Weighted approximation with varying weight (Q1323814)

The central theme of this monograph concerns the following general problem: given a sequence \(\{W_ n(x)\}\) of ``weight functions'' and a continuous function \(f(x)\), under what circumstances does there exist a sequence of polynomials \(\{P_ n\}\), where \(P_ n\) is of degree at most \(n\), such that the sequence \(\{W_ n(x) P_ n(x)\}\) converges uniformly to \(f(x)\) on some interval \(I\). The monograph is divided into four parts. Part I deals with so-called Freud weights. The basic result for Freud weights, that is, weights of the form \(w_ \alpha(x)= \exp(-\gamma_ \alpha| x|^ \alpha)\), where the constants \(\gamma_ \alpha\) are appropriately chosen, states that for each continuous function \(f\) on the real line which vanishes outside the open interval \((-1,1)\) there exists a sequence of polynomials \(\{P_ n\}\), where \(P_ n\) is of degree at most \(n\), such that the sequence \(\{(w_ \alpha(x))^ n P_ n(x)\}\) converges uniformly to \(f\) on the whole real line. The condition that \(f\) vanishes outside the open interval \((-1,1)\) is necessary for this convergence. A fairly elementary proof of this result, and other related results about Freud weights, is presented. These proofs are modifications of previously published proofs, giving a somewhat shorter path to the main results. Part II of the monograph deals with more general weights \(w(x)\). The results given apply to the support \(s_ w\) of the equilibrium measure \(\mu_ w\) associated with the weight function \(w\). As is the case with Freud weights, if the sequence \(\{(w(x))^ n P_ n(x)\}\) converges uniformly to \(f(x)\) at \(x_ 0\), where \(x_ 0\) is not in \(S_ w\), then it is necessary that \(f(x_ 0)= 0\). While the basic result here is similar to that for Freud weights, there are some difficulties for functions which are non-zero at points where the measure \(\mu_ w\) has vanishing density. Complete proofs and some illuminating examples are given. Part III deals with the situation where the weights \(W_ n(x)\) are not of the form \((w(x))^ n\), where \(w(x)\) is a fixed function. Again, if the measures \(\mu_{W_ n}\) are sufficiently nice, there is a sequence \(\{W_ n(x) P_ n(x)\}\) to approximate any continuous function. In order to deal with points where the measures \(\mu_{W_ n}\) may not be ``well-behaved'', the usual technique of using \(n\) equally spaced nodes to approximate the measure by a measure with support consisting of a finite number of points can be modified by a slight shift in the nodes. This method is used in the proofs in Part III, and also in Part IV, which consists of a number of applications of the results. The monograph gives a state of the art presentation dealing with the general problem in a number of contexts. However, the potential reader should be aware that the treatment is technical with quite lengthy proofs, and there is very little discussion which might give an overview to the more causal reader.