ABSTRACT. In the space l~ of vector sequences, we consider the symmetric operator L generated by the expression (lu)j := Bjuj+l + Ajuj + B~_luj_l , where u-1 --0, u0, ut .... E C p, Aj and Bj are p x p matrices with entries from C, A~ = Aj, and the inverses B~ -1 (j = 0, 1,...) exist. We state a necessary and sufficient condition for the deficiency numbers of the operator L to be maximal; this corresponds to the completely indefinite case for the expression l. Tests for incomplete indefiniteness and complete indefiniteness for I in terms of the coefficients Aj and Bj are derived.KEY WOADS: sequence space, difference expression, matrix polynomial, deficiency numbers of an operator.w Introduction Let C p (p _~ 1) be the p-dimensional Euclidean space of column vectors with the standard inner P product y*x = )"~1=1 xjyj, where xj and yj (j = 1, 2,...~ p) are the coordinates of the vectors x and y, respectively, and the asterisk indicates the adjoint matrix. Further, let Aj and Bj (j = 0, 1, ...) be p x p matrices with complex entries; we assume that Aj is a self-adjoint matrix and Bj is invertible (j = 0, 1,... ). Consider the second-order difference expressionwhere u-1 := 0, u0, ul, .-. E C p 9 Let the symbol I denote the operation that acts on sequences of p x p matrices U0, U1,... with entries from C in the same way as l, i.e., according to (1). The following Green formula is valid: ~t A similar expression is also valid for the case in which the vectors uj and vj are replaced by the matrices Uj and V~ and the operation 1 is replaced by I.
~-~((lu)iv I -ui(lv)j) = [uvl(s)[~_,,Let s be the Hilbert space of infinite sequences of vectors u = (uo, ul, ... ), u i E C v, with inner product (u, v) w,+oo . = 2-.,i=o vjuj, and let L be the minimal closed operator generated by the expression (1) with the boundary condition u-1 = 0 in the space s (see [1, Chap. VII, w p. 509;2]). It is well known that the operator L is symmetric but is not self-adjoint in general, and the deficiency numbers n+ and n_ of this operator satisfy the inequalities 0 < n+, n_ <_ p. We say that we have a completely indefinite case for the expression l if n+ = n_ --p. It is readily derived from the subsequent argument (see Theorem 1 and its proof) that we have the completely indefinite case for 1 if and only if all solutions of the equation (lu)j=zuj, j=0,1,..., u_,=0,for z = 0 belong to the space d~2p.