Asymptotic behavior of the Dirichlet energy on Poisson point clouds (Q6168863)

The authors consider the family \(N_{s}\) of simple measures \( N_{s}=\{\sum_{i\in I\subset \mathbb{N}}\delta _{x_{i}}\in \mathcal{M}^{+}( \mathbb{R}^{2})\mid \{x_{i}\}_{i\in I}\in \mathbb{R}^{2}\), \(x_{i}\neq x_{j}\) for all \(i,j\in I\), with \(i\neq j\}\), where \(\delta _{x}\) is the Dirac delta at \(x\), and the \(\sigma \)-algebra \(\mathcal{N}\) generated by \(\{A_{E,k}\mid E\in Bor(\mathbb{R}^{2})\), \(k\in \mathbb{N}\}\), where \(A_{E,k}=\{\mu \in N_{s}\mid \mu (E)=k\}\). A Poisson point process on \(\mathbb{R}^{2}\) with intensity \(\gamma \) is a random element \(\eta \) on \((N_{s},\mathcal{N})\), that is a map from a probability space \((\Omega ,\mathcal{F},\mathbb{P})\) onto \((N_{s},\mathcal{N})\), such that \(\mathbb{P}(\eta \in A_{E,k})=\frac{ (\gamma \left\vert E\right\vert )^{k}}{k!}e^{\gamma \left\vert E\right\vert } \), and if \(\eta (B):(\Omega ,\mathcal{F},\mathbb{P})\rightarrow N\) is the random variable induced by \(\eta \) through \(\eta (B)=\eta (\omega )(B)\), for \(\omega \in \Omega \), for any pairwise disjoint Borel sets \(B_{1},\ldots ,B_{m}\), then \(\eta (B_{1}),\ldots ,\eta (B_{m,})\) are independent. Let \( \eta \) be a Poisson point process with intensity \(\gamma =1\) without loss of generality, the authors define the Dirichlet energy as: \(\mathcal{F} _{\varepsilon }(u;A)=\sum_{x\in \eta _{\varepsilon }\cap A}\sum_{y\in \eta _{\varepsilon }\cap B_{\lambda \varepsilon }(x)}\left\vert u(x)-u(y)\right\vert ^{2}\), where \(u\in L^{2}(Q;\eta _{\varepsilon })\), \( A\subset Q\) the unit square, \(\eta _{\varepsilon }(A)=\eta (\varepsilon ^{-1}A)\), \(\varepsilon ,\lambda >0\). They establish properties of this energy functional. The authors prove statistical properties of Poisson point processes and they define a notion of convergence for functions on Poisson point clouds. The first main result is a compactness theorem. If \( \{u_{\varepsilon }\in L^{2}(U;\eta _{\varepsilon })\}_{\varepsilon >0}\) is a sequence satisfying \(sup_{\varepsilon >0}\{\sum_{x\in \eta _{\varepsilon }\cap A}\sum_{y\in \eta _{\varepsilon }\cap B_{\lambda \varepsilon }(x)}\left\vert u(x)-u(y)\right\vert ^{2}+\sum_{x\in \eta _{\varepsilon }\cap Q}\varepsilon ^{2}u_{\varepsilon }^{2}(x)\}<+\infty \), where \(U\supset A\supset Q\) is any open set strictly containing \(Q\), there exists \(u\in W^{1,2}(Q)\) such that \(u_{\varepsilon }\) converge to \(u\), up to subsequences. Here \(\lambda \) is greater than some \(\lambda _{0}>0\) appropriately chosen. For the proof, the authors use Theorem 3.1 proved by \textit{R. Alicandro} and \textit{M. Cicalese} [SIAM J. Math. Anal. 36, No. 1, 1--37 (2004; Zbl 1070.49009)]. The second main result is a \( \Gamma \)-convergence result expressed in the following way. There exists a deterministic constant \(\Xi \) such that, for almost all realizations \(\omega \), the energy \(\mathcal{F}_{\varepsilon }(u;Q)\) \(\Gamma \)-converges to \( \mathcal{F}(u;Q)=\Xi \int_{Q}\left\vert \nabla u(x)\right\vert ^{2}dx\), in the topology induced by the convergence of sequences. For the proof, the authors build a cell problem and they separately prove the \(limsup\) and \( liminf\) inequalities using appropriate cut-off functions and the preceding compactness theorem.