Density-Functional Theory Group: Research Projects

Quantum-Mechanical Simulations

Computer simulations are playing an ever-increasing role as a complement to experiment in modern physics, chemistry, materials science and biology. Quantum mechanics describes the behaviour of electrons and nuclei, and the bonding between them, that is common to all these fields. While methods based on empirically determined classical interatomic potentials may be sufficient to describe familiar situations, such assumptions cannot be relied upon when pioneering new fields or predicting the properties of new materials. In these cases the quantum-mechanical equations must be solved from first principles using only well-controlled approximations.

Our group is at the forefront of the development and application of new techniques for quantum mechanical simulations. Foremost of these has been the development, within TCM and in collaboration with Imperial College and University of Southampton, of methods for linear-scaling density-functional theory. This research has resulted in the ONETEP code, which now enables simulations of hundreds to tens of thousands of atoms to be performed with unparalleled accuracy on systems ranging from biological macromolecules to nanostructures. Current research into applications of linear-scaling DFT centres around biological physics, strongly-correlated systems, and solvation models. These advances promise to bring the power of quantum-mechanical simulations to bear on systems of an unprecedented scale, for use in applications as diverse as the design of new drug molecules to specifically target particular diseases to the characterisation of nanomaterials for photovoltaic solar cells.

We also collaborate with the Department of Engineering in the development of hybrid modelling schemes. In these approaches, accurate quantum simulations are embedded within a fast empirical scheme dynamically, where the extent of the quantum-mechanical region is determined on the fly.

¹Imperial College London, ²University of Southampton, ³Department of Engineering

Sponsors: EPSRC, Royal Society

Biological Physics

Computational methods that are capable of elucidating features of molecular recognition, binding affinities and structural stability are likely to drive experimental approaches to studying macromolecules. Such methods may aid determination of structure-activity relationships by revealing the behaviour of systems derived from experimental structures and, more excitingly, systems unamenable to experimental structure determination. We are using approaches that combine high accuracy, linear-scaling DFT methods with long time scale classical molecular dynamics simulations to investigate the properties of macromolecules of genuine biological interest. Examples of our work include the determination of protein-ligand binding affinities at a large receptor interface and the stability of various structures of G-quadruplex DNA (pictured below).

Despite their many successful applications, conventional molecular dynamics simulations are generally limited to submicrosecond time scales and to systems of a few hundred thousand atoms. This makes the exploration of conformational changes of large systems over high kinetic barriers infeasible. We are exploring the use of "coarse-grained" simulations in the study of systems such as ligand-gated ion channels (pictured right) and assembly of proteins on material surfaces.

¹University of Southampton, ²University of Bremen

Sponsor: EPSRC

Examples of Current work of Ph.D. students

Linear-scaling methods for calculating excited-state properties of strongly-correlated nanoclusters and organometallic molecules

This work is focused on developing broadly applicable techniques for computing both ground-state properties and excitation spectra in large systems which are challenging for conventional DFT methods. A rich and promising area of application of DFT, as well as providing valuable insight into experimental results, is the ab initio design of functional organometallic biomolecules tailored for particular optical or magnetic properties. In order to tackle such problems, linear-scaling algorithms for treating many-body correlation effects due to localised electrons on transition-metal ions, and for the accurate calculation of excited-state properties, are required.

Linear scaling of computational effort with system-size, such as that afforded by the ONETEP code, allows for important effects due to substrates and solvent media surrounding an optically-active or catalytic site to be explicitly included in calculations. DFT+U is an efficacious method for improving the description of correlation effects which are traditionally problematic for DFT, those associated with the localised electrons on transition-metal ions which are crucial to the function of many systems. Time-Dependent Density Functional Theory (TDDFT) is a rigorous formulation for treating excited-state properties such as optical-absorption, optical conductivity, dichroism etc. within DFT. This effort entails developing novel linear-scaling DFT+U and TDDFT functionality for ONETEP and applying these methods to selected nanoclusters and technological biomolecules.

¹Imperial College London

Implicit Solvent Models for Electronic Structure Calculations

Implicit solvation is an approach to simulating solvated systems by replacing the complex arrangement of molecules comprising the solvent with a continuous dielectric medium that has the electrostatic properties of the bulk solvent. This introduction requires a different approach to determining the electrostatic contributions to a density functional theory calculation. In exchange for this added complexity, we achieve a realistic representation of the electrostatic environment of solvated molecules and thus can perform ab initio studies of biomolecular systems with greater accuracy.

¹Imperial College London

First Principles Nuclear Magnetic Resonance

Experimental NMR techniques have established themselves over the past four decades as one of the main tools for studying structural information. In these experiments the absorption of an oscillating magnetic field by the atoms is measured. This absorption has a maximum at a characteristic resonant frequency. The resonance of each atom can be measured (for an example of a typical experimentally measured spectrum see below) and structural information can be worked out from these measurements. However, by using NMR experiments alone it is difficult to work out the detailed structure of a system under study. By combining NMR experiments with first principles calculations we are able to obtain a detailed structure of a given system and deduce its function.

First-principles approaches rely only on the most fundamental theory that describes physics at the atomic level. By solving these quantum mechanical equations we can calculate various NMR parameters for model structures. Hence, these calculations provide us with the link between experimental data and the underlying structure.

TCM has been involved in the recent development of the GIPAW approach implemented in the CASTEP code for calculating NMR parameters. The combination of our theoretical with experimental approaches makes NMR an extremely powerful tool applicable to various fields of research tackling issues from radioactive waste storage to improving healthcare. Among others, we have used this approach to study small biological molecules such as sugars, amino acids and nucleosides (below), induced currents in carbon nanotubes with applications in drug delivery (right), highly disordered glasses with application in optical data transfer, microelectronics and radioactive waste storage.

¹University of Oxford