Algebraic conditions and the sparsity of spectrally arbitrary patterns (Q2236391)

Let \(A\) be a \(m\times n\) matrix with entries in a field \(F\). \(A\) is said to be a zero-nonzero matrix if the entries belong to the set \(\{\ast ,0\},\) where we replace the nonzero entries with the symbol \(\ast.\) The zero-nonzero pattern of a matrix \(B\) is the zero-nonzero pattern matrix \(\mathcal{B}\) such that the \((i,j)\) entry of \(\mathcal{B}\) is \(\ast\) if and only if the \((i,j)\) entry of \(B\) is nonzero. \(B\) is said to be a realization of the pattern \(\mathcal{B}.\) The zero-nonzero pattern of a matrix gives a combinatorial description leading to the question concerning the information about the operator-theoretic properties of the matrix. For square matrices a pattern is said to be spectrally arbitrary when it carries essentially no information about the eigenvalues of the matrix. This means that the zero-nonzero pattern is consistent with any choice of characteristic polynomial. When one deals with matrices over on an ordered field, such as the set of real numbers, a sign pattern is a matrix such their entries belong to the set \(\{+, -, \ast\}.\) In [\textit{T. Britz} et al., SIAM J. Matrix Anal. Appl. 26, No. 1, 257--271 (2004; Zbl 1082.15016)], the so called \emph{\(2n\) conjecture} is formulated as follows: every \(n\times n\) sign pattern with entries in the field of real numbers is spectrally arbitrary if it has at least \(2n\) nonzero entries. On the other hand, the question about when a particular zero-nonzero pattern is spectrally arbitrary on any field \(F\) is also formulated therein. In particular, the assumption about the irreducibility of the matrix is posed in order to solve the \emph{generalized \(2n\) conjecture}: every irreducible zero-nonzero pattern that is spectrally arbitrary over \(F\) has at least \(2n\) zero entries. In order to solve this conjecture, it suffices to determine whether or not there exists an \(n\times n\) irreducible zero-nonzero pattern with \(2n-1\) nonzero entries that is spectrally arbitrary over some field, according to a result in [\textit{Y. Shitov}, Linear Algebra Appl. 534, 73--77 (2017; Zbl 1371.15033)]. Here, the authors prove that for matrices with real entries is sufficient to establish that \emph{generalized \(2n\) conjecture} holds and this implies that the \emph{ \(2n\)} conjecture holds for irreducible matrices with real entries, when \(1\leq n\leq 7.\) Notice that for all spectrally arbitrary patterns, the lower bound of \(2n\) nonzero entries was proved for \(1\leq n\leq 4\) in [\textit{L. Corpuz} and \textit{J. J. McDonald}, Linear Multilinear Algebra 55, No. 3, 249--273 (2007; Zbl 1112.15015)], and extended to \(n=5\) for both reducible and irreducible zero-nonzero patterns in [\textit{L. M. DeAlba} et al., Linear Algebra Appl. 423, No. 2--3, 262--276 (2007; Zbl 1118.15007)]. Furthermore, in this last paper the bound for reducible zero-nonzero patterns is established for \(n=6,7.\) The validity of the first conjecture is achieved via a computer-assisted verification that exploits an algebraic condition. On the other hand, the results of the paper apply to every field with characteristic \(0\) or \(p\), up to for some finitely many nonzero values of \(p.\) The first conjecture does hold over all such fields when \(1\leq n\leq6,\) and in the case \(n=7\) there are only two patterns that could be counterexamples. They are analyzed and it is shown that they are not spectrally arbitrary over the field of real numbers. However, the conjecture remains open over the field of complex numbers for \(n=7.\) Essentially, the only patterns of this size that could be counterexamples are the two mentioned above.