In this note we study perturbations of a J-nonnegative operator A in a KREIN space which are such that the difference of the resolvents of A and of the perturbed operator B is of rank one. Here B is also supposed to be J-selfadjoint. With the pair A, B we associate a one-parameter family {Br),,eR of J-selfadjoint operators (or linear relations) which extend the "common part" of A and B. In particular, A and B belong to this family. Furthermore, we characterize those opemtors in the family which are J-nonnegative (Theorem 2.3.2). The description of the operators in this family through the corresponding resolvents is similar to M. G. KREIN'S description of the orthogonal generalized resolvents of a EhsmTian operator with defect numbers one in HILBHRT space (see e.g.[l]). The Theorem 2.3.2 gives, in particular, a description of the generalized resolvents of a densely defined J-nonnegative HrPmmian operator A, with defect one on some KBEIN space S, which are generated by J-nonnegative J-selfadjoint extensions of A, in S.
( c q 1~2 & ( A ) g, t p 2 & ( L l ) g) 5 0-1.Hence the sequence G1'zgc(Ll) 8 converges to aome h E .#' 0 N ( C ) if d tenda to (0, oo), and llhllz 5 c-l. Making use of the second relation in (8) we find (Ci/z)h = fi, and the assertion follows. Proof of Proposition2.1. Sincetheseta9(Aa) andO(H) aredeneein.#'~~lz =.#'+l/2(H) the operator A,, is J-nonnegative if and only if 0 5 [Jz + a[z, { I f , 21 = (J& z) + a I(Jf, 41' (z E Jf'+1,z(H)), and this holds if and only if (9) ( H z , 5) 2 --d~ I(J1, %)I8 (2 E OW)). We have Jf E 2-l/2(H). Then, according to Lemma 2.2, the relation (9) holds if and only if either u 2 0 or Jf E B((Hi")-) and 0 < -u 5 l\((Hi'z)-)-l Jfll-8. Now observe that Hi/2 = JKi12J. Hence Jj = ( H i / 2 )z, where z E 2 0 J ( H ) r ie equivalent to f = (KiI2)-Jz. It follows that ll((Hi/z)-)-l Jfll = ll((Ki/2)-)-1 1 1 1 and the Proposition 2.1 is proved. y > IR(i) 11: if 111-= 00 y > IB(i) fl; or y 5 fl', -lfl! if lfl-< 00. Moreover the relation f = a8(i) f -i8(i) f implies that IfI-< 00 if and only if @(i) fl-< 00. A little computation yields in thie caw the relation By (6) the inequality (6) is equivalent to and to I~I ? = IB(i) fI: + IR(i) 11: . Thus, the Proposition 2.1 has the following Corollary. Zf f E then B, ie J-mntnegatave ij and only if y E R \ (-I &) f t , I&) 11;).Here the nonnegativity of BI;ci,fl: follows from Lemma 1.2.3. The general m e . In order to find the valuea of the parameter y for which B, is J-nonnegative in the general case 1 E #'-we shall use the following proposition. Proposition 3.1. Let A be a J-nonnegative J-eelfadjoint opemtor, e(A) $; 0 and put H := J A . Then for arbitrary x E .#' the function (0, 00) 3 7 -.ra[AR(iq) 2, R(irl) 4 ia nondecreasing. Z j z E 94(H1I2) then lim qz[AR(iq) z, R(iq) z] = ( H l k , EI1l2z), st-