An index theorem is a tool for computing the change of the index (i.e., the number of negative eigenvalues) of a symmetric monotone matrix-valued function when its variable passes through a singularity. In 1995, the first author proved an index theorem in which a certain critical matrix coefficient is constant. In this paper, we generalize the above index theorem to the case when this critical matrix may be varying, but its rank, as well as the rank of some additional matrix, are constant. This includes as a special case the situation when this matrix has a constant image. We also show that the index theorem does not hold when the main assumption on constant ranks is violated. Our investigation is motivated by the oscillation theory of discrete symplectic systems and Sturm-Liouville difference equations with nonlinear dependence on the spectral parameter, which was recently developed by the second author and for which we obtain new oscillation theorems.
Motivation.In 1995, the first author proved a result called an "index theorem," which shows how to compute the change of the index of a symmetric monotone matrix-valued function when its variable passes through a singularity; see [19, Theorem 2] or [20, Theorem 3.4.1], and for comparison also [24, Proposition 2.5]. By the index of a matrix we mean the number of its negative eigenvalues. This result has been utilized in many applications, in particular in the oscillation theory of SturmLiouville differential equations, linear Hamiltonian systems, and discrete symplectic systems. The relevant references are [20, sections 4.2, 5.2, 7.2] and [5, 10, 22, 24]. More precisely, the index theorem in [19, Theorem 2 and Corollary 2] or [20, Theorem 3.4.1 and Corollary 3.4.2] is the crucial auxiliary result for the continuous and discrete oscillation theory as presented in the references [20] (see the proof of Theorem 7.1.2 therein) and [10] (see the proof of Theorem 1 in section 4.4 therein).One of the key assumptions of this index theorem is that one of the considered coefficients is constant, i.e., it does not depend on t. In our main result (Theorem 2.1 below) this would be the matrix R 2 ≡ R 2 (t). It is the aim of this paper to let R 2 (·) depend on t and to analyze how this dependence can be incorporated, so that the assertion of the index theorem still holds. The assumption on the constancy of R 2 (·) implies, of course, limitations in the applications, as, e.g., in the above-mentioned references [20, Assumption (7.1.3)] and [10], where particularly the matrices B k do not depend on the spectral parameter λ. A similar requirement regarding the independence on λ in the coefficients B k (λ) was recently observed in [24, section 6.4]. On the *