A team led by EPFL has built a laser on a single photonic chip that fires femtosecond pulses with energy and speed that, until now, required bench-top instruments occupying an entire optical table. The device produces pulses compressible to approximately 147 femtoseconds at pulse energies of roughly 1.05 nanojoules and a repetition rate near 176 MHz. Published in Nature, the result represents the first wafer-scale manufactured chip-scale femtosecond mode-locked laser, a development that could carry ultrafast optics out of specialized laboratories and into portable instruments for metrology, imaging, and sensing.
Why a 147-femtosecond chip laser changes the ultrafast optics equation
Femtosecond lasers sit at the heart of precision frequency combs, biological imaging, and industrial micromachining. The systems that deliver those pulses have historically been large, expensive, and sensitive to vibration. Shrinking one onto a chip while preserving performance has been a goal for roughly two decades, dating back to early demonstrations of erbium-doped microdisk lasers on silicon. Those early devices proved that rare-earth gain media could work inside photonic circuits, but they fell far short of the pulse energy and duration that tabletop titanium-sapphire or fiber oscillators could reach.
The EPFL team closed that gap by adopting a Mamyshev oscillator architecture on an erbium-ion-implanted silicon nitride photonic integrated circuit. The Mamyshev design uses two spectrally offset bandpass filters separated by nonlinear gain sections, which together enforce stable self-starting mode locking and allow high pulse energies without the saturable absorbers that limit many chip lasers. The cavity length is folded on-chip using waveguide Bragg grating offset filtering and a specific pump configuration, details described in the team’s openly accessible technical preprint.
The practical consequence is direct: a device small enough to integrate into handheld or rack-mounted equipment can now generate pulses that match what researchers previously obtained only from instruments costing tens of thousands of dollars and requiring dedicated vibration-isolated tables. For engineers designing portable optical coherence tomography scanners, compact lidar units, or field-deployable spectroscopy tools, this changes the design space significantly. Instead of routing fiber from a remote, delicate source, designers can place the femtosecond engine millimeters away from detectors, modulators, or nonlinear conversion stages.
Performance numbers and the manufacturing path behind them
Three figures anchor the claim. The laser output is compressible to approximately 147 fs, pulse energies reach approximately 1.05 nJ, and the repetition rate sits near 176 MHz, according to the peer-reviewed Nature study. Those numbers place the chip laser in the same performance class as commercial erbium fiber oscillators, which typically produce pulses between 100 and 200 fs at sub-nanojoule to low-nanojoule energies. Critically, the EPFL device demonstrates clean spectral profiles and low-noise operation indicative of genuine mode locking rather than noisy Q-switching or relaxation-oscillation artifacts.
The distinction is manufacturing method. The EPFL device is fabricated on a silicon nitride wafer using processes compatible with standard photonic foundry workflows. The European Commission’s Horizon Europe PI-MOLL project record describes the effort as the first wafer-scale manufactured chip-scale femtosecond mode-locked laser and lists intended applications in metrology, imaging, and spectroscopy. Wafer-scale fabrication matters because it means hundreds or thousands of nominally identical laser chips could be produced in a single run, driving unit costs down in ways that hand-assembled fiber lasers cannot match. It also opens a path toward co-integrating the laser with modulators, filters, and detectors on the same substrate.
Independent work by other groups has already shown that chip-based photonic circuits can handle femtosecond pulses at serious power levels. Separate peer-reviewed research on integrated amplifiers, documented for example in a PubMed-indexed report, demonstrated femtosecond pulse amplification on a chip reaching hundreds of watts of peak power. That finding, while not from the same EPFL group, confirms that the broader silicon photonics ecosystem can support the high peak intensities that femtosecond operation demands. Together, these results suggest a roadmap from integrated oscillators to fully integrated ultrafast systems.
Open questions around stability, tunability, and real-world packaging
The published results establish performance at a single operating point. What they do not yet show is how the laser behaves over extended continuous operation, across temperature swings, or after packaging into a commercial module. Long-term stability data and failure-rate statistics under sustained use are absent from the primary sources. For any manufacturer considering this technology for a product, those numbers will determine whether the chip laser is a laboratory proof of concept or a production-ready component.
A second open question involves repetition-rate flexibility. The Mamyshev oscillator architecture, in principle, should allow tuning of the repetition rate by adjusting the on-chip cavity length or pump parameters. Whether the EPFL design can achieve tuning over a range wide enough to be useful, for instance across tens of megahertz, without requiring external cavities or feedback electronics, has not been demonstrated in the published record. Direct spectral and timing measurements on future fabricated batches will be needed to confirm or rule out that capability.
Thermal management is likely to be another key factor. Silicon nitride and silicon substrates conduct heat differently from bulk glass fiber, and the localized heating from pump absorption and nonlinear propagation can shift refractive indices and resonance conditions. The preprint provides extended figures and detailed pump configuration data that the journal version does not include, which means the full engineering picture is split across two documents. Researchers attempting to replicate or build on the work will need both, and they will have to add their own measurements of temperature sensitivity, mode-hop behavior, and environmental robustness.
Packaging will ultimately decide how far this technology moves beyond the lab. In a research setting, fiber pigtails, active alignment stages, and open-air heat sinking are acceptable. In a commercial context, the same chip must survive board-level soldering, mechanical shock, and years of on-off cycling. Hermetic sealing, integrated temperature control, and standardized optical and electrical interfaces will all add complexity. None of those engineering layers appear in the current publications, leaving a significant translation step from demonstrated physics to deployable hardware.
What comes next for integrated femtosecond sources
Even with those caveats, the EPFL result marks a turning point for integrated ultrafast photonics. A wafer-scale manufacturable chip laser operating in the 150-fs, nanojoule, hundred-megahertz regime brings several long-discussed applications closer to reality. Frequency-comb-based metrology could migrate from national labs to factory floors if the light source fits into a compact module. Portable biomedical imaging systems might incorporate on-board femtosecond excitation without relying on bulky fiber lasers. Quantum photonics platforms could benefit from stable, chip-level pulsed pumping for entangled photon generation and manipulation.
On the research side, the demonstration is likely to spur efforts to push performance further: shorter pulses through broader on-chip bandwidth, higher energies via improved dispersion engineering and larger mode areas, and different wavelength bands by swapping gain media. Competing platforms, including indium phosphide and thin-film lithium niobate, may see renewed interest as alternative hosts for integrated mode-locked oscillators. Each material system offers its own balance of nonlinearity, loss, and manufacturability.
For now, the most important message is that the core physics of high-performance femtosecond generation no longer demands a room-filling optical setup. By showing that a Mamyshev oscillator can be realized in an erbium-implanted silicon nitride circuit with wafer-scale processes, the EPFL team has moved ultrafast lasers into the same integration conversation as modulators and detectors. The remaining questions-stability, tunability, and packaging-are difficult but fundamentally engineering problems. As those are solved, femtosecond pulses may become as commonplace on chips as continuous-wave telecom signals are today, reshaping what designers can assume about light sources in everything from precision clocks to handheld diagnostic tools.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
