A Photonic Chip Physically Separates Hidden Modes in Mixed Light

A silicon photonic chip has demonstrated a physical operation that conventional cameras and software reconstruction do not perform: it learned the hidden coherence modes inside overlapping light and routed those components into separate output beams. The laboratory device could eventually give microscopes and telescopes access to information that ordinary intensity images discard.

What the chip separates

Coherent light maintains a stable relationship between its waves, while incoherent light does not. Real scenes often contain partially coherent mixtures whose underlying sources overlap in space. A camera records the combined intensity, but that two-dimensional image can lose correlations that reveal how many independent optical components produced it.

The team treated those components as coherence modes, mathematically distinct patterns that are mutually uncorrelated. Earlier techniques could reconstruct such modes numerically through tomography. The new circuit instead separates them physically, leaving each mode at its own waveguide output where another optical system could measure or process it.

Key Number: The nine-input chip separated its two dominant modes with more than 98% fidelity against on-chip tomography.

A circuit that configures itself

The foundry-made device uses layers of Mach-Zehnder interferometers and tunable phase shifters on silicon. Two independent infrared lasers near a wavelength of 1,535 nanometers illuminated nine grating-coupler inputs. Electronic optimization adjusted the circuit to maximize one optical component at the first output, then repeated the process for the remaining light.

The separated outputs reached measured cross-talk as low as minus 13.0 decibels, with an uncertainty of 2.6 decibels. The researchers also verified operation with three coherence modes on a reconfigurable 9 by 9 circuit. Because the hardware grows linearly with both the number of sampled spatial channels and the number of modes, the authors argue that the architecture offers a scalable alternative to a brute-force optical processor.

Why cameras may be leaving information behind

Two fluorescent particles can overlap in a microscope image and appear as one spot even when their light carries different correlation patterns. A coherence-mode separator could, in principle, route those contributions apart before detection. The same principle might help astronomy distinguish unresolved emitters or allow communications and optical-computing systems to use partially coherent light more deliberately.

This is more than a new image-processing algorithm. The chip performs the decomposition in the optical domain and keeps the outputs available, so a downstream sensor or processor does not have to recreate the original field from intensity measurements alone. Stanford researchers say they are working with NASA to explore whether the approach can extract additional information from distant astronomical sources.

What researchers need to demonstrate next

The next step is to connect the chip to real partially coherent scenes and show that the separated outputs improve a concrete measurement. Microscopy tests would need to resolve overlapping emitters under realistic noise, while astronomy would require stable coupling across many spatial channels and long observations. Those experiments will determine whether physical coherence sorting becomes a new imaging instrument or remains an elegant laboratory analysis of light.