Kanae Oguchi and Toshio Suzuki
Department of Materials Engineering, School of Engineering, The University of Tokyo, Tokyo 113-8656
The Computational efficiency of a conventional phase-field simulation was compared with a modified simulation in which the thermal field away from the interface was calculated by a random-walk algorithm. The growth velocity and computational time were compared for the simulations by two programs. The error and computational time were mainly dependent on the distance from the interface, at which the random-walk calculation was started. The distance was given by the number of calculation mesh intervals between the interface and the edge of the random-walk calculation area, and the error in growth velocity decreased but the computational time increased with increasing the number of mesh intervals. When the number of mesh intervals was set to be 80, the error in growth velocity was negligible and the computational time was reduced by 20%. The computational efficiency was sigficantly improved at small undercooling conditions, where the necessary calculation area for thermal diffusion became large.
phase-field model, dendrite growth, random-walk, computational efficiency
Please do not copy without permission.