The bond-orbital model is reformulated taking explicit account of the large overlap of two hybrids in the same bond rather than absorbing it in a pseudopotential. This gives a correction to the cohesive energy but other changes are absorbed in the $V_1$, $V_2$, and $V_3$ for each material. These parameters are reevaluated taking band calculations as the standard for $V_1$ (but using the atomic term values to scale from material to material) and taking the energy of the optical-absorption peak as the standard for $V_2$ and $V_3$ (but using the dielectric constant to scale from material to material). Using the new parameters we test the model by comparison of predictions with experiment (or accurate calculations) for various valence-band gaps, the pressure dependence of the dielectric constant, the macroscopic transverse charge, the piezoelectric constant, the photoelectric threshold, and the cohesive energy. Discrepancies in the cohesive energy are found to scale with a reasonable form for the interbond correlation energy, giving a semiempirical expression for the cohesive energy in terms of parameters of the bond-orbital model. The covalent energy is found in this study to scale with a kinetic energy (with the inverse square of the bond length) both under pressure and for variation from material to material.

• Received 8 April 1974

DOI:https://doi.org/10.1103/PhysRevB.10.1516