A team at EPFL has built an ultrafast laser on a single photonic chip that fires femtosecond pulses at energy levels previously confined to bulky tabletop systems. The device, based on an erbium-ion-implanted silicon nitride circuit and a Mamyshev oscillator design, delivers a 176 MHz pulse train at 1.05 nJ per pulse. That performance bracket matters because femtosecond lasers drive applications from corrective eye surgery to precision spectroscopy and quantum information experiments, all of which have depended on equipment too large and expensive for widespread deployment.
Why a chip-scale femtosecond laser changes the cost equation
Tabletop ultrafast lasers occupy entire optical benches, require careful alignment by trained technicians, and cost tens of thousands of dollars. Shrinking that capability onto a photonic integrated circuit strips away most of those barriers. A chip can be manufactured using standard semiconductor fabrication steps, shipped without delicate free-space optics, and potentially mass-produced at a fraction of the cost. For fields like terahertz time-domain spectroscopy, portable medical diagnostics, and on-chip quantum photonics, access to nanojoule-class femtosecond pulses in a compact form factor could shift what is practical outside a well-funded research lab.
The EPFL result did not appear in isolation. It sits at the end of a deliberate research arc. An earlier demonstration showed that erbium-doped waveguides could provide meaningful optical gain inside a photonic integrated circuit, establishing the material platform. A separate effort then proved that femtosecond pulses could be amplified on-chip to hundreds of watts of peak power. The new work closes the loop by generating those pulses on the chip itself, rather than feeding them in from an external source.
Mamyshev oscillator performance on erbium-implanted silicon nitride
The core innovation reported in the Nature paper is the use of a Mamyshev oscillator architecture inside a photonic integrated circuit. A Mamyshev oscillator uses two stages of spectral broadening followed by offset spectral filtering to enforce mode-locking, the mechanism that forces a laser to emit ultrashort pulses instead of continuous light. In bulk optics, this design is valued for its self-starting behavior and tolerance of high pulse energies. Translating it onto a chip required erbium ions implanted directly into silicon nitride waveguides, giving the circuit optical gain at telecommunications wavelengths near 1550 nm without bonding separate III-V semiconductor materials.
The Nature paper reports a 176 MHz repetition rate and 1.05 nJ pulse energy from the integrated device. Those numbers place the chip-scale laser in the same operating regime as many commercial benchtop mode-locked oscillators used in research and industrial metrology. The repetition rate is set by the round-trip length of the on-chip cavity, while the pulse energy reflects how much gain the erbium-doped waveguides can supply before nonlinear effects or damage thresholds intervene.
The erbium gain platform itself was validated in a peer-reviewed study published in Science, which demonstrated an erbium-doped amplifier capable of supporting the signal levels needed for pulsed operation. That earlier work gave the EPFL group confidence that erbium-implanted silicon nitride could handle the optical intensities inside a mode-locked cavity, not just amplify weak continuous-wave signals. Together, the amplifier and the new oscillator show that both gain and pulse shaping can be integrated into the same material system, simplifying future chip-scale ultrafast platforms.
From a systems perspective, the Mamyshev design offers several advantages over more traditional ring or linear cavity oscillators. The dual-filter configuration naturally suppresses continuous-wave lasing and noisy pulse regimes, improving stability. It also supports higher pulse energies before nonlinear distortions degrade the waveform, which is essential when the goal is to match or exceed the nanojoule-class performance of tabletop lasers. Embedding this architecture in a low-loss silicon nitride platform further reduces cavity attenuation, allowing the erbium gain sections to operate efficiently.
Open data, open questions for chip-scale ultrafast sources
The research team deposited all experimental datasets and the laser simulation code in a public Zenodo archive referenced by the Nature paper. That transparency allows other groups to reproduce the pulse characterization, verify coherence measurements, and test the simulation against their own fabrication results. Open data deposits of this kind are still uncommon for integrated photonics demonstrations, and the decision lowers the barrier for independent validation.
Several gaps remain between the published results and a device ready for deployment. The primary records do not include fabrication yield data or wafer-scale statistics. A single device producing 1.05 nJ pulses is a strong proof of concept, but semiconductor photonics lives or dies on reproducibility across hundreds or thousands of chips per wafer run. Without yield numbers, the path from laboratory demonstration to volume manufacturing stays unclear, especially when ion implantation and precise waveguide geometries must be controlled simultaneously.
Long-term stability data is another absent piece. Femtosecond lasers used in surgery or industrial inspection must hold their pulse characteristics over thousands of hours of continuous or cycled operation. The peer-reviewed papers and the Zenodo deposit do not contain extended reliability testing or packaging studies. Those results typically come later in the development cycle, but they will determine whether the chip can survive real operating environments with temperature swings, mechanical vibration, and optical contamination. Packaging, thermal management, and integration with fiber connectors or free-space outputs will all influence the final stability envelope.
Direct head-to-head benchmarks against specific commercial tabletop lasers also do not appear in the primary literature. News coverage has drawn those comparisons, but the published data focuses on internal metrics such as pulse energy, duration, and spectral bandwidth rather than brand-name matchups. That makes sense for a first demonstration, yet end users will eventually want to know how a packaged chip compares in noise, timing jitter, and maintenance requirements to the lasers they already own.
There are also open questions about how far the architecture can be pushed. The current device operates at a repetition rate set by the on-chip cavity length, but many applications require either much higher repetition rates for frequency-comb metrology or lower rates with higher pulse energies for nonlinear microscopy and micro-machining. Adjusting the cavity design, dispersion engineering, and gain distribution could in principle move the operating point, but each change introduces new fabrication and design trade-offs.
Integration with other photonic functions is another frontier. A chip-scale femtosecond laser becomes more powerful when combined with on-chip frequency conversion, pulse compression, or beam steering. The same silicon nitride platform has already supported supercontinuum generation and low-loss routing in other experiments, suggesting that complex photonic circuits could eventually host both the source and the processing stages. However, co-locating high-intensity pulse propagation with sensitive quantum or sensing elements will demand careful layout to avoid crosstalk and damage.
Despite these uncertainties, the EPFL work marks a clear inflection point. By demonstrating nanojoule-level femtosecond pulses from an integrated Mamyshev oscillator on erbium-implanted silicon nitride, the team has shown that performance once reserved for room-filling optical benches can emerge from a chip-scale platform. The combination of a mature gain medium, a robust mode-locking scheme, and open dissemination of data and code gives other groups a concrete foundation to build on. As fabrication processes mature and reliability studies fill in the missing pieces, chip-integrated ultrafast lasers are poised to move from laboratory curiosities toward practical engines for sensing, communications, and precision measurement.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
