TY - JOUR
T1 - OpenPFC
T2 - An open-source framework for high performance 3D phase field crystal simulations
AU - Pinomaa, Tatu
AU - Aho, Jukka
AU - Suviranta, Jaarli
AU - Jreidini, Paul
AU - Provatas, Nikolas
AU - Laukkanen, Anssi
N1 - Publisher Copyright:
© 2024 The Author(s). Published by IOP Publishing Ltd.
PY - 2024/6
Y1 - 2024/6
N2 - We present OpenPFC (https://github.com/VTT-ProperTune/OpenPFC), a state-of-the-art phase field crystal (PFC) simulation platform designed to be scalable for massive high-performance computation environments. OpenPFC can efficiently handle large-scale simulations, as demonstrated by our strong and weak scaling analyses up to an 81923 grid on 65 536 cores. Our results indicate that meaningful PFC simulations can be conducted on grids of size 20483 or even 40963, provided there is a sufficient number of cores and ample disk storage available. In addition, we introduce an efficient implementation of moving boundary conditions that eliminates the need for copying field values between MPI processes or adding an advection term to the evolution equations. This scheme enhances the computational efficiency in simulating large scale processes such as long directional solidification. To showcase the robustness of OpenPFC, we apply it to simulations of rapid solidification in the regime of metal additive manufacturing using a recently developed quantitative solid-liquid-vapor PFC model, parametrized for pure tungsten (body-centered cubic) and aluminum (face-centered cubic).
AB - We present OpenPFC (https://github.com/VTT-ProperTune/OpenPFC), a state-of-the-art phase field crystal (PFC) simulation platform designed to be scalable for massive high-performance computation environments. OpenPFC can efficiently handle large-scale simulations, as demonstrated by our strong and weak scaling analyses up to an 81923 grid on 65 536 cores. Our results indicate that meaningful PFC simulations can be conducted on grids of size 20483 or even 40963, provided there is a sufficient number of cores and ample disk storage available. In addition, we introduce an efficient implementation of moving boundary conditions that eliminates the need for copying field values between MPI processes or adding an advection term to the evolution equations. This scheme enhances the computational efficiency in simulating large scale processes such as long directional solidification. To showcase the robustness of OpenPFC, we apply it to simulations of rapid solidification in the regime of metal additive manufacturing using a recently developed quantitative solid-liquid-vapor PFC model, parametrized for pure tungsten (body-centered cubic) and aluminum (face-centered cubic).
KW - high performance computing
KW - phase field crystal modeling
KW - rapid solidification
UR - http://www.scopus.com/inward/record.url?scp=85188344558&partnerID=8YFLogxK
U2 - 10.1088/1361-651X/ad269e
DO - 10.1088/1361-651X/ad269e
M3 - Article
AN - SCOPUS:85188344558
SN - 0965-0393
VL - 32
JO - Modelling and Simulation in Materials Science and Engineering
JF - Modelling and Simulation in Materials Science and Engineering
IS - 4
M1 - 045002
ER -