Extrapolating B splines for interpolation (Q1073281)

Following Schoenberg, the author in this paper studies the problem of cardinal spline interpolation. For \(h>0\), let \(x_ j=jh\) \((j\in {\mathbb{Z}})\). Given a function f defined on \({\mathbb{R}}\), let \(\tilde f(x,h)=\sum^{\infty}_{j=-\infty}f(x_ j)M_ n(x-x_ j,h),\) where \(M_ n(\cdot,h)\) is the n-th central B spline of order n. Usually, the interpolation error \(\tilde f(\cdot,h)-f\) is only \(O(h^ 2)\). The author uses Richardson extrapolation to obtain an interpolation formula with accuracy of higher degree than that obtained with the variation diminishing spline approximation. More precisely, the author introduced the interpolation kernel \(W_ n(\cdot,h):=[3M_ n-h(\partial M_ n/\partial h)]/2,\) and shows that the interpolation formula \(\tilde f(x,h)=\sum^{\infty}_{j=-\infty}f(x_ j)W_ n(x-x_ j,h)\) gives accuracy of \(O(h^ 4)\). Moreover, the author indicates the applicability of spline interpolation to particle simulations.