The nature of the band gap in the semiconducting material $\beta -{\mathrm{FeSi}}_2$ is still under some dispute. Although most experimental results indicate the band gap to be direct, ab initio work generally reports the material to be an indirect semiconductor with the direct transition a few tens of millielectron volts higher than the indirect gap. However, $\beta -{\mathrm{FeSi}}_2$ is commonly grown epitaxially on a diamond-structure Si substrate, and as a consequence, the $\beta -{\mathrm{FeSi}}_2$ unit cell is strained. Here we report the results of ab initio density-functional calculations, which we have performed on $\beta -{\mathrm{FeSi}}_2$ where its lattice parameters are constrained according to the heteroepitaxial system $\beta -{\mathrm{FeSi}}_2\mathrm{}\left(100\right)/\mathrm{Si}\mathrm{}\left(001\right).$ This forms two types of lattice matching: (A) $\beta -{\mathrm{FeSi}}_2\left[010\right]||\mathrm{Si}〈110〉$ and (B) $\beta -{\mathrm{FeSi}}_2\left[010\right]||\mathrm{Si}〈001〉.$ We find that the $\beta -{\mathrm{FeSi}}_2$ band gap is highly sensitive to its lattice parameters and therefore to the orientation at which the material is grown on silicon. We find that type A favors a more direct band gap, while type B has an indirect gap.

• Received 13 May 1998

DOI:https://doi.org/10.1103/PhysRevB.58.10389