There's a thing you can build in a garage that, if you do it right, makes a tiny star.
I'm not exaggerating. People do this. They post the videos. The device is called a Farnsworth fusor, and it's been quietly humming away in basements since 1964. The kid who first patented it was nineteen. The whole apparatus fits on a workbench.
I should be honest up front: a fusor doesn't make useful energy. Nobody has ever gotten one to produce more power than it consumes, and there are deep reasons to think nobody ever will. But it does make real fusion reactions. Real neutrons. Real, controlled bursts of "we just smashed two hydrogen nuclei together hard enough to fuse them into helium." In a glass jar. On a workbench. In some guy's basement in Ohio.
I built the simulator above to figure out why this works, why it doesn't quite work enough, and — the part that surprised me most — what happens when you change the shape of the cage in the middle. This is the story of what I found.
The bulb
The recipe is hilariously short. Take a glass vacuum chamber, pump almost all the air out, put a tiny puff of deuterium gas back in. (Deuterium is just hydrogen with an extra neutron. It's stable. You can mail-order it.) Suspend a small metal cage inside the chamber — a few centimeters across, made of thin tungsten wire bent into a sphere or a geodesic or really whatever shape you feel like. Apply a very negative voltage to the cage. Tens of thousands of volts, negative. The chamber walls stay grounded. Turn the lights off.
If you did it right, you see a soft pink glow filling the bulb. The cage in the middle is also glowing. And if you look carefully, you can see bright beams of light shooting out from inside the cage, through the gaps between the wires, like spokes on a glowing bicycle wheel.
That glow is plasma. Those beams are ions, moving at roughly a million meters per second. And a tiny number of them are colliding hard enough in the middle to fuse.
What is actually going on?
The field, or, why ions roll downhill
Here's the trick.
The cage has thirty thousand volts on it. Negative thirty thousand. Negative voltage means "this place repels electrons and attracts anything positive." Hydrogen ions are positive things.
So if you take a deuterium ion — somewhere in the chamber, doesn't matter where — and let it go, it does what any attracted thing does. It falls. Toward the cage. Picking up speed as it goes.
By the time it reaches the cage, it has roughly thirty thousand electron-volts of kinetic energy. (One eV is what an electron picks up falling through one volt. Thirty thousand volts means thirty thousand eV. Per ion. There's no free lunch — that energy comes straight out of the wall socket.) This is fast. Faster than escape velocity from the Sun.
Now here's the part that took me a while to wrap my head around: the ion doesn't stop at the cage.
The cage isn't a solid sphere. It's a few wires arranged into the shape of a sphere, with most of the volume being empty space between wires. So the ion screams toward the cage, threads the needle through one of the gaps, and finds itself on the other side of the cage, still moving fast.
But now the cage is behind it. Which means the cage is still pulling it back. So the ion slows, stops, reverses, and falls back through the cage again. And again. And again. Each oscillation taking about a microsecond.
Until it either (a) clips a wire and dies, or (b) collides with another ion in the dense middle hard enough to actually fuse.
The dance
Now imagine doing this not with one ion but with a hundred million.
They're all falling toward the same point — the center of the cage. They all reach the middle at roughly the same time, all moving very fast, all converging on the same tiny volume of space.
That's where the fusion happens. When two deuterium ions slam into each other head-on at high enough energy, occasionally — very, very occasionally, like one in a million collisions — they tunnel through their mutual Coulomb repulsion, stick together, become a heavier nucleus, and release a neutron. That neutron carries off some of the energy and goes flying out of the chamber.
That's fusion. The thing the Sun does. Just at much smaller scale.
The whole reason the cage works is because of its gaps. If the cage were a solid sphere, ions would hit it and stop. The fact that it's mostly empty space lets ions oscillate through it indefinitely. Each ion gets thousands of chances to find a partner before something stops it.
Which means the shape of the cage really, really matters. And this is the part nobody told me. I had to build a simulator to actually see it.
Energy donuts versus cages
Quick detour, because if you've heard of fusion you've probably heard the word "tokamak" and you're wondering how it relates.
There are two big families of fusion devices on Earth.
Energy donuts (tokamaks) use magnetic fields to confine plasma in a torus — a literal donut shape. The plasma swirls around the donut, steered by enormous superconducting coils. ITER, the big international fusion project, is a tokamak. SPARC, the MIT spinoff, is a tokamak. These are the billions-of-dollars, will-probably-work-someday projects.
Cages (electrostatic confinement, like our fusor) use voltage instead of magnets. They are absurdly cheaper, simpler, and inherently leaky — the cage wires keep absorbing ions, which is the fundamental reason a fusor will (probably) never give net energy. But they work, you can build one for a few thousand dollars, and you can change the cage shape on a Saturday afternoon to see what happens.
Both shapes are honest attempts at the same problem: how do you get a bunch of positively-charged particles to stay close enough to each other for long enough that some of them fuse, despite Coulomb's law screaming at them to fly apart?
The donut crowd answers with magnets. The cage crowd answers with electricity. Both work. Neither has won.
Shape matters: what I found
Here's the simulator's whole reason for existing: the cage doesn't have to be a sphere.
People have built fusors with cages shaped like single rings, helices, double helices, hourglasses, cones, cubes, and tiny toroids. They all work. They all behave differently. I implemented all of these in code and watched.
Geodesic sphere. The classic. Ions converge to a tight focus in the middle. You see spokes of light through every gap. This is called star mode. It's what most fusors look like in operation.
Single ring. The focus collapses to a line through the middle of the ring instead of a point. Beautiful, but the ions oscillate along the axis instead of converging spherically. Less central density. Less fusion.
Helix and helical cross. The ions get an angular twist on the way in. The focus is fuzzier, the dynamics are wilder, and the cross-helix gives some funky standing-wave patterns I genuinely did not expect.
Hourglass and cone. Asymmetric cages. These can lock into tight jet mode — instead of star-mode spokes going every direction, all the plasma streams out one preferred opening as a single collimated beam. It is striking on camera.
Cube. Surprisingly bad. Ions get hung up at the edges of the faces. Lots of grid losses.
Toroid. Basically a tiny tokamak grid. Star mode forms in two ring-shaped focal zones instead of one central point. The "donut shape inside a cage" concept does work in miniature, but you lose the central convergence that makes the spherical fusor good at making neutrons.
The biggest lesson: there is no single best shape. Every cage trades focus quality against grid losses against where in the chamber the energy goes. The reason every amateur uses a sphere or geodesic is because, for the basement-budget version of this experiment, central convergence beats everything else.
What the sim does, and doesn't
Quick honesty section, because I owe it to you.
The sim solves real Coulomb's law. The electric field around each charged wire segment is computed from the textbook closed-form expression for a finite line of charge — no shortcuts, no approximations of the field shape. The ions move under real Newtonian mechanics with a symplectic integrator that conserves energy properly over long runs.
What the sim doesn't model: space charge (the ions don't repel each other), ionization (real fusors continuously make new ions out of the neutral gas), electrons (the sim only tracks positive ions), and the actual fusion reactions themselves (zero neutrons are produced — you'd need a quantum cross-section calculation for that). It also uses reduced units, so you can't read off "this is 30 kV deuterium" from the screen — only relative behavior between geometries is meaningful.
This is a qualitative tool. It tells you why one cage shape focuses ions tighter than another, and what star mode and jet mode and halo mode look like from the inside. It does not predict how many neutrons per second a real build would produce. For that, you need much heavier machinery — particle-in-cell codes that take days of supercomputer time to run.
Why this matters (a little)
Net-positive fusion is not coming out of a hobbyist's basement. The energy-donut crowd will get there first, if anyone does, and even they have to keep moving the date.
But understanding why a fusor works gives you a free pass into thinking about every other fusion device on Earth. Magnetic confinement, inertial confinement, the tabletop neutron sources used in oil-well logging, the giant lasers at NIF in California — they all reduce to the same single question, which is the question I started this thing with: how do you get charged particles to stick together long enough to fuse, when Coulomb's law is doing everything it can to fling them apart?
A fusor is the simplest possible answer to that question: just pull them together with electricity. The whole rest of fusion engineering is asking "okay, but cheaper, and without the cage absorbing all of them."
Play with it
The simulator above is open source. Pick a cage shape. Crank the voltage. Watch the modes change. Fork it, try a shape I didn't think of, post your screenshots.
If you build something interesting, tag me. I want to see what star mode looks like inside a Klein bottle.
— Ali