Lower bounds of Dirichlet eigenvalues for some degenerate elliptic operators (Q889731)

In this paper, the authors consider the eigenvalue problem \[ \begin{cases} -\Delta_X u = \lambda u \quad & \text{in } \Omega, \\ u= 0 \quad &\text{on } \partial\Omega, \end{cases} \eqno{(1.1)} \] in \(H^1_{X,0}(\Omega)\), where the real smooth vector field \(X = (X_1, X_2, \ldots , X_m)\) is defined on an open domain \(W\) in \(\mathbb R^n\) and \(\Omega \subset W\) is a bounded open subset with smooth boundary \(\partial \Omega\) which is non-characteristic for \(X\). To prove the main results, the authors use the following inequality. \(\quad\) { Proposition 2.1.} The system of vector fields \(X = (X_1, \ldots, X_m)\) satisfies Hörmander's condition on \(\Omega\) with Hörmander index \(Q \geq 1\) if and only if the subelliptic estimate \[ \||\nabla|^{1/Q}u\|^2_{L^2(\Omega)} \leq C(Q) \left( \|Xu\|^2_{ L^2(\Omega)}+\widetilde{C}(Q)\|u\|^2_{L^2(\Omega)}\right) \eqno{(2.1)} \] holds for all \(u \in C^\infty_0 (\Omega )\), where \(\nabla = (\partial_{x_1}, \ldots , \partial_{x_n}), |\nabla|^{1/Q}\) is a pseudo-differential operator with symbol \(|\xi |^{1/Q}\), and the constants satisfy \(C(Q) > 0\) and \(\widetilde{C}(Q) \geq 0\). \(\quad\) Let \(B_n\) be the volume of the unit ball in \(\mathbb R^n\), \(|\Omega |_n\) the volume of \(\Omega\), and \(\omega_{n-1}\) the area of the unit sphere in \(\mathbb R^n\) and let \(\lambda_j\) be the \(j\)-th Dirichlet eigenvalue for \(-\Delta_X\) on \(\Omega\). \(\quad\) First, by using the sub-elliptic estimate directly, the authors give a lower bound on \(\lambda_j\) for general finitely degenerate elliptic operators. \(\quad\) { Theorem 1.1} \, Let \(X\) be finitely degenerate on \(\Omega\) with Hörmander index \(1 \leq Q < +\infty\). Then, \[ \sum^k_{j=1} \lambda_j \geq C_1k^{1+\frac{2}{Qn}} - \widetilde{C}(Q)k \quad\text{for all } k \geq 1, \] where \[ C_1 = \frac{nQ(2\pi)^{\frac{2}{Q}}}{C(Q)\cdot(nQ+2)(|\Omega |_n B_n)^{\frac{2}{nQ}}}, \] and \(C(Q)\) and \(\widetilde{C}(Q)\) are the constants in Proposition 2.1. \(\quad\) Second, if \(\Delta_X\) is a so-called Grushin-type degenerate elliptic operator, then authors give a precise lower bound for \(\lambda_j\). \(\quad\) { Theorem 1.2} \, Let \(X = (\partial_{x_1}, \ldots , \partial_{x_{n-1}} , x_1^l \partial_{x_n})\), \(l \in \mathbb Z^+\), \(n \geq 2\), be a Grushin-type vector field on \(\Omega\) and \(\Omega \cap \{x_1 = 0\}\not= \emptyset\). Then, \(X\) satisfies Hörmander's condition with Hörmander index \(Q = l + 1\). Also, the generalized Métivier index \(\tilde{\nu} = Q + n - 1\). Then, \[ \sum^k_{j=1} \lambda_j \geq C_2k^{1+\frac{2}{\tilde{\nu}}} -\widetilde{C}(Q)k\quad\text{for all } k \geq 1, \] where \[ C_2 = \frac{A_Q}{\widehat C(Q)n(n+Q+1)}\left( \frac{(2\pi)^n}{|\Omega|_n\omega_{n-1}Q}\right)^{\frac{2}{n+Q-1}}(n+Q-1)^{\frac{n+Q+1}{n+Q-1}} \] and \[ \widehat{C}(Q) = C(Q) + \min\{1, Q - 1\} > 0,\quad A_Q = \left\{ {\min\{1, n^{\frac{3-Q}{2}}\}, \; Q \geq 2,}\atop{ n, \qquad\qquad\quad \;\, Q = 1,} \right. \] where \(C(Q)\) and \(\widetilde{C}(Q)\) are the constants in Proposition 2.1. \(\quad\) In the third part of the paper, the authors study the problem with \(X\) being an infinitely degenerate vector field and they suppose that the infinitely degenerate vector field \(X\) satisfies the logarithmic regularity estimate \[ \|(\log \Lambda)^su\|^2_{L^2(\Omega )}\leq C_0 \left[\int_\Omega |Xu|^2\,dx + \|u\|^2_{L^2(\Omega )}\right]\quad\text{for all } u \in C^\infty_0 (\Omega ), \eqno{(1.9)} \] where \(s > 1, C_0 > 0\) and \(\Lambda = (e^2 + |\nabla|^2)^{\frac{1}{2}}\). The next result tells us that in this case the Dirichlet eigenvalues \(\lambda_k\) will be at least logarithmic increasing in \(k\). \(\quad\) {Theorem 1.4} Suppose that the infinitely degenerate vector field X satisfies the logarithmic regularity estimate (1.9) with \(s > 1\). Then, \[ \sum^k_{j=1} \lambda_j \geq C_4k(\log k)^{2s} - k\quad\text{for all } k \geq k_0, \] where \[ k_0 = [\frac{2^{2s} e^n B_n |\Omega |_n}{ C_0\pi^n} ] + 1, \] \[ C_4 = (2^n - 1)\left(C_02^{n+4s} (| \log \frac{|\Omega |_n B_n}{(2\pi)^n} |^{2s} + n^{2s} )\right)^{-1}, \] and \(s\) and \(C_0\) are given in (1.9).