The axion is a hypothetical electrically neutral, spin-zero particle associated with a proposed solution to the strong CP problem: the unexplained near-absence of a particular kind of symmetry violation in quantum chromodynamics (QCD), the theory of the strong interaction. Introduced in 1978, the axion arises when a new symmetry makes a seemingly arbitrary parameter of particle physics dynamically relax toward a value consistent with experiment. Its existence would therefore explain a feature of the Standard Model that otherwise requires an exceptionally small parameter.
Axions are also candidates for dark matter. Although an individual axion could be extremely light, a population produced coherently in the early universe could behave as cold, nonrelativistic matter. This connection has made axions a major subject spanning particle physics, cosmology, stellar astrophysics, and precision measurement. Searches use microwave resonators, strong magnets, optical instruments, spin sensors, and astronomical observations. No axion has been established experimentally. The term is sometimes extended to axion-like particles, but these need not possess the defining relationship between mass and interactions that characterizes the QCD axion.
The strong CP problem and the origin of the proposal
QCD describes quarks interacting through gluons. Besides the terms that govern familiar strong-interaction processes, its most general renormalizable Lagrangian permits a term proportional to
Here is the strong coupling, is the gluon field strength, and its dual is . The coefficient is an angular parameter. Although this expression is a total derivative locally, the nontrivial topology of the QCD vacuum makes it physically consequential.
The term violates parity and CP symmetry, the combined transformation of charge conjugation and spatial reflection. CP violation is not generally forbidden in nature: it occurs in weak-interaction processes and is incorporated into the Standard Model through quark mixing. The puzzle is why its strong-interaction counterpart is so small. The observable parameter combines the gluonic vacuum angle with the phase of the quark mass matrix, conventionally written . Setting the gluonic angle alone to zero therefore does not remove the problem; the physically relevant combination must be small.
A nonzero would induce an electric dipole moment in the neutron. Such a moment would describe a permanent separation of electric charge aligned with its spin, despite its zero total charge. Calculations relate the neutron dipole moment to , with uncertainties from strong-interaction physics. Experimental limits at the level of imply that is roughly below . For example, a neutron experiment reported in 2020 a limit of at 90 percent confidence. Why an otherwise unrestricted angular parameter should be so close to zero is the strong CP problem.
Roberto Peccei and Helen Quinn proposed a dynamical solution in 1977. Steven Weinberg and Frank Wilczek independently recognized in 1978 that this mechanism predicts a new light particle. Wilczek named it the axion after a detergent brand, invoking its role in “cleaning up” the strong CP problem. The original electroweak-scale realization was soon excluded, but models with a much higher symmetry-breaking scale remained viable.
Other proposed solutions have different implications. A massless up quark would make the strong CP angle unobservable, but determinations of quark masses, including lattice QCD calculations, do not support this possibility. Models with fundamental CP symmetry and carefully arranged symmetry breaking offer another approach. The axion proposal is distinctive because it replaces a fixed parameter with a dynamical field and provides a potentially observable new particle.
Peccei–Quinn symmetry and dynamical relaxation
The Peccei–Quinn symmetry is usually introduced as an approximate global symmetry, often denoted . It acts on new fields and, depending on the model, on some Standard Model fields. When this symmetry undergoes spontaneous symmetry breaking, the vacuum chooses one point from a continuous family of equivalent states. The axion corresponds to fluctuations along this angular direction.
For a simple complex scalar field, a schematic parametrization is
The radial excitation is generally heavy, whereas the angular field would be massless if the symmetry were exact. This is the mechanism underlying a Goldstone boson. The physical axion decay constant is related to the symmetry-breaking scale, with normalization factors determined by the model’s anomaly coefficient. Consequently, and should not always be identified without qualification.
The crucial feature is that the Peccei–Quinn symmetry has a QCD anomaly: a symmetry of the classical theory is not preserved by the quantum theory. Its consequence is an axion–gluon interaction that replaces the fixed strong CP angle by a field-dependent combination,
Nonperturbative QCD generates a potential for this combination. Near its minimum, the potential can be expanded as
where is the QCD topological susceptibility. The vacuum settles at , modulo the field’s periodicity. Thus the effective strong CP angle vanishes at the minimum rather than requiring a specially chosen initial value.
The same potential gives the axion a mass. It is therefore a pseudo-Goldstone boson: light because it originates from an approximate continuous symmetry, but not massless because QCD breaks that symmetry. The relaxation mechanism is conceptually different from screening an electric field or simply assigning a small coupling. The axion changes the vacuum state so that the offending CP-violating combination is minimized.
The mechanism also has a theoretical vulnerability. Additional interactions that explicitly violate Peccei–Quinn symmetry can shift the minimum away from the CP-conserving point. Generic high-energy symmetry-breaking operators may therefore spoil the solution unless their effects are strongly suppressed. This is known as the axion quality problem. Proposed constructions use discrete symmetries, gauge structure, compositeness, or other protective mechanisms, but the required protection depends on the scale and field content of the model.
Mass, couplings, and particle models
For the QCD axion, mass and interaction strength are linked. In units with ,
Combining low-energy QCD with lattice calculations gives the representative relation
The numerical coefficient has a small theoretical uncertainty; it is not an exact defining constant. A larger decay constant produces a lighter axion with generally weaker interactions. This inverse relationship distinguishes QCD axions from more general particles whose mass and couplings can vary independently.
An axion is a pseudoscalar boson: it has zero spin and changes sign under spatial inversion. One especially important interaction couples it to two photons:
With a standard choice of conventions, this is proportional to . For a QCD axion,
where is the electromagnetic fine-structure constant and is the ratio of electromagnetic to color anomaly coefficients. The approximate number 1.92 comes from low-energy QCD. The two contributions can partially cancel, so an axion need not couple to photons with the same strength in every model.
Two widely used model families illustrate this variation. In KSVZ models, named for Jihn E. Kim, Mikhail Shifman, Arkady Vainshtein, and Valentin Zakharov, new heavy colored fermions communicate Peccei–Quinn symmetry breaking to QCD. Ordinary quarks and leptons can be neutral under the new symmetry, and the electron coupling can vanish at tree level. A commonly used benchmark has , although this is not compulsory for every KSVZ construction.
In DFSZ models, associated with Michael Dine, Willy Fischler, Mark Srednicki, and independently Alexander Zhitnitsky, the Standard Model fermions participate through an extended Higgs sector. These models generally contain two Higgs doublets and a singlet involved in Peccei–Quinn breaking. A standard anomaly ratio is , while fermion couplings depend on Higgs-sector parameters.
At low energies, axions can also interact with electrons, protons, and neutrons through derivative couplings, schematically
The dimensionless coefficient is model dependent. Translating quark couplings into nucleon couplings requires hadronic physics. Consequently, an experimental exclusion of a particular photon or electron coupling does not automatically exclude every axion model at that mass.
Production in the early universe and dark matter
Axion dark matter need not arise from a gas of particles in thermal equilibrium. Its most characteristic production mechanism is vacuum misalignment. In an expanding universe, a nearly homogeneous axion field approximately satisfies
where is the cosmic expansion rate and is the temperature-dependent potential. At sufficiently early times, expansion strongly damps the field’s motion. As QCD effects strengthen and the axion mass grows, the field begins oscillating when the mass becomes comparable to the expansion rate, conventionally estimated by .
These oscillations store energy. Once the potential is approximately quadratic and the mass has settled, their average energy density decreases in proportion to the inverse cube of the cosmic scale factor, as ordinary nonrelativistic matter does. This explains how a particle with a tiny rest mass can constitute cold dark matter: the relevant axions are produced with very small momenta rather than inheriting a high thermal temperature.
The predicted abundance depends strongly on the timing of Peccei–Quinn symmetry breaking relative to cosmic inflation. If the symmetry breaks before or during inflation and is not subsequently restored, the observable universe can inherit one nearly uniform initial angle, . The abundance then depends on both and . A sufficiently small initial displacement can permit larger decay constants than a generic order-one angle would allow. Inflationary fluctuations in the axion can also create isocurvature perturbations—variations in matter composition distinct from ordinary total-density perturbations—which are constrained by the cosmic microwave background.
If Peccei–Quinn symmetry breaks after inflation, different regions choose different initial angles. The resulting field configuration includes topological defects: cosmic strings and, when the QCD potential becomes important, domain walls. Their evolution and decay contribute additional axions. For domain-wall number one, the string–wall network can decay. For larger domain-wall number, a stable network is generally cosmologically problematic unless additional physics removes or destabilizes it. Simulating these systems involves enormous differences between microscopic and cosmological scales, producing significant uncertainty in abundance predictions.
Post-inflationary inhomogeneities may seed dense structures called axion miniclusters. Their abundance and survival affect the local distribution of dark matter and potentially the interpretation of laboratory searches. Axions produced thermally form a separate population, whose cosmological effects differ from those of the cold coherent field.
There is therefore no single experimentally mandatory “dark-matter axion mass.” A prediction requires assumptions about the initial angle, inflation, thermal history, and defect evolution. Estimates in common cosmologies motivate microelectronvolt and higher mass ranges, but the strong CP mechanism alone does not fix which cosmological history occurred.
Stellar production and astrophysical constraints
Weakly interacting particles can carry energy out of stellar interiors. Axions would be produced through several processes whose importance depends on their couplings. The Primakoff effect converts a photon into an axion in the electromagnetic field of a charged particle. Electron couplings permit production through bremsstrahlung, Compton-like scattering, and atomic processes. In dense nuclear matter, axion emission can involve nucleon–nucleon collisions and additional hadronic reactions.
These channels change stellar energy transport. In the usual weak-coupling regime, axions escape rather than repeatedly scattering, acting as an extra cooling mechanism. A star may respond by increasing its nuclear burning rate, changing its lifetime, core structure, or evolutionary path. Constraints therefore come not just from measuring an instantaneous luminosity, but from comparing stellar populations and evolutionary stages.
Horizontal-branch stars in globular clusters provide an important probe of axion–photon interactions. Their helium-burning cores can produce axions through the Primakoff process; excessive energy loss would shorten the horizontal-branch lifetime. Representative analyses constrain the photon coupling to values of order a few for sufficiently light axions. The precise inferred bound depends on stellar modeling and quantities such as the helium abundance. Red-giant evolution and white-dwarf cooling constrain axion–electron interactions, often at the level of a few for the dimensionless coupling .
Some stellar datasets have shown a preference for additional cooling compatible with weak axion-like interactions. Such preferences are not detections. Stellar ages, composition, opacities, convection, and population properties introduce uncertainties, and a proposed particle interpretation must consistently fit several observations and laboratory limits. Different datasets also probe different combinations of couplings rather than one universal axion parameter.
The neutrino burst from supernova 1987A supplies another major constraint. Axions emitted from the hot proto-neutron star could remove energy and alter the duration and character of the burst. The resulting limits on axion–nucleon interactions are important but depend on dense-matter physics, transport, and the supernova model. Moreover, energy-loss bounds do not necessarily become monotonically stronger as coupling increases: sufficiently strong interactions can trap the new particles, changing how they transport energy.
Compact objects also offer possible conversion signals. In a neutron-star magnetosphere, axion dark matter might convert resonantly to radio photons where the plasma conditions allow matching of dispersion relations. Predicted spectra and fluxes depend on the plasma distribution, magnetic geometry, and dark-matter environment. These uncertainties distinguish such searches from laboratory experiments with controlled magnets and detectors.
Haloscopes and searches for Galactic axion dark matter
A haloscope searches for axions presumed to occupy the Galactic dark-matter halo. Pierre Sikivie proposed the resonant microwave-cavity technique in 1983. A strong static magnetic field enables axion–photon conversion, while a resonator enhances the signal when one of its electromagnetic modes matches the axion’s oscillation frequency.
The approximate conversion frequency follows from mass–energy equivalence and the Planck constant:
An axion with rest energy corresponds to about megahertz. Galactic velocities are typically of order , adding a small kinetic-energy contribution and producing a fractional spectral width of order . A conventional halo signal is therefore a narrow spectral excess, although streams and other substructure can modify its shape.
For a cavity experiment, the expected power scales schematically as
Here is the local axion energy density, the applied field, the cavity volume, the mode-overlap factor, the loaded cavity quality factor, and the effective quality factor associated with the axion linewidth. Extraction efficiency and geometry introduce additional factors. Depending on parameters, expected powers can be near watts or smaller.
The Axion Dark Matter eXperiment, ADMX, established the feasibility of searching at benchmark QCD-axion sensitivities over selected microelectronvolt mass intervals. It combines a tunable microwave cavity, a superconducting magnet, cryogenic operation, and extremely low-noise amplification. Other programs, including HAYSTAC, CAPP, ORGAN, and QUAX, explore complementary frequencies and technologies. These experiments scan: a resonator samples a limited frequency interval at each tuning step, so coverage accumulates over time.
Noise reduction is central to the method. Cooling suppresses thermal photons, but quantum fluctuations become limiting at low temperatures. Josephson-junction amplifiers and related quantum devices approach the relevant noise limits, while squeezed-state receivers can increase scan speed under suitable operating conditions. Calibration of the receiver gain, noise, cavity properties, and magnetic-field overlap is essential for turning a missing spectral excess into a coupling limit.
Different architectures address different mass scales. At higher frequencies, ordinary cavities become smaller, motivating coupled resonators, photonic structures, and dielectric haloscopes. At lower frequencies, lumped-element circuits and toroidal magnet geometries seek axion-induced electromagnetic signals. All halo limits assume a local axion density and usually a velocity distribution. They therefore constrain the combination of coupling and axion abundance, rather than coupling alone independently of cosmology.
Solar helioscopes and laboratory photon conversion
An axion helioscope looks for axions produced in the Sun and reconverted into X-ray photons in a laboratory magnet. Unlike a haloscope, it does not require axions to constitute Galactic dark matter. The solar flux instead follows from thermal production in the solar interior. For Primakoff production, the spectrum lies mainly at kiloelectronvolt energies, with a characteristic mean energy of about 4 keV.
In a homogeneous transverse magnetic field, the axion–photon conversion probability can be written, in natural units and neglecting absorption, as
where is the field length and is the momentum mismatch. For relativistic axions,
The effective photon mass is determined by the medium. Coherent conversion requires the mismatch to remain small enough that contributions from different parts of the magnet add rather than cancel.
The CERN Axion Solar Telescope, CAST, used a roughly 9-tesla, 9-metre decommissioned accelerator magnet to track the Sun. Its vacuum operation constrained below at 95 percent confidence for axion masses below approximately eV, as reported in 2017. Introducing helium gas gave the photons an adjustable effective mass, allowing searches at higher axion masses by restoring coherence. Those measurements covered discrete mass settings rather than applying one uniform sensitivity across the entire interval.
Helioscope sensitivity depends on magnetic field, length, aperture, tracking time, detector background, and X-ray optics. The proposed International Axion Observatory, IAXO, and its precursor BabyIAXO pursue larger acceptance and dedicated instruments. If solar production proceeds through electron interactions rather than the Primakoff effect, the detected rate involves a different product of couplings. This matters when comparing helioscope results with stellar constraints.
Purely terrestrial experiments avoid both a solar-production model and a Galactic-density assumption. In light-shining-through-a-wall experiments, laser photons convert into axions in one magnetic region. An opaque barrier blocks the photons, while axions can pass through and reconvert into photons in a second region. The unenhanced rate is suppressed by the product of two conversion probabilities and thus scales as . Optical resonators can substantially enhance production and regeneration. ALPS II is a prominent implementation of this approach.
Axion interactions can also affect light polarization in magnetic fields. Such measurements must separate hypothetical particle effects from instrumental backgrounds and the much smaller vacuum-optical effects predicted by quantum electrodynamics.
Spin-based searches and precision measurements
Axions need not be sought through photons. Their interactions with fermions can produce tiny spin-dependent effects, and an oscillating axion background can act as a coherent external field. This motivates methods based on nuclear magnetic resonance, atomic magnetometry, spin-precession measurements, and precision comparisons of different materials or isotopes.
For a nonrelativistic classical axion field, a useful approximation in natural units is
Its coherence time is set by the velocity-induced frequency spread, not simply by one oscillation period. For a smooth virialized halo, the inverse fractional linewidth corresponds to roughly oscillation cycles, up to convention-dependent numerical factors. Experiments use this structure to design coherent integrations and searches for narrowband stochastic signals.
The derivative axion coupling to fermions produces a nonrelativistic interaction involving spin and the spatial gradient of the field. Because a detector moves through the Galactic axion distribution, this can resemble a tiny oscillating effective magnetic field, commonly called the “axion wind.” Its directional response can vary with Earth’s rotation and orbital motion. These signatures help distinguish a possible particle signal from laboratory-fixed disturbances.
The QCD coupling provides a different route. An oscillating axion field corresponds to an oscillating effective strong CP angle and can therefore induce oscillating nucleon electric dipole moments. In a suitable electric environment, this can drive spin motion. Nuclear and atomic structure complicate the observable response: screening, nuclear moments, and material properties must be treated explicitly. The Cosmic Axion Spin Precession Experiment, CASPEr, develops approaches using both electric-dipole and axion-wind interactions. GNOME, a geographically distributed network of optical magnetometers, searches for correlated spin disturbances, including signals from hypothetical axion-like field structures.
Laboratory forces provide another class of tests. Axion exchange can generate short-range spin-dependent forces. Interactions combining scalar and pseudoscalar couplings can produce a force between an unpolarized source and a polarized detector, often termed a monopole–dipole interaction. Its range is set by the reduced Compton wavelength,
The ARIADNE concept uses a rotating source mass and a polarized helium-3 sample to seek a periodically driven nuclear-spin response. Interpreting such searches requires identifying which couplings are present: a scalar interaction is not automatically fixed by the same assumptions as a pseudoscalar spin interaction. Accordingly, a bound on a coupling product should not be presented as a model-independent exclusion of all QCD axions.
Axion-like particles, related phenomena, and unresolved questions
An axion-like particle (ALP) is a light pseudoscalar with interactions resembling those of an axion but without the required QCD relationship between mass and decay constant. An ALP can couple to photons or fermions without solving the strong CP problem. Conversely, a QCD axion is defined by its role in the Peccei–Quinn mechanism, not merely by being a light particle with a two-photon coupling. This distinction is essential when interpreting experimental parameter plots.
Theories extending the Standard Model can contain several axion-like fields. Some constructions in quantum gravity, including string-based models, yield many such degrees of freedom with widely separated masses. This possibility is sometimes called an axiverse. It motivates searches beyond the conventional QCD-axion relation, but neither a particular particle spectrum nor the existence of these fields follows uniquely from such broad theoretical frameworks.
Astrophysical propagation probes part of this larger parameter space. Photons and ALPs can interconvert in cosmic magnetic fields, potentially changing gamma-ray spectra or polarization. Interpretation depends on the magnetic field along the line of sight, plasma effects, the emitted spectrum, and conventional absorption. Spectral irregularities alone are therefore not a unique particle signature. Polarization measurements likewise require careful separation of a possible field-induced rotation from calibration uncertainties and astrophysical effects.
Rotating black holes provide another probe through black-hole superradiance. Under suitable conditions, a bosonic field can extract rotational energy and build a cloud around the hole. The effect is strongest when the particle’s Compton wavelength is comparable to the gravitational size of the object. Measurements of black-hole spins and searches for associated gravitational radiation can test ranges of ultralight-boson masses. The interpretation depends on the age and environment of the black hole and on particle self-interactions; it is not restricted to QCD axions.
A separate use of the word occurs in condensed-matter physics. Certain materials admit an electromagnetic response described by a term mathematically analogous to . In appropriate systems, collective excitations can be described as axion-like modes. These are emergent material phenomena, not evidence that the fundamental particle proposed by Weinberg and Wilczek has been found.
For the QCD axion itself, unresolved questions include the origin and protection of Peccei–Quinn symmetry, its breaking scale, its relation to inflation, and the contribution of defects to the relic abundance. A convincing detection would require more than an unexplained narrow feature: reproducibility, controlled backgrounds, and the predicted response to magnetic fields, tuning, orientation, or independent instruments would be crucial. Even discovery of a new pseudoscalar would leave the further task of establishing whether its interactions possess the QCD structure needed to solve the strong CP problem.