$\mathrm{Se}$ Inter-Diffusion Limits Absorber Layer Grain Growth in $\mathrm{Cd}\mathrm{Se}$-$\mathrm{Cd}\mathrm{Te}$ Photovoltaics

Diffusion of $\mathrm{Se}$ from the $\mathrm{Cd}\mathrm{Se}$ window layer into the $\mathrm{Cd}\mathrm{Te}$ absorber improves the short circuit current density by narrowing the band gap and increasing the carrier lifetime. Thicker $\mathrm{Cd}\mathrm{Se}$ layers, however, show a dramatic loss in photocurrent collection due to $\mathrm{Se}$ over-alloying. Electron microscopy investigations show that this decrease in performance is due to the formation of small grains (∼783 nm average diameter), which exhibit grain boundary porosity in the $\mathrm{Se}$ inter-diffusion region. The larger grain boundary area and void free surfaces give rise to higher levels of nonradiative recombination, and therefore, a lower photocurrent. It is proposed that the small grain size is due to a drag force exerted by segregated $\mathrm{Se}$ solute atoms on a moving grain boundary, while faster $\mathrm{Se}$ diffusion along the grain boundaries results in vacancy build up and porosity due to the Kirkendall effect. The results indicate that the device processing conditions must be carefully controlled such that the negative effects of $\mathrm{Se}$ alloying (i.e., smaller grains, Kirkendall voids) do not undermine its benefits.

Popular Summary

Cadmium telluride ($\mathrm{Cd}\mathrm{Te}$) solar cells are attractive for their cost-effectiveness and relatively high efficiency. By incorporating a cadmium selenide ($\mathrm{Cd}\mathrm{Se}$) window layer, selenium can diffuse into the $\mathrm{Cd}\mathrm{Te}$ absorber, enhancing the short circuit current density. In this study, the authors explore the impact of varying $\mathrm{Cd}\mathrm{Se}$ layer thickness on the structure and chemistry of $\mathrm{Cd}\mathrm{Te}$ solar cells. Thicker $\mathrm{Cd}\mathrm{Se}$ layers lead to over-alloying of selenium, causing a notable decline in photocurrent. Electron microscopy shows that the decline can be attributed to the formation of small grains with porous grain boundaries in the selenium inter-diffusion region. The small grain size may arise from a drag force exerted by segregated selenium solute atoms on moving grain boundaries. Faster diffusion along the grain boundaries lead to accumulation of vacancies and subsequent porosity. In conclusion, device processing conditions should be carefully considered to optimize selenium alloying for maximum performance benefits in $\mathrm{Cd}\mathrm{Te}$ solar cells.

Figure 1

Schematic side-on view of a focused ion-beam milled bevel cross section. Due to the shallow 8° bevel angle, blue and gray thin-film layers appear elongated in cross section. Diagram not drawn to scale.

Figure 2

(a) External quantum efficiency for devices with variable $\mathrm{Cd}\mathrm{Se}$ layer thickness (50, 100, 200, and 400 nm). Results are shown for the device with highest efficiency measured at a given $\mathrm{Cd}\mathrm{Se}$ layer thickness. Corresponding current density-voltage graphs are shown in (b). Short circuit current density (J_SC), open circuit voltage (V_OC), fill factor (FF), and efficiency for all devices are plotted in (c)–(f).

Figure 3

Simultaneously acquired secondary electron (a) and panchromatic CL (b) images for the 100 nm $\mathrm{Cd}\mathrm{Se}$ bevel cross section. Horizontal dashed lines in (a) indicate the boundaries of the porosity region (${\mathrm{Sn}\mathrm{O}}_2:\mathrm{F}$ is the transparent conducting oxide). (c) Secondary electron image with the region of hyperspectral CL mapping indicated with a red box. CL spectra extracted from the porosity region (labeled “CST”) and from close to the absorber layer back surface (labeled “$\mathrm{Cd}\mathrm{Te}$”) are superimposed in (d). Maximum intensity of each spectrum is normalized for a direct visual comparison. (e) Band gap grading and (f) $\mathrm{Se}$ concentration profiles extracted from the hyperspectral map.

Figure 4

STEM bright field (a) and dark field (b) images of the 100 nm $\mathrm{Cd}\mathrm{Se}$ device. $\mathrm{Pt}$ is a protective layer used during FIB TEM specimen preparation, while ${\mathrm{Sn}\mathrm{O}}_2:\mathrm{F}$ is the transparent conducting oxide. Note that part of the TEM cross section has been damaged by FIB milling during specimen preparation. Arrows in (b) indicate porosity in the absorber layer. Selected area electron diffraction patterns acquired from the top and bottom red circle regions in (a) are shown in (c),(d), respectively.

Figure 5

EBSD pattern quality (a) and phase map (b) for the 100 nm $\mathrm{Cd}\mathrm{Se}$ absorber layer (bevel cross section). EBSD patterns were matched against cubic zinc blende $\mathrm{Cd}\mathrm{Te}$ (red) and hexagonal wurtzite $\mathrm{Cd}\mathrm{Se}$ (yellow) reference phases. All grains in the absorber layer are identified as being cubic.

Figure 6

(a) STEM dark field image of the 100 nm $\mathrm{Cd}\mathrm{Se}$ device with EDX false color image superimposed over the map region. $\mathrm{Te}$ Lα, $\mathrm{Cd}$ Lα, $\mathrm{Se}$ Kα, $\mathrm{Sn}$ Lα, and $\mathrm{O}$ Kα EDX x-ray maps are shown in (b). Presence of $\mathrm{Te}$ in the ${\mathrm{Sn}\mathrm{O}}_2:\mathrm{F}$ layer is an artefact due to energy overlap between $\mathrm{Te}$ Lα and $\mathrm{Sn}$ Lα x-ray peaks. Scale bar for each of the EDX maps is 500 nm.

Figure 7

Simultaneously acquired secondary electron (a) and panchromatic CL (b) images for the 400 nm $\mathrm{Cd}\mathrm{Se}$ bevel cross section. Horizontal dashed lines in (a) indicate the boundaries of the porosity region (${\mathrm{Sn}\mathrm{O}}_2:\mathrm{F}$ is the transparent conducting oxide). (c) Secondary electron image with the region of hyperspectral CL mapping indicated with a red box. (d) Band gap grading and (e) $\mathrm{Se}$ concentration profiles extracted from the hyperspectral map. Some of the data points within the porosity region in (e) are not shown due to complications in calculating the $\mathrm{Se}$ concentration; see text for further details.

Figure 8

STEM bright field (a) and dark field (b) images of the 400 nm $\mathrm{Cd}\mathrm{Se}$ device. Arrows in (b) indicate porosity in the absorber layer. (c) Many-beam bright field TEM image of an absorber layer grain (image center) close to the ${\mathrm{Sn}\mathrm{O}}_2:\mathrm{F}$ interface. Two selected area electron diffraction patterns acquired from this grain at different specimen tilts are shown in (d),(e).

Figure 9

EBSD pattern quality (a) and phase map (b) for the 400 nm $\mathrm{Cd}\mathrm{Se}$ absorber layer (bevel cross section). EBSD patterns were matched against cubic zinc blende $\mathrm{Cd}\mathrm{Te}$ (red) and hexagonal wurtzite $\mathrm{Cd}\mathrm{Se}$ (yellow) reference phases. All indexed grains in the absorber layer are cubic, although for some small grains at the bottom it was not possible to identify the phase due to poor pattern quality.

Figure 10

(a) STEM dark field image of the 400 nm $\mathrm{Cd}\mathrm{Se}$ device with EDX false color image superimposed over the map region. $\mathrm{Te}$ Lα, $\mathrm{Cd}$ Lα, $\mathrm{Se}$ Kα, $\mathrm{Sn}$ Lα, and $\mathrm{O}$ Kα EDX x-ray maps are shown in (b). Presence of $\mathrm{Te}$ in the ${\mathrm{Sn}\mathrm{O}}_2:\mathrm{F}$ layer is an artefact due to energy overlap between $\mathrm{Te}$ Lα and $\mathrm{Sn}$ Lα x-ray peaks. Scale bar for each of the EDX maps is 500 nm.

Figure 11

(a) Panchromatic CL image of the transition region between the small interfacial grains (image bottom) and the rest of the absorber layer in the 400 nm $\mathrm{Cd}\mathrm{Se}$ sample. Red box denotes the area used for hyperspectral CL mapping. Gaussian functions were fitted to each pixel in the hyperspectral map. In (b), the amplitude of the Gaussian is mapped, where white represents 683 counts and black 97 counts. (c) Peak wavelength of the fitted Gaussian, where white and black correspond to 842 and 808 nm, respectively. Images appear pixelated due to fewer electron beam scan positions being used in hyperspectral mapping. (d) CL spectra extracted from a grain boundary and neighboring grain interior labeled “GB” and “GI,” respectively, in (c). Peak intensity for each spectrum has been normalized for direct visual comparison.

Figure 12

Schematic illustration of the proposed grain growth mechanism in $\mathrm{Cd}\mathrm{Se}$-$\mathrm{Cd}\mathrm{Te}$ photovoltaics during close space sublimation. $\mathrm{Cd}$ and $\mathrm{Te}$ atoms in the vapor phase and as adatoms are denoted in gray. They form $\mathrm{Cd}\mathrm{Te}$ islands, which eventually merge into grains (also colored gray). $\mathrm{Cd}\mathrm{Se}$ layer and $\mathrm{Se}$ atoms are denoted in yellow. Underlying substrate is blue. See text for further details.