The finite element method developed by engineers in 1950 was considere
d in the first three decades as an h-version, that means the degree p
of the polynomials used was fixed at a low rate (e.g. p = 1, 2) and th
e mesh size h was reduced to obtain a better approximation of the exac
t solution. In 1981, I. BABUSKA, B. A. SZABO and I. N. KATZ introduced
the p-version: they fixed the mesh size h and increased the degree p
of the polynomials in order to reduce the approximation error. In the
same year, I. BABUSKA and M. R. DORR investigated a combination of bot
h versions and tried it the hp-version. Here both the mesh size h and
the polynomial degree p are changed to improve accuracy. This was done
with the help of an error estimator needed to decide where the degree
p should be increased (p --> p + 1) or which finite element must be r
efined (h --> HBAR < h). The three versions lead to different approxim
ation results. While the h-version and the p-version have only a polyn
omial rate of convergence with respect to the number of degrees of fre
edom, the hp-version with a geometric mesh refinement has an exponenti
al rate.