post from book < Functional Analysis and Boundary Value Problems: an Introductory Treatment>
An important aspect of the theory of BVP is that concerned with the nature of the solution to the problem. Given the problem of finding a function f that satisfies a PDE, and one or more B.C., it's clearly of great value to know beforehand whether such a solution exists and if so, whether it is unique, and finally how smooth this function is.
When approaching such questions one is essentially dealing with the properties of an operator ( in this case a partial differential operator) A from one space of functions U to another space V, so to start with it is necessary to choose suitable spaces for U and V. The C^m appear to be appropriate since they are spaces of m-times continuously differentiable functions and we are dealing with differential equations. However, they suffer from the drawback that they are not complete.
The Sobolev space H^m provide a very natural setting for B.V.P. First of all, they are complete, secondly, it is possible to obtain quite general results regarding existence and uniqueness of solutions to differential equations, using these spaces. Also Sobolev space provide a means of characterizing the degree of smoothness of functions. Finally, and perhaps of most importance is the fact that approximate solution methods such as the Galerkin and FEM are most conveniently and correctly formulated in finite dimensional subspaces of Sobolev spaces.