Inclusion relations of Gårding's cones and \(k\)-convex cones (Q2302014)

The authors prove different properties of Gårding's cones defined as \(\Gamma_{k}=\{\lambda \in \mathbb{R}^{n}\mid S_{j}(\lambda)>0\), \(1\leq j\leq k\}\), for \(k=1,\dots ,n\), \(\lambda =(\lambda_{1},\dots\lambda_{n})\), where \(S_{j}(\lambda)\) is the \(j\)th-order elementary symmetric polynomial given by \(S_{j}(\lambda)=\sum_{i_1,\dots,i_j}\prod_{i=1}^{j}\lambda_{i_{s}}\) where \(i_{1},\dots ,i_{j}\in \{1,\dots ,n\}\), and of the \(k\)-convex cones defined as \(\mathcal{P}_{k}=\{\lambda \in \mathbb{R}^{n}\mid \sum_{s=1}^{k}\lambda_{i_{s}}>0\}\), for \(k=1,\dots ,n\), \(\lambda =(\lambda_{1},\dots \lambda_{n})\), \(i_{1},\dots ,i_{j}\in \{1,\dots ,n\}\), and \(i_{s}\neq i_{t}\) if \(s\neq t\). The authors recall that \(K^{+}\subset \Gamma_{n}\subset \Gamma_{n-1}\subset \dots \subset \Gamma_{1}\subset H^{+}\) and \(K^{+}\subset \mathcal{P}_{1}\subset \mathcal{P}_{2}\subset \dots\subset \mathcal{P}_{n}\subset H^{+}\), with \(K^{+}=\{\lambda \in \mathbb{R}^{n}\mid \lambda_{i}>0\), \(i=1,\dots ,n\}\) and \(H^{+}=\{\lambda \in \mathbb{R}^{n}\mid \sum_{s=1}^{n}\lambda_{i}>0\}\). The main results of the paper prove that for \(2\leq n\leq 4\), \(K^{+}=\Gamma_{n}=\mathcal{P}_{1}\subset \Gamma_{n-1}\subset \mathcal{P}_{2}\dots \subset \mathcal{P}_{n-1}\subset \Gamma_{1}=\mathcal{P}_{n}=H^{+}\), that for \(n\geq 5\), \(\mathcal{P}_{k}\varsubsetneq \Gamma_{n-k}\), for \(k=2,\dots ,n-2\), that for \(n\geq 2\), \(\Gamma_{n-k}\varsubsetneq \mathcal{P}_{k}\), for \(k=1,\dots ,n-1\), and that for \(n\geq 2\), \(\Gamma_{n-k+1}\subset \mathcal{P}_{k}\), for \(k=1,\dots ,n\), \(\Gamma_{n}=\mathcal{P}_{1}=K^{+}\), \(\Gamma_{1}=\mathcal{P}_{n}=H^{+}\) and \(\Gamma_{n-k+1}\varsubsetneq \mathcal{P}_{k}\), for \(k=2,\dots ,n-1\). The proofs of the results are obtained essentially through direct computations. The paper ends with an example in \(\mathbb{R}^{5}\) and with further considerations on fully nonlinear partial differential equations of the type \(f(\lambda (M[u]))=B\) posed in a domain \(\Omega\subset \mathbb{R}^{n}\). Here, \(M[u]=D^{2}u-A(\cdot ,Du)\) is the augmented Hessian matrix and \(\lambda (M[u])\) denotes the eigenvalues of the matrix \(M[u]\). The authors recall the link between admissibility of \(C^{2}(\Omega)\)-solutions and ellipticity.