Our research focuses on theory development and predictive modeling on materials behaviors at complex environments. This site provides some scripts related to the modeling algorithms that I have contributed to, and I am happy to share those with other researchers so that we could together build a reproducible research environment in our community.
1. Exploring potential energy landscape (PEL) to reveal the unfaulting mechanism of irradiated defects
(reference: Phys. Rev. B 81 (2010) 104102)
A self-written code (in Fortran) to explore the defect's PEL by adding a series of Gaussian penalty functions and then relax the system. The algorithm is mimicking the process of filling water into a basin: by adding a number of water droplets, the system can be gradually lifted and simultaneously find the neighboring minimum once the water level exceeds the lowest saddle point. Below is an example showing how such algorithm allows one to probe the detailed unfaulting mechanisms of a trapped defect in bcc-iron.
2. Probing the dislocation-obstacle interaction mechanisms at low strain rates
(reference: Proc. Natl. Acad. Sci. USA 110 (2013) 17756)
To extend the timescales to realistic strain rates in experiments, we combined the Autonomous Basin Climbing (ABC) method with transition state theory (TST) in a dynamics pattern to investigate the dislocation-obstacle interactions. As illustrated in the figure, the consecutive iterations can be described below:
(i) Under a given strain condition (including the initial state with no strain), employ ABC to explore the PEL structure and thus obtain the dominant reaction pathway at this moment.
(ii) Apply nudged elastic band (NEB) method
to accurately calculate the energy barrier between the current and next minimum states in PEL that are identified in step (i).
(iii) Determine the thermal activation time defined as Dt shown in the figure, where n0 is the attempt frequency (of Debye order 1013 s-1). For a prescribed strain rate, The corresponding strain increment at a prescribed strain rate is then applied to the system, as illustrated in the figure. Go back to step (i) under the new strain condition.
The current algorithm is built in the DL-POLY framework. In future it will be implemented in the LAMMPS framework.
3. Exploring the multiple transition pathways in PEL
(reference: J. Phys.: Condensed Matter 26 (2014) 365402)
The original ABC method is only able to find out the most dominant mechanism, i.e. the pathway in the PEL with the lowest activation energy. However, in many cases there are multiple competitive pathways in the PEL, and it is important to uncover all of those. Such challenge is addressed by implementing a 2-steps extension: (i) Add a repulsive blocking Gaussian function centered at the identified saddle state (the green profile in the figure), so that the system will not jump out of the current basin, which would enable the identification of other saddle states; (ii) In addition to the blocking function, a penalty function will also be added to the initial minimum state in the PEL, as illustrated by the purple profile in the figure. The benefit of doing so is that the successive searches later on will not start from the very bottom of the basin. Instead, the successive searches start from the current level of the partially filled basin, so that the searching process will be significantly fastened.
The script is written in a compatible interface with LAMMPS.