Substitution of indium by zinc and tin in CuIn(S, Se)2 leads to the quaternary compound semiconductor Cu2ZnSn(S, Se)4 (CZTSSe). The record efficiency of thin-film solar cells using a CZTSSe absorber layer is above 10 % [47].

Both compounds, CZTS and CZTSe, belong to the family of tetrahedrally — coordinated adamantine semiconductors [48]. Here each anion is tetrahedrally coordinated by four cations (two copper, one zinc and one tin), whereas each cation is coordinated by four anions (sulfur or selenium). Thus, the structure is charac­terized by a well-defined framework of tetrahedral bond arragnements, which is advantageous for the properties of the material.

For the quaternary a2BiiCivXJi chalcogenides with A-Cu; B-Zn; C-Si, Ge, Sn and X-S, Se, different crystal structures are discussed in literature: the kesterite — type structure (space group I4), the stannite-type structure (space group I42m), as well as the wurtz-stannite (space group Pmn21) and the wurtz-kesterite type structure (space group Pc). The same tetrahedral metal-coordination (2Cu, one II — and one IV-element surrounding each S-atom) is possible in all four space groups. A clear decision can only be made by a detailed structure analysis for each compound.

At room temperature CZTS adopts the space group /4. The structure can be described as a cubic close-packed array of sulfur atoms, with metal atoms occu­pying one half of the tetrahedral voids. CZTSe was reported to crystallize in the space group 142m in a topologically-identical structure, but different cation distri­bution of A1 and B11 among the positions (0, 0, 0), (0, XA, ^), and (0, XA, %) [49]. Both structures are represented in Fig. 5.13.

Neutron powder diffraction and the method of the average neutron-scattering length were used to clarify possible differences of the cation distribution in CZTS and CZTSe, especially with respect to the electronically-similar elements copper and zinc [50, 51].

According to the general formula for the calculation of the average neutron­scattering length (see also Sect. 5.3.2)

bj = Aj • bA + Bj • bB + Vj (5.13)

where A and B are two different cations, V represent possible vacancies and j stands for the Wyckoff position, the following equations were derived for the calculation of the experimental average neutron-scattering lengths 4bjexp

exp

b2a occ2a ‘ bCu

b2xp = OCC2c ‘ bZn b2dp = occ2d • bCu

b2x/ = Occ2a • bZn

b4d^ = occ4d • bCu

The cation site-occupancy values occ, resulted from the Rietveld analysis of the neutron diffraction data. Because the study was performed using stoichiometric powder samples (the chemical composition of the samples was determined by wavelength-dispersive X-ray spectroscopy), vacancies were not taken into account.

The Rietveld analysis of the neutron diffraction data was performed by applying different cation-distribution models within both structure types as the starting model for the crystal structure (see Table 5.4). The cation site occupancies were taken as free parameters in the refinement. It was found that the refined cation site-occu­pancy values differ from their nominal values (Table 5.4), with the exception of the tin. Thus, one can conclude that the site is not only occupied by the initially — supposed model cation, but also by a mixture of different elements. For example, a decrease of the average neutron-scattering length of the 2d site in the kesterite-type structure can be attributed to a partial Zn occupation due to bCu > bZn.

The refined cation site occupancies were used to calculate the experimental average neutron-scattering lengths (see Fig. 5.14). The first result obtained was a confirmation of the occupancy of the 2b site by tin in both compounds. Con­cerning the sites 2a, 2c and 2d for the space group /4 as well as 2a and 4d for the space group /42m, a similar picture for both compounds appears. Irrespective of the structure model used in the Rietveld analysis, the average neutron-scattering length of the 2a position indicates that this position is occupied by copper only. The average neutron-scattering lengths of the sites 2c and 2d indicate a mixed occu­pancy of these both sites by copper and zinc. Approximately 50 % of the zinc site 2c is occupied by copper and vice versa concerning the copper site 2d, forming

CuZn and ZnCu anti-sites. Nevertheless, this disorder is limited to the positions 2c and 2d and hence to the lattice planes at z = 1/4 and 3/4. The 2a site is always occupied by copper only, and there are no indications of zinc occupying that site.

These facts bring us to the conclusion that both compounds, CZTS and CZTSe, adopt the kesterite-type structure, but exhibit Cu-Zn disorder in the (00l) lattice planes at z = 1/4 and 1/3. This disorder creates CuZn and ZnCu anti-site defects. These results are supported by ab initio calculations carried out on these systems [52, 53]. The authors report that the CuZn anti-site acceptor can be predicted as the most probable defect and thus can be easily formed. Moreover, the kesterite-type structure was indicated as the ground-state structure for both CZTS and CZTSe, though the stannite-type structure has only a slightly-lower binding energy (CZTS: 2.9 meV/atom [52], 1.3 meV/atom [53], CZTSe: 3.8 meV/atom [52], 3.3 meV/atom [53]).

Using the approach of the average neutron-scattering length, and applying dif­ferent structural models with different cation distributions in the structural description in the Rietveld refinement procedure, the crystal structure of CZTS and CZTSe was determined and intrinsic point-defects were identified experimentally. In contradiction with earlier X-ray diffraction studies, the kesterite-type structure was also proved for CZTSe.