Origin of the Optical Contrast in Phase-Change Materials

Several chalcogenide alloys exhibit a pronounced contrast between the optical absorption in the metastable rocksalt and in the amorphous phase. This phenomenon is the basis for their application in optical data storage. Here we present ab initio calculations of the optical properties of GeTe and ${\mathrm{Ge}}_1{\mathrm{Sb}}_2{\mathrm{Te}}_4$ in the two phases. The analysis of our computations and experimental data reveal the correlation between local structural changes and optical properties as well as the origin of the optical contrast in these materials. We find that the change in optical properties cannot be attributed to a smearing of transition energies as commonly assumed for amorphous semiconductors: the optical contrast between the two phases can only be explained by significant changes in the transition matrix elements.

Figure 1

Optical absorption of GeTe and ${\mathrm{Ge}}_1{\mathrm{Sb}}_2{\mathrm{Te}}_4$ (a): experimental, $(\mathrm{b})+(\mathrm{c})$: calculation, here two amorphous models for ${\mathrm{Ge}}_1{\mathrm{Sb}}_2{\mathrm{Te}}_4$ are displayed. For both alloys a decrease, a broadening and a blueshift of the absorption, is reproduced for the $a$ state. Lattice distortions upon structural relaxation and Ge vacancies lead to a better agreement with the experiment in $c$ GeTe.

Figure 2

(a) $\mathrm{JDOS}/{\omega }^2$ of ${\mathrm{Ge}}_1{\mathrm{Sb}}_2{\mathrm{Te}}_4$ in the $c$- and the $a$ state in number of $\mathrm{\text{transitions}}/{\mathrm{eV}}^3$ per cell and $k$ point. Up to 1.4 eV the JDOS in the $c$ phase is stronger than in the $a$ state, while ${ϵ}_2$ is stronger in the $c$ phase up to 1.8 eV. For GeTe (c) up to 1.7 eV more transitions are found in the $a$ state compared to the relaxed $c$ phase. For both materials the decrease in the absorption upon amorphization can only be explained taking into account the velocity matrix elements shown in (b) (${\mathrm{Ge}}_1{\mathrm{Sb}}_2{\mathrm{Te}}_4$) and (d) (GeTe) with $M_{v,c,\mathbf{k}}=2\frac{4{\pi }^2}{\Omega }\underset{\mathbf{q}\to 0}{lim}\frac{1}{{\mathbf{q}}^2}|{\mathbf{m}}_{v,c,\mathbf{k}}{|}^2$.

Figure 3

The number of matrix elements for excitation energies up to 2.5 eV vs the overlap $\int |{\varphi }_c(\mathbf{r})||{\varphi }_v(\mathbf{r})|d\mathbf{r}$.