sGDML: Constructing accurate and data efficient molecular force fields using machine learning

in Computer Physics Communications (2019), 240

We present an optimized implementation of the recently proposed symmetric gradient domain machine learning (sGDML) model. The sGDML model is able to faithfully reproduce global potential energy surfaces ... [more ▼]

We present an optimized implementation of the recently proposed symmetric gradient domain machine learning (sGDML) model. The sGDML model is able to faithfully reproduce global potential energy surfaces (PES) for molecules with a few dozen atoms from a limited number of user-provided reference molecular conformations and the associated atomic forces. Here, we introduce a Python software package to reconstruct and evaluate custom sGDML force fields (FFs), without requiring in-depth knowledge about the details of the model. A user-friendly command-line interface offers assistance through the complete process of model creation, in an effort to make this novel machine learning approach accessible to broad practitioners. Our paper serves as a documentation, but also includes a practical application example of how to reconstruct and use a PBE0+MBD FF for paracetamol. Finally, we show how to interface sGDML with the FF simulation engines ASE (Larsen et al., 2017) and i-PI (Kapil et al., 2019) to run numerical experiments, including structure optimization, classical and path integral molecular dynamics and nudged elastic band calculations. [less ▲]

The Flexible Rare Event Sampling Harness System (FRESHS)

in Computer Physics Communications (2014)

We present the software package FRESHS (http://www.freshs.org) for parallel simulation of rare events using sampling techniques from the ‘splitting’ family of methods. Initially, Forward Flux Sampling ... [more ▼]

We present the software package FRESHS (http://www.freshs.org) for parallel simulation of rare events using sampling techniques from the ‘splitting’ family of methods. Initially, Forward Flux Sampling (FFS) and Stochastic Process Rare Event Sampling (SPRES) have been implemented. These two methods together make rare event sampling available for both quasi-static and full non-equilibrium regimes. Our framework provides a plugin system for software implementing the underlying physics of the system of interest. At present, example plugins exist for our framework to steer the popular MD packages GROMACS, LAMMPS and ESPResSo, but due to the simple interface of our plugin system, it is also easy to attach other simulation software or self-written code. Use of our framework does not require recompilation of the simulation program. The modular structure allows the flexible implementation of further sampling methods or physics engines and creates a basis for objective comparison of different sampling algorithms. Our code is designed to make optimal use of available compute resources. System states are managed using standard database technology so as to allow checkpointing, scaling and flexible analysis. The communication within the framework uses plain TCP/IP networking and is therefore suited to high-performance parallel hardware as well as to distributed or even heterogeneous networks of inexpensive machines. For FFS we implemented an automatic interface placement that ensures optimal, nearly constant flux through the interfaces. We introduce ‘ghost’ (or ‘look-ahead’) runs that remedy the bottleneck which occurs when progressing to the next interface. FRESHS is open-source, providing a publicly available parallelized rare event sampling system. [less ▲]

Nucleation in suspensions of anisotropic colloids

in Computer Physics Communications (2005), 169(1-3), 117-121

We report Monte Carlo studies of liquid crystal nucleation in two types of anisotropic colloidal systems: hard rods and hard ellipsoids. In both cases we find that nucleation pathways differ strongly from ... [more ▼]

We report Monte Carlo studies of liquid crystal nucleation in two types of anisotropic colloidal systems: hard rods and hard ellipsoids. In both cases we find that nucleation pathways differ strongly from the pathways in systems of spherical particles. Short hard rods show an effect of self-poisoning. This part of the article is based on a previous publication [T. Schilling, D. Frenkel, Self-poisoning of crystal nuclei in hard-rod liquids, Phys. Rev. Lett. 92 (2004) 085505]. When a crystallite forms, its surfaces are covered preferentially by rods which align perpendicular to the surface. Therefore subsequent growth is stunted. Hard, almost spherical ellipsoids can be compressed to very high densities without crystallization—in contrast to hard spheres, which crystallize easily. When forced to crystallize, ellipsoids form very loosely aggregated nuclei. In both cases nucleation pathways are complex and it is therefore difficult to define an appropriate reaction coordinate. The common strategies of investigation of nucleation problems (i.e. definition of a coordinate and then sampling barrier crossing with well-known techniques) fail in these systems. [less ▲]