One Small CASTEP for Archer, One Giant Leap for First-Principle Prediction of Crystal Form

New project developing quantum mechanics code to bridge the gap between organic crystal modelling and observation

A new collaboration catalyzed by the Advanced Digital Design of Pharmaceutical Therapeutics (ADDoPT) project aims to deliver a step change in accuracy for first-principle predictive models of organic and biochemical systems, offering the prospect of greatly reduced experimental workloads when designing and developing crystal structures important to industry including active pharmaceutical ingredients (APIs).

Researchers at the Science and Technology Facilities Council’s (STFC) Rutherford Appleton Laboratory and Hartree Centre have joined forces with co-investigators from the Universities of Leeds, York and Dublin in a new 14-month code development project backed by the UK ARCHER supercomputing service to enhance the long-range dispersion interaction functionality of CASTEP, a widely-used code for calculating the properties of materials from first principles.

CASTEP is based upon Density Functional Theory (DFT), used successfully to explore many materials and solid-state condensed matter systems, but conventional DFT neglects long-range van der Waals or “dispersion” interactions. These forces are vital to a fundamental understanding of, and exact prediction of properties for layered materials such as graphene, and large organic molecular crystals prevalent in the pharmaceutical industry. The new project will address the current disparity between DFT-based predictions and observation identified collaboratively by Hartree Centre and Leeds researchers in ADDoPT, by optimizing and extending the current state-of-the-art in long-range dispersion algorithms – the many-body dispersion (MBD) calculation scheme.

According to STFC Hartree Centre Project Scientist Dr. Dawn Geatches,

“While directional trends are useful to theoreticians, near-absolute predictions from first-principles models make experimentalists sit up.”

The research team will benchmark calculated properties such as lattice energies and elastic constants of molecular crystals against experimental sublimation enthalpies and mechanical property values, important in predicting solubility and hardness, aiming to achieve chemical accuracy of calculated energies to within 4 kJ/mol.

ADDoPT Project Manager, Dr. Sean Bermingham comments,

“Ultimately for ADDoPT, accurate, fundamental calculations lead to a robust pipeline in drug dosage form design and to an ideal synergy between experiment and theory.”

Newly appointed post-doctoral researcher Peter Byrne, working under Principal Investigator Dominik Jochym has been funded for 8 months by ARCHER eCSE support, leveraging a further 6 months of STFC’s effort under ADDoPT. Hopes are high for an improvement of more than an order of magnitude in time-to-science on ARCHER, with desktop and PC cluster CASTEP users also set to benefit from the more efficient MBD algorithm.

For further information please contact Dawn Geatches at the STFC Hartree Centre.