Passive mixing in microchannels with geometric variations

This research project was part of the microfluidic program in the CRC for Microtechnology, Australia, during 2000 to 2003. The aim of this research was to investigate the feasibility of applying geometric variations in a microchannel to create effects other than pure molecular diffusion to enhance microfluidic mixing. Geometric variations included the shape of a microchannel, as well as the various obstacle structures inside the microchannel. Generally, before performing chemical or biological analysis, samples and reagents need to be mixed together thoroughly. This is particularly important in miniaturized Total Analysis Systems (μTAS), where mixing is critical for the detection stage. In scaling down dimensions of micro-devices, diffusion becomes an efficient method for achieving homogenous solutions when the characteristic length of the channels becomes sufficiently small. In the case of pressure driven flow, it is necessary to use wider microchannels to ensure fluids can be pumped through the channels and the volume of fluid can provide sufficient signal intensity for detection. However, a relatively wide microchannel makes mixing by virtue of pure molecular diffusion a very slow process in a confined volume of a microfluidic device. Therefore, mixing is a challenge and improved methods need to be found for microfluidic applications. In this research, passive mixing using geometric variations in microchannels was studied due to its advantages over active mixing in terms of simplicity and ease of fabrication. Because of the nature of laminar flow in a microchannel, the geometric variations were designed to improve lateral convection to increase cross-stream diffusion. Previous research using this approach was limited, and a detailed research program using computational fluid dynamic (CFD) solvers, various shapes, sizes and layouts of geometric structures was undertaken for the first time. Experimental measurements, published experimental data and analytical predictions were used to validate the simulations for selected samples. Mixing efficiency was evaluated by using mass fraction distributions. It was found that the overall performance of a micromixer should include the pressure drop in a microdevice, therefore, a mixing index criterion was formulated in this research to combine the effect of mixing efficiency and pressure drop. The mixing index was used to determine optimum parameters for enhanced mixing, as well as establish design guidelines for such devices. Three types of geometric variations were researched. First, partitioning in channels was used to divide fluids into mixing zones with different concentrations. Various designs were investigated, and while these provided many potential solutions to achieving good mixing, they were difficult to fabricate. Secondly, structures were used to create lateral convection, or secondary flows. Most of the work in this category used obstacles to disrupt the flow. It was found that symmetric layouts of obstacles in a channel had little effect on mixing, whereas, asymmetric arrangements created lateral convection to enhance crossstream diffusion and increase mixing. Finally, structures that could create complex 3D advections were investigated. At high Reynolds numbers (Re = 50), 3D ramping or obstacles generated strong lateral convection. Microchannels with 3D slanted grooves were also investigated. Mixers with grooved surfaces generated helicity at low Reynolds numbers (Re ≤ 5) and provided a promising way to reduce the diffusion path in microchannels by stretching and folding of fluid streams. Deeper grooves resulted in better mixing efficiency. The 3D helical advection created by the patterned grooves in a microchannel was studied by using particle tracing algorithms developed in this research to generate streaklines and Poincare maps, which were used to evaluate the mixing performance. The results illustrated that all the types of mixers could provide solutions to microfluidic mixing when dimensional parameters were optimized