Re: P vs h-type
In principle, any problem which can be solved by the h-version can be solved by the p-version and, conversely, any problem which can be solved by the p-version can be solved by the h-version. There are large differences in convergence rates, however. For example, it was demonstrated in one well-documented plane elastic model problem that to achieve one percent relative error in energy norm (which is similar to the root-mean-square measure of error in stress), approximately 1000 degrees of freedom were needed with the p-version and properly designed mesh, whereas 10 million degrees of freedom would have been required with the h-version, utilizing 8-noded quadrilaterals and uniform mesh refinement. For details, refer to p.190 in Szabo and Babuska, Finite Element Analysis (1991). There are other important areas where the p-version has clear and substantial advantages: adhesively bonded joints (where very large aspect ratios are required), structural plates and shells, fracture mechanics, etc.
