Axions & the strong CP problem

Count the free parameters of the Standard Model and you reach twenty-five: twelve fermion masses, three gauge couplings, two Higgs parameters, and the eight mixing angles and phases of the CKM and PMNS matrices. But there is a twenty-sixth number hiding in plain sight. Quantum chromodynamics — the exact same non-Abelian gauge theory that gives us confinement and asymptotic freedom — permits one more term in its Lagrangian, a term built from the gluon field and its dual, weighted by an angle $\theta$ . That term violates parity and CP in the strong interaction. Nothing in the theory forbids it, and nothing fixes its size; on dimensional and symmetry grounds $\theta$ should be a number of order one, like every other angle in the theory. Experiment says otherwise. The strong CP angle would give the neutron a permanent electric dipole moment, and decades of ever-more-sensitive experiments have found none, forcing the physically observable angle $\bar\theta$ below $10^{-10}$ . A parameter free to be anything is measured to be zero to ten decimal places. That is the strong CP problem: not a contradiction, but a spectacular and unexplained fine-tuning, and one of the sharpest open questions the Standard Model leaves on the table. This article develops the problem from the QCD Lagrangian, then follows the most compelling solution — the Peccei–Quinn mechanism and its inevitable by-product, a new light particle called the axion, which turns out to be one of the best-motivated candidates for the dark matter of the universe.

Key takeaways

QCD allows a CP-violating term $\bar\theta\,\frac{g_s^2}{32\pi^2}G\tilde G$ ; the physical angle is $\bar\theta=\theta+\arg\det M_q$ , combining the gluon $\theta$ and the quark-mass phases.
A non-zero $\bar\theta$ gives the neutron an electric dipole moment $d_n\simeq3\times10^{-16}\,\bar\theta\ e\,\mathrm{cm}$ ; the measured bound $|d_n|<1.8\times10^{-26}\,e\,\mathrm{cm}$ forces $\bar\theta\lesssim10^{-10}$ .
The strong CP problem is why a parameter that should be of order one is instead essentially zero — an unexplained fine-tuning, not a Standard Model prediction.
The Peccei–Quinn mechanism promotes $\bar\theta$ to a dynamical field: a new broken U(1) symmetry whose pseudo-Goldstone boson, the axion, rolls to the CP-conserving minimum, setting the effective angle to zero automatically.
The axion has a QCD-fixed mass $m_a\simeq5.7\,\mu\mathrm{eV}\,(10^{12}\,\mathrm{GeV}/f_a)$ and a coupling to two photons that scales as $1/f_a$ — the lighter it is, the more weakly it couples.
Key idea: QCD hides a free CP-violating angle that experiment pins to zero to ten decimals; making that angle a dynamical field — the axion — solves the puzzle and simultaneously supplies a natural dark-matter candidate.

The core insight

QCD allows a CP-violating term whose strength $\bar\theta$ is unconstrained by the theory, yet the neutron electric dipole moment forces it below $10^{-10}$ ; promoting $\bar\theta$ to a dynamical field — the axion of a broken Peccei–Quinn symmetry — relaxes it to zero automatically and leaves behind a light particle that can be the dark matter.

Where this lives today

The axion has moved from a theorist's fix to an experimental frontier attacked from every side. The flagship dark-matter search is ADMX (the Axion Dark Matter eXperiment), a resonant microwave cavity in an eight-tesla magnet that has reached the sensitivity to detect DFSZ-strength axions in the few- $\mu$ eV band — the first experiment to probe the theoretically favoured coupling. Complementary haloscopes push to other masses: HAYSTAC and CAPP use quantum-limited amplifiers at higher frequencies, MADMAX proposes dielectric-disk stacks for the $100\,\mu$ eV range, and DMRadio and ABRACADABRA target much lighter axions. Solar axions are hunted by helioscopes: CAST at CERN set the leading laboratory bound $g_{a\gamma}<0.66\times10^{-10}\,\mathrm{GeV}^{-1}$ , and its successor IAXO aims to go an order of magnitude further. Meanwhile next-generation neutron-EDM experiments push directly on $\bar\theta$ , lattice QCD sharpens the topological susceptibility, and astrophysics supplies strong limits — stellar cooling and the SN 1987A neutrino burst constrain the couplings, and black-hole superradiance excludes bands of ultralight masses. Beyond the strict QCD axion, string-motivated "axion-like particles" (ALPs) have become a generic dark-matter target, making this one of the most active corners of physics beyond the Standard Model.

The physical picture

The strong CP problem is a story about the vacuum. In a non-Abelian gauge theory the vacuum is not unique: there are infinitely many topologically distinct field configurations, labelled by an integer winding number, and the true ground state is a superposition of them weighted by a phase $e^{i\theta}$ . That phase is physical, and it is CP-odd — it multiplies the combination $\vec E^a\!\cdot\vec B^a$ of the chromo-electric and chromo-magnetic fields, exactly the structure that flips sign under parity. If $\theta$ were of order one, the neutron would carry a measurable electric dipole moment aligned with its spin. The Peccei–Quinn idea replaces the fixed angle with the value of a new field, which QCD then pushes to the CP-conserving bottom of a cosine potential. The first figure shows that potential and the relaxation; the second shows where the resulting axion sits in the experimentally hunted landscape.

Figure

The QCD-generated axion potential V of a equals minus Λ⁴ cos of the effective angle. The field rolls from any initial misalignment down to the minimum where θ̄ plus ⟨a⟩ over f_a equals zero, dynamically cancelling strong CP. The curvature at the bottom fixes the mass m_a ≈ 5.7 μeV × 10¹² GeV over f_a

Because the axion's potential comes from QCD, its mass is not a free parameter: it is tied to the pion mass and decay constant through the same non-perturbative physics, and it scales inversely with the symmetry-breaking scale $f_a$ . A heavier axion means a lower $f_a$ and a stronger coupling; a feather-light axion means a huge $f_a$ and couplings so weak the particle earned the nickname "invisible". The second figure places the QCD axion on the map that experiments actually search — the plane of photon coupling versus mass — where the model prediction is not a region but a narrow diagonal band.

Figure

The QCD-axion band in the photon-coupling versus mass plane. Because both g_aγ and m_a scale as 1 over f_a, the KSVZ and DFSZ models trace a diagonal g_aγ ≈ 0.15 to 0.39 ×10⁻¹⁵ times m_a in μeV, in GeV⁻¹. Haloscopes such as ADMX scan the μeV window where f_a ≈ 10¹² GeV lets the axion be the dark matter

The core equations

Start from the one term the previous QCD articles left out. Alongside the kinetic and quark-mass terms, gauge invariance and renormalisability permit a further gauge-invariant, Lorentz-invariant operator built from the gluon field strength $G^a_{\mu\nu}$ and its dual $\tilde G^{a\,\mu\nu}$ . It is the topological $\theta$ -term.

\mathcal{L}_\theta=\theta\,\frac{g_s^{2}}{32\pi^{2}}\,G^{a}_{\mu\nu}\tilde G^{a\,\mu\nu},\qquad \tilde G^{a\,\mu\nu}=\tfrac12\epsilon^{\mu\nu\rho\sigma}G^{a}_{\rho\sigma} \qquad (1)

where: $\theta$ — the dimensionless vacuum angle, periodic mod $2\pi$ ; $g_s$ — the strong coupling; $G^a_{\mu\nu}$ — the gluon field strength ( $a=1,\dots,8$ ); $\tilde G^{a\,\mu\nu}$ — its dual, so $G\tilde G\sim\vec E^a\!\cdot\vec B^a$ . The combination is P-odd and CP-odd but gauge invariant — it cannot be forbidden by any symmetry QCD respects. The angle in $\mathcal{L}_\theta$ is not the whole story, because rotating the phases of the quark fields shifts $\theta$ through the chiral anomaly while simultaneously changing $\arg\det M_q$ . Only the sum is invariant, and it is the sum that observable physics depends on.

\bar\theta=\theta+\arg\det\!\big(M_uM_d\big),\qquad d_n\simeq3\times10^{-16}\,\bar\theta\ \ e\,\mathrm{cm} \qquad (2)

where: $\bar\theta$ — the physical strong-CP angle; $M_u,M_d$ — the up- and down-type quark mass matrices; $d_n$ — the induced neutron electric dipole moment. The measured bound $|d_n|<1.8\times10^{-26}\,e\,\mathrm{cm}$ turns directly into $\bar\theta\lesssim10^{-10}$ . The Peccei–Quinn resolution introduces a new pseudoscalar field $a$ , the axion, whose shift symmetry is broken only by the QCD anomaly. It appears exclusively in the combination $\bar\theta+a/f_a$ , so its vacuum value can absorb $\bar\theta$ .

\mathcal{L}\supset\tfrac12(\partial_\mu a)^2+\Big(\bar\theta+\frac{a}{f_a}\Big)\frac{g_s^{2}}{32\pi^{2}}\,G\tilde G \qquad (3)

where: $a$ — the axion field; $f_a$ — the axion decay constant (the Peccei–Quinn symmetry-breaking scale). The kinetic term makes $a$ dynamical; the anomaly term is what QCD converts into a potential. Non-perturbative QCD — the instantons that made $\theta$ physical in the first place — generate a periodic potential for $a$ , minimised where the effective angle vanishes.

V(a)\simeq-\Lambda_{\mathrm{QCD}}^{4}\cos\!\Big(\bar\theta+\frac{a}{f_a}\Big),\qquad \Big\langle\frac{a}{f_a}\Big\rangle=-\bar\theta \qquad (4)

where: $\Lambda_{\mathrm{QCD}}\approx200\,\mathrm{MeV}$ — the QCD scale; the minimum sits at $\bar\theta+\langle a\rangle/f_a=0$ , so the effective strong-CP angle is dynamically driven to zero and $d_n\to0$ automatically. Expanding the potential about that minimum gives the axion a mass, fixed by QCD in terms of the pion mass and decay constant and inversely proportional to $f_a$ .

m_a=\frac{\sqrt{m_um_d}}{m_u+m_d}\,\frac{f_\pi m_\pi}{f_a}\simeq5.7\,\mu\mathrm{eV}\left(\frac{10^{12}\,\mathrm{GeV}}{f_a}\right) \qquad (5)

where: $m_u,m_d$ — up and down quark masses; $f_\pi\approx92\,\mathrm{MeV}$ , $m_\pi\approx135\,\mathrm{MeV}$ — pion decay constant and mass. The single product $m_af_a\approx f_\pi m_\pi\times\sqrt{m_um_d}/(m_u+m_d)$ is a QCD constant: fix one of $m_a$ , $f_a$ and the other follows. The axion also couples to two photons — the interaction every laboratory search exploits — with a strength set by $1/f_a$ and a model-dependent anomaly ratio.

g_{a\gamma}=\frac{\alpha}{2\pi f_a}\Big(\frac{E}{N}-1.92\Big),\qquad \mathcal{L}_{a\gamma}=-\tfrac14 g_{a\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu}=g_{a\gamma}\,a\,\vec E\cdot\vec B \qquad (6)

where: $\alpha$ — the fine-structure constant; $E/N$ — the ratio of electromagnetic to colour anomaly (0 for KSVZ, $8/3$ for DFSZ); the $-1.92$ is the model-independent pion-mixing contribution. In a magnetic field the axion converts to a photon (the Primakoff effect) — the basis of every haloscope and helioscope.

Derivation AWhy theta-bar and not theta is the physical CP-violating angle

The goal is to establish that the strong-CP problem cannot be argued away, from either of the two directions people first try. One protests that the $\theta$ -term is a total derivative and so should drop out; the other protests that the quark-mass phases are a separate matter of flavour physics. Both objections fail, and understanding why fixes the one combination — $\bar\theta=\theta+\arg\det M_q$ — that experiment measures through the neutron electric dipole moment.

Step 1. Write $G\tilde G=\partial_\mu K^\mu$ , where $K^\mu$ is the Chern–Simons current; naively a total derivative integrates to a surface term and vanishes. But in a non-Abelian theory the gauge field need not vanish at infinity — it can approach a pure gauge with a non-trivial winding, and then $\int d^4x\,\frac{g_s^2}{32\pi^2}G\tilde G=n$ , an integer topological charge. Why: the QCD vacuum is a sum over these winding sectors, and $\theta$ is the phase weighting them — a physical property of the ground state, not a discardable boundary term.

Step 2. Now perform a chiral rotation of the quark fields, $q\to e^{i\alpha\gamma_5}q$ , to try to remove a phase from the mass matrix. The path-integral measure is not invariant under this rotation — the Adler–Bell–Jackiw anomaly — and the Jacobian shifts $\theta\to\theta-2\alpha\,N_f$ at the same time as it rotates $\arg\det M_q$ by the opposite amount. Why: this is the crucial coupling — you cannot move a phase between the gluon sector and the quark-mass sector freely, because the anomaly links them, and only their sum stays put.

Step 3. Therefore the invariant, and the only thing physics can depend on, is $\bar\theta=\theta+\arg\det(M_uM_d)$ . As a corollary, if any single quark were exactly massless, $\det M_q=0$ and $\bar\theta$ could be rotated to any value — the strong-CP problem would evaporate. Lattice QCD, however, finds $m_u\neq0$ at many standard deviations, closing that escape. Why: the massless-quark loophole is the cleanest logically possible solution, and ruling it out is what leaves the problem genuinely open.

Step 4. Finally, $\bar\theta$ shows up in a P- and T-odd observable. Because $G\tilde G\sim\vec E\cdot\vec B$ reverses under parity and time reversal, a non-zero $\bar\theta$ induces a neutron electric dipole moment aligned with the neutron spin, $d_n\simeq3\times10^{-16}\,\bar\theta\ e\,\mathrm{cm}$ from chiral estimates. Why: this converts an abstract vacuum angle into a laboratory number, and its non-observation is exactly what makes the smallness of $\bar\theta$ a quantitative puzzle.

Go deeper · Derivation B · 2 minHow the Peccei-Quinn mechanism relaxes theta-bar to zero

Derivation BHow the Peccei-Quinn mechanism relaxes theta-bar to zero

The aim is to show that the fine-tuning disappears if $\bar\theta$ is not a constant of nature but the frozen value of a dynamical field. We add a global symmetry with a QCD anomaly, spontaneously break it, and follow its Goldstone boson — the axion — as QCD grows it a potential whose minimum sits precisely at the CP-conserving point. The outcome is that strong-CP conservation becomes automatic, an output of the dynamics rather than a number dialled in by hand.

Step 1. Postulate a global $U(1)_{PQ}$ symmetry that is (i) spontaneously broken at a scale $f_a$ , giving a massless Goldstone boson $a$ with a shift symmetry $a\to a+\text{const}$ , and (ii) anomalous under QCD, so the shift changes the coefficient of $G\tilde G$ . Then $a$ can only ever appear as $(\bar\theta+a/f_a)$ , and by redefining the field we absorb $\bar\theta$ entirely into $a$ . Why: an exact shift symmetry would forbid any potential, but the anomaly breaks it just enough to let QCD generate one — this is the whole trick.

Step 2. The same instanton physics that made $\theta$ observable now generates a potential for the axion. Summing over the vacuum sectors gives $V(a)\simeq-\Lambda_{\mathrm{QCD}}^4\cos(\bar\theta+a/f_a)$ , periodic and bounded. Why: non-perturbative QCD is the only interaction that feels the anomalous combination, so it alone can produce the potential — and it necessarily produces a cosine, the leading harmonic of a $2\pi$ -periodic function.

Step 3. Minimise: $\partial V/\partial a=0$ gives $\sin(\bar\theta+\langle a\rangle/f_a)=0$ , so $\langle a\rangle=-f_a\bar\theta$ and the effective angle $\bar\theta+\langle a\rangle/f_a=0$ . The neutron electric dipole moment vanishes with it. Why: whatever value $\bar\theta$ started with — natural, tuned, or random — the field slides to cancel it, converting a $10^{-10}$ coincidence into a consequence of rolling downhill.

Step 4. The Vafa–Witten theorem guarantees that vector-like QCD conserves parity in its vacuum, so the minimum sits exactly at the CP-conserving point rather than nearby — the mechanism does not merely make $\bar\theta$ small, it makes it zero (up to tiny calculable electroweak corrections far below the experimental bound). Why: this elevates Peccei–Quinn from "another small number" to a real explanation — the answer is pinned by a theorem, not by luck.

Go deeper · Derivation C · 1 minThe axion mass from the QCD topological susceptibility

Derivation CThe axion mass from the QCD topological susceptibility

The purpose here is to turn the shape of the potential into a number, and to expose why the axion is simultaneously very light and very weakly coupled. Expanding $V(a)$ about its minimum gives a mass set by the curvature, which current algebra relates to the pion sector, yielding the benchmark $m_a\simeq5.7\,\mu\mathrm{eV}\,(10^{12}\,\mathrm{GeV}/f_a)$ and the fixed product $m_af_a$ that defines the axion line.

Step 1. Expand $V(a)=-\Lambda^4\cos(a/f_a)$ about the minimum: $m_a^2=V''(\langle a\rangle)=\Lambda^4/f_a^2\equiv\chi_{\rm top}/f_a^2$ , where $\chi_{\rm top}=\int d^4x\,\langle Q(x)Q(0)\rangle$ is the QCD topological susceptibility and $Q=\frac{g_s^2}{32\pi^2}G\tilde G$ is the topological charge density. Why: the mass is just the curvature of the well, and that curvature is a single, calculable QCD correlator — nothing about the axion sector itself enters except $f_a$ .

Step 2. Compute $\chi_{\rm top}$ in the chiral Lagrangian. At leading order the topological susceptibility is saturated by the pion and equals $\chi_{\rm top}=\dfrac{m_um_d}{(m_u+m_d)^2}\,m_\pi^2 f_\pi^2$ . Substituting, $m_a=\dfrac{\sqrt{m_um_d}}{m_u+m_d}\dfrac{m_\pi f_\pi}{f_a}$ . Why: the axion mixes with the neutral pion through the anomaly, so its mass inherits the pion scale — this is why $m_a f_a\approx m_\pi f_\pi$ up to a quark-mass factor, and why the answer is a QCD constant, not a free input.

Step 3. Put in numbers: with $m_u/m_d\approx0.47$ , $m_\pi=135\,\mathrm{MeV}$ , $f_\pi=92\,\mathrm{MeV}$ , the reduced factor is $\sqrt{m_um_d}/(m_u+m_d)\approx0.47$ , so $m_af_a\approx5.8\times10^3\,\mathrm{MeV}^2$ . Dividing by $f_a=10^{12}\,\mathrm{GeV}=10^{15}\,\mathrm{MeV}$ gives $m_a\approx5.8\,\mu\mathrm{eV}$ ; the full next-to-leading and lattice treatment sharpens this to the quoted $5.7\,\mu\mathrm{eV}$ . Why: this is the master calibration of the whole field — every experiment converts its mass reach into an $f_a$ reach through this one line.

Step 4. Read off the scaling. Because $m_a\propto1/f_a$ and $g_{a\gamma}\propto1/f_a$ , a lighter axion is more weakly coupled: raising $f_a$ from the electroweak scale to $10^{12}\,\mathrm{GeV}$ shrinks both the mass and the coupling by ten orders of magnitude, turning the originally "visible" axion into the "invisible" axion that survives all laboratory and astrophysical bounds. Why: this single inverse scaling is what pushed the field from the ruled-out original model to the viable KSVZ and DFSZ invisible-axion models still searched today.

Limiting cases

Case	Result	Meaning
$\bar\theta=0$	$d_n=0$ , strong CP exactly conserved	the observed situation; no principle in the SM requires it
$\bar\theta\sim1$ (natural)	$d_n\sim3\times10^{-16}\,e\,\mathrm{cm}$	excluded by the measured bound by ten orders of magnitude — the puzzle
$m_u\to0$ (massless up quark)	$\bar\theta$ unphysical, rotatable away	would solve strong CP for free, but lattice QCD finds $m_u\neq0$
$f_a\sim v_{\rm EW}\ (250\,\mathrm{GeV})$	visible axion, $m_a\sim100\,\mathrm{keV}$	the original PQWW model — ruled out by beam-dump and rare-decay searches
$f_a\sim10^{9}\text{–}10^{12}\,\mathrm{GeV}$	invisible axion, $m_a\sim\mu\mathrm{eV}$ – $\mathrm{meV}$	the classic open window: evades astrophysics, can be the dark matter
$f_a\gtrsim10^{12}\,\mathrm{GeV}$	misalignment overproduces axions	$\Omega_a>\Omega_{\rm DM}$ unless the initial angle $\theta_i$ is tuned small

Common trapThe theta-term is a total derivative, so it can just be dropped

This is the most seductive mistake in the subject, and it is wrong. Yes, $G\tilde G=\partial_\mu K^\mu$ is a total derivative, and in an Abelian theory like electrodynamics the analogous term genuinely is unphysical — you can integrate it away. But QCD is non-Abelian, and its vacuum contains gauge configurations (instantons) that do not fall off fast enough at infinity for the surface term to vanish. Their winding number makes $\int G\tilde G$ a non-zero integer, so $\theta$ weights physically distinct vacuum sectors and has real, CP-violating consequences — most directly, a neutron electric dipole moment. The correct statement is subtler: $\theta$ has no effect in perturbation theory (order by order in Feynman diagrams it never appears), but it is a genuine non-perturbative parameter. Dismissing the strong-CP problem because "it's a total derivative" is exactly the reasoning 't Hooft's 1976 instanton analysis overturned.

Worked examples

1 · From the neutron EDM to the bound on theta-bar (fully worked)

The most sensitive neutron-EDM experiment (PSI, 2020) reports $|d_n|<1.8\times10^{-26}\,e\,\mathrm{cm}$ . Using the chiral estimate $d_n\simeq3\times10^{-16}\,\bar\theta\ e\,\mathrm{cm}$ , find the bound on $\bar\theta$ and quantify the fine-tuning.

Step 1. Invert the estimate. Setting $|d_n|=3\times10^{-16}\,\bar\theta\ e\,\mathrm{cm}$ equal to the experimental ceiling gives $\bar\theta<\dfrac{1.8\times10^{-26}}{3\times10^{-16}}=\mathbf{6\times10^{-11}}$ , i.e. $\bar\theta\lesssim10^{-10}$ .

Step 2. Translate to a physical length. At the bound the charge separation inside the neutron is $d_n/e\approx1.8\times10^{-26}\,\mathrm{cm}$ . The neutron's own radius is about $10^{-13}\,\mathrm{cm}$ , so the implied displacement is smaller than the neutron by a factor of $\sim10^{13}$ — comparable to sliding the poles of the Earth apart by the width of a single atom.

Step 3. Quantify the puzzle. On symmetry grounds $\bar\theta$ combines two a priori unrelated sectors — the QCD vacuum angle $\theta$ and the phase of the quark mass matrix $\arg\det M_q$ — each expected to be of order one. That they cancel to better than one part in $10^{10}$ is the fine-tuning. The bound is $\bar\theta\lesssim6\times10^{-11}$ ; a "natural" $\bar\theta\sim1$ is excluded by ten orders of magnitude, which is the strong CP problem stated as a number.

2 · Axion mass, Compton frequency, and why haloscopes use microwaves

Take the benchmark decay constant $f_a=10^{12}\,\mathrm{GeV}$ . Find the axion mass, its Compton frequency, and comment on the detector this implies.

Step 1. Mass from equation (5): $m_a=5.7\,\mu\mathrm{eV}\times(10^{12}\,\mathrm{GeV}/f_a)=\mathbf{5.7\,\mu eV}$ . In SI this is $m_a c^2=5.7\times10^{-6}\,\mathrm{eV}$ , about $10^{11}$ times lighter than the electron.

Step 2. Convert to a frequency via $\nu=m_ac^2/h$ . Using $1\,\mathrm{eV}\leftrightarrow2.418\times10^{14}\,\mathrm{Hz}$ , $\nu=5.7\times10^{-6}\times2.418\times10^{14}\,\mathrm{Hz}=\mathbf{1.38\ GHz}$ — squarely in the microwave band, near the frequencies of mobile-phone and radar electronics.

Step 3. Interpret the detector. A dark-matter axion of this mass, converting to a photon in a strong magnetic field (the Primakoff effect), deposits its rest energy plus a tiny kinetic energy as a $1.38\,\mathrm{GHz}$ photon. To catch it you build a resonant microwave cavity tuned to that frequency inside a several-tesla magnet and slowly scan the resonance across the mass range — precisely the ADMX haloscope strategy. A $\mu$ eV-scale axion rings a GHz cavity; the mass is unknown, so the experiment is a slow scan, which is why the search takes decades.

3 · Misalignment dark matter — fixing f_a from the relic abundance

The vacuum-misalignment mechanism gives $\Omega_a h^2\simeq0.12\,(f_a/9\times10^{11}\,\mathrm{GeV})^{7/6}\,\theta_i^2$ . For an order-one initial angle $\theta_i=1$ , what $f_a$ (and $m_a$ ) reproduces the observed dark-matter density $\Omega_{\rm DM}h^2=0.12$ ?

Step 1. Set $\Omega_a h^2=0.12$ with $\theta_i=1$ : the bracket must equal one, so $f_a=\mathbf{9\times10^{11}\ GeV}$ — remarkably, close to the grand-unification-adjacent scale, with no tuning of $\theta_i$ .

Step 2. Corresponding mass from equation (5): $m_a=5.7\,\mu\mathrm{eV}\times(10^{12}/9\times10^{11})=5.7\times1.11=\mathbf{6.3\ \mu eV}$ , giving a Compton frequency of about $1.5\,\mathrm{GHz}$ .

Step 3. Read the coincidence. The very same $f_a\sim10^{12}\,\mathrm{GeV}$ that makes the invisible axion cosmologically safe also makes it, without extra ingredients, all of the dark matter for a natural misalignment angle. A QCD axion near $6\,\mu$ eV solves strong CP and supplies the dark matter at once — which is why the microwave window is the prime hunting ground.

Physics of the portrait

The strong CP problem crystallised in the mid-1970s once 't Hooft showed the QCD vacuum had a non-trivial topology, and its most enduring answer came from a single 1977 paper by a theorist then at Stanford and his student.

Origins · 1 minRoberto Peccei and how this was discovered

Roberto Peccei

1942–2024 · the dynamical solution to the strong CP problem

Roberto Peccei was an Italian-American theoretical physicist whose name is permanently paired with that of Helen Quinn through the mechanism they proposed together at Stanford in 1977. The problem was fresh: Gerard 't Hooft had just shown in 1976 that instantons make the QCD $\theta$ -angle physical, which solved the old " $U(1)_A$ problem" of the heavy $\eta'$ meson but created a new one — why $\bar\theta$ is so absurdly small. Peccei and Quinn refused to accept $\bar\theta$ as a constant at all: they introduced a new global chiral symmetry, now called $U(1)_{PQ}$ , whose spontaneous breaking lets the effective angle relax to zero dynamically. Within months Steven Weinberg and Frank Wilczek independently noted what this implied — that a spontaneously broken global symmetry must leave behind a light pseudo-Goldstone boson. Wilczek named it the axion, after a brand of laundry detergent, because it "cleaned up" the strong-CP problem. The original electroweak-scale axion was quickly excluded, but the KSVZ (Kim; Shifman–Vainshtein–Zakharov) and DFSZ (Dine–Fischler–Srednicki–Zhitnitsky) models pushed the symmetry-breaking scale far higher, making the axion nearly invisible yet still a solution. Pierre Sikivie then showed in 1983 how even an invisible axion could be detected through its conversion to photons in a magnetic field, launching the experimental program that continues today. Peccei led the field for decades from UCLA; the axion he and Quinn made necessary is now among the most actively hunted particles in physics.

History

1976

— 't Hooft shows instantons make the QCD θ-vacuum physical, sharpening the strong CP problem

1977

— Peccei and Quinn propose a new U(1) symmetry that relaxes θ̄ to zero

1978

— Weinberg and Wilczek independently identify the light boson it implies: the axion

1979

— Crewther, Di Vecchia, Veneziano and Witten compute the neutron EDM from θ̄

1983

— Preskill, Wise, Wilczek and others show a light axion can be the cold dark matter

1983

— Sikivie proposes the haloscope and helioscope, making the invisible axion detectable

2020

— the PSI experiment sets the neutron-EDM bound at 1.8×10⁻²⁶ e·cm, so θ̄ ≲ 10⁻¹⁰

Going further

Texts. Peskin & Schroeder, An Introduction to Quantum Field Theory, §19.1–19.3 (the axial anomaly and the $\theta$ -vacuum) · Weinberg, The Quantum Theory of Fields, Vol. II, §23.6 (the strong CP problem and the axion) · Schwartz, Quantum Field Theory and the Standard Model, Ch. 30 (anomalies) and §29.6 · Thomson, Modern Particle Physics, §18.1–18.2 (Standard Model parameters and open questions) · Cheng & Li, Gauge Theory of Elementary Particle Physics, Ch. 16. The key numerical checks in this article are: (1) $\bar\theta<6\times10^{-11}$ follows from $d_n\simeq3\times10^{-16}\bar\theta\,e\,$ cm and $|d_n|<1.8\times10^{-26}\,e\,$ cm; (2) $m_a=5.7\,\mu$ eV at $f_a=10^{12}\,$ GeV from $m_af_a=\sqrt{m_um_d}/(m_u+m_d)\times m_\pi f_\pi$ with $m_u/m_d\approx0.47$ ; (3) the Compton frequency $1.38\,$ GHz; (4) the misalignment scale $f_a\approx9\times10^{11}\,$ GeV for $\Omega_a h^2=0.12$ , $\theta_i=1$ ; (5) KSVZ/DFSZ couplings $g_{a\gamma}\approx(1.5\text{–}3.9)\times10^{-16}(m_a/\mu\mathrm{eV})\,$ GeV $^{-1}$ .

Prerequisites. Colour and QCD · CP violation in the Standard Model · Standard Model summary. Leads to. Dark matter · Open questions in particle physics · Grand Unification.

Ask the tutor: Ask me to show exactly how the anomaly ties the quark-mass phase to $\theta$ , or to work through how a $\mu$ eV axion becomes a signal in a microwave cavity.

If θ̄ is just a small number, why call it a "problem" rather than an accident?

Because it is unnaturally small in a technical sense. A parameter is called natural if setting it to zero enhances the symmetry of the theory — then quantum corrections keep it small and its tininess explains itself. But $\bar\theta$ enjoys no such protection: setting it to zero buys you nothing, because $\bar\theta$ combines two independent sectors (the gluon vacuum angle and the phase of the quark mass matrix) that have no reason to know about each other. Both should be of order one, yet their sum is below $10^{-10}$ . That is a coincidence at the one-part-in-ten-billion level with no symmetry reason behind it — what physicists mean by a fine-tuning problem, as opposed to a merely small measured number like the electron mass.

What is the difference between θ and θ̄?

$\theta$ is the bare coefficient of the $G\tilde G$ term in the QCD Lagrangian, but it is not by itself observable. A chiral rotation of the quark fields shifts $\theta$ (through the axial anomaly) while simultaneously changing the phase of the quark mass matrix, $\arg\det M_q$ ; only the combination $\bar\theta=\theta+\arg\det M_q$ is invariant under these rotations, so only $\bar\theta$ can enter physical predictions. Think of $\theta$ and $\arg\det M_q$ as two coordinates that rotate into each other, with $\bar\theta$ the true, rotation-invariant physical angle. Everything measurable — above all the neutron EDM — depends on $\bar\theta$ , never on $\theta$ alone.

In what sense does the axion solve the problem "dynamically"?

In the Standard Model you resolve the puzzle by simply declaring $\bar\theta=0$ — an input you dial in by hand and hope nature respects. The Peccei–Quinn mechanism instead makes $\bar\theta$ the value of a field, the axion. QCD generates a potential $V\propto-\cos(\bar\theta+a/f_a)$ for that field, and like a ball in a bowl the field rolls to the bottom, which sits exactly where the effective angle is zero. So the answer $\bar\theta_{\rm eff}=0$ is produced by the equations of motion rather than assumed. Whatever value the angle would otherwise have had, the axion field slides to cancel it. That is the meaning of "dynamical": the smallness is an output of relaxation, not a tuned initial condition.

Couldn't a massless up quark solve strong CP without any axion?

Logically, yes — and it would be the most economical solution. If any single quark were exactly massless, then $\det M_q=0$ , and the phase $\arg\det M_q$ becomes meaningless: you can rotate $\bar\theta$ to any value you like, so there is no physical strong-CP angle at all. For years the " $m_u=0$ " hypothesis was a live contender precisely because it needs no new particles. What killed it is data: modern lattice QCD determines the up-quark mass to be about $2\,$ MeV, non-zero at many standard deviations. With $m_u\neq0$ established, that escape route is closed, which is a large part of why the axion — costing a new symmetry and a new particle — is taken so seriously as the remaining natural option.

How can a particle proposed to fix strong CP also be the dark matter?

Through a cosmological accident of timing. In the early universe, before QCD switches on the axion potential, the field is frozen at some random initial angle $\theta_i$ . When the Hubble rate drops to the axion mass, the field begins to oscillate, and these coherent oscillations are a condensate of enormous numbers of extremely cold axions — energy density but negligible pressure, exactly the definition of cold dark matter. The relic abundance matches the observed dark-matter density for $f_a\sim10^{12}\,$ GeV and an order-one $\theta_i$ , the same scale that makes the axion invisible to laboratory and stellar bounds. So one hypothesis, introduced only to cancel a vacuum angle, doubles as a complete explanation of the dark matter — the reason the axion is doubly motivated.

Have axions been discovered, and how would we know?

Not yet. No experiment has produced a confirmed signal; current results are exclusion limits carving away parameter space. Detection would most likely come from a haloscope: a resonant microwave cavity in a strong magnetic field, in which a dark-matter axion converts to a single photon at a frequency set by its mass (a few $\mu$ eV axion gives a $\sim1.4\,$ GHz photon). Because the mass is unknown, the cavity must be tuned slowly across a range, so a search is inherently a long scan; ADMX has now reached the theoretically motivated coupling in the few- $\mu$ eV window. A discovery would appear as a narrow excess of power at one frequency, tracking the local dark-matter velocity distribution — a signal so specific it would be hard to fake, and one of the cleanest possible dark-matter detections.