The light sees a perfectly flipped material, and conversion efficiency is maximized.
For decades, ultraviolet light has resisted being tamed at the scale of a computer chip, its energy dissipating before it could be made useful. Now, a Harvard team led by Marko Lončar has demonstrated a photonic device — built from lithium niobate and a novel fabrication technique called sidewall poling — that generates 120 times more UV light on a chip than any previous effort, crossing a threshold from scientific curiosity into practical possibility. The achievement, published in Nature Communications, arrives at a moment when quantum computers, atomic clocks, and environmental sensors all hunger for precisely this kind of compact, powerful light source. It is a reminder that the most consequential breakthroughs often live not in grand leaps of theory, but in the patient solving of a manufacturing puzzle.
- UV light has long been essential to quantum computing and precision sensing, yet chip-scale sources have remained too weak — often only microwatts — to be practically useful.
- The core obstacle was poling: flipping the internal crystal structure of lithium niobate in a precise repeating pattern without sacrificing the ability to correct for manufacturing imperfections.
- Harvard's team broke the deadlock by placing metal electrodes directly on the sides of waveguides with 50-nanometer precision, allowing full control of the crystal domains across the entire light-guiding channel.
- The device now produces 4.2 milliwatts of UV light at 390 nanometers — a 120-fold leap over prior demonstrations — moving the technology from proof-of-concept into the range of real-world application.
- Trapped-ion quantum computers, ultra-precise atomic clocks, and greenhouse gas sensors are all immediate beneficiaries, as each depends on compact UV light sources that until now simply did not exist at chip scale.
- Unexplained nonlinear effects observed in the material suggest lithium niobate still holds undiscovered properties, leaving the door open for further breakthroughs from the same platform.
Ultraviolet light is everywhere in nature, but confining and directing it through the microscopic channels of a photonic chip has long defeated researchers. The energy bleeds away, the manufacturing tolerances are punishing, and the applications — quantum computing, atomic clocks, environmental sensing — have waited. A Harvard team led by Marko Lončar has now cleared that barrier, reporting in Nature Communications a chip-scale device that produces 120 times more UV light than any previous attempt using the same material, thin-film lithium niobate.
The approach is indirect but elegant. Rather than generating UV light from scratch, the device converts red light into UV on the chip itself — two red photons merging inside the crystal to produce one higher-energy UV photon, a process called frequency upconversion. Lithium niobate's exceptional nonlinear optical properties make it a natural candidate, but exploiting those properties efficiently required solving a fabrication problem that had blocked progress for years.
The solution the team calls sidewall poling. To convert light efficiently, the crystal structure inside the waveguide must be flipped in a precise, repeating pattern — a process called poling. Previous methods forced a choice between poling accuracy and the ability to correct for imperfections introduced during manufacturing. The Harvard team escaped this bind by placing metal electrodes directly on the sides of the waveguide, positioned with 50-nanometer precision. Applying a small voltage during fabrication, they could flip crystal domains across the full cross-section of the channel and fine-tune the pattern to compensate for variations in film thickness and waveguide geometry.
The outcome is a device producing 4.2 milliwatts of UV light at 390 nanometers — where earlier demonstrations managed only tens of microwatts. For trapped-ion quantum computers, which require near-UV light to manipulate individual atoms, this means light sources could finally shrink to chip scale, making compact quantum hardware conceivable. The same wavelength range serves researchers building sensors to monitor greenhouse gases with high precision.
The team attributes part of their success to keeping theory, design, fabrication, and testing integrated within a single group — a unity of perspective that proved as important as any individual technique. And the work is not finished: the researchers have observed nonlinear effects in lithium niobate they do not yet fully understand, suggesting the material still has more to offer.
Ultraviolet light has always been abundant from the sun, but harnessing it on a computer chip has proven stubbornly difficult. The applications are everywhere—disinfection, biological imaging, the precise etching of circuit patterns—and in the near future, pinpricks of UV light embedded in photonic chips are expected to unlock advances in quantum computers and atomic clocks of unprecedented precision. The problem is that UV light bleeds away as it travels through the microscopic channels that guide it, making chip-scale sources seem perpetually out of reach.
Until now. A team led by Marko Lončar at Harvard has cracked the problem using a material called thin-film lithium niobate, a crystalline substance that has become a workhorse in integrated photonics. Their device generates 120 times more ultraviolet light on a chip than any previous attempt with the same material. The research, published in Nature Communications, suggests that lithium niobate—a platform most researchers associate with infrared wavelengths and telecommunications—can be coaxed into producing powerful UV light at scales small enough to matter.
The trick is elegant: rather than trying to generate UV light directly, the Harvard team converts red light into UV light right there on the chip. Two red photons combine inside the lithium niobate crystal to produce a single, higher-energy UV photon. This process, called frequency upconversion, relies on the crystal's exceptional ability to bend light in nonlinear ways. But making it work required solving a manufacturing puzzle that had stymied previous efforts.
The heart of the solution is something the team calls sidewall poling. In a photonic chip, light travels through waveguides—microscopic channels etched into the material. To convert red light to UV efficiently, the crystal structure inside the lithium niobate must be flipped in a precise, repeating pattern along the entire length of the waveguide. This pattern, called poling, had always been the bottleneck. Earlier approaches forced researchers into a bind: pole the entire film and lose the ability to correct for manufacturing imperfections, or fabricate the waveguides first and sacrifice efficiency because the electrodes that do the poling were too far away.
The Harvard team broke the deadlock by placing metal electrodes directly on the sides of the waveguide itself, rather than only on top. By positioning these "fingers" with 50-nanometer precision and applying a small voltage during fabrication, they could flip the crystal domains across the entire cross-section of the waveguide. The light traveling through sees a perfectly inverted material, and the conversion efficiency jumps dramatically. The team could also fine-tune the poling pattern along the device to compensate for variations in film thickness and waveguide shape, using techniques developed in Lončar's lab.
The results speak for themselves. The device produces 4.2 milliwatts of ultraviolet light at a wavelength of 390 nanometers—about 120 times more power than the best previous demonstrations with thin-film lithium niobate at that wavelength. Earlier versions of the technology generated only tens of microwatts, enough to prove the concept but nowhere near practical use. This new output opens doors.
For quantum computing, the implications are immediate. Trapped-ion quantum computers, which use individual atoms held in place by electromagnetic fields, often require light at near-ultraviolet wavelengths to manipulate those atoms. If quantum computers are ever going to be compact enough to be useful—rather than the size of a truck—the light sources need to shrink to the chip level. The same wavelength band also appeals to researchers building environmental sensors, devices that could monitor greenhouse gases and atmospheric pollutants with unprecedented precision.
Lončar's team succeeded, according to the researchers themselves, because they kept all the work under one roof: the theory, the design, the fabrication in the cleanroom, and the optical testing. That integration of perspective—understanding both the deep physics and the practical engineering—proved essential. As the work moves forward, there remain mysteries to solve. The team has observed nonlinear effects in the material that they don't yet fully understand, suggesting that lithium niobate still has surprises to offer.
Notable Quotes
If you want a scalable quantum computer that isn't the size of a truck, you need to scale everything down to the chip level, and this includes the light sources.— Kees Franken, co-first author and former research fellow
The hands-on intuition that we gained for how to make a device, while also keeping the zoomed-out view of what this device is for—that's a big part of what enabled this project for us.— Soumya Ghosh, co-first author and former graduate student
The Hearth Conversation Another angle on the story
Why does ultraviolet light lose power so quickly on a chip, when other wavelengths don't?
UV photons are high-energy, and the materials we use to guide light tend to absorb them more readily. It's like trying to contain something volatile in a small space—the shorter the wavelength, the more it wants to escape.
So instead of generating UV directly, you convert red light to UV on the chip itself. Why is that easier?
Because red light travels through lithium niobate with almost no loss. We've been using this material for infrared applications for years. Once we get the red light where it needs to be, the crystal can do the conversion right there, locally, before the UV has a chance to degrade.
The sidewall poling sounds like a small tweak. Was it really that hard to get right?
It required 50-nanometer precision in electrode placement over centimeter-long devices. That's positioning something smaller than a virus with absolute consistency. The real insight was asking: why are we keeping the electrodes away from the waveguide? Once we brought them to the sides, everything changed.
What does 120 times more power actually mean for someone building a quantum computer?
It means the difference between a proof-of-concept and something you can actually scale. Previous versions generated microwatts. This generates milliwatts. That's the threshold where you can start thinking about real systems.
Are there still unknowns?
Yes. We're seeing nonlinear effects in the material that we don't fully understand yet. Lithium niobate keeps revealing new properties. That's both humbling and exciting.