Light now travels a multi-lane highway on a chip — protected by pure math

University of Pennsylvania School of Engineering and Applied Science / Nature Physics

A donut and a coffee mug walk into a math department. They're told they're the same thing. The mathematician isn't wrong: both objects have exactly one hole, and no amount of stretching, squishing, or warping changes that fact. This is topology, the branch of mathematics concerned not with shape but with properties that survive deformation.

It sounds abstract. It is abstract. But topology has quietly become one of the most powerful tools in photonics — the science of moving information using light. And a team led by engineers at the University of Pennsylvania has just demonstrated something the field has been chasing for years: a way to send multiple streams of light through the same chip-based network, all protected by topological principles that make the signals nearly impervious to defects.

The work, published in Nature Physics, marks the first time anyone has achieved multi-channel topological protection in a reconfigurable photonic system. Until now, topology could shield only one signal at a time — a fundamental bottleneck that limited its practical value.

One hole, one protected lane — until now

To understand why this matters, consider what topology actually does for light. In a conventional optical chip, photons travel through carefully fabricated waveguides — tiny channels etched into silicon or similar materials. The problem is that fabrication is never perfect. Microscopic defects scatter light, degrade signals, and introduce errors. The more complex the circuit, the worse it gets.

Topological photonic systems sidestep this by routing light along pathways defined not by the physical perfection of the structure, but by its mathematical topology. "Those defects don't change the topology of the system," says Li Ge, a professor of physics and astronomy at the City University of New York and co-author of the study. The signal keeps moving even when the structure is flawed.

Think of it as a highway where potholes literally cannot affect traffic. That's not hyperbole — it's a direct consequence of the math. Topologically protected light follows edge states that are, by definition, robust against local perturbations. The signal doesn't scatter because, from topology's perspective, the defect doesn't exist.

But there was a catch. Each protected pathway could carry only a single mode of light — one lane, one stream of data. For a technology promising to underpin next-generation computing and communications, that's a serious constraint.

The pseudo-spin trick

The breakthrough came from rethinking how different pseudo-spin states of light behave at the boundaries within a photonic lattice. Pseudo-spin is an engineered property that mimics the quantum spin of electrons, giving photons a kind of internal degree of freedom that can be manipulated.

In previous topological photonic systems — including one demonstrated by senior author Liang Feng's lab at Penn in a 2019 Science paper — each interface between lattice regions supported only one protected mode traveling in a given direction. That earlier work proved topology could guide light through a network of microring resonators (tiny circular waveguides that trap and redirect photons), rerouting beams around corners without scattering. But the single-mode limitation remained.

The new insight was theoretical at first. Tianwei Wu, a postdoctoral fellow in Penn's Department of Electrical and Systems Engineering and co-first author, realized that by engineering the coupling between pseudo-spin states at the boundary, it should be possible to create conditions where multiple protected channels emerge simultaneously along the same interface.

"In conventional topological systems, each interface usually supports only one protected mode in a propagation direction," Wu explains. The team's mathematical analysis showed this wasn't a fundamental law — it was a design limitation that could be overcome.

Building the lattice

Translating theory into silicon demanded precision. The photonic chip consists of a lattice of microring resonators, each one a few tens of micrometers across, fabricated on a chip roughly the size of a fingernail. The resonators must be spaced and coupled with exacting tolerances to produce the desired multi-channel behavior.

Xilin Feng, a doctoral student in ESE and the study's other co-first author, led much of the fabrication effort. "The coupling between the resonators had to be engineered very carefully," Xilin Feng says. "But once everything was in place, we were able to observe multiple protected channels propagating along the same interface, even when defects were introduced."

That last clause is the key result. The team deliberately introduced imperfections into the lattice — the photonic equivalent of digging potholes — and the multiple channels kept transmitting. Topology held.

Why photonics needs more lanes

The practical stakes are high. Light already carries the vast majority of the world's long-distance data through fiber-optic cables, but the next frontier is on-chip photonic communication — replacing some of the electrical wiring inside and between computer chips with optical interconnects. Light is faster, consumes less power, and generates less heat than electrons pushing through copper.

But chip-scale photonic networks are notoriously fragile. Manufacturing imperfections, thermal fluctuations, and material aging all degrade performance. Topological protection addresses these problems at a fundamental level, but only if the protected channels can carry enough data to be useful. A single-mode system is, in networking terms, a single-lane country road. Modern data infrastructure needs highways.

The Penn team's demonstration transforms that single lane into multiple parallel lanes, each individually protected. Liang Feng, who holds appointments in both Materials Science and Engineering and Electrical and Systems Engineering at Penn, frames it directly: "We can bring the robustness of topology into systems that carry many signals at once."

What this doesn't do yet

Some caveats are worth stating plainly. This is a laboratory demonstration, not a commercial product. The number of simultaneous protected channels, while greater than one, is still modest — scaling to dozens or hundreds of channels will require further engineering. The system operates under controlled conditions; real-world deployment would subject it to temperature variations, mechanical stress, and other perturbations beyond what was tested.

There's also the question of integration. A standalone photonic chip demonstrating multi-channel topological protection is one thing. Embedding that capability into a larger computing or communications architecture — alongside electronic components, modulators, detectors, and the rest of the photonic toolkit — is a different engineering challenge entirely.

Still, the conceptual barrier has been broken. The field now knows that multi-mode topological protection is physically achievable, not just theoretically possible. That's the kind of result that redirects research programs.

The road from here

Future work from the group will likely focus on three fronts: increasing the number of protected channels, integrating the design into larger photonic circuits, and exploring whether similar topological principles can be adapted for other applications — routing light in quantum computing systems, for instance, or building more resilient sensors.

The study was supported by the Army Research Office, the Office of Naval Research, and the National Science Foundation. It was conducted at the University of Pennsylvania School of Engineering and Applied Science, with contributions from the City University of New York.

If the designs can be scaled, they could become a foundational technology for photonic networks — systems where data travels as light, protected not by better manufacturing, but by mathematics that doesn't care about imperfection.