› Traveling Wave Magnetophoresis

After I enrolled at Duke, I began working with the Magnetic Nanosystems Group to explore the non-linear dynamics of superparamagnetic beads in potential energy landscapes generated through superimposition of a rotating external magnetic field onto a micro-magnet array. Aside from demonstration of intriguing mechanics, systems like this have promising applications to micro-organism sorting and development of high-precision lab-on-a-chip devices.

Here is a visualization of the system I initially focused on.

By varying the direction of the external field with respect to the substrate magnetization, my objective was to be able to design a system where beads of differing sizes could be driven at varying angles across the substrate, or even lead to orthogonal separation. My simulations confirmed this possibility, though experiments proved more challenging, and require ongoing effort to fully realize this behavior. If implemented, it is a very efficient sorting mechanism for biological cargo in arbitrary quantities. 

Click here to read the paper, Physical Review E, 84, 011403 (2011)

During my summers, I traveled to Shanghai Jiaotong University, and studied the effect of varying the frequency ratio and phase difference between the horizontal and vertical fields on the dynamics of these magnetic beads. The system in this situation has potential to drive beads of two different sizes into opposite directions, and I used numerical simulations to guide my experimental efforts to realize this important behavior. 

Click here to read the paper, Lab on a Chip, 11, 4214-4220 (2011)

Click here to read the paper, Physical Review E, 85, 041407 (2012)

An illustration of how phase difference impacts the direction of the beads' motion on the substrate is shown below for various bead sizes. I am currently focusing on experiments with a slightly modified system that would allow even more enhanced control of directionality among different bead sizes.

As a part of my work within the group, I also developed a code that could automatically and robustly tag the locations of colloids within a micrograph.

This code was then utilized in various projects, such as in the study of a simple method to direct the self-assembly of colloids into numerous complex structures, selection among which is governed by two experimental parameters, ferrofluid concentration and number ratio of the two bead types undergoing assembly.

Click here to read the paper, Nature Communications, 3, 794 (2012)

A phase diagram of the rich possibilities that this system allows is shown below:

I am currently focusing on new substrate geometries for the same rotating field setup, and exploring further avenues and applications that this system can provide.

› Soft Active Materials

I have also worked in the Soft Active Materials Laboratory at Duke on studies relating to real-time observation of electromechanical instability formation in dielectric elastomers.

› Non-Linear Dynamics

Another project I work on (in the Duke University Non-Linear Dynamics Group) is exploring the non-linear dynamics of a rocking block on an oscillating foundation.


› Porous Media and Multiphase Flows

I worked in the Porous Media and Multiphase Flows lab at the Virginia Commonwealth University from September 05. 2008 to August 2. 2009. Prior to my research studies, I formulated scripts to generate computational models of non-woven fibrous media structures for various CFD codes like GeoDict (Fraunhofer ITWM) and Fluent (ANSYS). My primary research study was to investigate the sensitivity of numerical permeability computations to a media's fiber orientation. I discovered that transverse permeability calculations are independent of planar fiber orientation, but proportional to deviation of fiber angles from a planar setup. The results were published in Physics of Fluids.

Click here to read the paper, Phys. Fluids 21, 083604 (2009)

My structure generation code was utilized by the Porous Media lab for a study on the impact of fiber orientation on fluid spread in fibrous sheets, the results of which were published in the International Journal of Heat and Mass Transfer.

Click here to read the paper, Int. Heat & Mass Transfer 53, 1750 (2010)

A highlight figure from the Phys. Fluids paper, which depicts the relation between transverse permeability and through-plane fiber angle deviation (from zero), is shown below. The results were in disagreement with those presented in a former Phys. Fluids article on a related study.


› Radiative Heat Transfer

While at the Porous Media lab in VCU, I also did modeling work relating to heat flow in fibrous materials. This work has extensive relevance to industrial applications, including the design of household and spaceship (re-entry) insulation materials. For fibrous materials with fiber diameter much larger than the wavelength of incoming rays, I used ray-tracing to accurately model the radiative mode of heat transfer as a function of various microstructural parameters.The results were published in the International Journal of Heat and Mass Transfer.

Click here to read the paper, Int. Heat & Mass Transfer 53, 4629 (2010)

An example of a structure generated by my code, frozen at a particular instant through the simulation, is shown below:


› The TinyHPC  Cluster and Operating System

TinyHPC is a Class I beowulf high performance computing cluster of the shared memory architecture that I built in the summer following my junior year in high school. It uses MPI for inter-node communication. It has a theoritical peak of 15.6 GFLOPS


The first day I got a cluster running (ParallelKnoppix) is June 13. 2008. This version was composed of a Gateway e3400 Desktop, and two Toshiba Satellite laptops connected by D-Link DI-524. The next version connected by an Intel Express 330T hub and having four nodes running PelicanHPC (one Satellite laptop, and three desktops being from Gateway, Tangent, and eMachines) was created on July 08. 2008. A similar version with six nodes (the other Toshiba laptop, and another desktop) was finished on July 10. 2008. Up till here, these clusters did not do well on the HPL due to their hetrogeneous nature. As such, I decided to build a homogeneous cluster. I obtained five Intel 815EGEW motherboads and five Intel Celeron FCPGA2 1.3 GHz processor, as well as 512 Mb RAM sticks, power supplies, and network cards. The next version was ready by July 29. 2008. This version is now the TinyHPC project.

› Computational Intelligence

I worked in the Compuational Intelligence and Simulation lab at VCU since about September 20. 2008 to May 20. 2008. My work related to neural network based prediction of corrosion fatigue.

› Shell Access Over Email


A lot of us need access to our home computers, but for one reason or another, cannot transport it everywhere with us. SSH might be an answer to some, but not everyone, mainly due to lack of resources (such as a computer to SSH from!). 

Since I am able to email from my cell phone to my computer, I decided to set something up to let me email my computer shell commands that it can execute, while maintain a sensible degree of security that's enough for an average user - of course, this email-terminal setup can never be as secure as SSH as it is now. Anyways, here is the source if anyone wants to use it or play around with it. The code is in Python. It uses LibGMAIL, so you'll have to install that first to make any use out of the application. 



› Natural Language Programming

An attempt was made to design and implement a system that allows users to program a computer through limited domain natural language statements, thus extending the ability to interact with machines at a system level to the relatively novice population. The proposed system design combines a range of artificial intelligence related techniques, including natural language parsing, noun phrase chunking, named entity recognition, and neural network classification to translate raw natural language statements into a custom backend language that is designed to make the mapping of natural language parses to it as direct as possible. The entire system is also unique in its ability to “learn” from experience, made possible by the use of adaptive neural networks for the translation phase. As such, the system steadily improves over time in its ability to understand natural language statements, and coupled with the fact that it allows the user to program in an interactive manner, gives highly accurate results. A screenshot of the interface is presented below.


This project was done over the course of my junior year. It won many awards at regional and state science competitions, including recognition from Intel, SAIC, the Army, and the Air Force.