Trace formulae for perturbations of class \(s_m\) (Q1933404)

The authors obtain a very general trace formulae in the case of perturbations of self-adjoint operators by self-adjoint operators in the Schatten class \(S_m\), where \(m\) is a positive integer. In particular, they recover the known formulae for \(m=1\) due to Lifshits and Krein [\textit{M. G. Krejn}, Mat. Sb., N. Ser. 33(75), 597--626 (1953; Zbl 0052.12303)] and \textit{L. S. Koplienko} [Sib. Math. J. 25, 735--743 (1984); translation from Sib. Mat. Zh. 25, No. 5(147), 62--71 (1984; Zbl 0574.47021)] for \(m=2\). To extend to other values \(m\in \mathbb N\), they use techniques from a paper by \textit{D. Potapov}, \textit{A. Skripka} and \textit{F. Sukochev} [``Spectral shift function of higher order'', Invent. Math. 193, No. 3, 501--538 (2013; \url{doi:10.1007/s00222-012-0431-2}), \url{arXiv:0912.3056}], where it has been shown that, for every \(m\in \mathbb N\) and every pair \((A,K)\) of self-adjoint operators with \(K\) in the Schatten class \(S_m\), there exists a unique function \(\eta_m\in L^1(\mathbb R)\) such that the operator Taylor polynomial \[ {\mathcal T}^{(m)}_{A,K}f= f(A+K)-f(A)- \sum_{k=1}^{m-1}\frac{1}{k!}\frac{d^k}{dt^k} f(A+tK)|_{t=0} \] satisfies \[ \operatorname{trace}({\mathcal T}^{(m)}_{A,K}f)=\int_{\mathbb R} f^{(m)}(x)\eta_m(x)\,dx \] for functions \(f\) such that \({\mathcal F} f^{(k)}\in L^1(\mathbb R)\) for \(1\leq k\leq m\). The authors of the present paper extend, among other things, the result to functions in the Besov spaces \(B^m_{\infty,1}(\mathbb R)\) given by functions such that \(\int_{\mathbb R} \frac{\|\Delta_t^{m+1}f\|_\infty}{|t|^{m+1}}\,dt<\infty\) and \(\lim_{|x|\to \infty}\frac{|f(x)|}{(1+ |x|)^m}=0\). They prove a formula for operator Taylor polynomials of functions in \(B_{\infty,1}^m\) in terms of multiple operator integrals and present a general formula which permits to compute, for any measure \(\mu\) and \(f\in B^m_{\infty,1}(\mathbb R)\), the value of \[ \operatorname{trace}\langle f(A+tK), \mu^{(m)}\rangle \] . This is then applied to recover a number of interesting results.