Patterned shape changes in liquid crystal elastomers: when the material is the machine - John Biggins
Liquid crystal elastomers are a remarkable new class of active soft solid, often advertised as artificial muscles. This name arrises because LCEs reversibly elongate and contract, often by hundreds of percent, when they go through a solid-solid phase transition from isotropic to aligned. This transition can be tuned to occur around room temperature, or can be driven by illumination. Recently, experimentalists have learnt to 3-D print LCE structures in which this elongation can be patterned so that, at each point in the structure, it points in any desired direction. Upon heating/illumination, these 3-D structures will morph into entirely different shapes. Such shape-morphing via patterned shape-change is common in biology, where complex shapes arise via patterned growth during development and are actuated by patterns of muscular contraction to drive the key processes of life, including digestion, locomotion, and respiration, but is novel in human science and technology. In this project, we will develop a numerical platform to design and analyze LCE structures with patterned elongations, learning the basic vocabulary of effects achievable by patterning shape changes into materials. Specifically, we will use our code to design soft machines, including heart like soft pumps, snapping shells and non-inertial oscillators. We will work closely with experimentalists in th US, who will fabricate and test our designs.
Surface-tension driven deformations in soft solids - John Biggins
Molecules at a condensed phase’s surface have fewer neighbors than those in the bulk, so all materials suffer an energy proportional to their exposed area, known as surface tension. The resulting tendency to reduce area is familiar in fluids, explaining why droplets are round, taps drip, and pond skaters do not drown. We are less familiar with solids distorting to minimize their surface area, and with good reason: in conventional stiff solids the elastic energy of the such deformations always outweighs the surface energy gains. However, soft solids, such as elastomers, gels, can be distorted by their surface energy, taking on dramatically different shapes. These effects are particularly large in the small soft systems that abound in biology, and emerge from a fundamental and ubiquitous trade off between elasticity and surface tension. In this project we will build a finite element code that combines surface tension and large deformation elasticity, to predict and understand the distorted shapes that emerge from this trade off, and apply it to soft fibers, sheets, balloons, ellipsoids and platonic solids. We will then use this code to study how solid-surface tension modifies classic JKR adhesion/contact theory between a rigid sphere and a soft solid substrate.
Tensile elastic instabilities - John Biggins
The prototypical elastic instability is the Euler buckling of a slender column under compression. This instability arrises a geometric reason: buckling allows the column to return to its original contour length, and hence to save elastic energy. Soft solids, are capable of undergoing large deformations, and so are subject to a much richer geometric effects, and can undergo a correspondingly rich set of elastic instabilities, including aneurism instabilities, folding and wrinkling instabilities, and solid cavitation. These instabilities cause simple shapes to spontaneously turn into complex shapes, and have been harnessed by biology to sculpt organs during development - for example my own work has shown the brain folds via a compressive buckling instability. There is clearly an opportunity for human engineers to use these instabilities to make devices that switch shape during use. In this project we will examine the instability of the surface of a soft solid when subject to an E or B field - the solid analogue of the famous ferrofluid Rosensweig instability. The motivating idea is that a flat surface will, at sufficient field strength, become unstable and form a rough topography (spikes), allowing the manipulation of surface properties such as friction, adhesion and wettability.
Machine Learning and Drug Discovery - Gabor Csanyi
Accurate computational methods for predicting the strength of binding between a candidate drug molecule and its therapeutic target have the potential to revolutionise the drug discovery process. Our goal is to improve the accuracy of this approach by combining free energy calculations with custom-made force fields which employ machine learning techniques to faithfully reproduce the quantum mechanical potential energy surface of the drug molecule. The project will be coordinated by two supervisors with expertise in machine learning for molecules (Prof Gábor Csányi, University of Cambridge) and computer-aided drug design (Dr Daniel Cole, Newcastle University) aspects of this research. The student will work closely with drug discovery programmes at AstraZeneca (Dr Graeme Robb) with the goal of establishing these computational methods as part of the standard tool kit in the drug discovery pipeline. The project will provide highly sought-after training in the fields of computational medicinal chemistry and machine learning. As such, it will provide excellent experience for a future career in either academia or the pharmaceutical industry.
Understanding material failure at all length scales - Gabor Csanyi
It is well known that material failure is ultimately controlled by the interaction of atoms with their neighbours because that determines how defects in the material form, propagate and interact with one another. However, traditional engineering models of failure are phenomenological, and typically operate on the length scale of dislocations at which the atomic structure of matter is homogenised. This is often a fine approach, but can break down for example at high temperatures where many of usual assumptions no longer hold, and also when we consider stresses high enough that dislocations break through grain boundaries. On the other hand, detailed models of atomic structure are only available for "idealised" systems, and the accurate "chemistry based" techniques cannot be scaled up to treat extended defects. In order to create predictive models, a more joined-up approach is needed - and there is a dearth of people with the required training. In this project we intend to solve this by having two supervisors: Professors Deshpande (large length scale) and Csanyi (atomic length scale) will train and guide a student through an interdisciplinary and multi-length scale research that tackles some of the most prominent problems in fundamental materials engineering.
Machine learning quantum mechanics - Gabor Csanyi
Much of the high performance computing capacity available to academics today is spent on solving the equations of quantum mechanics (in the context of solid state physics and quantum chemistry) for a diverse set of atomic configurations, e.g. when doing high through-put materials discovery, searching "chemical composition space", or running long time trajectories using molecular dynamics of crystals, liquids, surfaces and molecules. These costly calculations can in principle be speeded up by many orders of magnitude if the potential energy of atoms is modelled using analytical potential functions, rather than obtained from the quantum mechanical equations each time. However, such empirical potentials (also known as force fields) are not particularly accurate. This forced tradeoff between accuracy and speed can be "broken" by using machine learning methods, which can be very accurate and still much faster than direct quantum mechanics, thus opening up new areas for such simulations. It is proposed that similar methods of molecular representation and function fitting that work well for approximating the energies and forces on atoms might also work well for "learning" other, much more subtle observables, such as electronic charge, polarisability, band structure, effective Hamiltonians, spin-spin interactions, etc. This project will explore a variety of physical properties of matter and how they can be quickly and efficiently calculated using machine learning techniques.