Abstract
The three-magnon confluence relaxation frequency 1τ is calculated without making assumptions of the previous calculation of Sparks, Loudon, and Kittel. The results are such that the discrepancies between the experiments of Comstock, LeCraw, Nilsen, Remeika, Spencer, and Walker and the previous theory are removed. In particular, the exponentially small (rather than linear) dependence of 1τ on k1 and T is explained by the new theory. A calculation of the bending down of 1τ below linearity in k1 at large values of k1 is also given.