This paper deals with the finite element displacement method for approximating isolated solutions of general quasilinear elliptic systems. Under minimal assumpt,ions on the structure of the continuous problems it is shown that the discrete analogues also have locally unique solutions which converge with quasi-optimal rates in L, and L,. The essential tools of the proof are a deformation argument and n technique using weighted L,-norms.
Q 1. IntroductionLet .Q be a bounded region in EucLrDean space Rfl, n 2 2. We consider the quasilinear system : k'ind u = (uk)k=;,...,r such that -i. a i~f ( . , u, ~u ) + F;(.) u, ~u ) = o in Q, ~k = gk on ail,,, 2 P:(-, u, c/u) ni = c i+zi on a~, , i = l n i= 1 i = 1 where Ss, qk and 8: are given functions and {aQ,, aQ,) is a partition of the boundary aQ.Below Lp(Q)' and Hm-p(Q)r, m E N, 1 5 p 5 00, denote the usual LEBESGUE and SOBOLEV spaces of (real) vector valued functions 2, = (2)k)k=;,,,,,r on 9. The corresponding norms are rising inult,i-index notation, with the usual modification for p = 00. €ZFJ'(aQ,; l2)' is the closure in H",p(Q)* of the subspace of Cw-vector functions vanishing on aQ,. We shall use the usual summation convention for repeated indices k = 1, . . ., r. For example the L2-scalar product is written as follows :(u, w) = J U k W k d X .
RThe above boundary value problem has the usual weak formulation in the EEZLBEBT