Bounds for the \(\alpha\)-adjacency energy of a graph (Q6569657)

Spectral graph theory deals with finite graphs, mostly the simple undirected ones, by considering the eigenvalues and eigenvectors of various kinds of naturally associated matrices, such as the adjacency matrix \(A(G)\) or the signless Laplacian matrix \(D(G)+A(G)\), with \(D(G)\) being the diagonal matrix given by degrees of vertices. Clearly to study the ``deformation'' given by the family of convex combinations \(A_\alpha(G):=\alpha D(G)+(1-\alpha)A(G)\), \(0\le\alpha\le 1\), is of interest and the authors acknowledge \textit{V. Nikiforov} [Appl. Anal. Discrete Math. 11, No. 1, 81--107 (2017; Zbl 1499.05384)] of having introduced the generalized adjacency matrices \(A_\alpha(G)\). In this situation, the notion of energy of a graph, whose introduction in connection with theoretical chemistry is ascribed to \textit{I. Gutman} [Ber. Math.-Stat. Sekt. Forschungszent. Graz 103, 22 S. (1978; Zbl 0402.05040)], has been generalized by other authors.\N\NTheorem 3.2 shows an upper bound to that generalized adjacency energy for a connected \(G\), in terms of the vertex covering number \(\tau\) and an integer parameter \(\sigma\) related with the eigenvalues of \(A_\alpha(G)\). In the range \(0\le\alpha\le 0.5\) some numbers \(m_i\) that are related to the minimum covering set are also involved (their definition is given in the course of the proof of the theorem). The statement also includes that for star graphs the bound is reached. Two corollaries to the theorem add some information on previously known facts in the literature.\N\NTheorem 3.5 shows an upper bound and a lower bound in terms of the (first) Zagreb index, the maximum degree \(\Delta\), and the Frobenius norm of the adjacency matrix \(A(G)\). The statement also characterizes graphs for which ``equality occurs in both inequalities'', but in the proof it is more precisely stated that to have equality in each one of them suffices to have those extremal graphs. Theorem 3.6 is similar. Theorem 3.7 presents a relatively simple upper bound in terms of the mentioned parameter \(\sigma\) and the vertex degrees and also says that for regular graphs equality occurs.\N\NThe paper is organized and written with care, and only occasionally I found small issues that might make reading a bit uncomfortable. For instance, on p.~134, line~3, \(\alpha(m_i+1)\pm\sqrt{\alpha^2(m_i+1)^2+4m_i(1-2\alpha)}<0\) for \(\alpha>\frac12\) should be \(\alpha(m_i+1)-\sqrt{\alpha^2(m_i+1)^2+4m_i(1-2\alpha)}>0\) for \(\alpha>\frac12\); and similarly in the next line. Or also, on p.~133, line~1, \(\rho_n=0\) should be \(\rho_n\le 0\). Besides, I found ``unique eigenvalue'' obscure on p.~128, line~13.\N\NIt seems to me that a more severe issue may have occurred on p.~132, lines~12--14. Indeed, for the graph of order \(5\) with edges \(\{0,1\}\), \(\{0,2\}\), \(\{0,3\}\), \(\{0,4\}\), \(\{1,2\}\), \(\{3,4\}\), and \(\alpha=0.5\) the equality in (3.4) occurs (with \(\{v_i,v_j\}\) equal to \(\{1,2\}\) or \(\{3,4\}\)), but the graph is not of the form \(K_2\vee H\). That assertion is used in the proof of Theorem~3.1 in conjunction with \(\sigma=n-1\) (where \(n\) is the order of \(G\)), which does not hold in our example, so in my opinion that theorem and previous discussion would have deserved some more clean-up.