnSnSe₄ precursor.^{Reference Schwarz, Cojocaru-Mirédin, Mousel, Redinger, Raabe and Choi45} (f) Schematic drawing of the Cu₂ZnSnSe₄ precursor before (top) and after (bottom) annealing.^{Reference Schwarz, Cojocaru-Mirédin, Mousel, Redinger, Raabe and Choi45} Reproduced with permission from Refs. Reference Schwarz, Marques, Botti, Mousel, Redinger, Siebentritt, Cojocaru-Mirédin, Raabe and Choi43 and Reference Schwarz, Cojocaru-Mirédin, Mousel, Redinger, Raabe and Choi45.

Moreover, using APT and STEM-EDX, Schwarz et al. carried out an advanced study on the direct detection and quantification of nanosized Cu–Sn–Se compounds and provided further verification of the existence of ZnSe randomly distributed throughout CZTSe as shown in Figs. 7(d) and 7(e).^{Reference Schwarz, Cojocaru-Mirédin, Mousel, Redinger, Raabe and Choi45} In the same work, they also confirmed the existence of the Cu_2−xSe phase as wedge-shaped grains at the surface of CZTSe using STEM-EDX. Consequently, they proposed a growth model of CZTSe, especially under Cu-rich conditions [see Fig. 7(f)]. In this model, excess Cu leads to the formation of Cu_2−xSe at the surface of the Cu-rich precursor while the Zn excess yields a nanoscale ZnSe network across the entire film of CZTSe before annealing. However, annealing leads to the disappearance of the Cu_2−xSe and nanoscale distribution of the Cu–Sn–Se phase at 200 nm below the surface with the ZnSe network still present in the CZTSe. The locations of these compounds in the space charge region enhance the recombination and limits the device performance. Another performance loss that is related to the V _oc deficit is a narrow band gap of Cu–Sn–Se compounds (Cu₂SnSe₃) in a range of 0.4–0.8 eV.^{Reference Takeshi, Tsuyoshi, Kouji, Masaru and Takahiro53,Reference Rau and Werner54} Thus, they concluded that the existence of these detrimental secondary phases close to the CZTSe/CdS interface results in a nonideal heterojunction.^{Reference Schwarz, Cojocaru-Mirédin, Mousel, Redinger, Raabe and Choi45}

A. Design of novel earth-abundant photovoltaic material with an ordered lattice: Cu₂–II²⁺–Sn–(S,Se)₄ chalcogenide (II = earth-alkali element)

Given that the main driving force of Cu–Zn antisite formation is the similarity of the atomic radii and chemical valences of Cu and Zn ions, researchers reasoned that the introduction of an ionic mismatch around the constituents would be an effective way to suppress the formation of cation antisite defects in the CZTSSe lattice. Following this line of thought, researchers have substituted various cations with ionic radii different from Cu or Zn: for example, Ag (1.14 Å)^{Reference Shannon55–Reference Gershon, Lee, Antunez, Mankad, Singh, Bishop, Gunawan, Hopstaken and Haight60} as a substitute for Cu (0.74 Å), and Cd (0.92 Å)^{Reference Yuan, Chen, Xiang, Gong, Walsh, Park, Repins and Wei56,Reference Su, Tan, Li, Zeng and Batabyal61,Reference Fu, Tian, Zhou, Kou, Meng, Zhou and Wu62} as a substitute for Zn (0.74 Å). According to the calculated band structure as shown in Fig. 8(a), however, the VBM of CZTS is mainly composed of Cu 3d and S 3p states, while the conduction band minimum (CBM) is formed by the hybridization of S 3p and Sn 5s states.^{Reference Chen, Walsh, Gong and Wei19,Reference Paier, Asahi, Nagoya and Kresse63} This suggests that there is no significant contribution of Zn 3d states to either VBM or CBM, i.e., the Zn 3d states do not play a critical role in the near-band-edge electronic structure. In terms of maintaining good optical properties for photovoltaic applications (such as band gap) of kesterite, divalent cation substitution for the Zn site is more advantageous than monovalent cation substitution for the Cu site as long as the new divalent cation has a smaller or larger ionic size compared to Cu.

FIG. 8.

(a) The HSE-based band structure (left) and molecular interaction diagram (right) for kesterite-structured Cu₂ZnSnS₄.^{Reference Paier, Asahi, Nagoya and Kresse63} (b) Crystal structures and HSE-based band structures and density of states for Cu₂BaSnS_4−xSe_x (x = 0, 3, and 4). The Cu, Ba, Sn, S, and Se atoms of the crystal structure illustrations are shown in blue, gray, green, yellow, and light green, respectively.^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68} (c) Evolution of calculated band measured band gap values as a function of Se content in Cu₂BaSnS_4−xSe_x.^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68} Reproduced with permission from Refs. Reference Paier, Asahi, Nagoya and Kresse63 and Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68.

Researchers have tried to substitute Zn sites with various transitional metals such as Mn, Fe, and Co. Unfortunately, no improvement in the CZTSSe photovoltaic performance was reported by such cation substitutions. Recently, Lie et al. observed that when the Mn content in CZTSSe exceeded 5 at.%, Mn-substituted device efficiencies significantly decreased; pure Cu₂MnSn(S,Se)₄-based photovoltaic devices even showed less than 1% power conversion efficiency (PCE) because of high carrier concentrations of about 10¹⁷–10¹⁸ cm⁻³.^{Reference Lie, Rui Tan, Li, Leow, Tay, Bishop, Gunawan and Wong64} In addition, Cu₂FeSnS₄-based and Cu₂CoSnS₄-based photovoltaic devices yielded a PCE of 0.08% and 1.78%, respectively, even when their band gaps are in the optimal range of 1.40–1.55 eV.^{Reference Dong, Ashebir, Qi, Chen, Wan, Chen and Wang65,Reference Maldar, Gaikwad, Mane, Nikam, Desai, Giri, Sarkar and Moholkar66} Such a low device performance seems to be associated with partially filled 3d states that lead to multiple charge states (e.g., the possible number of Mn oxidation states is 10) and semimetallic behaviors.^{Reference Lie, Rui Tan, Li, Leow, Tay, Bishop, Gunawan and Wong64,Reference Shin, Saparov and Mitzi67} Therefore, transitional metals with multiple charge states are not ideal candidates as cation substitutes for Zn sites.

More recently, several groups have attempted to use earth-alkali metals as an alternative cation to replace Zn in the CZTSSe since these elements have a stable divalent charge state (i.e., a fully filled 3d state) and an ionic size larger than that of Cu.^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68–Reference Zhu, Huhn, Wessler, Shin, Saparov, Mitzi and Blum70} Based on these anticipated advantages, Wang et al. calculated the phase stability of a kesterite-based Cu₂–II–Sn–S₄ system with II = Mg, Ca, Sr, and Ba elements.^{Reference Wang, Chen, Yang, Lang, Xiang, Gong, Walsh and Wei71} However, it was found that the earth-alkali elements induce phase separation in binary and ternary compounds (e.g., Cu₂MgSnS₄ → Cu₂SnS₃ + MgS). Despite this theoretical prediction, Teske et al. successfully synthesized Cu₂SrSnS₄ (CSTS) and Cu₂BaSnS₄ (CBTS) from a mixture of three binary sulfides and the quaternary compounds. CSTS and CBTS crystallize in a trigonal structure with space group P3₁.^{Reference Teske72,Reference Teske and Vetter73} The discrepancy between theoretical and experimental studies originates from the crystal structure considered for the phase stability estimation of Cu₂–II–Sn–S₄ in a calculation that was different from the real crystal structure (i.e., a tetragonal structure with I-4 instead of a trigonal structure with P3₁ was applied for the DFT calculation.). A DFT study performed by Shin et al. verified that trigonal-structured Cu₂–II–Sn–S₄ systems with II = Sr and Ba are thermodynamically stable.^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68,Reference Hong, Lin, Meng and Yan69} Particularly, the phase stability zone of CBTS is wider than that of CSTS.^{Reference Hong, Lin, Meng and Yan69} In this regard, the CBTS material system was considered as a novel earth-abundant chalcogenide with less antisite disordering.

Using X-ray diffraction data of solid-state-synthesized Cu₂BaSnS_4−xSe_x powders, Shin et al. observed that when the Se content in Cu₂BaSnS_4−xSe_x exceeds x > 3, there is a phase transition from a trigonal (0 ≤ x ≤ 3) to an orthorhombic structure (x = 4).^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68} In addition to the phase transition [Fig. 8(b)], the estimated band gap of Cu₂BaSnS_4−xSe_x (0 ≤ x ≤ 3) decreases with Se content from 1.74 to 1.33 eV, whereas that of pure selenide (x = 4) shows a larger band gap of 1.50 eV.^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68} The measured band gaps of Cu₂BaSnS_4−xSe_x (0 ≤ x ≤ 3) are in the range of 1.5–2.0 eV [Fig. 8(c)]. The two results are in good agreement. The band gap values of Cu₂BaSnS_4−xSe_x are suitable for single-junction and multiphotovoltaic applications.

Hong et al. carried out a theoretical study on point defects of the CBTS.^{Reference Hong, Lin, Meng and Yan69} As can be seen in Fig. 9(a), under Cu-deficient compositions of CBTS (i.e., μ_Cu < 0), the Cu vacancy (denoted as V_Cu) has the lowest defect formation energy, which means that the V_Cu is prevalent in the CBTS.^{Reference Hong, Lin, Meng and Yan69} In addition, the transitional energy level of V_Cu is a shallow acceptor that makes them an effective acceptor-like intrinsic defect [Fig. 9(b)].^{Reference Hong, Lin, Meng and Yan69} On the other hand, Cu–Ba antisite defects possess a very high formation energy due to the large ionic size of Ba.^{Reference Hong, Lin, Meng and Yan69} Therefore, cationic antisite defects would be unlikely to form in the material. Interestingly, the overall defect properties of CBTS are similar to those of CIGSe rather than CZTSSe.

FIG. 9.

(a) The calculated formation energies of intrinsic point defects for Cu₂BaSnS₄ as a function of chemical potentials.^{Reference Hong, Lin, Meng and Yan69} (b) The calculated transition energy levels of acceptor-like (left) and donor-intrinsic point defects (right) in Cu₂BaSnS_4.^{Reference Hong, Lin, Meng and Yan69} Reproduced with permission from Ref. Reference Hong, Lin, Meng and Yan69.

B. Potentials for earth-abundant Cu₂BaSn(S,Se)₄-based solar cells

Recently, several groups demonstrated earth-abundant chalcogenide photovoltaic devices using trigonal Cu₂BaSnS_4−xSe_x (CBTSSe) absorbers.^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68,Reference Hong, Lin, Meng and Yan69,Reference Ge, Koirala, Grice, Roland, Yu, Tan, Ellingson, Collins and Yan74,Reference Shin, Zhu, Huang, Gunawan, Blum and Mitzi75} As a proof-of-concept, the first reported CBTS-based photovoltaic device yielded 1.62% PCE using the same device architecture as CIGSe and CZTSSe solar cells [Fig. 8(a)].^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68} A pure sulfide CBTS-based device with a CdS/CdS:O bilayer buffer achieved a higher PCE of 2.03% by Ge et al.^{Reference Ge, Koirala, Grice, Roland, Yu, Tan, Ellingson, Collins and Yan74} Low device efficiencies are due to the relative large band gap of CBTS, ∼2.0 eV, which is not optimal for single-junction solar cells. As previously stated, Se incorporation in CBTSSe leads to a reduction in the band gap; Cu₂BaSnS_4−xSe_x (x = 3) possesses a band gap of 1.55 eV as measured by PL [Fig. 10(b)].^{Reference Shin, Zhu, Huang, Gunawan, Blum and Mitzi75} An additional benefit of the selenization process is the enlargement of the CBTSSe grain size up to more than >1 μm, which is suitable for high-performance photovoltaic absorbers.^{Reference Shin, Zhu, Huang, Gunawan, Blum and Mitzi75–Reference Zhou, Shin, Ngaboyamahina, Han, Parker, Mitzi and Glass77} Because of the reduced band gap and improved microstructure of CBTSSe, Se-incorporated CBTSSe photovoltaic devices demonstrated the highest PCE of 5.2% with V _oc = 611 mV, J _sc = 17.4 mA cm², FF = 48.9% as shown in Fig. 10(b).^{Reference Shin, Zhu, Huang, Gunawan, Blum and Mitzi75} The abrupt absorption transition in the EQE spectrum and sharp PL feature suggest that band tailing associated with antisite defects is not a substantial issue in this material. In terms of control over the detrimental antisite defects and associated band-tailing, Ba substitution makes the Se-rich CBTSSe absorber an attractive candidate for the use within earth-abundant photovoltaic devices.

FIG. 10.

The light (solid line) and dark (dashed line) current density–voltage characteristics and external quantum efficiencies spectrum of (a) Cu₂BaSnS₄-based and (b) Cu₂BaSnS_4−xSe_x (x ≈ 3)-based PV devices.^{Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68,Reference Shin, Zhu, Huang, Gunawan, Blum and Mitzi75} In (b), the PL spectrum for Cu₂BaSnS_4−xSe_x (x ≈ 3) film is also displaced. Reproduced with permission from Refs. Reference Shin, Saparov, Zhu, Huhn, Blum and Mitzi68 and Reference Shin, Zhu, Huang, Gunawan, Blum and Mitzi75.