Abstract
The integrals are estimated, considering the following contributions. (i) Deviations from the muffin-tin potential assumed in Mattheiss' APW calculation, both outside and inside the muffin-tin spheres. (ii) Lack of selfconsistency. Changes in the populations of the various sub-bands of the 3d band cause changes of these integrals. (iii) Different approximations for the exchange interaction, such as the Slater or Kohn-Sham rho 1/3 approximations are shown to have only small effect. (iv) The effect of screening by the vanadium 4s band is estimated. (v) The effect of covalency with the gallium 4p band is estimated and shown to be small. A comparison of the crystal-field integrals estimated this way, with those guessed empirically, shows differences of the order of 20-40 mRyd. A proper selfconsistent calculation is probably required to get a still better accuracy.