(January 25) Dr. Ari Stern:
Hilbert complexes, finite element exterior calculus, and problems on
Riemannian hypersurfaces
—
In recent years, the success of "mixed" finite element methods has
been shown to have surprising connections with differential geometry
and algebraic topology---particularly with the calculus of exterior
differential forms, de Rham cohomology, and Hodge theory. In this
talk, I will discuss how the notion of "Hilbert complex," rather than
"Hilbert space," provides the appropriate functional-analytic setting
for the numerical analysis of these methods. Furthermore, I will
present some recent results that extend this analysis from polyhedral
regions in Euclidean space to arbitrary Riemannian manifolds. As a
direct consequence, our analysis also generalizes several key results
on "surface finite element methods" for the approximation of elliptic
PDEs on hypersurfaces (e.g., membranes undergoing geometric evolution).
(February 1) Prof. Elisabeth Ullman:
Efficient Iterative Solvers for Stochastic Galerkin Discretizations of
the Lognormal Diffusion Problem —
We consider stochastic Galerkin discretizations of a steady-state
diffusion problem with stochastically nonlinear, lognormally distributed
diffusion coefficient. The iterative solution of the resulting large,
coupled system of Galerkin equations is computationally demanding, since
matrix-vector products with the Galerkin matrix are expensive and the
Galerkin matrix is ill-conditioned with respect to the statistical
parameters of the problem. We consider a reformulated version of the
lognormal diffusion problem in terms of a stochastic
convection-diffusion problem with random convective velocity which
depends linearly on a fixed number of statistically independent Gaussian
random variables. The associated Galerkin matrix is nonsymmetric but
sparse and allows for fast matrix-vector multiplications with optimal
complexity. We construct and analyze two block diagonal preconditioners
for this Galerkin matrix for use with the generalized minimal residual
method (GMRES). We test the efficiency of the proposed preconditioning
approaches and compare the iterative solver performance for a lognormal
diffusion problem posed as convection-diffusion and diffusion problem. (February 8) Dr. Markus Schmuck:
A porous media approximation of the Stokes-Poisson-Nernst-Plank equations —
We first introduce a classical continuum model which allows to
describe essential electrokinetic phenomena as electro-phoresis
and -osmosis. Applications and corresponding theory range from
design of microfluidic devices, energy storage devices,
semiconductors over modeling communication in biological cells by
nanopores. Based on this classical formulation, we derive
effective macroscopic equations which describe binary symmetric
electrolytes in porous media. The inhomogeneous structure of
porous materials naturally induces additional nonlinearities
which are called "material tensors". A better understanding of
geometric effects in heterogeneous media on ionic transport is
expected by the new formulation. The numerical advantage bases on
a strong reduction of the degrees of freedom by upscaling the
microstructure. The results are gained by the two-scale
convergence method. In the remaining part of the talk, we
motivate the derivation of first error bounds between the exact
microscopic solution and its upscaled macroscopic approximation.
(February 10) Prof. Peter Constantin:
Regularity issues for the Surface Quasi-Geostrophic and related
equations —
"The Surface Quasi-Geostrophic Equation" is a nonlinear PDE which has
a geophysical origin, but is mainly used as a theoretical testing
ground for ideas about singularities in incompressible fluids. I'll
motivate the model and present recent results, some numerical,
mostly analytical. The main problem is that of global regularity for
the inviscid equation, and that is still open.
(February 11) Prof. Peter Constantin:
Complex Fluids —
The talk will be about some of the models used to describe fluids with
iparticulate matter suspended in them. Some of these models are very
complicated. After a bit of history and a review of known results, I
will try to point out some open problems, isolate some of the
mathematical difficulties, and illustrate some of the phenomena on
simpler didactic models. (February 15) Prof. A. Javier Sayas:
Energy estimates in semidiscrete time-domain boundary integral equations —
Boundary integral operators in the time domain
offer competitive ways to solve exterior scattering problems
for different kinds of transient (acoustic, elastic,
visco-elastic,?electromagnetic,...) waves. They also provide
exact absorbing boundary conditions that can be placed
arbitrarily close to the support of source terms or inhomogeneities.
In this talk I will address the problem of how space
discretization with Galerkin BEM redistributes the energy,
leaking energy to the interior of the scatterer. In the case of
transmission problems, BEM-FEM space discretization creates a
ghost wave in the interior domain. This effect can be shown by
considering abstract wave equations with exotic transmission
condition that are exactly satisfied by the semidiscrete
equations. This novel energy analysis can be used to better
understand the correct balance of energy of Boundary Integral
Methods and might lead to evidence for or against the
possibility of using non-symmetric BEM-FEM coupled methods
in the time domain. Because of its generality, the conclusions
are valid both for Galerkin or Convolution Quadrature
discretizations in time. (February 22) Prof. Petr Plechak:
Born-Oppenheimer approximation and accuracy of molecular dynamics —
I shall discuss recent results from the joint work with C. Bayer, H. Hoel, A. Szepessy
and R. Tempone concerning accurate approximation of Schroedinger observables in
molecular systems with an electron spectral gap. We present error estimates that can
be used for estimating accuracy of molecular dynamics as an approximation of the
evolution of heavy nuclei in a many-body quantum system. The presented approach is
based on the study of the time-independent Schroedinger equation and thus differs from
a more standard analysis derived from time-dependent Schroedinger equation. It gives a
different perspective on the Born-Oppenheimer approximation, Schroedinger Hamiltonian
systems and numerical simulations in molecular dynamics at micro-canonical ensemble. (March 1) Prof. Adrian Lew:
Discontinuous Galerkin Methods in Solid Mechanics —
Some problems in Solid and Structural mechanics require special care
when analyzed via finite element approximations. Typical examples are
problems that involve kinematic constraints, such as in incompressible
elasticity or Reissner-Mindlin plate models, and problems involving
moving boundaries, such as evolving cracks, phase transition
interfaces or shape optimizations. Nonlinearities in the material
behavior only exacerbate these difficulties, and immediately rule out
many of the proposed solutions. In this talk I will show how under these circumstances Discontinuous
Galerkin methods provide an attractive and advantageous
alternative. The overarching idea I will convey is that by relaxing
the constraint of having continuous displacements across element
boundaries, Discontinuous Galerkin methods are able to impose other
kinematic constraints in the problem and still provide accurate
solutions. I will demonstrate it by showcasing the performance of a
class of Discontinuous Galerkin methods we introduced in a variety of
circumstances. First, in nonlinear elasticity problems involving
different kinematic constraints. Second, in a class of immersed
boundary methods, which sidestep the need for automatic remeshing in
problems with evolving boundaries by embedding the boundary in any
mesh. And finally, in the accurate solution of the stress and
displacement fields around cracks when cracks are "embedded" in the
mesh, as in extended finite element methods. In the three cases,
I will comment on recent convergence results we
obtained. In all cases, a special interpolant taking advantage of the
discontinuities in the finite element space had to be constructed,
since standard interpolants in conforming finite element spaces can be
guaranteed to converge only at suboptimal rates. I will conclude the
talk by briefly commenting on an adaptive stabilization technique we
created to enhance the robustness of the method in highly nonlinear
problems. Some of these ideas carry over to a wider variety of
popular methods in Solid Mechanics, such as Enhanced Strain Methods.
(March 8) Prof. Christian Klingenberg:
A numerical scheme for ideal MHD and its application to turbulence simulations
in astrophysics —
The reported work gives an overview of a code for ideal MHD developed at the
University of Oslo applied to astrophysical simulations done at Wrzburg University
both of which I have contributed to.
The main ingredients of the code are: finite volume scheme, an appoximate Riemann
solver based on relaxation, using a Powell source term, second order positivity
preserving reconstructions. The application consist in doing driven turbulence
simulations in order to study their influence on star formation. (March 15) Prof. Alina Chertock:
Sinking, merging and stationary plumes in a coupled chemotaxis-fluid model: a
high-resolution numerical approach —
Aquatic bacteria like Bacillus subtilis are heavier than water yet they are able to
swim upwards an oxygen gradient and concentrate in a layer below the water surface, which will
undergo Rayleigh-Taylor type instabilities for sufficiently high concentrations. Recently, a
simplified chemotaxis-fluid system has been proposed as a model for modestly diluted
cell-suspensions. It couples a convective chemotaxis system for the oxygen-consuming and
oxyotactic bacteria with the incompressible Navier-Stokes equations subject to a gravitational
force proportional to the relative surplus of the cell-density compared to the water-density. In this talk, I will present a high-resolution vorticity-based hybrid finite-volume
finite-difference scheme, which allows to investigate the nonlinear dynamics of a
two-dimensional chemotaxis fluid system. In particular, I will show selected numerical examples,
which illustrate (i) the formation of sinking plumes, (ii) the possible merging of neighboring
plumes and (iii) the convergence towards numerically stable stationary plumes. The examples with
stable stationary plumes show how the surface-directed oxyotaxis feeds continuously cells into a
high-concentration layer near the surface, from where the fluid-flow (recurring upwards in the
space between the plumes) transports the cells into the plumes, where then gravity makes the
cells sink and constitutes the driving force in maintaining the fluid-convection and, thus, in
shaping the plumes into (numerically) stable stationary states. Joint work with K. Fellner, A. Kurganov, A. Lorz and P. Markowich. (March 29) Prof. Georg Dolzmann:
Modeling and simulation of vectorfields on membranes —
The fundamental models for lipid bilayers are curvature based
and neglect the internal structure of the lipid layers. In this
talk, we explore models with an additional order parameter
which describes the orientation of the lipid molecules in the
membrane and compare their predictions based on numerical
simulations. This is joint work with Soeren Bartels (Bonn) and
Ricardo Nochetto (College Park).
(April 5) Prof. Jay Gopalachisnan:
Design and analysis of the new Discontinuous Petrov-Galerkin schemes —
Starting with simple examples we will show how to design numerical methods of the DPG
(Discontinuous Petrov-Galerkin) type. Petrov-Galerkin method find approximations in a
trial space by weakly imposing the equations using a test space. Our design principle
is that while trial spaces must have good approximation properties, the test space
must be chosen for stability. When this idea is applied to discontinuous finite
element spaces, we obtain methods that exhibit remarkable stability properties with
respect to mesh size and approximating polynomial degrees. (April 12) Prof. K.G. van der Zee:
Goal-Oriented Error Estimation and Adaptivity of Interface
Problems: Free Boundaries and Diffuse Interfaces —
Problems with evolving interfaces arise in many applications in
science and engineering. To obtain numerical approximations
various approaches have been developed, for example
moving-domain (ALE) methods and diffuse-interface models.
Interface problems are inherently nonlinear, displaying
multiple length and time scales, and therefore would benefit
highly from error estimation and adaptive techniques. In this talk, we consider goal-oriented error estimation and
adaptivity of interface problems. Goal-oriented error estimates
are dual-based a posteriori estimates of the error in output
functionals of the solution (the quantities of interest). The
estimates rely on the computation of a dual
(linearized-adjoint) problem, which can be highly nontrivial
for interface problems. We consider two examples: a
free-boundary problem and a time-dependent diffuse-interface
model. For free-boundary problems, such as fluid-structure
interaction, the linearization of the geometric nonlinearity
required for the dual problem poses a significant complication.
We show that the linearization using shape derivatives yields a
dual problem corresponding to a curvature-dependent
boundary-value problem. For diffuse-interface problems, such as
the Cahn-Hilliard equation of phase separation, the
linearized-adjoint problem is a backwards-in-time evolution
equation with unstable growth in the phase-separating regime.
For both problems we present numerical experiments that show
the effectivity of the dual-based error estimates.
(April 19) Dr. Irene Kyza:
Error analysis of time-discrete higher order ALE formulations —
ALE formulations are useful when approximating solutions of problems in deformable
domains, such as fluid-structure interactions. For realistic simulations involving
fluids in 3d, it is important that the ALE method is at least of second order of
accuracy. Second order ALE methods in time without any constraint on the time step do
not exist so far in the literature and the role of the so called geometric
conservation law (GCL) for stability and accuracy is not clear. We propose
discontinuous Galerkin (dG) methods of any order in time for a time dependent
advection-diffusion model problem in moving domains. We prove that our proposed
numerical schemes are unconditionally stable and that the conservative and non
conservative formulations are equivalent, when the involved integrals are computed
exactly. Both a priori and a posteriori error analyses are provided. The final error
estimates are of optimal order of accuracy. When using quadrature for the evaluation
of the integrals, we give a sufficient condition for preserving the stability and the
accuracy of the numerical schemes, without any constraint on the time-step. This
condition is a generalization of the GCL. Stability and optimal order error estimates,
under a mild restriction on the time step, are also obtained when applying the natural
for the dG method in time Radau quadrature. Numerical examples confirm our theoretical
results. This is joint work with Andrea Bonito and Ricardo H. Nochetto
(April 26) Dr. Josef Sifuentes:
The stability of GMRES convergence, with application to
preconditioning by approximate deflation —
How does GMRES convergence change when the coefficient matrix is
perturbed? Through resolvent estimates we develop simple, general
bounds to quantify the lag in convergence such a perturbation can induce.
This analysis is particularly relevant for preconditioned systems,
where an ideal preconditioner is only approximately applied
in practical computations. To illustrate the utility of this
theory, we combine our analysis with Stewart's invariant
subspace perturbation theory to develop rigorous bounds
on the performance of approximate deflation preconditioning
using Ritz vectors.
(May 3) Prof. Florian Potra:
Interior-Point Methods —
Interior point methods have revolutionized the field of mathematical
programming over the past three decades. They have been used for
proving polynomial complexity for different classes of mathematical
programming problems, and they have been implemented in very
efficient software packages for solving large scale optimization
problems arising in a variety of applications. While the implemented
interior point methods may not always have proven computational
complexity, they typically posses superlinear convergence. The talk
highlights the most relevant results on the polynomial complexity
and superlinear convergence of interior point methods in the
literature, and presents some recent results obtained by the speaker
and his collaborators.
(May 10) Prof. Bernardo Cockburn:
Superconvergent DG methods for elliptic problems —
We propose a new error analysis of a large class of finite element methods for
second order elliptic problems which includes the hybridized version of the main mixed
methods and the hybridizable discontinuous Galerkin methods. The main feature of this approach is that it reduces the main difficulty of the
analysis to the verification of some properties of an auxiliary, locally defined
projection and of the local spaces defining the methods. They guarantee the optimal
convergence of the approximate flux and the superconvergence of an element-by-element
postprocessing of the scalar variable. New mixed and hybridizable discontinuous Galerkin methods with these properties are
devised which are defined on squares, cubes and prisms. (May 17) Prof. Yoichiro Mori:
A Three-Dimensional Model of Electrodiffusion and Osmosis —
We introduce a three-dimensional model of cellular electrical activity. This model
takes into account the three-dimensional geometry of biological tissue as well as
ionic concentration dynamics, both of which are neglected in conventional models of
electrophysiology. This model is applied to study an anomalous mode of cardiac action
potential propagation: cardiac propagation without gap junctions. Then, we shall
discuss an extension of this model to incorporate osmotic and membrane mechanical
effects. The salient feature of this extension is that it satisfies a natural free
energy identity. We discuss how this free energy identity has led to the resolution of
a longstanding question on the stability of steady states of pump-leak models of cell
volume control.
webmaster@math ||
Math Department ||
Numerical Analysis ||
Seminars
|