Antimatter Factories: How We Make, Trap, and Lose the Universe's Most Dangerous Substance

The usual line about antimatter is that a microgram of it would level a city, and that CERN makes the stuff every day, as if a lab outside Geneva had a vat of doomsday in the basement. The first half is wrong by six orders of magnitude, and the second is technically true while giving exactly the wrong impression. You can check both with one equation and a calculator. The true story is stranger. Antimatter is not dangerous because there is a lot of it. It is hard because there is almost none, and keeping even a hundred particles alive for a year is one of the finest pieces of engineering humans have pulled off.

§Do the arithmetic on the apocalypse

When a particle meets its antiparticle, the rest mass of both turns into other particles and energy. This is the one place in physics where $E = mc^2$ runs at full efficiency: all of the mass goes, not the tenth of a percent you get from fission. So take one microgram of antimatter and let it meet one microgram of ordinary matter. Two micrograms of mass disappear:

A ton of TNT is defined as 4.184 × 10⁹ joules.^[1]1. The conversion factor is 4.184 × 10⁹ joules per ton of TNT, listed as exact in the NIST Guide to the SI, Appendix B.8. It is a convention, not a measurement of any particular explosive. Divide:

Forty-three kilograms. That is a serious bomb, the kind that takes the front off a building and kills everyone near it. It is not a “localised apocalypse” by any definition I would accept, and it is an upper bound: it assumes every joule of the annihilation turns into blast, which it would not. To get into nuclear-weapon territory you need about a gram, a million times more. The microgram myth is off by six orders of magnitude, and the reason it survives is that “mass to energy” sounds infinite. It isn't. $c^2$ is about 9 × 10¹⁶, which is big, and a microgram is 10⁻⁹ kilograms, which is small, and when you multiply them the small one wins by more than people expect.

So the headline on this piece is wrong in the way most antimatter headlines are wrong. A microgram is a bad day. But the more interesting correction is that nobody has ever had a microgram, and at current rates nobody ever will.

§A factory that produces almost nothing

CERN does call its antiproton facility the Antimatter Factory, and it does run routinely. CERN says it delivers about 400 million antiprotons an hour.^[2]2. The production rate, the one-extra-particle-per-billion asymmetry, and the estimate that antihydrogen traps could in principle hold about 100,000 atoms all come from CERN's own explainer, “Antimatter”. That sounds like a lot until you weigh it. An antiproton has the same mass as a proton, 1.67 × 10⁻²⁷ kg, so a microgram is

and at 4 × 10⁸ an hour, that takes

of the machine running without a break. Not a production problem you solve with a bigger budget. A production problem you solve with a different civilisation.

Walter Oelert, who led the team that made the first antihydrogen atoms, ran a similar calculation in 2005: around 10¹² antiprotons produced a day, about 10¹⁶ over fifty years of operation.^[3]3. Walter Oelert, who led the team that made the first antihydrogen atoms, interviewed by Tomasz Rozek in 2005: “Setting the record straight”, CERN Courier. The 10¹²-per-day, 10¹⁶-over-50-years and 10⁻¹⁷ mbar figures are his. His own figure for the mass was “one millionth of a milligram” — a nanogram. When I redo it from his inputs I get closer to 17 nanograms (10¹⁶ × 1.67 × 10⁻²⁷ kg). Whichever you take, the unit is nanograms, over half a century, and nearly all of it was annihilated within minutes of being made, because that is what experiments do with it. Annihilate all 17 nanograms at once and you get about 3 megajoules, less than a kilogram of TNT. Every antiproton CERN has ever made, if you could somehow have kept them in one box, would add up to a large firework.

Oelert's other point is the one that kills antimatter as a fuel or a weapon: making it costs vastly more energy than you get back. The antiprotons come from slamming a proton beam from the Proton Synchrotron into a block of metal.^[4]4. CERN, “The Antiproton Decelerator”, which describes the target, the cooling cycles and ELENA's reduction from 5.3 MeV to 0.1 MeV. The collisions spray out a zoo of secondary particles; a small fraction are antiprotons; a smaller fraction of those have the right energy and direction to be caught. Everything else is heat in a target. As an energy store, antimatter is a battery you charge at an efficiency so low it rounds to zero.

§Slowing down something that wants to go fast

The antiprotons that come off the target are far too hot to keep. They are relativistic, sprayed in a spread of directions and energies, and the job of the Antiproton Decelerator — a ring of bending and focusing magnets installed in 2000 — is to tame them. It uses electric fields to brake the beam, and between braking stages it “cools” it: squeezes the spread of energies and the wobble of the trajectories so the beam stays tight as it slows. Several rounds of cooling and deceleration bring the antiprotons down to about a tenth of light speed.

That was still too fast for most experiments, which could catch only a sliver of what arrived. ELENA, a 30-metre ring coupled to the AD, takes the beam from 5.3 MeV down to 0.1 MeV, a factor of fifty, with an electron cooler to keep it dense. CERN says this raises the number of antiprotons the experiments can trap by a factor of ten to a hundred. I find this the most quietly impressive number in the whole chain. Not more antimatter: the same antimatter, wasted less.

§How to hold something that can touch nothing

Here is the constraint that makes the engineering fun. You cannot put antimatter in a container, because the container is matter. You cannot let it touch a stray gas molecule, because the gas molecule is matter. You have to hold it in empty space with fields alone, and the space has to be very, very empty.

The obvious idea, an electrostatic bowl, does not work. Earnshaw's theorem says that no arrangement of static electric fields has a stable point of equilibrium for a charge in free space: if the potential curves up in one direction, it must curve down in another. The Penning trap cheats by using two kinds of field at once. A strong, uniform magnetic field along the trap's axis makes a charged particle circle instead of drifting outwards, at the cyclotron frequency

which for an antiproton in a one-tesla field works out to (1.60 × 10⁻¹⁹ × 1)/(2π × 1.67 × 10⁻²⁷) ≈ 15 MHz. So the magnet holds the particle in the radial plane. Along the axis, where the magnet does nothing, a set of electrodes makes a shallow electrostatic saddle that pushes the particle back toward the middle. The potential is unstable radially, exactly as Earnshaw demands, and the magnetic field turns that instability into a slow circular drift. Each half of the trap covers for the other's blind spot.

Then the vacuum. Oelert quoted a gas pressure of 10⁻¹⁷ millibar for CERN's antimatter traps. Atmospheric pressure is about 1,000 millibar, so this is twenty orders of magnitude down. You do not get there with better pumps alone. You get there with cold: the trap sits at liquid-helium temperatures, where nearly every stray atom that touches a wall freezes onto it and stays. The cold wall is the pump.

The results are absurd in the best way. The BASE experiment keeps a reservoir of typically 50 to 100 antiprotons for months, and CERN reports a trapping record of 600 days.^[5]5. CERN, “BASE”: a reservoir of typically 50–100 antiprotons kept for months, a 600-day trapping record, and proton–antiproton comparisons at 16 parts per trillion (charge-to-mass) and 1.5 parts per billion (magnetic moment). A hundred antiprotons weigh about 1.7 × 10⁻²⁵ kg. That is the inventory: a handful of particles that have outlived most houseplants, in a vacuum twenty orders of magnitude below the air in your room, held by a magnet in a cryostat. In March 2026 the team took 92 of them, disconnected the trap — a roughly one-tonne package of superconducting magnet, liquid helium, vacuum chamber and backup power — loaded it on a truck and drove it about eight kilometres around the CERN site with the experiment still running.^[6]6. CERN, “BASE experiment at CERN succeeds in transporting antimatter”, 24 March 2026; distance from Science News. The magnet has to stay below 8.2 K for the entire journey. The reason for the trip is my favourite detail in the story: the machines of the Antimatter Factory itself produce magnetic noise of around a billionth of a tesla, and that is now the thing limiting BASE's precision. The factory is too loud for its own products.

§Why you can't just keep a bigger jar

Suppose production were free. Could you fill a Penning trap with a microgram? No, and the reason has nothing to do with annihilation. A microgram of antiprotons is 6 × 10¹⁷ particles, each with charge −1.6 × 10⁻¹⁹ coulombs, so about 0.096 coulombs of like charge that very much does not want to be in one place. Pack it into a ball one centimetre in radius and the electrostatic self-energy is

That is about 27 times the energy you would get by annihilating it. The thing that would blow up first is not the antimatter but the charge. Penning traps hold plasmas of charged particles, and a plasma has a maximum density set by how hard the magnetic field can push back against its own repulsion. Bulk storage of antiprotons is not a bigger version of what BASE does. It is a different problem.

The way around it is to make the antimatter neutral. Put an antiproton together with a positron and you get antihydrogen, which has no net charge and so no space-charge problem. But it also has no charge for a Penning trap to grab. All you have left is the atom's tiny magnetic moment, which a carefully shaped magnetic field can pull on only very weakly, so you can only hold atoms that are already extraordinarily cold. CERN's estimate is that current technology could in principle store up to about 100,000 antihydrogen atoms this way. That is still around 10⁻¹⁹ grams.

§What a few hundred atoms are for

None of this is about stockpiling. It is about asking antimatter a handful of very precise questions. ALPHA, set up in 2005, mixes antiprotons and positrons to make antihydrogen and holds it magnetically; in 2011 it kept atoms trapped for over sixteen minutes, long enough to start shining lasers and microwaves at them.^[7]7. CERN, “ALPHA”. The 2023 free-fall result is summarised by the University of British Columbia, one of the collaborating institutions. In 2023 it stood a trap on its end, slowly weakened the magnets at top and bottom, and counted which way the atoms escaped. About 80 percent went out of the bottom. Antimatter falls down. (If you want the long version of why that question is less silly than it sounds, there is one on gravity not being a force.) BASE, meanwhile, has measured the antiproton's charge-to-mass ratio against the proton's to 16 parts per trillion and found them the same.

Which is the point, and the problem. CERN's own summary of the mystery is that the Big Bang should have produced matter and antimatter in equal amounts, and that what we live in is the leftover of an imbalance of about one extra particle per billion. Everything you have ever touched is that rounding error. The matter–antimatter differences physicists have actually measured, in the decays of certain mesons and baryons, are real but far too small to account for it, so the search moves to the places where no one has looked carefully enough. Antihydrogen is the cleanest of these: if its spectrum, its magnetism or its falling differs from hydrogen's at the tenth decimal place, that is a crack in physics.

So far, no crack. I find that more interesting than any number of imaginary bombs. The real antimatter story is a factory that makes nanograms per lifetime, decelerates them with absurd care, holds a hundred of them for more than a year in a near-perfect vacuum, and drives them around on a truck because its own hum is too loud — all to check whether the universe made a bookkeeping error. It probably did. We just haven't found the line.