Past

H. Johnston and J.G. Liu proposed in 2004 a numerical scheme for approximating numerically solutions of the incompressible Navier-Stokes system. The scheme worked very well in practice but its analytic properties remained elusive.\newline

In order to understand these analytical aspects they considered together with R. Pego a continuous version of it that appears as an extension of the incompressible Navier-Stokes to vector-fields that are not necessarily divergence-free. For divergence-free initial data one has precisely the incompressible Navier-Stokes, while for non-divergence free initial data, the divergence is damped exponentially.\newline

We present analytical results concerning this extended system and discuss numerical implications. This is joint work with R. Pego, G. Iyer (Carnegie Mellon) and J. Kelliher, M. Ignatova (UC Riverside).

Saturated fusion systems are a next generation approach to the theory of finite groups- one major motivation being the opportunity to borrow techniques from homotopy theory. Extending work of Broto, Levi and Oliver, we introduce a new object - a 'tree of fusion systems' and give conditions (in terms of the orbit graph) for the completion to be saturated. We also demonstrate that these conditions are 'best possible' by producing appropriate counterexamples. Finally, we explain why these constructions provide a powerful way of building infinite families of fusion systems which are exotic (i.e. not realisable as the fusion system of a finite group) and give some concrete examples.

Although the importance of hydrologic uncertainty is widely recognized it is rarely considered in control problems, especially real-time control. One of the reasons is that stochastic control is computationally expensive, especially when control decisions are derived from spatially distributed models. This talk reviews relevant control concepts and describes how reduced order models can make stochastic control feasible for computationally demanding applications. The ideas are illustrated with a classic problem -- hydraulic control of a moving contaminant plume.

Let $X$ be a variety and $Z$ be a sub-variety. Can one glue vector bundles on $X-Z$ with vector bundles on some small neighborhood of $Z$? We survey two recent results on the process of gluing a vector bundle on the complement of a sub-variety with a vector bundle on some 'small' neighborhood of the sub-variety. This is joint work. The first with M. Temkin and is about gluing categories of coherent sheaves over the category of coherent sheaves on a Berkovich analytic space. The second with J. Block and is about gluing dg enhancements of the derived category of coherent sheaves.

In an Erdős--R\'enyi random graph above the phase transition, i.e.,

where there is a giant component, the size of (number of vertices in)

this giant component is asymptotically normally distributed, in that

its centred and scaled size converges to a normal distribution. This

statement does not tell us much about the probability of the giant

component having exactly a certain size. In joint work with B\'ela

Bollob\'as we prove a `local limit theorem' answering this question

for hypergraphs; the graph case was settled by Luczak and Łuczak.

The proof is based on a `smoothing' technique, deducing the local

limit result from the (much easier) `global' central limit theorem.

High-pressure freezing processes are a novel emerging technology in food processing,
offering significant improvements to the quality of frozen foods. To be able to simulate
plateau times and thermal history under different conditions, a generalized enthalpy
model of the high-pressure shift freezing process is presented. The model includes
the effects of pressure on conservation of enthalpy and incorporates the freezing point
depression of non-dilute food samples. In addition, the significant heat-transfer effects of
convection in the pressurizing medium are accounted for by solving the two-dimensional
Navier–Stokes equations.
The next question is: is high-pressure shift freezing good also in the long run?
A growth and coarsening model for ice crystals in a very simple food system will be discussed.

Topological quantum error correcting codes, such as the Toric code, are
ideal candidates for protecting a logical quantum bit against local noise.
How are we to get the best performance from these codes when an unknown
local perturbation is applied? This talk will discuss how knowledge, or lack
thereof, about the error affects the error correcting threshold, and how
thresholds can be improved by introducing randomness to the system. These
studies are directed at trying to understand how quantum information can be
encoded and passively protected in order to maximise the span of time between successive rounds of error correction, and what properties are
required of a topological system to induce a survival time that grows
sufficiently rapidly with system size. The talk is based on the following
papers: arXiv:1208.4924 and Phys. Rev. Lett. 107, 270502 (2011).

The Preisach model of hysteresis has a long history, a convenient algorithmic form and "nice" mathematical properties (for a given value of nice) that make it suitable for use in differential equations and other dynamical systems. The difficulty lies in the fact that the "parameter" for the Preisach model is infinite dimensional—in full generality it is a measure on the half-plane. Applications of the Preisach model (two interesting examples are magnetostrictive materials and vadose zone hydrology) require methods to specify a measure based on experimental or observed data. Current approaches largely rely on direct measurements of experimental samples, however in some cases these might not be sufficient or direct measurements may not be practical. I will present the Preisach model in all its glory, along with some history and applications, and introduce an open inverse problem of fiendish difficulty.

This presentation concerns five topics in computational partial differential equations with the overall goals of reliable error control and efficient simulation.

The presentation is also an advertisement for nonstandard discretisations in linear and nonlinear Computational PDEs with surprising advantages over conforming

finite element schemes and the combination

of the two. The equivalence of various first-order methods is explained for the linear Poisson model problem with conforming

(CFEM), nonconforming (NC-FEM), and mixed finite element methods (MFEM) and others discontinuous Galerkin finite element (dGFEM). The Stokes

equations illustrate the NCFEM and the pseudo-stress MFEM and optimal convergence of adaptive mesh-refining as well as for guaranteed error bounds.

An optimal adaptive CFEM computation of elliptic eigenvalue

problems and the computation of guaranteed upper and lower eigenvalue bounds based on NCFEM. The obstacle problem and its guaranteed error

control follows another look due to D. Braess with guaranteed error bounds and their effectivity indices between 1 and 3. Some remarks on computational

microstructures with degenerate convex minimisation

problems conclude the presentation.

I will discuss recent joint work with S. Galatius, in which we

generalise the Madsen--Weiss theorem from the case of surfaces to the

case of manifolds of higher even dimension (except 4). In the simplest

case, we study the topological group $\mathcal{D}_g$ of

diffeomorphisms of the manifold $\#^g S^n \times S^n$ which fix a

disc. We have two main results: firstly, a homology stability

theorem---analogous to Harer's stability theorem for the homology of

mapping class groups---which says that the homology groups

$H_i(B\mathcal{D}_g)$ are independent of $g$ for $2i \leq g-4$.

Secondly, an identification of the stable homology

$H_*(B\mathcal{D}_\infty)$ with the homology of a certain explicitly

described infinite loop space---analogous to the Madsen--Weiss

theorem. Together, these give an explicit calculation of the ring

$H^*(B\mathcal{D}_g;\mathbb{Q})$ in the stable range, as a polynomial

algebra on certain explicitly described generators.

In the absence of background fluxes and sources, the compactification of string theories on Calabi-Yau threefolds leads to supersymmetric solutions.Turning on fluxes, e.g. to lift the moduli of the compactification, generically forces the geometry to break the Calabi-Yau conditions, and to satisfy, instead, the weaker condition of reduced structure. In this talk I will discuss manifolds with SU(3) structure, and their relevance for heterotic string compacitications. I will focus on domain wall solutions and how explicit example geometries can be constructed.

We consider a class of stochastic control problems with fuel constraint that are closely connected to the problem of finding adaptive mean-variance-optimal portfolio liquidation strategies in the Almgren-Chriss framework. We give a closed-form solution to these control problems in terms of the log-Laplace transforms of certain J-functionals of Dawson-Watanabe superprocesses. This solution can be related heuristically to the superprocess solution of certain quasilinear parabolic PDEs with singular terminal condition as given by Dynkin (1992). It requires us to study in some detail the blow-up behavior of the log-Laplace functionals when approaching the singularity.

In this talk, I will show that Brauer classes of a locally solvable degree 4 del Pezzo surface X are vertical, that is, that every Brauer class is obtained by pullback from an element of Br k(P^1) for some rational map f : X ----> P^1. As a consequence, we see that a Brauer class does not obstruct the existence of a rational point if and only if there exists a fiber of f that is locally solvable. The proof is constructive and gives a simple and practical algorithm, distinct from that in [Bright,Bruin,Flynn,Logan (2007)], for computing all nonconstant classes in the Brauer group of X. This is joint work with Anthony V\'arilly-Alvarado.

We study a thin liquid film on a vertical fibre. Without gravity, there

is a Rayleigh-Plateau instability in which surface tension reduces the

surface area of the initially cylindrical film. Spherical drops cannot

form because of the fibre, and instead, the film forms bulges of

roughly twice the initial thickness. Large bulges then grow very slowly

through a ripening mechanism. A small non-dimensional gravity moves the

bulges. They leave behind a thinner film than that in front of them, and

so grow. As they grow into large drops, they move faster and grow

faster. When gravity is stronger, the bulges grow only to finite

amplitude solitary waves, with equal film thickness behind and in front.

We study these solitary waves, and the effect of shear-thinning and

shear-thickening of the fluid. In particular, we will be interested in

solitary waves of large amplitudes, which occur near the boundary

between large and small gravity. Frustratingly, the speed is only

determined at the third term in an asymptotic expansion. The case of

Newtonian fluids requires four terms.

I will give an introduction to how SU(3)-structures appear in heterotic string theory and string compactifications. I will start by considering the zeroth order SU(3)-holonomy Calabi-Yau scenario, and then see how this generalizes when higher order effects are considered. If time, I will discuss some of my own work.

A triangulated category admits a strong generator if, roughly speaking,

every object can be built in a globally bounded number of steps starting

from a single object and taking iterated cones. The importance of

strong generators was demonstrated by Bondal and van den Bergh, who

proved that the existence of such objects often gives you a

representability theorem for cohomological functors. The importance was

further emphasised by Rouquier, who introduced the dimension of

triangulated categories, and tied this numerical invariant to the

representation dimension. In this talk I will discuss the generation

time for strong generators (the least number of cones required to build

every object in the category) and a refinement of the dimension which is

due to Orlov: the set of all integers that occur as a generation time.

After introducing the necessary terminology, I will focus on categories

occurring in representation theory and explain how to compute this

invariant for the bounded derived category of the path algebras of type

A and D, as well as the corresponding cluster categories.

Partial differential equations (PDEs) with random coefficients are used in computer simulations of physical processes in science, engineering and industry applications with uncertain data. The goal is to obtain quantitative statements on the effect of input data uncertainties for a comprehensive evaluation of simulation results. However, these equations are formulated in a physical domain coupled with a sample space generated by random parameters and are thus very computing-intensive.

We outline the key computational challenges by discussing a model elliptic PDE of single phase subsurface flow in random media. In this application the coefficients are often rough, highly variable and require a large number of random parameters which puts a limit on all existing discretisation methods. To overcome these limits we employ multilevel Monte Carlo (MLMC), a novel variance reduction technique which uses samples computed on a hierarchy of physical grids. In particular, we combine MLMC with mixed finite element discretisations to calculate travel times of particles in groundwater flows.

For coefficients which can be parameterised by a small number of random variables we employ spectral stochastic Galerkin (SG) methods which give rise to a coupled system of deterministic PDEs. Since the standard SG formulation of the model elliptic PDE requires expensive matrix-vector products we reformulate it as a convection-diffusion problem with random convective velocity. We construct and analyse block-diagonal preconditioners for the nonsymmetric Galerkin matrix for use with Krylov subspace methods such as GMRES.