Interlacing inequalities for eigenvalues of discrete Laplace operators (Q1941761): Difference between revisions

The Laplacian matrix of a graph \(G\) is the matrix \(L(G) = QQ^t\) where \(Q\) is the vertex-edge incidence matrix of \(G\): rows are indexed by vertices \(v_i\), columns are indexed by edges \(e_j\) and \(q_{ij} = \pm 1\) (following an orientation previously chosen) if \(e_i\) is an end of \(v_j\), while \(q_{ij}= 0\) otherwise. By using classical results on interlacing spectra of hermitian matrices, various results have been obtained concerning interlacing of the eigenvalues of \(L(G)\) with the eigenvalues of \(L(H)\) where \(H\) is a graph obtained from \(G\) by removing an edge or a vertex; others interlacing results concern the normalized Laplacian and the case of weighted graphs ; we refer to the paper under review for statements and references. A graph may be seen as a 1-dimensional simplicial complex and, in this paper, the authors generalize considerably these results by considering weighted simplicial complexes. For any simplicial complex \(K\), the incidence matrices with \(k\)-dimensional faces in rows and \((k+1)\)-dimensional faces in columns give higher dimensional Laplacian matrices which generalize the usual construction in graphs and, following the discrete Hodge approach initiated by \textit{B. Eckmann} [Comment. Math. Helv. 17, 240--255 (1945; Zbl 0061.41106)], the Laplace operator decomposes into two parts: \({\mathcal L}_k(K)={\mathcal L}_k^{\mathrm{up}}(K)+{\mathcal L}_k^{\mathrm{down}}(K)\); most of the results in this paper will be about Laplace operators \( {\mathcal L}_k^{\mathrm{up}}(K)\) whose spectrum determines the spectrum of the two others as it is explained in the companion paper [\textit{D. Horak} and \textit{J. Jost}, Adv. Math. 244, 303--336 (2013)]. The main technical result is an interlacing result (Theorem 1.1) between the eigenvalues of \({\mathcal L}_n^{\mathrm{up}}(K)\) and that of \({\mathcal L}_n^{\mathrm{up}}(L)\) where \(L\) is a weighted \((n+1)\)-dimensional subcomplex of a weighted \((n+1)\)-dimensional simplicial complex \(K\). It is well known that the Laplace operator encode interesting information about the geometry of the space on which it us defined and, as emphasized by the authors, their result gives \(\scriptscriptstyle\langle\!\langle\) a systematic scheme how to derive such analytical inequalities from topological considerations \(\scriptscriptstyle\,\rangle\!\rangle\). More precisely, the inequalities obtained are given by a homological type information. For example, if \({\mathcal L}_n^{\mathrm{up}}(K)\) has spectrum \(\lambda_1\leq \lambda_2 \leq \dots \leq \lambda_N\) and if \(L\) is the proper difference of \(K\) and \(H\) (where \(H\) is a weighted \((n+1)\)-dimensional subcomplex of \(K\)), then the eigenvalues \(\theta_1\leq \theta_2 \leq \dots \leq \theta_N\) of \({\mathcal L}_n^{\mathrm{up}}(L)\) verify: \[ \theta_k \leq \lambda_{k+D_H} \] for \(k=1,2,\dots,N-D_H\) with \(D_H = \dim C^n(H, \mathbb{R})\) where \(C^*(H, \mathbb{R})\) denotes the cochain groups of \(H\) with real coefficients (we refer to the paper for all missing definitions and to Theorem 1.1 for the complete set of inequalities). By this way, the authors get interlacing results between eigenvalues of complexes \(K\) and \(L\) when \(L\) is a covering complex of \(K\) (section 3), when \(L\) is obtained from \(K\) by an elementary contraction or by an elementary collapse (section 4) and they also get an interlacing result concerning relative Laplacians (section 5). Concerning the discretization of the notion of covering space, the authors introduce the notion of strong covering map and stress the fact that their definition gives a discretization much closer to its continuous counterpart than the notion of covering complex used, for example, in [\textit{J. Rotman}, Rocky Mt. J. Math. 3, 641--674 (1973; Zbl 0272.55003)].