Optimality of the relaxed polar factors by a characterization of the set of real square roots of real symmetric matrices

TL;DR

This paper characterizes the set of real square roots of symmetric matrices to determine optimal rotations minimizing a Cosserat shear-stretch energy, confirming formulas in 2D and 3D and introducing new matrix root characterizations.

Contribution

It provides a novel characterization of all real square roots of symmetric matrices, crucial for identifying optimal Cosserat rotations in nonlinear continuum models.

Findings

Characterization of all real square roots of symmetric matrices.

Proof of optimal Cosserat rotations in 2D and 3D.

Validation of previously derived formulas for optimal rotations.

Abstract

We consider the problem to determine the optimal rotations $R \in {\rm SO}(n)$ which minimize $$W: {\rm SO}(n) \to \mathbb{R}^+_0,\quad W(R\,;D) := ||{\rm sym}(RD - 1)||^2$$ for a given diagonal matrix $D := {\rm diag}(d_1, ..., d_n) \in \mathbb{R}^{n \times n}$. The function $W$ subject to minimization is the reduced form of the Cosserat shear-stretch energy, which, in its general form, is a contribution in any geometrically nonlinear, isotropic and quadratic Cosserat micropolar (extended) continuum model. We characterize the critical points of the energy $W(R\,;D)$, determine the global minimizers and the global minimum. This proves the correctness of previously obtained formulae for the optimal Cosserat rotations in dimensions two and three. The key to the proof is a characterization of the entire set of (possibly non-symmetric) real matrix square roots of (possibly non-positive definite) real symmetric matrices which does not seem to be known in the literature.