Antihydrogen
An antiproton and a positron make an atom of antihydrogen. CERN’s ALPHA collaboration traps about a hundred, then slowly lets them go: from a trap symmetric top to bottom, more leave through the bottom.
Scroll
It falls
In September 2023 the ALPHA collaboration reported in Nature that antihydrogen atoms are pulled down by the Earth’s gravity. The measured acceleration is a = (0.75 ± 0.13 ± 0.16) g: the first uncertainty is statistical and systematic, the second comes from the simulation, and g is 9.81 m/s². Within the uncertainties this agrees with the 1 g of ordinary matter.
From this the paper rules out “antigravity” of magnitude 1 g, pointing up. The precision is still rough: about 20 % of g, in CERN’s words. For ordinary matter the same question has been tested in Earth orbit to the order of 10⁻¹⁵.
(0.75 ± 0.13 ± 0.16) g · 1 g antigravity ruled out · 2023
Antimatter
Every particle has an antiparticle with the same mass and the opposite charge. In 1932 Carl Anderson saw the positron in cosmic rays, which won the 1936 Nobel Prize in Physics; in 1955 Chamberlain, Segrè and colleagues made antiprotons at Berkeley’s Bevatron accelerator, which won the prize in 1959.
A particle that meets its antiparticle annihilates into something else: an electron and a positron at rest give two photons of 511 keV each; an antiproton meeting a nucleus gives a few pions, on average about 2.6 of them charged. An antiproton and a positron together make an atom of antihydrogen.
positron 1932 · antiproton 1955 · e⁺e⁻ → 2 × 511 keV
Holding it
Charged antiparticles can be held by electric and magnetic fields, but stray fields push them far harder than gravity: the paper works out that the Earth’s pull on a proton equals an electric field of only about 10⁻⁷ V per metre. To see gravity, the antimatter has to be neutral: antihydrogen atoms.
In 1995 CERN’s LEAR made the first antihydrogen: 9 atoms, each lasting about 40 ns. In 2002 ATHENA and ATRAP made cold antihydrogen, but it soon hit the walls. In 2010 ALPHA trapped antihydrogen for the first time, 38 atoms; in 2011 it held them for 1,000 s; in 2025 it reported more than 15,000 held at once.
1995: 9 atoms · 2010: 38 · 2011: 1,000 s · 2025: more than 15,000
A magnetic bottle
Antihydrogen has no charge, but it is a tiny magnet. The atoms that can be trapped are pushed towards weaker field. So the trap is a field weak in the middle and strong all round: an octupole coil makes it stronger towards the wall, and two mirror coils raise it at the ends. Magnetic traps also appear in the Bose–Einstein condensate plate.
In the bottom trap of ALPHA-g the atoms can move in a cylinder 4.4 cm across and 25.6 cm tall. The background field is 1 T, about 1.74 T at the ends. The trap is about 0.5 K deep: a depth, as an energy in temperature units, not the trap’s temperature; only atoms with less energy stay. The coils are superconducting, in liquid helium near 4 K.
4.4 cm across · 25.6 cm tall · 1 → 1.74 T · about 0.5 K deep
Cut open: a trap on end
ALPHA-g stands upright. In the paper’s drawing the main solenoid is about 3 m tall, and the whole apparatus, from the entrance at the foot to the detectors at the top, is several metres. The main solenoid, outermost, has a 510 mm bore and makes a vertical field of 1 T.
Inside it, 64 plastic scintillator bars 2.6 m long form a ring; inside them, the radial time projection chamber, from 10.9 to 19.0 cm in radius and 230 cm tall; then the insulating vacuum, the liquid helium and the superconducting coils; and at the centre a stack of ring electrodes round the ultra-high vacuum. The apparatus has three trapping regions; only the bottom one was used. The tower here is at 1:20, with that region enlarged and cut open beside it.
1 T · 510 mm bore · 64 scintillator bars · rTPC r = 10.9–19.0 cm
Two ingredients
Antiprotons are knocked out of a target: CERN’s Proton Synchrotron fires 26 GeV protons at iridium, about 300,000 protons for each antiproton collected. The 182 m Antiproton Decelerator (AD) slows them to 5.3 MeV in about 100 s, and the 30.4 m ELENA ring takes them to 100 keV. ELENA sends ALPHA-g about 7.5 million every 120 s, and about 500,000 are caught.
Positrons come from radioactive sodium-22, with a half-life of 2.6 years. Solid neon slows them, and a Surko-type accumulator filled with nitrogen gathers them; each mixing uses about 3 million. Both kinds enter the tower at the foot, one from each side.
26 GeV → 5.3 MeV → 100 keV · 7.5 million every 120 s, about 500,000 caught · sodium-22
Making atoms
The charged particles are held in a Penning–Malmberg trap: the vertical field keeps them from moving sideways, and voltages on the electrodes stop them at either end. Electrons loaded in first cool the antiprotons. The positrons and the antiprotons each form a cloud; their size and density set, they are brought together in the middle electrodes, and some combine into antihydrogen.
Most new atoms are too “hot” to stay: in the 2017 data about 1 in 2,000 was trapped. ALPHA-g traps only a few per mixing, so it gathers them: 50 rounds, about 4 hours, about 100 atoms. In 2024 ALPHA-2, with positrons cooled by beryllium ions, trapped 163 in a 4-minute cycle, about 8 times its previous best.
about 3 million positrons a round · 50 rounds, about 4 hours, about 100 atoms
Seeing it annihilate
When an atom touches the wall it annihilates with a nucleus there and sends out a few charged pions. The radial time projection chamber holds 70 % argon and 30 % carbon dioxide; a passing pion frees electrons, which drift outwards to the outer wall and are recorded by 256 anode wires and 18,432 pads. Tracks traced back meet where the annihilation was, to about 2 cm in height.
The main nuisance is cosmic rays, about 70 a second. The 64 scintillator bars round the chamber help judge each event’s shape, and a machine-learning classifier cuts the cosmic rays about 350-fold.
256 wires · 18,432 pads · about 2 cm in height · cosmic rays cut about 350-fold
Its light
By CPT symmetry, antihydrogen’s energy levels should be exactly hydrogen’s. Hydrogen’s most precise spectral line is 1S–2S: two ultraviolet photons of 243 nm lift the atom to its first excited state. These spectra were all measured in ALPHA-2, the other trap next door.
In 2018 ALPHA measured antihydrogen’s 1S–2S frequency as 2,466,061,103,079.4 ± 5.4 kHz, in agreement with hydrogen’s value for the same magnetic field to 2 × 10⁻¹². The ground state’s hyperfine splitting was measured in 2017 and to 4 ppm in 2026; in 2018 ALPHA also saw the Lyman-α transition at 121.6 nm. No difference between antihydrogen and hydrogen has been found so far.
1S–2S: agreement to 2 × 10⁻¹² (2018) · hyperfine splitting 4 ppm (2026)
Cooled by light
Colder atoms give sharper measurements. In 2021 ALPHA laser-cooled antihydrogen for the first time, with Lyman-α light at 121.6 nm: an atom running into the light absorbs photons more readily and is pushed back. The median transverse energy fell from 15.1 to 1.3 µeV, more than tenfold. Laser cooling also appears in the Bose–Einstein condensate plate.
Once cooled, the 1S–2S line narrowed from 57.6 to 14.4 kHz, about a quarter. The gravity paper says colder atoms would make the escape curve in the next chapter steeper and the measurement more precise, and counts laser cooling as a step that way.
15.1 → 1.3 µeV · line 57.6 → 14.4 kHz (2021)
Weighing it
In the trap gravity is very weak: for a hydrogen atom, the 25.6 cm of height is worth only about 0.3 mK, against a well about 0.5 K deep, over a thousand times more. So the field is not switched off at once: the currents in the top and bottom mirror coils are lowered together over 20 s, and in the last 10 s the atoms leak out of whichever end is lower and annihilate on the wall. Simulations say that ordinary hydrogen, with equal fields at the two ends, would leave through the bottom about 80 % of the time.
A tiny difference between the two end fields can pass for gravity: 1 g is a difference of only 4.53 × 10⁻⁴ T. ALPHA-g added such a bias on purpose, from −3 g to +3 g, six or seven trials at each, over about 30 days. With no bias, after subtracting cosmic rays, it counted 94.5 going down and 36.7 going up, about 72 % through the bottom; the balance point lies near −1 g.
released over 20 s · 1 g ≙ 4.53 × 10⁻⁴ T · no bias: about 72 % down
What it does not say
The result shows that antimatter falls, with an acceleration consistent with 1 g; it finds no difference between antimatter and matter, and does not explain why the universe is left almost entirely matter. Only antigravity of 1 g is ruled out, with an uncertainty of about 20 %, far coarser than the tests on ordinary matter. ALPHA-g’s design goal is better than 1 %, and the paper also imagines fountain-style interferometry in a vertical trap.
Nor is antimatter a source of energy. By this page’s arithmetic, 1 g of antimatter annihilating with 1 g of matter releases about 1.8 × 10¹⁴ J, some 43 kilotonnes of TNT; at the rate CERN gives, up to about 3,000 an hour, making 1 ng would take about 23 million years. The more than 15,000 atoms of 2025, all annihilated, give about 4.5 µJ.
about 20 % vs 10⁻¹⁵ for matter · 1 ng would take about 23 million years
Taken apart
The scintillator, the time projection chamber, the insulating-vacuum wall, the helium vessel, the mirror coils, the two octupoles, the ultra-high-vacuum tube and the electrodes rise out of the section one by one.
Point at a part (tap it on a phone) and it lifts on its own, with its name; the rest fade.
Let them go
The lever is the bias: the extra field at the top mirror coil G over the bottom one, A, in the g it imitates. “Release” lowers both coils’ currents over 20 s; the atoms leak out one by one and annihilate on the wall, and the readouts count up and down.
This is this page’s simplified model, not the paper’s simulation: each atom leaves over the lower of the two barriers, and the trap’s three-dimensional effects are stood in for by a random spread about 1.4 g wide. The dots on the chart are the down-shares from Table 1’s counts. With “no gravity” or “antigravity” the model’s curve moves right; the measured dots stay put.
Try thisSet the bias to −1 g and let the atoms go: how many leave up, how many down? Switch to “antigravity”: which way does the curve move?
What the model leaves out
The tower is at 1:20, measured off the paper’s Fig. 1a and scaled by the time projection chamber’s published radii; the only published sizes are the trap (4.4 × 25.6 cm), the chamber, the 64 scintillator bars and the main solenoid’s 510 mm bore, and the rest is read off the drawing to within about 10 %. The vacuum chambers, valves, pumps and supports at the top and foot are schematic. The enlarged section is at 1:4.2; the coils’ heights are from Fig. 1b, while the radii of the coils, electrodes and walls are estimates; the electrodes are drawn gold-plated, as ALPHA’s are. The windings on the octupoles are a drawn pattern. The section’s top and bottom faces are the cut, painted red. The liquid helium and the gas are not drawn. The atoms, positrons, antiprotons and annihilations are drawn far larger and slower than they are: a real atom moves at about 100 m/s, the pions near the speed of light; the tracks’ curves and the lit bars are schematic. The field curve comes from this page’s simple coil model, scaled to the paper’s 1.74 T at the ends. The laboratory is a toy model, not the paper’s simulation.
Common misconceptions
“Antimatter falls up” is wrong: ALPHA-g ruled out antigravity of 1 g. “Antimatter’s gravity is exactly the same as matter’s” claims too much: the result agrees with 1 g to about 20 %. a = 0.75 g does not mean three quarters of the pull; it is within the uncertainties. The atoms were not dropped from a height: they move back and forth in the trap and leak out of the ends over 20 s, and what is counted is which end they leave by. 0.5 K is the trap’s depth, not a temperature; the walls are near 4 K. The magnetic trap holds antihydrogen by its magnetism, not by charge; it has none. The decelerator does not make antiprotons: they come from the target, and the decelerator slows them. Antiproton annihilation gives mostly pions, not gamma rays. The spectra were measured in ALPHA-2, the gravity in ALPHA-g. No difference between antihydrogen and hydrogen has been measured so far, and these measurements do not explain why the universe holds more matter than antimatter. Antimatter is not an energy source: making it takes far more energy than it gives back.
Common questions · 5
- Does antimatter fall down?
In September 2023 the ALPHA collaboration reported in Nature that antihydrogen atoms are pulled down by the Earth’s gravity.
01It falls ↑- How big is the space the antihydrogen is held in?
In the bottom trap of ALPHA-g the atoms can move in a cylinder 4.4 cm across and 25.6 cm tall.
04A magnetic bottle ↑- How weak is gravity inside the magnetic trap?
In the trap gravity is very weak: for a hydrogen atom, the 25.6 cm of height is worth only about 0.3 mK, against a well about 0.5 K deep, over a thousand times more.
11Weighing it ↑- Where do the antiprotons come from?
Antiprotons are knocked out of a target: CERN’s Proton Synchrotron fires 26 GeV protons at iridium, about 300,000 protons for each antiproton collected.
06Two ingredients ↑- Is antihydrogen any different from hydrogen?
No difference between antihydrogen and hydrogen has been found so far.
09Its light ↑
Sources
- Anderson et al. (ALPHA, 2023): Observation of the effect of gravity on the motion of antimatter, Nature 621, 716 (a = (0.75 ± 0.13 ± 0.16) g; Table 1; Fig. 1; 20 s; 4.4 × 25.6 cm; rTPC and scintillator)
- CERN (2023): ALPHA experiment at CERN observes the influence of gravity on antimatter (about 20 %)
- Amole et al. (ALPHA, 2014): The ALPHA antihydrogen trapping apparatus, NIM A 735, 319 (octupole and mirror coils; gold-plated electrodes; sodium-22 and the accumulator; silicon detector)
- Baker et al. (ALPHA, 2023): Design and performance of a novel low energy multispecies beamline for an antihydrogen experiment, Phys. Rev. Accel. Beams 26, 040101 (510 mm solenoid bore; entry from below)
- Hori, Walz (2013): Physics at CERN’s Antiproton Decelerator, Prog. Part. Nucl. Phys. 72, 206 (26 GeV protons; iridium target; antiproton yield)
- Oelert (2015): The ELENA project at CERN, Acta Phys. Pol. B 46, 181 (AD 182 m; ELENA 30.4 m; 5.3 MeV → 100 keV)
- Ponce et al. (2022): ELENA – from commissioning to operation, IPAC2022
- Ahmadi et al. (ALPHA, 2017): Antihydrogen accumulation for fundamental symmetry tests, Nat. Commun. 8, 681 (about 14 trapped per cycle, about 1 in 2,000)
- ALPHA Collaboration (2025): Be⁺ assisted, simultaneous confinement of more than 15000 antihydrogen atoms, Nat. Commun. 16, 10106 (163 per 4 minutes; about 8 times)
- Andresen et al. (ALPHA, 2010): Trapped antihydrogen, Nature 468, 673 (38 atoms)
- Andresen et al. (ALPHA, 2011): Confinement of antihydrogen for 1,000 seconds, Nat. Phys. 7, 558
- Ahmadi et al. (ALPHA, 2016): An improved limit on the charge of antihydrogen from stochastic acceleration, Nature 529, 373
- Ahmadi et al. (ALPHA, 2017): Observation of the 1S–2S transition in trapped antihydrogen, Nature 541, 506 (243 nm)
- Ahmadi et al. (ALPHA, 2017): Observation of the hyperfine spectrum of antihydrogen, Nature 548, 66
- Ahmadi et al. (ALPHA, 2018): Characterization of the 1S–2S transition in antihydrogen, Nature 557, 71 (2 × 10⁻¹²)
- Ahmadi et al. (ALPHA, 2018): Observation of the 1S–2P Lyman-α transition in antihydrogen, Nature 561, 211
- Baker et al. (ALPHA, 2021): Laser cooling of antihydrogen atoms, Nature 592, 35 (15.1 → 1.3 µeV; 57.6 → 14.4 kHz)
- ALPHA Collaboration (2026): Four ppm measurement of the antihydrogen ground-state hyperfine splitting, Nature 653, 1022
- CERN (1996): First atoms of antimatter produced at CERN (9 atoms, about 40 ns)
- Amoretti et al. (ATHENA, 2002): Production and detection of cold antihydrogen atoms, Nature 419, 456
- Gabrielse et al. (ATRAP, 2002): Background-Free Observation of Cold Antihydrogen…, Phys. Rev. Lett. 89, 213401
- Anderson (1933): The Positive Electron, Phys. Rev. 43, 491
- Chamberlain, Segrè, Wiegand, Ypsilantis (1955): Observation of Antiprotons, Phys. Rev. 100, 947
- Nobel Prize: The Nobel Prize in Physics 1936 (Carl D. Anderson)
- Nobel Prize: The Nobel Prize in Physics 1959 (Emilio Segrè, Owen Chamberlain)
- Parthey et al. (2011): Improved Measurement of the Hydrogen 1S–2S Transition Frequency, Phys. Rev. Lett. 107, 203001
- Touboul et al. (MICROSCOPE, 2022): Final Results of the Test of the Equivalence Principle, Phys. Rev. Lett. 129, 121102 (10⁻¹⁵)
- Bertsche (2018): Prospects for comparison of matter and antimatter gravitation with ALPHA-g, Phil. Trans. R. Soc. A 376, 20170265 (the 1 % goal)
- CERN: Antimatter (annihilation; up to about 3,000 an hour)
- CERN: The Antiproton Decelerator
Data
The tower follows the ALPHA collaboration’s drawing of the ALPHA-g apparatus (Anderson et al., Nature 621, 716 (2023), Fig. 1a, CC BY 4.0), scaled by the radial time projection chamber’s published radii; the enlarged trap follows Fig. 1b’s coil heights. The published sizes are the trap (4.4 cm × 25.6 cm), the rTPC, the 64 scintillator bars and the solenoid’s 510 mm bore; the rest is read off the drawing, and the notes list what is simplified. The escape counts are the paper’s Table 1. The on-axis field, the down-shares, the unit conversions and the release laboratory’s toy model are computed by this page (docs/plates/antihydrogen/computed-*.py, tests/antihydrogen.test.js); the toy is not the collaboration’s simulation. Material presets, parts and the softbox studio of the site’s model kit are ported from AirsupHQ/machine-skill (MIT, © 2026 Air Supply Technologies Inc.).
The text, pictures and models on this page are licensed CC BY-NC 4.0: free to use and adapt for teaching and other non-commercial purposes, with credit to world.lapla.si. Third-party material listed above keeps its own licence.
Cite this plate
Laplace’s Mini World, Plate F52, “Antihydrogen”, https://world.lapla.si/en/antihydrogen/, CC BY-NC 4.0, updated 2026-10-26.
Follow the question
Plate F28ATLAS and the Higgs BosonThe Large Hadron Collider smashes protons head-on, and the energy of the collision becomes new particles. A Higgs boson decays as soon as it is made; ATLAS records the photons and leptons it leaves and works its mass back out.
Plate F38Muon g−2A superconducting storage ring 15 m across keeps muons circling until they decay, counts the positrons they throw off, and pins down the small extra in the muon’s magnetism to 127 parts per billion.
Plate F27Bose–Einstein Condensation of AtomsLasers cool atoms to some tens of microkelvin above absolute zero; a magnetic trap then lets the hottest ones “evaporate”, a few at a time. Near 170 nanokelvin, a few thousand rubidium atoms fell into the same quantum state and showed up as a peak in a picture.
