mathematics of computation
volume 38,number 158
april 1982, pages 401-413
Optimal Error Properties of Finite Element Methods
for Second Order Elliptic Dirichlet Problems
By Arthur G. Werschulz
Abstract. We use the informational approach of Traub and Wozniakowski [9] to study the
variational form of the second order elliptic Dirichlet problem Lu = f on ü C RN. For
/e Hr(Q), where r> -1, a quasi-uniform finite element method using n linear functional
Jaf^i nas T7'(ß)-norm error 0(n~<r+1)/'v). We prove that it is asymptotically optimal among
all methods using any information consisting of any n linear functionals. An analogous result
holds if L is of order 2m: if / € Hr(ü), where r 3» -m, then there is a finite element method
whose //"(fi)-norm error is %(n~(2m+r~a)/N) for 0 « a « m, and this is asymptotically
optimal; thus, the optimal error improves as m increases. If the integrals jaf'p, are approxi-
mated by using n evaluations off, then there is a finite element method with quadrature with
7i'(ß)-norm error 0(n~r/N) where r > N/2. We show that when N = 1, there is no method
using n function evaluations whose error is better than ñ(n~r); thus for N = 1, the finite
element method with quadrature is asymptotically optimal among all methods using n
evaluations of /.
1. Introduction. This paper deals with the optimal solution of second order elliptic
partial differential equations. We wish to consider the variational form of the
problem
(1.1) Lu=f infiCR", u = 0 on3fi,
(see Section 2). Suppose that we evaluate information of the form
<"> [u*.//*•}•
If / E Hr(Q), where r > -1, there exists a finite element method using (1.2) whose
error is 0(n~(r+ X)/N) when measured in the i/'(ß) norm.
We first wish to answer two questions. First, is there another method using the
information (1.2) whose error is better than that of this finite element method?
Second, is the information (1.2) the best possible information using n linear
functionals? That is, is there another set of n linear functionals such that the best
algorithm using this new information is better than the best algorithm using (1.2)? In
Section 3, we asymptotically answer these questions in the negative. Thus,this finite
element method is of asymptotically optimal error among all algorithms, linear or
nonlinear, using any n linear functionals whatsoever.
We also report some results on 2mth order elliptic problems which indicate that
as m increases, the same number of evaluations yields smaller Hm(ü) error.
Received May 20, 1980; revised June 15, 1981.
1980 Mathematics Subject Classification. Primary 65N30, 68C25; Secondary 65J10, 65N05, 65N15.
401
©1982 American Mathematical Society
0025-5718/81/0000-0127/S04.25
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use