(612e) Optimal Design of Droplet Based Microfluidic Ladder Networks Using Genetic Algorithm

Droplet microfluidic technology helps is achieving spatio-temporal control of chemicals in a pre-defined network topology. The next generation droplet based lab-on-chip devices can possibly revolutionize combinatorial drug screening and high throughput analysis of chemicals and cells.  Ladder network is an example of a microfluidic device that is designed to achieve synchronization of pairs of drops. Ladder networks have two parallel channels that are interconnected by bypass channels. The droplets travel in the main channels and the bypass channels are inaccessible to droplets. To optimally design ladder networks there is a need to develop appropriate modeling and design algorithms. Using the concepts of a network model, a generic simulation approach for the ladder networks is developed. Using this simulation model we predict that the ladder networks with fore-aft structural symmetry are limited to reducing droplet spacing. Additionally, we show that asymmetry in ladder networks provide additional functionalities such as expansion, contraction, flipping and synchronization of drops. 

Typical objective functions for optimizing ladder network designs are binary and discontinuous. Therefore, conventional gradient based approaches for design are difficult to implement. In this talk, we propose a genetic algorithm based approach to design these networks. A novel encoding scheme is proposed to handle experimental constraints. In this work, we discuss the design of ladder networks for two different objectives using genetic algorithms: (i) Ladder networks that encode and decode droplets as they travel through the device, (ii) Ladder networks that synchronize drops that enter with a mean and a variance in delay.  In the first case, the design generated by the genetic algorithm transforms the exit spacing of the drops to a sigmoid function. In the second case, the ladder networks act at as a filter, where the inlet spacing of droplets is amplified by the first few bypasses and reduced by the other bypasses.