Processing math: 40%

Research reports

High order Galerkin approximations for parametric second order elliptic partial differential equations

by V. Nistor and Ch. Schwab

(Report number 2012-21)

Abstract
Let DIRd, d=2,3, be a bounded domain with piecewise smooth boundary D and let U be an open subset of a Banach space Y. We consider a parametric family Py of uniformly strongly elliptic, parametric second order partial differential operators Py on D in divergence form, where the parameter y ranges in the parameter domain U so that, for a given set of data fy, the solution u and the coefficients of the parametric boundary value problem Pyu=fy are functions of (x,y)D×U. Under suitable regularity assumptions on these coefficients and on the source term f, we establish a regularity result for the solution u:D×UIR of the parametric, elliptic boundary value problem Pyu(x,y)=fy(x)=f(x,y), xD, yU, with mixed Dirichlet-Neumann boundary conditions. Let D=dDnD denote decomposition of the boundary into a part on which we assign Dirichlet boundary conditions and the part on which we assign Neumann boundary conditions. We assume that dD is a finite union of closed polygonal subsets of the boundary such that no adjacent faces have Neumann boundary conditions (ie.,~there are no Neumann-Neumann corners or edges). Our regularity and well-posedness results are formulated in a scale of weighted Sobolev spaces Km+1a+1(D) of Kondrat'ev type in D. We prove that the parametric, elliptic PDEs (Py)yU admit a shift theorem which is uniform in the parameter sequence yU. Specifically, if coefficients aijpq(x,y) depend on the parameter sequence y=(yk)k1 in an affine fashion, ie.~aijpq=a¯ijpq0+k1ykψijpqk, and if the sequences are p-summable for some 0 < p < 1, then the parametric solution u_y admit an expansion into tensorized Legendre polynomials L_\nu (y) such that the corresponding coefficient sequence {\bf u} = (u_\nu) \in \ell^p ( {\cal F}; {\cal K}^{m+1}_{a+1} (D)). Here, we denote by {\cal F} \subset IN^{IN}_0 the set of sequences \{k_n\}_{n\in IN} with k_n \in IN_0 with only finitely many non-zero terms, and by Y = \ell^\infty(IN) and U = B_1 (Y), the open unit ball of Y. We identify the parametric solution u with its coefficient vector {\bf u} = (u_\nu)_{\nu \in {\cal F}}, u_\nu \in V, in the ''polynomial chaos'' expansion with respect to tensorized Legendre polynomials on U. We also show quasioptimal algebraic orders of convergence for Finite Element approximations of the parametric solutions u(y) from suitable Finite Element spaces in two and three dimensions. Let t = m/d and s = 1/p - 1/2 for some p \in (0, 1] such that {\bf u} = (u_\nu) \in \ell^p({\cal F}; {\cal K}^{m+1}_{a+1}(D)). We then show that, for each m \in IN, exists a sequence \{S_\ell\}_{\ell \ge 0} of nested, finite dimensional spaces S_\ell \subset L ^2 (U, \mu; V) such that M_\ell = {\rm dim}(S_\ell) \rightarrow \infty and such that the Galerkin projections u_\ell \in S_\ell of the solution u onto S_\ell satisfy \begin{equation*} \|u - u_\ell\|_{L^2 (U,\mu;V)} \le C\, {\rm dim}(S_\ell)^{- \min {s,t}} \|f\|_{H^{m -1} (D)} . \end{equation*} The sequence S_\ell is constructed using a nested sequence V_\mu \subset V of Finite Element space in D with graded mesh refinements toward the singular boundary points of the domain D as in [7, 9, 27]. Our sequence V_\mu is independent of y. Each subspace S_\ell is then defined by a finite subset \Lambda_\ell \subset {\cal F} of ''active polynomial chaos'' coefficients u_\nu \in V, \nu \in \Lambda_\ell in the Legendre chaos expansion of u which, in turn, are approximated by u_\nu \in V_{\mu(\ell,\nu)} for each \nu \in \Lambda_\ell, with a suitable choice of \mu(\ell,\nu).

Keywords:

BibTeX
@Techreport{NS12_464,
  author = {V. Nistor and Ch. Schwab},
  title = {High order Galerkin approximations for parametric second order elliptic partial differential equations},
  institution = {Seminar for Applied Mathematics, ETH Z{\"u}rich},
  number = {2012-21},
  address = {Switzerland},
  url = {https://www.sam.math.ethz.ch/sam_reports/reports_final/reports2012/2012-21.pdf },
  year = {2012}
}

Disclaimer
© Copyright for documents on this server remains with the authors. Copies of these documents made by electronic or mechanical means including information storage and retrieval systems, may only be employed for personal use. The administrators respectfully request that authors inform them when any paper is published to avoid copyright infringement. Note that unauthorised copying of copyright material is illegal and may lead to prosecution. Neither the administrators nor the Seminar for Applied Mathematics (SAM) accept any liability in this respect. The most recent version of a SAM report may differ in formatting and style from published journal version. Do reference the published version if possible (see SAM Publications).