WorksPhysics and Astronomy
Alleviating CMB-BAO tension with a large cosmic birefringence angle during reionization induced by energy injection to axion-like particles
Abstract
The recent measurement of baryon acoustic oscillations by the Dark Energy Spectroscopic Instrument reveals a mild tension with observations of the cosmic microwave background (CMB) within the standard CDM cosmological model. Another interesting hint beyond standard cosmology arises from recent analyses of Planck polarization data which has been interpreted as cosmic birefringence. One possible solution to explain both observations is to introduce cosmic birefringence induced by axion-like particles (ALPs) that has a much larger rotation angle during the reionization epoch than that during the recombination: the large rotation suppresses the reionization signal, allowing an elevated value of the CMB optical depth that alleviates the CMB-BAO tension. In this paper, we explore a scenario where the energy transfers from matter to ALPs before the reionization. Using Planck low- , , and data, we show that, for energy injection at , there exist parameter regions where is compatible with the data.
Alleviating CMB-BAO tension with a large cosmic birefringence angle during reionization induced by energy injection to a
Abstract
Introduction
Recent measurements of the cosmic microwave background (CMB) and baryon acoustic oscillations (BAO) exhibit a mild tension within spatially flat cold dark matter (CDM) cosmology [1, 2]. In particular, cosmological analyses prefer a sum of neutrino masses below the minimum allowed by oscillation experiments [3]. When the effective neutrino mass is extended to negative values, the minimal mass is disfavored at approximately , although the precise significance depends on the adopted Planck likelihood [3, 4, 5, 6, 7, 8, 9]. One possible explanation is a shift in the CMB reionization optical depth, [10, 11, 12, 13]. If the Planck low- polarization data are excluded, a larger value, such as , substantially reduces the neutrino-mass tension and weakens the preference for dynamical dark energy, as well as the lensing and curvature anomalies [10, 12]. However, this value is inconsistent with the Planck low- measurement at [12].
Another intriguing hint of beyond the CDM model is cosmic birefringence, a rotation of the polarization plane of light during propagation (see Refs. [14, 15] for reviews). Minami & Komatsu [16] first reanalyzed Planck polarization data and reported a hint for isotropic birefringence at statistical significance. See also Refs. [17, 18, 19] for other analyses of the birefringence angle with Planck data. The analysis is improved by adding Wilkinson Microwave Anisotropy Probe polarization data, finding a rotation angle at a preference [20]. The latest constraint is obtained by a joint analysis of the Planck and ACT data that yields a birefringence angle of , which is non-zero with a statistical significance of [21]. Because cosmic birefringence violates parity, it provides a sensitive probe of new physics beyond both CDM and the Standard Model [22]. Cosmic birefringence can be generated by a pseudoscalar field , such as an axion-like particle (ALP), coupled to electromagnetism through a Chern–Simons interaction,
where is the coupling constant, is the electromagnetic field-strength tensor, and is its dual. This mechanism has been studied in models of axion dark energy [23, 24, 25, 26, 27, 28, 29, 30, 31, 32], early dark energy [27, 33, 34, 35], and axion dark matter [36, 37, 38, 39, 40, 41]. Other proposed sources include topological defects [42, 43, 44, 45, 46] and quantum-gravity effects [47, 48, 49].
One possible scenario to explain the above two anomalies is to introduce a large rotation of cosmic birefringence during the reionization epoch [50]. If cosmic birefringence is induced by a single ALP, the ambiguity of the phase of the observed cosmic birefringence [51] could lead to a large rotation angle during the reionization while keeping the recombination birefringence signal unchanged. Alternatively, we can also consider a scenario originally proposed in Graham et al. (2025) [52] where part of the energy of matter transfers to ALPs.
In this paper, we develop the scenario of Ref. [52] to quantitatively constrain that scenario. We use Planck data to search for parameter regions that reproduce the observed data and to show whether this model explains the recent CMB-BAO tension.
This paper is organized as follows. In Sec. II, we review cosmic birefringence induced by ALPs and introduce the energy injection model. In Sec. III, we present constraints on cosmic-birefringence scenarios using data from Planck. Finally, Sec. IV is devoted to a summary of our results and conclusions.
Cosmic Birefringence and Energy Injection Model
CMB angular power spectra
Let us first introduce the CMB angular power spectra induced by cosmic birefringence from ALPs [24, 33, 36, 53, 54, 55, 56, 57, 58, 59]. When cosmic birefringence is sourced by an ALP field, the rotation angle becomes time-dependent. The net rotation angle for a photon observed today and emitted at conformal time is given by [60, 61, 62]
where is the present conformal time and is the initial value of . Throughout this paper, the overall amplitude is fixed so that the present-day rotation angle of the CMB photons emitted at recombination is [20]. Since scales linearly with , varying within its uncertainty would only shift the allowed values of the energy injection, without changing our results qualitatively. To compute the impact of this time-dependent rotation on the CMB polarization, we solve the Boltzmann equation for the polarized photon distribution [24, 36, 53, 54, 63]:
where are the Fourier modes of the linear polarization and are functions of conformal time , wavevector magnitude , and angle cosine between line-of-sight direction and wavevector . The other symbols are: (scale factor), (electron number density), (Thomson scattering cross-section), (polarization source term) [64], and (spin-two spherical harmonics). Primes denote derivatives with respect to conformal time.
To derive the angular power spectra, is expanded as [64]:
where the formal solution for the - and -mode sources is given as a line-of-sight integral [24]:
Here, is the spherical Bessel function, , and . Note that the -mode generated at time is multiplied by . Thus, a large variation of during reionization leads to cancellations of signals among different epochs, resulting in a suppression of the reionization bump signal in the power spectrum and mimicking a lower [50]. The CMB angular power spectrum is then calculated as
where and is the dimensionless primordial scalar curvature power spectrum. We systematically compute all CMB polarization auto- and cross-power spectra in Eq. (6) by appropriately modifying the polarization transfer function, , within the CLASS code [65, 66].
Energy injection from CDM to ALPs
We now incorporate a large rotation angle realized by energy injection to ALPs from CDM. We consider the following phenomenological model approach often used for CDM / dark energy interactions [52, 67, 68, 69, 70, 71]; we modify the continuity equation that transfers part of the DM energy density to ALPs:
where is the energy density of CDM, is the energy density of ALPs, is the pressure of ALPs, , is the rate of energy transfer.
We rewrite the continuity equation in terms of the pseudo-scalar field as
where is the partial derivative of the potential with respect to . In the following, we assume a cosine potential where is the decay constant. We assume that ALP is a spectator field and that the energy density of ALPs is negligible in the background evolution. Thus, even a small amount of energy transfer from matter can significantly enhance the energy density of ALPs, and the amplitude of . Note that in the absence of the energy transfer, we obtain the usual Klein-Gordon equation:
We assume the following energy transfer rate (e.g., [69, 70])
Instead of using a specific functional form for , we constrain in a model independent way; we assume that is a constant during a specific short period starting at the redshift of and ending at . Substituting the above equation into Eq. (9) yields
when . In our calculation, is always satisfied during the energy injection period.
To realize a large rotation angle, the ALP field is initially located near the bottom of the potential. We specifically assume the initial condition that satisfies
This choice, together with , implies , i.e., an anomaly coefficient for . Such large values have been considered in previous models of cosmic birefringence [28]. In this case, the initial ALP energy density is , where the last equality holds at the bottom of the potential well. Considering that , and denoting , the evolution equation is rewritten as
where is the present energy density of DM and .
It is convenient to define
For fixed , Eq. (15) depends on and only through : the evolution of , the birefringence angle, and the CMB spectra are all functions of alone. (and also ) is instead a free parameter that the CMB spectra cannot determine. only rescales the physical injection rate and the initial ALP energy, without affecting .
Figure 1 shows an example of the evolution of the ALP field. We choose that leads to a large birefringence angle. In the following, we focus on the rescaled energy injection rate instead of .
Figure 1. Time evolution of the cosmic birefringence angle for and an energy-injection of (blue curve), with injection occurring in the redshift interval (red band). The blue band marks the reionization epoch as reference. The rotation angle is null before the injection takes place; once the injection begins, is kicked out of its potential minimum and starts to oscillate, producing the large swing in . The oscillations are subsequently damped by Hubble friction, and asymptotically approaches the present-day birefringence angle . The phase model (black curve) used in [50] is shown for comparison.
We note that has an upper bound to realize a large birefringence angle. To see this, let us consider the case
if too much energy is injected into ALPs by increasing , and crosses over the peak of the cosine potential, , . In this situation, the ALP no longer oscillates around and instead settles into oscillations around a different minimum, . Then, the resulting cosmic birefringence angle for the CMB photons emitted at recombination becomes approximately . Since the field value varies within , we cannot realize a large rotation angle. Thus, there is a critical value of the energy transfer efficiency, , beyond which this kick-out occurs. Because the driving term is time-dependent and the injection acts only over a finite redshift interval, cannot be obtained in closed form and is instead determined numerically for each ALP mass and each injection window. Figure 2 shows as a function of the injection redshift window for . For the three representative masses shown, is non-monotonic, exhibiting a minimum around –. This minimum marks the redshift bin where energy injection couples most efficiently to the field, corresponding to the epoch when the field starts to be released from Hubble friction, , shortly before the onset of coherent oscillations at : injecting energy at this epoch requires the smallest to kick the field over the potential barrier, while injection well before (field still frozen by Hubble friction) or well after (field already oscillating and decaying) is less efficient and requires larger . Consistently, the minimum shifts to lower redshift for the lighter ALP, tracking the mass dependence of the epoch .
Figure 2. Critical energy-injection efficiency as a function of the injection redshift window, for . The non-monotonic trend, with a minimum around – that shifts to lower redshift for lighter masses, reflects the efficient coupling between the injection and the epoch , when the field starts to be released from Hubble friction.
Constraints on and
As discussed in Sec. II B, the CMB spectra depend on the energy injection only through , where is a free parameter. Consequently, and are degenerate and cannot be independently constrained by the CMB spectra alone. However, the physical energy exchanged between CDM and ALPs depends on both and . Since the energy densities of CDM and ALPs are constrained by CMB and large-scale structure observations, these constraints impose additional restrictions on the allowed values of and . We consider the following two conditions:
CDM depletion: The energy transferred to the ALP is extracted from CDM. Since the CDM energy density is tightly constrained to be close to that in the best-fit CDM cosmology, the fraction of CDM depleted must be smaller than the observational uncertainty in the CDM energy density. As the CDM energy density is proportional to the physical density parameter , we conservatively require
Here, the uncertainty is taken from Ref. [72].
Spectator condition: The ALP energy density must remain negligible compared to that of CDM; otherwise, it could significantly modify the background expansion history, potentially conflicting with large-scale structure observations. Following Ref. [27], we impose , where denotes the present-day ALP density parameter. This condition can be more conservatively expressed as
where we have used [72].
These two conditions do not alter the predicted CMB spectra at fixed , but they restrict the values of and for which a given is physically viable.
Satisfying these conditions also justifies the approximations adopted in our numerical calculations: the background CDM energy density is fixed to its best-fit CDM value, and the ALP is treated as a spectator field whose contribution to the background expansion is negligible.
In the following, we derive constraints on and from these two conditions.
CDM depletion. Integrating Eq. (7) with of Eq. (11) gives for and otherwise. The ratio to the CDM density after the injection is therefore
and Eq. (18) becomes a limit on only,
which depends only on the injection window and is independent of and .
Spectator condition. The ALP energy density has two contributions: the initial energy density , frozen in the potential until the field starts to oscillate, and the energy injected by the source term . The ratio grows while the field is frozen, since dilutes as , once the field starts to oscillate, also dilutes as and the ratio settles to a plateau. We therefore evaluate Eq. (19) today, , where the plateau value is reached and which is representative of the maximum of the ratio over the whole evolution. Using the results in Appendix A, we find
This ratio is not simply proportional to at fixed : Eq. (15) depends on and only through , so that is a function of . Defining the transfer function
which depends on and only through their ratio , Eq. (22) reads
The function includes non-linear effects of the injection, such as the kinetic phase that follows a strong kick and the delay in the onset of oscillations, which analytical approximations cannot reproduce (Appendix A). We therefore compute it numerically, solving Eq. (15) once for each on a grid , and then map it onto the plane using . The allowed region is the intersection of:
the kick-out limit, ;
the spectator limit, ;
the depletion limit .
Figure 3 shows the result for . Since the CMB spectra depend only on , the choice of only moves the physical value of . For the whole range satisfies both conditions for the three masses considered, and all the results of Sec. III are therefore presented in terms of , up to .
Parameter Constraints
Having established the background physics of the injection term, here we want to present a systematic study on how the energy injection and the large cosmic birefringence angle generated from it is actually a model able to suppress the reionization bump in the case of a high value of the optical depth . We have performed this analysis for axion masses , in steps of , and for all the injection redshift windows shown in Figure 2, finding qualitatively similar results. In the following, we present the results for three representative mass values, , injecting energy only within the redshift window .
Data and methodology
In this work we have considered the Planck NPIPE low- lollipop data [73, 74, 75], considering the , , and spectra and the corresponding NPIPE low- covariance matrix. To compute the theoretical angular power spectra including the energy-injection mechanism described in Sec. II B, we have developed birefINJ2, a modified version of the Boltzmann code CLASS [65, 66]. birefINJ implements the birefringence-induced source term of Eq. (5) in the computation of the angular power spectra, Eq. (6), following the approach of birefCLASS [56], together with the axion-field dynamics under energy injection described in Eq. (12).
Once the field dynamics and the resulting angular power spectra are computed, we evaluate the results through a statistics performed in a -only case or in a joint case involving , , and spectra. We adopt as fiducial cosmology the best-fit CDM model of Ref. [12], obtained without low- polarization data. When varying , we keep fixed to its fiducial value, where is the scalar amplitude, so that the high- CMB spectra are unchanged and only the low- reionization signal is affected. We do not include the spectrum in the analysis. Cosmic birefringence does not affect the temperature anisotropies, so that is independent of ; moreover, with fixed, increasing from to changes the low- spectrum by less than , well below the cosmic variance at these multipoles. We also do not include the low- spectrum, whose constraining power on is limited by its large uncertainties.
Denoting by the set of spectra entering a given analysis — either for the -only case or for the joint analysis — we define the data vector
and the concatenated vector , obtained by stacking each spectrum in . Here is the angular power spectrum in the standard convention. The is then
where is the NPIPE low- covariance matrix restricted to the spectra in : for the -only case reduces to the single block, while for the joint analysis is the full block covariance correlating , , and , with the number of multipoles. In this analysis , having restricted the analysis to the multipole range . The corresponding likelihood is . We evaluate using two independent methods: a Markov Chain Monte Carlo (MCMC) sampler and a direct evaluation on a two-dimensional grid at fixed ALP mass. The two approaches have been cross-validated against each other, yielding consistent results; in the following we present the results obtained from the grid-based likelihood evaluation.
Power spectra
As recently proposed in the literature [10, 11, 12, 13], raising the optical depth to can help alleviate the CMB-BAO tension. However, at this value of and in the absence of injection, the resulting reionization bump clearly overshoots the Planck data (blue curve in Figure 4). Introducing a large cosmic birefringence angle sourced by the energy-injection mechanism of Eq. (12) suppresses the bump back down to a level consistent with the data (green curve in Figure 4), closely tracking the standard, low- case (orange curve in Figure 4). The same qualitative behavior, but with different values of and , is observed for the other masses considered in this work and is therefore not shown separately.
Figure 4. (upper left), (upper right), and (center) angular power spectra for , compared to the Planck NPIPE low- data (black points). Orange: with no injection. Blue: with no injection where the resulting bump overshoots the observed reionization signal. Green: but with an injection of , below the kick-out limit for this mass, ; in this case, the injection suppresses the bump back to a level consistent with the data. The same injection also imprints a distinctive dip in and a corresponding bump in around –, the direct signature of the late-time rotation sourced by the axion dynamics.
The and panels of Figure 4 show that the injection produces a genuine, visible change in these two channels as well. However, given the current NPIPE low- uncertainties, all three curves remain statistically consistent with the data in both channels: no single spectrum, taken individually, visually excludes either scenario. This suggests that and can carry complementary information to that of alone, which tightly constrains the shape of the reionization bump but leaves highly degenerate. As shown quantitatively in Sec. III C, combining , , and reshapes the allowed region into a set of distinct sub-regions rather than a single, simply-connected area; a residual degeneracy nonetheless persists, as we discuss in Sec. III C.
Altogether, Figure 4 shows that there exist values of for which a large optical depth remains compatible with the observed , , and spectra, thereby strongly motivating the quantitative parameter analysis presented in the remainder of this section. We note that all these results closely resemble those obtained by Ref. [50] within the -phase model.
Contour plots
Having shown in Sec. III B that a large optical depth can, for suitable values of , remain qualitatively compatible with the observed , , and spectra, we now turn to a quantitative likelihood analysis of the full parameter space. We compare the results with the -only and joint cases under both a uniform and a Gaussian prior on .
Top block of Figure 5 shows the and confidence contours in the plane obtained assuming a uniform prior on . In the -only case (top row), all three masses display a common qualitative structure: a broad, continuous region around –, together with one or more narrow features extending toward higher at specific, discrete values of . This multimodal structure is not a numerical artifact — it reflects the oscillatory nature of the ALP dynamics under a finite-window energy injection, whereby the accumulated rotation phase can reach similar values for several distinct values of , producing comparably good fits to the reionization bump at those discrete points.
Figure 5. (dark) and (light) confidence contours in the plane for the EE-only (top row of each block) and joint EE+BB+EB (bottom row of each block) analyses, for the three ALP masses (columns). Top block: uniform prior . Bottom block: Gaussian prior , motivated by high- CMB+DESI-BAO constraints [12]. Note that the range differs across columns, as depends on the ALP mass (cf. Figure 2).
Combining , , and (bottom row) does not simply narrow this structure into a single, better-constrained region. Instead, for all three masses the allowed parameter space is reorganized into a set of distinct, sometimes disconnected sub-regions, some of which persist at values of higher than the dominant low- feature of the -only case. This is consistent with the fact that , , and are not independent probes: all three spectra are sourced by the same underlying rotation history where , so the additional information carried by and reshapes rather than simply shrinks the -only degeneracy.
Bottom block of Figure 5 shows the same analysis with a Gaussian prior on centered at , motivated by external CMB+BAO constraints. As expected, for all three masses, this prior consistently selects and reinforces the higher- sub-regions that were only marginal features under a uniform prior, while suppressing the low- region that dominates the -only, uniform-prior case. In all three cases, the surviving sub-regions cluster around –, directly tracking the external prior value . This result is robust across the three representative masses considered in this work: there consistently exist regions of the parameter space, compatible with the joint likelihood, in which a large optical depth is achieved through the energy-injection mechanism.
Because the allowed parameter space is multimodal rather than unimodal, a single symmetric credible interval does not adequately summarize the joint constraint; we therefore refrain from quoting a single best-fit value with symmetric errors, and instead rely on the sub-regions shown in the two blocks of Figure 5 to convey the structure of the allowed parameter space. To further characterize this multimodal structure, Figure 6 shows the profile likelihood in , obtained by maximizing the likelihood over at each value of . Unlike marginalization, this procedure is independent of the prior volume in and is therefore not affected by volume effects in the degenerate regime. In the joint analysis the profile exhibits several distinct peaks, including some at , consistent with the large- regions of Figure 5 being genuine features of the likelihood.
Figure 6. Profile likelihood in , , normalized to its maximum, for the EE-only (blue) and joint EE+BB+EB (red) analyses and the three ALP masses of Figure 5.
Summary and Discussion
We quantitatively explored the energy injection scenario in which a tiny fraction of the CDM energy density is transferred to the ALP energy density before the reionization epoch. Using a specific model of energy injection, we numerically showed that the ALP field value is enhanced after the energy injection. We then showed that the power spectrum at low- is suppressed due to a large rotation angle of cosmic birefringence after the reionization epoch. We found a parameter region of the model that simultaneously explains the Planck low- polarization data, a nonzero value of the rotation angle of cosmic birefringence obtained from power spectrum at high-, and a large value of suggested as a possible solution to the CMB-BAO tension. Although we have shown results for a single injection window and three representative masses, we have verified that similar parameter regions exist for all ALP masses and injection windows within .
In our constraints on the model parameters, we did not include the instrumental miscalibration angle. The miscalibration angle could introduce a nonzero power spectrum. However, both the -only and cases allow a larger value of at a similar value of
, suggesting that the impact of the miscalibration angle is not significant. The miscalibration angles at each frequency channel constrained from Ref. [20] are also within sub-degree, indicating that the impact of the miscalibration angle is much smaller than the observational errors of the power spectra at low-.
Our model predicts a large rotation angle of cosmic birefringence at low redshift, . Such a large rotation angle can be tested by observations of radio galaxies [76, 77]. Measurements of the low- polarization by future CMB experiments, such as LiteBIRD [78], are also important to validate the measurement of low- polarization power spectra.
Acknowledgments
We thank Elisa G. M. Ferreira for valuable suggestions at the early stage of this project. We also thank Luca Visinelli for useful comments on the draft. We acknowledge support from JSPS KAKENHI Grant No. JP22K03682, No. JP24KK0248, and No. JP25K00996, and from JST EXPERT-J, Japan Grant No. JPM-JEX2508. The Kavli IPMU is supported by World Premier International Research Center Initiative (WPI Initiative), MEXT, Japan. This work made use of the idark computing cluster at the Kavli IPMU.
Data availability
The data that support the findings of this article are publicly available at the following GitHub repository: https://github.com/SirlettiSS/birefINJ.
Energy injection and the spectator condition
In this appendix, we derive the evolution of the CDM and ALP energy densities during and after the injection, which are the basis for the depletion and spectator conditions discussed in Sec. II C. We denote by and the scale factors at the beginning and at the end of the injection window, and by the scale factor at which the initial conditions are set.
Dark matter evolution and depletion limit
The CDM density obeys Eq. (7) with the source of Eq. (11), i.e.
where is the step-function, equal to 1 inside the injection window and 0 outside. Since , the Hubble rate cancels when the equation is written in terms of the scale factor,
The equation can then be solved separately in the three intervals, requiring continuity of at and . Before the injection, there is no source, and coincides with the CDM solution , where is the present-day CDM value. We obtain
In the window, the density therefore scales as , while after the injection it dilutes again as , but with a reduced normalization.
Dividing by the CDM density, the depletion factor after the injection is (cf. Eq. (18))
which depends only on and on the width of the injection window, and is independent of and . Imposing the upper bound from the data as discussed in Sec. II C, , i.e. , gives Eq. (21). For the window used throughout the paper, .
ALP energy density
The energy density of the ALP obeys the continuity equation with the source of Eq. (11);
with . Using the scale factor as a time variable, this reads
with given by Eq. (29), and the initial condition . Assuming constant, the equation can be solved in the three intervals.
a. Interval 1, . There is no source and
b. Interval 2, . The equation to solve is
It is linear and can be solved exactly with the integrating factor , which turns the left-hand side into . The solution is
with the boundary values
At this gives
c. Interval 3, . The source vanishes again, and
Spectator ratio and the function
The ratio between the ALP and CDM densities follows from writing and ,
which for is Eq. (22). The dependence of , defined in Eq. (23), on and follows from the solution in the third interval, Eq. (38), once the constant term of Eq. (37) is expressed in terms of and . Evaluating it at and using we obtain
where for the second term we have used (there is no depletion before ) and to obtain
The first term in Eq. (40) is the free evolution of the initial energy and does not depend on or , while the injected energy is proportional to . Therefore, depends on and only through , up to the explicit dependence on in and , which is at the percent level for . The ratio at is then , which is the expression used in Eq. (22) to compute the ALP spectator limit in Eq. (19).
Limits of the analytical approximation
Equation (A14) is obtained for constant . To first approximation, the evolution of the field can be described by a sharp and instantaneous transition, at a scale factor , from , when the field is frozen in the potential by Hubble friction, to , when it oscillates around the minimum and behaves as matter. A natural choice is to place the transition at the end of the injection, . For up to and afterwards, Eq. (40) becomes
This is a good approximation of the exact result. As an example, Figure 7 compares the analytical ratio with the transition at to the numerical solution.

Figure 7. Evolution of the ratio for and , for (left), (center) and (right). The shaded band marks the injection window , the dash-dotted line marks , and the dotted horizontal line the spectator limit, Eq. (19). The thick blue line is the direct numerical solution of Eq. (15) at the given . The dotted yellow line is obtained by solving the equation once at fixed and rescaling the result as , as done to compute the contours of Figure 3. Its agreement with the direct numerical solution (thick blue) validates this construction. The dashed orange line is the analytical solution with an instantaneous transition from to at . It reproduces the numerical result well for small , but progressively underestimates it as increases. This is due to the intermediate phase between and , in which varies and is not captured by the sharp transition: its duration grows with , and for approaching it extends beyond , as shown by the oscillations of the numerical curve, which are larger and last longer in the right panel.
In reality, the transition from to is smooth. After the end of the injection, the field is still frozen in the potential until the Hubble friction becomes comparable to the mass, . In the matter era, , which gives
Between and , the equation of state is not constant but varies with redshift . This transition phase can end or extend even beyond , depending on how close is to its critical value. A strong injection gives the field a large velocity, such that its kinetic energy dominates over the potential energy, and can reach (kination). In this regime, the kinetic energy dilutes as , faster than matter, so that the ALP energy density decreases quickly until the field is again frozen by Hubble friction (), and the onset of the oscillations is delayed. Since this behavior is not captured by a constant , the sharp-transition approximation can both overestimate and underestimate the ratio, depending on .
Although the analytical approximation works well in most of the parameter space, we generate the exclusion plot shown in Figure 3 from the numerical solution of Eq. (15), which treats the evolution of exactly for every , following the procedure described in Sec. II C.
Figure 3. Allowed region in the plane for and . Black: spectator limit, ; dashed red: kick-out limit with (gray: broken dynamics); dash-dotted blue: CDM depletion limit .
References
- [1]DESI Collaboration , M. Abdul Karim et al. Phys. Rev. D 112 (2025), no. 8 083515, arXiv:2503.14738.
- [2]SPT-3G Collaboration , E. Camphuis et al. Phys. Rev. D 113 (2026), no. 8 083504, arXiv:2506.20707.
- [3]W. Elbers et al. Phys. Rev. D 112 (2025), no. 8 083513, arXiv:2503.14744.arxiv.org/abs/2503.14744
- [4]N. Craig, D. Green, J. Meyers, and S. Rajendran J. High Energy Phys. 09 (2024) 097, arXiv:2405.00836.arxiv.org/abs/2405.00836
- [5]D. Green and J. Meyers Phys. Rev. D 111 (2025), no. 8 083507, arXiv:2407.07878.arxiv.org/abs/2407.07878
- [6]W. Elbers, C. S. Frenk, A. Jenkins, B. Li, and S. Pascoli Phys. Rev. D 111 (2025), no. 6 063534, arXiv:2407.10965.arxiv.org/abs/2407.10965
- [7]P. G. Lynch and L. Knox Phys. Rev. D 112 (2025), no. 8 083543, arXiv:2503.14470.arxiv.org/abs/2503.14470
- [8]L. Herold and M. Kamionkowski Phys. Rev. D 111 (2025), no. 8 083518, arXiv:2412.03546.DOI
- [9]D. Naredo-Tuero, M. Escudero, E. Fernández-Martínez, X. Marcano, and V. Poulin Phys. Rev. D 110 (2024), no. 12 123537, arXiv:2407.13831.
- [10]W. Giarè, E. Di Valentino, and A. Melchiorri Phys. Rev. D 109 (2024), no. 10 103519, arXiv:2312.06482.DOI
- [11]M. Loverde and Z. J. Weiner J. Cosmol. Astropart. Phys. 12 (2024) 048, arXiv:2410.00090.DOI
- [12]N. Sailer, G. S. Farrens, S. Ferraro, and M. White Phys. Rev. Lett. 136 (2026), no. 8 081002, arXiv:2504.16932.DOI
- [13]T. Jhaveri, T. Karwal, and W. Hu Phys. Rev. D 112 (2025), no. 4 043541, arXiv:2504.21813.
- [14]S. di Serego Alighieri Proceedings of the XI Multifrequency Behaviour of High Energy Cosmic Sources Workshop (MULTIF15) (2015) 009, arXiv:1507.02433.
- [15]E. Komatsu Nature Rev. Phys. 4 (2022), no. 7 452–469, arXiv:2202.13919.DOI
- [16]Y. Minami and E. Komatsu Phys. Rev. Lett. 125 (2020), no. 22 221301, arXiv:2011.11254.DOI
- [17]A. Gruppuso, G. Maggio, D. Molinari, and P. Natoli JCAP 05 (2016) 020, arXiv:1604.05202.DOI
- [18]Planck Collaboration , N. Aghanim et al. Astron. Astrophys. 596 (2016) A110, arXiv:1605.08633.
- [19]M. Ballardini, A. Gruppuso, S. Paradiso, S. S. Sirletti, and P. Natoli JCAP 09 (2025) 075, arXiv:2507.16714.arxiv.org/abs/2507.16714
- [20]J. R. Eskilt and E. Komatsu Phys. Rev. D 106 (2022), no. 6 063503, arXiv:2205.13962.DOI
- [21]J. R. Eskilt arXiv:2608.06480.
- [22]Y. Nakai, R. Namba, I. Obata, Y.-C. Qiu, and R. Saito JHEP 01 (2024) 057, arXiv:2310.09152.DOI
- [23]S. M. Carroll Phys. Rev. Lett. 81 (1998) 3067–3070, astro-ph/9806099.
- [24]G.-C. Liu, S. Lee, and K.-W. Ng Phys. Rev. Lett. 97 (2006) 161303, astro-ph/0606248.
- [25]S. Panda, Y. Sumitomo, and S. P. Trivedi Phys. Rev. D 83 (2011) 083506, arXiv:1101.5877.
- [26]T. Fujita, Y. Minami, K. Murai, and H. Nakatsuka Phys. Rev. D 103 (2021), no. 6 063508, arXiv:2008.02473.DOI
- [27]T. Fujita, K. Murai, H. Nakatsuka, and S. Tsujikawa Phys. Rev. D 103 (2021), no. 4 043509, arXiv:2011.11894.DOI
- [28]G. Choi, W. Lin, L. Visinelli, and T. T. Yanagida Phys. Rev. D 104 (2021), no. 10 L101302, arXiv:2106.12602.DOI
- [29]I. Obata J. Cosmol. Astropart. Phys. 09 (2022) 062, arXiv:2108.02150.DOI
- [30]S. Gasparrotto and I. Obata J. Cosmol. Astropart. Phys. 08 (2022), no. 08 025, arXiv:2203.09409.
- [31]M. Galaverni, F. Finelli, and D. Paoletti Phys. Rev. D 107 (2023), no. 8 083529, arXiv:2301.07971.DOI
- [32]W. Lin, L. Visinelli, and T. T. Yanagida JCAP 10 (2025) 023, arXiv:2504.17638.DOI
- [33]K. Murai, F. Naokawa, T. Namikawa, and E. Komatsu Phys. Rev. D 107 (2023) L041302, arXiv:2209.07804.DOI
- [34]J. R. Eskilt, L. Herold, E. Komatsu, K. Murai, T. Namikawa, and F. Naokawa Phys. Rev. Lett. 131 (2023), no. 12 121001, arXiv:2303.15869.
- [35]J. Kochappan, L. Yin, B.-H. Lee, and T. Ghosh Phys. Rev. D 112 (2025), no. 6 063562, arXiv:2408.09521.
- [36]F. Finelli and M. Galaverni Phys. Rev. D 79 (2009) 063002, arXiv:0802.4210.DOI
- [37]G. Sigl and P. Trivedi (11, 2018) arXiv:1811.07873.arxiv.org/abs/1811.07873
- [38]G.-C. Liu and K.-W. Ng Phys. Dark Univ. 16 (2017) 22–25, arXiv:1612.02104.arxiv.org/abs/1612.02104
- [39]M. A. Fedderke, P. W. Graham, and S. Rajendran Phys. Rev. D 100 (2019) 015040, arXiv:1903.02666.DOI
- [40]D. Zhang, E. G. M. Ferreira, I. Obata, and T. Namikawa Phys. Rev. D 110 (2024), no. 10 103525, arXiv:2408.08063.arxiv.org/abs/2408.08063
- [41]T. Namikawa, K. Murai, and F. Naokawa Phys. Rev. D 112 (2025), no. 12 123513, arXiv:2506.20824.arxiv.org/abs/2506.20824
- [42]F. Takahashi and W. Yin J. Cosmol. Astropart. Phys. 04 (2021) 007, arXiv:2012.11576.DOI
- [43]N. Kitaima, F. Kozai, F. Takahashi, and W. Yin J. Cosmol. Astropart. Phys. 10 (2022) 043, arXiv:2205.05083.
- [44]M. Jain, R. Hagimoto, A. J. Long, and M. A. Amin J. Cosmol. Astropart. Phys. 10 (2022) 090, arXiv:2208.08391.
- [45]D. Gonzalez, N. Kitajima, F. Takahashi, and W. Yin Phys. Lett. B 843 (2023) 137990, arXiv:2211.06849.DOI
- [46]J. Lee, K. Murai, F. Takahashi, and W. Yin Phys. Rev. D 112 (2025), no. 4 043538, arXiv:2503.18417.DOI
- [47]R. C. Myers and M. Pospelov Phys. Rev. Lett. 90 (2003) 211601, hep-ph/0301124.
- [48]K. R. S. Balaji, R. H. Brandenberger, and D. A. Easson J. Cosmol. Astropart. Phys. 12 (2003) 008, hep-ph/0310368.
- [49]A. Arvanitaki, S. Dimopoulos, S. Dubovsky, N. Kaloper, and J. March-Russell Phys. Rev. D 81 (2010) 123530, arXiv:0905.4720.DOI
- [50]T. Namikawa Phys. Rev. Lett. 135 (2025), no. 16 161004, arXiv:2506.22999.arxiv.org/abs/2506.22999
- [51]F. Naokawa, T. Namikawa, K. Murai, I. Obata, and K. Kamada Phys. Rev. Lett. 136 (2026), no. 4 041003, arXiv:2405.15538.arxiv.org/abs/2405.15538
- [52]P. W. Graham, D. Green, and J. Meyers Phys. Rev. D 113 (2026), no. 4 043514, arXiv:2508.20999.DOI
- [53]G. Gubitosi, M. Martinelli, and L. Pagano J. Cosmol. Astropart. Phys. 12 (2014), no. 2014 020, arXiv:1410.1799.DOI
- [54]S. Lee, G.-C. Liu, and K.-W. Ng Phys. Rev. D 89 (2014), no. 6 063010, arXiv:1307.6298.DOI
- [55]B. D. Sherwin and T. Namikawa Mon. Not. R. Astron. Soc. 520 (2021) 3298–3304, arXiv:2108.09287.DOI
- [56]H. Nakatsuka, T. Namikawa, and E. Komatsu Phys. Rev. D 105 (2022) 123509, arXiv:2203.08560.DOI
- [57]F. Naokawa and T. Namikawa Phys. Rev. D 108 (2023), no. 6 063525, arXiv:2305.13976.arxiv.org/abs/2305.13976
- [58]K. Murai Phys. Rev. D 111 (2025), no. 4 043514, arXiv:2407.14162.DOI
- [59]S. Arcari, N. Bartolo, A. Greco, A. Gruppuso, M. Lattanzi, and P. Natoli JCAP 10 (2024) 101, arXiv:2407.02144.
- [60]S. M. Carroll, G. B. Field, and R. Jackiw Phys. Rev. D 41 (1990) 1231.
- [61]S. M. Carroll and G. B. Field Phys. Rev. D 43 (Jun, 1991) 3789–3793.
- [62]D. Harari and P. Sikivie Phys. Rev. B 289 (1992) 67–72.
- [63]M. Kamionkowski and R. R. Caldwell arXiv:2609.36000.
- [64]M. Zaldarriaga and U. Seljak Phys. Rev. D 55 (1997) 1830–1840, astro-ph/9609170.
- [65]J. Lesgourgues arXiv e-prints (2011) arXiv:1104.2932.arxiv.org/abs/1104.2932
- [66]D. Blas, J. Lesgourgues, and T. Tram JCAP 07 (2011) 034, arXiv:1104.2933.DOI
- [67]M. B. Gavela, D. Hernandez, L. Lopez Honorez, O. Mena, and S. Rigolin JCAP 07 (2009) 034, arXiv:0901.1611. [Erratum: JCAP 05, E01 (2010)].
- [68]V. Salvatielli, A. Marchini, L. Lopez-Honorez, and O. Mena Phys. Rev. D 88 (2013), no. 2 023531, arXiv:1304.7119.DOI
- [69]E. Di Valentino, A. Melchiorri, and O. Mena Phys. Rev. D 96 (2017), no. 4 043503, arXiv:1704.08342.DOI
- [70]B. Wang, E. Abdalla, F. Atrio-Barandela, and D. Pavon Rept. Prog. Phys. 79 (2016), no. 9 096901, arXiv:1603.08299.DOI
- [71]E. Di Valentino, R. Z. Ferreira, L. Visinelli, and U. Danielsson Phys. Dark Univ. 26 (2019) 100385, arXiv:1906.11255.DOI
- [72]Planck Collaboration Astron. Astrophys. 641 (2020) A6, arXiv:1807.06209.
- [73]Planck Collaboration, R. Adam et al. Astron. Astrophys. 596 (2016) A108, arXiv:1605.03507.
- [74]M. Tristram et al. Astron. Astrophys. 647 (2021) A128, arXiv:2010.01139.DOI
- [75]M. Tristram et al. Phys. Rev. D 105 (2022), no. 8 083524, arXiv:2112.07961.DOI
- [76]S. M. Carroll and G. B. Field Phys. Rev. Lett. 79 (1997) 2394–2397, astro-ph/9704263.
- [77]F. Naokawa Phys. Rev. Lett. 136 (2026), no. 4 041004, arXiv:2504.06709.arxiv.org/abs/2504.06709
- [78]LiteBIRD Collaboration Prog. Theor. Exp. Phys. (2, 2022) arXiv:2202.02773.





