Text from the interactive explanation

Characterization of impulse and plasma expansion dynamics in a magnetic nozzle for laser fusion propulsion applications

T. Ogawa, K. Koba, N. Yamamoto, T. Morita

Laser fusion propulsion · magnetic nozzle experiment

A laser fusion rocket could reach Mars in about 90 days, according to a design study, compared with more than 180 days by chemical rocket. It would fly on tiny fusion explosions, caught and thrown backwards by a magnetic field. Nobody can build one yet, so this team measured how hard a magnet gets kicked by lab-sized blasts.

Fire the laser, then turn up the magnet.

coil (copper, 99 turns)

magnetic field lines

plasma glow

laser

shield

carbon target

ε compares the magnetic energy in the plasma's volume with the laser energy. The five settings are the ones the team imaged at 480 mJ. Animation: illustrative model tuned to the measured bubble size. Numbers: measured (Ogawa et al. 2026, Figs. 8–9).

t = 0 ns after the laser pulse

Slow motion: 1 µs of plasma time takes 1.8 s here, about 2 million times slower than real life.

Bubble radius toward the coil at 1 µs: 4.23 mm

Kick on the coil: 4.84 µN·s

01 · The promise

A rocket that rides on explosions

Every rocket works the same way: throw stuff backward, get pushed forward. What differs is how hard it pushes, and how frugally it uses what it throws.

Engineers rate an engine with two numbers. Thrust is how hard it pushes, in newtons. Specific impulse is how frugal it is: how many seconds each kilogram of propellant could keep up a push equal to its own weight. Higher means the exhaust leaves faster, so less propellant is wasted.

Today's options force a trade-off. Chemical rockets push hard (a few meganewtons) but burn through propellant (about 500 s), so a trip to Mars takes more than 180 days. Nuclear electric engines are frugal (about 10,000 s) but push gently (about 100 N). Nuclear fission engines sit in between.

Where would a laser fusion rocket sit on this map?

Pick one to see what the team found.

Earth to Mars, as the design studies tell it

Day 0

How the engine would work

Fire. A powerful laser hits a small pellet of fusion fuel (deuterium and tritium) wrapped in propellant.

Fuse. The fuel fuses. Its fast particles heat the propellant into plasma: gas so hot that electrons are torn off the atoms, leaving a cloud of charged particles.

Catch. The plasma balloons into the field of a large coil, the magnetic nozzle. Electric currents flow in the plasma's surface, and the field squeezes back on them.

Throw. That squeeze slows the plasma and throws it out along the coil's axis. The push back on the coil is the rocket's thrust.

The catch: nobody can build a full-size one yet, and its explosions would be far bigger than any laboratory can set off. The only way to predict its thrust is to measure small blasts and find scaling laws: rules for how the push grows as the blast grows. An earlier experiment with a weak permanent magnet found no clear trend. This team used a strong pulsed electromagnet and measured each piece carefully.

02 · The trick

A nozzle made of magnetic field

A plasma's particles are electrically charged, and charged particles can't cut straight across magnetic field lines; they spiral around them. So a plasma blast doesn't simply pass through a magnetic field. Currents flow around the plasma's surface, and those currents shove the field out of the way, carving out a bubble with almost no field inside. Physicists call it a diamagnetic cavity.

The field lines don't vanish. They get pushed aside and bunched up around the bubble, like rubber bands pressed by a fist, and bunched field pushes back harder. The bubble stops growing when the plasma's outward push matches the field's push-back.

Double the field strength. The bubble's radius becomes…

Pick one to see what the team found.

Bubble radius ×1.00 (starting setup)

Pressure-balance rule from the paper (Eq. 14): radius ∝ (energy ÷ field²)1/3, in a uniform field. Sizes are relative. Field-line spacing shows field strength.

Remember the cube root: to make the bubble twice as wide you need eight times the energy. It comes back at the end.

03 · The instrument

Weighing a whisper

To test the idea on a tabletop, the team at Kyushu University fired an infrared laser (10-nanosecond pulses of 15 to 480 millijoules) at a 0.9 mm carbon rod inside a vacuum chamber. Each pulse blasts off a puff of plasma, a stand-in for the rocket's fusion plasma. Fifteen millimetres away sits a 99-turn coil, pulsed with current so its field (0.05 to 0.55 tesla at the target) peaks just as the laser fires.

The push to be caught is tiny. Measured as an impulse (force × time, in micronewton-seconds, µN·s), the biggest kicks here were about 6 µN·s: about the momentum of a 1-milligram grain of sand dropped from head height (our comparison).

So the coil, its capacitor and its switch ride on a thrust stand: a horizontal pendulum on a torsion spring. A kick sets it swinging by micrometres; displacement sensors track the swing, and a fitted damped wave gives its size. Knocking the stand with a small pendulum fitted with a force sensor calibrates swing against impulse: 152 µN·s per millimetre for the heavy stand.

Try the three shots the team compared:

Magnet + plasma: the stand swings ±8.0 µm

Traces re-created from the shape of the paper's Fig. 5 (period ≈ 1.7 s). Pendulum swing hugely exaggerated.

The magnet's own current pulse shakes the stand too: its wires push on each other. Plasma without a field does nothing. So the nozzle's kick is the extra swing when plasma meets field.

04 · The evidence, frame by frame

Watch a real catch

A gated camera, open for just 10 nanoseconds per picture, photographed the plasma's glow from above, 200 to 1600 ns after the laser hit (480 mJ, ε = 0.37).

As the field slows the plasma, where will its glow pile up?

Pick one to see what the team found.

200 ns: a blinding spot sits on the target, with a faint glow spreading out around it.

Images reproduced from Ogawa et al. (2026), Fig. 6, CC BY 4.0, cropped into single frames; axes redrawn. Each frame shows the upper half of the blast: target at bottom (z = 0, r = 0), coil 15 mm off to the right.

05 · Result 1

Turn up the magnet

Now the central experiment. With the laser fixed at 480 mJ, the team stepped up the coil's field and photographed the plasma at 1000 ns, when it is biggest. They measured two things: how far the bright arch sits from the target along the line toward the coil's windings (50° from the axis), and the kick on the coil.

The knob is the energy ratio ε: the magnetic energy stored in the space the plasma fills, divided by the laser energy.

From ε = 0.05 to 0.21, the kick grows by 1.9 µN·s. Going on from 0.21 to 0.60 adds more than twice as much field energy. The extra kick will be…

Pick one to see what the team found.

ε = 0.37: arch at 4.23 mm from the target; kick 4.84 µN·s.

Two patterns. A stronger field gives a smaller bubble and a bigger kick. And the two changes slow down together: from ε 0.05 to 0.21 the radius drops by 0.73 mm while the kick rises 1.9 µN·s; from 0.21 to 0.60 the radius drops only 0.44 mm, and the kick again rises 1.9 µN·s. The authors suggest that a stronger field turns more of the plasma's sideways motion into motion along the axis, which is what thrust needs. What the plasma does after the first microsecond couldn't be seen.

Images: Ogawa et al. (2026), Fig. 8, CC BY 4.0, cropped. Values: Fig. 9 (read from the figure). Dashed line: the 50° direction toward the coil windings; tick: measured arch radius.

06 · Result 2

How much magnet is enough?

If more field always meant more kick, you would simply build the biggest magnet possible. But a nozzle can only redirect momentum the plasma already has, so the kick must level off somewhere. The team repeated the field sweep at five laser energies, from 15 to 480 mJ.

Switch on all five energies, then press Compare shapes . Slide the cursor to compare the field's energy with the plasma's.

At ε = 0.40 the field holds 1.0× the plasma's energy: inside the bend.

Every curve bends over around ε ≈ 0.3–0.6. Roughly 40 % of the laser's energy ends up as plasma motion, so at the bend the field holds 0.75 to 1.5 times the plasma's energy. The rule of thumb: the field must store about as much energy as the blast carries . Beyond that, extra magnet buys little, like fitting ever-stronger brakes to a car that already stops.

Values read from the paper's Fig. 7 (points up to ε = 1.3 shown; error bars omitted). The paper reports that the impulse begins to saturate at ε ≈ 0.3–0.6 for 15–240 mJ; at 480 mJ the gains are already small and it is expected to saturate with a little more field.

07 · Result 3

Bigger blasts, bigger kicks, but not evenly

For a rocket, the real question is how the kick grows as the blast grows. The team took the saturated kick at each energy (ε ≈ 0.3–0.6). Separately, they measured how much forward momentum the plasma carries in the first place, by mounting the carbon target itself on the stand with no coil: its recoil equals the plasma's initial momentum.

Double the laser energy. How many times bigger is the nozzle's kick?

Set your guess, then press Check.

The puzzle. If the kick grows faster than the momentum available, the fraction caught, the momentum efficiency (kick ÷ initial momentum), should creep up smoothly with energy, as E0.35. Does it?

Tap the button to compare the measurements with a smooth trend.

Values read from the paper's Figs. 10 and 13. The paper reports fitted exponents of 1.2 ± 0.1 (kick) and 0.85 ± 0.08 (initial momentum); guide lines use those exponents. Initial momentum counts only the component along the coil axis.

08 · The explanation

Why small blasts slip through

Compare the plasma at its biggest (1000 ns) for each laser energy, with the field near saturation. Pick an energy, then drag the probe line (or use the angle slider) and watch the brightness along it.

480 mJ along 50°: a sharp peak 4.5 mm out.

At 240 and 480 mJ the brightness has a sharp peak about 5 mm out: a thin, dense shell with a steep edge. At 112 mJ the peak is softer and closer, near 3 mm. At 58 and 15 mJ there is no peak at all, just an even glow.

That matters because the currents the field pushes against flow where the plasma's density changes steeply. A sharp edge carries a strong current and gets a strong push; a fuzzy cloud carries a weak current and gets a weak one. So the nozzle grips big blasts better. The authors suggest this comes from using the same coil at every energy: for the small blasts it was effectively too big.

09 · Back to Mars

Size the nozzle to the blast

The fix the authors point to: grow the coil with the blast. In the simulations the paper builds on, the coil's radius grows as the cube root of the plasma energy, the same cube root as the bubble in chapter 2. Then coil and bubble keep the same shape at every size, the magnetic push and the plasma's momentum both grow in step with the energy, and the fraction caught stays the same.

Slide from this experiment's sparks toward the giant pulses of a full-size rocket.

480 mJ: the coil in this experiment, 12.5 mm inner radius.

Illustrative arithmetic of the similarity rule (radius ∝ energy1/3), starting from this experiment's coil at 480 mJ. Not a design. The simulations the paper cites span 3.9 J to 62 MJ.

What this experiment brings closer

A rule for the magnet. The field should store about as much energy as the plasma; here that meant ε ≈ 0.3–0.6.

Measured growth laws. The plasma's momentum grows as E0.85, close to theory; the caught kick as E1.2.

A design lesson. A fixed coil grips small blasts poorly. The nozzle has to grow with the blast.

These are the ingredients needed to extrapolate from millijoule sparks to the enormous pulses of a rocket that might one day reach Mars in about 90 days. The paper names the next questions: what the plasma does after the first microsecond, and how coils sized to each blast perform.

Back to the nozzle ↑

Source & credits

The paper

T. Ogawa, K. Koba, N. Yamamoto, T. Morita. “Characterization of impulse and plasma expansion dynamics in a magnetic nozzle for laser fusion propulsion applications.” Vacuum 250 (2026) 115350. doi:10.1016/j.vacuum.2026.115350. Open access under CC BY 4.0.

Figures 1, 2, 4 and 12 and the camera images of Figs. 6, 8 and 11 are reproduced from the paper under the CC BY 4.0 licence. Camera images were cropped into single panels and their axes redrawn; brightness profiles in chapter 8 were re-derived from the published image colours.

Chart values were read from the paper's Figs. 7, 9, 10 and 13 and carry small reading errors.

The animated nozzle and the camera-view model are illustrative, tuned so the bright arch sits at the measured radius. They are not the authors' simulations. Rocket performance and trip times are design-study figures quoted in the paper's introduction.

The sand-grain comparison and the coil sizes in chapter 9 are our own arithmetic.