Research

Interfacial mass and heat transfer

The group studies the interfacial mass and heat transfer that underlies environmental separations: the processes that govern how energy, water, and dissolved matter are exchanged between engineered systems and the natural environment. Our work combines bench-scale experiment, multiscale modeling, and data-driven analysis.

Thermal management

Regenerating dielectric coolants for next-generation cooling systems

Schematic of an immersion cooling loop coupled to a pervaporation regeneration unit. Used dielectric coolant leaves the tank holding the computer chip, is pumped through a heater into a pervaporation membrane module, and splits into a permeate vapor stream carrying water, light volatiles, and dissolved gases to a condenser and vacuum pump, and a retentate stream of renewed coolant that returns to the tank.

Why coolants have to be regenerated

Immersion cooling fluids take up water and dissolved gases, accumulate particulates and dissolved metals, and generate their own degradation products. They are costly and hard to dispose of responsibly, so the environmental case for immersion cooling rests on reuse rather than replacement.

Contaminant transport in a nonaqueous fluid

These species partition and diffuse very differently in a dielectric fluid than in water. We measure their behavior at separation interfaces to identify what actually limits removal.

Coupled heat and mass transfer

Regeneration happens in a fluid whose purpose is to carry heat, so separation and thermal transport are not independent. We study how temperature gradients and phase change interact with the transport of dissolved species.

Performance over a service life

Whether regeneration is worth doing depends on behavior after many cycles, not one. We track how separation efficiency and coolant properties evolve with repeated use.

Water security and resource recovery

Self-powered capture of metal ions in aquatic environments

Exploded schematic of an integrated water treatment system built into a portable water container. A motion-powered cap module holds an induction coil with a moving magnet, end magnets, a plastic tube, a supercapacitor, and a power management circuit. Wires run from the cap to an electrode cassette inside the container, shown in detail as a frame holding an arsenic adsorbing carbon felt, an insulating separator, a counter electrode, current collectors, and electrode terminals.

A hazard and a resource at once

The ions that make a water supply unsafe sit alongside metals whose supply chains are strained. Either way the target has to be pulled out of a background of chemically similar ions that are often far more abundant.

What sets ion selectivity

Hydration energy, charge, size, and specific binding compete to determine which ion crosses a membrane or binds a sorbent. We work to resolve which of these controls selectivity for a given target ion.

Capture without an external power supply

The communities that most need selective capture often have the least infrastructure for it. We study how spontaneous chemical potential differences, concentration gradients, and ambient energy can drive capture where a reliable grid cannot be assumed.

Recovery, not only removal

Taking a metal out of water and recovering it in usable form are different problems. We study the release and concentration steps that decide whether a captured metal becomes a product or another waste stream.

Engineering education

Engineering judgment in an era of abundant knowledge

Two-panel comparison titled Engineering judgment: era of abundance. The old way, friction-based, shows search, calculation errors, and assumption defense connected by a long winding resistive path that leads to robust judgment. The new way, agent-assisted, shows an AI agent delivering instant knowledge along a straight arrow that leads to fragile judgment.
Three-step sequence starting from an AI agent and ending in robust judgment. Step 1, verifying naive intuition through instant doubt training and sandbox testing. Step 2, adversarial boundary testing: expose hidden assumptions, vary conditions, and identify when the solution fails. Step 3, knowledge internalization: public reasoning, defending the decision with evidence, and applying the judgment in a new context.

Knowledge is no longer the constraint

A student can retrieve a derivation, a design precedent, or working code faster than they can judge whether any of it applies. What is scarce now is the judgment to tell.

Judgment was built on friction

An agent removes exactly the experiences that used to build it: the search that dead-ends, the estimate wrong by three orders of magnitude, the assumption that has to be defended out loud.

What develops, what atrophies

We want to know which parts of engineering judgment still form when an agent is always at hand, which quietly weaken, and which mature faster because harder problems come within reach sooner.

Designing courses around it

The engineers we train will practice with these tools, not without them. We treat this as an open empirical question in our own courses and lab rather than a settled one.