Quantum computers can compute faster than supercomputers, and with photons - the smallest particles of light - they will become even faster and more economical. But mass production of chips where photons work as qubits (quantum bits that simultaneously represent zero and one) remains a difficult task. Engineers at Purdue University are looking for solutions for quantum photonic integrated circuits.
Photons have advantages over other particles: they travel at the speed of light and are less susceptible to heating and interference, so qubits based on them can operate at room temperature. Other types of qubits require intense cold, which demands additional energy. However, practical calculations require hundreds or thousands of qubits, and current photonic chips do not yet reach that scale.
The main difficulty is a reliable source of single photons. In 2021, researchers showed in Science Advances how to emit single photons using silicon nitride, a material that has become a standard for quantum photonic platforms. The traditional method of spontaneous parametric down-conversion produces photons unstably and with low efficiency. The team's approach emits exactly one photon at room temperature from dozens of points each time a laser beam passes through the platform. These points are artificial defects in the atomic structure that confine emission to a single photon. Later, the scientists showed they could control the placement and number of defects by embedding them in nanoscale pillars, which simplifies manufacturing.
Another problem is that photons interact poorly with each other, making them hard to control with switches. Last year, in Nature Nanotechnology, the team reported a switch operating at the level of single photons. It is based on optical nonlinearity: one photon of a control beam knocks an electron out of silicon, and that triggers an avalanche of electrons that changes the material's refractive index. A probe beam passes through the device only when one photon has triggered the avalanche, which is what makes it possible to switch the signal.
Another line of work is combating decoherence, that is, the loss of quantum properties due to interference. In May this year, in Nature Communications, a group led by Alexander Kildishev and Pranshu Maan, with the participation of Shalaev and Boltaseva, showed that decoherence over time has a pattern that can be identified by machine learning. The researchers developed an algorithm with a memory mechanism that predicts future shifts in a photon's frequency (spectral diffusion) and makes it possible to correct them in advance, reducing such shifts in the systems studied. It is impossible to completely eliminate decoherence, but predicting and managing it can reduce the number of errors and the need for additional qubits. In the future, the team plans to use methods of statistical physics to better understand the causes of decoherence and improve the models.