Deep Learning
Recently, I have become interested in deep learning and artificial intelligence approaches to exploring complex systems. While machine learning has traditionally played an important role in particle physics and astronomy, other areas of physics (e.g., condensed matter, networks, chaotic systems) are now benefiting from advances in AI. This is an exciting research direction that, moving forward, I am enthusiastic about pursuing.
To begin exploring this research area, I built various neural networks capable of classifying the high- and low-temperature phases of the 2-dimensional Ising model. I have also investigated generative models (GANs and cGANs) that can produce Ising configurations for specified temperatures that are nearly indistinguishable from simulation results. This ongoing project is documented on my GitHub page.
Particle Levitation
Large or heavy particles suspended in a moving fluid can exhibit surprisingly complex behavior because their motion can significantly differ from that of the fluid. Unlike tracer particles (such as colored ink), heavy particles have inertia, which implies that they do not necessarily follow the streamlines of the fluid flow. From this property, we discovered a new form of dynamical particle levitation, whereby interacting vortices in a fluid can trap and carry particles (which are thousands of times denser than the fluid) against the direction of both the flow and gravity. See the video below for more explanation or read the paper!
Microfluidics
A primary topic of my Ph.D. thesis research is on the design of microfluidic networks through mathematical modeling and computational fluid dynamics. Microfluidic systems are typically composed of networks of tiny pipes embedded in a coin-sized plastic chip. The pipes are approximately the width of a strand of hair and carry only a few nanoliters of fluid. I design the network of pipes in these systems so that fluids flow through them in an automated or predefined way, thus reducing or removing the need for external control systems, such as computer-operated pumps. By using concepts from network science and complex systems, I have invented microfluidic systems that exhibit surprising flow behavior, including spontaneous oscillations, negative resistance, and bistability. Microfluidic technology is used widely to perform small scale experiments in chemistry and biology but also has a growing number of biomedical applications, including rapid bed-side diagnostic testing and wearable health monitoring devices.