Bounded deformation

In mathematics, a function of bounded deformation is a function whose distributional derivatives are not quite well-behaved-enough to qualify as functions of bounded variation, although the symmetric part of the derivative matrix does meet that condition. Thought of as deformations of elasto-plastic bodies, functions of bounded deformation play a major role in the mathematical study of materials, e.g. the Francfort-Marigo model of brittle crack evolution.

More precisely, given an open subset Ω of Rⁿ, a function u : Ω → Rⁿ is said to be of bounded deformation if the symmetrized gradient ε(u) of u,

${\displaystyle \varepsilon (u)={\frac {\nabla u+\nabla u^{\top }}{2}}}$

is a bounded, symmetric n × n matrix-valued Radon measure. The collection of all functions of bounded deformation is denoted BD(Ω; Rⁿ), or simply BD, introduced essentially by P.-M. Suquet in 1978. BD is a strictly larger space than the space BV of functions of bounded variation.

One can show that if u is of bounded deformation then the measure ε(u) can be decomposed into three parts: one absolutely continuous with respect to Lebesgue measure, denoted e(u) dx; a jump part, supported on a rectifiable (n − 1)-dimensional set J_u of points where u has two different approximate limits u₊ and u₋, together with a normal vector ν_u; and a "Cantor part", which vanishes on Borel sets of finite Hⁿ⁻¹-measure (where H^k denotes k-dimensional Hausdorff measure).

A function u is said to be of special bounded deformation if the Cantor part of ε(u) vanishes, so that the measure can be written as

${\displaystyle \varepsilon (u)=e(u)\,\mathrm {d} x+{\big (}u_{+}(x)-u_{-}(x){\big )}\odot \nu _{u}(x)H^{n-1}|J_{u},}$

where H ⁿ⁻¹ | J_u denotes H ⁿ⁻¹ on the jump set J_u and ${\displaystyle \odot }$ denotes the symmetrized dyadic product:

${\displaystyle a\odot b={\frac {a\otimes b+b\otimes a}{2}}.}$

The collection of all functions of special bounded deformation is denoted SBD(Ω; Rⁿ), or simply SBD.