Book of Abstracts - Dipartimento di Matematica
Book of Abstracts - Dipartimento di Matematica
Book of Abstracts - Dipartimento di Matematica
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Workshop on<br />
Non-Standard Numerical Methods for PDE’s<br />
Pavia (Italy), June 29th - July 2nd, 2010<br />
<strong>Book</strong> <strong>of</strong> <strong>Abstracts</strong>
Organizing Commetee<br />
Daniele B<strong>of</strong>fi<br />
Annalisa Buffa<br />
Carlo Lova<strong>di</strong>na<br />
Ilaria Perugia<br />
Giancarlo Sangalli<br />
Fun<strong>di</strong>ng<br />
Istituto <strong>di</strong> Alta <strong>Matematica</strong> “F. Severi”<br />
c<strong>of</strong>ounded by:<br />
ERC Project GeoPDEs<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong> “F. Casorati”,<br />
Università <strong>di</strong> Pavia
Program<br />
Tuesday Wednesday Thursday Friday<br />
9:00-9:50 D. N. Arnold P. Monk F. Brezzi E. Cohen<br />
9:50-10:20 M. Costabel T. Betcke G. Manzini B. Jüttler<br />
Break Break Break Break<br />
10:50-11:20 S. Christiansen T. Luostari M. F. Wheeler A. Reali<br />
11:20-11:50 J. Schoeberl G. Gabard I. Yotov J. A. Evans<br />
11:50-12:20 R. Falk J. M. Melenk<br />
Lunch<br />
L. Beirão da<br />
Veiga<br />
J. Rivas<br />
14:00-14:50 L. Demkowicz R. Hiptmair K. Lipnikov T. Dokken<br />
14:50-15:20 F. Gar<strong>di</strong>ni A. Moiola P. Bochev C. Manni<br />
15:20-15:50 R. Winther S. A. Sauter<br />
15:20-15:40<br />
D. Svyatskiy R. Duvigneau<br />
15:40-16:00<br />
A. Cangiani<br />
Break Break Break<br />
16:00-16:20<br />
A. Quaglino<br />
16:10-16:30 M. Dauge C. H<strong>of</strong>reither F. Cirak<br />
Break<br />
16:30-16:50 H. Heumann J. Phillips P. N. Nielsen<br />
16:50-17:10 A. Palha A. M. Marica A. Ratnani<br />
17:10-17:30 J. Kreeft M. Ruess<br />
16:40-20:00<br />
Panel Session C. Heinrich<br />
17:30-17:50 P. Gatto<br />
20:30<br />
Social <strong>di</strong>nner
Contents<br />
Plenary lectures 1<br />
Doug Arnold<br />
Hilbert Complexes and the Finite Element Exterior Calculus . . . . . . . 3<br />
Franco Brezzi<br />
Scalar products and reconstructions in MFD . . . . . . . . . . . . . . . . 4<br />
Elaine Cohen<br />
Analysis Aware Modeling for Isogeometric Analysis . . . . . . . . . . . . . 5<br />
L. Demkowicz, J. Bramwell and W. Qiu<br />
Solution <strong>of</strong> dual-mixed elasticity equations using Arnold-Falk-Winther element<br />
and <strong>di</strong>scontinuous Petrov-Galerkin method, a comparison . . . . . 6<br />
Tor Dokken<br />
Locally Refined Splines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8<br />
R. Hiptmair, C. Gittelson, A. Moiola, and I. Perugia<br />
Plane Wave Discontinuous Galerkin Methods . . . . . . . . . . . . . . . . 9<br />
K. Lipnikov<br />
Mimetic finite <strong>di</strong>fference method for solving PDEs on polygonal and polyhedral<br />
meshes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11<br />
Peter Monk<br />
The solution <strong>of</strong> time harmonic wave equations using complete families <strong>of</strong><br />
elementary solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12<br />
Invited talks 13<br />
L. Beirão da Veiga<br />
Mimetic <strong>di</strong>scretization <strong>of</strong> the <strong>di</strong>ffusion problem: a higher order method<br />
and a posteriori error estimates . . . . . . . . . . . . . . . . . . . . . . . . 15<br />
Timo Betcke<br />
Nonpolynomial FEM with MPSPACK . . . . . . . . . . . . . . . . . . . . 16<br />
Pavel Bochev<br />
Optimization-based computational models, or how to let someone else do<br />
the hard work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17<br />
Snorre Christiansen<br />
Mixed finite elements as inverse systems <strong>of</strong> <strong>di</strong>fferential forms . . . . . . . . 18<br />
Martin Costabel<br />
Regularized Poincaré operator and p version approximation <strong>of</strong> the Maxwell<br />
eigenvalue problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
R. Duvigneau<br />
Aerodynamic Analysis and Design using an Isogeometric Approach . . . . 20<br />
Annalisa Buffa, John A. Evans, Thomas J.R. Hughes, Giancarlo Sangalli<br />
Isogeometric Analysis in Fluid Dynamics: Divergence-Free Discretizations<br />
for the Stokes and Navier-Stokes Equations . . . . . . . . . . . . . . . . . 21<br />
R. Falk<br />
Canonical Families <strong>of</strong> Finite Element Differential Forms and Their Properties<br />
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22<br />
G. Gabard, R.J. Astley, P. Gamallo, G. Kennedy<br />
A brief review <strong>of</strong> wave-based computational methods for aero-acoustics . . 23<br />
Francesca Gar<strong>di</strong>ni<br />
Mimetic finite <strong>di</strong>fferences for eigenvalue problems . . . . . . . . . . . . . 24<br />
Bert Jüttler<br />
Dimensions and Bases <strong>of</strong> Bivariate Hierarchical Tensor-product Splines . 26<br />
Teemu Luostari, Tomi Huttunen and Peter Monk<br />
A 3D plane wave basis for elastic wave problems . . . . . . . . . . . . . . 27<br />
Carla Manni, Francesca Pelosi and Maria Lucia Sampoli<br />
Beyond NURBS: Non-Standard CAGD Tools in Isogeometric Analysis . . 29<br />
G. Manzini<br />
High-order accurate nodal formulation <strong>of</strong> the Mimetic Finite Difference<br />
Method for Elliptic Problems . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />
J.M. Melenk<br />
Mapping properties <strong>of</strong> Helmholtz integral operators and their application<br />
to the hp-BEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31<br />
Ralf Hiptmair, Andrea Moiolaand Ilaria Perugia<br />
Approximation by plane waves . . . . . . . . . . . . . . . . . . . . . . . . 32<br />
Alessandro Reali<br />
Isogeometric Analysis in Pavia . . . . . . . . . . . . . . . . . . . . . . . . 34<br />
L. Beirão da Veiga, A. Buffa, J. Rivasand G. Sangalli<br />
Some estimates for hpk-refinement in Isogeometric Analysis . . . . . . . . 35<br />
Stefan A. Sauter<br />
A Posteriori Error Estimation for Highly Indefinite Helmholtz Problems . 36<br />
Joachim Schöberl<br />
Nitsche-type Mortaring for Maxwell’s Equations . . . . . . . . . . . . . . 37<br />
Mary F. Wheeler<br />
A Convergent MFMFE for Highly Distorted Hexahedra . . . . . . . . . . 38<br />
R. Winther<br />
Cochain projections in finite element exterior calculus . . . . . . . . . . . 39<br />
Ivan Yotov<br />
A Multiscale Mortar Multipoint Flux Mixed Finite Element Method . . . 40<br />
Contributed talks 41
Andrea Cangiani<br />
Mimetic Finite Difference methods for convection-<strong>di</strong>ffusion problems . . . 43<br />
Fehmi Cirakand Quan Long<br />
A unified sub<strong>di</strong>vision <strong>di</strong>scretization approach for thick and thin shells . . 44<br />
Monique Dauge<br />
Skin effect in electromagnetism . . . . . . . . . . . . . . . . . . . . . . . . 45<br />
Paolo Gatto, Leszek Demkowicz<br />
Modeling <strong>of</strong> bone conduction <strong>of</strong> sound in the human head using hp-finite<br />
elements: code design and verification. . . . . . . . . . . . . . . . . . . . . 47<br />
Ch. Heinrich, A.-V. Vuong, B. Simeon<br />
Finite Volume Methods on NURBS Geometries with Application to Fluid<br />
Flow and Fluid-Structure Interaction . . . . . . . . . . . . . . . . . . . . . 48<br />
Holger Heumann<br />
High order Galerkin Methods for General Advection Problems . . . . . . 49<br />
Clemens H<strong>of</strong>reither<br />
A non-standard finite element method based on boundary integral operators 51<br />
J.J. Kreeft, A. Palha, M.I. Gerritsma<br />
Higher-order <strong>di</strong>scretization <strong>of</strong> the Laplace-Beltrami operator . . . . . . . . 52<br />
Aurora-Mihaela Marica(joint work with Enrique Zuazua)<br />
Localized solutions and filtering mechanisms for the <strong>di</strong>scontinuous Galerkin<br />
semi-<strong>di</strong>scretizations <strong>of</strong> the 1 − d wave equation . . . . . . . . . . . . . . . 53<br />
Peter Nørt<strong>of</strong>t Nielsen, Manh Dang Nguyen, Anton Evgrafov,<br />
Allan Roulund Gersborg and Jens Gravesen<br />
Isogeometric Shape Optimization . . . . . . . . . . . . . . . . . . . . . . . 55<br />
A. Palha, J.J. Kreeft, M.I. Gerritsma<br />
Higher order cochain interpolation . . . . . . . . . . . . . . . . . . . . . . 56<br />
Joel Phillips<br />
Quadrature for Finite Elements on Pyramids . . . . . . . . . . . . . . . . 57<br />
Joachim Linn, Alessio Quaglino, Max Wardetzky and Clarisse Weischedel<br />
Towards a Mimetic Nonlinear Shearable Shell . . . . . . . . . . . . . . . . 58<br />
Ahmed Ratnani, Eric Sonnendrucker and Nicolas Crouseilles<br />
High-Order Spline Finite Element Solver for the Time Domain Maxwell<br />
equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61<br />
M. Ruessand E. Rank<br />
A Fictitious Domain Method for Dynamic Analyses . . . . . . . . . . . . 62<br />
D. Svyatskiy<br />
Analysis <strong>of</strong> the Discrete Maximum Principle in the Mimetic Finite Difference<br />
Method for Elliptic Problems . . . . . . . . . . . . . . . . . . . . . . 64<br />
List <strong>of</strong> participants 65
Plenary lectures<br />
1
Plenary lectures<br />
Hilbert Complexes and the Finite Element Exterior Calculus<br />
Doug Arnold<br />
Institute for Mathematics and its Applications, University <strong>of</strong> Minnesota,<br />
Minneapolis, MN 55455, USA<br />
arnold@ima.umn.edu<br />
The finite element exterior calculus (FEEC) provides a framework in which to understand<br />
and develop stable finite element methods for a variety <strong>of</strong> problems. Hilbert<br />
complexes provide a natural framework for FEEC. In this talk we will survey the basic<br />
theory <strong>of</strong> Hilbert complexes and their <strong>di</strong>scretization, <strong>di</strong>scuss their applications to finite<br />
element methods, and draw connections to results from geometry and topology.<br />
3
Plenary lectures<br />
Scalar products and reconstructions in MFD<br />
Franco Brezzi<br />
Istituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Iinformatiche, CNR,<br />
via Ferrata 1, 27100 Pavia, Italia<br />
brezzi@imati.cnr.it<br />
It is well known that in many cases one has a fair amount <strong>of</strong> freedom in the choice <strong>of</strong><br />
scalar products for Mimetic Finite Differences. The best use <strong>of</strong> such freedom in many<br />
circumstances is still to be assessed. Here in particular we will <strong>di</strong>scuss the cases in<br />
which the MFD scalar product could be interpreted as a tra<strong>di</strong>tional scalar product in<br />
conforming or nonconforming finite element spaces. The usual consistency requirements<br />
for the scalar products will be related to the classical “patch test” used by engineers in<br />
order to assess the convergence <strong>of</strong> a given finite element <strong>di</strong>scretization.<br />
4
Analysis Aware Modeling for Isogeometric Analysis<br />
Elaine Cohen<br />
School <strong>of</strong> Computing, University <strong>of</strong> Utah,<br />
50 S. Central Campus Drive, 3190 MEB, Salt Lake City, UT, 84112, USA<br />
cohen@cs.utah.edu<br />
Plenary lectures<br />
Isogeometric Analysis (IA) has been proposed as a methodology for bridging the gap<br />
between Computer Aided Design (CAD) and Finite Element Analysis (FEA). While<br />
CAD typically focuses on a boundary representation, FEA has focused on volumetric<br />
representations that are low order (linear) approximations to the geometry. Creating<br />
high quality representations for just one <strong>of</strong> these goals can be challenging. However,<br />
proposed representations for IA must create parameterizations and elements suitable<br />
for supporting both good geometric computations and have good qualities for this new<br />
mode <strong>of</strong> analysis. This presentations <strong>di</strong>scusses some <strong>of</strong> the desirable attributes, challenges<br />
in moving from current representational and modeling methodologies, and some<br />
initial modeling/reconstruction methodologies towards creating models that satisfy both<br />
domains.<br />
5
Plenary lectures<br />
Solution <strong>of</strong> dual-mixed elasticity equations using Arnold-<br />
Falk-Winther element and <strong>di</strong>scontinuous Petrov-Galerkin<br />
method, a comparison<br />
L. Demkowicz, J. Bramwell and W. Qiu<br />
Institute for Computational Engineering and Sciences, The University <strong>of</strong> Texas at Austin,<br />
Austin, TX 78712, USA<br />
demlow@ms.uky.edu<br />
Construction <strong>of</strong> stable Finite Element (FE) <strong>di</strong>scretizations for the dual-mixed system<br />
<strong>of</strong> linear elasticity has been an area <strong>of</strong> research for three decades. The recent element<br />
<strong>of</strong> Arnold, Falk and Winther [1] builds upon reproducing on the <strong>di</strong>screte level an exact<br />
sequence for the elasticity equations (with weakly imposed symmetry) that provides a<br />
basis for stability <strong>of</strong> the continuous problem. The situation is more complicated than<br />
for the classical grad-curl-<strong>di</strong>v sequence where the <strong>di</strong>screte spaces are accompanied by<br />
the operators from the continuous setting. Construction <strong>of</strong> the <strong>di</strong>screte elasticity complex<br />
involves not only definition <strong>of</strong> appropriate <strong>di</strong>screte polynomial spaces but also a<br />
necessary mo<strong>di</strong>fication <strong>of</strong> the operators that involve besides the grad-curl-<strong>di</strong>v operator<br />
purely algebraic operations. In the end, the pro<strong>of</strong> <strong>of</strong> <strong>di</strong>screte stability boils down to the<br />
construction <strong>of</strong> appropriate interpolation operators that make simultaneously commute<br />
three particular <strong>di</strong>agrams related to the elasticity complex.<br />
The AFW element has been recently extended to elements <strong>of</strong> variable order [6,7,8].<br />
Whereas the generalization <strong>of</strong> polynomial spaces is standard, the construction <strong>of</strong> the<br />
interpolation operators is not and rather technical. 2D numerical experiments confirm<br />
the stability not only for h-refined meshes <strong>of</strong> variable order, but for hp-refinements as<br />
well. Showing the independence <strong>of</strong> stability constants upon order p, however, remains<br />
open.<br />
The main promise <strong>of</strong> the recently introduced Discontinuous Petrov Galerkin (DPG)<br />
method with optimal test functions is that it guarantees the <strong>di</strong>screte stability, if the underlying<br />
continuous problem is well-posed. The method delivers the best approximation<br />
in a problem-dependent energy (residual) norm [2,3,4]. Moreover, with an appropriate<br />
choice <strong>of</strong> norm for the test space, one can deliver the best approximation in a norm <strong>of</strong><br />
choice, starting with the L 2 -norm [5,9].<br />
The presentation will be devoted to a numerical comparison and illustration <strong>of</strong> the two<br />
methods using 2D examples. After a brief introduction <strong>of</strong> both methodologies, we shall<br />
use an L-shape domain problem to study the stability <strong>of</strong> both methods by comparing<br />
the actual approximation error with the best approximation error in context <strong>of</strong> uniform<br />
as well as h- and hp-adaptive refinements.<br />
6
References<br />
Plenary lectures<br />
[1] D. N. Arnold, R. S. Falk, and R. Winther. Mixed finite element methods for linear elasticity<br />
with weakly imposed symmetry. Mathematics <strong>of</strong> Computation, 76:1699-1723, 2007.<br />
[2] L. Demkowicz and J. Gopalakrishnan. A class <strong>of</strong> <strong>di</strong>scontinuous Petrov-Galerkin methods.<br />
Part I: The transport equation. Comput. Methods Appl. Mech. Engrg., 2009. accepted, see<br />
also ICES Report 2009-12.<br />
[3] L. Demkowicz and J. Gopalakrishnan. A class <strong>of</strong> <strong>di</strong>scontinuous Petrov-Galerkin methods.<br />
Part II: Optimal test functions. Technical Report 16, ICES, 2009. Numer. Meth. Part. D.<br />
E., in review.<br />
[4] L. Demkowicz, J. Gopalakrishnan, and A. Niemi. A class <strong>of</strong> <strong>di</strong>scontinuous Petrov-Galerkin<br />
methods. Part III: Adaptivity. Technical Report 1, ICES, 2010.<br />
[5] A.H. Niemi, J.A. Bramwell, and L.F. Demkowicz. Discontinuous Petrov-Galerkin method<br />
with optimal test functions for thin-body problems in solid mechanics. Technical Report 13,<br />
ICES, 2010.<br />
[6] W. Qiu and L. Demkowicz. Mixed hp-finite element method for linear elasticity with weakly<br />
imposed symmetry. Comput. Methods Appl. Mech. Engrg., 198:3682-3701, 2009.<br />
[7] W. Qiu and L. Demkowicz. Mixed hp-finite element method for linear elasticity with weakly<br />
imposed symmetry II: Curvilinear elements in 2D. Technical Report 6, ICES, 2010.<br />
[8] W. Qiu and L. Demkowicz. Mixed hp-finite element method for linear elasticity with weakly<br />
imposed symmetry III: 3D elements. Technical report, ICES, 2010, in preparation.<br />
[9] J. Zitelli, I. Muga, L. Demkowicz, J. Gopalakrishnan, D. Pardo, and V. Calo. A class <strong>of</strong><br />
<strong>di</strong>scontinuous Petrov-Galerkin methods. Part IV: Wave propagation problems. Technical<br />
Report 17, ICES, 2010.<br />
7
Plenary lectures<br />
Locally Refined Splines<br />
Tor Dokken<br />
Department <strong>of</strong> Applied Mathematics, SINTEF ICT,<br />
Forskningsveien 1, Oslo<br />
tor.dokken@sintef.no<br />
The central idea <strong>of</strong> isogeometric analysis is to replace tra<strong>di</strong>tional Finite Elements<br />
by NonUniform Rational B-splines (NURBS) to provide accurate shape description and<br />
higher order elements in analysis. Since the introduction <strong>of</strong> the idea by T. Hughes in 2005<br />
excellent results have been obtained documenting the potential <strong>of</strong> isogeometric analysis.<br />
However, it has also been demonstrated that NURBS do not support the local refinement<br />
needed in efficient isogeometric analysis. T-splines <strong>of</strong>fer some local refinement but have<br />
been shown to be linearly dependent in specific configurations, a situation not acceptable<br />
in analysis. Rather than adapting technology such as T-splines developed for Computer<br />
Aided Geometric Design (CAGD), Locally Refined Splines (LR-splines) are developed<br />
from the dual viewpoints <strong>of</strong> CAGD and analysis. T-splines represent the refinement<br />
in the grid <strong>of</strong> control vertices, a natural approach in CAGD. However, this hides the<br />
<strong>di</strong>mensionality <strong>of</strong> the spline space, gives refinement a fairly large footprint, and can in<br />
certain situations trigger an unwanted growth in the number <strong>of</strong> coefficients. LR-splines<br />
code the refinement <strong>di</strong>rectly in the parameterization <strong>of</strong> the spline space by insertion <strong>of</strong><br />
knot line segments. This approach gives refinement a small footprint, provides control <strong>of</strong><br />
the <strong>di</strong>mensionality <strong>of</strong> the spline space, and supplies a basis composed <strong>of</strong> tensor product<br />
B-spline basis functions with <strong>di</strong>fferent levels <strong>of</strong> refinement. The LR-spline basis also<br />
satisfies the important partition <strong>of</strong> unity property without resorting to rational scaling<br />
as in T-splines. The talk will report on the current status <strong>of</strong> the development <strong>of</strong> LRsplines,<br />
and show the first examples <strong>of</strong> the flexible refinement <strong>of</strong> LR-splines.<br />
8
Plane Wave Discontinuous Galerkin Methods<br />
R. Hiptmair † , C. Gittelson † , A. Moiola † , and I. Perugia ‡<br />
† Seminar for Applied Mathematics, ETH,<br />
Rämistrasse 101, 8092 Zürich, Switzerland<br />
‡ <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia,<br />
Via Ferrata 1, 27100 Pavia, Italia<br />
Plenary lectures<br />
{hiptmair,gittelson,moiola}@sam.math.ethz.ch, ilaria.perugia@unipv.it<br />
A natural idea to improve the accuracy <strong>of</strong> Galerkin methods for wave propagation<br />
problems in the frequency domain is to incorporate qualitative information about the<br />
oscillatory behavior <strong>of</strong> solutions into the trial spaces. For the homogeneous Helmholtz<br />
equation −∇u − ω 2 u = 0 this idea naturally resorts to plane wave functions x →<br />
exp(iωd · x), |d| = 1, and it has given rise to various plane-wave based methods. Prominent<br />
examples <strong>of</strong> such methods are the plane wave partition <strong>of</strong> unity finite element<br />
method [2], the ultraweak Galerkin <strong>di</strong>scretization (UWVF) [4, 5], the <strong>di</strong>scontinuous enrichment<br />
method [1, 13], and the VTCR (variational theory <strong>of</strong> complex rays) approach<br />
[12]. All perform fairly well in computations, see [8-10] for computational results for the<br />
ultra-weak approach. This presentation will focus on plane wave <strong>di</strong>scontinous Galerkin<br />
methods (PWDG), <strong>of</strong> which the UWVF is a special incarnation, see [3, 6]. In the framework<br />
<strong>of</strong> Trefftz-DG methods the formulation is derived and, subsequently, the available<br />
theoretical results on convergence <strong>of</strong> the h-version and the p-version will be surveyed [6,<br />
7], resorting to refined plane wave approximation estimates from [11], cf. the talk by A.<br />
Moiola. Numerical results on the performance <strong>of</strong> PWDG will be <strong>di</strong>scussed along with<br />
the extension <strong>of</strong> convergence theory to electromagnetic wave propagation. An outlook<br />
will address the potential <strong>of</strong> adaptive plane wave methods and Trefftz-DG methods in<br />
for other singularly perturbed boundary value problems.<br />
References<br />
[1] M. Amara, R. Djellouli, and C. Farhat, Convergence analysis <strong>of</strong> a <strong>di</strong>scontin- uous Galerkin<br />
method with plane waves and lagrange multipliers for the solution <strong>of</strong> Helmholtz problems,<br />
SIAM J. Numer. Anal., 47 (2009), pp. 1038–1066.<br />
[2] I. Babuˇska and J. Melenk, The partition <strong>of</strong> unity method, Int. J. Numer. Methods Eng., 40<br />
(1997), pp. 727–758.<br />
[3] A. Buffa and P. Monk, Error estimates for the ultra weak variational formulation <strong>of</strong> the<br />
Helmholtz equation, Math. Mod. Numer. Anal., 42 (2008), pp. 925–940. Published online<br />
August 12, 2008, DOI 10.1051/m2an:2008033.<br />
9
Plenary lectures<br />
[4] O. Cessenat and B. Després, Application <strong>of</strong> an ultra weak variational formulation <strong>of</strong> elliptic<br />
PDEs to the two-<strong>di</strong>mensional Helmholtz equation, SIAM J. Numer. Anal., 35 (1998), pp.<br />
255–299.<br />
[5] — , Using plane waves as base functions for solving time harmonic equations with the ultra<br />
weak variational formulation, J. Computational Acoustics, 11 (2003), pp. 227–238.<br />
[6] C. Gittelson, R. Hiptmair, and I. Perugia, Plane wave <strong>di</strong>scontinuous Galerkin methods:<br />
Analysis <strong>of</strong> the h-version, Math. Model. Numer. Anal., 43 (2009), pp. 297– 331.<br />
[7] R. Hiptmair, A. Moiola, and I. Perugia, Plane wave <strong>di</strong>scontinuous Galerkin methods for<br />
the 2D Helmholtz equation: Analysis <strong>of</strong> the p-version, Report 2009-20, SAM, ETH Zürich,<br />
Zürich, Switzerland, 2009. Submitted to SINUM.<br />
[8] T. Huttunen, M. Malinen, and P. Monk, Solving Maxwell’s equations using the ultra weak<br />
variational formulation, J. Comp. Phys., 223 (2007), pp. 731–758.<br />
[9] T. Huttunen and P. Monk, The use <strong>of</strong> plane waves to approximate wave propagation in<br />
anisotropic me<strong>di</strong>a, J. Computational Mathematics, 25 (2007), pp. 350–367.<br />
[10] T. Huttunen, P. Monk, and J. Kaipio, Computational aspects <strong>of</strong> the ultra-weak variational<br />
formulation, J. Comp. Phys., 182 (2002), pp. 27–46.<br />
[11] A. Moiola, R. Hiptmair, and I. Perugia, Approximation by plane waves, Report 2009-27,<br />
SAM, ETH Zürich, Zürich, Switzerland, August 2009.<br />
[12] P. Rouch and P. Ladeveze, The variational theory <strong>of</strong> complex rays: A pre<strong>di</strong>ctive tool for<br />
me<strong>di</strong>um-frequency vibrations, Comp. Meth. Appl. Mech. Engr., 15 (2003), pp. 3301–3315.<br />
[13] R. Tezaur and C. Farhat, Three-<strong>di</strong>mensional <strong>di</strong>scontinuous Galerkin elements with plane<br />
waves and lagrange multipliers for the solution <strong>of</strong> mid-frequency Helmholtz problems, Int.<br />
J. Numer. Meth. Engr., 66 (2006), pp. 796–815.<br />
10
Plenary lectures<br />
Mimetic finite <strong>di</strong>fference method for solving PDEs on polygonal<br />
and polyhedral meshes<br />
K. Lipnikov<br />
Los Alamos National Laboratory, Theoretical Division, MS B284<br />
Los Alamos, NM, USA, 87545<br />
lipnikov@lanl.gov<br />
A successful <strong>di</strong>scretization method inherits or mimics fundamental properties <strong>of</strong> the<br />
PDE such as conservation laws, symmetries, positivity structures and maximum principles.<br />
Construction <strong>of</strong> such a method is made more <strong>di</strong>fficult when the mesh is <strong>di</strong>storted<br />
so that it can conform and adapt to the physical domain and problem solution. The<br />
talk is about one such method - the mimetic finite <strong>di</strong>fference (MFD) method. The MFD<br />
method can be applied for solving problems with full tensor coefficients on unstructured<br />
polygonal and polyhedral meshes. These meshes may include arbitrary elements:<br />
tetrahedrons, pyramids, hexahedrons, degenerated and non-convex polyhedrons, and<br />
generalized polyhedrons.<br />
The classical MFD method has many similarities with other low-order <strong>di</strong>scretization<br />
methods, in particular, with the low-order mixed finite element method. Both methods<br />
try to preserve fundamental properties <strong>of</strong> physical and mathematical models. The essential<br />
<strong>di</strong>fference is that the MFD method uses only the surface representation <strong>of</strong> <strong>di</strong>screte<br />
unknowns to build elemental stiffness and mass matrices. Since no basis functions (no<br />
extension inside the mesh element) is required, practical implementation <strong>of</strong> the MFD<br />
method is simple for complex elements. Existence (but not uniqueness!) <strong>of</strong> real basis<br />
functions correspon<strong>di</strong>ng to stiffness and mass matrices is used only in the convergence<br />
analysis.<br />
The talk introduces basic elements <strong>of</strong> the MFD method. Its various generalizations<br />
developed over the last 5 years will be presented in other talks at this session. As an illustration<br />
<strong>of</strong> the mimetic methodology, I present construction <strong>of</strong> the MFD method for <strong>di</strong>ffusion<br />
problem on generalized polyhedral meshes (joint work with F.Brezzi, M.Shashkov<br />
and V.Simoncini). I describe families <strong>of</strong> <strong>di</strong>screte MFD schemes, summarize known theoretical<br />
results and verify them with numerical experiments.<br />
11
Plenary lectures<br />
The solution <strong>of</strong> time harmonic wave equations using complete<br />
families <strong>of</strong> elementary solutions<br />
Peter Monk<br />
Department <strong>of</strong> Mathematical Sciences, University <strong>of</strong> Delaware<br />
Newark, DE 19716, USA<br />
monk@math.udel.edu<br />
This presentation is devoted to <strong>di</strong>scussing plane wave methods for approximating the<br />
time-harmonic wave equation paying particular attention to the Ultra Weak Variational<br />
Formulation (UWVF). This method is essentially an upwind Discontinuous Galerkin<br />
(DG) method in which the approximating basis functions are special traces <strong>of</strong> solutions<br />
<strong>of</strong> the underlying wave equation. In the classical UWVF, due to Cessenat and Després,<br />
sums <strong>of</strong> plane wave solutions are used element by element to approximate the global<br />
solution. For these basis functions, convergence analysis and considerable computational<br />
experience shows that, under mesh refinement, the method exhibits a high order <strong>of</strong><br />
convergence depen<strong>di</strong>ng on the number <strong>of</strong> plane wave used on each element. Convergence<br />
can also be achieved by increasing the number <strong>of</strong> basis functions on a fixed mesh (or<br />
a combination <strong>of</strong> the two strategies). However ill-con<strong>di</strong>tioning arising from the plane<br />
wave basis can ultimately destroy convergence. This is particularly a problem near a<br />
reentrant corner where we expect to need to refine the mesh.<br />
The presentation will start with a summary <strong>of</strong> the UWVF and some typical analytical<br />
and numerical results for the Hemholtz equation. It may be that <strong>di</strong>fferent basis functions<br />
need to be used in <strong>di</strong>fferent parts <strong>of</strong> the domain. I shall present some numerical results<br />
investigating convergence on an L-shaped domain using singular Bessel functions near<br />
the corner. An alternative, that also extends to 3D, is to use polynomial basis functions<br />
on small elements. Using mixed finite element methods, we can view the UWVF as a<br />
hybri<strong>di</strong>zation strategy and I shall also present theoretical and numerical results for this<br />
approach.<br />
Although neither the Bessel function or the plane wave UWVF are free <strong>of</strong> <strong>di</strong>spersion<br />
error (pollution error) they can provide a method that can use large elements and small<br />
number <strong>of</strong> degrees <strong>of</strong> freedom per wavelength to approximate the solution. Extensions<br />
to Maxwell’s equations and elasticity will be briefly <strong>di</strong>scussed. Perhaps the main open<br />
problems are how to improve on the bi-conjugate gra<strong>di</strong>ent method that is currently used<br />
to solve the linear system, and how to adaptively refine the approximation scheme.<br />
12
Invited talks<br />
13
Invited talks<br />
Mimetic <strong>di</strong>scretization <strong>of</strong> the <strong>di</strong>ffusion problem: a higher<br />
order method and a posteriori error estimates<br />
L. Beirão da Veiga<br />
Department <strong>of</strong> Mathematics “F.Enriques”, University <strong>of</strong> Milan,<br />
Via Sal<strong>di</strong>ni 50, 20133 Milano, Italy<br />
lourenco.beirao@unimi.it<br />
The main characteristic <strong>of</strong> the Mimetic Finite Difference (MFD) method, when compared<br />
to a more standard finite element approach, is that the basis functions related to<br />
the <strong>di</strong>screte degrees <strong>of</strong> freedom are not explicitly defined. As a consequence, the operators<br />
and other quantities appearing in the problem must be approximated by <strong>di</strong>screte<br />
counterparts that satisfy finite <strong>di</strong>mensional analogs <strong>of</strong> some fundamental property. This<br />
approach allows for a greater flexibility <strong>of</strong> the mesh and the possibility to mimic intrinsic<br />
properties <strong>of</strong> the <strong>di</strong>fferential problem under study. In particular, general polyhedral (or<br />
polygonal in 2 <strong>di</strong>mensions) meshes, even with non convex and non matching elements,<br />
can be adopted. This talk covers two developements on the mimetic approximation <strong>of</strong><br />
the <strong>di</strong>ffusion problem in mixed form.<br />
The first one is a higher order version <strong>of</strong> the original approximation proposed by Brezzi<br />
et al. in 2005, which guarantees an O(h 2 ) approximation both for pressures and fluxes. A<br />
post-processing procedure allows morevoer to build a piecewise quadratic pressure with<br />
improved approximation properties. We note that a mo<strong>di</strong>fied consistency con<strong>di</strong>tion must<br />
be adopted in order to mantain the order <strong>of</strong> accuracy also when the <strong>di</strong>ffusivity tensor is<br />
variable across the domain.<br />
In the second part <strong>of</strong> the talk we introduce a local a-posteriori error estimator for the<br />
method, which applies both to the low and the high order method. The error estimator is<br />
shown to be both reliable and efficient with respect to an energy type norm involving the<br />
post-processed pressure. Finally, the error in<strong>di</strong>cator is combined with a simple adaptive<br />
process and a set <strong>of</strong> numerical tests for the low order case is presented.<br />
15
Invited talks<br />
Nonpolynomial FEM with MPSPACK<br />
Timo Betcke<br />
Department <strong>of</strong> Mathematics, University <strong>of</strong> Rea<strong>di</strong>ng, Whiteknights,<br />
PO Box 220, Berkshire, RG6 6AX, United Kingdom<br />
The understan<strong>di</strong>ng <strong>of</strong> nonpolynomial finite element methods for Helmholtz problems<br />
has seen tremendous progress over recent years. A common feature <strong>of</strong> most implementations<br />
is that they are based on meshes with element sizes <strong>of</strong> a few wavelengths<br />
in <strong>di</strong>ameter, having local approximation spaces consisting <strong>of</strong> plane waves or sometimes<br />
Fourier-Bessel functions. This follows the philosophy <strong>of</strong> standard finite element methods<br />
<strong>of</strong> having many elements and few basis functions per element, lea<strong>di</strong>ng to large and sparse<br />
matrices with relatively small dense blocks.<br />
In this talk we describe implementations <strong>of</strong> nonpolynomial finite elements based on<br />
the philosophy <strong>of</strong> using as few elements as possible and to achieve convergence by pure<br />
p-refinement on each <strong>of</strong> these global elements, where the types <strong>of</strong> basis functions can<br />
<strong>di</strong>ffer from one element to another. The goal is to use on each element basis functions<br />
that locally guarantee fast exponential convergence. This global element approach leads<br />
to moderately sized, essentially dense matrices.<br />
We present several examples based on the freely available Matlab Toolbox MPSPACK<br />
developed by Barnett and Betcke for the solution <strong>of</strong> Helmholtz problems using global<br />
elements and various types <strong>of</strong> nonpolynomial basis functions. In particular, we show<br />
how to implement a simple and fast nonpolynomial finite element method for scattering<br />
from convex and nonconvex polygons, which gives exponentially accurate results.<br />
16
Invited talks<br />
Optimization-based computational models, or how to let<br />
someone else do the hard work<br />
Pavel Bochev<br />
Applied Mathematics and Applications, San<strong>di</strong>a National Laboratories 1<br />
MS 1320, Albuquerque, NM 87185, USA<br />
pbboche@san<strong>di</strong>a.gov<br />
Discretization is a model reduction process which converts infinite <strong>di</strong>mensional mathematical<br />
models into finite <strong>di</strong>mensional algebraic equations that can be solved on a<br />
computer. Consequently, <strong>di</strong>scretization is accompanied by inevitable losses <strong>of</strong> information<br />
which can adversely affect the pre<strong>di</strong>ctiveness <strong>of</strong> the <strong>di</strong>screte models. Compatible<br />
and regularized <strong>di</strong>scretizations control these losses <strong>di</strong>rectly by choosing suitable field<br />
representations and/or by mo<strong>di</strong>fications <strong>of</strong> the correspon<strong>di</strong>ng variational forms. Such<br />
methods excel in control- ling “structural” information losses responsible for the stability<br />
and well-posedness <strong>of</strong> the <strong>di</strong>screte equations. However, they encounter considerable<br />
<strong>di</strong>fficulties in at least two cases: multi-physics models which combine constituent components<br />
with fundamentally <strong>di</strong>fferent mathematical properties, and loss <strong>of</strong> “qualitative”<br />
properties such as maximum principles, monotonicity and local bounds preservation.<br />
In this talk we show optimization ideas can be used to control externally information<br />
losses which are <strong>di</strong>fficult (or impractical) to manage <strong>di</strong>rectly in the <strong>di</strong>scretization<br />
process. This allows us to improve pre<strong>di</strong>ctiveness <strong>of</strong> computational models, increase<br />
robustness and accuracy <strong>of</strong> solvers, and enable efficient reuse <strong>of</strong> code. Two examples<br />
will be presented: an optimization-based framework for multi-physics coupling, and an<br />
optimization-based algorithm for constrained interpolation (remap). In the first case,<br />
our approach allows to synthesize a robust and efficient solver for a coupled multiphysics<br />
problem from simpler solvers for its constituent components. To illustrate the<br />
scope <strong>of</strong> the approach we derive a robust and efficient solver for nearly hyperbolic PDEs<br />
from standard, <strong>of</strong>f-the-shelf algebraic multigrid solvers for the Poisson equation, which<br />
by themselves cannot solve the original equations. The second example demonstrates<br />
how optimization ideas enable design <strong>of</strong> high- order conservative, monotone, bounds<br />
preserving remap and transport schemes which are linearity preserving on arbitrary<br />
unstructured grids, inclu<strong>di</strong>ng grids with polyhedral and polygonal cells.<br />
This is a joint work with D. Ridzal , G. Scovazzi (SNL) and M. Shashkov (LANL)<br />
17
Invited talks<br />
Mixed finite elements as inverse systems <strong>of</strong> <strong>di</strong>fferential forms<br />
Snorre Christiansen<br />
CMA c/o Department <strong>of</strong> Mathematics, University <strong>of</strong> Oslo,<br />
P.O. Box 1053 Blindern, NO-0316 Oslo, Norway<br />
snorrec@math.uio.no<br />
Conforming mixed finite elements for the grad, curl and <strong>di</strong>v operators can be thought<br />
<strong>of</strong> in terms <strong>of</strong> <strong>di</strong>fferential forms. More precisely we argue that they are inverse systems<br />
<strong>of</strong> finite <strong>di</strong>mensional complexes <strong>of</strong> <strong>di</strong>fferential forms, i.e. special functors from the set<br />
<strong>of</strong> geometric objects in the mesh (ordered by inclusion) to the category <strong>of</strong> <strong>di</strong>fferential<br />
complexes (equipped with pull-back). The Galerkin spaces are then recovered from the<br />
local data as inverse limits in the sense <strong>of</strong> category theory. We show how the theory <strong>of</strong><br />
mixed finite elements can be developed in this setting, which naturally accommodates<br />
polyhedral meshes as well as non-polynomial functions.<br />
18
Invited talks<br />
Regularized Poincaré operator and p version approximation<br />
<strong>of</strong> the Maxwell eigenvalue problem<br />
Martin Costabel<br />
IRMAR, Université de Rennes 1, Campus de Beaulieu,<br />
35042, Rennes Cedex, France<br />
martin.costabel@univ-rennes1.fr<br />
The regularized Poincaré integral operators are a new tool in the analysis <strong>of</strong> <strong>di</strong>fferential<br />
forms that have a variety <strong>of</strong> <strong>di</strong>fferent applications. In particular, they allow the<br />
construction <strong>of</strong> vector and scalar potentials <strong>of</strong> optimal regularity in terms <strong>of</strong> fractional<br />
order Sobolev norms. This can be used to complete a crucial step in the pro<strong>of</strong> <strong>of</strong> the<br />
<strong>di</strong>screte compactness <strong>of</strong> the p version edge elements, and therefore <strong>of</strong> their spectrally<br />
correct approximation <strong>of</strong> the Maxwell eigenvalue problem.<br />
In the talk, these integral operators and their companions, the Bogovskiĭ integral<br />
operators, are presented with their main properties. Their role for the p version edge<br />
element approximation <strong>of</strong> the Maxwell eigenproblem is explained.<br />
Collaboration with:<br />
Daniele B<strong>of</strong>fi (Pavia), Monique Dauge (Rennes), Leszek Demkowicz (Austin), Ralf Hitmair<br />
(Zürich), Alan McIntosh (Canberra).<br />
References<br />
[1] M. Costabel, A. McIntosh: On Bogovskiĭ and regularized Poincaré integral operators<br />
for de Rham complexes on Lipschitz domains. Math. Z. 265, No 2 (2010), 297–320.<br />
[2] D. B<strong>of</strong>fi, M. Costabel, M. Dauge, L. Demkowicz, R. Hiptmair: Discrete compactness<br />
for the p-version <strong>of</strong> <strong>di</strong>screte <strong>di</strong>fferential forms. IRMAR-Preprint 09-39, Universite<br />
Rennes 1 2009. HAL : hal-00420150 ; arXiv : 0909.5079<br />
19
Invited talks<br />
Aerodynamic Analysis and Design using an Isogeometric<br />
Approach<br />
R. Duvigneau<br />
INRIA Sophia-Antipolis Mé<strong>di</strong>terranée,<br />
BP 93, 2004 route des lucioles 06902 Sophia-Antipolis, France<br />
Automated design optimization procedures are now commonly used in aerodynamics,<br />
for drag minimization exercises or target pressure reconstruction problems. However,<br />
these procedures are <strong>of</strong>ten cumbersome, since they are achieved by coupling existing sophisticated<br />
numerical tools, originating from Computer Aided Design (CAD), optimization<br />
(descent or evolutionary methods), simulation (finite-element, finite-volume), grid<br />
generation (structured or unstructured), etc. Moreover, these tools are based on <strong>di</strong>fferent<br />
geometrical representations, ranging form NURBS (Non-Uniform Rational B-Splines)<br />
used in CAD, grids used by physical solvers, to ad-hoc design parameterizations (angles,<br />
lengths, volumes, etc) <strong>of</strong>ten used in optimization. Hence, several geometrical transformations<br />
are required in practice to couple these tools, that could have severe consequences:<br />
loss <strong>of</strong> accuracy, generation <strong>of</strong> spurious noise, local optima, non-<strong>di</strong>fferentiability, etc.<br />
To definitely overcome these <strong>di</strong>fficulties, T. Hughes proposed a few years ago to consider<br />
a design process entirely based on NURBS, in which the simulation is achieved<br />
using parameterized surfaces and volumes instead <strong>of</strong> grids.<br />
In this study, the application <strong>of</strong> isogeometric analysis to aerodynamic problems will<br />
be explored. Especially, a variational approach is carried out using a NURBS basis to<br />
define the geometry and solution fields for compressible Euler equations, yiel<strong>di</strong>ng highorder,<br />
adaptive and hierarchical numerical schemes. The <strong>di</strong>fferences with respect to<br />
the classical approach are underlined and some examples are demonstrated, concerning<br />
aerodynamic analysis and design problems.<br />
20
Invited talks<br />
Isogeometric Analysis in Fluid Dynamics: Divergence-Free<br />
Discretizations for the Stokes and Navier-Stokes Equations<br />
Annalisa Buffa, John A. Evans, Thomas J.R. Hughes, Giancarlo Sangalli<br />
Institute for Computational Engineering and Sciences, University <strong>of</strong> Texas at Austin,<br />
3517 North Hills Drive, #T301, Austin, TX, USA 78731<br />
evans@ices.utexas.edu<br />
In this talk, I <strong>di</strong>scuss recently developed isogeometric <strong>di</strong>scretizations for the Stokes and<br />
Navier-Stokes equations. These <strong>di</strong>scretizations are a smooth generalization <strong>of</strong> classical<br />
Raviart-Thomas elements in two <strong>di</strong>mensions and Raviart-Thomas-Nedelec elements in<br />
three <strong>di</strong>mensions. When coupled with a weak enforcement <strong>of</strong> slip boundary con<strong>di</strong>tions,<br />
these <strong>di</strong>scretizations lead to provably inf-sup stable and high-order convergent numerical<br />
methods. As such, they are a natural can<strong>di</strong>date for use in the simulation <strong>of</strong> Stokes or<br />
Navier-Stokes flow. In fact, when applied to incompressible flows, these <strong>di</strong>scretizations<br />
provide pointwise <strong>di</strong>vergence-free velocity fields. This is an especially useful property<br />
for coupled flow-transport problems where mass conservation is <strong>of</strong> great importance.<br />
Furthermore, these <strong>di</strong>scretizations admit a natural decomposition <strong>of</strong> the velocity basis<br />
into <strong>di</strong>vergence-free and complementary sub-bases, a useful property for fast and robust<br />
solution <strong>of</strong> the resulting <strong>di</strong>screte system <strong>of</strong> equations. Numerical examples will be shown<br />
illustrating optimal convergence rates for model problems and the effectiveness <strong>of</strong> this<br />
technology for general incompressible flows.<br />
21
Invited talks<br />
Canonical Families <strong>of</strong> Finite Element Differential Forms and<br />
Their Properties<br />
R. Falk<br />
Department <strong>of</strong> Mathematics, Rutgers University,<br />
110 Frelinghuysen Road, Piscataway, NJ 08854<br />
falk@math.rutgers.edu<br />
We consider two families <strong>of</strong> finite element <strong>di</strong>fferential forms defined on simplicial<br />
meshes in R n . The first family, which we denote by PrΛ k , is based on the space Pr<br />
<strong>of</strong> polynomials <strong>of</strong> degree less than or equal to r and is a generalization <strong>of</strong> the BDM and<br />
second kind Nédélec elements, while the second, which we denote by P − r Λ k , is based on<br />
a polynomial space between Pr−1 and Pr and is a generalization <strong>of</strong> the Raviart-Thomas<br />
and first kind Nédélec elements. We consider a number <strong>of</strong> important properties <strong>of</strong> these<br />
spaces and especially their connections.<br />
22
Invited talks<br />
A brief review <strong>of</strong> wave-based computational methods for<br />
aero-acoustics<br />
G. Gabard, R.J. Astley, P. Gamallo, G. Kennedy<br />
ISVR, University <strong>of</strong> Southampton,<br />
University Road, Southampton, SO17 1BJ, UK<br />
gabard@soton.ac.uk<br />
Several wave-based methods developed for pre<strong>di</strong>cting sound propagation in non- uniform<br />
flows are reviewed. The basic principle <strong>of</strong> wave-based methods is to incorporate<br />
some known properties <strong>of</strong> the underlying physics into the numerical model. For instance,<br />
instead <strong>of</strong> using standard polynomials or Chebyshev polynomials to approximate the solution,<br />
wave-based methods generally use canonical solutions such as Green’s functions<br />
or plane waves to construct a local description <strong>of</strong> the solution. The methods described<br />
in this paper include the Green’s function <strong>di</strong>scretisation, the partition <strong>of</strong> unity finite<br />
element method and the wave-based <strong>di</strong>scontinuous Galerkin method. The principles<br />
<strong>of</strong> these methods are described and their performance and shortcomings are <strong>di</strong>scussed.<br />
While the methods are intended to be used with strongly inhomogeneous propagation<br />
me<strong>di</strong>a, the canonical solutions used as interpolating functions are only valid for uniform<br />
coefficients. Another issue that will be <strong>di</strong>scussed is that <strong>of</strong> mesh generation.<br />
23
Invited talks<br />
Mimetic finite <strong>di</strong>fferences for eigenvalue problems<br />
Francesca Gar<strong>di</strong>ni<br />
Università degli Stu<strong>di</strong> <strong>di</strong> Pavia,<br />
via Ferrata 1, 27100 Pavia, Italy<br />
francesca.gar<strong>di</strong>ni@unipv.it<br />
In the last two decades, a variety <strong>of</strong> mimetic <strong>di</strong>scretizations for the numerical approximation<br />
<strong>of</strong> Partial Differential Equations has been developed. These numerical methods<br />
are designed to mimic some fundamental properties <strong>of</strong> the continuous problems through<br />
the definition <strong>of</strong> <strong>di</strong>screte operators that satisfy <strong>di</strong>screte analogs <strong>of</strong> the fundamental relations<br />
<strong>of</strong> vector and tensor calculus, e.g., Gauss-Green’s identities.<br />
In this talk, we extend the mimetic finite <strong>di</strong>fference method to the solution <strong>of</strong> eigenvalue<br />
problems. For more general results and exhaustive bibliography we refer to [3].<br />
The convergence analysis is set in the framework <strong>of</strong> the classical spectral approximation<br />
theory for compact operators; see [1]. The correct spectral approximation, i.e., a good<br />
approximation <strong>of</strong> eigenmodes and the absence <strong>of</strong> spurious modes, is thus obtained by<br />
proving the uniform convergence <strong>of</strong> the <strong>di</strong>screte solution operators to the continuous one.<br />
The uniform convergence is a <strong>di</strong>rect consequence <strong>of</strong> a new a priori error estimate for the<br />
source problem, which improves the original a priori error analysis in [2] by relaxing the<br />
H 1 -regularity assumption for the forcing term. Here, we obtain an improved error bound<br />
in terms <strong>of</strong> the L 2 -norm <strong>of</strong> the right-hand side <strong>of</strong> the source problem by reformulating<br />
the latter using an elemental lifting operator. Numerical results confirming the optimal<br />
behavior <strong>of</strong> the method are presented.<br />
The talk is based on a joint work with Andrea Cangiani (Università <strong>di</strong> Milano Bicocca)<br />
and Gianmarco Manzini (Istituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Informatiche -<br />
CNR).<br />
References<br />
[1] Babuˇska I., and Osborn J. 1991 Eigenvalue Problems. In Handbook <strong>of</strong> Numerical Analysis<br />
(ed. P. Ciarlet and J. Lions), vol. 2, pp. 641–787. North Holland.<br />
[2] Brezzi F., Lipnikov K., and Shashkov M. 2005 Convergence <strong>of</strong> the mimetic finite <strong>di</strong>fference<br />
method for <strong>di</strong>ffusion problems on polyhedral meshes. SIAM J. Numer. Anal. 43, 1872–1896.<br />
24
Invited talks<br />
[3] Cangiani A., Gar<strong>di</strong>ni F., and Manzini G. 2009 Convergence <strong>of</strong> the mimetic finite <strong>di</strong>fference<br />
method for eigenvalue problems in mixed form. Technical Report 31PV09/24/0, IMATI-CNR<br />
(submitted to Comput. Methods Appl. Mech. Engrg.).<br />
25
Invited talks<br />
Dimensions and Bases <strong>of</strong> Bivariate Hierarchical Tensorproduct<br />
Splines<br />
Bert Jüttler<br />
Institute <strong>of</strong> Applied Geometry, Johannes Kepler University,<br />
Altenberger Str. 69, 4040 Linz, Austria<br />
bert.juettler@jku.at<br />
Due to their potential applications for isogeometric analysis, tensor-product spline<br />
spaces with local refinability have recently become an active topic <strong>of</strong> research. Most<br />
<strong>of</strong> the existing and ongoing work focuses on the so-called T-splines <strong>of</strong> Sederberg et al.<br />
These spline spaces, however, possess certain limitations, in particular with respect to<br />
the locality <strong>of</strong> refinement operations. The talk will recall the classical construction <strong>of</strong><br />
hierarchical tensor-product splines. In ad<strong>di</strong>tion, it will <strong>di</strong>scuss <strong>di</strong>mensions and bases <strong>of</strong><br />
hierarchically refined tensor-product spline spaces and it will show how do construct<br />
bases which are, non-negative, locally supported, and form a partition <strong>of</strong> unity.<br />
26
A 3D plane wave basis for elastic wave problems<br />
Teemu Luostari † , Tomi Huttunen † and Peter Monk ‡<br />
† Department <strong>of</strong> Physics and Mathematics, University <strong>of</strong> Eastern Finland,<br />
Kuopio Campus, P.O. Box 1627, FIN-70211, Finland.<br />
‡ Department <strong>of</strong> Mathematical Sciences, University <strong>of</strong> Delaware,<br />
Newark, DE 19716, USA.<br />
teemu.luostari@uef.fi<br />
Invited talks<br />
Elastic wave problems and the Navier equation arise in many practical applications<br />
in engineering. However, the modeling <strong>of</strong> elastic waves is more challenging and time<br />
consuming than the modeling <strong>of</strong> the acoustics since the solution consists <strong>of</strong> <strong>di</strong>fferent<br />
wave modes (e.g. P-, SH- and SV-wave components). These modes can propagate with<br />
<strong>di</strong>fferent speeds in a me<strong>di</strong>um. Therefore dense meshes are <strong>of</strong>ten needed in the standard<br />
finite element method (FEM) to model the problem accurately. We shall consider a<br />
competitive method to FEM, the ultra weak variational formulation (UWVF), which is<br />
a special form <strong>of</strong> the <strong>di</strong>scontinuos Galerkin method (DGM), see [2,3]. The UWVF was<br />
first introduced and developed to the Helmholtz equation and Maxwell’s equation by<br />
Cessenat and Després [1]. The feasibility <strong>of</strong> the UWVF for the elastic wave problems in<br />
2D was stu<strong>di</strong>ed in [4].<br />
The UWVF is a non-polynomial volume based method that uses plane wave basis<br />
functions, i.e. physical basis functions. In the elastic-UWVF the plane wave basis<br />
consists <strong>of</strong> pressure and shear waves. Therefore in the UWVF the number <strong>of</strong> basis<br />
functions is chosen separately for <strong>di</strong>fferent wave modes. Using the plane wave basis the<br />
computational burden is reduced because the integrals encountered in the method can<br />
be computed efficiently in closed form. However, the plane wave basis may produce an<br />
ill-con<strong>di</strong>tioned matrix system which deteriorates the accuracy.<br />
We shall consider the UWVF for the 3D elastic wave problem (Navier equation). The<br />
model problem is simple elastic wave propagation in a cubic domain. We investigate<br />
the performance <strong>of</strong> the UWVF with <strong>di</strong>fferent wave numbers and <strong>di</strong>fferent combinations<br />
<strong>of</strong> the P-, SH- and SV-wave basis function ratios. We shall show the convergence with<br />
<strong>di</strong>fferent number <strong>of</strong> basis functions and <strong>di</strong>scuss the con<strong>di</strong>tioning <strong>of</strong> the method.<br />
References<br />
[1] Cessenat O, Després B. Application <strong>of</strong> an ultra weak variational formulation <strong>of</strong> elliptic<br />
PDEs to the two-<strong>di</strong>mensional Helmholtz problem. SIAM Journal <strong>of</strong> Numerical Analysis<br />
1998; 35(1):255–299.<br />
27
Invited talks<br />
[2] Gabard G. Discontinuous Galerkin methods with plane waves for time-harmonic problems.<br />
Journal <strong>of</strong> Computational Physics, 225(2):1961–1984, 2007.<br />
[3] Huttunen T, Malinen M, Monk P. Solving Maxwell’s equations using the ultra weak variational<br />
formulation. Journal <strong>of</strong> Computational Physics 2007; 223(2):731–758.<br />
[4] Huttunen T, Monk P, Collino F, Kaipio JP. The ultra weak variational formulation for<br />
elastic wave problems. SIAM Journal on Scientific Computing, 25(5):1717–1742, 2004.<br />
28
Invited talks<br />
Beyond NURBS: Non-Standard CAGD Tools in Isogeometric<br />
Analysis<br />
Carla Manni † , Francesca Pelosi † and Maria Lucia Sampoli ‡<br />
† Department <strong>of</strong> Mathematics, Università <strong>di</strong> Roma “Tor Vergata”,<br />
Via della Ricerca Scientifica 1, 00133 Roma, Italy<br />
‡ Department <strong>of</strong> Mathematics and Computer Sciences, Università <strong>di</strong> Siena,<br />
Pian dei Mantellini 44, 53100 Siena, Italy<br />
manni@mat.uniroma2.it<br />
NURBS-based isogeometric analysis can be seen as a superset <strong>of</strong> FEM and <strong>of</strong>fers a<br />
powerful tool in the area <strong>of</strong> numerical methods for <strong>di</strong>fferential problems. The NURBSbased<br />
approach leads to remarkable results, not only in the context <strong>of</strong> structural analysis,<br />
but also in advection dominated flow phenomena, <strong>of</strong>ten characterized by sharp layers<br />
involving very strong gra<strong>di</strong>ents.<br />
Nevertheless the rational model (NURBS) presents several drawbacks both from the<br />
geometrical and the analytical point <strong>of</strong> view. Therefore our attention has been focused on<br />
overcoming the problems <strong>of</strong> NURBS standards, analyzing Isogeometric Analysis schemes<br />
with <strong>di</strong>fferent bases, but still equipped with refinement properties, geometric features<br />
and classical algorithms as degree elevation, knot insertion, <strong>di</strong>fferentiation formulas,<br />
etc. . . Such spaces admit a representation in term <strong>of</strong> functions which are a natural extension<br />
<strong>of</strong> polynomial B-splines so that they are properly referred to as generalized<br />
B-splines.<br />
In this talk, after a general presentation, we focus on suitable spaces <strong>of</strong> generalized Bsplines<br />
whose <strong>di</strong>stinguishing property is the ability to efficiently describe sharp variations<br />
without introducing extraneous oscillations. Thus, they seem particularly well suited to<br />
face advection dominated flow phenomena.<br />
29
Invited talks<br />
High-order accurate nodal formulation <strong>of</strong> the Mimetic Finite<br />
Difference Method for Elliptic Problems<br />
G. Manzini<br />
Istituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Informatiche, CNR,<br />
via Ferrata 1, 27100 Pavia, Italy,<br />
Centro per la Simulazione Numerica Avanzata CeSNA, IUSS,<br />
27100 Pavia, Italy<br />
marco.manzini@imati.cnr.it<br />
The nodal formulation <strong>of</strong> the Mimetic Finite Difference Method for two-<strong>di</strong>mensional<br />
elliptic problems is based on a consistency con<strong>di</strong>tion that reproduces exactly the integration<br />
by parts formula for linear polynomials. Increasing the degrees <strong>of</strong> the polynomials<br />
for which a consistency con<strong>di</strong>tion holds is a possible way to achieve higher order formulations,<br />
but requires some care in the proper formulation <strong>of</strong> the method and in its<br />
implementation. Basically, a new kind <strong>of</strong> degrees <strong>of</strong> freedom, moments <strong>of</strong> polynomials,<br />
are introduced in the mimetic formulation to represent the elemental integral originated<br />
by the integration by parts formula. Likewise, the edge integrals requires new ad<strong>di</strong>tional<br />
nodes and high order quadrature formulas <strong>of</strong> Gauss-Lobatto type. For example, when<br />
quadratic polynomials are considered to implement a formally third-accurate method we<br />
need one more degree <strong>of</strong> freedom per cell and one more degree <strong>of</strong> freedom per edge.<br />
From a theoretical standpoint, optimal convergence rate is proved in a mesh-dependent<br />
H 1 norm for polynomials <strong>of</strong> any order for meshes with very general shaped elements.<br />
Experimental results confirm this behavior and show the optimal convergence rate in<br />
this norm for polynomials <strong>of</strong> degree up to five when the method is applied to meshes <strong>of</strong><br />
quadrilateral and (possibly non-convex) polygonal cells. On such meshes, we have also<br />
an experimental in<strong>di</strong>cation <strong>of</strong> optimal behavior when approximation errors are measured<br />
using the L 2 norm.<br />
This is a joint work with L. Beirao da Veiga and K. Lipnikov.<br />
30
Invited talks<br />
Mapping properties <strong>of</strong> Helmholtz integral operators and<br />
their application to the hp-BEM<br />
J.M. Melenk<br />
Institut für Analysis und Scientific Computing (E101), Technische Universität Wien<br />
Wiedner Hauptstraße 8-10, A-1040 Wien, Austria<br />
melenk@tuwien.ac.at<br />
For the Helmholtz equation (with wavenumber k) on analytic curves and surfaces Γ<br />
we analyze the mapping properties <strong>of</strong> the single layer, double layer as well a combined<br />
potential boundary integral operator. A k-explicit regularity theory for the single layer<br />
and double layer potentials is developed, in which these operators are decomposed into<br />
three parts: the first part is the single or double layer potential for the Laplace equation,<br />
the second part is an operator with finite shift properties, and the third part is an<br />
operator that maps into a space <strong>of</strong> piecewise analytic functions. For all parts, the kdependence<br />
is made explicit. We also develop a k-explicit regularity theory for the<br />
inverse <strong>of</strong> the combined potential operator A = ±1/2 + K ′ − ikV and its adjoint, where<br />
V and K are the single layer and double layer operators for the Helmholtz kernel. Under<br />
the assumption that A −1 L 2 (Γ)←L 2 (Γ) grows at most polynomially in k, the inverse A −1<br />
is decomposed into an operator A1 : L 2 (Γ) → L 2 (Γ) with bounds independent <strong>of</strong> k and<br />
a smoothing operator A2 that maps into a space <strong>of</strong> analytic functions on Γ. The kdependence<br />
<strong>of</strong> the mapping properties <strong>of</strong> A2 is made explicit. We show quasi-optimality<br />
(in an L 2 (Γ)-setting) <strong>of</strong> the hp-version <strong>of</strong> the Galerkin BEM applied to A under the<br />
assumption <strong>of</strong> scale resolution, i.e., the polynomial degree p is at least O(log k) and<br />
kh/p is bounded by a number that is sufficiently small, but independent <strong>of</strong> k. Under this<br />
assumption, the constant in the quasi-optimality estimate is independent <strong>of</strong> k. Numerical<br />
examples in 2D illustrate the theoretical results.<br />
This work is joint with M. Löhndorf (Vienna).<br />
31
Invited talks<br />
Approximation by plane waves<br />
Ralf Hiptmair † , Andrea Moiola † and Ilaria Perugia ‡<br />
† Seminar for Applied Mathematics, ETH,<br />
Rämistrasse 101, 8092 Zürich , Switzerland<br />
‡ <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia,<br />
Via Ferrata 1, 27100 Pavia - Italy<br />
hiptmair@sam.math.ethz.ch, moiola@sam.math.ethz.ch,<br />
ilaria.perugia@unipv.it<br />
The Trefftz methods are special finite element methods where the trial and test functions<br />
are solutions <strong>of</strong> the underlying PDE in the interior <strong>of</strong> each element. In the case<br />
<strong>of</strong> the homogeneous Helmholtz equation −∆u − ω 2 u = 0, plane waves (x ↦→ e iωx·d ) or<br />
circular/spherical waves are usually used as basis functions.<br />
One <strong>of</strong> the main steps in the convergence analysis <strong>of</strong> any <strong>of</strong> these methods is the pro<strong>of</strong><br />
<strong>of</strong> a best approximation estimate for the trial space. The only results <strong>of</strong> this kind for<br />
plane wave spaces available in literature (see [?] and [?]) are limited to two <strong>di</strong>mensional<br />
domains and are not completely satisfactory.<br />
In order to improve and generalize these estimates we proceed in two steps. In the<br />
first one, we show how u can be approximated by the so-called generalized harmonic<br />
polynomials, i.e. the circular (in two <strong>di</strong>mensions) and the spherical (in three <strong>di</strong>mensions)<br />
waves:<br />
x = re iψ ↦→ e ±ilψ <br />
x<br />
Jl(ωr), x ↦→ Yl,m |x|<br />
jl(ω|x|), 0 ≤ |m| ≤ l ∈ N.<br />
The main tool used is the Vekua transform, a bijective integral operator that maps<br />
Helmholtz solutions into harmonic functions on the same domain. This allows to reduce<br />
the problem to the simpler case <strong>of</strong> the approximation <strong>of</strong> harmonic functions by harmonic<br />
polynomials.<br />
In the second step we approximate these functions with plane waves. The link between<br />
plane waves and circular/spherical waves is provided by the Jacobi-Anger expansion. We<br />
truncate the expansion and give a bound on the solution <strong>of</strong> the linear system obtained;<br />
in three <strong>di</strong>mensions this requires a careful choice <strong>of</strong> the propagating <strong>di</strong>rections <strong>of</strong> the<br />
plane waves.<br />
We show that, for any solution u ∈ H K+1 (D) <strong>of</strong> the homogeneous Helmholtz equation,<br />
D ⊂ R N starshaped with <strong>di</strong>ameter h, N = 2, 3, it is possible to approximate u with a<br />
linear combination <strong>of</strong> p plane waves <strong>of</strong> given <strong>di</strong>rections {dk}k=1,...,p with error<br />
inf<br />
α∈C p<br />
<br />
u − p <br />
iωx·dk<br />
k=1αke ≤ C h j,ω,D K+1−j q −λ(K+1−j) uK+1,ω,D 32
Invited talks<br />
where the Sobolev norms are weighted with ω, the constant C depends on ωh (explicitly),<br />
j, k, {dk} and the shape <strong>of</strong> D. The number p <strong>of</strong> plane waves is linked to the parameter q<br />
as p = 2q + 1 in two <strong>di</strong>mensions and as p = (q + 1) 2 in three <strong>di</strong>mensions. The parameter<br />
λ depends only on the shape <strong>of</strong> D: in two <strong>di</strong>mensions it can be chosen arbitrarily close<br />
to 1 while in three <strong>di</strong>mensions we can only prove its positivity.<br />
The pro<strong>of</strong> <strong>of</strong> analogous bounds for Maxwell equations (curl curl u − ω 2 u = 0) is more<br />
<strong>di</strong>fficult because the Vekua operators are not appropriate for these equations. We use<br />
Herglotz functions, vector spherical harmonics and a special Jacobi-Anger formula to<br />
separate the vectorial plane/spherical waves that are solution <strong>of</strong> Maxwell equations from<br />
the ones that are not <strong>di</strong>vergence free. With these tools we can prove a h-estimate for<br />
spherical waves with the same order <strong>of</strong> convergence proved for the Helmholtz case but<br />
using only a few more basis functions.<br />
References<br />
[1] O. Cessenat and B. Despres, Application <strong>of</strong> an ultra weak variational formulation <strong>of</strong> elliptic<br />
PDEs to the two-<strong>di</strong>mensional Helmholtz equation, SIAM J. Numer. Anal. 35 1 (1998), 255–<br />
299.<br />
[2] R. Hiptmair, A. Moiola and I. Perugia, Approximation by plane waves, SAM Report 2009-<br />
27, ETH Zürich.<br />
[3] J.M. Melenk, On Generalized Finite Element Methods, Ph.D. thesis (1995), University <strong>of</strong><br />
Maryland, USA.<br />
33
Invited talks<br />
Isogeometric Analysis in Pavia<br />
Alessandro Reali<br />
Structural Mechanics Department, University <strong>of</strong> Pavia,<br />
Via Ferrata 1, 27100 Pavia, Italy<br />
alessandro.reali@unipv.it<br />
Isogeometric analysis (IGA), introduced by the pioneering work <strong>of</strong> Hughes et al. in<br />
2005, can now be considered more than just a promising technique in the framework <strong>of</strong><br />
numerical methods for PDEs, with many groups around the world contributing to its<br />
development and <strong>di</strong>ffusion. Some researchers from Pavia started working on IGA from<br />
its infancy and now an entire research group is active on this topic, studying problems<br />
ranging from mathematical analysis to code generation, from element technology to<br />
engineering applications in solid mechanics, fluid dynamics, electromagnetics, etc. The<br />
aim <strong>of</strong> this talk is to show some <strong>of</strong> the results obtained by our research group within<br />
this context.<br />
34
Invited talks<br />
Some estimates for hpk-refinement in Isogeometric Analysis<br />
L. Beirão da Veiga ∗ , A. Buffa † , J. Rivas ‡ and G. Sangalli §<br />
∗ <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong> “F. Enriques”; Via Sal<strong>di</strong>ni, 50, 20133 Milano, Italy.<br />
† IMATI, CNR; Via Ferrata 1, 27100 Pavia, Italy<br />
‡ Departamento de Matemáticas, Universidad del País Vasco-Euskal Herriko Unibertsitatea,<br />
Barrio Sarriena s/n, 48940 Leioa (Bizkaia), Spain<br />
§ <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia; Via Ferrata 1, 27100 Pavia, Italy<br />
lourenco.beirao@unimi.it, annalisa@imati.cnr.it, ju<strong>di</strong>th.rivas@ehu.es,<br />
giancarlo.sangalli@unipv.it<br />
New results on the approximation properties <strong>of</strong> two-<strong>di</strong>mensional Non-Uniform Rational<br />
B-splines (NURBS) spaces will be presented. More precisely, we will define a new<br />
projection operator into certain spaces <strong>of</strong> NURBS and give arror estimates in Sobolev<br />
norms which are explicit in the three <strong>di</strong>scretization parameters: mesh size, h, degree<br />
p and regularity k. We firstly construct a one-<strong>di</strong>mensional projector onto the space <strong>of</strong><br />
splines <strong>of</strong> degree p with continuous derivatives up to the order k − 1, with the restriction<br />
that 2k − 1 ≤ p, and obtain the correspon<strong>di</strong>ng error estimates. The extension to two<strong>di</strong>mensional<br />
spline approximations is done through a tensor product construction and,<br />
finally, the projector onto a NUBRS space is defined by multiplication with the weight<br />
and composition with the mapping associated to the considered NURBS space. These<br />
results are a first step in the analysis <strong>of</strong> the approximation properties <strong>of</strong> NURBS spaces<br />
in terms, not only <strong>of</strong> the mesh size, but also <strong>of</strong> the other <strong>di</strong>scretization parameters.<br />
However, the restiction over the regularity, k, still leaves the most interesting cases <strong>of</strong><br />
higher regularity (up to k = p) open.<br />
35
Invited talks<br />
A Posteriori Error Estimation for Highly Indefinite<br />
Helmholtz Problems<br />
Stefan A. Sauter<br />
Institut für Mathematik, Universität Zürich,<br />
Winterthurerstrasse 190, 8057 Zürich<br />
stas@math.uzh.ch<br />
In this talk, we will consider Galerkin <strong>di</strong>scretizations <strong>of</strong> highly indefinite Helmholtz<br />
problems. Standard a priori and a posteriori error estimates, typically, suffer from the<br />
ill-con<strong>di</strong>tioned behaviour <strong>of</strong> the <strong>di</strong>screte Galerkin operator for large wave-numbers k.<br />
Recently, it has been shown that - by choosing the polynomial order p in an hp-finite<br />
element space accor<strong>di</strong>ng to p = O(log(k)) - the optimal a priori error estimates are<br />
preserved. The result is based on a new regularity theory which employs a wave-number<br />
dependent frequency splitting <strong>of</strong> the solution. In this talk, we generalize this regularity<br />
theory for the posteriori error estimation <strong>of</strong> highly indefinite Helmholtz problems.<br />
This talk comprises joint work with Willy Dörfler.<br />
36
Nitsche-type Mortaring for Maxwell’s Equations<br />
Joachim Schöberl<br />
Institute for Analysis and Scientific Computing, Vienna University <strong>of</strong> Technology,<br />
Wiedner Hauptstrasse 8-10, Wien, Austria<br />
joachim.schoeberl@rwth-aachen.de<br />
Invited talks<br />
We propose a new method for treating transmission con<strong>di</strong>tions on non-matching<br />
meshes. The basic method is a hybrid version <strong>of</strong> Nitsche’s method. By introducing<br />
a scalar potential on the interface we obtain a robust method for the low frequency<br />
limit. In order to simplify numerical integration, we use smooth B-spline basis functions<br />
<strong>of</strong> Nedelec-type on the interface. We present numerical results for scalar equations and<br />
Maxwell’s equations in frequency domain.<br />
37
Invited talks<br />
A Convergent MFMFE for Highly Distorted Hexahedra<br />
Mary F. Wheeler<br />
Center for Subsurface Modeling, The University <strong>of</strong> Texas at Austin,<br />
1 University Station C0200, Austin, Texas 78712, USA<br />
mfw@ices.utexas.edu<br />
In this presentation, we develop a new mixed finite element method for elliptic problems<br />
on general quadrilateral and hexahedral grids that reduces to a cell-centered finite<br />
<strong>di</strong>fference scheme. A special non-symmetric quadrature rule is employed that yields a<br />
positive definite cell-centered system for the pressure by eliminating local velocities. In<br />
ad<strong>di</strong>tion, this scheme is shown to be accurate on highly <strong>di</strong>storted quadrilateral and hexahedral<br />
grids. Theoretical and numerical results in<strong>di</strong>cate first-order convergence for the<br />
pressure and face fluxes. Extensions to parabolic and two phase flow in porous me<strong>di</strong>a<br />
are <strong>di</strong>scussed.<br />
This work has been done in collaboration with G. Xue and I. Yotov.<br />
38
Invited talks<br />
Cochain projections in finite element exterior calculus<br />
R. Winther<br />
Centre <strong>of</strong> Mathematics for Applications (CMA), University <strong>of</strong> Oslo,<br />
CMA, P.O. Box 1053, Blindern, 0316 Oslo, Norway<br />
ragnar.winther@cma.uio.no<br />
We will focus on the role <strong>of</strong> cochain projections, i.e. projections which commute with<br />
the exterior derivative, in finite element exterior calculus. In particular, we will <strong>di</strong>scuss<br />
the relations between various uniform bounds on such projections and the convergence<br />
properties <strong>of</strong> the associated finite element methods for boundary value problems and<br />
eigenvalue problems. Furthermore, a new construction <strong>of</strong> local cochain projections <strong>of</strong><br />
Clément type will be presented.<br />
39
Invited talks<br />
A Multiscale Mortar Multipoint Flux Mixed Finite Element<br />
Method<br />
Ivan Yotov<br />
Department <strong>of</strong> Mathematics, University <strong>of</strong> Pittsburgh,<br />
301 Thackeray Hall, Pittsburgh, PA 15260, USA<br />
yotov@math.pitt.edu<br />
We present a multiscale mortar multipoint flux mixed finite element method for second<br />
order elliptic problems. The equations in the coarse elements (or subdomains) are<br />
<strong>di</strong>scretized on a fine grid scale by a multipoint flux mixed finite element method that<br />
reduces to cell-centered finite <strong>di</strong>fferences on irregular grids. The subdomain grids do<br />
not have to match across the interfaces. Continuity <strong>of</strong> flux between coarse elements is<br />
imposed via a mortar finite element space on a coarse grid scale. With an appropriate<br />
choice <strong>of</strong> polynomial degree <strong>of</strong> the mortar space, we derive optimal order convergence<br />
on the fine scale for both the multiscale pressure and velocity, as well as the coarse<br />
scale mortar pressure. Some superconvergence results are also derived. The algebraic<br />
system is reduced via a non-overlapping domain decomposition to a coarse scale mortar<br />
interface problem that is solved using a multiscale flux basis. Numerical experiments<br />
are presented to confirm the theory and illustrate the efficiency and flexibility <strong>of</strong> the<br />
method.<br />
This is joint work with Mary F. Wheeler and Guangri Xue, the University <strong>of</strong> Texas<br />
at Austin.<br />
40
Contributed talks<br />
41
Contributed talks<br />
Mimetic Finite Difference methods for convection-<strong>di</strong>ffusion<br />
problems<br />
Andrea Cangiani<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong> e Applicazioni, Università <strong>di</strong> Milano Bicocca,<br />
via Cozzi 53, 20125 Milano, Italy<br />
andrea.cangiani@unimib.it<br />
We present recent results on the extension <strong>of</strong> the Mimetic Finite Difference method<br />
to elliptic equations comprising convection terms. We consider mimetic <strong>di</strong>scretizations<br />
based on both the nodal and mixed formulation <strong>of</strong> the problem. These are low order<br />
methods that extend, respectively, the linear and Raviart-Thomas finite element to general<br />
polygonal meshes. The analysis relies on the existence <strong>of</strong> appropriate elemental<br />
lifting operators and, in the case <strong>of</strong> the mixed formulation, on the connection between<br />
the mimetic formulation and the lowest order Raviart-Thomas mixed finite element<br />
method. Stabilization techniques in the convection-dominated regimes will also be presented.<br />
Numerical experiments on general polygonal meshes, also inclu<strong>di</strong>ng non-convex<br />
shaped elements, confirm the theoretical results and in<strong>di</strong>cate a superconvergence effects<br />
in some particular cases.<br />
This a joint work with G. Manzini and A. Russo.<br />
43
Contributed talks<br />
A unified sub<strong>di</strong>vision <strong>di</strong>scretization approach for thick and<br />
thin shells<br />
Fehmi Cirak and Quan Long<br />
Department <strong>of</strong> Engineering, University <strong>of</strong> Cambridge,<br />
Trumpington Street, Cambridge CB2 1PZ , UK<br />
http://www-g.eng.cam.ac.uk/csml/<br />
fc286@eng.cam.ac.uk<br />
We present a unified model and an associated <strong>di</strong>scretization scheme for thin (shearrigid)<br />
and thick (shear-flexible) shells. The equilibrium equations in weak form are<br />
derived following a standard semi-inverse approach. Thereby, any configuration <strong>of</strong> the<br />
shell is assumed to be defined as<br />
ϕ = x + ζ(n + w) with − t t<br />
2 ≤ ζ ≤ 2<br />
where ϕ is the position vector <strong>of</strong> a material point within the shell volume, x is the<br />
position vector <strong>of</strong> a material point on the mid-surface, n is the unit normal to the midsurface,<br />
w is a vector for allowing out-<strong>of</strong>-plane shear deformations and t is the thickness<br />
<strong>of</strong> the shell. In the assumed kinematics, x and w are the independent variables and n is<br />
a function <strong>of</strong> x. Importantly, in the thin limit (t → 0) the Kirchh<strong>of</strong>f-Love model with<br />
|w| ≡ 0 is recovered.<br />
The conforming finite element <strong>di</strong>scretization <strong>of</strong> the shell equations with the assumed<br />
kinematics requires C 1 -continuous shape functions. In the present method, smooth<br />
sub<strong>di</strong>vision shape functions are used for interpolating the mid-surface position vector<br />
x and the shear vector w. In the thin-shell limit, the resulting finite elements do not<br />
exhibit shear locking simply because the concurrent interpolation <strong>of</strong> x = 0 and w ≡ 0<br />
with the same shape functions does not lead to any compatibility problems.<br />
44
Skin effect in electromagnetism<br />
Monique Dauge<br />
IRMAR, Université de Rennes 1, Campus de Beaulieu,<br />
35042, Rennes Cedex, France<br />
monique.dauge@univ-rennes1.fr<br />
Contributed talks<br />
We consider the Maxwell equations in a domain made <strong>of</strong> two parts, <strong>di</strong>electric and<br />
conductor. We are interested in the behavior <strong>of</strong> solutions when the conductivity in the<br />
conductor part is large.<br />
In a first stage, we assume smoothness for the <strong>di</strong>electric–conductor interface and we<br />
exhibit a multiscale asymptotic expansion with pr<strong>of</strong>ile terms rapidly decaying inside<br />
the conductor. We measure this skin effect by introducing a skin depth function that<br />
turns out to depend on the mean curvature <strong>of</strong> the boundary <strong>of</strong> the conductor. We<br />
then confirm these asymptotic results by numerical experiments in various axisymmetric<br />
configurations.<br />
In a second stage, motivated by numerical experiments in cylindrical configurations,<br />
we also investigate the case <strong>of</strong> nonsmooth interfaces.<br />
We bring extension or complements to previous works [10,6,7,5].<br />
From collaboration with: Victor Péron (Bordeaux, and soon, Pau), Gabriel Caloz<br />
(Rennes), Erwan Faou (Rennes), and Clair Poignard (Bordeaux), see [9,3,1,4].<br />
Computations done with Finite Element Library Mélina [8].<br />
References<br />
[1] G. Caloz, M. Dauge, V. Péron. Uniform estimates for transmission problems<br />
with high contrast in heat conduction and electromagnetism. To appear in Journal<br />
<strong>of</strong> Mathematical Analysis and Applications (2010). http://hal.archives-ouvertes.fr/hal-<br />
00422315/en/http://hal.archives-ouvertes.fr/hal-00422315/en/.<br />
[2] G. Caloz, M. Dauge, E. Faou, V. Péron. On the influence <strong>of</strong> the geometry on skin<br />
effect in electromagnetism. (In preparation), 2010.<br />
[3] M. Dauge, E. Faou, V. Péron. Comportement asymptotique à haute conductivité de<br />
l’épaisseur de peau en électromagnétisme. C. R. Acad. Sci. Paris Sér. I Math. (348) (2010)<br />
385–390.<br />
[4] M. Dauge, V. Péron, C. Poignard. Asymptotic expansion for the solution <strong>of</strong> a stiff<br />
transmission problem in electromagnetism with a singular interface. (In preparation), 2010.<br />
45
Contributed talks<br />
[5] H. Haddar, P. Joly, H.-M. Nguyen. Generalized impedance boundary con<strong>di</strong>tions for<br />
scattering problems from strongly absorbing obstacles: the case <strong>of</strong> Maxwell’s equations.<br />
Math. Models Methods Appl. Sci. 18(10) (2008) 1787–1827.<br />
[6] R. C. MacCamy, E. Stephan. Solution procedures for three-<strong>di</strong>mensional eddy current<br />
problems. J. Math. Anal. Appl. 101(2) (1984) 348–379.<br />
[7] R. C. MacCamy, E. Stephan. A skin effect approximation for eddy current problems.<br />
Arch. Rational Mech. Anal. 90(1) (1985) 87–98.<br />
[8] D. Martin. Mélina, bibliothèque de calculs éléments finis. Source code. http://anummaths.univ-rennes1.fr/melinahttp://anum-maths.univ-rennes1.fr/melina,<br />
1990-2010.<br />
[9] V. Péron. Modélisation mathématique de phénomènes électromagnétiques dans des<br />
matériaux à fort contraste. PhD thesis, Université Rennes 1 2009. http://tel.archivesouvertes.fr/tel-00421736/fr/http://tel.archives-ouvertes.fr/tel-00421736/fr/.<br />
[10] E. Stephan. Solution procedures for interface problems in acoustics and electromagnetics.<br />
In Theoretical acoustics and numerical techniques, volume 277 <strong>of</strong> CISM Courses and<br />
Lectures, pages 291–348. Springer, Vienna 1983.<br />
46
Contributed talks<br />
Modeling <strong>of</strong> bone conduction <strong>of</strong> sound in the human head<br />
using hp-finite elements: code design and verification.<br />
Paolo Gatto, Leszek Demkowicz<br />
Inst. for Comp. Engineering and Sciences, University <strong>of</strong> Texas at Austin,<br />
Austin, TX 78712, USA<br />
We focus on the development <strong>of</strong> a reliable numerical model for investigating the boneconducted<br />
sound in the human head. The main <strong>di</strong>fficulty <strong>of</strong> the problem arises from the<br />
lack <strong>of</strong> fundamental knowledge regar<strong>di</strong>ng the transmission <strong>of</strong> acoustic energy through<br />
non-airborne pathways to the cochlea, i.e. the most sensitive organ <strong>of</strong> the middle ear.<br />
A fully coupled model based on the acoustic/elastic interaction problem with a detailed<br />
resolution <strong>of</strong> the cochlea region and its interface with the skull and the air pathways,<br />
should provide an insight into this fundamental, long stan<strong>di</strong>ng research problem. To this<br />
aim, we have developed a 3-D finite element code that supports elements <strong>of</strong> all shapes –<br />
tetrahedra, prisms, and pyramids – in order to fully capture the complicated geometry<br />
<strong>of</strong> the problem. We have tested our code employing the case <strong>of</strong> a multilayered sphere,<br />
for which an analytical solution in terms <strong>of</strong> Hankel’s functions is known.<br />
47
Contributed talks<br />
Finite Volume Methods on NURBS Geometries with Application<br />
to Fluid Flow and Fluid-Structure Interaction<br />
Ch. Heinrich, A.-V. Vuong, B. Simeon<br />
Centre for Mathematical Sciences, Technische Universität München,<br />
Fakultät für Mathematik (M2), Lehrstuhl für Numerische Mathematik, Boltzmannstraße 3 D,<br />
85748 Garching bei München, Germany<br />
heinrich@ma.tum.de<br />
Isogeometric Analysis extends isoparametric finite elements to more general basis functions<br />
such as B-splines and NURBS. In this way, it is possible to fit exact geometries<br />
at the coarsest level <strong>of</strong> <strong>di</strong>scretization and eliminate geometry errors from the very beginning.<br />
In CFD simulations, however, the Finite Volume Method (FVM) is still the<br />
method-<strong>of</strong>-choice. This contribution <strong>di</strong>scusses approaches to incorporate the idea <strong>of</strong> an<br />
exact boundary representation and NURBS geometries in connection with the FVM. In<br />
particular, we transform the problem formulation to the parametric domain and analyze<br />
the <strong>di</strong>scretization <strong>of</strong> the arising integral terms. Furthermore, the convergence behaviour<br />
is stu<strong>di</strong>ed for numerical examples in the field <strong>of</strong> flow simulations. Finally we present a<br />
partitioned Fluid-Structure Interaction (FSI) coupling algorithm combining the solver<br />
described above with a structural solver based on Isogeometric Analysis solving the<br />
equations <strong>of</strong> linear elasticity. In this context, we also <strong>di</strong>scuss an approach to incorporate<br />
a matching interface.<br />
48
Contributed talks<br />
High order Galerkin Methods for General Advection Problems<br />
Holger Heumann<br />
Seminar for Applied Mathematics, ETH Zürich<br />
Rämistrasse 101, 8092 Zürich, Switzerland<br />
Holger.Heumann@sam.math.ethz.ch<br />
Co-or<strong>di</strong>nate free <strong>di</strong>fferential forms <strong>of</strong>fer considerable benefits for the construction <strong>of</strong><br />
compatible <strong>di</strong>scretizations. We adopt here the perspective <strong>of</strong> <strong>di</strong>fferential forms to derive<br />
Galerkin schemes for general advection problems:<br />
∗ω + ∗Lβω = f in Ω ⊂ R n ,<br />
ı ∗ ω = ı ∗ ω0 on inflow boundary,<br />
iβω = iβω0 on inflow boundary.<br />
(3.1)<br />
These are boundary value problems for unknown l-forms ω, 0 ≤ l ≤ n, in the domain<br />
Ω. The symbol ∗ stands for the Hodge operator mapping a l-form to a (n − l)-form.<br />
Boundary con<strong>di</strong>tions on the inflow boundary are imposed via the trace ı ∗ ω and the<br />
contraction iβω <strong>of</strong> ω. By Lβ we denote the Lie derivative for a given velocity field β.<br />
The Lie derivative generalizes the concept <strong>of</strong> <strong>di</strong>rectional derivatives <strong>of</strong> scalar functions to<br />
<strong>di</strong>fferential forms, hence it models advection. For n = 3 the vector analytic incarnations<br />
<strong>of</strong> the Lie derivative for unknown fields u or functions u are:<br />
• l=0: β · grad u;<br />
• l=1: curl u × β + grad(u · β);<br />
• l=2: β<strong>di</strong>vu + curl(β × u);<br />
• l=3: <strong>di</strong>v(βu).<br />
While the cases l = 0 and l = 3 have been object <strong>of</strong> intense research, the cases l = 1 and<br />
l = 2 have received scant attention, though they are relevant for numerical modeling in<br />
e.g. symmetrized MHD [2] or magnetic <strong>di</strong>ffusion convection problems [4].<br />
By means <strong>of</strong> <strong>di</strong>fferential forms calculus we derive a consistent and stable Galerkin<br />
<strong>di</strong>scretization <strong>of</strong> (1) for general, not necessarily conforming, approximation spaces. For<br />
<strong>di</strong>scontinuous approximation spaces we even obtain stability in some mesh dependent<br />
norm, which is stronger than the L 2 -norm. In the cases l = 0 and l = n these schemes are<br />
49
Contributed talks<br />
the usual Galerkin schemes for scalar advection and globally continuous or <strong>di</strong>scontinuous<br />
approximation spaces.<br />
1<br />
k+ We then prove for the case l = 1 and n = 3 optimal convergence <strong>of</strong> order O(h 2 )<br />
for either fully <strong>di</strong>scontinuous, normally continuous or tangentially continuous piecewise<br />
polynomial approximation spaces. The pro<strong>of</strong> is very much inspired from the correspon<strong>di</strong>ng<br />
pro<strong>of</strong> for l = 2, n = 2 [3] and the principle ideas can be extended to the curl- and<br />
<strong>di</strong>v-conforming approximation spaces due to the build-in finite element de Rham complex<br />
[1].<br />
References<br />
[1] D.N. Arnold, R.S. Falk and R. Winther: Finite element exterior calculus, homological techniques,<br />
and applications, Acta Numerica, vol. 15 (2006), 1-155.<br />
[2] T. Barth: On the role <strong>of</strong> involutions in the <strong>di</strong>scontinous Galerkin <strong>di</strong>scretization <strong>of</strong> Maxwell<br />
and magnetohydrodynamic systems, Compatible spatial <strong>di</strong>scretizations, IMA Vol. Math.<br />
Appl.,vol 142 (2006), 69-88<br />
[3] F. Brezzi, L.D.Marini, E. Süli: Discontinuous Galerkin methods for first-order hyperbolic<br />
problems, Math. Models Methods Appl. Sci., vol 14 (2004), 1893–1903<br />
[4] Holger Heumann, Ralf Hiptmair and Jinchao Xu: A semi–Lagrangian method for convection<br />
<strong>of</strong> <strong>di</strong>fferential forms, SAM-Report 2009-9, 2009.<br />
50
Contributed talks<br />
A non-standard finite element method based on boundary<br />
integral operators<br />
Clemens H<strong>of</strong>reither<br />
Institute <strong>of</strong> Computational Mathematics,<br />
Altenberger Strasse 69, A-4040 Linz, Austria/Europe<br />
clemens.h<strong>of</strong>reither@dk-compmath.jku.at<br />
Joint work Ulrich Langer and Clemens Pechstein.<br />
We present a non-standard finite element method based on the use <strong>of</strong> boundary integral<br />
operators that permits polyhedral element shapes as well as meshes with hanging<br />
nodes. The method employs elementwise PDE-harmonic trial functions and can thus be<br />
interpreted as a local Trefftz method. The construction principle requires the explicit<br />
knowledge <strong>of</strong> the fundamental solution <strong>of</strong> the partial <strong>di</strong>fferential operator, but only locally,<br />
i.e. in every polyhedral element. This allows us to solve PDEs with elementwise<br />
constant coefficients. In this talk we consider the <strong>di</strong>ffusion equation as a model problem,<br />
but the method can be generalized to convection-<strong>di</strong>ffusion-reaction problems and to<br />
systems <strong>of</strong> PDEs like the linear elasticity system with elementwise constant coefficients.<br />
We provide a rigorous error analysis <strong>of</strong> the method for the three-<strong>di</strong>mensional case under<br />
quite general assumptions on the geometric properties <strong>of</strong> the elements. Some numerical<br />
tests confirm the theoretical results.<br />
51
Contributed talks<br />
Higher-order <strong>di</strong>scretization <strong>of</strong> the Laplace-Beltrami operator<br />
J.J. Kreeft, A. Palha, M.I. Gerritsma<br />
Delft University <strong>of</strong> Technology<br />
Kluyverweg 1, 2629 HS Delft, The Netherlands<br />
J.J.Kreeft@tudelft.nl<br />
The Laplace-Beltrami operator acting on <strong>di</strong>fferential forms is given by ∇ := δd + dδ :<br />
Λ k ↦→ Λ k . This operator resembles among others the Laplace <strong>of</strong> scalars as well as the<br />
Laplace acting on vectors. An example <strong>of</strong> the former is in the Poisson equation and the<br />
later can be found e.g. in fluid dynamics and electro-magnetism [1].<br />
In this talk we show how this operator fits in the double DeRham-complex [2] and the<br />
two <strong>di</strong>fferent ways to <strong>di</strong>scretize/approximate the Laplace-Beltrami operator. The first is<br />
in approximating the Hodge-⋆ operator, resulting in a staggered-grid finite-volume like<br />
method, and the second is in approximating the dual DeRham-complex, resulting in a<br />
(mixed) finite-element like formulation [3].<br />
The presented formulations are combined with the mimetic spectral element method,<br />
a higher-order interpolation method for cochains, the <strong>di</strong>screte counterpart <strong>of</strong> <strong>di</strong>fferential<br />
forms. The robustness <strong>of</strong> the proposed method is demonstrated using highly curved<br />
grids while still maintaining higher-order convergence.<br />
References<br />
[1] F. Brezzi, A. Buffa, Innovative mimetic <strong>di</strong>scretizations for electromagnetic problems, Journal<br />
<strong>of</strong> Computational and Applied Mathematics, 234, 1980-1987, (2010).<br />
[2] M. Desbrun, E. Kanso, Y. Tongy, Discrete <strong>di</strong>fferential forms for computational modeling,<br />
in procee<strong>di</strong>ngs <strong>of</strong> International conference on computer graphics and interactive techniques,<br />
(2005).<br />
[3] D.N. Arnold, R.S. Falk, R. Winther, Finite element exterior calculus: from Hodge theory<br />
to numerical stability, Bulletin <strong>of</strong> the American mathematical society, 47, 281-354, (2010).<br />
52
Contributed talks<br />
Localized solutions and filtering mechanisms for the <strong>di</strong>scontinuous<br />
Galerkin semi-<strong>di</strong>scretizations <strong>of</strong> the 1−d wave equation<br />
Aurora-Mihaela Marica (joint work with Enrique Zuazua)<br />
BCAM - Basque Center for Applied Mathematics,<br />
Bizkaia Technology Park, 500, E-48160, Derio, Basque Country, Spain<br />
marica@bcamath.org<br />
We perform a complete Fourier analysis <strong>of</strong> the semi-<strong>di</strong>screte 1 − d wave equation<br />
obtained through P1 <strong>di</strong>scontinuous Galerkin (DG) approximations <strong>of</strong> the continuous<br />
wave equation on an uniform grid. The resulting system exhibits the interaction <strong>of</strong><br />
two types <strong>of</strong> components: a physical one and a spurious one, related to the possible<br />
<strong>di</strong>scontinuities that the numerical solution allows. Each <strong>di</strong>spersion relation contains<br />
critical points where the correspon<strong>di</strong>ng group velocity vanishes. Following previous<br />
constructions in [5], we rigorously build wave packets with arbitrarily small velocity <strong>of</strong><br />
propagation concentrated either on the physical or on the spurious component. We also<br />
develop filtering mechanisms aimed at recovering the uniform velocity <strong>of</strong> propagation <strong>of</strong><br />
the continuous solutions. Finally, some applications to numerical approximation issues <strong>of</strong><br />
control problems and connections to the <strong>di</strong>screte estimates for the Schrö<strong>di</strong>nger equation<br />
are presented.<br />
Key words: wave equation, <strong>di</strong>scontinuous Galerkin methods, Fourier analysis, <strong>di</strong>spersion<br />
relation, group velocity, filtering, bigrid algorithm.<br />
References<br />
[1] D.N.Arnold, F.Brezzi, B.Cockburn, L.D.Marini, Unified analysis <strong>of</strong> Discontinuous Galerkin<br />
Methods for Elliptic Problems, SIAM J. Numer. Anal., 39 (2002), 1749-1779.<br />
[2] R.Glowinski, J.-L.Lions, J.He, Exact and approximate controllability for <strong>di</strong>stributed parameter<br />
systems: a numerical approach, Encyclope<strong>di</strong>a <strong>of</strong> Mathematics and its Applications,<br />
Cambridge University Press, 117 (2008).<br />
[3] L.Ignat, E.Zuazua, Convergence <strong>of</strong> a multi-grid method for the control <strong>of</strong> waves, J. Eur.<br />
Math. Soc., 11 (2009), 351391.<br />
[4] L.Ignat, E.Zuazua, Numerical <strong>di</strong>spersive schemes for the nonlinear Schrö<strong>di</strong>nger equation,<br />
SIAM. J. Numer. Anal., 47(2) (2009), 1366-1390.<br />
[5] A.Marica, E.Zuazua, Localized solutions for the finite <strong>di</strong>fference semi-<strong>di</strong>scretization <strong>of</strong> the<br />
wave equation, C.R. Acad. Sci. Paris, to appear.<br />
53
Contributed talks<br />
[6] A.Marica, E.Zuazua, Localized solutions and filtering mechanisms for the <strong>di</strong>scontinuous<br />
Galerkin semi-<strong>di</strong>scretizations <strong>of</strong> the 1 − d wave equation, C.R. Acad. Sci. Paris, submitted.<br />
[7] E.Zuazua, Exponential decay for the semilinear wave equation with localized damping in<br />
unbounded domains, J. Math. Pures Appl., 70 (1991), 513-529.<br />
[8] E.Zuazua, Propagation, Observation, Control and Numerical Approximations <strong>of</strong> Waves,<br />
SIAM Review, 47(2)(2005), 197-243.<br />
54
Isogeometric Shape Optimization<br />
Peter Nørt<strong>of</strong>t Nielsen ∗† , Manh Dang Nguyen ∗ , Anton Evgrafov ∗ ,<br />
Allan Roulund Gersborg † and Jens Gravesen ∗<br />
∗ DTU Mathematics, Technical University <strong>of</strong> Denmark,<br />
Bygning 303S, DK-2800 Kgs. Lyngby, Denmark<br />
† DTU Mechanical Engineering, Technical University <strong>of</strong> Denmark<br />
Nils Koppels Allé, Buil<strong>di</strong>ng 404, DK-2800 Kgs. Lyngby, Denmark<br />
p.n.nielsen@mat.dtu.dk<br />
Contributed talks<br />
Isogeometric analysis unites the power to analyse complex engineering problems from<br />
finite element analysis (FEA) with the ability to represent complicated shapes from<br />
computer aided design (CAD)[1, 2]. Being a mixture <strong>of</strong> CAD and FEA, isogeometric<br />
analysis serves as an ideal basis for shape optimization [1, 3, 4].<br />
In the first part <strong>of</strong> the talk we apply the method <strong>of</strong> isogeometric shape optimization<br />
to 2-<strong>di</strong>mensional, steady-state Stokes flow problems with Dirichlet boundary con<strong>di</strong>tions.<br />
We present implementation details and results for a simple optimization problem in<br />
which the objective is to find the most energy efficient shape <strong>of</strong> a 2-<strong>di</strong>mensional pipe<br />
bend.<br />
In the second part, we present isogeometric shape optimization towards the problem<br />
<strong>of</strong> designing vibrating membranes where their first few eigenfrequencies are prescribed.<br />
Some nice shapes have been obtained by <strong>di</strong>fferent methods. They thus show the robustness<br />
<strong>of</strong> our approach and support the integration <strong>of</strong> isogeometric analysis into shape<br />
optimization.<br />
References<br />
[1] T.J.R. Hughes, J.A. Cottrell, Y. Bazilevs, Isogeometric analysis: CAD, finite elements,<br />
NURBS, exact geometry and mesh refinement, Comput. Methods Appl. Mech. Engrg., 194,<br />
4135–4195, 2005.<br />
[2] J.A. Cottrell, T.J.R. Hughes, Y. Bazilevs, Isogeometric Analysis. Toward Integration <strong>of</strong><br />
CAD and FEA, John Wiley and Sons, 2009.<br />
[3] W.A. Wall, M.A. Frenzel, C. Cyron, Isogeometric structural shape optimization, Comput.<br />
Methods Appl. Mech. Engrg., 197, 2976–2988, 2008.<br />
[4] S. Cho, S.-H. Ha, Isogeometric shape design optimization: exact ge- ometry and enhanced<br />
sensitivity, Struct. Multi<strong>di</strong>sc. Optim, 38, 53–70, 2009.<br />
55
Contributed talks<br />
Higher order cochain interpolation<br />
A. Palha ∗ , J.J. Kreeft † , M.I. Gerritsma †<br />
∗ Delft University <strong>of</strong> Technology<br />
Kluyverweg 1, 2629 HS Delft, The Netherlands<br />
{a.palha,j.j.kreeft,m.i.gerritsma}@tudelft.nl<br />
Algebraic topology can be considered as the natural <strong>di</strong>screte analogue <strong>of</strong> <strong>di</strong>fferential<br />
geometry. This analogy is used in so-called mimetic <strong>di</strong>scretization schemes. The kforms<br />
from <strong>di</strong>fferential geometry have a natural <strong>di</strong>screte counterpart in the k-cochains<br />
in algebraic toplogy. The exterior derivative, d, from <strong>di</strong>fferential geometry with all its<br />
properties can be represented <strong>di</strong>scretely by the coboundary operator, δ, with similar<br />
properties. The DeRham map, R, maps k-forms onto k-cochains, whereas the Whitney<br />
map, I, maps k-cochains onto k-forms. The DeRham map consists <strong>of</strong> integration <strong>of</strong> kforms<br />
over oriented k-<strong>di</strong>mensional geometric objects, whereas the Whitney map consists<br />
<strong>of</strong> interpolation, [1]. The maps should be constructed such that<br />
R · I ≡ I and I· = O(hp), (3.2)<br />
where I is the identity map, h is the characteristic mesh size and p the order <strong>of</strong> the<br />
Whitney map. Furthermore, we require that<br />
R · d = δ · R and d · I = I · δ. (3.3)<br />
Given a cell complex, K, the DeRham map is trivial. This talk will present a Whitney<br />
map <strong>of</strong> arbitrary order satisfying the above requirements. In ad<strong>di</strong>tion, it will be shown<br />
that if these interpolation functions are used as basis functions for finite element formulations<br />
for conservation laws, well-known finite volume <strong>di</strong>scretizations are obtained.<br />
Since all operators are purely topological, all these relations remain valid in severely<br />
<strong>di</strong>storted domains, [2].<br />
References<br />
[1] P.B. Bochev and J.M. Hyman, Principles <strong>of</strong> mimetic <strong>di</strong>scretizations <strong>of</strong> <strong>di</strong>fferential equations,<br />
IMA, 142, Eds. D. Arnold, P. Bochev, R. Nicolaides and M. Shashkov, (2006).<br />
[2] M. Gerritsma, Edge functions for spectral element methods, in procee<strong>di</strong>ngs <strong>of</strong> ICOSAHOM<br />
2009, (2010).<br />
56
Quadrature for Finite Elements on Pyramids<br />
Joel Phillips<br />
Mathematics, Rea<strong>di</strong>ng University,<br />
Whiteknights, PO Box 220, Rea<strong>di</strong>ng RG6 6AX, UK<br />
joel.phillips@rea<strong>di</strong>ng.ac.uk<br />
Contributed talks<br />
High order finite elements on pyramidal domains that conform to and commute with<br />
the spaces <strong>of</strong> the de Rham sequence have recently been constructed. A practical implementation<br />
<strong>of</strong> a finite element method requires a quadrature rule, <strong>of</strong> which the basic<br />
requirement is that it not decrease the order <strong>of</strong> convergence <strong>of</strong> the method.<br />
The analysis <strong>of</strong> the effect <strong>of</strong> quadrature on polynomial elements is classical. One<br />
rule <strong>of</strong> thumb is that for the approximation <strong>of</strong> an elliptic problem using finite elements<br />
consisting <strong>of</strong> polynomials <strong>of</strong> degree k, it is sufficient to use a quadrature rule that exactly<br />
integrates all polynomials <strong>of</strong> degree 2k − 2. However, useful conforming pyramidal finite<br />
elements cannot be constructed with purely polynomial approximation spaces - it is<br />
essential to include certain rational functions.<br />
Quadrature rules for pyramids based on conical product formulae that exactly integrate<br />
polynomials <strong>of</strong> a given degree have been known for over half a century. Promisingly,<br />
these rules also exactly integrate the rational functions included in each <strong>of</strong> the pyramidal<br />
finite element families. However, the classical theory requires that the approximation<br />
space is a subspace <strong>of</strong> H k (K) for each element K. This con<strong>di</strong>tion is not satisfied by the<br />
rational functions present in the pyramidal elements and so the classical rule <strong>of</strong> thumb<br />
does not imme<strong>di</strong>ately generalise.<br />
In this talk, we shall show how to overcome this problem via a refinement <strong>of</strong> the<br />
Bramble-Hilbert lemma.<br />
57
Contributed talks<br />
Towards a Mimetic Nonlinear Shearable Shell<br />
Joachim Linn ∗ , Alessio Quaglino † , Max Wardetzky † and Clarisse Weischedel †<br />
∗ Fraunh<strong>of</strong>er Institute for Industrial Mathematics, Kaiserslautern, Germany,<br />
† Department <strong>of</strong> Numerical and Applied Mathematics, University <strong>of</strong> Göttingen, Germany<br />
quaglino@math.uni-goettingen.de<br />
Linear shearable plates have received wide attention due to the effort <strong>of</strong> deriving compatible<br />
finite element <strong>di</strong>scretizations for the Reissner-Mindlin model [1], a topic which<br />
by now enjoys a mature literature. The inherent kinematic assumptions, allowing for<br />
transverse shear strain, have been successfully employed in the geometrically nonlinear<br />
setting, lea<strong>di</strong>ng to the well-known Nagh<strong>di</strong> shell model.<br />
The fact that the FE view <strong>of</strong>ttimes occludes the geometry behind a given physical<br />
model has stirred recent interest in <strong>di</strong>screte geometric models, e.g., grounded in exterior<br />
calculus [2]. By incorporating a geometric viewpoint for the case <strong>of</strong> shells, we propose a<br />
mimetic FD <strong>di</strong>scretization, with the aim to provide a unified language for geometrically<br />
linear and nonlinear deformations <strong>of</strong> shells and plates. We argue that this goal can be<br />
achieved by <strong>di</strong>scretizing the geometric quantities underlying the smooth theory in an<br />
invariant, parameterization-free manner.<br />
We present in detail one particular triangular element that is derived from these<br />
considerations. We show that this <strong>di</strong>scretization <strong>of</strong>fers an independent derivation <strong>of</strong> a<br />
shell element proposed by Flores et al. [3] and generalizes a previously considered <strong>di</strong>screte<br />
thin-shell model by Grinspun et al. [4]. We further argue that any low-order finite<br />
element for the Reissner-Mindlin model can be translated into our geometric language.<br />
Ad<strong>di</strong>tionally, we provide a geometric intuition for shear locking.<br />
Overview<br />
In contrast to the the Nagh<strong>di</strong> description <strong>of</strong> shells, the geometric view describes curvatures<br />
by the change <strong>of</strong> normals instead <strong>of</strong> derivatives <strong>of</strong> positions. Within this framework,<br />
moderately thick shells may be treated by the Cosserat model [5] that allows for midsurface<br />
“normals”—referred to as <strong>di</strong>rectors—to undergo transverse shear, i.e., to deviate<br />
from true midsurface normals. By attaching <strong>di</strong>rectors to surfaces, the Cosserat model<br />
imme<strong>di</strong>ately lends itself to a <strong>di</strong>fferential-geometric treatment, based on the theory <strong>of</strong><br />
Cartan’s moving frames. If the <strong>di</strong>rectors coincide with surface normals, it recovers the<br />
Koiter model <strong>of</strong> thin shells.<br />
58
Contributed talks<br />
We represent the first and second fundamental forms <strong>of</strong> the deformed surface, enco<strong>di</strong>ng<br />
intrinsic metric and extrinsic curvature properties, respectively, by<br />
I = dφ T dφ, II = 1<br />
<br />
dd<br />
2<br />
T dφ + dφ T <br />
dd , (3.4)<br />
where φ : ¯ S → R3 denotes the (orientation preserving) deformation <strong>of</strong> the midsurface,<br />
d : ¯ S → R3 denotes the <strong>di</strong>rector field, and d denotes the (metric-free) Cartan outer<br />
derivative. To get a low-order model, we choose φ to be a P1 -mapping between the<br />
undeformed and deformed configuration, so that dφ (and therefore the first fundamental<br />
form) is rea<strong>di</strong>ly obtained in the <strong>di</strong>screte case. A Cosserat shell’s elastic potential energy<br />
is the sum over membrane, ben<strong>di</strong>ng, and shearing contributions:<br />
W = t<br />
2<br />
<br />
¯S<br />
I − Ī2 Km + t2 II − ĪI2 Kb + ksd tan 2 . (3.5)<br />
Here Ī, ĪI denote the fundamental forms <strong>of</strong> the undeformed surface, respectively, and<br />
dtan is the <strong>di</strong>rector’s tangential component with respect to the deformed surface. Further,<br />
· K denotes a norm enco<strong>di</strong>ng material stiffnesses, respectively for membrane and<br />
ben<strong>di</strong>ng, and ks represents the shearing stiffness, which can be treated as a scalar in the<br />
isotropic case. While our invariant energy formulation is equivalent to a parameterizationdependent<br />
description considered by, e.g., Simo and coworkers [6], we are not aware <strong>of</strong> a<br />
similarly compact representation as ours—a formulation that almost imme<strong>di</strong>ately carries<br />
over to the <strong>di</strong>screte setting.<br />
For a <strong>di</strong>screte treatment <strong>of</strong> shear, we assign surface normals and <strong>di</strong>rectors to edge<br />
degrees <strong>of</strong> freedom <strong>of</strong> a triangulated surface mesh. While it might appear that <strong>di</strong>screte<br />
surface normals are naturally defined as angle-bisecting ones, we present numerical evidence<br />
that this choice leads to shear locking when the <strong>di</strong>screte mesh interpolates a given<br />
smooth surface. Instead, we allow the unit midedge normal to adjusts itself accor<strong>di</strong>ng<br />
to minimizing the elastic energy, and show that this choice is equivalent to reduced<br />
integration in the FE language.<br />
For obtaining a <strong>di</strong>screte second fundamental form II, we observe that dd is a (multivalued)<br />
1-form in the language <strong>of</strong> exterior calculus. In the <strong>di</strong>screte view, dd is an<br />
integrated quantity, which can be applied to 1-<strong>di</strong>mensional cells. We choose midedge<br />
connecting lines as such requisite 1-cells, giving three <strong>di</strong>rections per triangle for which II<br />
can be evaluated. Fortunately, in <strong>di</strong>mension two, any quadratic form (and hence II) is<br />
uniquely determined by its action on three <strong>di</strong>fferent vectors, yiel<strong>di</strong>ng a constant ben<strong>di</strong>ng<br />
strain per triangle, which can be expressed in a similarly compact and invariant manner<br />
as the strain <strong>of</strong> a constant strain triangle.<br />
References<br />
[1] F. Brezzi, M. Fortin, Mixed and Hybrid Finite Element Methods, Springer-Verlag, 1991.<br />
[2] D. Arnold, R. Falk, R. Winther, Finite Elements Exterior Calculus, Acta Numerica, 2006.<br />
[3] F. G. Flores, E. Oñate, F. Larate, New assumed strain triangles for non linear shell analysis,<br />
J. Comp. Mech., pp. 107-114, Vol. 17, 1995.<br />
59
Contributed talks<br />
[4] E. Grinspun, Y. Gingold, J. Reisman, D. Zorin, Computing <strong>di</strong>screte shape operators on<br />
general meshes, Eurographics (Com- puter Graphics Forum), Vol. 25, 2006.<br />
[5] M. Bisch<strong>of</strong>f, W. A. Wall, K.-U. Bletzinger, E. Ramm, Models and Finite Elements for Thinwalled<br />
Structures, Encyclope<strong>di</strong>a <strong>of</strong> Computational Mechanics Vol. 2, Wiley, 2004.<br />
[6] J. C. Simo, D. D. Fox, On Stress Resultant Geometrically Exact Shell Model. Part I: Formulation<br />
and Optimal Parametrization, J. Comp. Meth. Appl. Mech. Eng., pp. 267-304, Vol.<br />
72, 1989.<br />
60
Contributed talks<br />
High-Order Spline Finite Element Solver for the Time Domain<br />
Maxwell equations<br />
Ahmed Ratnani ∗ , Eric Sonnendrucker † and Nicolas Crouseilles ∗<br />
∗ INRIA and University <strong>of</strong> Strasbourg, France,<br />
† University <strong>of</strong> Strasbourg, France<br />
ahmed.ratnani@math.unistra.fr<br />
Isogeometric analysis has been developed recently in order to be able to use the CAD<br />
description <strong>of</strong> the computational domain to define Finite Element basis functions. In the<br />
same spirit, we developed an exact sequence <strong>of</strong> Finite Element spaces based on splines<br />
having the same properties as the Whitney Finite Element spaces tra<strong>di</strong>tionnally used<br />
for the Finite Element solution <strong>of</strong> Maxwell’s equations. As with the Whitney elements,<br />
one <strong>of</strong> Ampere’s or Faraday’s law can be <strong>di</strong>scretized with a relation between the spline<br />
coefficients <strong>of</strong> the eletric and magnetic fields independent <strong>of</strong> the topology <strong>of</strong> the mesh.<br />
The metric comes in through a <strong>di</strong>screte Finite Element Hodge operator which appears as<br />
the mass matrix involved in the other equation. All domains represented by NURBS can<br />
be represented exactly and arbitrary high degree splines can be used. We shall describe<br />
a Time Domain Maxwell solver based on these Finite Elements and a few validation test<br />
cases.<br />
References<br />
[1] Ratnani, Sonnendrucker, Crouseilles. Arbitrary High-Order Spline Finite Element Solver for<br />
the Time Domain Maxwell equations (In preparation)<br />
61
Contributed talks<br />
A Fictitious Domain Method for Dynamic Analyses<br />
M. Ruess and E. Rank<br />
Center for Simulation Technology in Engineering - CeSIM, Technische Universität München,<br />
Arcisstrasse 21, 80333 München, Germany<br />
ruess@tum.de<br />
The analysis <strong>of</strong> the dynamic response <strong>of</strong> complex structures from solid mechanics <strong>of</strong>ten<br />
requires an essential abstraction and simplification <strong>of</strong> the true structural properties due<br />
to a tremendous <strong>di</strong>scretization effort and <strong>of</strong>ten is limited with tra<strong>di</strong>tional methods.<br />
With the Finite Cell Method (FCM) [1, 2], a novel high-order fictitious domain method<br />
was proposed that fully embeds the structure in an extended, coarse orthogonal cell<br />
structure and identifies its true geometry on the numerical integration level. The method<br />
has been intensively stu<strong>di</strong>ed for linear analyses and recently showed very promising<br />
results also for nonlinear analyses [3]. Dynamic analyses, however, haven’t yet gained<br />
much attention and will be thoroughly examined in this contribution.<br />
The Finite Cell Method dramatically simplifies the modelling effort for complex geometries<br />
and/or multi-material interfaces and essentially keeps the advantageous properties<br />
<strong>of</strong> the p-version <strong>of</strong> the Finite Element Method, thus combining high accuracy, exponential<br />
convergence rates and pre<strong>di</strong>ctable error properties. Even more, the method turns out<br />
to be highly suited for non-homogeneous material models where the constitutive properties<br />
change continuously within the computation domain. The quality <strong>of</strong> the solution<br />
for the proposed method depends on several factors, such as the numerical integration<br />
scheme and a balanced h/p refinement strategy that will be <strong>di</strong>scussed and analysed with<br />
regard to dynamic structure response.<br />
The numerical integration scheme requires a dense <strong>di</strong>stribution <strong>of</strong> integration points<br />
to accurately capture the structural inhomogeneity and to confine the integration error.<br />
An adaptive integration scheme is used to counteract a tremendous integration effort,<br />
that still is an essential criteria for the efficiency <strong>of</strong> the method if several large and dense<br />
element matrices (e.g. stiffness, mass) are required, as they appear for higher polynomial<br />
Ansatz functions. In the case <strong>of</strong> isotropic material properties the orthogonality <strong>of</strong> the<br />
cell structure can be favorably exploited since the mapping <strong>of</strong> a global cell element to a<br />
normalized hexahedral reference element is considerably simplified and allows a subcellbased<br />
pre-computation <strong>of</strong> an element-independent stiffness and mass matrix except for<br />
scaling factors representing material constants <strong>of</strong> each subcell. An optimal configuration<br />
<strong>of</strong> polynomial degree and mesh density will be derived from several benchmark tests and<br />
62
Contributed talks<br />
Figure 3.1: Convergence plot for the first three circular frequencies <strong>of</strong> a cantilever beam<br />
with circular cross-section and a moderate number <strong>of</strong> cells and p-elements,<br />
respectively. A consistent mass <strong>di</strong>stribution is applied. The numerical results<br />
are compared to the analytical Bernoulli solution as an approximation to the<br />
exact 3D solution as in<strong>di</strong>cated by the curves levelling <strong>of</strong>f at p > 3.<br />
systematically analysed concerning accuracy, numerical stability and efficiency.<br />
This contribution will give a brief overview about the basic idea <strong>of</strong> the Finite Cell<br />
Method for dynamic problems inclu<strong>di</strong>ng mass lumping strategies and the proper choice<br />
<strong>of</strong> boundary con<strong>di</strong>tions. A frequency analysis on cell and domain level is presented and<br />
compared to results from classical h- and p-FEM computations to reveal the potentials<br />
and limitations <strong>of</strong> the method for dynamic analyses. The proposed method is applied<br />
to multi-material continuum problems.<br />
References<br />
[1] A. Düster, J. Parvizian, Z. Yang, E. Rank. “The finite cell method for three- <strong>di</strong>mensional<br />
problems <strong>of</strong> solid mechanics”, Comput. Methods Appl. Mech. Engrg, v. 197, p. 3768-3782,<br />
2008<br />
[2] J. Parvizian, A. Düster, E. Rank. “Finite cell method – h- and p-extension for embedded<br />
domain problems in solid mechanics”, Comput. Mech., v. 41, p. 121-133, 2007<br />
[3] D. Schillinger, S. Kollmannsberger, R.-P. Mundani, E. Rank. “The Hierarchical B- Spline<br />
Version <strong>of</strong> the Finite Cell Method for Geometrically Nonlinear Problems <strong>of</strong> Solid Mechanics”,<br />
in Proc. IV ECCM, Paris, France, May 16-21, 2010<br />
63
Contributed talks<br />
Analysis <strong>of</strong> the Discrete Maximum Principle in the Mimetic<br />
Finite Difference Method for Elliptic Problems<br />
D. Svyatskiy<br />
Applied Mathematics and Plasma Physics Group, Theoretical Division,<br />
Los Alamos National Laboratory, Los Alamos, NM 87545, USA<br />
dasvyat@lanl.gov<br />
The Maximum Principle is one <strong>of</strong> the most important properties <strong>of</strong> solutions <strong>of</strong> partial<br />
<strong>di</strong>fferential equations. Its numerical analog, the Discrete Maximum Principle (DMP), is<br />
one <strong>of</strong> the properties that are very <strong>di</strong>fficult to incorporate into numerical methods, especially<br />
when the computational mesh contains <strong>di</strong>storted or degenerate cells or the problem<br />
coefficients are highly heterogeneous and anisotropic. A violation <strong>of</strong> DMP may lead to<br />
numerical instabilities such as oscillations and to non-physical solutions, for example,<br />
when heat flows from a cold material to a hot one. Some physical quantities, like concentration<br />
and temperature, are non-negative by their nature and their approximations<br />
should be non-negative as well.<br />
The family <strong>of</strong> the Mimetic Finite Difference (MFD) methods provides flexibility in the<br />
choice <strong>of</strong> parameters which define a particular member <strong>of</strong> the family. For example, the<br />
MFD <strong>di</strong>scretization scheme for quadrilateral meshes depends on three parameters. The<br />
correct choice <strong>of</strong> these parameters may guarantee that the resulting numerical scheme<br />
satisfy the DMP principle. The analysis <strong>of</strong> this strategy is based on the properties <strong>of</strong> Mmatrices.<br />
The monotonicity limits <strong>of</strong> MFD method are investigated in several practically<br />
important cases inclu<strong>di</strong>ng meshes generated using the Adaptive Mesh Refinement (AMR)<br />
strategy.<br />
64
List <strong>of</strong> participants<br />
65
Jade Gesare Abuga jadeabuga@yahoo.com<br />
Mathematics, University <strong>of</strong> Eastern Africa, Baraton<br />
P.O.Box 2500, 30100, Eldoret, Kenya<br />
Paola Antonietti paola.antonietti@polimi.it<br />
MOX-<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Politecnico <strong>di</strong> Milano<br />
Via Bonar<strong>di</strong> 9, 20133 Milano<br />
Blanca Ayuso de Dios bayuso@crm.cat<br />
Centre de Recerca <strong>Matematica</strong> (CRM), UAB Science Faculty<br />
08193 Bellaterra Barcelona, Spain<br />
Douglas Arnold arnold@ima.umn.edu<br />
List <strong>of</strong> participants<br />
Institut for Mathematics and its Applications, University <strong>of</strong> Minnesota<br />
Minneapolis, MN 55455, USA<br />
Fer<strong>di</strong>nando Auricchio auricchio@unipv.it<br />
<strong>Dipartimento</strong> <strong>di</strong> Meccanica Strutturale, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia<br />
Lourenco Beirão da Veiga lourenco.beirao@unimi.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong> “F. Enriques”, Università <strong>di</strong> Milano<br />
Via Sal<strong>di</strong>ni 50, 20133 Milano, Italy<br />
Timo Betcke t.betcke@rea<strong>di</strong>ng.ac.uk<br />
Department <strong>of</strong> Mathematics, University <strong>of</strong> Rea<strong>di</strong>ng<br />
Whiteknights, PO Box 220, Rea<strong>di</strong>ng, RG6 6AX, UK<br />
Michel Bercovier berco@cs.huji.ac.il<br />
School <strong>of</strong> Computer Science and Engineering, Hebrew University <strong>of</strong> Jerusalem<br />
E.J. Safra Campus, Givat-Ram Jerusalem 91904, Israel<br />
Pavel Bochev pbboche@san<strong>di</strong>a.gov<br />
Applied Mathematics and Applications, San<strong>di</strong>a National Laboratories<br />
P.O. Box 5800, MS 1320, Albuquerque, NM 87185-1320<br />
Daniele B<strong>of</strong>fi daniele.b<strong>of</strong>fi@unipv.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia<br />
Luca Bonaventura luca.bonaventura@polimi.it<br />
MOX - <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Politecnico <strong>di</strong> Milano<br />
Via Bonar<strong>di</strong> 9, 20133 Milano<br />
Francesca Bonizzoni francesca1.bonizzoni@mail.polimi.it<br />
MOX, <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong> “F. Brioschi”, Politecnico <strong>di</strong> Milano<br />
Via Bonar<strong>di</strong> 9, e<strong>di</strong>ficio 14 “La Nave”, VI piano, 20133 Milano<br />
67
List <strong>of</strong> participants<br />
Andrea Bressan andrea.bressan@unipv.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia<br />
Franco Brezzi brezzi@imati.cnr.it<br />
Istituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Informatiche, C.N.R.<br />
Via Ferrata 1, 27100 Pavia<br />
Annalisa Buffa annalisa@imati.cnr.it<br />
Istituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Informatiche, C.N.R.<br />
Via Ferrata 1, 27100 Pavia<br />
Andrea Cangiani andrea.cangiani@unimib.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong> e Applicazioni<br />
Via Roberto Cozzi 53, E<strong>di</strong>ficio U5, 20125 Milano<br />
Fausto Cavalli fausto.cavalli@ing.unibs.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Brescia<br />
Via Valotti 9, Brescia Italia<br />
Zhiming Chen zmchen@lsec.cc.ac.cn<br />
LSEC, Institute <strong>of</strong> Computational Mathematics, Chinese Academy <strong>of</strong> Sci-<br />
ences<br />
Beijing 100190, China<br />
Clau<strong>di</strong>a Chinosi clau<strong>di</strong>a.chinosi@mfn.unipmn.it<br />
<strong>Dipartimento</strong> <strong>di</strong> Scienze e Tecnologie Avanzate, Università del Piemonte<br />
Orientale<br />
viale T. Michel 11, Alessandria<br />
Durkbin Cho durkbin@imati.cnr.it<br />
Istituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Informatiche, C.N.R.<br />
Via Ferrata 1, 27100 Pavia, Italy<br />
Lagat Robert Cheruiyot lagat.robert@yahoo.co.uk<br />
Mathematics and physics, University <strong>of</strong> Eastern Africa<br />
Baraton, P.O Box 2500 Eldoret<br />
Snorre Christiansen snorrec@math.uio.no<br />
CMA and Dept. Math., University <strong>of</strong> Oslo<br />
PO Box 1053 Blindern, NO-0316 Oslo, Norway<br />
Fehmi Cirak fc286@eng.cam.ac.uk<br />
Department <strong>of</strong> Engineering, University <strong>of</strong> Cambridge<br />
Trumpington Street, Cambridge CB2 1PZ, U.K.<br />
68
Elaine Cohen cohen@cs.utah.edu<br />
School <strong>of</strong> Computing, University <strong>of</strong> Utah<br />
List <strong>of</strong> participants<br />
50 S. Central Campus Drive, 3190 MEB, Salt Lake City, UT, 84112, USA<br />
Martin Costabel costabel@univ-rennes1.fr<br />
IRMAR, Institut Mathématique, Université de Rennes 1<br />
Campus de Beaulieu, 35042 Rennes Cedex, France<br />
Catterina Dagnino catterina.dagnino@unito.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Torino<br />
Via Carlo Alberto 10, 10124 Torino, Italia<br />
Monique Dauge monique.dauge@univ-rennes1.fr<br />
IRMAR, Université de Rennes 1<br />
Campus de Beaulieu, 35042 Renne Cedex, France<br />
Carlo de Falco carlo.defalco@polimi.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Politecnico <strong>di</strong> Milano<br />
P.zza Leonardo da Vinci 32<br />
Leszek Demkowicz leszek@ices.utexas.edu<br />
ICES, The University <strong>of</strong> Texas at Austin<br />
ACE 6.326, 201 E.24th Street, Austin, TX 78712, USA<br />
Alan Demlow demlow@ms.uky.edu<br />
Department <strong>of</strong> Mathematics, University <strong>of</strong> Kentucky<br />
715 Patterson Office Tower, Lexington, Kentucky 40506, USA<br />
Tor Dokken tor.dokken@sintef.no<br />
Department <strong>of</strong> Applied Mathematics, SINTEF ICT<br />
Forskningsveien 1, Oslo<br />
Régis Duvigneau Regis.Duvigneau@sophia.inria.fr<br />
OPALE Project-Team, INRIA Sophia-Antipolis<br />
2004 route des Lucioles, 06902 Sophia-Antipolis, France<br />
John Evans evans@ices.utexas.edu<br />
Institute for Computational Engineering and Sciences, University <strong>of</strong> Texas<br />
at Austin<br />
3517 North Hills Drive, #T301, Austin, TX, USA 78731<br />
Richard Falk falk@math.rutgers.edu<br />
Marco Favino favinom@usi.ch<br />
Department <strong>of</strong> Mathematics, Rutgers University<br />
110 Frelinghuysen Road, Piscataway, NJ 08854<br />
Institute <strong>of</strong> Computational Science, University <strong>of</strong> Lugano<br />
Via Giuseppe Buffi 13, Lugano, 6900<br />
69
List <strong>of</strong> participants<br />
Algiane Froehly algiane.froehly@inria.fr<br />
INRIA<br />
351,cours de la liberation,Bat A29 bis,33405 Talence Cedex, France<br />
Gwenael Gabard gabard@soton.ac.uk<br />
ISVR, University <strong>of</strong> Southampton<br />
University Road, Southampton, SO17 1BJ, UK<br />
Francesca Gar<strong>di</strong>ni francesca.gar<strong>di</strong>ni@unipv.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia, Italy<br />
Lucia Gastal<strong>di</strong> lucia.gastal<strong>di</strong>@ing.unibs.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Brescia<br />
Via Valotti 9, 25133 Brescia, Italy<br />
Paolo Gatto gatto@ices.utexas.edu<br />
ICES, University <strong>of</strong> Texas at Austin<br />
ACE 4.102, 201 E 24th Street, Austin, TX 78712, U.S.A.<br />
Loredana Gau<strong>di</strong>o loredana.gau<strong>di</strong>o@polimi.it<br />
MOX - <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong> “F. Brioschi”, Politecnico <strong>di</strong> Milano<br />
Via Bonar<strong>di</strong> 9, 20133 Milano<br />
Ivan Georgiev john@parallel.bas.bg<br />
Institute <strong>of</strong> Mathematics and Informatics, Bulgarian Academy <strong>of</strong> Sciences<br />
Acad. G. Bonchev St. 8, 1113 S<strong>of</strong>ia, Bulgaria<br />
Irina Georgieva irina@math.bas.bg<br />
Mathematical Modeling, Institute <strong>of</strong> Mathematics and Informatics, Bulgar-<br />
ian Academy <strong>of</strong> Sciences<br />
Bulgaria, S<strong>of</strong>ia 1113, ul. Acad. Georgi Bonchev 8<br />
Marc Gerritsma M.I.Gerritsma@TUDelft.nl<br />
Faculty <strong>of</strong> Aerspace Engineering, TU Delft<br />
Kluyverweg 1, 2629 HS Delft, The Netherlands<br />
Antti Hannukainen antti.hannukainen@tkk.fi<br />
Department <strong>of</strong> Mathematics and Systems Analysis, Aalto University<br />
P.O. Box 11100, FI-00076 Aalto, Finland<br />
Christoph Heinrich heinrich@ma.tum.de<br />
Centre for Mathematical Sciences, Technische Universität München<br />
Fakultät für Mathematik (M2), Lehrstuhl für Numerische Mathematik, Boltz-<br />
mannstraße 3 D, 85748 Garching bei München, Germany<br />
70
Holger Heumann hheumann@math.ethz.ch<br />
Seminar for Applied Mathematics, ETH Zürich<br />
Rämistrasse 101, 8092 Zürich, Switzerland<br />
Ralf Hiptmair hiptmair@sam.math.ethz.ch<br />
SAM, D-MATH, ETH Zürich<br />
Rämistrasse 101, CH 8092 Zürich<br />
Anil Hirani hirani@cs.illinois.edu<br />
List <strong>of</strong> participants<br />
Department <strong>of</strong> Computer Science, University <strong>of</strong> Illinois at Urbana-Champaign<br />
201 N. Goodwin Ave., Urbana, IL 61801, USA<br />
Clemens H<strong>of</strong>reither clemens.h<strong>of</strong>reither@dk-compmath.jku.at<br />
DK Computational Mathematics, JKU University Linz<br />
Altenbergerstr. 69, 4040 Linz, Austria<br />
Akhlaq Husain akhlaq@iitk.ac.in<br />
Mathematics, In<strong>di</strong>an Institute <strong>of</strong> Technology Kanpur<br />
Department o Mathematics and Statistics, IIT Kanpur, Kanpur, Pin: 208016,<br />
In<strong>di</strong>a<br />
Bert Jüttler bert.juettler@jku.at<br />
Institute <strong>of</strong> Applied Geometry, Johannes Kepler University<br />
Altenberger Str. 69, 4040 Linz, Austria<br />
Clement Jourdana clement@imati.cnr.it<br />
<strong>Dipartimento</strong> <strong>di</strong> matematica, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia, Italy<br />
Peter Knabner knabner@am.uni-erlangen.de<br />
Department <strong>of</strong> Mathematics, Friedrich-Alexander Universität Erlangen-Nürn-<br />
berg<br />
Alexander Kolpakov algk@ngs.ru<br />
Martensstrasse 3 D 91058 Erlangen<br />
DAEIMI, Cassino University<br />
Via Di Biasio, 43 03043 Cassino (FR), Italy<br />
Jasper Kreeft j.j.kreeft@tudelft.nl<br />
Aerospace Engineering - Aerodynamics, Delft University <strong>of</strong> Technology<br />
Kluyverweg 1, 2629 HS Delft, The Netherlands<br />
Mukesh Kumar mukesh@imati.cnr.it<br />
Instituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Iinformatiche, CNR<br />
Via Ferrata 1, 27100 Pavia<br />
71
List <strong>of</strong> participants<br />
Joachim Linn joachim.linn@itwm.fraunh<strong>of</strong>er.de<br />
Mathematical Methods in Dynamics and Durability, Fraunh<strong>of</strong>er Institute<br />
for Industrial Mathematics (ITWM)<br />
Konstantin Lipnikov lipnikov@lanl.gov<br />
Fraunh<strong>of</strong>er Platz 1 D, 67663 Kaiserslautern, Germany<br />
Applied Mathematics and Plasma Physics, Los Alamos National Laboratory<br />
MS B284, Los Alamos, NM 87545, USA<br />
Carlo Lova<strong>di</strong>na carlo.lova<strong>di</strong>na@unipv.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia<br />
Teemu Luostari teemu.luostari@uef.fi<br />
Department <strong>of</strong> Physics and Mathematics, University <strong>of</strong> Eastern Finland<br />
University <strong>of</strong> Eastern Finland, Department <strong>of</strong> Physics and Mathematics,<br />
Kuopio Campus, P.O. Box 1627, FI-70211 Kuopio, Finland<br />
Carla Manni manni@mat.uniroma2.it<br />
<strong>Matematica</strong>, Universita <strong>di</strong> Roma “Tor Vergata”<br />
Via della Ricerca Scientifica 1, 00133 Roma<br />
Gianmarco Manzini marco.manzini@imati.cnr.it<br />
IMATI, CNR<br />
Via Ferrata 1, 27100 Pavia, Italia<br />
Aurora Mihaela Marica marica@bcamath.org<br />
Ikerbasque, Basque Center for Applied Mathematics<br />
Bizkaia Technology Park, Buil<strong>di</strong>ng 500, 48160, Derio, Basque Country, Spain<br />
Ilario Mazzieri ilario.mazzieri@mail.polimi.it<br />
Politecnico <strong>di</strong> Milano<br />
Via Bonar<strong>di</strong> 9, 20133 Milano<br />
Azaiez Mej<strong>di</strong> azaiez@ipb.fr<br />
Mecanics, Institut Polytechnique de Bordeaux<br />
Jens Markus Melenk melenk@tuwien.ac.at<br />
Lab TREFLE, 16 Av Pey Berland 33607 Pessac Cedex<br />
Institut für Analysis und Scientific Computing (E101), Technische Univer-<br />
sität Wien<br />
Wiedner Hauptstraße 8-10, A-1040 Wien, Austria<br />
Andrea Moiola andrea.moiola@sam.math.ethz.ch<br />
Seminar for Applied Mathematics, ETH Zürich<br />
HG J 45, Rämistrasse 101, 8092 Zürich, CH<br />
72
Peter Monk monk@math.udel.edu<br />
List <strong>of</strong> participants<br />
Department <strong>of</strong> Mathematical Sciences, University <strong>of</strong> Delaware<br />
Newark, DE 19716, USA<br />
Dang Manh Nguyen d.m.nguyen@mat.dtu.dk<br />
Department <strong>of</strong> Mathematics, Technical University <strong>of</strong> Denmark<br />
Matematiktorvet, Buil<strong>di</strong>ng 303S, Kgs. Lyngby, Denmark<br />
Peter Nørt<strong>of</strong>t Nielsen p.n.nielsen@mat.dtu.dk<br />
DTU Mathematics, Technical University <strong>of</strong> Denmark<br />
Bygning 303S, DK-2800 Kgs. Lyngby, Denmark<br />
Artur Palha a.palha@tudelft.nl<br />
Aerodynamics Department, Faculty <strong>of</strong> Aerospace Engineering, TUDelft<br />
Kluyverweg 2, 2629 HT Delft, The Netherlands<br />
Asieh Parsania asieh.parsania@math.uzh.ch<br />
Department <strong>of</strong> Mathematics, Zürich University<br />
Winterthurerstrasse 190, CH-8057 Zürich, Switzerland<br />
Luca Pavarino luca.pavarino@unimi.it<br />
Department <strong>of</strong> Mathematics, Università <strong>di</strong> Milano<br />
Via Sal<strong>di</strong>ni 50, 20133 Milano<br />
Ilaria Perugia ilaria.perugia@unipv.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia<br />
Jorg Peters jorg@cise.ufl.edu<br />
CISE, University <strong>of</strong> Florida<br />
Gainesville FL 32605<br />
Joel Phillips joel.phillips@rea<strong>di</strong>ng.ac.uk<br />
Mathematics, Rea<strong>di</strong>ng University<br />
Whiteknights, PO Box 220, Rea<strong>di</strong>ng RG6 6AX, UK<br />
Paola Pietra pietra@imati.cnr.it<br />
IMATI, CNR<br />
Via Ferrata 1, 27100 Pavia<br />
Valentina Poletti valentina.poletti@gmail.com<br />
Computational Science Institute, University <strong>of</strong> Lugano<br />
Via Giuseppe Buffi 13, Lugano, 6900<br />
Alessio Quaglino quaglino@math.uni-goettingen.de<br />
Numerical and Applied Maths, Göttingen University<br />
Lotzestr. 16 Göttingen 37073 Germany<br />
73
List <strong>of</strong> participants<br />
Ahmed Ratnani ahmed.ratnani@math.unistra.fr<br />
Applied mathematics, IRMA University <strong>of</strong> Strasbourg, INRIA Grand-Est<br />
7 rue René-Descartes, 67084 Strasbourg Cedex, France<br />
Alessandro Reali alessandro.reali@unipv.it<br />
Structural Mechanics Department, University <strong>of</strong> Pavia<br />
Via Ferrata 1, 27100, Pavia, Italy<br />
Elisabetta Repossi elisabetta.repossi@tiscali.it<br />
MOX- <strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Politecnico <strong>di</strong> Milano<br />
Via Bonar<strong>di</strong>, 9 20133 Milano, Italia<br />
Ju<strong>di</strong>th Rivas ju<strong>di</strong>th.rivas@ehu.es<br />
Martin Ruess ruess@tum.de<br />
Mathematics, University <strong>of</strong> the Basque Country<br />
Barrio Sarriena s/n, 48940 Leioa, Spain<br />
Center for Simulation Technology in Engineering, Technische Universität<br />
München<br />
Arcisstraße 21, 80333 München<br />
Gianni Sacchi gianni.sacchi@imati.cnr.it<br />
IMATI, CNR<br />
Via Ferrata 1, 27100 Pavia, Italy<br />
Giancarlo Sangalli giancarlo.sangalli@unipv.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Pavia<br />
Via Ferrata 1, 27100 Pavia<br />
Marcus Sarkis msarkis@wpi.edu<br />
Mathematical Sciences , Worcester Polytechnic Institute<br />
100 Institute Road, Worcester, MA 01609, USA<br />
Stefan Sauter stas@math.uzh.ch<br />
Institut für Mathematik, Universität Zürich<br />
Winterthurerstrasse 190, 8057 Zürich<br />
Simone Scacchi simone.scacchi@unimi.it<br />
Department <strong>of</strong> Mathematics, University <strong>of</strong> Milan<br />
Via Sal<strong>di</strong>ni 50, 20133 Milano, Italia<br />
Joachim Schöberl joachim.schoeberl@rwth-aachen.de<br />
Institute for Analysis and Scientific Computing, Vienna University <strong>of</strong> Tech-<br />
nology<br />
Wiedner Hauptstrasse 8-10, Wien, Austria<br />
74
Corina Simian corina.simian@math.uzh.ch<br />
List <strong>of</strong> participants<br />
Numerical Analyses, Institute for Mathematics, Zürich University<br />
Winterthurerstrasse 190, CH-8057 Zürich<br />
Daniil Svyatskiy dasvyat@lanl.gov<br />
Theoretical Division, Los Alamos National Laboratory<br />
Mail Stop B284, Los Alamos National Laboratory, Los Alamos, NM 87545,<br />
U.S.A.<br />
Pragnesh Thakkar pragnesh.thakkar@nirmauni.ac.in<br />
Mathematics Scection, Institute <strong>of</strong> Diploma Stu<strong>di</strong>es, Nirma University<br />
4,Varsha Apartment, Plot#184, Nehru Park, Opp.Vastrapur Lake, Vastra-<br />
pur, Ahmedabad, Gujarat, In<strong>di</strong>a<br />
Satyendra Tomar satyendra.tomar@ricam.oeaw.ac.at<br />
RICAM, Austrian Academy <strong>of</strong> Sciences<br />
Altenbergerstrasse 69, 4040 Linz, Austria<br />
Rafael Vázquez vazquez@imati.cnr.it<br />
Instituto <strong>di</strong> <strong>Matematica</strong> Applicata e Tecnologie Informatiche, CNR<br />
Via Ferrata 1, 27100 Pavia, Italia<br />
Alexander Veit alexander.veit@math.uzh.ch<br />
Institut für Mathematik, University <strong>of</strong> Zürich<br />
Winterthurerstrasse 190, CH-8057 Zürich<br />
Marco Verani marco.verani@polimi.it<br />
MOX-Department <strong>of</strong> Mathematics, Politecnico <strong>di</strong> Milano<br />
Via Bonar<strong>di</strong>, 9 20133 Milano<br />
Clau<strong>di</strong>o Ver<strong>di</strong> clau<strong>di</strong>o.ver<strong>di</strong>@unimi.it<br />
<strong>Dipartimento</strong> <strong>di</strong> <strong>Matematica</strong>, Università <strong>di</strong> Milano<br />
Via Sal<strong>di</strong>ni 50, 20133 Milano<br />
Anh-Vu Vuong vuong@ma.tum.de<br />
Centre for Mathematical Sciences, Technische Universität München<br />
Fakultät für Mathematik, Lehrstuhl für Numerische Mathematik (M2), Boltz-<br />
mannstraße 3 D, 85748 Garching, Germany<br />
Mary F. Wheeler mfw@ices.utexas.edu<br />
Center for Subsurface Modeling, The University <strong>of</strong> Texas at Austin<br />
1 University Station C0200, Austin, Texas 78712<br />
Ragnar Winther ragnar.winther@cma.uio.no<br />
Centre <strong>of</strong> Mathematics for Applications (CMA), University <strong>of</strong> Oslo<br />
CMA, P.O. Box 1053, Blindern, 0316 Oslo, Norway<br />
75
List <strong>of</strong> participants<br />
Gang Xu gxu@sophia.inria.fr<br />
GALAAD Team, INRIA Sophia-antipolis<br />
2004, route des Lucioles B.P. 93 F, 06902 Sophia Antipolis Cedex, France<br />
Ivan Yotov yotov@math.pitt.edu<br />
Department <strong>of</strong> Mathematics, University <strong>of</strong> Pittsburgh<br />
301 Thackeray Hall, Pittsburgh, PA 15260, USA<br />
76