Resonant x-ray ptychographic nanotomography of kesterite solar cells

The ${\mathrm{Cu}}_2{\mathrm{ZnSnS}}_4$ kesterite is currently among the most promising inorganic, nontoxic, earth-abundant materials for a new generation of solar cells. Interfacial defects and secondary phases present in the kesterite active layer are, however, detrimental to the performance of the device. They are typically probed with techniques that are destructive or limited to the surface, and x-ray diffraction cannot reliably distinguish small amounts of zinc sulfide or copper tin sulfide from kesterite. Conversely, resonant ptychographic tomography, which relies on electron density contrast, overcomes these limitations. Here, we demonstrate how this technique can enable localization and quantification of secondary phases, along with measurements of adherence at the interfacial layers, on complete and functioning devices. In our experiment, we utilize an x-ray energy value far from absorption edges as well as three single energies corresponding to the absorption edges of Cu, Zn, and Sn, to gain elemental sensitivity to these elements and enhance contrast between phases with similar electron density. As a result, we image and identify in the active layer grains of a secondary phase, namely, zinc sulfide, which is not easily discriminated by other standard characterization techniques. In addition, we are able to observe Cu diffused from the active layer into the CdS buffer layer as well as Cu in the form of copper sulfide at their interface. Other relevant morphological features are best resolved off-resonance at the optimal energy for the synchrotron beamline with ∼20 nm resolution.

Figure 1

Histograms of electron density value for voxels in a ${\mathrm{MoS}}_2$ slice of E1 (a) and comparison with calculated $\mathit{\delta }$ of ${\mathrm{MoS}}_2$ from tabulated values (b). The measured values are about 95% of the expected ones.

Figure 2

(a) Volume rendering of the layer stack from the resulting tomographic reconstruction of sample E1. Layers are enumerated in ascending order from bottom to top and their expected size are reported in parenthesis. False colors are assigned as in the colorbar to the main peaks of the histogram of electron densities to distinguish between the layers. (b) Apparent electron density profile of E1 sample at the different energies. The profile is evaluated across the blue slab in the center of (a), averaging on an area of 8 × 8 pixels ($200\times 200\mathrm{n}{\mathrm{m}}^2$). Profile measurements can be compared to the actual electron densities of the compounds (dashed black lines), also reported in Table S2. The area is chosen within a regular grain, with flat surface and no additional features. The dip of electron density in proximity of the CdS layer measured at Cu edge is interpreted as evidence of Cu migration into the layer.

Figure 3

Morphological features from phase tomograms of the CZTS layer. (a) Sagittal slices of E1 and E2 highlighting different types of defects: pinholes (d1), CdS precipitates (d2), thinning of CZTS layer (d3), discontinuity of ITO (d4). The dashed green lines in E1 indicate the range of depth of axial slices illustrated next. (b) Grain morphology of the active layer illustrated by four consecutive axial slices of the active layer, from bottom (slice 47) to top (slice 50). Distance between each slice equals the voxel size (25 nm). Grain size is around 1 µm and the grain boundaries are clearly resolvable. (c) A comparison of the same slice at different edge energies showing that morphological features are best highlighted at the off-resonance energy. The colorbar is truncated in (b) and (c) to enhance gray level contrast.

Figure 4

Chemical features revealed by contrast variation at different energies. Each row reports the same slice or area at the four different energies (columns). ZnS grains (yellow arrows) are found in E1 and E2 (row 1 and 2). Cu- and Sn-rich clusters (circled in yellow) from E2 are highlighted in row 3 and 4, respectively. Colorbars are truncated to enhance gray level contrast.

Figure 5

Bivariate histograms of ZnS in E1 (a) and E2 (b). The apparent electron density at the Zn edge is reported on the $x$ axis vs. the other energies. The active layer of E1 contains significantly larger amounts of ZnS. Clusters in the top row plots are asymmetrical because of the lower dispersion off-resonance. The apparent electron density of ZnS is shifted down to ca. 0.95 ${\AA }^{-3}$ at the Zn-edge. Expected shift values are reported in Table S3 of Supplemental Material [28].

Figure 6

Images of the CdS/CZTS interface. (a) Axial slice of E2 at the four different energies. The dashed green line refers to the cross section of the segmented interface in (b). The image taken at the Cu K-edge appears darker than the others highlighting Cu migration into the CdS layer. That is illustrated by the trail of points (yellow arrows) around the CZTS cluster in the bivariate histograms of (c). These points show a lower electron density at the Cu edge and similar or higher density at Zn or Sn edge (nominal electron density of CdS is 1.28 ${\AA }^{-3}$).

Figure 7

Top row: comparison of an axial slice of the active layer of E2, phase $\delta$ (a) vs. absorption $\beta$ (b) off-resonance tomograms. The slice contains a feature (yellow arrow) illustrated in Fig. 4. In this case the dark feature in the absorption-based tomogram rules out ${\mathrm{Cu}}_3{\mathrm{SnS}}_4$ in favor of CuS. The higher contrast of phase tomograms is illustrated by the histogram of occurrences of $\delta$ (c) and $\beta$ (d) in the total volume. The dispersion of measured values is small enough for the $\delta$ to allow high contrast between layers (c), whereas $\beta$ values are roughly centered around two values only (d). In (e) the dispersion of measured values is quantified on a slice of ${\mathrm{MoS}}_2$. The dispersion of measurement is only slightly higher for $\delta$ values, which are an order of magnitude higher than $\beta$, providing superior relative accuracy.