Supercomputer modelling of advanced materials

The potential energy landscape provides a conceptual and computational framework for investigating structure, dynamics and thermodynamics in condensed matter and molecular science. This talk will summarise new approaches for global optimisation, calculating thermodynamic properties in systems exhibiting broken ergodicity, and rare event dynamics. Applications will be presented that exploit generalised basin-hopping for global optimisation in continuous and discrete metric spaces, with examples ranging from biomolecules tomachine learning and quantum computing. Design principles for self-assembly of mesoscopic structures extend the single funnel paradigm to multifunnel landscapes, with the potential for encoding multifunctional materials.

Selected Publications:

DJ Wales, Annu Rev Phys Chem (2018) 69, 401

JA Joseph, K Röder, D Chakraborty, RG Mantell, and DJ Wales, Chem Comm 53, 6974, 2017.

DJ Wales, "Energy Landscapes", Cambridge University Press, Cambridge, 2003

New chemical materials and compounds serve as the foundation for the modern technology of our civilisation, where these materials should be environmentally friendly, with stable and controllable properties for a multitude of applications, and at the same time energy efficient in their synthesis. Traditionally, such materials are generated in bulk and then machined down to whatever size is needed. But such materials are also often grown in a bottom-up approach on a substrate, and in the extreme we are dealing with and aiming for low-dimensional systems with bespoke properties, where the stability of such materials often becomes an issue of concern. Among these systems, we find monolayers on surfaces or inside layered compounds, ultrathin films, nanotubes and nanowires, just to name a few, which are becoming increasingly important from a technological point of view. The ability to predict such kinetically stable and/or thermodynamically (meta)stable nanomaterials, followed by a computation of their properties and evaluation of their stability, is clearly of great value in their design and synthesis.[1] For the past three decades, the structure prediction of three-dimensional bulk crystalline compounds and their modifications[1,2] on the one hand, and of single atom clusters[3] and (bio)molecules[4] on the other hand, has shown great progress, and the computational approaches used are expected to be also applicable to low-dimensional systems.[5,6] In this presentation, Professor Schön will discuss the methodological features specific to the prediction of the latter systems,[7] together with examples of structure prediction for one-dimensional (eg, atom chains),[7] quasi-one-dimensional (eg, nanotubes),[7] two-dimensional (eg, monolayers),[5,8] quasi-two-dimensional (eg, layer-like building blocks for layered compounds),[9] and composite (eg, multi-molecule patterns on substrates)[10] systems.

[1] JC Schön, M Jansen, Angew. Chem Int Ed, 35:1286 (1996)
[2] SM Woodley, CRA Catlow, Nature Mater, 7:937 (2008)
[3] DJ Wales, H Scheraga, Science, 285:1368 (1999)
[4] GM Day et al, Acta Cryst B, 61:511 (2005)
[5] JC Schön, Process Appl Ceram, 9:157 (2015)
[6] SM Woodley, GM Day, CRA Catlow, Phil Trans Royal  Soc A, 378:20190600 (2020)
[7] JC Schön, in: Energy landscapes of nanoscale systems, Ed: DJ Wales, chapter 12 (Elsevier, Amsterdam, 2022)
[8] R Gutzler, JC Schön, Z Anorg Allg Chem, 643:1368 (2017)
[9] A Mahmoodabadi, M Modarresi, JC Schön, in preparation
[10] S Abb et al, Angew Chem Int Ed, 58:8336 (2019)

Controlling the formation of specific polymorphs during crystallisation is of great interest in the design of materials with specific properties. However, the initial stages of nucleation and growth during solidification are still not fully understood, even for seemingly simple systems such as unary metals. Using transition path sampling, the researchers have studied the microscopic mechanisms during homogeneous nucleation in nickel. Their analysis of the path ensemble reveals that the emergence of the crystalline phase is preceded by the formation of pre-structured regions in the liquid that act as precursors. Extending their study to heterogeneous nucleation, they likewise observe a precursor-mediated crystallisation mechanism. Small seeds impact the structural characteristics of the liquid and induce the formation of precursors with specific structural hallmarks. The nucleating ability and polymorph selectivity of these templates is, therefore, not simply given by lattice mismatch and translational order but strongly linked to the seed’s ability to promote the formation of suitable precursors in the liquid. Furthermore, they have analysed the dynamical heterogeneity in the liquid which is revealed to play a key role in the nucleation mechanism. Crystallisation occurs preferentially in regions of low mobility in the supercooled liquid. These low mobility regions form before and spatially overlap with the regions of structural precursors, revealing a clear link between dynamical and structural heterogeneity in the liquid. The relationship between structural and dynamical heterogeneity in the liquid and the nucleation mechanism appears to be key in controlling essential parameters during crystallisation, including nucleation rates and polymorph selectivity. 

The delivery of the next generation of medical treatments such as regenerative medicine hinges on our ability to store biological material. Cryopreservation - that is, the process of freezing biological material whilst retaining its function - is our best bet to achieve that, but the state-of-the-art suffers from a number of crippling limitations. One such limitation is the rather scanty portfolio of cryoprotectants available to us: these compounds are added into the mix so as to control the formation of ice in biological matter and thus mitigate the many detrimental effects caused to cells and tissues by the growth of ice crystals. We need more effective, safer, less toxic, cheaper cryoprotectants - and yet, in the last few decades we have failed to design any valid alternative to the very few compounds (glycerol being a prominent example) that we have more or less serendipitously discovered in the by now distant past. This is because we lack the molecular-level understanding of how exactly cryoprotectants work. In this talk,  Dr Sosso will present some recent advances specific to the mechanism of ice re-crystallisation inhibition agents, ie compounds that limit the growth of ice crystals - one of the many aspects that a cryoprotectant can act upon. In particular, Dr Sosso will discuss the picture emerging from our findings in terms of the ice re-crystallisation inhibition activity of selected polymers, peptides and surprisingly small molecules as well. Whilst the ultimate goal of achieving the rational design of novel cryoprotectants might still be beyond our grasp, molecular simulations are playing an important part in getting us all there - and, the further we progress, the greater the computational demands that we will need to meet. 

Ion binding at surfaces from solution is an important process in a variety of contexts from geochemistry through to the crystallisation of materials. Therefore, knowledge of the thermodynamics of such events is vital for predicting how systems will grow or dissolve under a set of conditions. With the advent of supercomputers and enhanced sampling methods, the determination of free energies for such adsorption processes is becoming increasingly possible, even for water where solvent exchange rates can be slow on the molecular dynamics timescale. Although numerous examples of solid-solution ion transfer can already be found in the literature, especially based on pathway-driven techniques, such as metadynamics or umbrella sampling, this presentation will examine the question of whether such information is actually meaningful or useful in the form often given? To explore the above question, this talk will focus on the widely studied example of NaCl growth from aqueous solution. Specifically, Professor Gale will consider the free energies of the kink sites relative to aqueous solution, as the correctness of the thermodynamics can be readily checked via the bulk solubility. By comparing the free energies of these sites as determined by a range of approaches, including pathway or alchemical techniques, it will be shown that wildly varying answers can be arrived at. However, after careful correction for all relevant simulation factors, such as system size and state, it will be shown that consistent and valid results can be reached by both alchemy and pathway-based methods.