Following the parametrization scheme first introduced by Goodwin, Skinner,
and Pettifor, we propose here a model of the empirical tight-binding Hamilt
onian for selenium, based on the fitting of cohesive energy curves obtained
from density-functional calculations for solid phases, rings, and chains s
tructures. We have assessed the model by means of various tight-binding mol
ecular-dynamics calculations performed in liquid and amorphous states. Comp
arisons with ab initio calculations and experimental results indicate that
the model is fairly accurate for the pair structure at low and medium tempe
ratures, but tend to overestimate bonding at high temperatures. The number
of valence alternation pair defects also seems to be overestimated. On the
other hand, the band structure derived from the tight-binding density of st
ates predicts the occurrence of a semiconductor-to-metal transition when ap
proaching the critical temperature in good agreement with ab initio calcula
tions and experimental evidence. [S0163-1829(99)02333-4].