1. The enhanced sampling in parallel finite-time dynamics method with replica exchange.
- Author
-
Xu, Chudong, Lu, Shengdong, Kong, Yongfeng, and Xiong, Wanjie
- Subjects
- *
PARALLEL algorithms , *ISING model , *LANGEVIN equations , *MONTE Carlo method , *PHASE transitions - Abstract
The finite-time dynamical method is an effective tool studying phase transition in many-body systems. We design a parallel version of this method based on replica exchange whose probability is deduced from Langevin equation. The configuration swaps between different replicas enhance the importance sampling in parallel and control the fluctuation, leading to the conspicuous improvement for the simulation efficiency. The computational performance with configuration exchange is above 3 times as high as the calculation effect without the exchange on five current CPU cores. In addition, this parallel algorithm is used to simulate the two- and three-dimensional Ising model and obtain compatible exponents with that from theoretical solutions and numerical simulations from references, verifying the reliability of the scheme. • We develop a parallel version of the finite-time dynamics method based on MPI. • The replica exchanges enhance the importance sampling and suppress the fluctuation. • The speed-up reaches 3 times on 5 CPU cores as opposed to the simulation without the swap. [ABSTRACT FROM AUTHOR]
- Published
- 2021
- Full Text
- View/download PDF