Contrary to a broadly accepted assumption we show that random (Ga,In)P is not an exception with respect to the crude classification of the phonon mode behavior of random mixed crystals in terms of $1\text{−}\text{bond}\to 1\text{−}\text{mode}$ systems or $2\text{−}\text{bond}\to 1\text{−}\text{mode}$ systems, as established from the simple criterion derived by Elliott et al. [R. J. Elliott et al., Rev. Mod. Phys. 46, 465 (1974)]. Consistent understanding of the puzzling Raman/infrared behavior of (Ga,In)P, that has been a subject of controversy, is achieved via a basic version of our $1\text{−}\text{bond}\to 2\text{−}\text{mode}$ model originally developed for (Zn,Be)-chalcogenides, that exhibit a large contrast in the bond properties, and recently extended under a simplified form to the usual (Ga,In)As alloy. The Raman/infrared features from (Ga,In)P are accordingly re-assigned, with considerable change with respect to the previous approaches. In particular the In impurity mode, previously assigned within $(\sim 390{\mathrm{cm}}^{-1})$ the optical band of the host GaP compound $(368–403{\mathrm{cm}}^{-1})$, is re-assigned below it $(\sim 350{\mathrm{cm}}^{-1})$. Accordingly the Ga-P and In-P transverse optical branches do not overlap, which reconciles (Ga,In)P with the Elliott’s criterion. Besides, we show that the idea of two bond lengths per species in alloys, supported by our $1\text{−}\text{bond}\to 2\text{−}\text{phonon}$ picture, opens an attractive area for the discussion of spontaneous ordering in $\mathrm{Ga}\mathrm{In}{\mathrm{P}}_2$, and mixed crystals in general. Essentially this is because it allows to play with the related competition effects regarding the minimization of the local strain energy due to the bond length mismatch between the parent compounds. In particular the unsuspected issue of intrinsic limit to spontaneous ordering comes out ($\eta \sim 0.5$ in $\mathrm{Ga}\mathrm{In}{\mathrm{P}}_2$). The whole discussion is supported by detailed re-examination of the (Ga,In)P Raman/infrared data in the literature, full contour modeling of the transverse and longitudinal optical Raman lineshapes via our phenomenological $1\text{−}\text{bond}\to 2\text{−}\text{mode}$ model, and first-principles bond length calculations concerned with the minority bond species close to the impurity limits ($\mathrm{Ga}\sim 0$, 1) and to the Ga-P $(\mathrm{Ga}\sim 0.19)$ and In-P $(\mathrm{Ga}\sim 0.81)$ bond percolation thresholds. In the latter case we discriminate between connected and isolated bonds, not in the usual terms of next-nearest neighbors.

• Received 15 September 2005

DOI:https://doi.org/10.1103/PhysRevB.73.165206