A polarization camera used to be a piece of furniture. To measure how light waves are oriented as they bounce off a surface, an optics lab would mount a rotating polarizing filter in front of a sensor, take a sequence of photos as it spun, and reconstruct the answer in software. Astronomers used the technique to map magnetic fields on the sun. Defense labs used it to spot the unnatural glint of a vehicle hiding under a camouflage net. Materials engineers used it to see invisible stress frozen into a pane of glass. It was powerful, slow, and big. Nobody was putting it in a phone.

Now it fits under your screen.

The quiet version of this story lives in the US patent record. Across 9.3 million US utility grants, the number that describe a metasurface or metalens — a flat optic that bends light with rows of nanoscale pillars instead of a curve ground into glass — has gone from a handful before 2016 to 143 last year, 633 all-time. Most of that is the obvious stuff: tiny lenses for proximity sensors, beam shapers for lidar, optics for augmented-reality goggles. But a sharper signal hides inside it. Grants that pair a metasurface with the word polarization went from 3 in 2019 to 24 in 2025, an eightfold run. And two names sit at the top of that narrow list: Harvard, and a company Harvard spun out of one professor’s lab.

The trick is reading the twist

Light has an orientation. When a wave reflects off a surface, the geometry of that surface tilts the wave’s electric field in a way the Fresnel equations have described since the 1820s. Read the tilt across an image and you are reading the shape and material of everything in front of the camera, not just its color and brightness. A printed photo, a phone screen playing a video, a silicone mask — each scrambles polarization differently from living skin. The information was always there. The problem was always the hardware.

Federico Capasso’s group at Harvard spent the 2010s collapsing that hardware. Capasso is not a small figure in optics; he co-invented the quantum cascade laser at Bell Labs in 1994, a light source now sitting in industrial gas sensors worldwide. His Harvard lab’s 2011 paper in Science showed that a flat sheet patterned with subwavelength antennas could steer light as well as a shaped lens. A 2016 cover paper showed those flat sheets taking genuinely sharp images. In 2016 he, his graduate student Rob Devlin, and entrepreneur Bart Riley founded Metalenz to commercialize it, with an exclusive Harvard license to the flat-optics portfolio.

Read the company’s recent grants and the polarization thread is unmistakable. One 2025 grant, “Polarization sorting metasurface microlens array device,” describes an array of tiny metasurface lenslets where each lenslet splits incoming light by polarization and throws the two states onto different pixels of an ordinary image sensor. No spinning filter, no time sequence. The whole polarization measurement happens in one exposure, in a layer attached directly to the sensor with a spacer or an air gap. Another, “Transmissive metasurface lens integration,” co-designs the part that projects light and the part that collects it into a single optical system built on a semiconductor substrate. The Harvard patents underneath are the physics layer: “Systems and methods for parallel polarization analysis” claims a single birefringent metasurface grating that reads multiple polarization orders from arbitrarily polarized light at once.

Mentally delete the word “metasurface” from that set and the inventions still belong together. They are all variations on one engineering move: take an optical bench that needed moving parts and multiple shots, and freeze it into a single flat chip you can make in a chip fab. That is the coherence test the convergence keyword stories fail, and this one passes.

Who cares: the notch is the target

Here is the “so what.” The most valuable face-unlock hardware on Earth is the structured-light module Apple introduced with Face ID — a dot projector that sprays thousands of infrared points onto your face and a camera that reads the distortion. It works, and it is expensive, power-hungry, and physically bulky enough that it has dictated the industrial design of the front of every iPhone since 2017. You cannot hide it behind a powered OLED screen, which is why the notch and the pill-shaped cutout exist.

Polarization unlock attacks exactly that constraint. Because the metasurface reads the polarization signature of real skin rather than projecting a dot pattern, it needs no projector, and the polarization signal survives passing through a lit OLED display. Metalenz’s Polar ID system claims a 0% spoof acceptance rate and, per the company, runs under a fully powered phone screen — something structured light cannot do. The pitch to a phone maker is blunt: payment-grade face authentication at roughly half the size and cost of the dot projector, with the front of the phone finally free of cutouts.

The supply chain has been quietly assembling around it. Samsung supplies the image sensor the system runs on, its ISOCELL Vizion 931, and Korean leakers say Samsung is testing Metalenz’s face biometrics for a future flagship. STMicroelectronics, which already ships time-of-flight sensors built on Metalenz optics, expanded its licensing deal to manufacture the metasurfaces in its own fabs. In November 2025, Metalenz and United Microelectronics Corporation announced Polar ID had reached mass production. By the company’s count, roughly 100 million of its metasurfaces are already installed in shipping consumer devices, mostly in humble proximity and distance sensors. The biometrics piece is the upgrade.

This is why the patent concentration matters more than the patent count. Metalenz holds nine US grants; that is nothing next to Samsung, which leads broad metasurface filings with 26 since 2021, or Harvard’s 21. But in the polarization-imaging slice — the one that turns a flat lens into a spoof-resistant biometric and, in the new Polar 3D variant unveiled in February, into a single-shot 3D face capture that runs on-device with no cloud round trip — Harvard and Metalenz are the two leaders, with eight and six grants since 2020. The big consumer electronics companies own the wide field. The Harvard spinout owns the specific corner that is about to ship inside their phones.

The adjacent door

Polarization imaging spent decades as a specialist’s instrument: remote sensing, microscopy, quality control on factory lines, the kind of thing that lived in a $10,000-plus camera or a university optics bench. The science never needed to improve for it to reach a phone. What needed to change was the optics getting small, cheap, and manufacturable in the same fabs that already stamp out image sensors by the hundred million. The metasurface did that. The moment a measurement that required moving parts becomes a static chip, every product that could not afford the bench suddenly can.

Devlin frames the near-term use as authentication for AI agents and payments, where, in his words, “the interaction needs to be truly seamless.” But the more interesting frontier is everything else polarization sees. The same chip that tells a living face from a photo can in principle read the gloss of paint for a defect, the stress in a windshield, the sheen of skin for a health signal, the difference between water and black ice on a road. Aptiv, the automotive supplier, is already filing metasurface patents. The face-unlock sensor is the wedge. What follows it through the door is a polarization camera cheap enough to put on anything.

Method note. Counts come from 9.3M US utility patent grants sourced from USPTO bulk grant XML, searched by full text for metasurface, metalens, and flat-optics terms, and the subset that also mentions polarization, over grants issued from 2016 through early June 2026. Assignee tallies combine variant spellings and subsidiary filings for each organization and reflect grants only, not pending applications, so they undercount very recent activity that has not yet issued. Patent counts measure where work is being protected, not market share or revenue. Company, product, and partnership details — Polar ID, Polar 3D, the Samsung sensor, the STMicroelectronics and UMC manufacturing deals, the ~100 million metasurfaces shipped, and the company’s founding history — come from Metalenz, Harvard’s Office of Technology Development, optics.org, Optica’s Optics & Photonics News, and Biometric Update. Spoof-rate and under-display performance figures are the company’s own claims.