Magnetic anisotropy is a difficult topic to describe by electronic structur
e theory. The results of ab initio calculations for iron and nickel are dis
appointing, especially since these calculations require high numerical prec
ision and are very time consuming. For iron the value of the energy is too
small by a factor of 3, while for nickel the sign is wrong and the value is
too small by a factor of 5. The local density approximation can be improve
d by adding corrections that mimic the inclusion of Hund's second rule. Thi
s is equivalent to increasing the effective spin-orbit parameter. For iron
a small increase in the effective spin-orbit parameter is sufficient to rec
oncile theory and experiment. For nickel, this is not possible. (C) 2000 Am
erican Institute of Physics. [S0021-8979(00)66108-X].