Physics-based approach to computational design of molecular solids
Abstract
Advanced technologies developed by the DoD as well as improvements in capabilities of existing technologies are often based on novel molecular materials. If the properties of such systems could be predicted computationally, the development processes might be shortened by screening the candidate materials, and new classes of materials could be suggested. The properties of molecular materials are governed by intermolecular forces, an area in which the University of Delaware (UD) group has had an impact through the development of methods for ab initio calculations of intermolecular force fields that are both accurate and applicable to large molecules. In particular, forces computed using symmetry-adapted perturbation theory (SAPT) have been used in crystal structure prediction (CSP) protocols for several energetic materials, including blind predictions. In all cases, excellent agreement with available experimental data was achieved. This work includes prime examples of a successful in silico material design. The investigations were enabled by an automated method of generating force fields which shortened times of the development process by two orders of magnitude. The New York University (NYU) group is developing enhanced sampling and structure generation and ranking tools that yield accurate results at experimentally relevant condition, leveraging the interaction models of the UD group. The UD and NYU groups now have established a track record of successful collaboration. The two main methods currently used to model materials are based either on empirical force fields or on density-functional theory (DFT) calculations performed on-the-fly. Physics-based force fields, derived using ab initio quantum mechanical methods, are more accurate than either of the aforementioned approaches and applicable to much larger systems than the latter of these. We propose here to develop a comprehensive multi-layer physics-based computational method for modeling of molecular solids that will dramatically expand the current capabilities of in-silico predictions. While ab initio interaction energies have been used so far to generate system-specific force fields, a method is proposed to create class-universal force fields, applicable to classes of similar molecules. In contrast to universal empirical force fields, no experimental data need be used in such development. Next, we propose to improve the current treatment of non-rigid monomers by a combination of flexible-monomer force fields and on-the-fly approaches. The latter will use the dispersionless density functional plus dispersion (dlDF+D) method, created by the UD group, which will be tuned for crystal structure predictions. Tight-binding and machine-learning variants of this method will allow calculations for very large unit cells. System-specific physics-based force fields can be benchmarked in CSP applications for which experimental data exists. Ultimately, these force fields can be leveraged to predict structures of notional crystals and trends in molecular solids, including polymorphism, co-crystallization with co-formers, and thermodynamic properties, providing the basis for rationally planned crystallization experiments. Thus, the proposed computational suite will also include the creation of a robust CSP pipeline that includes a structure generation and optimization module and free-energy based enhanced sampling algorithms for polymorph ranking. Machine learning techniques will be introduced to support the structure generation protocols by learning energy surfaces Òon-the-flyÓ during the structure exploration, dramatically improving the efficiency of structure generation and thereby allowing large sets of space groups and more monomers per unit cell or symmetry-independent ones to be explored in a given amount of computational time. All these methods will be applied, in collaboration with scientists from the Army Research Laboratory, to systems of current DoD interest.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 20, 2019
- Source ID
- W911NF1910117
Entities
People
- Krzysztof Szalewicz
Organizations
- Army Contracting Command
- United States Army
- University of Delaware