Treffer: Optimization of microfluidic mixing using multiobjective strategies

It is known that rapid mixing in microchannels overcomes the inherent diffusion-limited mixing of laminar flow. Consequently, many techniques that enhance microfluidic mixing are under development: slanted wells, shallow grooves, electrokinetic instability mixing and surface layers, etc. The long-term goal of this work is the optimization of surface-property distributions to control mixing and provide surface-directed flows. In this work, binary fluid mixing is explored using the lattice Boltzmann Method (LBM) to simulate flows in two-dimensional, microfluidic channels having surface temperature variations. Previous work has shown the advantages of controlled wall temperature distributions (i.e., flow-through PCR devices). Over 100 mixing scenarios were simulated by varying the Reynolds number, wall temperature distributions, binary fluid density ratios and interaction strengths, and the coupling strength between momentum and temperature. If one adds channel geometry variations, we are optimizing a mixing function over a multidimensional parameter space of large dimension. This vector-valued mixing function contains two scalar-valued objective functions. Each objective function measures the mixing obtained for fluid 1 and fluid 2. The optimization problem is to find designs that simultaneously attain an optimal mix of both fluids. Consequently, we have a massive, computationally-intensive, multiobjective optimization problem. In multiobjective optimization problems, many acceptable designs can be obtained by trading one objective function against the other. For example, one might accept a slightly worse mixing of fluid 1 for a much better mix of fluid 2. We demonstrate optimal mixing as a function of these designs, that is the variation of the wall temperatures and the heater lengths.