ACCURACY OF THE CHEMICAL-PSEUDOPOTENTIAL METHOD FOR TETRAHEDRAL SEMICONDUCTORS

Simple two-center tight-binding models have often been used in electro nic-structure calculations, although uncertainty about the foundations of the tight-binding method has severely limited progress. This uncer tainty has been much reduced following the recent successes of the non -self-consistent Harris scheme but several important questions still r emain. In particular, the accuracy of the (almost ubiquitous) two-cent er approximation is still in doubt. It has often been argued that alth ough this is a poor approximation when applied crudely, it can be just ified within the more sophisticated context of chemical-pseudopotentia l theory. The argument looks sensible, but clear quantitative tests on ly became possible with the advent of the Harris scheme. This paper re ports a careful series of tests of the accuracy of the two-center appr oximation for germanium, and shows that neither the crude version nor the more sophisticated chemical-pseudopotential version is accurate en ough to be useful. The remarkable success of two-center tight binding in predicting the structures and surface reconstructions of semiconduc tors is, therefore, still mysterious.