On the inequalities of Babuška-Aziz, Friedrichs and Horgan-Payne (Q2354696)

The authors consider the smoothness of a domain \(\Omega\) necessary for several relations appearing in a paper by \textit{C. O. Horgan} and \textit{L. E. Payne} [Arch. Ration. Mech. Anal. 82, 165--179 (1983; Zbl 0512.73017)]. First, they investigate the smoothness of \(\Omega\) needed to ensure the relation \(C(\Omega)=1+\Gamma(\Omega)\), where \(C(\Omega)\) is the smallest constant in \(\|u\|^2_{1,\Omega}\leq C\|q\|^2_{0,\Omega}\) for solutions \(u\in H^1_0(\Omega)\) of \(\mathrm{div}(u)=q\). Moreover, \(C(\Omega)=1/\beta(\Omega)^2\) holds for the inf-sup constant \(\beta(\Omega)\) which is important, e.g., for pressure stability in hydromechanics, for proving the Korn inequality in elasticity, for the well-posedness of the Stokes problem and for its finite element approximations and their numerical solution. On the other side, \(\Gamma(\Omega)\) is the best constant in an inequality considered by Friedrichs between conjugate harmonic functions. The relation between this \(\Gamma\) and \(C\) was proved originally for bounded smooth domains in two dimensions, later, by other authors, for Lipschitz and more general domains in any dimension. Here, the present authors show the relation to be true independent on smoothness in dimension 2, further, if one constant is \(\infty\) for certain domains, then the other is \(\infty\), too. For two-dimensional domains, an upper bound for \(F\) also had been given in the paper by Horgan and Payne for star-shaped domains with respect to a ball. This means an upper bound for the constant \(C\) (and a lower bound for \(\beta\)). Here, the authors give examples of star-shaped domains (with respect to a ball) for which the upper bound by Horgan and Payne is not true but prove that it holds for a number of domains (like regular polygons). They give a new upper bound for all two-dimensional star-shaped domains with respect to a ball. Finally, they mention open questions like the minimal smoothness needed for the relation \(K(\Omega)=2C(\Omega)\) to hold which also appears in the paper of Horgan and Payne [loc. cit.].