WorksPhysics and Astronomy
A Universal Axion in a Twin-QCD Sector
Abstract
Dark matter may consist of multiple components with distinct particle-physics origins. We investigate a scenario in which a universal axion couples to both visible and twin QCD, with twin baryons providing a possible dominant dark-matter component. The presence of a second strongly interacting sector can substantially enhance the axion mass while preserving its role in resolving the strong-CP problem, provided the physical vacuum angles are appropriately aligned. For a representative uniformly rescaled twin spectrum with a confinement scale five times that of ordinary QCD, the axion mass is enhanced by approximately a factor of 25. Using lattice-informed QCD results and a two-temperature description of the early universe, we investigate the cosmological evolution and relative abundances of the dark-sector components. In a representative conditional numerical benchmark, axions account for a small percentage of the observed dark matter, while twin baryons could constitute the remainder for a suitable primordial baryon asymmetry. We also examine the implications for additional relativistic species, strong-CP alignment, and the interactions of stable twin matter. Our results provide a quantitative framework for exploring mixed axion–twin-baryon dark matter and identify the particle-physics and cosmological conditions necessary for a viable realization, including constraints from the thermal history and formation of cosmic structure.
Scope and assumptions
Consider a twin sector with Standard Model gauge and matter content and a universal axion that couples to both color groups. A motivation is the visible/twin-sector structure of the Clemens–Raby F-theory construction [1]. The phenomenological benchmark examined here is
This note starts from the normalization of the proposed axion locus and extends the analysis to the thermal history and composition of the dark sector. The results are conditional effective-theory statements; they do not establish a complete construction or a fully tested cosmology.
The assumptions are that the axion is canonically normalized, the two color anomaly coefficients are equal, the physical strong-CP minima are aligned, and additional contributions to the axion potential are negligible. The visible photon relation below also assumes , large relative to hadronic scales, and no additional portal-induced correction to the visible photon coupling. Shared-axion constructions illustrate both the mass-enhancement mechanism and the importance of alignment [2, 3].
The crucial ultraviolet requirement is stronger than equality of the bare color anomalies. The canonically normalized light axion must couple to both color groups with the specified relative coefficient, and the two physical strong-CP angles, including quark mass determinant phases and heavy thresholds, must coincide to neutron-EDM accuracy. For the uniformly rescaled benchmark, , a small mismatch produces ; the twin susceptibility does not suppress the residual visible CP angle. Similarly, the conventional is an explicit phenomenological input until the light axion’s electromagnetic anomaly is derived from the specified construction. These questions are distinct from the low-energy mass enhancement and from the cosmological population calculations.
Independent inputs and the meaning of a twin sector
The equality of the two gauge and matter representations should not be confused with equality of their masses, temperatures, or baryon asymmetries. In particular, a visible-sector copy is not necessarily an exact exchange-symmetric vacuum. We use “twin” to denote the second Standard-Model-like sector, without assuming the Higgs-sector relations of a specific twin-Higgs model. The phenomenological inputs include
where controls masses only when the corresponding Yukawa couplings are specified, is the low-energy twin-to-visible electromagnetic coupling ratio, denotes a baryon-asymmetry yield, and is the initial angular displacement from the common axion minimum. Neither nor follows from the axion anomaly coefficients. Their precise conventions are defined below.
The F-theory construction of Ref. [1, 4] motivates studying visible and twin gauge sectors with independently adjustable scales. Its explicit example involves a very different, high-scale supersymmetric twin spectrum. Thus, that example is not itself a derivation of the low-energy spectrum or of used here. Obtaining this benchmark requires a separate choice of vacuum, thresholds, and axion direction. This distinction is important when turning a phenomenological relation into a compactification prediction.
There are three logically separate questions. The vacuum theory fixes the mass and couplings; the cosmological history fixes the populations; and the distributions of those populations fix the interpretation of dark-matter searches. A successful treatment must answer all three without replacing one by an assumption about another.
Mass enhancement: susceptibility rather than confinement alone
Normalize the color interactions as
Equal anomaly coefficients in this convention do not require equal running strong couplings. The zero-temperature mass follows from the curvature of the two vacuum energies. For aligned minima,
Here and are the visible and twin topological susceptibilities, and is the ordinary QCD axion mass at the same [3, 5].
If all QCD dimensionful parameters are rescaled together, the dimensionless ratios are unchanged and dimensional analysis gives . Thus,
The two contributions add in mass squared; adding to obtain 26 is not the appropriate combination.
Identical gauge and matter content does not, however, enforce identical mass ratios. To display this dependence, define . In the leading light-quark approximation,
so the scaling estimate yields
The light-quark expression and its connection to mirror-sector scales are discussed in Refs. [2, 5]. For unchanged light-quark masses, and give , rather than 25. These are scaling benchmarks; quark thresholds and the applicable chiral regime must be specified for a precision calculation. The equal-mass-ratio benchmark also requires the relevant heavier-quark thresholds to rescale consistently.
In ongoing work on the underlying F-theory construction, the visible and twin Yukawa couplings appear to coincide at the ultraviolet matching scale. This is a preliminary property of the construction, not yet a completed matching calculation. Equality of ultraviolet Yukawa couplings does not imply equal low-energy fermion masses: schematically, in each sector and , with the renormalized Yukawas evaluated at the appropriate scales. Differences in Higgs expectation values, renormalization-group running, and heavy-particle thresholds can all affect . In particular, a choice can motivate the uniform light-fermion rescaling only if the running and matching preserve the requisite mass ratios; it does not by itself produce .
The effective potential and its range of validity
Let and be the physical angles after including the phases of the quark mass determinants and any integrated-out colored states. With the convention in Eq. (3), define
The vacuum potential has the general form
Here denotes explicit breaking beyond the two color anomalies. Equation (4) follows by expanding around and neglecting this extra term. It does not require assuming that either QCD potential is an exact cosine.
For example, at leading order in the two-flavor chiral expansion, the visible contribution, shifted to vanish at its minimum, is
Its curvature is , while its large-angle shape differs from a cosine. The twin contribution is obtained by replacing its hadronic parameters, provided the twin theory remains in the same chiral regime [5]. Large-angle misalignment therefore requires more information than alone. In particular, a susceptibility fixes the quadratic term but not all higher derivatives of the potential.
For nonuniform light-quark masses, the two-flavor scaling generalizes to
Strange-quark effects, running-mass conventions, and higher-order low-energy constants should be included in a quantitative evaluation. If every twin quark is heavy compared with twin confinement, neither Eq. (6) nor its simple rescaling applies; the appropriate susceptibility is that of the corresponding approximately pure-glue theory, with its own matching coefficient. We are in the process of calculating the Yukawa couplings in our theory. At the moment it is apparent that the Yukawa couplings in the visible and twin sectors are identical.
Confinement-scale matching is a separate calculation
The confinement scale is exponentially sensitive to the running gauge coupling. At one loop, in a nonsupersymmetric interval with active Dirac quarks,
Continuity at a perturbative heavy-quark threshold gives
Applying this relation at top, bottom, and charm yields, within the same one-loop convention,
These equations make explicit the threshold sensitivity underlying mirror-axion mass estimates [2].
For identical six-flavor boundary couplings and a common rescaling , Eq. (14) would give , not . Thus produces only about 1.43 in this restricted example. The uniformly rescaled spectrum is perfectly meaningful as a low-energy benchmark, but it is not obtained merely by multiplying all quark masses by five while holding the ultraviolet coupling fixed. Different boundary couplings or additional thresholds can change this conclusion. Supersymmetric thresholds require the appropriate beta functions and cannot be inserted into Eq. (14) without modification.
The relation to the visible photon coupling
At masses well below the pion scale, the visible photon coupling is
while the ordinary QCD axion mass is
We use the central values from Ref. [5]. The ratio in Eq. (15) refers to the visible electromagnetic and visible color anomalies in the normalization of Eq. (3); it is not obtained by summing the two color-group anomalies in the denominator.
Using to eliminate gives
This is the relation that should be placed on a mass–coupling plot. For fixed , an enhancement moves the ordinary QCD axion locus rightward by a factor of 25 at fixed coupling, or downward by a factor of 25 at fixed physical mass. The displacement in either coordinate is decades. In the supplied figure, successive labeled mass prefixes, such as eV, meV, and eV, are separated by three decades.
A conventional unified benchmark
Complete multiplets with a common Peccei–Quinn charge within each multiplet give the familiar benchmark with standard hypercharge normalization [5]. Its applicability to the particular universal axion must be checked rather than assumed. Taking this value gives . With , Eq. (17) becomes
For example, at ,
These numerical values are conditional predictions of the stated benchmark, not a determination of the compactification parameters.
Decay-constant benchmarks and other interactions
Table I gives a useful conversion between the decay constant and the physical mass. All entries use the same intrinsic visible coupling; no relic-density assumption is built into the table. The two mass columns isolate the susceptibility dependence at fixed . Conversely, at a fixed measured mass, the larger- benchmark has a larger and consequently weaker interactions proportional to .
| [GeV] | [eV], | [eV], | [GeV] |
Table I. Particle-physics benchmarks for and . Masses and couplings use central values in Eqs. (15)–(16). The column is a leading chiral scaling estimate.
For an unbroken twin electromagnetic group, write
Equal color anomalies do not imply or . The twin electromagnetic anomaly, its quark-mass ratios, and possible kinetic mixing must be matched separately. The parameter , defined in Eq. (2), depends on the ultraviolet gauge couplings, charged-particle thresholds, and their running. Distinct unified gauge couplings can arise from independent geometric moduli in the motivating F-theory construction [1]; a larger twin GUT scale alone does not fix the sign or magnitude of . Equality is an additional benchmark when the relevant charge assignments and dimensionless hadronic parameters coincide. Twin-photon interactions can affect the thermal history even though the displayed experimental plane contains only the visible coupling [3].
Derivative fermion interactions may similarly be written as
Their coefficients are not fixed by . A model below an axion–photon exclusion can still be tested through other interactions. Conversely, suppressing the visible photon anomaly does not remove the color coupling used to generate the mass.
The intrinsic axion–photon locus
The visible mass–coupling relation in Eq. (17) refers to the intrinsic particle interaction, not the abundance-weighted sensitivity of an individual dark-matter experiment. For the unified benchmark and uniform rescaling , it predicts Eq. (18). At meV, the ordinary QCD axion has in this benchmark, while the uniformly rescaled twin sector gives . The alternative leading chiral estimate gives approximately at the same mass.
Figure 1 displays an illustrative compilation of axion–photon constraints attributed to the AxionLimits resource of O’Hare [6], with the uniform-scaling benchmark highlighted in red. Its marked point at agrees, to the resolution of the plot, with Eq. (19). The slope and normalization of the red segment are therefore consistent with Eq. (18); an additional near-cancellation in is not implied by the plotted benchmark. The background compilation draws on limits with different statistical, cosmological, and astrophysical assumptions. Its exact plot version has not been established, so the overlay should not be used to assign precision exclusion statements to individual regions [6].

Figure 1. Experimental and astrophysical constraints on the axion–photon coupling, adapted from the FIPS Workshop 2022 compilation by C. A. J. O’Hare, available through the AxionLimits project [6]. The superimposed red line represents our universal-axion benchmark with , corresponding to an axion mass enhancement . The red point indicates the benchmark with . Individual exclusion regions depend on the assumptions of the corresponding experiments, including the assumed local axion dark-matter abundance where applicable.
For comparison, Figure 2 shows only the particle-physics relations, without overlaid experimental regions. At fixed , the enhanced susceptibility increases the physical axion mass while leaving the assumed visible photon coupling unchanged. At fixed , the coupling is reduced by relative to the ordinary QCD axion when is held fixed.
Figure 2. Intrinsic axion–photon loci for : ordinary QCD, the leading chiral benchmark with , and the uniformly rescaled benchmark. The marked point corresponds to GeV in the uniform-scaling case. These curves show intrinsic couplings and not the relic abundance; the relation is approximate, and neither case is derived from a specified compactification.
Intrinsic coupling versus dark-matter abundance
The vertical axis of Figure 1 is the intrinsic particle coupling, not an abundance-weighted signal strength. This distinction is important if twin baryons supply most of the dark matter. Coupling compilations combine laboratory, stellar, cosmological, and dark-matter searches whose assumptions are not interchangeable [6].
For fixed apparatus and halo velocity distribution, the haloscope signal power scales schematically as
Defining
the rescaling of an all-axion-dark-matter coupling limit is
This relation holds with other experimental and halo assumptions fixed. It changes the inferred exclusion, not the intrinsic model line.
For example, under the illustrative assumption that only three percent of the local halo density is in axions, , Eq. (24) weakens an all-axion haloscope coupling limit by a factor . This assumes a specified halo density and does not change the intrinsic line in Figure 2. The local axion fraction need not equal the cosmological fraction if the dominant twin matter is dissipative.
Stellar-production constraints do not acquire the same halo-abundance rescaling. Cosmological and decay limits require a separate calculation of the axion population, lifetime, and decay channels. In particular, a common axion can couple to twin photons as well as visible photons; mirror-axion cosmologies illustrate the consequences for radiation and entropy transfer [3]. A white region in the supplied compilation should therefore not be read as proof that a complete twin-sector cosmology is viable.
Ultraviolet matching, strong-CP alignment, and axion quality
The first ultraviolet matching task is to determine the coefficients of , , and for the canonically normalized light axion. Together with the twin quark masses and thresholds, these determine the quantities that fix the locus: , , and . The twin-photon coupling and any additional axion potential must also be checked. An equal color coupling by itself does not specify the visible photon coefficient.
Alignment of the physical strong-CP angles
Strong-CP alignment remains a separate consistency condition. Near the aligned limit, expanding the two vacuum energies gives
Thus, when dominates, a mismatch of the physical strong-CP angles is not diluted. Their alignment, including mass-matrix phases, must follow from the underlying theory rather than from the mass formula alone [2, 3].
This result can be seen directly by expanding Eq. (9):
Minimization gives , reproducing Eq. (25). For , the coefficient is essentially unity. A representative strong-CP target therefore requires comparably accurate alignment of the physical angles. This order-of-magnitude target reflects the neutron electric-dipole constraint and its hadronic interpretation; it is not a new precision extraction of that constraint [5, 7].
Equal bare theta parameters are insufficient if the quark determinants have different phases. The matching must include
with a corresponding primed expression. Exchange or CP symmetries may control these terms, but symmetry breaking and complex heavy thresholds must be assessed in the specified model. There is one dynamical axion variable and, generically, two independent CP angles; simultaneous relaxation is not automatic.
Additional sources of the axion potential
Choose the angular origin at the aligned QCD minimum and parameterize one explicit-breaking contribution by
The allowed values of depend on the actual axion periodicity. For a perturbatively small contribution and small ,
Thus the condition is on the slope of the unwanted potential, not only its mass contribution. In the absence of a tuned cancellation,
A vanishing slope does not by itself guarantee an irrelevant correction to the mass: must also be compared with .
Using gives for the uniformly rescaled benchmark. The right side of Eq. (30) is then about . Enhancement strengthens the restoring potential by a factor of 626 relative to visible QCD alone, but does not justify ignoring arbitrary ultraviolet breaking. The improvement of axion quality from an aligned mirror potential is a separate issue from the alignment itself [8].
Canonical axion normalization and anomaly matching
A geometric axion coordinate need not be the canonically normalized light mass eigenstate. Let dimensionless periodic coordinates have kinetic matrix , with mass dimension two, and interactions
If all orthogonal modes can be integrated out, parameterize the light direction by , with
The coefficients have inverse-mass dimension. Choosing sets , and the benchmark requires . The visible ultraviolet electromagnetic anomaly ratio is in the stated normalization. A sum over two independent color groups is not the electromagnetic anomaly denominator.
More generally, around a common minimum the color contribution satisfies
This is useful for testing whether small changes in the light eigenvector alter the benchmark. If another axion is comparably light, the single-field reduction is insufficient: the full mass matrix must be diagonalized, and the photon coupling projected onto each eigenstate. Likewise, stabilization must not give the putative QCD axion a larger, misaligned potential. These are matching questions, not quantities that can be fixed by adjusting the line on a coupling plot.
Two-temperature thermal history and extra radiation
Temperature ratio, entropy, and the expansion rate
Let and denote the visible and twin photon/plasma temperatures at the same cosmic time, and define . Use for energy and for entropy degrees of freedom, with primed functions for the twin sector. After species decouple, these functions include their individual temperatures relative to the sector’s photon temperature. During radiation domination, neglecting the axion and nonrelativistic components in the background,
The twin-sector clock is therefore not given by a universe filled only with twin radiation at . For a colder, subdominant twin bath, the visible bath largely determines .
In the absence of intersector transfer and late entropy production, and are separately conserved, where is the cosmological scale factor, distinct from the axion field . With , this gives
The reheating values carry a subscript . Equation (35) is implicit when depends on . It shows why the temperature ratio at axion oscillation cannot in general be identified with the late photon temperature ratio. In the uniformly rescaled QCD limit the two confinement crossovers occur at different visible temperatures, which can temporarily change .
For instantaneous reheating by a common parent with branching fractions and into internally equilibrated radiation baths,
Here and are, respectively, the fractions of the parent energy deposited into visible and twin radiation, so . Since and similarly for the twin bath, the factor in Eq. (36) is , not its inverse. Thus a small can encode asymmetric reheating. The subsequent entropy history still has to be evolved; Eq. (36) is not generally a relation between branching fractions and today’s photon temperatures.
The twin contribution to the radiation budget
After visible electron–positron annihilation, define extra radiation by
Suppose the relativistic twin population consists of a massless twin photon and neutrino species, each with two populated helicity states. Put and . Their contribution follows directly from their thermal energy densities:
For three effectively massless twin neutrinos with , the coefficient is approximately 7.40. This is the late-time photon-temperature convention; it should not be confused with the familiar pre-annihilation coefficient near 6.14. The neutrino temperature, number of populated helicities, and whether the neutrinos remain relativistic must all be checked in the actual thermal history.
As an illustration, gives in the restricted radiation model above. This is an outcome of the stipulated particle content and temperatures, not a fitted value for the twin sector. Planck supplies a useful reference for radiation constraints [9]. For a specified extension with extra free-streaming radiation, ACT DR6 combined with other cosmological data reports [10]. Consequently is not automatically allowed, nor is it necessarily excluded by that number alone. The relevant likelihood must take account of which twin species remain relativistic and free-streaming, their temperatures, and any interactions between twin radiation and twin baryons [11, 12]. A viable radiation budget is necessary but not sufficient for a viable twin-baryon cosmology.
Can the universal axion reconnect the two baths?
Separate entropy conservation must be tested against the same interactions that define the axion. Schematically, energy transfer between the baths obeys
where is positive for net transfer to the twin bath. A non-negligible axion population requires its own kinetic equation and consistent source terms rather than counting its energy in both baths. The shared-axion portal can materially affect the entropy budget [3].
At temperatures well above confinement and below the effective-theory cutoff, dimensional estimates distinguish two possibilities. Thermal axion production in one bath scales parametrically as
with a corresponding twin rate. Off-shell intersector scattering through two anomalous vertices instead has the schematic scaling
These are power-counting estimates, not thermally averaged collision integrals; and contain phase space, numerical anomaly factors, and plasma effects. Sequential production and absorption of real axions can dominate over off-shell scattering. Consequently, checking only Eq. (42) is not sufficient to establish thermal isolation. The appropriate comparison is between energy-transfer rates and throughout the specified reheating history.
At low temperatures, visible and twin photons can also contribute to the axion lifetime. For the convention , the two-body width is
for massless final photons. Additional kinematically allowed channels must be included before assigning a lifetime or decay branching ratio.
Axion production in the two-temperature universe
Finite-temperature susceptibility and the onset of motion
For aligned minima and negligible explicit breaking, the instantaneous small-amplitude mass is
Under the full uniform rescaling used to obtain Eq. (5), the finite-temperature relation is
This requires the dimensionless QCD parameters and all relevant threshold ratios to match. It is not valid merely because . Lattice calculations of and the equation of state provide the visible-sector input [13]; the changed-mass benchmark requires its own twin susceptibility or a controlled approximation.
The homogeneous field obeys
In the small-angle regime, the last term is . The conventional onset estimate is
This is an operational estimate of a continuous transition, not an exact boundary. The numerical solution fixes the corresponding order-unity onset correction.
A colder twin sector can turn on its potential at a larger visible temperature and start axion oscillations earlier. This mechanism is already present in mirror-axion cosmology [14]. However, describes a zero-temperature enhancement, while Eq. (47) depends on both temperature-dependent susceptibilities. Replacing the ordinary mass everywhere by generally misses this distinction.
Comoving number and the relic abundance
Once and , the action variable of the nearly harmonic oscillator is adiabatic. In this regime,
when visible comoving entropy is conserved. The onset estimate gives
Here accounts for matching through the onset and at small angle. Their separation is only a useful parameterization, not two independent physical sources. A large-angle evolution should instead use the specified full potential.
For a stable, nonrelativistic population today,
The factor in this expression is important. An earlier onset reduces the comoving number, but a larger final mass increases the energy per particle. Those two effects need not yield a net suppression. If visible comoving entropy increases by a factor after number conservation is established, is divided by . Oscillations during the decaying matter-dominated era itself require solving its background; postulating a dilution factor alone is not sufficient.
Thermally produced axions, decays of heavy fields, and string–wall production are not included in Eq. (49). They must be added with their momentum distributions when present. A relativistic component contributes to radiation rather than to the cold fraction used in a haloscope interpretation.
A controlled high-temperature scaling limit
A useful analytic limit exposes the competition just described. Approximate the ordinary high-temperature mass by
where and is a matching scale. Thus the corresponding susceptibility scales as . Assume constant , constant visible and , negligible twin contribution to , small initial angle, and oscillation onset above the respective crossover in both the ordinary and twin theories. If the twin term dominates at onset, Eq. (45) implies
Let denote the ordinary-QCD onset at the same . Dividing the two onset equations gives
Since and , Eq. (49) yields . At equal initial angle and equal onset corrections, the relic ratio is therefore
These are analytical estimates derived under the assumptions above, not the output of a numerical parameter scan.
For , , and , Eq. (54) is about 2.7: the axion abundance is larger than in the ordinary-QCD comparison despite the colder twin bath. Net suppression in this limit would require roughly , or for these parameters, provided the high-temperature-onset assumption continues to hold. The example demonstrates why a small axion fraction cannot be inferred from alone.
There is also a saturation regime. If twin confinement completes before oscillations start and its susceptibility dominates, the mass is already approximately constant. Then
at fixed and . Once this regime is reached, further lowering does not continue the power-law suppression. The two-temperature susceptibility and entropy evolution determine where the transition occurs.
Lattice-informed two-temperature evolution
The analytic power-law limit does not include the QCD crossover, changes in the relativistic degrees of freedom, or the two baths’ contributions to the Friedmann equation. We therefore integrate the small-angle homogeneous field using numerical QCD input. For the visible sector, the data file accompanying this manuscript reproduces the 20 temperatures and values from Table S7 of Ref. [13]. The conversion to natural units is
We use shape-preserving cubic Hermite interpolation in and . Below 100 MeV we use the zero-temperature susceptibility , while above the last tabulated point we continue with . The latter is an explicitly specified extrapolation, not a lattice determination beyond 3 GeV. As in Eq. (45), the uniform-spectrum twin benchmark is
The distinct unchanged-quark-mass () benchmark cannot be calculated by this rescaling and is not included in the present scan.
For the thermal background we sample the effective energy and entropy degrees of freedom tabulated by Saikawa and Shirai [15], interpolating the supplied sparse subset in the data directory. Our particular uniformly rescaled Standard Model proxy also adopts
This stronger homologous spectrum assumption is independent of, and not implied by, alone. It cannot substitute for a derived twin threshold spectrum or a calculation of twin neutrino decoupling. In particular, the proxy uses uniformly shifted thresholds rather than an explicitly calculated thermal history for varying . The numerical results below refer to the specified homogeneous background; changing the twin electromagnetic coupling requires re-evaluating thermal decoupling and, especially, twin atomic recombination before making cosmological claims. We impose separate comoving-entropy conservation and define :
with and . At each we compute from both radiation baths through Eq. (34), and evolve and using Eq. (80). At sufficiently rapid oscillation we extract the adiabatic number-to-visible-entropy ratio and multiply by the zero-temperature axion mass. Throughout this calculation we set intersector energy transfer, thermal axions, anharmonicity, late entropy production and a matter-dominated reheating stage to zero.
For the reference choice , , , and , the solution begins overdamped and reaches the operational onset at with . At the onset ; after the modeled entropy releases, . The resulting abundance is
Switching the numerical evolution to adiabatic-number conservation at or 500, instead of 300, changes by less than 0.1%; this is a convergence check on the integration, not a systematic error bar. When the twin contribution to the axion potential is switched off while holding the same two-bath expansion background fixed, the QCD reference abundance is 0.00126, so the calculated ratio is 2.48. This need not equal the 2.72 result of the fixed-, power-law toy model in Sec. VIII E; the input susceptibility and entropy histories differ.
Figure 4. Evolution of the photon/plasma temperature ratio for three initial choices , under separate entropy conservation and the homologous equation of state in Eq. (58). The late photon ratios differ from their initial values because the particle thresholds in the two baths occur at different temperatures. The neutrino treatment is inherited from the rescaled Standard Model background and is not a twin-neutrino Boltzmann calculation.
The same separately conserved entropies give a late entropy ratio for this benchmark. If the surviving light twin radiation consists of a massless photon and three relativistic twin neutrinos with the standard internal neutrino/photon temperature ratio, Eq. (38) gives . This is a conditional particle-content estimate and is not a CMB fit. For , the required remainder of the dark-matter budget, expressed as the twin baryon asymmetry per twin entropy relative to the visible yield, is
Figure 5. Misalignment fraction for , three decay constants, and 12 initial temperature ratios . The background uses the tabulated QCD susceptibility and a sparse resampling of the Standard Model equation of state. All plotted points are conditional background abundances, not cosmologically allowed regions or likelihood exclusions.
The required asymmetry is an input, not a prediction of the axion anomaly.
a. Cosmological status. These calculations close the homogeneous axion and baryon density budget only after imposing a chosen baryon mass and asymmetry. They do not evolve twin weak freeze-out, nuclear chemistry, recombination, acoustic oscillations, ionization, velocity-dependent scattering, or the linear CMB/matter perturbations. In particular, a dark-matter sector that is almost entirely ionized or atomic baryons generally cannot be treated as pressureless collisionless matter merely because is small. A consistent full test will require a specified twin spectrum, a recombination and perturbation solver, and joint comparison with CMB anisotropies, Lyman-, and galactic self-interaction observables [11, 16].
| Quantity | Value | Interpretation |
| , , , | See text | Chosen initial conditions |
| , | 8.90 GeV, 0.3994 | Numerical onset criterion |
| , | 0.3711, 0.05111 | Separate-entropy background |
| , | 0.00312, 0.0260 | Homogeneous axion misalignment |
| 2.48 | Same radiation background, twin potential removed in reference | |
| 0.140 | Assumed late relativistic twin content | |
| 20.4 | Required budget for |
Table 2. Numerical benchmark with tabulated visible-QCD input. All quantities are conditional on the homogeneous, uniformly rescaled, decoupled-bath assumptions specified in the text. The dark-radiation value additionally assumes a light twin photon and three light neutrinos.
Independent analytic-limit check with a smooth toy susceptibility
As an independent numerical regression test of the high-temperature scaling without imposing an abrupt oscillation onset, we integrate the linearized homogeneous axion field equation in the number of expansion e-folds, , with :
For this controlled illustration we set , with constant , and assume radiation domination with negligible twin radiation in , no intersector transfer, and no late entropy injection. These simplifying assumptions are not a self-consistent precision thermal history. We use the interpolating susceptibilities
with and . The system is initialized while and evolved into the adiabatic regime; is extracted from the oscillation-averaged energy density and the zero-temperature mass is used in converting to . The functions in Eq. (64) are useful continuity tests, not lattice-QCD fits [13].
For , , , and , the ordinary-QCD reference gives while the twin-enhanced case gives . Their ratio is 2.7165, in agreement with the analytic high-temperature estimate 2.7166. The close agreement is a check of the stated limit, not a claim that the complete twin-sector cosmology has been solved.
Figure 7. A numerical test of the high-temperature abundance scaling. Points show the ratio obtained by integrating the small-angle field equation with Eq. (64); the curve is for and . All calculations use fixed and , constant and , and neglect twin radiation in the expansion rate. This is a controlled toy benchmark, not an observationally calibrated prediction.
If we illustrate the abundance accounting with and , the above toy benchmark contains an axion fraction . Assuming further that and the two late-time entropy degrees of freedom are equal, , the density budget would require
This is an algebraically consistent abundance assignment, not a calculation of twin baryogenesis, recombination, or structure formation.
Twin baryons as the dominant dark matter
The abundance coincidence and its entropy dependence
For an asymmetry-dominated baryon population, neglecting binding-energy corrections, the present densities obey
Here and denote the conserved baryon-number densities remaining after symmetric annihilation. For mixtures of nuclei, should be interpreted as the mean mass per unit baryon number, including any relevant binding and accompanying lepton contribution. A fivefold baryon mass explains a fivefold energy density only when the number-density ratio is approximately unity.
Define asymmetry yields using each sector’s own entropy:
When both comoving entropies are conserved,
Thus equal yields relative to each bath’s entropy do not imply equal physical number densities in a colder twin sector. For equal late entropy degrees of freedom and , . Requiring then demands . Equivalently, equal yields relative to a common visible entropy would imply equal number densities. The distinction is a convention-dependent statement about “equal asymmetries,” but its consequence for the density is physical.
Allowing an axion component and assuming no other dark matter gives
The ratio is the motivating rounded benchmark; a data comparison should use the fitted ratio in its chosen cosmological model [9]. Equation (69) links axion production, reheating, and baryogenesis. These cannot all be fixed independently if the dark sector is to saturate the observed matter density.
Removal of the symmetric component
A conserved asymmetry fixes the surviving baryon number but does not automatically remove baryon–antibaryon pairs. Below confinement a schematic annihilation equation is
with the same annihilation term for . Their difference is conserved in the absence of baryon-number violation. The condition for the asymmetry-dominated approximation is that the residual symmetric component is small compared with that difference. Strong annihilation is a motivation for this possibility, not a substitute for computing it in the specified hadronic spectrum. Quark-level evolution above confinement and hadronic annihilation below it should be matched consistently.
Stable baryons, nuclei, and atoms
For light constituent quarks, the baryon mass is approximately set by confinement, but the identity of the lightest baryon depends on isospin and electromagnetic effects. A schematic separation is
with dimensionless coefficients determined by the hadronic theory. The competition of these contributions is established in visible QCD plus electromagnetism [17]. With uniform scaling and fixed dimensionless couplings, both terms scale together. With unchanged light-quark masses but enhanced confinement, their relative importance changes; the stable twin baryon cannot simply be assumed to remain the proton.
Ignoring neutrino masses, free-neutron beta decay requires , while free-proton beta-plus decay requires . Bound-state electron capture introduces a separate condition involving the atomic binding energy. Nuclear binding can change the surviving composition further. A neutron-dominated sector, an ionized proton–electron plasma, and neutral twin atoms have very different scattering and cooling properties even at the same .
For a hydrogen-like twin atom,
If the twin proton and electron are each five times heavier, and , these scales become and . The value of must be obtained from ultraviolet and threshold matching; is used only for the following numerical size estimate. Even if the twin unification scale differs from the visible one, the hierarchy alone does not uniquely determine . Recombination, residual ionization, and kinetic decoupling then depend on , the radiation temperature, and the twin baryon-to-photon ratio. The binding energy alone is not the recombination temperature. Atomic dark matter can exhibit dark acoustic oscillations and velocity-dependent self-interactions [11, 16]. Nuclear abundances also depend on twin weak freeze-out and nuclear reaction rates, not just rescaled baryon masses.
For purely hadronic scattering, uniform scaling suggests a dimensional estimate and hence at fixed dimensionless kinematics. This estimate is not a prediction near nuclear resonances and says nothing about potentially much larger atomic cross sections. Astrophysical self-interaction tests instead constrain a velocity- and composition-dependent transport cross section per unit mass; what is relevant for dwarf galaxies may differ from what is relevant for galaxy clusters. A neutron-dominated twin sector could be governed primarily by short-range hadronic scattering, whereas charged or neutral twin atoms can exhibit collisional, radiative, and acoustic dynamics. Neither the scaling estimate nor the axion abundance calculation supplies a predicted momentum-transfer cross section. Testing dominant twin-baryon dark matter requires the cross section at the relevant halo velocities, plus its ionized, atomic, nuclear, and dissipative fractions [11, 16, 18, 19].
An elementary atomic size estimate emphasizes the required next step. For the illustrative choice , , and , the hydrogen-like binding energy is approximately and . At general , these become and . The ratio of the geometric area to the assumed atomic mass is of order . This number is only a size-scale diagnostic: the momentum-transfer cross section is sensitive to collision velocity, quantum molecular potentials, ionization, and the actual atomic composition. In particular it must not be interpreted as a predicted self-interaction cross section or a direct exclusion. With a twin-baryon fraction, computing twin recombination and dark acoustic damping is necessary before declaring the example compatible with cosmological data.
For an unbroken twin electromagnetic group, a charged lightest baryon may bind with twin electrons. Such a dark component need not behave as pressureless collisionless matter, even if the homogeneous abundance budget closes. Dark acoustic oscillations and dissipative cooling are constrained by CMB, galaxy-formation, and small-scale-structure measurements [16, 18, 19]. A model-level constraint requires the stable twin hadron spectrum, recombination and residual ionization histories, velocity-dependent transport cross sections, and the resulting linear matter transfer function. Constraints inferred in simplified atomic-dark-matter models cannot be adopted as automatic exclusions of the present F-theory-motivated benchmark without matching the particle content and interaction history.
Inflationary initial conditions and topological defects
If the Peccei–Quinn symmetry is broken before inflation and never restored, our observable patch may have a single initial angle. For a light, canonically normalized axion during inflation and an unchanged decay constant,
In the harmonic misalignment regime, with the other dark matter adiabatic and uncorrelated with the axion fluctuation,
For a chosen upper limit this implies
Planck 2018 supplies reference likelihoods for specific isocurvature models [20]. These linearized expressions also assume a classical displacement larger than its fluctuations, ; the fluctuation-dominated regime requires a statistical treatment. The formula should not be used unchanged for a heavy inflationary axion, a different inflationary kinetic normalization, a large-angle abundance, or correlated twin-baryon isocurvature. In particular, selecting a baryogenesis mechanism can introduce an additional fluctuation rather than merely fixing the background abundance.
If the symmetry is instead broken after inflation, averaging over patches is not a complete abundance calculation: strings and domain walls can radiate axions. To identify the wall number, restore the fundamental periodic variable with period and integer color anomalies and . For aligned potentials with no additional identifications, the exact common periodicity gives
In particular, leaves , not . The normalization used in the color coupling hides this integer and must not be taken as proof that . Additional gauge identifications or explicit breaking can change the counting. In the equal-anomaly case, two confinement contributions do not by themselves remove a stable wall problem; a proposed bias must also satisfy the strong-CP condition in Eq. (30).
From benchmark relations to a testable model
The preceding equations define a sequence of calculations rather than a single exclusion plot. First, the ultraviolet matching fixes the light axion direction, its anomalies, and the twin spectrum. The matched spectrum fixes the susceptibilities and the identity of stable matter. The reheating and entropy history then fixes and the background expansion used in the axion and baryon calculations. Finally, the resulting matter fractions and interactions determine which cosmological and experimental likelihoods apply.
A useful first comparison keeps the two susceptibility benchmarks but varies , the initial angle, and the temperature ratio. The case can use a consistently rescaled visible-QCD thermal input; the case cannot use that rescaling without an additional approximation. A small observed axion signal could be explained by a small intrinsic coupling, a small local axion density, or both, but the first is a particle-physics output and the second is a population and halo output. They should be displayed separately.
Table III distinguishes earlier analytic or toy consistency checks from the single conditional numerical background benchmark in Table II. Neither table is a fitted, fully viable cosmology. In particular, at oscillation and at late times need not agree.
The minimal population requirement, assuming no other dark matter, is
It must be imposed together with radiation, isocurvature, stability, and structure-formation tests, not before calculating them. The local fraction appearing in Eq. (24) can then be related to only under a specified halo model. Dissipative twin matter makes that last step especially nontrivial.
| Input | Relation tested | Consequence |
| , | Eq. (18) | ; fixed intrinsic slope |
| Three relativistic twin neutrinos; | Eq. (38), with standard neutrino temperature ratio | |
| Equal late entropy degrees of freedom; | Eq. (68), with | |
| , ; high-temperature onset | Eq. (54), equal and | |
| , ; lattice-informed, separately adiabatic baths | Eqs. (61)–(62) | ; (conditional) |
| Small | Eq. (25) | Almost no suppression of the mismatch |
Table 3. Conditional consistency checks for the uniformly rescaled benchmark. No row asserts that all constraints have been met.
Conclusion
A universal axion coupled to visible and twin QCD can be heavier at fixed decay constant, provided the two physical strong-CP minima are aligned and the canonically normalized axion anomalies are correctly matched. The physical mass follows from the sum of the two susceptibilities. In the uniformly rescaled benchmark , and assuming gives the intrinsic visible-photon relation . It agrees with the marked benchmark in the supplied axion-coupling plot. The visible-photon coupling and a local dark-matter-signal rescaling must remain separate.
Beyond this particle-physics locus, we have advanced the cosmology from an idealized analytic estimate to a reproducible numerical background calculation. We interpolate published finite-temperature visible-QCD topological-susceptibility values, combine them with a sampled Standard Model equation of state, track separate visible and twin entropy release, and include both radiation baths in . Under a uniform full-spectrum rescaling and no intersector energy transfer, a benchmark with , and produces , about 2.6% of an illustrative . The evolved late photon ratio is , in contrast to the initial 0.40. A conditionally light twin photon and three light twin neutrinos would then yield . To fill the remaining abundance with five-times-heavier twin baryons would require when the asymmetries are normalized to each sector’s entropy.
These numbers are not a self-contained viable mixed-dark-matter model. The entropy evolution uses a sparse, interpolated Standard Model equation of state and a homologous twin proxy; the susceptibility is extrapolated above 3 GeV; and we have not included axion transfer between baths, thermal axions, nonlinear misalignment, twin baryogenesis, recombination, or matter perturbations. In particular, a dark sector dominated by stable charged or atomic twin baryons requires detailed checks of dark acoustic oscillations, dissipative dynamics, and velocity-dependent self-interactions. Preliminary equality of visible and twin ultraviolet Yukawa couplings, when established, would still require Higgs-VEV and threshold matching; likewise the ratio must be determined before the atomic benchmark becomes a prediction. These requirements are in addition to the strong-CP and axion-normalization consistency tests. The accompanying tabulated inputs and integration code make the conditional background calculation reproducible and provide a starting point for those necessary model-dependent tests.
Acknowledgments
We thank Gordon Kane and Bobby Acharya for useful conversations. This research was supported in part by U.S. Department of Energy grant DE-FG02-85ER40237, DE-GR102719 and the Simons Center.
Numerical evolution: a reproducible prescription
This appendix specifies the evolution equations used in both the lattice-informed background calculation in Sec. VIII D and the independent smooth-toy regression test in Sec. VIII E. The lattice-informed calculation uses a sparse, interpolated Standard Model equation of state and the stated homologous twin-sector assumptions; neither calculation includes linear cosmological perturbations or a joint observational fit. The natural independent variable is the expansion e-fold . In a separately adiabatic visible bath,
and the twin temperature follows from its own entropy equation or Eq. (35). If intersector transfer or a decaying matter component is important, the corresponding energy equations replace this adiabatic relation.
Write . The axion equation becomes
For radiation with constant degrees of freedom, , providing a useful check of the damping term. Initial conditions and can be imposed sufficiently early that , provided the field was initially overdamped. One must not impose this condition after the onset of motion merely because the visible sector has not yet confined.
For testing an integrator, a smooth toy susceptibility is
It interpolates between the assumed limiting powers, but is not a lattice fit. Choosing it and a cosine potential is a numerical benchmark that should be labeled as such. The changed-quark-mass case needs a separate thermal input. For publication-quality abundances, one should propagate uncertainties from the susceptibility, equation of state, and nonharmonic potential rather than varying only .
After many adiabatic oscillations, extract
The average must extend over several oscillations in a regime where the potential is effectively harmonic and has stabilized. Convergence should be checked against the integration tolerance, initial time, averaging interval, and the time of the switch to adiabatic number conservation.
a. Data and reproducibility. The complete project contains the original 20-row susceptibility values and quoted one-sigma logarithmic uncertainties in data/borsanyi2016_table_S7.csv, and a 29-row sparse sample of the published Saikawa–Shirai degrees of freedom in data/eos2018_sampled.csv. These are finite-temperature visible-sector inputs, not derived twin datasets. The file code/lattice_two_temperature_scan.py implements the monotone log-log interpolation, separate-entropy inversion, Hubble rate, harmonic field evolution, adiabatic number conversion, and scans underlying Figure 3–Figure 6. The illustrative numerical conversion uses and . Both the external table interpolation and all additional cosmological assumptions must be varied for a credible precision uncertainty estimate. Overleaf reads the precomputed PDF figures, so compilation does not require Python. A README documents how to reproduce the results locally.
Figure 3. Topological susceptibility used in the numerical evolution. Points with quoted errors are the visible-QCD input of Ref. [13]; the smooth line interpolates the tabulated data, with specified low- and high-temperature continuations. The dotted line marks the end of the published entries at 3 GeV. The model for is the conditional uniform rescaling in Eq. (57), not a separate twin lattice result.
Figure 6. Illustrative extra-radiation budget from the same evolving temperature ratio, assuming one massless twin photon and three light twin neutrinos with the standard neutrino-to-photon temperature ratio. The horizontal guide is not an observational likelihood. Interacting radiation and atomic dark matter require a joint perturbation calculation before model exclusion.
Three limiting checks are particularly useful. Removing the twin susceptibility must recover the single-sector evolution. Identical sectors at equal temperature give and require both baths in . Finally, a twin-dominated, power-law onset with negligible twin radiation must approach Eq. (54); a sufficiently early twin crossover must instead approach Eq. (55). These tests separate errors in the potential, the background, and the final mass conversion.
Finite-angle alignment in a cosine approximation
The small-mismatch relation in the main text does not assume cosine potentials. For illustration only, take
Define the complex quantity . Then , so the stable minimum and curvature are
Expanding at small reproduces . This example also illustrates why the aligned susceptibility sum cannot be applied unchanged at arbitrary relative phase. For large , the visible strong-CP angle is generally large even when the axion is heavy. The full chiral or nonperturbative potential, rather than Eq. (84), is required for quantitative large-angle claims in QCD.
References
References
- [1]C. H. Clemens and S. Raby, Relative scales of the GUT and twin sectors in an F-theory model, JHEP 2020 (04), 004, see revised version v2, arXiv:2001.10047 [hep-th].DOI
- [2]Z. Berezhiani, L. Gianfagna, and M. Giannotti, Strong CP problem and mirror world: the Weinberg–Wilczek axion revisited, Phys. Lett. B 500, 286 (2001), arXiv:hep-ph/0009290.DOI
- [3]D. I. Dunsky, L. J. Hall, and K. Harigaya, A heavy QCD axion and the mirror world, JHEP 2024 (02), 212, arXiv:2302.04274 [hep-ph].arxiv.org/abs/2302.04274
- [4]H. Clemens and S. Raby, Heterotic-F-theory duality with Wilson line symmetry-breaking, JHEP 2019 (12), 016, arXiv:1908.01913 [hep-th].DOI
- [5]G. Grilli di Cortona, E. Hardy, J. Pardo Vega, and G. Villadoro, The QCD axion, precisely, JHEP 2016 (01), 034, arXiv:1511.02867 [hep-ph].DOI
- [6]C. A. J. O’Hare, AxionLimits: Fips workshop 2022 axion–photon coupling compilation, https://github.com/cajohare/AxionLimits/blob/master/docs/fips.md (2022), fIPS Workshop 2022 figure. AxionLimits repository archived at DOI: 10.5281/zenodo.3932430.
- [7]C. Abel et al., Measurement of the permanent electric dipole moment of the neutron, Phys. Rev. Lett. 124, 081803 (2020), arXiv:2001.11966 [hep-ex].DOI
- [8]A. Hook, S. Kumar, Z. Liu, and R. Sundrum, The high quality QCD axion and the LHC, Phys. Rev. Lett. 124, 221801 (2020), arXiv:1911.12364 [hep-ph].DOI
- [9]N. Aghanim et al. (Planck), Planck 2018 results. VI. cosmological parameters, Astron. Astrophys. 641, A6 (2020), arXiv:1807.06209 [astro-ph.CO].DOI
- [10]E. Calabrese et al., The atacama cosmology telescope: Dr6 constraints on extended cosmological models, Journal of Cosmology and Astroparticle Physics 2025 (11), 063, arXiv:2503.14454.arxiv.org/abs/2503.14454
- [11]S. Bansal, J. H. Kim, C. Kolda, M. Low, and Y. Tsai, Mirror twin higgs cosmology: Constraints and a possible resolution to the H₀ and S₈ tensions, JHEP 2022 (05), 050, arXiv:2110.04317 [hep-ph].DOI
- [12]C. Trendafilova, F. Ge, A. R. Khalife, and C. L. Reichardt, Cosmological constraints on interacting dark matter with spt-3g d1 (2026), arXiv:2610.37604 [astro-ph.CO].arxiv.org/abs/2610.37604
- [13]S. Borsanyi et al., Calculation of the axion mass based on high-temperature lattice quantum chromodynamics, Nature 539, 69 (2016), arXiv:1606.07494 [hep-lat].
- [14]M. Giannotti, Mirror world and axion: Relaxing cosmological bounds, Int. J. Mod. Phys. A 20, 2454 (2005), arXiv:astro-ph/0504636.DOI
- [15]K. Saikawa and S. Shirai, Primordial gravitational waves, precisely: the role of thermodynamics in the standard model, JCAP 2018 (05), 035, a sparse numerical subset of the public EOS2018 table is supplied with this work, arXiv:1803.01038 [hep-ph].DOI
- [16]F.-Y. Cyr-Racine and K. Sigurdson, The cosmology of atomic dark matter, Phys. Rev. D 87, 103515 (2013), arXiv:1209.5752 [astro-ph.CO].
- [17]S. Borsanyi et al., Ab initio calculation of the neutron-proton mass difference, Science 347, 1452 (2015), arXiv:1406.4088 [hep-lat].DOI
- [18]J. Barron, R. Essig, M. H. McDuffie, J. Pérez-Ríos, and G. Suczewski, Pushing the limits of atomic dark matter: First-principles recombination rates and cosmological constraints, Journal of Cosmology and Astroparticle Physics 2026 (07), 090, arXiv:2602.10197.arxiv.org/abs/2602.10197
- [19]Z. Yuan, C. Gemmell, K. K. Rogers, J. Barron, S. Roy, D. Curtin, and N. Murray, Strongest constraints on dark acoustic oscillations from the lyman-alpha forest (2026), arXiv:2606.28482 [astro-ph.CO].arxiv.org/abs/2606.28482
- [20]Y. Akrami et al. (Planck), Planck 2018 results. X. constraints on inflation, Astron. Astrophys. 641, A10 (2020), arXiv:1807.06211 [astro-ph.CO].DOI





