Researchers at DTU Electro in Denmark have nearly doubled the spectral range achievable from stable, low-noise supercontinuum sources by baking a section of optical fiber at high temperature for close to an hour. The thermal treatment reshapes the fiber's dispersion profile — the parameter that decides whether a supercontinuum spans a broad but noisy band or a narrower, stable one — and the group claims the widest wavelength coverage in the low-noise class to date.

The work, published by A. Arduin, S. Rao D. S., A. B. Skov, and O. Bang in Optica (vol. 13, no. 6, pp. 1098–1103, 2026; doi: 10.1364/optica.595939), forms part of a larger DTU project called Tabletop Synchrotron, which aims to use optical fibers and supercontinuum generation to cover as much of the optical spectrum as a synchrotron facility does.

"Our group has been a world leader in supercontinuum lasers for decades now, driven by our desire for applications with an impact in the real world," says Andrea Arduin, a postdoc researcher at DTU Electro.

The physics in one paragraph. A supercontinuum arises when intense pump light drives nonlinear effects inside a fiber core, generating new wavelengths beyond the pump color. Dispersion governs how those colors propagate and interact, and it imposes a hard trade-off: engineers can have a wide, noisy, unstable spectrum or a narrower, coherent, repeatable one. The DTU team sidesteps part of that trade-off by heating a short fiber section for roughly an hour, which permanently alters the glass structure and its dispersion. The counterintuitive part is geometric: thermal treatment makes the fiber core larger, not smaller. Until now, tapering the fiber down was the only practical way to engineer local dispersion.

"Cooking it does the opposite process because it makes the fiber core larger, which isn't normally possible," says Arduin. "We've used the technique to improve supercontinuum generation, but now I'm curious to see if someone else comes up with another application for this."

The effect of thermal exposure had been ignored even though the underlying technique was used for other purposes. Simulations predicted success, and the first experimental confirmation prompted a fitting email. "I sent Ole an email with a subject line: Preliminary Eureka!" Arduin recalls.

Why noise matters for instrument builders. Every application that detects weak optical signals pays a direct penalty for source noise. The DTU group targets optical coherence tomography, where source noise maps straight into image noise. Gas-trace spectroscopy is another case: the signal is weak, and with a noisy supercontinuum, measurements can stretch to hours. A stable, broadband source cuts that integration time.

The current source works as delivered, and silica's transmission loss now sets the ceiling on spectral extension. Arduin identifies fluoride and chalcogenide fibers as the interesting next materials, while acknowledging the transition will be challenging.

The group's near-term roadmap points at the mid-infrared, where several low-noise supercontinuum systems are in development. Mid-IR coverage matters because molecular fingerprints concentrate there, and broadband coherent tabletop sources remain scarce relative to the visible and near-IR.

For labs specifying supercontinuum sources, the development raises a practical question: post-thermal processing of standard fiber offers a route to wide, low-noise coverage without exotic fiber designs, so buyers should ask vendors whether thermal dispersion engineering — not just tapering — figures in their product roadmaps.