Caltech researchers have developed ultralow-loss optical pathways directly on silicon wafers, bringing the efficiency of fiber-optic light transmission closer to chip-scale devices and opening possibilities for more precise lasers, optical clocks, quantum technologies, sensors, and energy-efficient data-center communications.
Modern communications depend heavily on optical fiber because it can carry enormous amounts of information over long distances while losing remarkably little light along the way.
Reproducing that performance on a tiny silicon chip has been far more difficult.
Now, researchers at the California Institute of Technology have developed a photonic platform made from germano-silicate glass โ a material closely related to the glass used in optical fiber โ that can be fabricated directly on standard silicon wafers.
The resulting optical circuits achieve extremely low signal loss across wavelengths ranging from violet light through the near-infrared and telecommunications bands. At visible wavelengths, where existing integrated photonic devices often struggle with substantial scattering losses, the improvement is particularly dramatic.
The research was published in Nature and demonstrates a route toward photonic integrated circuits, or PICs, whose light-guiding performance begins to approach that of conventional optical fiber.
Bringing fiber-optic performance onto a chip
A photonic integrated circuit performs many of the functions of an electronic integrated circuit, except it manipulates photons rather than primarily moving electrical current.
Tiny structures known as waveguides direct light through the chip.
Photonic chips are already used for telecommunications, sensors, lasers, computing, and other optical technologies. But every time light travels through a waveguide, some fraction of its energy can be absorbed or scattered away.
This loss becomes especially problematic when light must circulate for long periods or interact with a device many times.
Optical fiber performs exceptionally well because its glass is extremely transparent and manufactured with remarkably smooth interfaces.
Caltech physicist Kerry Vahala and his colleagues have spent years trying to reproduce those advantages on silicon wafers.
Their new method essentially transfers some of the material principles responsible for fiber optics' low losses into a planar chip-manufacturing process.
The team fabricated circuits from germano-silicate, or germanium-doped silica, on silicon wafers using a lithography-based process compatible with semiconductor manufacturing.
Why visible light has been especially difficult
Researchers have already achieved impressive optical performance on chips at telecommunications wavelengths, particularly around the near-infrared region used by fiber-optic communications.
Visible wavelengths present a tougher challenge.
As wavelength becomes shorter, tiny imperfections in a waveguide become increasingly important.
Surface roughness that has relatively little impact on longer-wavelength infrared light can scatter a much larger fraction of visible light.
Material absorption can also increase toward shorter wavelengths.
Together, these effects mean that photonic circuits operating in the visible and near-visible spectrum generally lose much more light than their telecommunications-band counterparts.
That is a major limitation because many valuable applications depend on visible or short near-infrared wavelengths.
These include atomic clocks, trapped-ion quantum systems, quantum sensors, biological imaging, precision navigation, lidar, astronomy, and communications technologies.
The Caltech platform was designed specifically to overcome this wavelength problem.
The same family of glass used in optical fiber
The researchers turned to germano-silicate because of its long history in optical fiber.
Adding germanium oxide to silica changes the refractive index of the glass, allowing light to remain confined within a carefully engineered waveguide.
More importantly, the material can exhibit exceptionally low intrinsic optical absorption.
The challenge was developing a practical way to manufacture high-quality germano-silicate waveguides on silicon wafers.
The team developed a deep-ultraviolet lithography process capable of patterning planar optical circuits from the material on wafers of the type used by the semiconductor industry.
This moves the technology away from long spools of fiber and toward compact integrated optical systems.
Heating the glass makes its surface extraordinarily smooth
One of the platform's most important advantages comes from the physical properties of germano-silicate itself.
Because the glass softens at comparatively accessible processing temperatures, researchers can heat the fabricated waveguides so their surfaces undergo a controlled reflow.
This smooths out microscopic imperfections.
Hao-Jing Chen, one of the study's lead authors, explained that the treatment can reduce surface roughness toward atomic-scale smoothness.
That matters because microscopic roughness is one of the principal causes of scattering loss, particularly for visible light.
By reducing those irregularities, far more of the light remains inside the waveguide instead of being scattered away.
Record-low losses across a wide range of wavelengths
The researchers measured waveguide performance at wavelengths extending from violet light to the standard telecommunications band.
After annealing, they reported losses of about 0.49 decibels per meter at 458 nanometers, 0.32 dB/m at 532 nm, 0.19 dB/m at 780 nm, and only 0.08 dB/m at 1,064 nm.
At the telecommunications wavelength of 1,550 nm, the reported loss was approximately 0.09 dB/m.
The 1,064-nanometer result approaches the low-loss performance achieved by some of the earliest breakthrough optical fibers.
More strikingly, the researchers report that their violet-band waveguide loss is approximately 13 decibels lower than that of previous integrated photonic platforms.
Caltech describes the visible-wavelength performance as roughly a 20-fold improvement over the previous silicon-nitride record.
Why lowering optical loss matters so much
A small reduction in waveguide loss can produce a disproportionately large improvement in certain optical devices.
One example is a ring resonator.
In a ring resonator, light enters a closed circular or spiral path and travels around it repeatedly.
Although the physical structure may be only millimeters across, the photons can effectively travel a much greater distance because they circulate again and again.
Each pass introduces a small amount of loss.
If that loss is high, the light disappears quickly.
If the waveguide is exceptionally efficient, the light can continue circulating for much longer.
This is why researchers working with a chip only a few centimeters wide can still care about losses measured over distances equivalent to meters or even kilometers.
The effective optical journey can be enormous even when the physical device is tiny.
Quality factors above 180 million
One measure of how efficiently an optical resonator stores light is its quality factor, commonly called the Q factor.
A higher Q means light remains trapped within the resonator for more cycles before its energy dissipates.
The germano-silicate devices demonstrated resonator Q factors exceeding 180 million across wavelengths ranging from violet to telecommunications frequencies.
Such high-Q resonators are valuable because they strengthen interactions between light and the material around it.
That can improve lasers, frequency combs, sensors, nonlinear optical devices, and other systems requiring extremely precise control of photons.
Better waveguides can produce much more coherent lasers
The benefits become particularly dramatic for lasers whose performance depends on high-quality optical resonators.
Laser coherence describes, in simplified terms, how consistently the light maintains a well-defined frequency and phase.
Higher coherence is essential for applications requiring exceptionally stable light, including precision measurements and atomic clocks.
According to the Caltech team, reducing optical loss by a factor of 10 can produce roughly a 100-fold improvement in laser coherence in relevant resonator-based systems.
Lasers demonstrated using the new platform showed more than a hundredfold improvement in coherence compared with earlier designs highlighted by the researchers.
That could enable much more stable chip-scale laser sources.
Optical clocks could become smaller
Among the technologies that stand to benefit are optical atomic clocks.
Atomic clocks measure time using extremely stable transitions between atomic energy levels.
Optical clocks operate at much higher frequencies than conventional microwave atomic clocks and can achieve extraordinary precision.
But today's most advanced systems often require large and complex optical setups.
Miniaturizing the lasers, resonators, frequency references, and supporting components onto photonic chips could help move such clocks beyond specialized laboratories.
Because different atomic species interact with particular visible or near-infrared wavelengths, a photonic platform capable of operating efficiently across a broad spectrum is especially useful.
The wide wavelength range demonstrated by the germano-silicate platform could support multiple kinds of atomic transitions and precision-measurement systems.
More precise gyroscopes and navigation systems
Ultralow-loss photonics could also improve optical gyroscopes.
These instruments measure rotation by comparing how light travels along different paths.
They are used in navigation systems where GPS may be unavailable or unreliable, including aircraft, spacecraft, submarines, and precision autonomous systems.
The longer light can circulate through an optical resonator without being lost, the more sensitive such measurements can potentially become.
Highly efficient on-chip resonators could therefore contribute to smaller and more precise inertial-navigation devices.
The Nature paper specifically identifies precision navigation and quantum sensing among the technologies that could benefit from fiber-like integrated photonics.
Quantum technologies could also benefit
Many quantum-computing and quantum-networking platforms rely on specific optical wavelengths.
Trapped ions, neutral atoms, quantum dots, and other quantum systems often require carefully controlled laser light to manipulate and measure quantum states.
Visible and near-visible photonic circuits have historically been more challenging to build with extremely low loss.
The Caltech platform substantially expands the wavelength range in which high-performance integrated photonics can operate.
That could make it easier to build compact optical components for ion traps, quantum sensors, quantum communications, and other technologies that need precise control of visible light.
Possible benefits for AI data centers
The technology may also have implications for data centers, where moving information is becoming a major energy challenge.
As AI systems grow larger, enormous quantities of data must travel between processors, memory, accelerators, racks, and facilities.
Electrical connections consume substantial energy and generate heat, particularly as data rates increase.
Optical links can move information more efficiently over certain distances, which is why photonics is receiving increasing attention for next-generation computing infrastructure.
Henry Blauvelt, a Caltech visiting associate and one of the paper's authors, noted that germano-silicate waveguides can also be engineered to transfer light efficiently between conventional optical fibers and semiconductor lasers.
Reducing losses at those interfaces and within photonic circuits could help lower the energy required for optical communication in server infrastructure.
Spirals squeeze long optical paths onto tiny chips
The researchers do not need meter-long chips to take advantage of ultralow-loss waveguides.
Instead, optical pathways can be wound into spirals and rings.
Much like wrapping a long length of fiber around a spool, a spiral waveguide allows light to travel a long distance while occupying only a small area of silicon.
This geometry is particularly useful for delay lines, resonators, lasers, and precision sensors.
The physical chip may be only around a few centimeters across, yet the light can effectively travel an enormous accumulated distance through repeated circulation.
That is why fiber-like losses matter even on very small devices.
The platform does more than guide light
The Nature paper also demonstrates that the material can support several important photonic functions beyond simple low-loss transmission.
The researchers demonstrated soliton microcomb generation, which creates a set of precisely spaced optical frequencies from a compact resonator.
Optical frequency combs are useful for clocks, spectroscopy, communications, and precision measurement.
They also demonstrated stimulated Brillouin lasing, a process in which light interacts with acoustic vibrations inside a material to produce highly coherent optical signals.
In addition, the researchers showed low-frequency-noise self-injection locking, a technique for stabilizing semiconductor lasers using an ultrahigh-Q resonator.
These demonstrations show that the germano-silicate platform is not simply a passive optical wire.
It can provide the foundation for complex integrated photonic systems.
Compatibility with existing chip manufacturing matters
A laboratory material can achieve spectacular performance and still be difficult to turn into useful technology if it requires completely unfamiliar manufacturing methods.
One of the notable aspects of the Caltech approach is its compatibility with wafer-scale processing.
The Nature study describes the fabrication process as CMOS-foundry compatible, meaning it is designed around manufacturing concepts used by the semiconductor industry.
Caltech says the circuits can be produced directly on the same general class of 8-inch and 12-inch wafers used for computer chips.
That compatibility could eventually make it easier to combine ultralow-loss photonic structures with lasers, modulators, detectors, electronics, and other materials on the same platform.
Performance is already strong without high-temperature treatment
Although thermal reflow provides the lowest reported losses, the researchers also found impressive performance without the post-fabrication annealing step.
In the telecommunications band, the unannealed germano-silicate devices achieved a quality factor about 10 decibels higher than previous comparable platforms highlighted in the study.
This is important because not every component that might be integrated onto a photonic chip can survive high temperatures.
Semiconductor lasers, electronic circuits, and other sensitive materials may be damaged by intense thermal processing.
A platform that retains very low loss without requiring a final high-temperature treatment could therefore simplify the integration of multiple technologies on a single chip.
A โSwiss Army knifeโ for photonics
Vahala describes the platform's appeal as its broad usefulness.
Rather than addressing only one narrow photonics problem, ultralow optical loss can improve many technologies simultaneously.
Better resonators can produce more coherent lasers.
Better waveguides can reduce communication energy.
Broad wavelength operation can support different atoms and ions.
Improved optical storage can increase sensor precision.
Efficient nonlinear resonators can generate new frequencies and optical combs.
And compatibility with conventional wafers creates possibilities for bringing these functions together inside increasingly complex integrated systems.
Fiber optics may finally be shrinking onto silicon
Conventional optical fiber transformed telecommunications by allowing light to travel vast distances with remarkably little attenuation.
Integrated photonics has long sought to bring that same efficiency into much smaller devices.
The challenge has been especially severe at visible wavelengths, where tiny imperfections cause light to scatter much more readily.
By combining the glass chemistry of optical fiber with semiconductor-style fabrication and an unusual surface-smoothing process, the Caltech researchers have pushed integrated waveguide losses much closer to fiber-like territory.
The work does not mean today's fiber networks can simply be compressed onto a chip without trade-offs.
The researchers themselves describe the technology as an important step toward fiber-like integrated photonics rather than the final limit.
But their results demonstrate that the enormous performance gap between optical fiber and chip-scale waveguides can be substantially narrowed.
If those losses continue to fall, technologies that currently require laboratory-scale optics could become dramatically smaller.
Precision clocks could fit onto chips.
Quantum systems could gain more efficient optical controls.
Lasers could become far more coherent.
Data centers could move information with less energy.
And a silicon wafer only centimeters wide could allow photons to behave more like they are traveling through kilometers of extraordinary optical fiber.
Journal reference
Hao-Jing Chen, Kellan Colburn, Peng Liu, Hongrui Yan, Hanfei Hou, Jinhao Ge, Jin-Yu Liu, Phineas Lehan, Qing-Xin Ji, Zhiquan Yuan, Dirk Bouwmeester, Christopher Holmes, James Gates, Henry Blauvelt, and Kerry Vahala. โTowards fibre-like loss for photonic integration from violet to near-infrared.โ Nature, 649, 338โ344 (2026), published online January 7, 2026. The DOI was verified against the official Nature article and Crossref record and resolves to this exact publication.