VOLUME 79, NUMBER 14 PHYSICAL REVIEW LETTERS 6OCTOBER 1997 A Lattice BGK Model for Viscoelastic Media Yue-Hong Qian 1 and Yue-Fan Deng 2 1 Department of Applied Physics, Columbia University, New York, New York 10027 2 Department of Applied Mathematics and Statistics, SUNY-Stony Brook, Stony Brook, New York 11794 (Received 11 December 1996) A two- and three-dimensional lattice Bhatnager-Gross-Krook model is proposed to simulate viscoelastic media. Large-scale equations are derived using the Chapman-Enskog expansion. A simple linear relationship between the parameter E which is introduced to characterize the elastic behavior and the transverse velocity is obtained. Numerical simulations further confirm the analytical predictions. [S0031-9007(97)04196-3] PACS numbers: 83.50.Fc, 05.50. + q We investigate a new lattice-based model for simulat- ing viscoelastic media with the Bhatnager-Gross-Krook (BGK) approximation [1]. A decade ago, lattice gas models for hydrodynamics were introduced by Frisch et al. [2–5] and much research effort has led to encourag- ing progress [6–9]. These models have several appealing advantages over conventional methods for complex flows such as multiphase flows [10,11], flows in porous media [12], and reaction-diffusion systems [13]. The inconve- niences such as statistical noise [14,15] and non-Galilean invariance [4,5,16] in the original lattice gas models have been overcome by the use of lattice Boltzmann equations [17–20]. Another difficulty in lattice gas models, the existence of extra conserved quantities [21–24] which are not physically meaningful, was recently solved by using a fractional propagation procedure [25]. The introduc- tion of the BGK approximation to lattice-based models [20,26–28] simplified the complicated collision pro- cesses, increased computing efficiency, and offered new flexibility. These models have been quite effective for solving fluid problems. However, they have not been paid enough attention to in tackling solid or fluid-solid prob- lems [29–31]. The study of viscoelastic behavior of ma- terials done by d’Humières and Lallemand [32] based on a lattice gas model successfully produced transverse waves and the propagation speed. Their prediction was also con- firmed nicely by numerical simulations. However, their model requires an internal variable to describe “legal” collisions responsible for propagation of transverse waves and is clumsy in extension of their model to three dimensions. The equation used in lattice BGK models has the following form [20,26–28]: f i x 1 c i , t 1 1f i x, t 1vf e i x, t 2 f i x, t  , (1) where f i is the particle distribution density with pre- defined discrete velocity c i and v the relaxation parameter 0 #v# 2and i runs over the discrete velocity set. A suitable equilibrium f e i leading to the Navier-Stokes equations is [20,26] f e i t p r 1 1 c ia u a c 2 s 1 c i a c i b 2 c 2 s d ab 2c 4 s u a u b , (2) where t p is a lattice weight factor (the index p is equal to c 2 i ) and c s a constant. Greek subscripts a and b denote the space directions in Cartesian coordinates. The hydrodynamic quantities p and u are defined as r X i f i X i f e i , r u X i f i c i X i f e i c i . (3) We propose a simple lattice BGK model for viscoelas- tic materials. Unlike the internal variable needed in [32], a parameter E is introduced to characterize the elastic- ity. Choosing an equilibrium distribution is much more flexible in lattice BGK models than in the lattice gas models. In fact, adding a term to the equilibrium (2) to model the transverse waves is convenient, f e i t p r 1 1 c i a u a c 2 s 1 c i a c i b 2 c 2 s d ab 2c 4 s u a u b 1 E c i a c i b 2 c 2 s d ab 2c 4 s u a I b 1 u b I a 2 . (4) The last term involving E stems from the fact that one simple elastic behavior is wavelike propagation, like a vibrating string. The vector symbol I a appearing in the above equilibrium is defined as I a 1, ;a x, y, z . The Chapman-Enskog expansion [16,20,33] is used to derive the large-scale dynamical equations up to second order of the Knudsen number, t r1≠ a ru a 0, (5) 2742 0031-90079779(14) 2742(4)$10.00 © 1997 The American Physical Society