Abstract
A well-known method of computing magnetostrains, as the strains necessary to minimize the sum of magnetic and elastic potential energies, is improved in detail, extended to include terms dependent upon the structure of atomic magnets, and is applied in the cases of body-centered cubic iron, face-centered cubic nickel and hexagonal close-packed cobalt. It is shown that magnetostriction is not important in fixing directions of easiest magnetization in these crystals.