Title: Optimal inequalities in multiplication of derivatives of positive definite matrices and their powers

Abstract: The simple formulae for a product of derivatives of a scalar function raised to different exponents are generalized for the functions which take values in the set of symmetric positive definite matrices. These formulae are fundamental in derivation of various non-linear estimates, especially in the PDE theory. To get around the non-commutativity of the matrix and its derivative, we use an observation of independent interest: The derivative of a matrix power is, in a suitable sense, a logarithmically convex (possibly strictly) function with respect to the exponent. This is directly related to the validity of a certain simple-looking inequality combining the integral averages and the inner product on matrices. The optimality of our results is illustrated on numerous examples.