Structure prediction

A generative deep learning model of molecular structure is combined with supervised deep learning models of molecular properties to achieve high-throughput (multi-)property-driven design of organic molecules.

While stimuli-responsive metal-organic frameworks have been widely investigated, much less is understood about structural flexibility in their covalent counterparts. Here, 3D diamondoid covalent-organic frameworks are studied via dynamic free energy simulations, revealing key insight into how the nature of the building blocks and the degree of interpenetration contribute to framework flexibility.

Current crystal structure prediction methods for anticipating polymorphism are lacking in their ability to provide information on the depth of each energy minimum and the possible transition paths and energy barriers between polymorph structures. Here, a Monte Carlo threshold algorithm is applied to four polymorphic organic systems to estimate the energy barriers that separate the different polymorph structures and characterise the global structure of the energy landscapes of their molecular crystal structures.

Exploration of metastable phases of a given elemental composition is a data-intensive task. Here the authors integrate first-principles atomistic simulations with machine learning and high-performance computing to allow a rapid exploration of the metastable phases of carbon.

Developing theoretical frameworks to predict new polymorphs is highly desirable. Here the authors present an ab initio based force-field approach for crystal structure prediction offering a dramatic computational speed-up over fully ab initio schemes.

The identification of active sites in cation exchanged-zeolites is computationally highly demanding. Here the authors employ a DFT-based computational framework to identify the energetically most favorable adsorption sites in any zeolite using a systematic high-throughput approach.