Reconstruction of polyharmonic functions from samples (Q2456671)

For a complex valued function \(f: \mathbb{R}^2 \to \mathbb{C}\), define the following Wirtinger differential operators: \[ D_{0}f = f, \;\;D_{1}f = \frac {\partial f} {\partial \bar{z}}, \;\;D_{2j}f = \frac {\partial^{2j}f} {\partial z^{j} \partial {\bar{z}}^{j}}, \;\;D_{2j+1} = \frac {\partial^{2j+1}f} {\partial z^{j} \partial {\bar{z}}^{j+1}}, \] for \(j = 1, 2, 3, \dots.\) Let \({\mathcal H}_{k} = \{f: \mathbb{R}^2 \to \mathbb{C}: f \in C^{k}(\mathbb{R}^2), \;D_{k}f \equiv 0 \}.\) Here, the space \({\mathcal H}_{1}\) is the space of all entire functions, and \({\mathcal H}_{2}\) is the space of all complex valued harmonic functions. The space \({\mathcal H}_{2k}\) is called the space of polyharmonic functions of order \(k\), and the space \({\mathcal H}_{2k+1}\) is called the space of polynarmonic functions of order \(k + \frac {1} {2}\). A function \(f(x,y)\) can be written in the form \[ f(x,y) = f \bigg(\frac {z + \bar{z}} {2}, \frac {z - \bar{z}} {2i} \bigg) = f^{w}(z, \bar{z}) \;. \] The author proves a number of results about the form that a polyharmonic function may take. Of particular interest is a collection of results which recovers a representation for a polyharmonic function from the values the function and its Wirtinger derivatives assume at a finite collection of points. For \(f \in {\mathcal H}_{k}\) it is proved that \(f\) has a Wirtinger form \[ f^{w}(z, \bar{z}) = \sum_{j=0}^{m-1} \big(\bar{z}^{j} g_{j}(z) + z^{j}h_{j}(\bar{z})\big) \] for \(k = 2m\), where the functions \(g_{j}\) and \(h_{j}\) are all entire functions with \(h_{j}(0) = h_{j}'(0) = \dots = h_{j}^{m-1}(0) = 0\), and, if \(k = 2m-1\), then the sum is taken from \(j = 0\) only to \(j = m - 2\) and a term of the form \(\bar{z}^{m-1} g_{m-1}(z)\) is added to the summation. For the results given here, some restrictions on the space \({\mathcal H}_{k}\) are needed. In particular, let \[ {\mathcal H}_{k, \sigma} = \{f \in {\mathcal H}_{k}: \limsup_{r \to \infty} \frac {\log{\max_{x^{2} + y^{2} = r^{2}} | f(x,y)| }} {r} \leq \sigma \}. \] Let \({\mathbf t}\) denote a \`\` sequence\'\' \ \(\{t_{n}\}\) and let \(L \i n (0, 1/4]\), where each \(t_{n}, -\infty < n < \infty\) is a real number and \(| t_{n} - n| \leq L\). For such a sequence \({\mathbf t}\) and number \(L\), we say that \({\mathbf t} \in {\mathcal N}(L)\). Let \[ G({\mathbf t}; z) = (z - t_{0}) \prod_{n=1}^{\infty} \big(1 - \frac {z} {t_{n}}\big) \big(1 - \frac {z} {t_{-n}}\big) \;, \] and let \[ G_{n}({\mathbf t};z) = \frac {G({\mathbf t};z)} {(z - t_{n}) G'({\mathbf t};t_{n})}, \;\;\;-\infty < n < \infty. \] Among the many results proved is that for \(k\) a positive integer and for each non-negative integer \(j \leq k - 1\), if the sequence \({\mathbf t}_{j} = \{t_{j,n}\}_{-\infty < n < \infty} \in {\mathcal N}(L_{j})\), where \(L_{j} \in [0, \frac {1} {4}), 0 \leq j \leq k\), and if \(f \in {\mathcal H}_{k,\pi} \cap L^{2}(\mathbb{R}^{2})\), then \[ f(x,y) = \sum_{j=0}^{k-1} \sum_{n=-\infty}^{\infty} D_{j}f(t_{j,n}, 0) C_{j,n}({\mathbf t}_{j};x,y), \] where the double series converges in the norm \(\| -\| _{2}\) and uniformly on every strip \(| y| \leq K\) for any \(K > 0\), and the functions \(C_{j,n}\) are defined as follows: \[ C_{0,n}^{w}({\mathbf t};z,\bar{z}) = G_{n}({\mathbf t};z) \;\;\text{ and } \;\;C_{j,n}^{w}({\mathbf t}; z, \bar{z}) = \int_{z}^{\bar{z}} C_{j-1,n}^{w}({\mathbf t}; \zeta, z) d \zeta \;, \;j \geq 1. \] An algorithm for approximating a function in \(L^{2}(R^{2}) \cap {\mathcal H}_{k,\pi}\) is given, based on the values of the function and its Wirtinger derivatives on sequences \({\mathbf t_{j}}\) for which \(\lim_{n \to \infty} (t_{j,n+1} - t_{j,n}) = \delta_{j} \in (0, 1]\). The idea is to take successive approximating functions \(f_{n}\), where \(f_{1}\) uses only values assumed by \(f\) and its Wirtinger derivatives at values in the sequence \({\mathbf t}_{1}\), and \(f_{n+1}\) is obtained by ``correcting'' \(f_{n}\) by using values assumed by the the function \(f - f_{n}\) and its Wirtinger derivatives on points of the sequence \({\mathbf t}_{n+1}\). An error estimate in the \(L^{2}\) norm is obtained for each of the approximating functions \(f_{n}\), where this error goes to zero as \(n\) goes to \(\infty\).