The accuracy of different total-energy methods to compute the properti
es of multicomponent oxides is studied. These materials have typically
large unit cells and consequently, computer-running time consideratio
ns become, important. We show that while highly sophisticated quantum-
mechanics techniques such as pseudopotentials or the full-potential li
nearized-augmented-plane-wave method can be used to accurately compute
materials properties, they may require prohibitively long computer ru
ns in oxides, On the other hand, simple potential models, or even fast
quantum-mechanics methods such as the spherical self-consistent atomi
c deformation or the linear muffin-tin orbital method (in the atomic s
phere approximation), are not always reliable to study oxides, Charge
transfer, breathing of the oxygen ions, and nonspherical charge relaxa
tions are some of the factors that can make any of these schemes fail.
However, it is not necessary to always use sophisticated techniques.
Ne show that the self-consistent tight-binding formalism can be used a
s an interpolation tool to extend the results of accurate calculations
for a few compounds in a system to the rest of them. This opens new p
ossibilities for the use of ab initio methods to study technologically
-relevant materials properties, such as the temperature behavior of ox
ides, since formation energies of many different compounds al 0 K art:
a crucial input to these models.