Molecular-based numerical schemes, such as the direct simulation Monte Carlo (DSMC) method, are more physically appropriate for rarefied gas flows in microelectromechanical systems (MEMS). It is difficult for them to be statistically convergent, however, because the statistical fluctuation becomes insurmountably large at the low Mach numbers that are characteristic of MEMS. An information preservation (IP) technique is proposed to address this issue. This technique assigns each simulated molecule in the DSMC method two velocities. One is the molecular velocity used to compute the molecular motion following the same steps as the DSMC method. The other is called information velocity. It corresponds to the collective velocity of an enormous number of real molecules that the simulated molecule represents. Using the information velocity to compute macroscopic velocity and shear stress may remove the statistical fluctuation source inherent in the DSMC method that results from the randomness of the thermal velocity. The IP technique has been applied to benchmark problems, namely Couette, Poiseuille, and Rayleigh flows, in the entire Knudsen regime. The characteristic velocities in these flows range from 0.01 to 1 m/s, much smaller than the thermal velocity of about 340 m/s at room temperature. The meaningful results are obtained at a sample size of 10 3 -10 4 , in comparison with a sample size of 10 8 or more required for the DSMC method at such a range of flow velocity. This results in a tremendous gain in CPU time. The velocity distributions, surface shear stress, and mass flux given by the IP calculations compare quite well with exact solutions at the continuum and free molecular limits, and with the numerical solutions of the linearized Boltzmann equation and experimental data in the transition regime.

Large statistical scatter and effective pressure boundary conditions are two critical problems in the computation of microchannel ows with the direct simulation Monte Carlo (DSMC) method. To address these issues, an extension of the DSMC-IP (information preservation) coupled method is developed from the one-dimensionalcase to the twodimensional case for microchannel ow. Simulation results in a microchannel ow from DSMC, IP, and numerical and analytical solutions to the Navier-Stokes equations are compared. The DSMC-IP coupled method successfully reduces the large statistical scatter usually obtained with DSMC in such low-speed ow systems. It also provides a suitable implementation of pressure boundary conditions. Introduction MICROCHANNELS are an important componentof many microelectrical mechanical systems (MEMS). 1 Successful numerical simulation of the ow eld inside these devices is required to understand small scale ow phenomena. Adopting a standard computational uid dynamics (CFD) technique is not appropriate because CFD is based on the continuum assumption, which is only good for the continuum regime (Kn < 0.001), and acceptable for the temperature jump and velocity slip region (0.001 < Kn < 0.1) if a slip wall condition is adopted instead of nonslip boundary conditions. However, for microchannel ows under experimental conditions, 1 ¡ 4 the ows are sometimes in the transition regime (0.1 < Kn < 10). Here, rare ed effects are signi cant, and CFD methods are not reliable. The direct simulation Monte Carlo method (DSMC) 5 is accurate for all ow regimes because it is based on kinetic theory and does not rely on the continuum assumption. Many researchers have already performed much work on simulation of microchannel ow with the DSMC method.6 ¡ 10 However, there are still many dif culties, and in some researchers ' belief 8 it is impossible to use DSMC to simulate microchannel ows under experimental conditions. Indeed, there are many experimental results, but no corresponding DSMC simulations are reported yet. To statistically simulate the ow under experimental conditions, we must overcome two particular dif culties: statistical scatter and proper implementation of pressure boundary conditions. In this study we will discuss these dif culties in detail. A new technique, the DSMC-IP (information presentation) coupled method, is presented to address these problems. Two cases are computed in the present study: a simpli ed test case and a case under experimental conditions. Dif culties Associated with DSMC Calculation Statistical ScatterUsually, the ow velocity in microchannels under typical experimental conditions is very low. For example, in the experiments of Pong et al.,2 the inlet velocity is about 20 cm/s. If we suppose the velocity obeys a Maxwellian distribution,then at room temperature for nitrogen, the standard deviation is r = p (2RT ) = 422 m/s. If we suppose the sampling processes are totally independent from step to step, then the statistical scatter in the nal DSMC result will be r 0 = r / p N , ...

No abstract

No abstract

No abstract