A Monash led team closes the on chip generation routing readout loop for valley polarized photons at room temperature, and measures a 0.97 polarization selectivity in the process.
Photonic processors have a long-standing bandwidth problem. Light is fast, but most on-chip schemes use just one property of a photon (wavelength, or polarization) as the information carrier. A team at Monash University argues there is room for a second one, encoded in a quantum property called the valley index, and that it has shown, on a single chip and at room temperature, the full loop of generation, selective routing, and electrical readout that such a scheme would require.
The result, published in Nature Photonics, is described by the authors as a "critical gap" closed: prior work had shown pieces of the chain, but not all three functions on one device. The team's on-chip nanocircuit integrates chirality-selective meta-waveguide photodetectors with transition metal dichalcogenide (TMD) monolayers. The TMD stack — an encapsulated tungsten disulfide (WS2) monolayer as the chiral-photon source and few-layer tungsten diselenide (WSe2) as a photodetector — is dropped onto a metasurface. The team reports a polarization selectivity of 0.97 for the valley-dependent waveguide modes, meaning light of one valley handedness is routed to one output and the opposite handedness to another, with very little leakage between them.
In the proof-of-concept experiment, the chip encoded and processed two different images at the same time, each carried on a different valley-polarized channel. The output was read out electrically, which is the part that brings the device closer to a circuit element than a lab curiosity.
Room temperature is the practical lever. Many valleytronic and quantum photonic demonstrations run at cryogenic temperatures, which limits where the technology can plausibly be deployed. The Monash device runs at room temperature, according to the institutional release, because the team routes the valley information into an electrical signal via a TMD photodetector, sidestepping the need to preserve optical quantum coherence end-to-end.
The application envelope, as the authors and Monash frame it, spans faster and lower-energy computing, secure communications, advanced imaging, next-generation optical communication, and quantum information processing. ScienceDaily's release groups "AI and quantum computing" as headline targets. The honest reading is more modest: this is a device-level result, and the photonic-computing relevance is downstream of the demonstration, not a claim the paper leads with.
Lead author Dr. Chi Li said the achievement addresses a major obstacle in valleytronics research. "Until now, we could generate or detect these signals, but not do everything in one integrated device," Dr. Li said in the Monash release. "What we've built is a complete on-chip system that can create, route and read this information with very high precision."
Co-first author Dr. Kaijian Xing, a Research Fellow at Monash, explained the integration approach: "We employ a straightforward stacking approach to integrate ultra-thin materials with metasurfaces, overcoming the technical challenges of direct material growth on photonic structures, and enabling further advances in valleytronics." ScienceDaily's report notes that the international project brought together researchers from Australia, China, Singapore, Germany, and Japan.
Senior author Dr. Haoran Ren, an ARC Future Fellow who leads the Monash NanoMeta Group, framed the security use case: "Valley is a quantum property of electrons, which cannot be cloned or intercepted without being destroyed. This makes it intrinsically secure for the next generation of communications." The Monash release frames this as an architectural potential, and the "cannot be cloned" claim is the no-cloning property of quantum states applied to valley-encoded photons — best read as a directional argument about a technology's potential rather than a guarantee of a deployable secure channel.
There are reasons to read the result carefully. The 0.97 polarization selectivity is a single-device number, not a wafer-scale statistic. Manufacturing scalability is unaddressed in the public materials: the stacking approach is described as straightforward by the authors, but yield, uniformity, and wavelength compatibility with existing silicon photonics are not characterized. The work also sits inside a larger valleytronics literature, traced in the paper's reference lineage to foundational treatments such as Schaibley et al. (2016) in Nature Reviews Materials and Xiao et al. (2007/2012), and to more recent bulk MoS2 valleytronics work such as Tyulnev et al. (2024) in Nature. A single proof-of-concept nanocircuit is one data point in that lineage, not a replacement for it.
What to watch: a third-party measurement of the polarization selectivity and routing efficiency at wafer scale, a clear statement of the operating wavelength and how it lines up with silicon photonics infrastructure, and a move from the two-image demo to a multi-channel encoding experiment that tests the valley-multiplexing assumption directly. If those data points land, the room-temperature result stops being a curiosity and starts being a building block for photonic processors that need every available data channel they can get.