A neural network interface for DL_POLY and its application to liquid water

ABSTRACT

After a general discussion of neural networks potential energy functions and their standing within the various approaches of representing the potential energy function of a system, we describe a new interface between the open source atomistic library aenet of Artrith and Urban and the DL_POLY 4 code. As an application example, the training of a neural network for liquid water is described and the network is used in a molecular dynamics simulation. The resulting thermodynamic properties are compared with those from a reference simulation with the same SPC/E model that has been used in the training.

KEYWORDS:

• Feedforward neural network
• molecular dynamics simulation
• liquid water simulation

Acknowledgments

We thank Nongnuch Artrith and Alexander Urban for many fruitful discussions and for sharing their aenet library. The computational results presented have been achieved using the HPC infrastructure LEO of the University of Innsbruck.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This study was supported by the Austrian Science Fund (FWF) project P28979-N27. It was also supported by the Austrian Ministry of Science BMWF as part of the University Infrastructure Program of the Scientific Computing Platform LFU Innsbruck and has received funding from the Euratom Research and Training Programme 2014-2018 under Grant Agreement No. 633053. The authors were also partially supported by the Doctoral Programme on Computational Interdisciplinary Modelling; FP7 Fusion Energy Research.