The global parametrix in the Riemann-Hilbert steepest descent analysis for orthogonal polynomials (Q643200): Difference between revisions

The authors consider the \(2\times 2\) matrix-valued model Riemann-Hilbert problem (``global parametrix'') that arises in the application of the Deift-Zhou steepest descent method to orthogonal polynomials. The problem is the following. We are given \(N\) intervals \([a_1,b_1],\dots, [a_N,b_N]\) on the real line, \(N-1\) real numbers \(\alpha_1,\dots, \alpha_{N-1}\) and integer \(N\). Does there exist a \(2\times 2\) matrix-valued function \(M:\mathbb{C}\setminus[a_1, b_N]\to \text{Mat}(2,\mathbb{C})\) such that (a) \(M\) is analytic on \(\mathbb{C}\setminus[a_1, b_N]\), (b) \(M\) has jumps \(M_+(x)= M_-(x) I_M(x)\) for \(x\in [a_1,b_N]\), where \[ \begin{aligned} I_M(x) &= \begin{pmatrix} 0 & 1\\ -1 & 0\end{pmatrix},\quad\text{for }x\in (a_k,b_k),\quad k= 1,\dots, N,\\ I_M(x) &= \begin{pmatrix} e^{-2\pi in\alpha_k} & 0\\ 0 & e^{2\pi in\alpha_k}\end{pmatrix},\quad\text{for }x\in (b_k, a_{k+1}),\quad k= 1,\dots, N-1,\end{aligned} \] (c) \(M\) has at most fourth-root singularities near the endpoints \(a_k\), \(b_k\), (d) \(M(z)= I+ O(1/z)\quad\text{as }z\to\infty\)? The known positive solution of this problem uses Riemann theta functions on hyperelliptic Riemann surfaces. Here, the authors present an alternative approach to the solution of the model Riemann-Hilbert problem, which is less constructive (it does not lead to explicit formulas) but rather simple (it uses meromorphic differentials only). The construction deals with the two-sheeted Riemann surface of the algebraic function \[ w^2= \prod^N_{k=1} (z- a_k)(z- b_k) \] and meromorphic differentials on it. They are uniquely defined by their poles (which are all simple), residues and the property of vanishing of their \(A\)-periods. The solution \(M\) of the model Riemann-Hilbert problem is constructed by integration of these differentials and exponentiation of the obtained functions. After proving the solvability of the problem the authors propose some useful properties of the solution \(M\) (such as its uniqueness, zeros distribution of its entries, uniform boundedness in \(n\) away from the endpoints \(a_k\), \(b_k\)). At the end of the article there is presented an alternative way to construct the second row of \(M\), on the assumption that the first row is known.