Purdue Researchers Tackle Key Barriers to Photon-Based Quantum Computing

A quantum computer built from light would, in principle, be both faster and cheaper to run than the machines now under development. That is the premise behind a body of work at Purdue University, where engineers have spent several years chipping away at the obstacles to building quantum photonic integrated circuits at scale. These are chips in which single photons, the smallest possible particles of light, serve as the qubits.

A qubit is the quantum counterpart of the bit. An ordinary bit is a zero or a one; a qubit, through superposition, represents both at once, which greatly expands the information a machine can process. Photons offer particular advantages as qubits. They would move through a circuit's gates at the speed of light, cutting delays in an algorithm's performance. They are also not easily disturbed by heat or changes in their surroundings, so they can stay stable at room temperature. Other qubit types need supercold temperatures, which costs more energy.

The catch is scale. Today's quantum photonic chips can handle a small number of qubits, but practical calculations would need hundreds, if not thousands. The work is led by Vladimir Shalaev, the Bob and Anne Burnett Distinguished Professor in Electrical and Computer Engineering, and Alexandra Boltasseva, the Ron and Dotty Garvin Tonjes Distinguished Professor in the same field. They work within Purdue's Elmore Family School of Electrical and Computer Engineering, the Birck Nanotechnology Center and the Purdue Quantum Science and Engineering Institute. Both have appeared on the Clarivate Highly Cited Researchers list for multiple years, and Boltasseva was on last year's Quantum 100 list. Shalaev says the biggest scaling problems come down to creating good single-photon sources.

The first problem is producing exactly one photon at a time. The industry's usual process, spontaneous parametric down-conversion, is inconsistent and inefficient. In a 2021 Science Advances study, Shalaev and Boltasseva showed an alternative on silicon nitride, the industry-standard material for quantum photonic platforms. Their team created atomic-scale defects in the material that limit emission to a single photon. Each time a laser beam passes through, dozens of locations emit exactly one photon, at room temperature. A later study, presented at the 2022 Conference on Lasers and Electro-Optics, showed they can control the placement and number of defects by embedding them in nanosized pillars, which should make manufacturing more uniform.

The second problem is control. Single photons barely interact with one another, which makes them hard to manage with switches, the quantum-chip equivalent of a conventional computer's transistors. In a Nature Nanotechnology paper last year, the team reported a proof-of-concept switch that works at the single-photon level. It adapts optical nonlinearity, a mechanism used in fiber-optic communications in which a control beam influences a probe beam, though normally at power levels far too high for single photons. In the Purdue version, one photon from the control beam strikes silicon and creates an electron, which triggers an avalanche of electrons. That changes silicon's refractive index, so the probe beam passes through only when a single photon starts the avalanche, letting one photon turn the beam on and off.

The third problem is decoherence, in which a disturbance knocks qubits out of superposition or entanglement and introduces errors. Alexander Kildishev, a Purdue professor of electrical and computer engineering, and graduate student Pranshu Maan led a team that included Shalaev and Boltasseva. Their findings appeared in Nature Communications in May. Statistical-physics analysis showed that decoherence unfolds over time in a pattern, and machine learning confirmed the patterns hold predictable information. Boltasseva said decoherence only seems random.

The team targeted spectral diffusion, a major cause of decoherence, in which an emitted photon's frequency, and so its energy, drifts because of natural disorder in its environment. Superposition requires the photons to be indistinguishable, so one shifted photon can collapse it. The researchers found that these shifts retain a memory of how the frequency has wandered. They trained an algorithm with a built-in memory mechanism on data from individual quantum emitters to forecast future shifts. Using those forecasts, anticipatory control reduced spectral shifts by as much as 20-fold in the systems studied.

Shalaev put the stakes in qubit counts. Decoherence cannot be eliminated, he said, but it can be minimized once its pattern is learned. An algorithm needing a hundred qubits might typically require something like 10,000 to absorb the errors; fewer errors mean fewer qubits.

The work so far covers individual emitters. Next, the team plans to use statistical physics to extract hidden properties of a fluctuating environment from measured dynamics and feed them into the machine-learning models. They will also test whether the models can identify the microscopic processes behind spectral shifts, with the long-term aim of suppressing decoherence before it disrupts the quantum state, and eventually across multiple systems.