Accurate prediction of the phase behaviour of crystalline materials over a wide range of temperatures and pressures.

Lead Research Organisation: University of Cambridge
Department Name: Physics


Title: Accurate prediction of the phase behaviour of crystalline materials over a wide range of temperatures and pressures including prediction of vibrational characteristics and transition pathways.

Explicit mapping/exploration of the Born-Oppenheimer surface (energy as a function of atom positions) allows the classification of the structure, stability and dynamics of crystals, which make up the vast majority of all solid materials. Minima in the Born-Oppenheimer surface correspond to structures which are stable at zero temperature and can be located using materials-discovery methods such as ab-initio random structure searching (AIRSS). However, this is not the end of the story. As well as the energy of a particular minimum, the shape of the Born-Oppenheimer surface around the minimum also contributes to the thermodynamic stability of a given structure. The shape of the surface determines the vibrational contribution to the free energy of a given structure, which is often important in determining the lowest free-energy phase (i.e the thermodynamically most favorable phase).
Whilst this contribution is present (and often important) at zero temperature as a result of quantum zero-point motion, it is of particular importance at finite temperature when thermal vibrations appear. Often, vibrations are significant enough to lead to temperature-induced phase transitions. We aim to use first-principles simulations to predict when these phase transitions will occur, and to understand how they occur. The extended shape of the surface also determines the exact distortion that the material undergoes during the transition (the transition pathway).

Novel methodology
Many approximate methods exist to determine the vibrational properties of materials. Chief amongst these is the harmonic approximation, where vibrations are treated as small perturbations to the crystal structure. This method breaks down when strong vibrations are present, which is often exacerbated at high temperatures. In order to remedy this, we directly map the Born-Oppenheimer surface using density functional theory (DFT) calculations. This allows us to directly calculate the free energy of a given structure at any temperature, even when the amplitude of vibration is large. As well as allowing us to calculate which is the
thermodynamically favorable structure, we can also explore the surface in order to determine the network of transition pathways connecting stable structures. We can also combine this vibrational information with electronic-structure calculations to determine the strength of electron-phonon (electron-vibration) coupling, a key phenomenon in the study of superconductivity. In order to do this we are developing and testing software to efficiently map, and explore, the Born-Oppenheimer surface.


10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512461/1 01/10/2017 30/09/2021
1948652 Studentship EP/R512461/1 01/10/2017 30/09/2021 Michael Hutcheon
Description Investigation into the structure of Lithium metal at low temperatures lead to the proposal of a mixed-phase structure, based on DFT calculations. Harmonic vibrations were shown to drive phase transitions with temperature and anharmonic corrections were calculated and shown to not affect the results.

The structure record-holding superconductor, LaH10, was investigated across a large pressure range (100 - 400 GPa), and a new hexagonal phase was found at high pressures, potentially explaining impurities found in cubic samples synthesised in experiment.

Machine learning approaches were used to design new hydride superconductors, with an aim of reducing the pressure required for their operation. This resulted in the proposal of a new material, Immm-RbH12, which is predicted to be a liquid-nitrogen temperature superconductor as low as 50 GPa.
Exploitation Route The structure-searching methods that we use can be readily applied to any target material property that is calculable from first principles. In particular, the work showing that machine learning can help target material properties is an important proof-of-concept.
Sectors Aerospace, Defence and Marine,Chemicals,Electronics,Energy,Manufacturing, including Industrial Biotechology