Unless otherwise specified, all talks are held starting at 1.15pm in the Theory Library on the 4th floor of Physics East.

Everyone is encouraged to ask questions during the talk. You are welcome to leave when you need to, otherwise the talk will usually wrap up by 2.30pm, at which point there will be biscuits.

Thursday 29th October 2026: Mike Blake (Bristol)

Many-body chaos, hydrodynamics and horizon symmetries

I will describe recent connections between two naively distinct aspects of the dynamics of finite-temperature quantum field theories. These are many-body chaos [as probed through OTOCs of local operators] and hydrodynamics [an effective description of conserved quantities, specifically energy density]. I will first review a proposal that in strongly coupled QFTs the exponential growth of OTOCs can be understood as arising from exponential zero modes of a hydrodynamic effective action. I will then describe recent work in which we explicitly derive hydrodynamic actions using the AdS/CFT correspondence. In this setting it has been argued that the conjectured exponential zero modes of hydrodynamics are associated with certain horizon symmetries – that is infra-red diffeomorphisms that preserve certain structures of the horizon. I will present our explicit constructions of hydrodynamic actions dual to black holes with bulk dimensions d=2,3,4, as well as a general argument for how to identify gravitational horizon symmetries with zero modes of hydrodynamic actions. For d=2,3, the hydrodynamic actions we construct explicitly realise a form of horizon symmetries argued for by Knysh, Liu and Pinzani-Fokeeva.

Wednesday 11th November 2026: Yan Fyodorov (King’s College London)

Exact spectral integrability of a finite-channel random Schrödinger operator and edge profile of random banded matrices

We show that a quasi-one-dimensional matrix Schrödinger operator with a Hermitian white-noise potential admits an exact free-fermion
representation at arbitrary sample length $L$. In the limit of $L\to\infty$, this representation yields exact expressions for both
the density of states and the total Lyapunov growth rate in terms of a single Airy Hankel $\tau$-function admitting a Kontsevich Airy
integral representation. Saddle-point and double-scaling analyses of this integral capture, within a unified framework, both the
Lifshitz tail of the density of states at finite channel number and the large-$N$ spectral edge. In particular, for $N\gg 1$, the
sharp $N=\infty$ edge is rounded over a spectral window of width $N^{-4/5}$, with a profile governed by the tritronquée solution
of Painlevé I. We conjecture that this quasi-one-dimensional edge-scaling profile is universal for a broad class of $\beta=2$
random band matrices, extending beyond the setting of Sasha Sodin’s earlier analysis [Ann. Math., 2010]. The talk is based
on joint work with Pierre Le Doussal, arXiv:2610.27154.

NB: Non-standard day

Thursday 12th November 2026: Paul Sutcliffe (Durham)

Magnetic monopoles

I shall review magnetic monopole solutions of SU(2) Yang-Mills-Higgs gauge theory. Most studies concern a limit of this theory in which the static monopole solutions are obtained as solutions of an integrable equation. After discussing the properties of integrable monopoles, some recent work will be presented on monopoles beyond the integrable limit.

Thursday 19th November 2026: Oliver Lunt (Oxford)

TBA

TBA

Thursday 26th November 2026: Thomas Elliott (Manchester)

TBA

TBA

Thursday 10th September 2026: Sean Mitchell (Birmingham)

Using a Multiple Sphere T-Matrix Method to Simulate Light Transport Deep in Biological Tissue

The T-matrix method may be used to efficiently calculate the scattered field from a particle and is extendible to calculating fields from ensembles containing many thousands of spheres or particles. We present the application of the multiple sphere T-matrix method to simulating light transport in thick biological tissue. This is done by finding distributions of spheres which broadly represent the scattering properties of tissue. Our work is motivated by challenges with using optical light to image through and inside of biological tissue due to the scattering of light, which limits imaging depths to a few centimetres. Optical techniques to revert scattering effects such as wavefront shaping, and biomedical imaging methods such as Optical Coherence Tomography (OCT), interferometric Near Infrared Spectroscopy (iNIRS), and speckle correlation depend on deterministic coherent light phenomena which are computationally expensive to accurately model. Using the T-matrix method, we can perform full-wave simulations of light transport through millimetre-scale sections of biological tissue – a depth comparable to the transport mean free path of tissue – which was prohibitively expensive to simulate with prior methods. Our model could be used to improve wavefront shaping algorithms and improve understanding of imaging methods, paving the way for optical imaging in deeper tissue.

Thursday 1st October 2026: Jack Binysh (Birmingham Mechanical Engineering)

Non-reciprocal Robotic Metamaterials

Robotics promises to improve our lives by automating repetitive, delicate and dangerous tasks. Traditional robots have made huge strides towards these goals, but struggle with unpredictable environments and are ill-suited to miniaturization. By contrast, brainless organisms like starfish can squeeze through gaps and locomote over complex terrain by exploiting feedback between their soft bodies and the environment.

In this talk I will argue that robotic metamaterials–architected lattices woven from many sensors and actuators–are a promising platform to capture such embodied intelligence. I will show that these machine materials can autonomously crawl, dig and walk without central control by collectively breaking reciprocal symmetries in their mechanical response. These materials may form the basis for smarter robot bodies that are robust, conformal and scalable.

Thursday 8th October 2026: Igor Rozhanskiy (Manchester)

Spin-orbit physics in two-dimensional semiconductors

Atomically thin transition-metal dichalcogenides (TMDs) combine the electrostatic tunability of graphene with strong spin-orbit coupling and a valley degree of freedom. Their broken inversion symmetry locks spin and valley at the band edges, making these materials an attractive platform for studying spin- and valley-dependent transport. I will begin with an introduction to this physics and then focus on a surprisingly basic unresolved question: how large is the spin–orbit splitting of the conduction band in monolayer MoS2?

Conventional density functional theory predicts a splitting of only a few meV, whereas magnetotransport experiments indicate a value well above 10 meV. Shubnikov–de Haas measurements in high-mobility hBN-encapsulated MoS2 allow the onset of occupation of the upper spin-split conduction band to be identified directly. Explaining this observation requires both many-body and single-particle physics. Exchange interactions in the finite-density electron gas substantially enhance the splitting and give it a pronounced density dependence. The remaining discrepancy can be traced to the microscopic orbital composition of the band edge, where competing contributions make the splitting unusually sensitive to metal-chalcogen hybridisation. A DFT+U+V description corrects this balance and brings the bare splitting into agreement with the value inferred from transport.

I will conclude by discussing how the same ideas extend across the TMD family and to spin-orbit coupling induced in graphene-TMD heterostructures.

Thursday 15th October 2026: Andrew Green (University College London)

Time-evolution of closed, thermalising quantum systems

In this nascent age of quantum computing, it is routine to question the complexity of quantum simulation. The quantum state space grows exponentially with the number of particles, implying that a quantum computer is required. There are several reasons why this argument is incomplete. Firstly, we do not need the full wavefunction to compute most material properties. E.g. conductivity is given by spatial and temporal correlations of local observables through the Kubo Greenwood formula. Moreover, the properties of materials remain the same for samples above a certain size, so scaling arguments certainly end at some point. Finally, all* Hamiltonians obey the eigenstate thermalisation hypothesis, implying the emergence of a finite correlation length and time beyond which the complexity of simulating local observables scales classically. Motivated by this, I introduce a new method of quantum simulation that optimises the local rather than global similarity of the wavefunction. The resulting algorithm can be run on both classical and quantum hardware, extending the timescales for which classical simulation of local observables is possible and giving a clearer view of when a quantum simulation is advantageous.

*Except an interesting zero-measure set