Advances in the understanding and modelling of the behaviour of matter have opened the way for the development of systematic methodologies for the computer-aided design of materials, products and devices, integrating decisions at the molecular scale with decisions at the bulk scale. My research group focuses on the development of fundamental modelling and optimisation tools and applications to specific problems of relevance to today''s industry. These include the selection of optimal processing materials such as solvents, the design of high-performance products such as polymers or crystals, and the design of devices such as solid-oxide fuel cells or electrolysers.
Given the impact of molecular-level information on the overall system, we refer to this class of problem as Molecular Systems Engineering; the molecular structure or the structure of the material and their interactions with the end-user/process/device are crucial in determining performance and hence in identifying best designs. The design methodologies we are developing are based on an integrated approach that requires i) predictive relationships that relate the structure of the material to its properties, ii) reliable models of the product/process/device that connect the material properties to overall system performance, iii) optimisation tools to identify the best design based on such models.
The main application areas currently addressed with this work are the pharmaceutical/agrochemical and fine chemicals industries (design of solvents for organic reactions, crystal structure prediction), and energy systems engineering (the design of solvents and processes for CO2 capture, oil and gas production and fuel cells).
The research carried out involves interdisciplinary work under the auspices of the Molecular Systems Engineering programme at Imperial and in collaboration with a number of groups, including those of Paul Taylor at Warwick University, Sarah L. Price at UCL and Alan Armstrong and Nigel Brandon at Imperial.
We are tackling the following issues:
i) predictive relationships between structure and properties: we use a range of predictive modelling tools such as group contribution techniques, advanced equations of state (SAFT), molecular mechanics and quantum mechanics. Where appropriate, we develop new predictive methods, for instance the SAFT-γ group contribution equation of state. We design efficient algorithms to deploy the predictive power of expensive computational tools such as quantum mechanics on large-scale problems, or to solve challenging property evaluation problems such as multi-component phase stability and equilibrium.
ii) reliable models of the product/process/device that connect the material properties to overall system performance: we develop models of the devices and processes of interest, such as fuel cells or separation processes (e.g., for CO2 capture), often involving partial differential algebraic equations, in order to obtain reliable predictions of performance based on the use of different materials (e.g. solvent).
iii) optimisation tools to identify the best design: most of the optimisation problems we tackle are mixed-integer due to the need to account for molecular or material structure. They are usually and nonconvex, thereby requiring global optimisation algorithms. Given the computational expense associated with many of the prediction techniques used in the models, there is also a need to focus on computational efficiency. Thus, much of our work in design is focussed on problem formulation and algorithm development. We also investigate the interplay between experiments, design and model building.