Research

Quantum error correction is how fragile quantum bits become reliable computers: information is spread across many physical qubits so that errors can be detected and undone. But the codes we use today were largely designed by hand, for idealized, textbook noise. Real devices are messier — their noise is structured, correlated, and changes under the periodic drives (Floquet dynamics) used to control them.

This project builds a numerical framework that first learns the actual noisy dynamics of driven quantum hardware from measured data, and then uses that learned model to automatically discover error-correction codes and quantum sensing schemes matched to the hardware as it really is, rather than as we wish it were. The aim is a design loop where the machine proposes, evaluates, and refines encodings against the device's true noise — a step toward error correction that is co-designed with the hardware. Results are in preparation.

Most superconducting quantum circuits live in the 4-8 GHz band, where qubits, amplifiers, and control electronics are mature. Below 1 GHz lies a comparatively unexplored regime that natural signals and other quantum systems actually occupy — which makes it attractive for quantum sensing and for transduction, the faithful conversion of quantum signals between frequency bands.

This project develops new sub-GHz superconducting wave-mixing devices: nonlinear circuits that can amplify and convert low-frequency quantum signals. The work spans design, simulation, nanofabrication, and millikelvin measurement, with the goal of extending the standard circuit-QED toolbox — amplification, frequency conversion, and low-noise readout — down into the sub-GHz regime. Results are in preparation.

A qubit's environment usually cannot be chosen: connect it to the outside world for control and readout, and that same connection lets energy leak away. In everyday electronics this is a wire's job; in quantum hardware it is the central tension between controllability and coherence.

A "giant atom" offers a way out. When an artificial atom couples to a waveguide at several separated points, the emitted waves from each coupling point interfere — so its interaction with the environment depends strongly on frequency. Tuned to destructive interference, the atom effectively decouples and keeps its energy; tuned away, it couples strongly. In this work, an artificial giant atom is used as an engineered environment for a target superconducting qubit: the interference condition acts as a knob that reshapes what the qubit's surroundings look like, protecting it in one setting while permitting fast interaction in another.

The experiment demonstrates that a qubit's dissipative environment can itself be an engineered, tunable circuit element — a useful primitive for protected qubits, controlled emission, and quantum networks. See the paper for the device and the measured environment spectra.

One of the big open questions in nuclear physics is whether the phase diagram of quark matter (QCD) has a critical point — an end point of a first-order transition line, like the one where water's liquid-gas boundary terminates. The difficulty is that direct simulation fails in exactly the region of interest (the famous "sign problem"), so all we can compute reliably is a short Taylor series of the equation of state, expanded far away from the critical region.

The reach of such a series is limited by its nearest singularities — the Lee-Yang singularities, which are mathematically tied to the critical point itself. This paper shows how to get around that limit: by combining Padé approximants with conformal and uniformizing maps that reshape the complex plane, a handful of Taylor coefficients can be turned into an accurate reconstruction of the equation of state well beyond its naive radius of convergence — including the critical region, and even across onto the first-order side of the transition.

Tested on two exactly solvable stand-ins for QCD (a chiral random matrix model and the Ising model), the method locates the Lee-Yang singularities and reconstructs the singular equation of state from as few as ten expansion coefficients — evidence that the limited data lattice QCD can produce may be enough to corner the real critical point.

Adiabatic processes are the tortoise of quantum control: change a system slowly enough and it faithfully follows along, immune to many imperfections — but painfully slow. "Shortcuts to adiabaticity" add a cleverly designed extra drive that cancels the errors caused by going fast, delivering the adiabatic result in a fraction of the time. Before this work such shortcuts had only been demonstrated in closed systems, isolated from their environment — yet many of the things worth speeding up (like a measurement resonator filling with photons) are intrinsically lossy, open systems.

This experiment realized shortcuts to adiabaticity for open quantum systems on a superconducting circuit. A counterdiabatic drive steered a lossy microwave mode along its slow adiabatic trajectory eight times faster — 800 ns compressed to 100 ns — while an optimal-control extension equilibrated multiple lossy modes quickly and independently of the qubit's state.

The practical payoff is speed where it matters: faster ring-up and reset of readout resonators translates directly into faster qubit measurement cycles. More broadly, the result shows that precise time-domain control does not stop at the boundary of an open system — dissipation can be part of the choreography rather than an obstacle to it.

Every quantum algorithm starts the same way: with all qubits in their ground state. Getting them there quickly and reliably is harder than it sounds — simply waiting for a qubit to relax takes hundreds of microseconds, while measurement-based resets need fast feedback electronics and can disturb neighboring qubits.

This work demonstrates a reset that is both fast and unconditional — it needs no measurement and no feedback. Modulating the magnetic flux through a tunable transmon at the right frequency parametrically activates a swap between the qubit and its own readout resonator, which is deliberately lossy: the qubit's excitation is handed to the resonator and immediately drains away. The excited-state population is suppressed to 0.08% within 34 ns — roughly a thousand times faster than waiting for natural decay — using only hardware every standard circuit-QED chip already has.

The scheme also empties the qubit's second excited state (a common leakage error), leaves neighboring qubits essentially untouched, and doubles as a way to launch the qubit's state as a traveling single photon — useful for quantum communication. It has since become a practical building block for initializing tunable superconducting processors.