Research

Dynamic assembly of active matter

How can self-propelled particles control the assembly of matter?

Active particles convert energy into motion, creating structures and collective behaviour far from equilibrium. We study their dynamics and use activity to direct the assembly of passive building blocks.

Our simulations and theory investigate glass transitions, crystallisation, and effective interactions in active matter. Work on self-propelled hard spheres has explored how activity shifts the glass transition, helps crystallise hard-sphere glasses, and mediates tunable long-range forces. We also study how chirality changes collective behaviour, including dynamical clustering that interrupts motility-induced phase separation. These studies seek physical principles for using active particles to help assemble functional materials.

Related publications

• Pushing glass transition towards random close packing using self-propelled hard spheres (2013)
• Crystallizing hard-sphere glasses by doping with active particles (2014)
• Tunable long range forces mediated by self- propelled colloidal hard spheres (2015)
• Dynamical Clustering Interrupts Motility Induced Phase Separation in Chiral Active Brownian Particles (2022)

Non-equilibrium hyperuniform fluids

Can a disordered fluid suppress density fluctuations like a crystal?

Hyperuniform materials suppress density fluctuations over large distances without requiring crystalline order. We investigate how driven particle systems generate this unusual form of organisation.

We identified strongly hyperuniform fluids in two-dimensional systems of chiral active particles moving along circular trajectories. Large density fluctuations remain below the orbit radius, while fluctuations are suppressed at larger scales. Our dynamic mean-field theory connects the local fluctuations to motility-induced microphase separation and the global hyperuniformity to diffusion with locally centre-of-mass-conserving noise. We also developed a hydrodynamic description of random-organising hyperuniform fluids and investigated how a hyperuniform fluid freezes.

Related publications

• Non-Equilibrium Strongly Hyperuniform Fluids of Circle Active Particles with Large Local Density Fluctuations (2019)
• Hydrodynamics of Random-Organizing Hyperuniform Fluids (2019)
• How does a hyperuniform fluid freeze? (2023)
• Non-equilibrium dynamic hyperuniform states (2025)

Programmable assembly of DNA-coated colloids

How can molecular linkers program where colloidal building blocks assemble?

DNA-coated colloids offer a route to materials whose building blocks assemble through designed interactions. We combine simulation and analytical theory to understand and control this assembly.

Our work on mobile DNA-coated colloids shows how entropy can stabilise floppy crystals. We have also developed linker-mediated systems that reduce the strong temperature sensitivity of specific colloidal interactions, offering a route towards experimentally addressable assembly. Related work examines superselective multivalent adsorption and the designed assembly of programmable colloidal atom–electron equivalents. Together, these studies connect molecular binding rules to the organisation of larger structures.

Related publications

• Entropy stabilizes floppy crystals of mobile DNA-coated colloids (2018)
• Linker-mediated self-assembly of mobile DNA coated colloids (2020)
• Designing superselectivity in linker-mediated multivalent nanoparticle adsorption (2024)
• Designed self-assembly of programmable colloidal atom-electron equivalents (2025)

Linker-mediated vitrimers

How does entropy control crosslinking in reprocessable polymer networks?

Vitrimers combine crosslinked polymer networks with bonds that can exchange. We study how linker concentration and temperature govern their structure and phase behaviour.

Our theoretical framework reveals a governing role for entropy in linker-mediated vitrimers. Increasing linker concentration can produce a reentrant gel–sol transition, consistent with experiment. Even at low temperature, the degree of crosslinking depends on linker concentration through competition between polymer conformational entropy and linker translational entropy. Quantitative agreement between theory and simulation provides guidance for controlling these networks; related work examines entropy-driven thermo-gelling and algorithms for simulating dynamic crosslinks.

Related publications

• Entropy-Controlled Crosslinking in Linker-Mediated Vitrimers (2020)
• Entropy driven thermo-gelling vitrimer (2022)
• A bond swap algorithm for simulating dynamically crosslinked polymers (2024)

Machine learning for rare-event sampling

Can a neural network efficiently sample transitions that almost never occur?

Rare transitions can determine a system’s behaviour while remaining difficult to observe in simulation. We develop physics-informed machine learning methods to sample their pathways efficiently.

Our normalising Flow enhanced Rare Event Sampler, FlowRES, uses unsupervised normalising-flow networks to generate non-local Monte Carlo proposals. We validated it on transition-path ensembles of equilibrium and non-equilibrium Brownian-particle systems in increasingly complex potentials. FlowRES requires neither prior training data nor predefined collective variables, accommodates multiple transition routes, and maintained sampling efficiency as events became rarer in the systems studied. Our broader machine learning work also includes analysis of anomalous diffusion in particle-tracking data.

Related publications

• Efficient Rare Event Sampling with Unsupervised Normalising Flows (2024)
• U-Net 3+ for Anomalous Diffusion Analysis enhanced with Mixture Estimates (U-AnD-ME) in particle-tracking data (2025)

Self-assembly of fibril-forming polypeptides

How do peptide folding and collective assembly shape protein fibrils?

Protein fibrils form through a hierarchy of molecular folding and assembly. We use computer simulation to understand how these coupled processes govern fibril formation.

Fibril-forming polypeptides connect fundamental questions of protein organisation with opportunities for biomaterials. In simulations of these systems, we found a strong allosteric effect linking peptide folding and fibril assembly. We also investigated competition between surface adsorption and folding. Understanding these interactions helps explain the physical mechanisms behind fibril formation and offers guidance for the experimental fabrication of protein fibres.

Related publications

• Interplay between folding and assembly of fibril-forming polypeptides (2013)
• The competition between surface adsorption and folding of fibril-forming polypeptides (2015)