Two-term Szegő theorem for generalised anti-Wick operators (Q906574)

Recall that for a suitable class of functions \(q\) on \(\mathbb{R}^{2n}\), the Weyl quantized pseudodifferential operator is defined as \[ (\operatorname{op}[q]u)(x):=(2\pi)^{-n}\int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{n}}e^{i(x-y)\cdot\xi}q(\tfrac{1}{2}(x+y),\xi)u(y)\,dy\,d\xi. \] The (Schwartz) kernel of such an operator is, at least formally, \[ K_q(x,y) = (2\pi)^{-n}\int_{\mathbb{R}^{n}}e^{i(x-y)\cdot\xi}q(\tfrac{1}{2}(x+y),\xi)\,d\xi, \] and if \(q(x,\xi)\) decays sufficiently rapidly at infinity, this operator is of trace class with the trace given by \[ \operatorname{tr}\,\operatorname{op}[q] := \int_{\mathbb{R}^{n}}K_q(x,x)\,dx= (2\pi)^{-n}\int_{\mathbb{R}^{2n}}q(x,\xi)\,dx\,d\xi. \] The author treats asymptotics for the trace of functions of such an operator with the added difficulty that its Weyl symbol is the convolution of a dilated discontinuous function and a smooth function on \(\mathbb{R}^{2n}\). We point out that there are a variety conditions on \(q\) and a function \(f\) under which \(f(\operatorname{op}[q])\) is also a pseudodifferential operator. If \(\operatorname{op}[q]\) is selfadjoint, we may assume that \(f\in C^{\infty}(\mathbb{R})\) and employ the spectral theorem or Fourier inversion. In the non-selfadjoint case, one typically assumes that \(f\) is complex analytic and then utilizes Cauchy's integral formula or a power series, as is done in the paper under review. The main result is the following two term expansion for the trace: \[ \operatorname{tr}f(\operatorname{op}[W\ast(a\chi_{\Omega})_r]) = r^{2n}A_0 + r^{2n-1}A_1 +\mathcal{O}( r^{2n-2});\quad\text{as\,\,}r\to\infty. \] In the above formula, {\parindent=0.7cm\begin{itemize}\item[--] \(\Omega\subset \mathbb{R}^{2n}\) is a \(C^2\) domain satisfying extra conditions and \(\chi_{\Omega}\) is its characteristic function; \item[--] the function \(a(x,\xi)\) belongs to \(C^2(\mathbb{R}^{2n})\cap W^{2,1}(\mathbb{R}^{2n}) \cap W^{2,\infty}(\mathbb{R}^{2n})\); \item[--] \(W(x,\xi)\) is rapidly decreasing and has integral 1; and finally, \item[--] \(q_r\) denotes a dilated symbol defined by \(q_r(x,\xi) := q(x/r,\xi/r)\). \end{itemize}}It is crucial here that the coefficients \(A_0, A_1\) are given by integrals of quantities depending on \(f,a,W\). Indeed, the novelty of the result is the explicit expression -- along with a very nice geometric characterization -- of the second coefficient \(A_1\), which turns out roughly to be a surface integral on \(\partial\Omega\). The proof is broken into digestible pieces and, at its core, relies on approximation and localization arguments. Despite some typos and some unclear statements, the paper is well written and even includes an appendix explaining relevant differential-geometric notions.