On the question ``Can one hear the shape of a group?'' and a Hulanicki type theorem for graphs (Q2190042)

In the year 1966, \textit{M. Kac} published an interesting paper entitled ``Can one hear the shape of a drum?'' [Am. Math. Mon. 73, 1--23 (1966; Zbl 0139.05603)]. The above question can be traced back to Lipman Bers and Hermann Weyl. It concerns a plane domain \(\Omega\) with piecewise smooth boundary (a drum) and the Laplace operator \[\Delta\ = - \Big(\frac{\partial^2}{\partial x^2}+ \ \frac{\partial^2}{\partial y^2}\Big).\] The question is whether or not the domain \(\Omega\) can be determined (up to isometry) from the spectrum of \(\Delta\). \textit{C. Gordon} et al. [Invent. Math. 110, No. 1, 1--22 (1992; Zbl 0778.58068)] gave a negative answer by constructing a pair of regions in the plane of different shapes but with identical eigenvalues of the Laplace operator. In this paper, the authors consider the analogous question for the Laplace operator on a Cayley graph of an infinite finitely generated group. A natural question, inspired by the classical question about the shape of a drum, is whether or not it is possible to determine the group \(G\), up to some notion of similarity, from the spectrum of \(G\). In 1994, \textit{A. Valette}, in his paper entitled ``Can one hear the shape of a group?'' [Rend. Semin. Mat. Fis. Milano 64, 31--44 (1996; Zbl 0865.43002)], points out, it is easy to see that it is not possible to determine \(G\) up to isomorphism just from the spectrum of \(G\). Many questions about spectra of groups remain open. One of them is: what can be the shape of the spectrum of a group? In 2017, the authors [J. Mod. Dyn. 11, 99--123 (2017; Zbl 1502.22003)], have shown that the spectrum of a finitely generated group \(G\) can be an interval or a union of two intervals. But it is not known whether it can be a union of \(n\geq 3\) disjoint intervals, a countable set of points accumulating to a point, or a Cantor set. \textit{G. Kuhn} [Nagoya Math. J. 128, 95--120 (1992; Zbl 0767.22001)] has constructed a non-amenable group \(G_n\) for every \(n\geq 3\) generated by a finite set \(S_n\) and a Markov operator corresponding to some non-uniformly distributed probabilities on \(S_n\) with the spectrum equal to a disjoint union of \(n\) intervals, but it remains an open question whether the same can be achieved with all probabilities equal to \(\frac{1}{|S_n|}\) The other questions are: \begin{itemize} \item given a certain closed subset \(\sum\) of \([-1, 1]\) how many finitely generated groups have the spectrum to \(\sum\)? \item how does the spectrum change under group coverings? \end{itemize} In this paper, the authors address certain aspects of the latter two questions. Their starting point is a construction of a continuum of groups generated by four involutions with spectra equal to the same union of two intervals. These groups are from the family of groups \(G_{\omega}\), \(\omega \in \Omega = \{0,1,2\}^\infty\) by the author A. Pèrez. The most known of these groups is the group \(\mathcal{G}= G_{\omega_0}\), where \(\omega_0\) is the periodic sequence \(012012\dots\). The groups \(G_{\omega}\) are amenable groups generated by four involutions \(a\), \(b_{\omega}\), \(c_{\omega}\), \(d_{\omega}\) which are introduced by the authors in Section 2. The result is that the collection of groups \(G_{\omega}\), \(\omega \in \Omega\), contains a continuum of groups with pairwise, nonequivalent in the Schwardz-Milnor sense, growth functions (and thus with pairwise non-quasi-isomertic Cayley graphs). Let \(\Omega_2\) be the subset of \(\Omega\) consisting of sequences with at least two symbols from \(\{0,1,2\}\) occurring infinitely many times. The authors prove the following result: For every \(\omega \in \Omega_2\) one has \[\operatorname{sp}(G_{\omega})=\left[-\frac{1}{2}, 0\right]\cup \left[\frac{1}{2},1\right].\] The above result shows that the answer to the question in the title of the paper is ``No'' even when instead of isomorphism one considers a very weak notion of equivalence of groups: quasi-isometry. The authors derive the above result as a corollary of the following result: Let \(G\) be an amenable group generated by four involutions \(\tilde{a}\), \(\tilde{b}\), \(\tilde{c}\), \(\tilde{d}\) such that \(\tilde{b}\tilde{c}\tilde{d}=1\). Assume that for some \(\omega \in \Omega_2\) there exists a surjection \(\varphi: G \rightarrow G_{\omega}\) such that \(\varphi(\tilde{a})=a\), \(\varphi(\tilde{b})=b_{\omega}\), \(\varphi(\tilde{c})=c_{\omega} \) and \(\varphi(\tilde{d})=d_{\omega}.\) Then \[ \operatorname{sp}(G)=\left[-\frac{1}{2}, 0\right]\cup \left[\frac{1}{2},1\right]. \] As the application of the above result the authors obtain: For every \(\omega \in \Omega_2\) there exists a continuum of amenable groups covering \(G_{\omega}\) and with the spectrum equal to \([-\frac{1}{2}, 0]\cup [\frac{1}{2},1].\) Each of these groups is generated by four involutions \(a\), \(b\), \(c\), \(d\) satisfying the condition \(bcd=1\). The authors also prove the follwing: There is a continuum of pairwise non-isomorphic 2-generated torsion free step-3 solvable groups with the spectrum \([-1,1]\).\\ In the proof of their results the authors use the Hulanicki theorem.\\ Theorem [\textit{A. Hulanicki}, Stud. Math. 24, No. 1, 37--59 (1964; Zbl 0134.12501); ibid. 27, No. 2, 87--194 (1966; Zbl 0165.48701)]: A locally compact group \(G\) is amenable if and only if any unitary representation of \(G\) is wealy contained in the regular representation of \(G\). In particular, the Hulanicki theorem implies that given a subgroup \(H < G\) of a countable amenable group \(G\), the quasi-regular representation \(\lambda_{G/H}\) of \(G\) is weakly contained in the regular representation \(\lambda_G\). It is known that weak containment \(\rho \prec \eta\) of two unitary representations of a group G is equivalent to the inclusion of the spectra \(\sigma(\rho(m))\subset \sigma(\eta(m))\) for any \(m\in \mathcal{C}[G]\). Notice that given a generating set \(S\) for \(G\), the Cayley graph \(\Gamma(G,S)\) covers the Schreier graph \(\Gamma(G,H,S)\) of the action of G on \(G/H\) and from the Hulanicki theorem it follows that for an amenable group \(G\) the spectrum of \(\Gamma(G,S)\) includes the spectrum of \(\Gamma(G,H,S)\). The above observation motivates the authors to consider a more general situation. Hence, we find the following in the paper: The authors study the question of whether or not it is possible to determine a finitely generated group \(G\) up to some notion of equivalence from the spectrum \(\operatorname{sp}(G)\) of \(G\). The authors show that the answer is ``No'' in a strong sense. as a first example the authors present the collection of amenable 4-generated groups \(G_{\omega}, \omega\in \{0,1,2\}^{\mathcal{N}}\), constructed by the second author in 1984 [Math. USSR, Izv. 25, 259--300 (1985; Zbl 0583.20023); translation from Izv. Akad. Nauk SSSR, Ser. Mat. 48, No. 5, 939--985 (1984)]. The authors show that among them there is a continuum of pairwise non-quasi- isometric groups with \(\operatorname{sp}(G_{\omega})=[-\frac{1}{2}, 0]\cup [\frac{1}{2},1].\) Moreover, for each of these groups \(G_{\omega}\) there is a continuum of 2-generated torsion-free step-3 solvable groups with the spectrum \([-1,1]\). In addition, in relation to the above results, the authors prove a version of the Hulanicki theorem about inclusion of spectra for covering graphs. The following are the remarks and open questions given in the paper which are very much useful for researchers. Given a finitely generated group \(G\) together with the spectrum of \(G\), it is natural to consider spectral measures of the Markov operator \(M\). Since \(M\) is a self-adjoint operator on \(l^2(G)\), it admits spectral decomposition \[M=\int_{\sigma(M)}^{}dE(\lambda),\] where \(E(\lambda)\) is a projection-valued measure. The spectral measure \(\mu\) of \(M\) is introduced by \[\mu(A)=(E(A)\delta_e, \delta_e),\] where \(A\subset \mathcal{R}\) is an E-measurable set \(\delta_e \in l^2(G)\) is the delta- function of the unit element in \(G\) \((.,.)\) is the scalar product of \(l^2(G)\). Then \[\operatorname{supp}(\mu)= \sigma(M)=\operatorname{sp}(G)\] and \(\mu\) may contain more information about \(G\) than just the spectrum \(\operatorname{sp}(G)\). The following questions are very interesting for researchers to do fruitful research. Question 1: Is it possible to determine the Cayley graph of a finitely generated group \(G\) up to isometry from the spectral measure \(\mu\) of the associated Markov operator? The above question is wide open. Question 2: (a) Is it correct that for any two groups from the family \(G_{\omega}\), \(\omega \in \Omega\), with non-isometric Cayley graphs, the corresponding spectral measures are distinct? (b) Same question for the groups constructed by \(P\). Hall and used in the proof of the results. Another natural question is whether or not the weak Hulanicki theorem for graphs can be proven in the full generality: Question 3: Let \(\Gamma_1\) be an amenable uniformly bounded connected weighted graph which covers a weighted graph \(\Gamma_2\). Let \(H_1\) and \(H_2\) be the Laplace-type operators associated with \(\Gamma_1\) and \(\Gamma_2\), respectively. Is it true that \(\sigma(H_2)\subset \sigma(H_1)?\) The paper is a blend of Banach algebra, combinatorial graph theory and group theory. It is a classic paper well written with proper introduction and developing it into a full-fledged comprehensive paper connecting various areas in mathematics. The paper is equivalent to a thesis. There are remarks and open problems at the end from which the readers should get motivation and inspiration to read this paper so that in turn the researchers write out a classic paper like this.