On the best constant in {G}affney inequality
Gyula Csato, Bernard Dacorogna, Swarnendu Sil
TL;DR
This work identifies the best constant in Gaffney's inequality for differential forms and links its vanishing to geometric boundary convexity. Using boundary-operator expressions $L^{ν}$ and $K^{ν}$ tied to principal curvatures, it proves that $C_T(Ω,k)=1$ is equivalent to the domain being $(n-k)$-convex, the sharper inequality, non-attainment of the supremum, and scale invariance, with a dual, analogous result for $C_N$. The authors develop algebraic and geometric tools to express boundary contributions purely in terms of boundary geometry, and they show that in general non-convex domains the constants can be arbitrarily large, while for polytopes the inequality is exact with constant 1. The results extend to piecewise-smooth and polyhedral domains via integral identities and curvature sums, linking spectral-type inequalities to boundary geometry and providing sharp Korn-type corollaries in special cases.
Abstract
We discuss the value of the best constant in Gaffney inequality namely $$ \lVert \nabla ω\rVert_{L^{2}}^{2}\leq C\left( \lVert dω\rVert_{L^{2}}^{2}+\lVert δω\rVert_{L^{2}% }^{2}+\lVert ω\rVert_{L^{2}}^{2}\right) $$ when either $ν\wedgeω=0$ or $ν\,\lrcorner\,ω=0$ on $\partialΩ.$
