The researchers asked whether light scattered by stimulated Raman scattering (SRS) in a laser-produced plasma can be bright enough to undergo Raman scattering a second time. Theorists had long predicted this 'rescattering' in simulations for inertial confinement fusion, but nobody had observed it. Firing a 500 J, 2 ns, 526 nm laser beam (average intensity about 1e14 W/cm2, speckles near 1e15 W/cm2) into a hydrogen gas jet at the LULI2000 facility, they measured time-resolved spectra at five angles. Near the laser axis they found an extra signal at frequency close to omega0 - 2 omega_p, the sig

Fire a powerful laser into a plasma and some of the light bounces back off waves it stirs up in the electrons. For years, simulations for laser fusion predicted that this bounced light could be bright enough to bounce again. A team at the LULI2000 laser in France has now caught it happening, for the first time in an experiment.

Why a fusion laser has to worry

One route to fusion energy, inertial confinement fusion, uses powerful lasers to crush a tiny capsule of fuel. The capsule’s surface boils off into a cloud of plasma, and the blast squeezes the fuel inward until, ideally, it ignites.

But the laser light has to cross that plasma cloud first, and the plasma pushes back. Some light is scattered away before it does its job, and the scattering sets the plasma’s electrons rippling. Those ripples can whip electrons up to high speed. If fast electrons reach the fuel they warm it too early, and warm fuel is harder to squeeze.

Theorists flagged a further twist years ago: light that has been scattered once might be bright enough to be scattered again. In computer simulations, the waves from this rescattering could pass electrons up to ever higher energies. Despite many studies, nobody had caught it in an experiment. This paper is the first to do so.

To recognise it, you first need to hear the plasma’s note.

Every plasma has a note

A plasma is a gas so hot that electrons have been torn from their atoms. The heavy, positive ions barely move; the light electrons roam. Shove a band of electrons sideways and the ions they leave behind pull them back. They overshoot, swing back, and keep ringing at one particular rate: the plasma frequency , written ωp. Think of it as the plasma’s note.

The note also sets a floor for light . Light is an oscillating electric field. If it oscillates more slowly than the plasma’s note, the electrons have time to rearrange and cancel it, and the light cannot travel through. Only light above the floor gets through. Keep that floor in mind.

Raman’s trade: one note per bounce

Here is the trade at the heart of the paper. In stimulated Raman scattering , a particle of laser light gives up part of its energy to set the electrons ringing, launching a plasma wave, and leaves as a new particle of light carrying what is left. A light particle’s energy is its frequency, and frequency is colour, so the scattered light comes out redder by exactly one plasma note:

ωRaman = ω₀ − ωp

Momentum must balance too, and scattering straight back toward the laser grows fastest. The plasma wave and the scattered light feed each other and grow together; that is the “stimulated” part.

Now suppose the bounced light is bright enough to make the same trade again. It would lose a second note, landing at ω₀ − 2ωp, deep in the infrared. (Infrared is invisible to the eye; on this page it is shown in magenta, the colour past red on the colour wheel.) And a bounce of a backward bounce heads forward again, back along the laser’s own direction.

Teaching model: cold plasma, ωp = ω₀·√(n/nc). The “critical density” nc is the density at which the plasma’s note equals the laser’s own frequency, so the laser itself would be blocked.

One bounce needs the density below a quarter of critical. Two bounces need it below about a ninth. The hydrogen jet in this experiment starts out too dense for either, at roughly 0.3 of critical, and thins as the laser heats it and it expands during the 2-nanosecond pulse. So the ladder slides during every shot, and the scattered colours sweep across the spectrum. That is why the signals in the next section are streaks, not dots.

Energy and momentum conservation for Raman scattering and for its rescattering, as written in the paper:

ω₀ = ωp + ωSRS, k₀ = kp + kSRS ωSRS = ωp′ + ωRe‑SRS, kSRS = kp′ + kRe‑SRS

The plasma wave obeys ωp = ωpe(1 + 3kp²λD²)1/2, where ωpe is the electron plasma frequency and λD the Debye length. Light can only travel if its frequency exceeds ωpe, so Raman scattering needs ne < nc/4, and the rescattered light (ω ≈ ω₀ − ωp − ωp′) must also stay above ωpe. In the experiment the two scatterings happen in nearly the same patch of plasma, so ωp′ ≈ ωp and the rescattered light sits close to ω₀ − 2ωp.

In the hottest speckles both processes are estimated to grow in the absolute regime: they grow in place rather than amplifying as the light travels, which helps keep the first and second scattering close together.

Fire the laser

At the LULI2000 laser facility (École Polytechnique, France), the team focused a 500-joule, 2-nanosecond pulse of green light (526 nm) into a jet of hydrogen gas half a millimetre wide. The beam was broken into many small hot spots, or speckles: about 1.3×10¹⁴ W/cm² on average and up to roughly 6.5×10¹⁴ W/cm² in the hottest ones, typical of direct-drive fusion designs.

Light leaving the plasma was caught at five angles by spectrometers feeding streak cameras , which spread light out by wavelength in one direction and by time in the other. Here, wavelength runs left to right and time runs down. A streak that slants left is light turning bluer as the plasma thins.

Simulated streak record: wavelengths computed from the paper’s conservation laws, using plasma density and temperature traced (approximately) from the authors’ hydrodynamic simulation. Brightness is schematic and each detector has its own brightness scale, as in the paper’s figure. The real measured spectra are in Fig. 2 of the paper.

What was measured. Looking back toward the laser (170°–180°), only the ordinary backward Raman streak appears, brightest around 650–700 nm. Near the laser’s own direction (25° and 35°) three signals show up: Raman light scattered sideways, the cyan 2ωp light from pairs of plasma waves merging, and a third streak shifted into the infrared. That third streak enters the detector’s window at about 1.2–1.3 ns, just as the other two leave it, slides to around 850 nm and fades at 1.5–1.6 ns. It is weaker at 40°–50° and absent at 80°–90° and 170°–180°.

Its colour fits ω₀ − 2ωp, light that has paid the plasma’s note twice. But a colour match alone is not proof.

Four fingerprints

The team checked the newcomer four independent ways. Each check is something ordinary Raman light, or the 2ωp light, would fail.

Two notes down, same patch of plasma

Assuming the streak sits at ω₀ − 2ωp, the team worked backwards to the plasma’s note at each moment. From that they predicted where the ordinary Raman and 2ωp light must be after leaving the detector’s window. Both predictions continue the earlier streaks exactly. The two scatterings use the same note to within 5 %, the width of the streak itself.

Try it above: choose “ω₀ − 2ωp · two bounces” at the 25° detector. Back to the shot ↑

Backward of backward is forward

Drag the detector around the plasma. Curves show how bright each signal is at each angle in a particle-in-cell simulation of one hot speckle (2.2 billion simulated particles; brighter = further out, on a ×100-per-ring scale). Dots on the rim mark the real detectors: ● streak seen, ● not seen.

Which way the light vibrates

The laser’s electric field vibrates at 45°. Raman light inherits that tilt. The 2ωp light, born from merging plasma waves, vibrates horizontally. Put a polarizer in front of the forward (35°) spectrometer:

No parent, no echo

Both Raman signals stop before the laser does. As the plasma thins and heats, the Raman plasma wave gets short compared with the Debye length, the distance over which electrons screen out charges. Past kλD ≈ 0.3, fast electrons drain the wave (Landau damping). The hydrodynamic simulation puts this at about 1.55 ns for the primary Raman wave.

Measured : the signal’s colour, direction, polarization and timing. Simulated : the plasma’s density and temperature (with the radiation-hydrodynamics code TROLL), which reproduce all three wavelength tracks, and the angular pattern and relative strengths (with the particle-in-cell code Smilei, on 32 GPUs for 48 hours). Together they make the case that this is light that was Raman-scattered twice. Rescattering had only ever been seen in simulations; this is the first direct experimental evidence.

Electrons surfing the wrong way

Plasma waves are travelling ripples of electric field. An electron moving at nearly the ripple’s speed can catch it and surf, gaining energy. Ordinary Raman plasma waves travel forward, with the laser, so they push electrons forward. The rescattered light’s plasma wave travels backward.

In the team’s particle-in-cell simulation, electrons were accelerated forward and backward alike, up to about a third of the speed of light (~35 keV). To find the source of the backward ones, they re-ran it with the ions frozen, removing another way to make backward waves (the Langmuir decay instability). The backward fast electrons stayed, so rescattering makes them.

Each dot is an electron: position along the laser on the left, speed upward for forward and downward for backward. Dashed lanes mark the speeds of forward and backward plasma waves; electrons that reach a lane get carried along it. The bars on the right count electrons at each speed on a log scale. Turn the rescattering waves off and the backward tail never grows.

That matters for reading experiments. Forward and backward Raman both send fast electrons forward, and the paper notes that the energy-balance diagnostics, including those that measure energetic electrons, are read with that in mind. Rescattering sends some of them the other way. Simulations by other groups (Winjum and colleagues, 2013) also showed electrons being passed from one rescattering wave to a faster one, reaching higher energies step by step.

The echo of an echo, revisited

Here is the opening scene again, now with labels. Every arrow is something you have now seen evidence for.

Probably not rare

The conditions here, about 10¹⁴ W/cm² on average with speckles near 10¹⁵ W/cm², are typical of direct-drive fusion. The authors suggest rescattering of backward Raman light could be common there.

Some backscatter never comes back

Rescattering feeds on the backward Raman light, so detectors that measure backscatter may underestimate how much Raman scattering really happened, especially with short, high-intensity pulses.

The energy books need a second look

By moving energy into forward infrared light and backward-travelling plasma waves, rescattering (Raman, and its cousin Brillouin) may change how measurements of the energy balance, fast electrons included, should be read.

Higher intensities next

Schemes such as shock ignition use around 10¹⁶ W/cm². Simulations by other groups, cited in the paper, suggest rescattering could matter even more there.

For years, the echo of an echo lived only in simulations. Now there is a measured fingerprint, a faint infrared streak at just the right colour, angle, polarization and moment, that models of laser–plasma interaction for fusion can be checked against.

Experimental Evidence of Stimulated Raman Rescattering in Laser-Plasma Interaction

J.-R. Marquès, F. Pérez, P. Loiseau, L. Lancia, C. Briand, S. Depierreux, M. Grech, C. Riconda. Physical Review Letters 135 , 105101 (2025). doi.org/10.1103/r7th-rq38

Open preprint with supplementary material: arXiv:2505.02547 (CC BY-NC-SA 4.0). This page redraws the science rather than reproducing the paper’s figures. Plasma density and temperature (supplementary Fig. 2) and the particle-in-cell angular distribution (Fig. 5) were traced approximately from the published plots. Wavelength tracks are recomputed from the paper’s conservation laws. Streak brightness, scattering rates and the electron-surfing model are illustrative. Animated scenes are illustrations, not simulations from the paper.

Related work mentioned: B. J. Winjum et al., Phys. Rev. Lett. 110, 165001 (2013). Typefaces: Space Grotesk, Newsreader and IBM Plex Mono (SIL Open Font License).