On Courant's nodal domain property for linear combinations of eigenfunctions. I (Q1617754): Difference between revisions

The authors consider the following eigenvalue problem \[ (\Omega,\mathfrak b):-\Delta u=\lambda u\text{ in }\Omega\text{ and }\mathfrak b(u)=0\text{ on }\partial \Omega, \] such that \(\mathfrak b\) is either the Dirichlet or the Neumann classical boundary condition, \(\Omega\) is a (compact) bounded domain and \(\partial\Omega\) represents the boundary of \(\Omega\). Let us denote by \(\mathcal E(\Omega,\mathfrak b,\lambda)\) the eigenspace associated to \(\lambda(\Omega,\mathfrak b)\), an eigenvalue of \((\Omega,\mathfrak b)\). The index of \(\lambda(\Omega, \mathfrak b)\), denoted by \(\kappa(\lambda(\Omega,\mathfrak b))\), is the minimum of \(k\) such that \(\lambda_k(\Omega,\mathfrak b)\), the \(k\)th eigenvalue of \((\Omega,\mathfrak b)\), is equal to \(\lambda(\Omega,\mathfrak b)\). The nodal set of a continuous function \(v\) on \(\Omega\), denoted by \(\mathcal Z(v)\), is the closure of the set \(v^{-1}(\{0\})\) and \(\beta_0(v)\) is the number of nodal domains of \(v\), i.e. the number of connected components of the complement of \(\mathcal Z(v)\) in \(\Omega\). We denote by \(\mathcal L(\lambda(\Omega,\mathfrak b))\) the space of linear combinations of eigenfunctions associated to eigenvalues of \((\Omega,\mathfrak b)\) which are less or equal to \(\lambda(\Omega,\mathfrak b)\). The \textit{Extended Courant Property} associated to \((\Omega,\mathfrak b)\) (\(\mathrm{ECP}(\Omega,\mathfrak b)\) for short) states that \(\beta_0(v)\leq\kappa(\lambda_n(\Omega,\mathfrak b))\leq n\) for \(v\in \mathcal L((\lambda_n(\Omega,\mathfrak b))\). Then, the authors state that \(\mathrm{ECP}(\Omega,\mathfrak b)\) is invalid for specific sets. Precisely, the authors affirm that \(\mathrm{ECP}(\Omega,\mathfrak b)\) is untrue, no matter the boundary condition, when \(\Omega\) is one of the following set: a hypercube, a regular parallelepiped, an equilateral triangle. Next, relying on the boundary condition, the authors state that \(\mathrm{ECP}(\Omega,\mathfrak b)\) is invalid when \(\Omega\) is one of the following set with cracks: a rectangle, the flat torus, the round two-sphere.