Brief research synopses

Photophysics and inference from images

Fluorescence microscopy has become an essential tool for biological research, allowing scientists to visualize subcellular processes (often in living cells) with protein specificity and impressive spatial and temporal precision. Despite this power, microscopic images are still imperfect representations of the positions of the light-emitting fluorophores in the sample. Optical diffraction and unavoidable statistical fluctuations in the measured signal lead to blurring and noise, even if other error sources are eliminated by careful measurement. Using optical and computational techniques to bring images closer to the underlying distribution of fluorophores has been an active research area for decades. In Andrew York’s lab at Calico, I have developed approaches to augment classic deconvolution techniques with extra information about the photophysical behavior of the emitters. The backbone of most of our methods is maximum likelihood estimation, but I’ve also applied other techniques (e.g. hidden Markov models, neural networks) to understand our problems.

In one project I incorporated photobleaching (a destructive process that extinguishes light-emitting fluorophores) as a prior for maximum likelihood deconvolution. Although we chose this as a first modification for its simplicity, it proved surprisingly rich. In the course of testing our algorithm, I posed a natural question: how can we fairly compare a multi-exposure method with a single exposure one? Investigating this led me to consider how to take the best image possible in the presence of photophysical noise. I developed theory to determine optimal single exposure times, and discovered the surprising result that this additional noise limits the expected signal-to-noise improvement from using bleaching resistant fluorophores. This has implications for nearly all fluorescent imaging methods. I also developed a nonlinear extension to the well-known Richardson-Lucy deconvolution algorithm and applied it to FRET microscopy, a technique where fluorophore-fluorophore interactions are responsible for key parts of the signal.

A more in-depth discussion of these projects is available here.

Quantum computing

Google’s quantum supremacy result is a spectacular example of the progress made in quantum computing in the last twenty years. That a roughly 50 qubit device is so noteworthy is itself a spectacular example of the immense engineering challenge quantum computing represents. Fifteen years ago, even within the world of superconducting quantum computing — where the qubits are built from small superconducting circuits — there was no consensus even on how one should build a single qubit. Many approaches were studied to learn how to build well performing qubits. My Ph.D. thesis was an experimental and theoretical study of one particular kind of superconducting qubit: the dc SQUID phase qubit.

After learning the ropes in the lab by doing measurements on some existing devices, I designed a new version of this qubit with a large shunt capacitor before the circuit. We expected this to allow projective microwave readout of the qubit state, in analogy to experiments in other labs. To test this I built several versions, using both e-beam and photolithographic techniques, and performed electronic and microwave measurements on these devices at temperatures near 50 millikelvin. When the new readout approach failed, I reviewed the theory we were using and found it used a key approximation that was no longer appropriate for those device parameters. I developed new theory to explain the quantum dynamics of our circuit. This theory became an important tool for our lab in designing the next generation of qubits to aim for lower loss. When we built and measured one of these to check the performance, we found unusual features that seemed explicable by quantum behavior in the shunt capacitor, in accordance with my theory.

A longer account of this work is available here.

Smaller projects

Before my dissertation, I worked on theoretical condensed matter research. My undergraduate thesis under Daniel Aalberts studied a model for the conjugated polymer family polyacetylene, a starting point for understanding the fast photoisomerization of light detecting molecules in the eye. I modified in-house C code to include steric effects in the model, looked at soliton behavior in odd-length chains, and studied how mechanically stretching these molecules changed the transition frequency. Working with Victor Yakovenko, I studied quasi-1D organic superconductors. We published a paper discussing the Aharonov-Bohm effect as a unifying theme explaining the angular magnetoresistance oscillations observed in these materials.

After my Ph.D., I did a brief postdoc in Ian Appelbaum’s lab. I implemented a proof-of-principle experiment to demonstrate the feasibility of a capacitive, nonlinear scheme for detecting bound Majorana fermion excitations in nanowire systems. Instead of nanowires, I fabricated superconducting-insulating-normal junctions as a test system to validate the measurement technique. I built a custom low-capacitance probe to use with a commercial low temperature measurement system, and performed conductivity measurements demonstrating the suitability of this approach.