A fully noncommutative Painlevé II hierarchy: Lax pair and solutions related to Fredholm determinants (Q2024539): Difference between revisions

For an \(r\) by \(r\) matrix \(U\) whose entries are real-valued functions of \(\vec{s}=(s_1,\ldots, s_r)\), let \(\mathcal{L}_n[U]\) be differential polynomials given by the recursive relations: \begin{align*} &\mathcal{L}_0[U] =(1/2) I_r, \\ & \frac{\mathrm{d}}{\mathrm{d}S} \mathcal{L}_n[U] = \Bigl( \frac{\mathrm{d}^3}{\mathrm{d}S^3} + [U, \cdot]_+ \frac{\mathrm{d}}{\mathrm{d}S} +\frac{\mathrm{d}}{\mathrm{d}S} [U, \cdot]_+ + [U,\cdot] { \frac{\mathrm{d}} {\mathrm{d}S} }^{-1} [U,\cdot] \Bigr) \mathcal{L}_{n-1}[U], \quad n \ge 1, \end{align*} where \(\mathrm{d}/\mathrm{d}S:= \sum^{r}_{k=1} \partial/\partial s_k,\) \((\mathrm{d}/\mathrm{d}S)^{-1}\) is the corresponding formal antiderivative, and \([U,\cdot]= U\cdot\, -\,\cdot U\) and \([U,\cdot]_+= U\cdot\, +\,\cdot U\) are the commutator and anti-commutator between two matrices. In this paper the noncommutative Painlevé II hierarchy is defined by \[ \mathrm{PII}^{(n)}_{\mathrm{NC}}: \quad \Bigl(\frac{\mathrm{d}}{\mathrm{d}S} + [W, \cdot ]_+ \Bigr) \mathcal{L}_n[U] =(-1)^{n+1} 4^n [S,W]_+, \quad U:= \frac{\mathrm {d}}{\mathrm{d}S} W-W^2, \] \(S\) on the right-hand side being \(S=\mathrm{diag}(s_1,\ldots, s_r).\) The author obtains an isomonodromic Lax pair for \(\mathrm{PII}^{(n)}_{\mathrm{NC}}\) of the form \[ \frac{\mathrm{d}}{\mathrm{d}S} \Psi^{(n)}(\lambda,\vec{s})=L^{(n)}(\lambda, \vec{s}) \Psi^{(n)}(\lambda,\vec{s}), \quad \frac{\partial}{\partial\lambda} \Psi^{(n)}(\lambda,\vec{s})=M^{(n)}(\lambda, \vec{s}) \Psi^{(n)}(\lambda,\vec{s}). \] In finding this Lax pair, \(\Psi^{(n)}\) is constructed by the use of the Riemann-Hilbert techniques related to the matrix Airy Hankel operator \(\mathcal{A}i_{2n+1},\) and furthermore it is shown that \(\mathrm{PII}^{(n)} _{\mathrm{NC}}\) admits a solution \(W\) expressed by the Fredholm determinant \[ -\mathrm{Tr}(W^2(\vec{s})) =\frac{\mathrm{d}^2}{\mathrm{d}S^2} \mathrm{ln} (\mathrm{det}(\mathrm{Id}_{\mathbb{R}_+} - \mathcal{A}i^2_{2n+1})). \] Here \(\mathcal{A}i_{2n+1}\) is such that \begin{align*} & (\mathcal{A}i_{2n+1} \mathbf{f})(x) :=\int_{\mathbb{R}_+} \mathbf{Ai}_{2n+1} (x+y,\vec{s}) \mathbf{f}(y) dy, \quad \text{for any \(\mathbf{f} \in L^2(\mathbb{R}_+, \mathbb{C}^r)\),} \\ & \mathbf{Ai}_{2n+1}(x,\vec{s})=(c_{j,k} \mathrm{Ai}_{2n+1}(x+s_j+s_k))_{j,k =1}^r, \quad \mathrm{Ai}_{2n+1}(x)=\int_{\gamma^n_+}\frac 1{2\pi} \exp\Bigl(\frac{\mathrm{i} \mu^{2n+1}}{2n+1} +\mathrm{i} x\mu \Bigr) \mathrm{d}\mu, \end{align*} where \(c_{j,k}\in \mathbb{C}, x\in \mathbb{R},\) and \(\gamma^n_+\) is a suitably chosen path.