Observation of the effect of gravity on the motion of antimatter

The study tests whether neutral antimatter (antihydrogen) experiences gravitational attraction toward Earth consistent with the weak equivalence principle (WEP). Direct ballistic tests for antimatter have been hindered by the lack of stable, neutral test particles and the dominance of electromagnetic forces on charged antiparticles in Earth's fields. With advances in producing and trapping antihydrogen, the ALPHA-g experiment—a vertical magnetic trap—enables controlled release of trapped anti-atoms to observe any gravitational influence on their motion. The central question is the sign and magnitude of the gravitational acceleration of antihydrogen, addressing longstanding speculation about possible repulsive gravity for antimatter and providing a direct test relevant to general relativity and WEP.

Prior precision tests of the WEP with ordinary matter (e.g., MICROSCOPE) have reached ~10^-15 precision. Indirect constraints involving charged antimatter (e.g., cyclotron-frequency and clock-based tests) were consistent with expected gravitational behavior but could not probe free fall due to electromagnetic forces. The ALPHA collaboration previously demonstrated production and trapping of antihydrogen and performed precision spectroscopy (hyperfine, 1S–2S, Lyman-α, fine structure) and recently laser cooling, establishing experimental control necessary for gravitational studies. Earlier proposals and a 2013 proof-of-principle by ALPHA suggested that controlled release from magnetic traps could reveal gravitational effects. Alternative experimental approaches (GBAR, AEgIS) aim to measure antimatter gravity through different techniques, underscoring community interest.

Apparatus and trapping: The ALPHA-g apparatus is a vertically oriented antihydrogen trap with a nominal trapping volume of 4.4 cm diameter and 25.6 cm height. Antiprotons from CERN’s AD/ELENA (7.5×10^9 at 100 keV every 120 s; ~5×10^7 captured) are cooled with co-trapped electrons and injected into ALPHA-g. Positrons (~3×10^8 per mixing cycle) from a Surko-type accumulator are injected and manipulated for size/density control. Antihydrogen is produced by mixing positron and antiproton plasmas within a superconducting anti-atom trap comprising octupole magnets for transverse confinement and two solenoidal mirror coils (A bottom, G top) for axial confinement. The region operates near 4 K. Trapped anti-atoms have kinetic energies up to the trap depth (~0.5 K). ALPHA-g typically traps a few atoms per mixing, which are accumulated (“stacked”) over many cycles (about 50 stacks over ~4 h yielding ~100 trapped atoms). Release protocol and magnetic biasing: After stacking, atoms are released by simultaneously ramping down currents in mirror coils A and G linearly over 20 s. Atoms can escape upward (through G) or downward (through A) and annihilate on the apparatus walls, with positions reconstructed by a radial time projection chamber (rTPC) and selected with a barrel scintillator. Gravity is expected to bias escape towards the bottom even in a symmetric trap (simulations predict ~80% downward escapes for hydrogen under ALPHA-g conditions when B_A = B_G). To probe and control the gravitational effect, a small, programmable magnetic bias is introduced by adding a differential current to mirror G while coils A and G are connected in series. The bias is defined as μ_B(B_G − B_A)/[m_H g (z_G − z_A)] and expressed in units of g. In a simple 1D model, −1g bias balances downward gravity. The laboratory g ≈ 9.81 m s^-2 corresponds to a vertical magnetic field gradient of 1.77×10^-3 T m^-1 for ground-state hydrogen; over the 25.6 cm separation of mirror peaks, a field difference of 4.53×10^-5 T mimics 1g. During ramp-down, the smaller of the mirror fields is not fully reduced to the well bottom but held ~5×10^-3 T above to allow atoms to pass minor axial bumps from octupole end turns. Field measurements and calibration: On-axis field magnitudes under each mirror are measured via electron cyclotron resonance (ECR) at two times: ~130 s after the LOC ramp-down (pre-release characterization) and ~96 s after the mirror ramp-down (final well characterization). Persistent currents and small asymmetries cause the on-axis field difference to vary during the ramp. A measurement-based magnetic field model combining as-built superconducting wire geometry with ECR and magnetron-frequency measurements is used to compute the instantaneous bias during the ramp. Each annihilation event is assigned the calculated bias corresponding to its escape time; biases for trials with the same programmed configuration are averaged to yield plotted bias values. Uncertainties in bias include ECR spectrum width, repeatability, peak field fitting, asymmetrical post-ramp decay, temporal variation during ramp, and field modeling (combined ~0.1g scale). Octupole configuration and LOC diagnostic: Two octupoles are used: a long octupole (LOC) spanning 1.5 m (covering additional trapping regions) and a bottom octupole (OCB) covering the active region. For trapping and stacking, both are at ~830 A. After stacking, the LOC is ramped down in 1 s to remove transverse confinement above mirror G, releasing roughly half the stacked atoms. The number of annihilations during this LOC ramp-down serves as a normalization and an estimate of stacked atom number per trial. Bias scan and data collection: Trials were interleaved over ~30 days for nominal biases: ±3g, ±2g, ±1.5g, ±1g, ±0.5g, and 0g, plus calibration runs at ±10g intended to force unidirectional release (top for −10g, bottom for +10g). For each bias, 6–7 trials were performed depending on event yields. Event selection applied spatial z-cuts to exclude vertices between the physical mirror centers and outside ±20 cm beyond them, and a time window accepting events from 10–20 s into the ramp (earlier events negligible). Detector relative efficiencies for up/down regions were obtained using ±10g calibration data and normalized to LOC-ramp counts. Cosmic-ray backgrounds were subtracted (per-trial backgrounds: top 0.18±0.01, bottom 0.21±0.01 events; LOC window 0.83±0.02 events over 13.1 s). Simulations: Extensive 3D trajectory simulations of trapped/released antihydrogen employed the measured current profiles and the field model, including on-axis and off-axis features (e.g., OCB end-turn bumps). Simulations generated escape probability curves P_dn versus applied bias for scenarios of attractive gravity (~1g), no gravity, and repulsive gravity, and for varied assumed values of the antihydrogen gravitational acceleration a_g to enable likelihood fits to data. Simulation uncertainty accounts for potential magnet winding misalignments, off-axis persistent fields, and uncertainties in initial longitudinal/transverse energy distributions. Additional studies: Cross-check runs with a slower 130 s ramp for biases 0g, −1g, −2g agreed with the 20 s data and simulations within uncertainties. Some atoms were released after the 20 s ramp, possibly due to slow axial–transverse energy mixing; their behavior appeared consistent with the main sample though detailed systematics remain for future work. Systematic considerations: Potential confounders were evaluated and found negligible: an antihydrogen charge limit <10^-28 C combined with electrode potentials stable to ±10 mV implies negligible electrostatic influence; antihydrogen magnetic moment uncertainty (<0.1%) negligibly affects potentials; ground-state polarizability implies even 100 V m^-1 electric fields are equivalent to <10^-15 T magnetic change; background gas collision probability <0.5% over 20 s (3% over 130 s).

- Escape asymmetry consistent with downward gravitational attraction: As the applied magnetic bias increases from −3g to +3g, the fraction escaping downward (P_dn) increases monotonically; near +3g almost all atoms exit bottom, near −3g almost all exit top. - Balance point near −1g: The P_dn = 0.5 point lies close to −1g, as expected from the 1D argument where −1g magnetic bias balances gravity. - Quantitative acceleration estimate: Likelihood analysis comparing data to simulations yields a_g = (0.75 ± 0.13 (statistical + systematic) ± 0.16 (simulation)) g, with g = 9.81 m s^-2, consistent with attractive gravity of magnitude ~1g within uncertainties. - Exclusions: Probability that data arise if gravity does not act on antihydrogen is 2.9×10^-15; data are overwhelmingly incompatible with repulsive gravity of magnitude 1g (p < 10^-15), ruling out such antigravity. - Representative counts (10–20 s window, z-cuts, background-corrected; 6–7 trials per bias): For nominal biases (g): −3.0: N_up≈151.7, N_dn≈16.5; −1.0: 69.7 up, 62.5 down; 0: 36.7 up, 94.5 down; +1.5: 13.9 up, 180.7 down; +3.0: 7.7 up, 147.5 down. Calibration ±10g produced essentially unidirectional escape (−10g: ~142.9 up, 0.7 down; +10g: ~0 up, 185.7 down).

The observed dependence of the downward escape probability on applied magnetic bias directly demonstrates that neutral antihydrogen responds to Earth’s gravitational field with attraction. The balance point near −1g and the agreement in shape and position of the escape curve with detailed 3D simulations support the interpretation that gravity biases the escape direction of anti-atoms during controlled release. The likelihood fit indicates a downward acceleration consistent with g within uncertainties, excluding no-gravity and repulsive gravity hypotheses for antihydrogen in this configuration. Systematic effects that could mimic gravity—residual charge, electric fields, dipole moment uncertainty, background gas collisions, persistent currents—were measured or bounded and shown to be negligible at the present sensitivity. Cross-checks with longer ramp times and analysis of late releases further support robustness. These results align antimatter behavior with general relativity’s WEP and provide a foundation for more precise future tests.

The experiment provides the first direct observation that neutral antimatter (antihydrogen) falls toward Earth, with a best-fit acceleration of (0.75 ± 0.13 ± 0.16) g, consistent with attractive gravity and incompatible with repulsive antigravity. This validates the expected sign of gravitational interaction for antimatter and supports the weak equivalence principle. The work establishes a practical method—controlled release from a vertical magnetic trap with tunable bias—to probe antimatter gravity and identifies key statistical and systematic factors governing precision. Future directions include producing colder antihydrogen (via laser cooling and adiabatic expansion) to sharpen the escape curve, improved magnetometry and field control (electrons, NMR probes, cold ions), use of a central trapping region less sensitive to stray fields, and exploration of fountain-type interferometry with potential precision at the 10^-8 level. Complementary experiments (GBAR, AEgIS) will broaden the experimental program.

- Precision limited by magnetic field control and modeling: uncertainties in on-axis/off-axis fields, persistent currents, and field asymmetries lead to bias uncertainties of order 0.1 g; simulation model uncertainty contributes ±0.16 g to a_g. - Dependence on simulations: extraction of a_g relies on matching data to simulated escape curves; unmeasured factors (winding misalignments, off-axis fields, initial energy distributions) are encompassed by a simulation uncertainty band but remain to be benchmarked. - Finite statistics and detector calibration: limited atom numbers per trial and uncertainties in relative up/down detection efficiencies contribute to the total uncertainty (e.g., ±0.12 g from efficiency calibration; ±0.06 g statistical/systematic; ±0.06 g finite data size). - Selection cuts and time window: results pertain to events escaping during 10–20 s of the ramp and within specified z-regions; some atoms escape after the ramp due to slow dynamics, whose detailed systematics are not yet fully characterized. - Not a precision measurement: the experiment aimed to determine sign and approximate magnitude, not a high-precision value; further work is required to reduce systematics and improve field metrology.