Fredholm determinant and Nekrasov sum representations of isomonodromic tau functions (Q1989917)

Consider a Fuchsian system with $n$ regular singular points $a:=\{ a_0, \dots, a_{n-2}, a_{n-1}\equiv \infty \}$ on $\mathbb{P}^1(\mathbb{C})$ of the form \[ \partial_{z}\Phi = \Phi A(z), \quad A(z)= \sum_{k=0}^{n-2} \frac{A_k} {z-a_k}, \tag{1} \] where $A_0,\dots, A_{n-2}$ are $N\times N$ matrices such that $A_0, \dots, A_{n-2}, A_{n-1}=-\sum_{k=0}^{n-2} A_k$ are non-resonant. Then the isomonodromic deformation with respect to $a_0,a_1, \dots, a_{n-2}$ is governed by the Schlesinger equation \[ \partial_{a_i}A_k=\frac{[A_i, A_k]}{a_k-a_i} \quad (i\not= k), \quad \partial_{a_i}A_i=\sum_{k\not=i} \frac{[A_i, A_k]}{a_i-a_k}, \tag{2} \] whose tau function of Jimbo-Miwa-Ueno [\textit{M. Jimbo} et al., Physica D 2, No. 2, 306--352 (1981; Zbl 1194.34167)] is defined by \[ d_a \ln \tau_{\text{JMU}}(a) := \frac 12 \sum_{k=0}^{n-2} \text{res}_{z=a_k} \text{Tr} \, A^2(z) da_k. \] When $N=2$ (respectively, $(N,n)=(2,4)$), Equation (2) is reduced to the Garnier system (respectively, the Painlevé VI equation). In this paper, for the tau function corresponding to prescribed generic monodromy data of (1) a Fredholm determinant representation is derived by Riemann-Hilbert approach. In constructing it, for a decomposition of the $n$-punctured sphere $\mathbb{P}^1(\mathbb{C})\setminus \{a_0,a_1,\dots, a_{n-1}\}$ into $n-2$ trinions (pairs of pants) by $n-3$ annuli, the jump matrices are made along the boundaries of these annuli and trinions, and an auxiliary Riemann-Hilbert problem of Fuchsian type with 3 regular singular points is assigned on each trinion. In the Fredholm determinant $\text{det}(\mathbb{I}-K)$ thus obtained, the integral operator $K$ acts on holomorphic vector functions on the annuli; the kernel of $K$ is written in terms of solutions to these auxiliary Fuchsian systems, and in the case of $N=2$ it is represented by hypergeometric functions. Furthermore, it is shown that the principal minor expansion of $\text{det}(\mathbb{I}-K)$, written in the Fourier basis of the functions on the annuli, yields the combinatorial series representation for the tau function. In particular, if $N=2$, the series expression for the Garnier system coincides with that obtained as a Fourier transform of Liouville conformal blocks (or Nekrasov functions) [\textit{O. Gamayun} et al., J. High Energy Phys. 2012, No. 10, Paper No. 38, 25 p. (2012; Zbl 1397.81307); \textit{N. Iorgov} et al., Commun. Math. Phys. 336, No. 2, 671--694 (2015; Zbl 1311.30029)].