A modified lattice Boltzmann model with a stochastic relaxation mechanism mimicking "virtual" collisions between free-streaming particles and solid walls is introduced. This modified scheme permits to compute plane channel flows in satisfactory agreement with analytical results over a broad spectrum of Knudsen numbers, ranging from the hydrodynamic regime, all the way to quasifree flow regimes up to Kn ∼ 30.
PACS numbers:The dynamic behaviour of flows far from hydrodynamic equilibrium is an important subject of nonequilibrium thermodynamics, with many applications in science and engineering. The non-hydrodynamic regime is characterized by strong departures from local equilibrium which are hardly handled on analytical means. Consequently, much work is being devoted to the development of computational techniques capable of dealing with the aforementioned non-perturbative and non-local effects. Recently, the lattice Boltzmann (LB) method has attracted considerable interest as an alternative to the discretization of the Navier-Stokes equations for the numerical simulation of a variety of complex flows [1]. Extending LB methods to non-hydrodynamic regimes is a conceptual challenge on its own, with a variety of microfluidic applications, such as flows in micro and nano electro-mechanical devices (NEMS, MEMS) [2]. Departures from local equilibrium are measured by the Knudsen number, namely the ratio of molecular mean free path, l m , to the shortest hydrodynamic scale, l h : Kn = l m /l h . Ordinary fluids feature Kn < 0.01, while high Knudsen numbers are typically associated to rarefied gas dynamics, where l m is large because density is small, typical case being aero-astronautics applications. More recently, however, the finite-Kn regime is becoming more and more relevant for a variety of microfluidics applications in which the Knudsen number is large because of the increasingly smaller size of the devices. It is also worth emphasizing that the finite-Knudsen regime may also bear relevance to the problem of modeling fluid turbulence [3,4]. Recent work indicates that LB may offer quantitatively correct information also in the finite-Kn regime [5]. This hints at the possibility that LB may complement or even replace expensive microscopic simulation techniques such as kinetic Monte Carlo and/or molecular dynamics. On the other hand, this is also puzzling because finite-Knudsen flows are in principle expected to receive substantial contributions from highorder kinetic moments, whose dynamics is arguably not quantitatively correct because of lack of symmetry of the discrete lattice [6,7]. In this work we point out that, irrespectively of symmetry considerations, any LB extension aiming at describing non-hydrodynamic flows in