Sensing the invisible: GaN devices for low-concentration detection

The hardest sensing problems are rarely about detecting whether something is there, they're about detecting it when there's almost none of it. Trace gases. Femtomolar concentrations of a pathogen. A few stray molecules of hydrogen leaking from a fuel cell. This is the challenge that drew me to gallium nitride (GaN) at Stanford, and to a deceptively simple question: how do you build a device that can reliably hear a whisper?

AlGaN/GaN high-electron-mobility transistors (HEMTs) are remarkably well-suited to this problem. The physics starts with the two-dimensional electron gas (2DEG), a thin sheet of highly mobile electrons that forms spontaneously at the interface between the AlGaN and GaN layers. Because this charge layer sits just nanometers from the device surface, it responds to chemical events at that surface with unusual sensitivity. When a molecule binds nearby, it perturbs the local electric field, and the 2DEG shifts in response. That shift is electrical, measurable, and fast. The transduction chain from chemistry to signal is short, which is part of why these devices work so well.

But sensitivity at trace concentrations demands more than a good transducer. At low analyte concentrations, the signal competes with noise from thermal fluctuations, surface traps, and ambient interference. My work focused on the structural and surface-engineering choices that tip that balance, understanding how gate geometry, surface functionalization, and passivation layers affect the signal-to-noise ratio when very few molecules are present. For gas sensing applications like hydrogen detection, this also meant thinking carefully about operating conditions: GaN devices are stable at elevated temperatures where many competing sensor technologies degrade, which makes them attractive for the demanding environments of hydrogen infrastructure and fuel cell monitoring.

The broader implication is that GaN isn't just a power electronics material. Its surface chemistry, thermal resilience, and native piezoelectric properties make it a compelling platform for sensing across domains, from biosensors detecting glucose or DNA in clinical settings to real-time gas monitors in industrial environments. Designing these devices to work at the edge of detection, where concentrations are lowest and the stakes are often highest, is where the interesting materials science lives.

What I found, and what continues to drive my interest in this space, is that reliable low-concentration sensing isn't a single problem, it's an intersection of epitaxial growth quality, surface chemistry, and device architecture. Getting all three right is hard. Getting them right together is what makes a sensor actually useful.