Program

For many years a discrepancy between the results of experiment and computer simulation has been observed in the melting of noble gas solids under extreme pressure. This talk presents an approach to such simulations based on the use of the isothermal, iso-stress ensemble, which generalises the isothermal isobaric ensemble (NPT) to include a constant tensorial stress. A discussion is given of how this generalisation is achieved and implemented in a parallel tempered Monte Carlo calculation, as well as the observed simulation results at a range of pressures and stresses.

Development of sustainable and multifunctional nanomaterials is essential for advancing environmentally benign technologies in sensing and biomedical applications. Herein, hydrothermal-based one-pot synthesis of carbon dots (GH-CDs) using glucono-d-lactone (GDL) and L-histidine is reported. The GH-CDs exhibit cyan-blue fluorescence and excellent aqueous dispersibility with abundant surface functionalities, as confirmed using UV-vis, fluorescence spectroscopy, HR-TEM, SAED, pXRD, XPS, FT-IR, TGA, and zeta potential measurements. These natural-source-derived CDs exhibited selective detection of Fe3+ and Ce4+ ions mainly due to static interaction and surface complexation, which was confirmed using fluorescence quenching mechanistic analysis. Specifically, GH-CDs demonstrated excellent LOD values of 0.0514 and 0.034 μM for Fe3+ and Ce4+ ions, respectively. The real-time validation of this selective metal ion detection was performed in real water samples, which confirmed the operational robustness and interference resistance of GH-CDs. Additionally, the GH-CDs demonstrated exceptionally high specificity for tryptophan compared to all other natural amino acids. Further, the multifunctional nanoprobe analysis of GH-CDs was extended for evaluating their antioxidant activity using a DPPH assay. This analysis revealed a concentration-dependent radical scavenging efficiency of GH-CDs with 73.2% inhibition at 500 μg/mL and an IC50 value of 86.6 μg/mL. The cytotoxicity analysis followed by cellular imaging studies using the normal HEK cell line further demonstrated the high biocompatibility and fluorescence imaging potential of GH-CDs. The reported versatile applications of GH-CDs make them a sustainable, cost-effective, and efficient nanoprobe material for future use.

An enduring question within condensed matter is the nature of the transition from a high temperature liquid to their ground state(s). While first-order freezing is well understood, glassy transitions are difficult due to complexity and degeneracy of the ground state.

Using a 2D lattice model by Ronceray & Harrowell, we can tune the ordering and complexity of the ground state by directly controlling the favoured local structures. With its wide range of ground states comes a corresponding multiplicity of transitions including simple first-order crystallization, multi-step transitions, and amorphous transitions that show no signs of a thermodynamic discontinuity. This talk will discuss the evolution of structural selectivity over the transition and preceding liquid, and its relation to the eventual ground state.

The origin of stick-slip flow has been attributed to various possible processes such as interlayer slip, interfacial slip, and strain-induced melting based on experiments and molecular dynamics simulation. The phase field model adopted in this study was originally developed to simulate melting induced stick-slip. We extend this model by including interfacial slip as a possible mechanism for stick-slip flow. Both simple Navier wall slip and a yield stress wall slip mechanism are incorporated into the phase field model. With the implementation of interfacial slip, we determined parameters responsible for wall slip induced stick-slip and a state diagram that illustrates three different types of stick-slip behaviors. We find that both wall slip and shear induced melting are viable mechanisms for stick-slip flow, but the parameter range where wall slip induced stick-slip flow is very narrow. At a fixed value of the Navier slip length above a certain threshold, as the critical yield stress for wall slip is increased from zero, we see a transition from simple slip flow to wall yield stress induced stick-slip flow, to melting-induced stick-slip flow.

In molecular dynamics simulations, using the microscopic expression of transport equations is essential for understanding a wide variety of atomistic transport phenomena. Our recent work(Phys. Rev. E 105, 034803, 2022) developed a method of transforming physical variables between the Eulerian and Lagrangian viewpoints within the classical molecular dynamics framework. In this talk, we focus on thermal transport phenomena at solid-liquid interfaces using simple LJ systems, and explore transport quantities based on the transformation between the two viewpoints. The method we developed will expand the possibilities of the microscopic expression of transport equations, providing new insights into transport phenomena at the atomic scale.

The crystal-melt interfacial free energy is a key quantity in theories of crystal morphology and nucleation. The fact that this free energy can be substantial in the absence of significant density difference is puzzling. Here we demonstrate that the interfacial free energy is a consequence of the uncompensated loss of entropy in the intergaial region.

Studying the dynamics of active suspensions has become a widespread topic in the fields of soft matter and biophysics. Typically, they consist of self-propelling agents, such as certain types of bacteria, in a carrier liquid, mostly water-based. Investigations of the analogue for systems involving elasticity has only rather recently caught substantially increasing interest. The topic is important, as many biological substances show pronounced viscoelastic behaviour.

We consider thin, active, viscoelastic or elastic films on a substrate. Our considerations are general, but the most immediate motivation in our mind are biofilms. For this purpose, we developed a theoretical continuum description that can be continuously tuned from viscous, via viscoelastic, to elastic systems by adjusting just one parameter.

As a result, we find that elasticity supports local circling of the active units, instead of directed collective migration. In a transitional regime between these two dynamic states, traveling stripe-like patterns emerge. Moreover, the circling motion becomes imperfect in larger systems, which leads to desynchronisation.

Finally, our study provides a basis to describe previous experimental observations on oscillatory motion in biofilms. We are currently working on an extension to discretised, particle-based descriptions.

Polymer informatics is a rapidly advancing field, yet its progress is constrained by the scarcity and inconsistency of experimental datasets. Polymer properties are strongly influenced by synthesis routes, processing history, and measurement techniques, which makes existing data often unsuitable for direct use in machine learning (ML) models. High-throughput molecular dynamics (MD) simulations are a promising strategy to overcome this limitation by generating large-scale, consistent datasets under standardized conditions. These datasets not only minimize noise arising from synthesis or measurement variations but also provide atomistic level resolution, enabling systematic exploration of structure-property relationships. However, a critical challenge with MD simulations is generating accurate parameters of polymers to ensure that the simulations yield physically realistic data, as well as in the generation of sufficiently diverse datasets to enable robust ML training. In this study, we present a workflow for integrating MD simulations with polymer informatics, using poly(N-isopropylacrylamide) (PNIPAM) as a model system. We demonstrate how accurate parameter sets were employed to reproduce the thermoresponsive behavior of PNIPAM and how large, standardized datasets were generated to address the requirements of ML applications. Finally, we discuss how we extracted key descriptors from MD trajectories that capture relevant structural features of the polymer linking those to their macroscopic behaviour. This work highlights how MD driven datasets can address the limitations associated with experimental data and accelerate the development of predictive polymer informatics tools.

Janus colloids exhibit complex dynamical behaviour that we can steer between different self-assembled structures by adjusting features such as interaction strength, temperature, or patch geometry, making them attractive for applications across a range of fields and in the design of novel materials. We conduct molecular dynamics simulations on a system of three triblock Janus particles in a fluid to analyse the tunability of their dynamics when adjusting geometric (patch size) and chemical (hydrophobic interaction strength) parameters. We create low-dimensional representations by combining atomistic descriptors with dimension reduction techniques to observe and classify the different assemblies this system can access, and estimate the time each simulation spends visiting a given state. Using metadynamics to construct free energy surfaces, we can map out pathways the system explores between local and global minima as we tune the physical and chemical characteristics. We find we can use the tuneability of these features to bias the favourability of transition pathways between competing minima on the free energy surface. Finally, we are developing a machine learning model that, once trained, will allow the fluid degrees of freedom to be integrated out of the simulation entirely, replaced by a coarse-grained potential paired with suitable dynamics. The first two methods above can then be applied to the trajectories using the new coarse-grained potential to compare the dynamics of the system to the original molecular dynamics simulations that include the fluid explicitly.

Passive membrane permeation is an important consideration in drug development. However, permeability coefficients obtained from atomistic molecular dynamics (MD) simulations are much too costly for routine use in the drug discovery pipeline. In this work, we demonstrate that fatty alcohols such as dodecanol can mimic phospholipid (POPC) bilayers to accelerate computational prediction of passive membrane permeability (by about 30-fold) using atomic-resolution MD simulations. Comparison of the two membrane models indicates that key features of their transmembrane free energy profiles are very similar (typically within 1 kcal mol-1) for a wide range of hydrophilic, hydrophobic, amphiphilic and charged solutes with molecular weights up to 400 Da. The strong correlation between permeability coefficients (logPm) and the transmembrane permeation barrier was exploited to develop a regression model that further reduced computational cost by about 300-fold. This regression model uses the dodecanol transmembrane barrier as input and is demonstrated to efficiently recover phospholipid MD-predicted logPm values with an average prediction error of approximately 0.4 log units across 41 diverse solutes. Our findings suggest complementary strategies to accelerate computational prediction of phospholipid membrane permeability using atomic-resolution MD simulations.

Repulsive power-law systems turn out to be unusual on both sides of the glass transition. I will begin in equilibrium, where the interaction range fundamentally reshapes the thermodynamics: long-ranged repulsions inherently suppress long-wavelength density fluctuations, with a sharp crossover separating regimes of qualitatively different behavior. I will then follow these systems out of equilibrium — upon cooling, they vitrify into Wigner glasses that retain, and even enhance, the hyperuniformity of the parent liquid, bearing unmistakable structural signatures in stark contrast to ordinary glasses.

The focus of the talk, however, will be on their mechanics. Under shear, disordered solids fail when localized plastic rearrangements organize into collective flow — and I will show that what controls this transition is not how much softness a solid contains, but how that softness is organized in space. Suppressed density fluctuations fragment the soft regions into disconnected domains, and remarkably, the connectivity of this landscape anticipates the eventual yield strain: local plasticity persists, but collective failure is held off, and the yield strain more than doubles. The spatial connectivity of softness thus links large-scale structure to macroscopic failure — and points toward crystal-like mechanical stability without crystallization.

We investigate phase transitions in argon under high pressure using parallel-tempering Monte Carlo simulations. Atomic configurations are sampled across a broad range of thermodynamic conditions and subsequently analysed to identify structural transformations. To characterize these changes, we combine common-neighbour analysis with quantitative similarity metrics, enabling the automated classification of distinct phases. In particular, we examine transitions between face-centred cubic (fcc) and hexagonal close-packed (hcp) structures, as well as the emergence of Barlow packings across a range of pressures in the GPa regime. This approach provides a systematic means of tracking phase evolution and assessing the effectiveness of replica exchange in exploring the configurational landscape. The resulting structural classifications offer deeper insight into phase-transition pathways, the stability of competing crystal structures, and the ergodicity of the simulations.

Active nematics are a widely-employed continuum viscoelastic fluid model for the dynamics of some biological materials, including tissue monolayers. They can be compared with discrete mechanical 'vertex models' that capture the cells and their interactions directly. I will describe some preliminary work on using ideas from topology, especially Morse theory, to make connections between the two kinds of models, and to capture the structural changes that underpin the transition from 'arrested' states to dynamical turbulence in the active nematic.

Classifying the morphology of self-assembled nanostructures in molecular dynamics simulations typically relies on geometric criteria such as aspect ratios, moments of inertia, and contact cutoffs. However, these metrics have clear structural blind spots: they cannot distinguish hollow tubes from solid fibres, fail to separate vesicles from spherical micelles due to the absence of cavity checks, and become unreliable under thermal noise.

To address these limitations, we apply topological data analysis to construct coordinate-free representations of supramolecular assemblies. Using alpha-complex filtrations on point clouds from coarse-grained molecular dynamics simulations, we extract persistent homology across dimensions 0, 1, and 2, capturing connected components, tunnels, and internal cavities. We vectorise these scale-dependent Betti numbers into Euler characteristic curves and combine them with coordinate covariance eigenvalues. When trained on benchmark geometries, a support vector classifier reliably identifies fibres, tubes, sheets, vesicles, and spheres across concentration series and noise regimes where standard geometric rules collapse. We further outline how these invariant topological features provide compact representations to evaluate and guide candidate screening in peptide self-assembly.

Methods based in response theory, such as the Green-Kubo relations and the transient time correlation function (TTCF) formalism, have proven their utility in molecular dynamics simulation. However, some difficulties remain in treating time-dependent dynamics and perturbations to nonequilibrium states. This talk discusses these difficulties and current progress towards resolving them.

Metallic glass nanoparticles are emerging as highly promising materials for diverse applications in biomedical sensing and catalysis. While bulk metallic glasses typically rely on multicomponent complexity to achieve vitrification, the structural criteria governing glass-forming ability (GFA) at the nanoscale remain poorly understood. To address this, we investigate the vitrification of Au-based alloy nanoparticles through molecular dynamics simulations. Au-Cu systems were simulated across the entire compositional spectrum (0-100%). Our results reveal that alloys containing 30-80% Cu exhibit strong glass formation, whereas compositions approaching pure Au or Cu (85-95%) demonstrate high crystalline nature. We quantify GFA from the post-quench crystalline fraction and the abundance of perfect and distorted icosahedral motifs. The compositions with highest GFA maintain a negligible crystalline fraction across various quench-rates and elevated icosahedral content. This demonstrates that the binary Au–Cu system is a strong candidate for nanoglass applications and provides quantitative design rules for noble-metal glassy nanoparticles.

Neuromorphic computing takes inspiration from the neural network structure of the brain to create physical computing system structures that rely on the intrinsic properties of materials to perform complex logic such as pattern recognition in a more native way. Among the most important properties that create neuromorphic behaviour is a change in electrical behaviour dependent on the ‘historic state’ of the material which simulates ‘memory’ in a processing device and allows for the avoidance of the Von Neumann bottleneck which hinders the progress of traditional Von Neumann architecture and has promise to reduce power consumption for complex calculations [1].

Many devices across different classes of materials have been constructed which experimentally verify this memristive behaviour [2, 3, 6], however the physical mechanisms underpinning this behaviour is not well understood or transferable between systems. Among these devices of particular interest are reservoir computing systems which self assemble, avoiding the large cost and difficulty of nanoscale precise manufacturing. Percolating systems of nanoparticles [5, 2] and carbon nanotube networks [6] are two such frameworks that macroscopically exhibit continuous changes in conductance in the low voltage regime and atomic atomic filament switching spiking in the high voltage regime. In order to stimulate more continuous conductance change memristance alongside atomic filament switching, other classes of memristor materials are added to these frameworks. One such example is ’Tripak’, a small organic molecule which has been found to experimentally change between 6 oxidation states with voltage [4].

We will present density functional theory results that give insight into the charge transport properties and hence memristive mechanism of the molecule Tripak within individual metallic junctions. These results will show how Tripak modulates the electronic environment within individual junctions to feed into a multiscale model of the aggregate self assembled networks.

[1] Kejie Guan et al. “Atomic Nb-doping of WS2 for high-performance synaptic transistors in neuromorphic computing”. In: Microsystems amp; Nanoengineering 10.1 (Sept. 2024). issn: 2055-7434. doi: 10.1038/ s41378-024-00779-1. url: http://dx.doi.org/10.1038/s41378-024-00779-1.

[2] Joshua B. Mallinson et al. “Experimental Demonstration of Reservoir Computing with Self-Assembled Percolating Networks of Nanoparticles”. In: Advanced Materials 36.29 (May 2024). issn: 1521-4095. doi: 10.1002/adma.202402319. url: http://dx.doi.org/10.1002/adma.202402319.

[3] Matthew J. Marinella and A. Alec Talin. “Molecular memristors offer a path to ultra-efficient computing”. In: Nature 597.7874 (Sept. 2021), pp. 36–37. issn: 1476-4687. doi:10.1038/d41586-021-02323-x. url: http://dx.doi.org/10.1038/d41586-021-02323-x.

[4] Pawel Pakulski et al. “A multifunctional pseudo-[6]oxocarbon molecule innate to six accessible oxidation states”. In: Chem 10.3 (Mar. 2024), pp. 971–997. issn: 2451-9294. doi: 10.1016/j.chempr.2023.12.024. url: http://dx.doi.org/10.1016/j.chempr.2023.12.024.

[5] Sofie J. Studholme and Simon A. Brown. “Integer Factorization with Stochastic Spiking in Percolating Networks of Nanoparticles”. In: ACS Nano 18.41 (Oct. 2024), pp. 28060–28069.issn: 1936-086X. doi:10.1021/acsnano.4c07200. url:http://dx.doi.org/10.1021/acsnano.4c07200.

[6] H Tanaka et al. “In-materio computing in random networks of carbon nanotubes complexed with chemically dynamic molecules: a review”. In: Neuromorphic Computing and Engineering 2.2 (May 2022),p. 022002. issn: 2634-4386. doi: 10.1088/2634- 4386/ac676a. url: http://dx.doi.org/10.1088/2634-4386/ac676a.

This study evaluates whether Patchy Brownian Cluster Dynamics [1] can be used to model the hexameric globulin plant proteins and their aggregation. Hexameric structure consists of six subunits, where subunits first assemble into trimers that subsequently combine to form the globulin protein [2]. Each protein subunit is modelled as a spherical colloidal particle with three patches, representing disulphide bonds and hydrophobic interactions. Initially, two patch simulations are performed to optimise parameters for disulphide bonds, followed by three-patch simulations to optimise the overall model. For two-patch simulations, at patch angle 20°, varying the inter-patch angle from 150° to 90° results in a transition from chain-like to trimeric structures. However, for three patch simulations, we observed few hexamers, suggesting the need to further tune the interaction range and strength. The current observation of hexamers along with smaller aggregates indicates that a patchy colloidal approach can be used to describe multi-subunit plant proteins. Future work will optimise the model and seek to use it to predict the structures arising during the gelation of plant proteins under different experimental conditions.

[1] Prabhu, et al., J. Chem. Phys. 141, 024904 (2014).

[2] Sharan et al., Compr. Rev. Food Sci. Food Saf. 20, 401-428 (2021).

Liquid metals are a fascinating material class that have recently garnered significant scientific attention, especially low temperature liquid metals, typically defined as melting below 300 ˙C. Such metals and alloys have a low energy cost to access their unique and highly tunable properties for practical applications that include catalysis, microfluidics, flexible electronics, energy storage, and much more. Gallium is a standout component of liquid metal alloys owing to its naturally low melting temperature of 30 ˙C. Recently, a eutectic alloy of bismuth doped gallium has exhibited intricate regular surface patterning (including dopant segregation) upon freezing. Towards better understanding this phenomenon, and others such as gallium’s tendency to supercool, this presentation will canvas recent simulations of liquid gallium at the interface with its own ambient solid phase, alpha-Ga. The simulations described utilise machine learned force fields, trained with active learning on data from density functional theory calculations, to enable much larger scale simulations. A heavy emphasis is placed on data visualisation, including for determining if a machine learned force field is sufficiently trained, and understanding the liquid dynamics of the system under different conditions.

Molecular simulations of tribology commonly employ nonequilibrium molecular dynamics (NEMD), in which a structured lubricant is confined between moving walls and its tribological and rheological properties are measured. However, conventional NEMD cannot access the realistic strain rates encountered experimentally. We have recently demonstrated that nonlinear response theory, implemented through the transient-time correlation function (TTCF) technique, can bridge this gap, although these studies were largely limited to simple Lennard-Jones fluids as proof-of-concept systems.

Here, we demonstrate the application of TTCF to NEMD simulations at experimentally accessible strain rates and examine two tribological systems: squalane confined between sheared iron oxide walls modified with organic surfactants, and water confined within sheared graphene nanochannels. At realistic strain rates, with the lowest of order 10⁵ s⁻¹, standard NEMD produces signals too noisy to accurately measure any property of interest. In contrast, TTCF yields exceptionally clean signals and high statistical accuracy for all measurable quantities. We determine slip velocity, slip length, shear stress and friction coefficient for both systems. This is the first demonstration of NEMD measurements of these properties at experimentally accessible strain rates for these systems. For water confined by graphene, we also compare our results with existing simulation and experimental measurements.

Proteins range from well-folded structures to floppy intrinsically disordered chains, thus modelling them appropriately requires methods that can meaningfully capture the different scales of dynamics they exhibit. All-atom molecular dynamics simulations are a gold standard for obtaining atomistic detail in molecular systems, but sampling intrinsically disordered proteins/regions (IDPs/IDRs) becomes increasingly difficult and computationally expensive. Coarse-grained (CG) Langevin simulations are therefore often combined with single-molecule experiments, to obtain experimentally validated conformational ensembles. Here, this approach was employed to characterise the conformational ensemble of the full-length pioneer transcription factor FoxA1 in its unbound, and DNA-bound states. The disordered tails of FoxA1 were found to expand when DNA bound, consistent with a ‘fly-casting’ mechanism for recruiting binding partners. When comparing FoxA1 bound to two different DNA oligonucleotides, FoxA1 showed increased remodelling of contacts between the DNA-binding domain (DBD) and neighbouring IDRs when bound to a less structurally stable oligonucleotide, suggesting adaptation to a more dynamic ligand. Atomistic simulations then provided molecular detail on the structured DBD, revealing that a breast cancer-linked mutation reduces overall DBD dynamics, restricting conformational changes that the wild-type undergoes upon binding DNA.