Foundry-friendly trick lets wide, low-loss microresonators do Kerr nonlinear optics
The paper shows a practical way to get strong on‑chip optical nonlinearity from microresonators made on a standard silicon nitride foundry process. The team uses “modal phase matching” — matching different spatial light patterns inside the ring — to meet the strict conditions needed for Kerr nonlinear interactions while keeping resonator waveguides wide. They designed and built devices on a 160‑nanometre‑thick silicon nitride platform and demonstrated Kerr optical parametric oscillation on a foundry chip.
The researchers designed and fabricated high‑quality (high‑Q) microrings in a commercial foundry and used multiple spatial mode families to satisfy dispersion requirements. In experiments they observed optical parametric oscillation, reported a conversion efficiency of about 20%, and achieved continuous (gap‑free) tuning of the generated signal over more than 1 nm near the cesium D1 atomic transition (roughly 895 nm). They also showed the same approach can be pumped in both the 1060‑nm and 795‑nm bands to generate wavelengths across roughly 600 nm to 1400 nm, without needing custom device layer thicknesses.
In plain terms, Kerr nonlinear optics on a chip converts light from one color to others through a process called four‑wave mixing. That process must conserve both energy and momentum, which in a resonator translates to tight frequency and phase matching conditions. The usual way to get the needed dispersion is to make tall, narrow waveguides, but those shapes increase scattering loss and complicate foundry processing. Modal phase matching meets the matching conditions by combining different spatial modes in the same wide waveguide, so the designers can keep the geometry that reduces sidewall scattering and raises Q while still enabling wavelength conversion.
Why this matters: it relaxes a long‑standing tradeoff between low loss (high Q) and the dispersion needed for Kerr processes. Because the devices are made on a thin, foundry‑compatible silicon nitride platform, the approach can be more easily scaled and integrated with other on‑chip components. The authors also report practical tuning: on‑chip heaters let them sweep the signal wavelength between about 894.5 nm and 895.9 nm without gaps, which is useful for applications tied to atomic transitions.