Functional integral representations and Golden-Thompson inequalities in boson-fermion systems (Q2855834)

The celebrated Golden-Thompson inequality originally says the trace inequality \(\text{tr}\, e^{A+B} \leq \text{tr}\, (e^Ae^B)\) for Hermitian matrices \(A\) and \(B\), which is useful in mathematics and physics, and especially in statistical mechanics and was first derived there. This paper deals with this inequality in supersymmetric quantum mechanics on \({{\mathbb R}^n}\).NEWLINENEWLINE\textit{S. Klimek} and \textit{A. Lesniewski} [Lett. Math. Phys. 21, No. 3, 237--244 (1991; Zbl 0723.58054)] proved, with Hamiltonian NEWLINE\[NEWLINE H_{KL} = -\tfrac{\hbar^2}2\Delta -\tfrac{\hbar}2\Delta P(x) + \tfrac12|\nabla P(x)|^2 + \hbar \sum_{j,k=1}^n \partial_j\partial_k P(x)\, b_j^*b_k NEWLINE\]NEWLINE acting on the tensor product Hilbert space \({\mathcal H} =L^2({\mathbb R}^n) \otimes \Lambda({\mathbb C}^n)\) of \(L^2({\mathbb R}^n)\) and the fermion Fock space \(\Lambda({\mathbb C}^n)\), in other words, on the Hilbert space \(L^2({\mathbb R}^n;\,\Lambda({\mathbb C}^n))\), a Golden-Thompson inequality NEWLINE\[NEWLINE \text{tr } e^{-tH_{KL}} \leq \tfrac1{(2\pi t)^{n/2}\hbar^n} \int_{{\mathbb R}^n} \text{det} [I+ e^{-t\hbar\nabla\otimes \nabla P(x)}] e^{-\tfrac{t}2 [|\nabla P(x)|^2-\hbar\Delta P(x)]}dx, \,\,\, t>0, \tag{1} NEWLINE\]NEWLINE if the right-hand side is finite. Here \(P(x)\) is a real polynomial in \(x =(x_1, \dots, x_n) \in {\mathbb R}^n\), and \(\nabla = (\partial_1, \dots, \partial_n)\) [with \(\partial_j = \partial/\partial {x_j}\)], \(\Delta = \nabla^2\), and \(\nabla\otimes\nabla\, P(x)\) denotes the \(n\times n\) matrix with \((j,k)\) entry \(\partial_j\partial_k P(x)\), while \(b_j/b_j^* \,\,(j=1, \dots, n)\) stands for the fermion annihilation/creation operator on \(\Lambda({\mathbb C}^n)\).NEWLINENEWLINEIn the paper under review, the author tries to improve the inequality (1) in the following two aspects. One is that it is not best possible, because (1) does not attain \textit{equality}, i.e. the left-hand side is less than the right one, even in the case for \(H_{KL}\) being the Hamiltonian of a supersymmetric quantum harmonic oscillator, by demonstrating it with an example. The other is that it has not a form suitable for infinite dimensional extensions. So another inequality should be found satisfying this requirement, one which would be more natural than (1). For a general class of boson-fermion Hamiltonians \(H\) on \({\mathcal H}\), he first establishes a functional integral representation for the trace \(\text{tr}(F\otimes z^{N_f}e^{-tH})\) and then proves a Golden-Thompson type inequality for its absolute value \(|\text{tr}(F\otimes z^{N_f}e^{-tH})|\), where \(F \in L^{\infty}({\mathbb R}^n)\), \(z\) is a non-zero complex number, and \(N_f:= \sum_{j=1}^n b_j^* b_j\) is the fermion number operator on \(\Lambda({\mathbb C}^n)\). The result is applied to a supersymmetric quantum mechanics with Hamiltonian NEWLINE\[NEWLINE H_{SS} := H_b + \Phi_W(x) + H_f + \hbar\sum_{j,k=1}^n \partial_j\partial_k W(x)\, b_j^* b_k NEWLINE\]NEWLINE on \({\mathcal H}\) with NEWLINE\[NEWLINE\begin{aligned} &H_b = -\tfrac{\hbar^2}2\Delta -\tfrac{\hbar}2 \text{tr}\, \Omega + \tfrac12(\Omega x)^2, \\ &\Phi_W(x) = (\Omega x) \cdot (\nabla W(x)) + \tfrac12 |\nabla W(x)|^2 -\tfrac{\hbar}2\Delta W(x), \quad H_f = \sum_{j=1}^n \hbar\omega_j b_j^*b_j, \end{aligned} NEWLINE\]NEWLINE where \(\Omega\) is the diagonal matrix \(\text{diag}(\omega_1,\dots, \omega_n)\) with constant \(\omega_j >0\) for each \(j\). The author proves, under some appropriate condition on a real-valued function \(W\) on \({\mathbb R}^n\) with \(\partial_jW \in L^4_{\text{loc}}\) and \(\partial_j\partial_kW \in L^2_{\text{loc}}\) for all \(j, k\), an improved version of Golden-Thompson inequality NEWLINE\[NEWLINE\begin{multlined} \text{tr}\, e^{- t H_{SS}} \leq \int_{{\mathbf R}^n} K_t(x,x) \det \big[I+ e^{-t\hbar(\Omega+\nabla\otimes\nabla\,W(x))}\big]\\ \times e^{-t[(\Omega x) \cdot (\nabla W(x)) + \tfrac12 |\nabla W(x)|^2 -\tfrac{\hbar}2\Delta W(x)]}\,dx, \quad t>0, \end{multlined}\tag{2} NEWLINE\]NEWLINE if the right-hand side is finite, where \(K_t(x,y)\) is the integral kernel of the heat semigroup \(e^{-tH_b}\). The inequality (2) with \(P\) in place of \(W\) includes (1) since \(H_{KL}\) is obtained as a limit of \(H_{SS}\) as \(\Omega\) tends to \(0\) in a suitable sense.NEWLINENEWLINEThe author uses the functional integral representation method, as well as Klimek and Lesniewski. The significant difference lies in that, to represent the bosonic part of \(e^{- t H_{SS}}\), the former is using the \(n\)-dimensional conditional oscillator measure, while the latter used the \(n\)-dimensional conditional Wiener measure. The paper is well-written and also instructive about the subject.