---
title: 'Model-Independent Axion: String Theory & QCD'
url: https://www.emergentmind.com/topics/model-independent-axion
type: topic
---

# Model-Independent Axion: String Theory & QCD

Searching arXiv for relevant papers on the model-independent axion and recent axion-photon coupling results.
The **model-independent axion** has two closely related but distinct meanings in the axion literature. In perturbative string theory, especially weakly coupled heterotic compactifications, it denotes the universal pseudoscalar descending from the ten-dimensional antisymmetric tensor field \(B_{MN}\); its defining feature is that it appears in every compactification and participates in Green–Schwarz anomaly cancellation [2401.17354]. In QCD axion phenomenology, the same phrase is also used for the **model-independent contribution** to the axion–photon coupling, namely the part generated purely by QCD dynamics and therefore independent of the ultraviolet realization of the Peccei–Quinn symmetry [2603.29153]. Both usages are standard, and both are central to current work connecting string compactifications, low-energy effective theory, lattice QCD, and axion searches.

## 1. Terminology and conceptual scope

In weakly coupled heterotic or Type II string vacua one always finds an NS–NS two-form gauge field \(B_2\) in ten dimensions. Its zero-mode in four dimensions is often called the “model-independent axion,” a name emphasizing that it appears in every compactification regardless of the details of the internal manifold [2401.17354]. By contrast, “model-dependent” axions arise from Ramond–Ramond \(p\)-form gauge fields \(C_p\) or from \(B_2\) reduced on non-trivial 2-cycles; their number and couplings depend on the topology and geometry of the compactification.

A distinct but equally standard usage appears in axion–photon phenomenology. The low-energy interaction between an axion field \(A(x)\) and the electromagnetic field is conventionally written as
\[
L_{a\gamma\gamma}=-\tfrac14\,g_{a\gamma\gamma}\,A\,F_{\mu\nu}\tilde F^{\mu\nu},
\]
with
\[
g_{a\gamma\gamma}=
\underbrace{\frac{E}{N}\,\frac{\alpha}{2\pi f_A}}_{\text{model-dependent}}
+
\underbrace{g_{a\gamma\gamma}^{\rm QCD}}_{\text{model-independent}}.
\]
Here \(E/N\) is the ratio of electromagnetic to color anomaly coefficients of the Peccei–Quinn current, \(\alpha=e^2/(4\pi)\), \(f_A\) is the axion decay constant, and the second term is generated purely by QCD dynamics [2603.29153]. The phrase “model-independent axion” therefore refers either to a universal axionic degree of freedom in string theory or to a universal contribution to an axion coupling in low-energy QCD. The overlap in terminology is historical rather than accidental: in both cases the emphasis is on universality.

## 2. String-theoretic origin from the ten-dimensional two-form

In the weakly coupled heterotic string one has a ten-dimensional antisymmetric tensor field
\[
B_{MN}\qquad (M,N=1\ldots 10).
\]
Upon compactification to four dimensions, the four-dimensional components \(B_{\mu\nu}\) give rise to a pseudoscalar axion field \(a_{\rm MI}(x)\) via
\[
B_{\mu\nu}(x)\propto \varepsilon_{\mu\nu\rho\sigma}\,\partial^\rho a_{\rm MI}(x),
\]
and because this mode descends from the universal two-form it is called the model-independent axion [1405.0221]. In the more general effective-field-theory presentation, the ten-dimensional action contains
\[
S_{10}=-\frac{1}{2\kappa_{10}^2}\int d^{10}x\,\sqrt{-G}\,e^{-2\Phi} H_{MNP}H^{MNP},
\]
with
\[
H_3=dB_2-\frac{\alpha'}{4}\left(\Omega_3^{\rm YM}-\Omega_3^{\rm Lorentz}\right),
\]
where \(\Omega_3^{\rm YM}\) and \(\Omega_3^{\rm Lorentz}\) are the Yang–Mills and gravitational Chern–Simons three-forms [2401.17354]. The Green–Schwarz term \(B_2\wedge (\mathrm{tr}\,F^2-\mathrm{tr}\,R^2)\) sits inside this structure and guarantees ten-dimensional anomaly cancellation.

In many string vacua there survives an anomalous Abelian gauge symmetry \(U(1)_{ga}\), whose gauge boson acquires a mass by absorbing the MI axion; this is a four-dimensional version of the Green–Schwarz mechanism [1405.0221]. Below that mass scale one is left with an exact global Peccei–Quinn-type symmetry \(U(1)_{PQ}\) under which the MI axion shifts. This distinguishes the MI axion from field-theoretic KSVZ or DFSZ axions: they do not arise from the 10D two-form, do not carry a universal coupling to every gauge anomaly, and do not participate in Green–Schwarz anomaly cancellation.

## 3. Effective couplings, decay constant, and universal anomaly structure

After compactification on a six-dimensional manifold \(X_6\) of volume \(V_6\), one may expand
\[
B_2(x,y)=a(x)\,\omega_2(y)+\cdots,
\]
where \(\omega_2\) is a harmonic two-form normalized so that \(\int_{X_6}\omega_2\wedge *\omega_2=1\), and define the four-dimensional axion
\[
a(x)\equiv \int_{C_2} B_2 .
\]
Inserting this ansatz into the ten-dimensional action gives the four-dimensional kinetic term
\[
S_4\supset -\frac12\int d^4x\, f_a^2\,(\partial_\mu a)(\partial^\mu a),
\]
with
\[
f_a^2 \simeq e^{-2\langle\Phi\rangle}V_6 M_s^8/\kappa_{10}^2
\simeq \frac{M_P^2}{4\pi V_6 M_s^6},
\]
or equivalently
\[
f_a \sim M_s e^{-\langle\Phi\rangle}\sqrt{V_6}\sim \frac{M_P}{\sqrt{V_6 M_s^6}}.
\]
In typical weakly coupled heterotic models one finds \(f_a\sim 10^{16}\!-\!10^{18}\,\mathrm{GeV}\), while in warped or large-volume Type IIB compactifications \(f_a\) can be lowered to the \(10^9\!-\!10^{16}\,\mathrm{GeV}\) window [2401.17354].

The anomaly couplings inherit their structure from the Green–Schwarz term. Dimensionally reducing \(B_2\wedge(\mathrm{tr}\,F^2-\mathrm{tr}\,R^2)\) gives
\[
L_{\rm int}=
\frac{a}{f_a}\frac{g^2}{32\pi^2}F^a_{\mu\nu}\tilde F^{a\mu\nu}
-
\frac{a}{f_a}\frac{1}{384\pi^2}R_{\mu\nu\rho\sigma}\tilde R^{\mu\nu\rho\sigma},
\]
and, in the four-dimensional Einstein-frame formulation used for the MI axion contribution to gravity,
\[
S_{\rm GS}\supset \int d^4x\,\sqrt{-g}\,
\frac{a_{\rm MI}}{f_a}
\left(
\frac{1}{32\pi^2}F\tilde F
-
\frac{1}{768\pi^2}R\tilde R
\right)
\]
[1608.02540]. The perturbative shift symmetry \(a\to a+\mathrm{const}\) is exact at the perturbative level and is broken to a discrete subgroup by non-perturbative effects such as worldsheet or D-brane instantons and gaugino condensation. The induced potential is of cosine form,
\[
V(a)\sim A e^{-S_{\rm inst}}\,[1-\cos(a/f_a+\delta)],
\]
with mass
\[
m_a^2 \simeq \frac{1}{f_a^2}\Lambda^4 e^{-S_{\rm inst}},
\qquad
m_a\sim \Lambda^2/f_a
\]
[2401.17354].

## 4. Mixing, anomalous \(U(1)\), and the domain-wall number

Below the \(U(1)_{ga}\) gauge-boson mass \(M_A\), the light fields may include the MI axion \(a_{\rm MI}\) with decay constant \(F_{\rm MI}\), a QCD axion \(a_2\) with decay constant \(f_2\), and a hidden-sector axion \(a_3\) with decay constant \(f_3\). The relevant kinetic and anomaly terms can be written as
\[
L \supset
\frac12(\partial_\mu a_{\rm MI})^2
+\frac12(\partial_\mu a_2)^2
+\frac12(\partial_\mu a_3)^2
-\frac{a_{\rm MI}}{32\pi^2F_{\rm MI}}(G\tilde G+F_h\tilde F_h)
-\frac{N\,a_2}{32\pi^2 f_2}G\tilde G
-\frac{N\,a_3}{32\pi^2 f_3}F_h\tilde F_h ,
\]
where \(N\simeq O(10^2)\) is a large model-dependent integer from the sum of PQ charges in the string spectrum [1405.0221]. One may equivalently work in the basis of three orthogonal pseudoscalar currents \(\theta_{e,f,g}\); only \(\theta_f\) and \(\theta_e\) couple to non-Abelian anomalies, while \(\theta_g\) is an exactly flat direction.

A central consequence is the reduction of the physical domain-wall number to unity. In a toy model with three heavy quark flavors, the low-energy QCD axion coupling appears at first sight to imply \(N_{DW}=3\). In the full string construction, however, the three vacua are related by shifts of the MI axion originating from the anomalous \(U(1)_{ga}\). The true physical domain-wall number is
\[
N_{DW}^{\rm phys}=\gcd(N_{\rm MI},N_{\rm QCD}),
\]
with \(N_{\rm MI}=1\) and \(N_{\rm QCD}=3\), hence
\[
N_{DW}^{\rm phys}=\gcd(1,3)=1.
\]
This is the “Choi–Kim mechanism” [1405.0221]. Under high-scale inflation, the distinction is cosmologically decisive: for \(N_{DW}=1\), string–wall systems become unstable once they acquire a tilt from the QCD potential at \(T\sim \Lambda_{\rm QCD}\) and collapse rapidly, whereas \(N_{DW}\ge 2\) leads to a network that cannot fully annihilate and disastrously dominates the energy density.

The same framework gives characteristic scales. The MI axion decay constant is generically at or above the GUT scale, \(F_{\rm MI}\gtrsim 10^{16}\,\mathrm{GeV}\), while the QCD axion decay constant from misalignment, including string radiation, satisfies
\[
f_a \lesssim (1.0\!-\!1.7)\times 10^{11}\,\mathrm{GeV}
\]
under the assumptions quoted in the source summary [1405.0221]. A plausible implication is that the MI axion is most naturally a high-scale universal mode whose principal cosmological role is often indirect, through mixing and vacuum identification, rather than as the dominant low-energy QCD axion itself.

## 5. Model-independent axion–photon coupling from QCD

For the QCD axion, the model-independent contribution to the photon coupling is a finite, renormalization-group-invariant number when expressed as \(g_{a\gamma\gamma}^{\rm QCD} f_A/e^2\). In the presence of a constant imaginary background electric field \(E\) and magnetic field \(B\) pointing in the \(z\)-direction, one may define
\[
g_{a\gamma\gamma}^{\rm QCD}
=
\frac{1}{V_4 f_A}\,
\lim_{E\!\cdot\!B\to 0}
\frac{\langle Q_{\rm top}\rangle_{E,B}}{iE\!\cdot\!B},
\]
where \(V_4\) is the Euclidean four-volume and \(Q_{\rm top}\) is the gluonic topological charge [2603.29153]. The 2026 lattice calculation presents the first non-perturbative determination of this QCD contribution using continuum extrapolated simulations.

The simulations employ \(N_f=2+1\) rooted staggered quarks, physical quark masses, and seven lattice spacings \(a\in[0.08\ldots 0.29]\,\mathrm{fm}\). Finite volumes satisfy \(m_\pi L\gtrsim 3.2\), and both volume effects and very low-temperature effects are explicitly checked. Imaginary uniform electric field and real magnetic field are implemented by quantized fluxes \(n_e,n_b\), and the ratio \(\langle Q_{\rm top}\rangle/(iE\!\cdot\!B)\) is measured on each ensemble. The continuum result is obtained by a two-step extrapolation: first a polynomial fit in the two invariants \((E\!\cdot\!B)\) and \((E^2+B^2)\to 0\) at fixed \(a\), and then a polynomial fit in \(a^2\to 0\). Statistical errors are estimated via jackknife, while systematic errors from varying fit ranges, orders, operator definitions, flow times, and related choices are combined through an AIC-weighted model average [2603.29153].

Two independent extraction methods are used. The **gluonic method** directly measures \(Q_{\rm top}\) via an improved lattice operator with tree-level \(O(a^4)\) discretization plus gradient flow, and also tests rounding \(Q_{\rm top}\) to the nearest integer on each configuration to assess \(a\to 0\) artifacts. The **fermionic method** uses the axial Ward identity in background fields,
\[
\partial_\mu\langle \bar\psi_f\gamma_5\gamma_\mu\psi_f\rangle
=
2m_f\langle \bar\psi_f\gamma_5\psi_f\rangle
+2Q_{\rm top}
+\frac{e^2 q_f^2}{4\pi^2}\,iE\!\cdot\!B ,
\]
integrated over \(V_4\) to relate \(Q_{\rm top}\) to the pseudoscalar condensate \(P_f=\langle \bar\psi_f\gamma_5\psi_f\rangle\). Four flavor combinations, \(u\), \(d\), \(s\), and \((u+d)/2\), yield consistent continuum results.

The final continuum value at physical \(m_u\ne m_d\) is
\[
g_{a\gamma\gamma}^{\rm QCD}\,\frac{f_A}{e^2}
=
-0.0224(2)_{\rm stat}(2)_{\rm def}(5)_a(8)_{EB}(1)_{\rm vol}(2)_m
=
-0.0224(10),
\]
or equivalently
\[
g_{a\gamma\gamma}^{\rm QCD}
=
-1.77(8)\,\frac{\alpha}{2\pi f_A}.
\]
The error brackets indicate statistics, choice of topological-charge definition, continuum extrapolation, \(E,B\to 0\) extrapolation, volume effects, and isospin-breaking correction. This establishes a universal nonzero QCD contribution and therefore a model-independent floor for the axion–photon coupling [2603.29153].

A closely related chiral-theory analysis pursues the model-independent component of the axion–photon–photon coupling in \(U(3)\) chiral perturbation theory up to next-to-leading order, emphasizing strong isospin breaking. In that treatment, including complete linear isospin-breaking terms in the \(\pi^0\)-\(\eta\)-\(\eta'\)-axion system raises the model-independent coupling from \(1.63\) to \(1.89\) in units of \(\alpha_{\rm em}/(2\pi f_a)\), corresponding to a \(+16\%\) shift [2411.06737]. This does not supersede the lattice determination; rather, it shows that isospin-breaking effects are quantitatively important in precision analyses of the model-independent contribution.

## 6. Phenomenology, searches, and related observables

Because every axion model satisfies
\[
g_{a\gamma\gamma}=
\frac{E}{N}\frac{\alpha}{2\pi f_A}
+
g_{a\gamma\gamma}^{\rm QCD},
\]
the lattice result fixes an irreducible QCD term of magnitude \(\simeq 0.0224\,e^2/f_A\), opposite in sign to many model-dependent contributions [2603.29153]. In particular, KSVZ \((E/N=0)\) and DFSZ \((E/N=8/3)\) shift the total coupling by approximately \(\mp(1.77-0.14)\,\alpha/(2\pi f_A)\), so the QCD piece reduces or enhances the bare anomaly term by about \(10\%\). Some values of \(E/N\) yield near-cancellation, most notably near \(E/N\approx 1.77\), producing “photophobic” axion models.

This sharpens the interpretation of laboratory and astrophysical searches. The lattice determination narrows the target region for helioscopes such as CAST and IAXO, haloscopes such as ADMX and HAYSTAC, and resonant-cavity or dielectric-plate searches such as DMRadio and successors [2603.29153]. Conversely, failure to detect an axion in the region bounded by \(|g_{a\gamma\gamma}|\ge |g^{\rm QCD}|\) would begin to rule out entire classes of Peccei–Quinn realizations, not merely benchmark models.

For higher axion masses, the PVLAS–Ferrara ellipsometer provides a model-independent laboratory upper bound on \(g_{a\gamma\gamma}\) above \(m_a=10^{-3}\,\mathrm{eV}\) [1410.4081]. The apparatus uses a high-finesse Fabry–Pérot cavity, two rotating permanent dipole magnets with \(B=2.5\,\mathrm{T}\) and length \(L=0.8\,\mathrm{m}\) each, and a Nd:YAG laser at \(\lambda=1064\,\mathrm{nm}\). After \(210\,\mathrm{h}\) of vacuum data, the baseline noise around \(2\nu_{\rm mag}\) is
\[
\psi^{\rm(noise)}\simeq 2.5\times 10^{-9},
\]
to be compared with the QED expectation
\[
\psi^{\rm(QED)}\approx 5\times 10^{-11}.
\]
Requiring the axion-induced ellipticity to remain below the noise floor yields, at \(95\%\) C.L., representative bounds
\[
m_a=10^{-3}\,\mathrm{eV}\ \Rightarrow\ g_{a\gamma\gamma}<3\times 10^{-7}\,\mathrm{GeV}^{-1},
\]
\[
m_a=10^{-2}\,\mathrm{eV}\ \Rightarrow\ g_{a\gamma\gamma}<5\times 10^{-7}\,\mathrm{GeV}^{-1},
\]
\[
m_a=10^{-1}\,\mathrm{eV}\ \Rightarrow\ g_{a\gamma\gamma}<1\times 10^{-6}\,\mathrm{GeV}^{-1}.
\]
The significance of this result is that it is entirely terrestrial and does not rely on assumptions about cosmic or solar axion-like-particle fluxes.

The MI axion also couples universally to \(R\tilde R\), which motivates gravitational applications. In the gravitational-wave context, the MI axion contribution to an apparent graviton mass is estimated by
\[
\Delta m_g^2 \simeq
\frac{1}{M_P^2}\,\frac{a_{\rm MI}}{f_a}\,
\bigl\langle R_{\alpha\beta\gamma\delta}R^{\alpha\beta\gamma\delta}\bigr\rangle .
\]
Near a Schwarzschild black hole of mass \(M\), the Riemann-squared invariant at the horizon is
\[
I|_{r_s}
=
\frac{3}{4(GM)^4}
=
\frac34\,\frac{M_P^8}{M^4}.
\]
For GW150914 with \(M\simeq 30\,M_\odot\approx 6\times 10^{31}\,\mathrm{kg}\), one finds \(\Delta m_g\sim 10^{-48}\,\mathrm{eV}\), completely negligible compared with the LIGO bound \(m_g<1.2\times 10^{-22}\,\mathrm{eV}\); an observable effect at the \(10^{-22}\,\mathrm{eV}\) level would instead require black holes of mass \(M\sim 2\times 10^{14}\,\mathrm{kg}\) [1608.02540]. This suggests that gravitational-wave probes of the MI axion, while conceptually clean, are relevant only in an extreme low-mass black-hole regime far from current stellar-mass merger observations.

Source: https://www.emergentmind.com/topics/model-independent-axion