The lattice Boltzmann method (LBM) is a specific discrete formulation of the Boltzmann equation. Since its first premises, thirty years ago, this method has gained some popularity and is now applied to almost all standard problems encountered in fluid mechanics including multi-component flows. In this work, we introduce the inter-molecular friction forces to take into account the interaction between molecules of different kinds resulting primarily in diffusion between components. Viscous dissipation (standard collision) and molecular diffusion (inter-molecular friction forces) phenomena are split, and both can be tuned distinctively. The main advantage of this strategy is optimizations of the collision and advanced collision operators are readily compatible. Adapting an existing code from single component to multiple miscible components is straightforward and required much less effort than the large modifications needed from previously available lattice Boltzmann models. Besides, there is no mixture approximation: each species has its own transport coefficients, which can be calculated from the kinetic theory of gases. In general, diffusion and convection are dealt with two separate mechanisms: one acting respectively on the species mass and the other acting on the mixture momentum. By employing an inter-molecular friction force, the diffusion and convection are coupled through the species momentum. Diffusion and convection mechanisms are closely related in several physical phenomena such as in the viscous fingering instability.
A simulation of the viscous fingering instability is achieved by considering two species in different proportions in a porous medium: a less viscous mixture displacing a more viscous mixture. The core ingredients of the instability are the diffusion and the viscosity contrast between the components. Two strategies are investigated to mimic the effects of the porous medium. The gray lattice Boltzmann and Brinkman force models, although based on fundamentally different approaches, give in our case equivalent results. For early times, comparisons with linear stability analyses agree well with the growth rate calculated from the simulations. For intermediate times, the evolution of the mixing length can be divided into two stages dominated first by diffusion then by convection, as found in the literature. The whole physics of the viscous fingering is thus accurately simulated. Nevertheless, multi-component diffusion effects are usually not taken into account in the case of viscous fingering with three and more species. These effects are non-negligible as we showcase an initial stable configuration that becomes unstable. The reverse diffusion induces fingering whose impact depends on the diffusion between species.