KCL-2020-56

Imprints of Axion Superradiance in the CMB

Diego Blas1, ∗ and Samuel J. Witte2, † 1Theoretical Particle and Cosmology Group, Department of Physics, King’s College London, Strand, London WC2R 2LS, UK 2Instituto de F´ısica Corpuscular (IFIC), CSIC-Universitat de Val`encia,Spain

−10 Light axions (ma . 10 eV) can form dense clouds around rapidly rotating astrophysical black holes via a mechanism known as rotational superradiance. The coupling between axions and photons induces a parametric resonance, arising from the stimulated decay of the axion cloud, which can rapidly convert regions of large axion number densities into an enormous flux of low- photons. In this work we consider the phenomenological implications of a superradiant axion cloud undergoing resonant decay. We show that the low energy photons produced from such events will be absorbed over cosmologically short distances, potentially inducing massive shockwaves that heat and ionize the IGM over Mpc scales. These shockwaves may leave observable imprints in the form of anisotropic spectral distortions or inhomogeneous features in the optical depth.

I. INTRODUCTION slowly decay via the emission of gravitational waves, with amplitudes that may be detectable in future ex- The evidence for the existence of dark matter is abun- periments [20–25]. Observations of highly rotating black dant in both cosmology and , however its un- holes have consequently been used to probe and con- derlying nature has evaded detection efforts for decades. strain light non-interacting bosons with masses in the −21 −10 The axion, a pseudo goldstone boson arising from the range 10 . mφ . 10 eV [9, 20, 21, 24–28]. The existence of self-interactions or couplings to other breaking of a global UPQ(1) symmetry, has recently emerged as one of the most favored dark matter candi- particles may complicate and quench the superradiant dates. Originally introduced as a solution to the strong- growth; for example, in the case of the axion, self- CP problem [1–4], it was soon realized that light axions interactions become important when the field value a −3 approaches the axion decay constant f , and may pro- ma . 10 eV were both cosmologically stable and could a be abundantly produced in the early Universe via the duce an explosion known as a ‘bosenova’ [9, 20]. Particle misalignment mechanism [5–8]. Ultralight bosonic fields production and scattering processes may dramatically in- have also been predicted to naturally appear from moduli crease the characteristic timescale required to extract the compactification in string theory [9, 10]; while these par- black hole’s angular , and thus provide a natu- ticles do not solve the strong-CP problem, they behave ral way for evading current superradiant constraints [29– in a comparable manner to the QCD axion, and thus 31]. More recently, it has been pointed out that the are frequently referred to as axion-like particles. In what parametric resonance arising from the repeated stimu- follows, we will use the word ‘axion’ interchangeably to lated decay of a superradiant axion cloud may quench refer to both. the growth, producing in the process an enormous flux −10 of low energy photons [26, 32–34]. While initial stud- Light axions with masses ma . 10 eV are known to trigger instabilities around rotating astrophysical black ies of this phenomenon assumed adiabatic evolution in a holes, even if their initial abundance is negligibly small. flat-spacetime, recent work has demonstrated that axion This processes, known as black hole superradiance, pro- fields coupled to electromagnetism in Kerr background duces an exponentially growing boson cloud at the robustly exhibit this phenomenon, so long as the axion- expense of the black hole’s rotational energy [11–16] photon coupling exceeds a threshold value [33, 34]. (see [17] for a recent review on rotational superradiance In this work we investigate the phenomenological im- and its application to black hole physics). In the limit plications that would arise from the resonant decay of a that the axion’s interactions can be neglected (that is, superradiant axion cloud. We show that the low energy arXiv:2009.10074v2 [astro-ph.CO] 3 Nov 2020 the limit that the axion’s equation of motion is governed photons produced in the resonant decay have short cos- by the Klein-Gordon equation), the axion cloud will grow mological mean free paths, and are typically absorbed until either the superradiant condition, given by ω < mΩ over ∼ O(pc) scales. This leads to enormous pressure – with ω the energy of the mode, m the azimuthal quan- gradients, which can induce a shockwave thats heats tum number and Ω the rotational frequency– is vio- and ionizes the surrounding medium as the shock front lated or a significant fraction of the black hole’s rota- pushes outward. We show that these shockwaves leave tional energy has been extracted1; this cloud will then anisotropic features in the spectrum and optical depth of the Cosmic Microwave Background (CMB); individual

∗ e-mail: [email protected] † e-mail: [email protected] bility is robust against gravitational emission, back-reaction on 1 Current work suggests the appearance and growth of the insta- the metric, and the presence of accretion disks [18, 19]. 2 features for a single black hole may be detectable with is [38–40] near-future spectral distortion experiments, while large- 2 scale features arising from the cumulative distribution Ma ∼ 1600 (GM) fa ; (4) of black holes may be near current detection thresholds, we emphasize that this relation was computed for however this latter point requires a more dedicated study. G maM ∼ 0.4 only, and the scaling with this parame- This manuscript is organized as follows. We review ter has not yet been rigorously tested. Should the axion the phenomenon of axion induced black hole superradi- field grow to this point, a bosenova may ensue in which ance in Sec.II, including possible outcomes (specifically axions are partially ejected and partially absorbed by the we focus on the possibility of either a bosenova or photon black hole. Axions, however, generically couple to elec- production via resonant decay). In Sec.III we discuss the tromagnetism via propagation of the low energy photons in the IGM, and gaγγ the subsequent formation and properties of the shock- L ⊃ − a F µν F˜ , (5) 2 µν wave. Observable features arising from such shockwaves are discussed in Sec.IV. We conclude in Sec.V. where F and F˜ are the electromagnetic field strength tensor and its dual; this coupling generates a two-body decay to photons, which in the case of non-relativistic II. AXION SUPERRADIANCE axions are approximately monochromatic with Eγ ∼ ma/2. If the axion field is sufficiently dense and the Rotational superradiance is a phenomenon that occurs energy distribution sufficiently narrow, the production of when a low energy boson with energy ω < mΩ scatters off photons at this frequency via axion decays may stimu- an absorbing surface rotating with frequency Ω (m here late the decay of the other axions in the cloud [41–48]3. being the azimuthal quantum number). In this scatter- This processes is typically suppressed in conventional as- ing process, the amplitude of the reflected boson field is trophysical and cosmological environments, due to either enhanced with respect to the incident amplitude. Should the finite width of the axion momentum distribution or there exist a confining mechanism for the reflected radi- gravitational de-tuning (see e.g. [46, 47]); superradiant ation, this process will extract the rotational energy at axion fields naturally evade these concerns, however, due an exponential rate until the growth is quenched by ad- to their enormous densities. Should the resonant decay of ditional interactions or the superradiant condition is no a superradiant axion cloud occur, an enormous burst of longer satisfied. monochromatic low-energy photons will be emitted that In the context of black hole superradiance, a non-zero will propagate away from the black hole and into the particle mass naturally acts to confine the light boson IGM. around the black hole, and exponential growth can occur. It has recently been shown numerically that an axion The timescale of the instability is directly related to the field coupled to electromagnetism in a does mass of boson ma and the black hole M, and for the indeed induce this parametric resonant decay, but only case of scalar particles can be as large as (in the limit if the axion-photon coupling is sufficiently large. The 2 G maM  1 )[20, 35–37] condition derived on axion-photons couplings for which this occurs is given by [33, 34]  M   0.2 9 τ ∼ 102 s . (1) r M 1 sr 10 M G m M −19 −1 a gaγγ & 10 2 GeV . (6) Ma (G ma M) The size of the generated axion cloud is approximately One can compare the fractional mass in the axion cloud given by at the time of a bosenova with that at the time of the (` + n + 1)2 GM electromagnetic burst (i.e. Eq. (4) with Eq. (6)); assum- 4 rcloud ∼ 2 , (2) ing fa ∼ 1/gaγγ one finds the axion decay occurs prior (G maM) which extends well beyond the horizon where rotational effects can be neglected. 3 It is worth mentioning that the difference between the resonant Assuming the axion potential is given by the conven- decay of the axion field (i.e. the focus of this work) and the stim- tional ulated decay discussed in [49, 50], is that the latter does not assume photons produced from the decays of axions contribute    2 2 a to the photon occupation number responsible for stimulating the V (φ) = ma fa 1 − cos , (3) decay; rather, the photon occupation number is assumed only fa to arise from CMB photons, galactic synchrotron and free-free non-linear self-interactions of the axion may become im- emission, etc. The approach of [49, 50] thus represents the con- servative limit of the resonant decay process. portant when the mass contained in the axion cloud Ma 4 In general there also exists an additional model-dependent pre- factor relating fa to gaγγ which may significantly alter this rela- tion. For the QCD axion, relation changes to fa = α/(4π) × (E/N) , with E and N given by anomaly coefficients (see 2 We work here with ~ = c = 1. e.g. [51]). 3

FIG. 1. Mean free path of photons produced from the resonant decay of superradiant axions of mass ma near a black hole of mass M, assumed to occur at z = 15 (left) or z = 6 (right). Interesting regions of parameter space are identified in color, while regions where energy injection is inefficient or impossible are heuristically identified. to the bosenova if is no longer valid or the timescale for superradiance is longer than the age of the Universe. The most likely 3 4 2 × 10 (G ma M) & 1 . (7) process for changing the black hole mass is accretion. The minimum timescale over which accretion is expected This relation is only valid at (G m M) ∼ 0.4 (note for a to change the black hole mass is given by τedd, i.e. the a 1 M black hole, this choice corresponds to an axion characteristic timescale for Eddington limited accretion, −11 mass of ma ∼ 6 × 10 eV) as Eq. (4) has only been given by [17] derived at this value – here, one finds the condition is, at least naively, safely satisfied. The peak luminosity was M σT 15 τedd ' ∼ ∼ 1.4  × 10 s. (10) estimated numerically, and was found (assuming ma ∼ M˙ 4πGmp 0.2/(GM) and a coupling gaγγ near the threshold value) to be roughly [34] Here,  is the radiative efficiency of the black hole, taken here to be ∼ 0.3 [52]. The constraint on the the mass of   dE Ma the axion cloud is thus given by ∼ 1066 eV/s . (8) dt M         Ma M τb s . β . (11) This emission is expected to occur in bursts; while the M M τ∆ τedd properties of the bursts are not known, we can approxi- mate the time between bursts as the timescale required to This value is maximally saturated near 10−6, suggesting 2 60 replenish the axion cloud τb ∼ 10 τsr, and the duration a maximal peak luminosity near 10 eV/s, although this of the bursts τ∆ ∼ O(ms) [26]. We caution the reader value may also be smaller. that large uncertainties are associated with both of these It is worth mentioning that Ref. [26] argued for the −2 numbers. We can also estimate the average luminosity existence of a maximum black hole mass Mc ∼ 10 M , over timescales t  τb, which is roughly given by above which the parametric resonance would not occur. This argument, however, was derived for the QCD ax- τ  L ∼ ∆ L. (9) ion, and does not generically apply to axion-like particles. τb Here, we focus on astrophysical (rather than primordial) black holes, and thus the remainder of this discussion The size of the axion cloud when these bursts occur has applies only to axion-like particles. also not been robustly determined. We can, however, de- Finally, a comment is in order regarding the environ- termine an upper limit by requiring that this emission mental dependence of these events. Photons acquire an does not significantly change the total black hole spin – effective mass in the IGM due to their interactions with in this way, this axion candidate can evade current su- the ambient plasma, given by the plasma frequency: perradiant constraints. We do this by imposing that the r time-integrated luminosity does not exceed a fraction β 4παne of the black hole’s rotational energy – for the results pre- ωp = , (12) me sented here, we will fix β = 0.5 in order to optimistically assess the detection thresholds. In general, axion super- where ne is the electron number density. The plasma radiance is expected to occur until the black hole mass frequency in IGM at redshifts z . 20 has typical val- grows to a point where either the superradiant condition ues of the order ∼ 10−14 eV, and thus axions of mass 4

III. SHOCKWAVES IN THE IGM

Axion clouds undergoing resonant decay can produce photons with short mean free paths in the IGM. Such low energy photons efficiently heat the gas via free-free absorption. The rate of absorption per unit volume is given by [71]

Γff = nγ ne σff , (13)

with the free-free cross section in the limit Eγ  T

4π2ασ rm g (E ,T ) √ T e ff γ σff ' np 2 . (14) 6π T TEγ

FIG. 2. Axion parameter space identified as capable of in- Here, T is the temperature of the gas, np the proton ducing shockwaves in the IGM. Shown for comparison are number density, and gff the gaunt factor, approximated constraints from helioscopes [65] and SN1987a [66] (red), and here as [71] haloscopes [67–69] (light blue). The edge of the QCD axion band is highlighted in green in lower right corner [70].   ν  g ' 4.691 1 − 0.118 ln γ . (15) ff 1010(T/104 K)3/2

For electron number densities and temperatures typical m 5 × 10−13 eV are kinematically forbidden from de- a . of the IGM, the mean free path of these particles, given caying to photons, unless they reside in a large under by λ = (n ×σ )−1, can be much smaller than parsec density. For this reason, we choose to focus our atten- mfp e ff scales. If one assumes ∼ 1057 eVs of energy (i.e. assum- tion on black holes with masses M 5 × 103M . It is . ing 1060 eV/s imparted over a burst width of 1 ms5) are important to bear in mind that black holes with masses instantaneously deposited in a parsec-sized volume, one near this threshold must reside in regions of the Uni- expects the ambient particles to be heated to tempera- verse where the baryon overdensity, ∆ ≡ n /n¯ , is not b b tures as high as ∼ 1012 K (in this estimation we have enormously different from one. While black holes are of taken z = 15 and x ∼ 10−4, where x ≡ n /n is the course expected to preferentially be found in over dense e e e H free electron fraction with n and n the electron and regions, simulations of black hole mergers have suggested e H hydrogen number densities). The large pressure gradient that remanent black holes may be produced with veloc- induced from this heating will induce an outward driven ities ∼ O(5000) km/s [53–60] – these ‘superkicks’ are shockwave that sweeps through the IGM, heating and expected to occur when the spins of the merging black ionizing large regions of space; we attempt to quantify holes are antiparallel and lying in the orbital plane. Even the effects of such a process here. larger kicks, potentially reaching v ∼ 0.1, may be gener- Before continuing, we identify the regions of parameter ated should the merging black holes have hyperbolic en- space for which a shockwave may develop. First, we are counters and the collision be ultra-relativistic [61]. Black interested in considering astrophysical black holes, and holes formed with such superkicks may easily escape their thus we restrict attention to M M . For a given black host galaxies, and if traveling at such speeds these objects & hole mass, there will exist upper and lower limit on the will have highly suppressed rate of accretion, meaning it axion mass, the former of which is set by requiring the may be possible to escape their local environment be- superradiant condition is valid, and the latter by requir- fore accumulating any additional matter. An alternative ing the superradiance timescale is sufficiently small so as possibility is that the supernova responsible for the cre- to extract energy. If the black hole mass is too large, ation of the black hole itself may be capable of ejecting axion decay may also be forbidden, and if the black hole sufficient material to reduce the local plasma mass be- is too small, the mean free path may be so large that low the kinematic threshold. Thus, the proposition of energy deposition is not efficient enough to drive a shock finding massive black holes in regions of ∆ ∼ 1 does not wave (we set this limit to be the point where the total seem unfathomable, although this problem presents an energy injection over the Eddington accretion timescale unavoidable uncertainty which may complicate an esti- only amounts to 102 eV / baryon, however observable ef- mate of the net rate at which shockwaves are formed. fects likely require a much greater energy injection). We Finally, the above concerns may be circumvented if the electromagnetic fields near the black hole are large. In this case, the photon dispersion relation can be modified, 5 and the effect of the plasma mass can be dramatically The luminosity is estimated here for a 10 M black hole with −7 reduced [62, 63]. This was recently pointed out in the Ma/M ∼ 10 , and taking a duration given by twice the de context of photon superradiance [64]. Broglie wavelength of the axion cloud. 5 plot in Fig.1 the mean free paths of the photons pro- blast wave is given by duced as a function of black hole mass (for a fixed axion   2 mass), assuming the event occurs either prior to reion- 8πGp 3 2 Ωb H R v˙ = 2 2 3 − (v − HR) − Ωcdm + , ization at z = 15 or post-reionization at z = 6. The Ωb H Rγ R 2 2 regimes in which the superradiant decay will be impossi- (17) ble or inefficient are identified heuristically. Using Fig.1, where the Lorentz factor γ3 accounting for the relativistic we identify the range of masses most likely to induce suppression must be included to suppress the outward −14 −11 shock waves as: 4 × 10 . ma . 6 × 10 eV. Us- force, since the initial wave travels very near the speed of ing the condition on the axion-photon coupling given in light. The equation of state for the interior of the plasma Eq. (6), we can identify the parameter space in mass and is given by axion-photon coupling potentially capable of generating 3 large shockwaves – this parameter space is highlighted in E = p Vsw , (18) Fig.2, where we show for comparison current constraints 2 from helioscopes [65], SN1987a [66], and haloscopes (light where Vsw is the volume contained in the shock bubble blue) [67–69], as well as the parameter space favored by and p the pressure. Energy conservation mandates the QCD axion [70] (green). dE dV In order to address the potential implications of a = L − p sw = L − p4πR2v . (19) shockwave induced from the absorption of escaping ra- dt dt diation, we adopt a formalism that has been developed The luminosity has contributions coming from: (1) the to trace the evolution of spherically symmetric explosions interior luminosity contributed by the superradiant event in the IGM, initially intended to treat supernovae winds LSR, (2) Compton cooling off CMB photons Lcomp, (3) from massive stars [72] (see e.g. also [73–78] for early Lbrem thermal bremsstrahlung emission, (4) Ld heating work on the subject). This treatment relies on the sim- from collisions between the shell and the IGM, and (5) ple assumption that there exists a three phase medium, Lion ionization of the surrounding medium. That is to consisting of: (1) a dense shell of radius R and thickness say the net luminosity is given by δR containing a fraction (1−fm) of the enclosed baryonic matter (assumed fm  1), (2) a uniform medium outside L = LSR + Ld − Lcomp − Lbrem − Lion . (20) the shell with mean IGM density, and (3) a hot isother- mal plasma inside the shell. The shell will be driven out- We account for the superradiant luminosity using the ward by the pressure induced from the absorption and time-averaged luminosity given in Eq. (9) and evolve the subsequent heating of the radiation escaping the black above equations from t = 0 to t = τedd; after the su- hole. As the shell expands, the plasma cools, ionizing perradiant energy injection has stopped, we convert the the newly encountered neutral hydrogen and dissipat- differential equations to redshift and use the final out- ing energy via Compton scattering with the CMB and put of the time-based solutions as initial conditions for bremsstrahlung emission. The medium may also undergo the cosmological evolution. We do not find significant heating via collisions with the surrounding IGM. Eventu- changes to our results if the equivalent total energy is ally, the shell velocity will asymptote to match the rate injected on over timescales t  τedd. Notice that within of expansion of the Universe, and the internal temper- these cosmological time scales, the heating of ions and of ature of the bubble will dissipate and asymptote to the neutral hydrogen can indeed be treated as instantaneous, background value. see e.g. [31]. The luminosity of Compton cooling of relativistic elec- Following [72], the total mass in the shell at a time t 4 3 trons off CMB photons is given by is given by m(t) = 3 πR(t) (1 − fm)ρb, and the rate of change is simply 4 π2 L = β2γ2σ T 4 n V (21) comp 3 T 15 γ e sw ˙ ! m˙ 1 d 3  R where Tγ is the temperature of the CMB, and that of = 3 R ρb = 3 − H (16) m R ρb dt R bremsstrahlung cooling is q −25 4 Lbrem = 1.422 × 10 Tg/10 K np ne g(T ) Vsw , (22) if R/R˙ > H, otherwise 0 (implying the shell simply expands outwards with the Hubble flow). The shell where Tg is the temperature of the gas and g(T ) is the will experience (1) a net braking force proportional to gaunt factor. Next, we need to address the fate of the (R˙ − HR˙ )m ˙ associated with the acceleration of the ma- energy created in the inelastic collisions between the shell terial encountered, (2) an outward pressure force 4πR2p and the IGM. This energy could be either radiated away induced from the hot interior, and (3) a gravitational de- in a shock cooling process or heat the shell (and subse- 4 2 celeration force 3 πGR(ρcdm + ρb/2) + GMBH /R , the quently the interior plasma). We adopt a conservative latter term being small at the scales of interest. One can approach to deal with this by parameterizing the uncer- show the resultant equation for the radial motion of the tainty in how much energy gets radiated / reabsorbed 6

3 FIG. 3. Shockwave solutions generated for M = 10 M and 30M , initiated at z = 15 and 6, respectively. Results are shown −2 for fiducial model of fd = 1 and fm = 0.1, and the effect of varying these parameters to fd = 0.5 and fm = 10 . All curves assume G maM = 0.2.

via a constant fd ∈ [0, 1], with the limit that fd → 1 and a 30 M black hole at z = 6, both with an axion representing the limit of total energy retention. The net mass ma ∼ 0.2/(GM). In our fiducial model, we take luminosity for this process is then given by fd = 1 and fm = 0.1, however we also consider fd = 0.5 −2 and fm = 10 to illustrate the effect of changing these 3m 3 Ld = fd (v − HR) . (23) parameters. The results, showing the comoving radius 2R Rc = (1 + z)R, velocity, pressure, and temperature evo- Lastly, the ionization luminosity simply accounts for the lution of each shockwave are shown in Fig.3. We evolve energy required to ionize the neutral hydrogen absorbed the system until the temperature of the bubble falls be- by the expanding wave, and is given by low that of the IGM. In the pre-reionization scenario, we assume the temperature of the external medium is given 2 Lion = fm fxe nb IH 4π R (v − HR) , (24) by that of ΛCDM, while lower redshifts we model the IGM temperature as [79] where IH = 13.6 eV is the ionization threshold of hy- drogen, and fxe is the fraction of neutral hydrogen. We assume f is the conventional tanh function used to de- xe ( 4 scribe reionization, with a central reionization redshift 10 K z > 3 T = 6 3.3 (26) taken to be zre = 7 and width ∆z = 0.5. Finally, the 10 /(1 + z) K z ≤ 3 . evolution of the internal pressure can be related to the net luminosity and the expansion of the shell via dp L v Under reasonable assumptions, these expanding pressure = − 5 p . (25) bubbles seem to be capable of expanding to comoving dt 2πR3 R distances as large as ∼ O(1) Mpc, and thus may leave With this formalism in hand, we restrict our atten- observable signatures in the form of heating and ioniza- 3 tion to two scenarios: a 10 M black hole at z = 15, tion. 7

IV. OBSERVABLE FEATURES in the optical depth. Inhomogeneous features in the op- tical depth alter the statistical features in the CMB via Finally, let us turn our attention to the observational the (i) screening of anisotropies (i.e. temperature and consequences of these shockwaves. The dominant im- polarization anisotropies are multiplied by a factor of −τ(ˆn) pact is a heating, and ionization in the case of the z = 15 e ), (ii) generation of polarization via Compton scat- superradiance, of the IGM over distance scales poten- tering, and (iii) the kinetic Sunyaev-Zel’dovich (kSZ) ef- tially as large as O(1) Mpc. While this may leave strong fect. The shockwaves identified here may naturally ap- signatures e.g. in the Ly-α forest and the evolution of pear as strong anomalies in future efforts to detect the structure, these signatures often rely on averaging over signatures from patchy reionization (see e.g. [89–93] for some set of statistics. Given the large uncertainties in the detection of patchy reionization in the CMB). The line rate and distribution of black holes undergoing superra- of sight contribution to the optical depth from a single diance, we would prefer to focus our attention on probes superradiant event is given by that may identify individual superradiant events, with Z 1 the understanding that prospects may improve should δτ = dz [ne(z) − ne,0(z)] σT , (28) the rate be larger. We believe the strongest of such sig- H(z)(1 + z) natures arises from angular features in the optical depth and spectral distortions in the CMB spectrum. where ne is the free electron fraction in the superradi- ant model and n is that from the default cosmologi- After recombination, heated gas can induce y-type e,0 cal model; we take the default reionization history to be spectral distortions in the CMB, arising from the effect of given by the conventional tanh reionization history, cen- Compton scattering. This effect is often parameterized tered at z = 7 with width ∆z = 0.5. The 3σ upper limit via the so-called y-parameter, given by [79] from [94] is shown for comparison. Telescopes measuring Z (T − T ) σ n small scale features in the optical depth may be capable y = dz cmb T e . (27) of identifying such features. c m H(z) (1 + z) e Thus far our discussion has been focused on the ef- fects arising from axion superradiance around a single It is well-known that the hot IGM after reionization is black hole. Needless to say, realistic population models expected to imprint a spectral distortion at the level of predict many black holes capable of efficiently inducing y ∼ 3×10−7 [80, 81]. The distortions introduced by hot c axion superradiance, and thus one may expect large scale gas in clusters may be larger, closer to 2 × 10−6, which features to appear; if the number of events is sufficiently may further complicate the identification of such effects large, it may be possible that FIRAS and Planck have (see e.g. [82]). Thus, if the generated shockwave can heat even already begun to probe interesting regions of pa- the IGM to a sufficiently high level (and over a sufficiently rameter space. We can estimate the approximate num- large angle in the sky), it may be possible to produce ber of events required to produce all-sky features simply strong signals on top of the astrophysical background. by looking at the ratio of the fractional sky coverage of We analyze the potential signal by looking at the rel- a single event; for ∼ 1 arc-minute scales, this fraction ative difference between the spectral distortion induced amounts to a factor of ∼ 10−9. Cosmologically, the exis- from the shockwave and that expected from astrophysical tence of a population of 109 rapidly rotating black holes sources. The ΛCDM contribution is modeled assuming in the mass range identified in Fig.1 is not unreason- the second ionization of helium occurred at z = 3.5 with able, and thus a dedicated study of the resonant decay of a width δz = 0.5, and the temperature of the IGM is de- superradiant axion clouds using black hole populations scribed in the previous section. This resultant difference SR ΛCDM would be of great interest. We leave this to future work. in spectral distortions is given by δyc = yc − yc , and is shown in the left panel Fig.4 as a function of an- gular degree on the sky. The leading constraints to date come from FIRAS, having measured the y−parameter V. CONCLUSIONS −5 to the level of yc . 10 ; FIRAS [85], however, had an angular resolution of ∼ 10◦, and thus would not Light axions can generate enormous clouds around have been capable of observing the induced heating from rapidly spinning black holes via a phenomenon known as axion superradiance (unless the rate of events is quite superradiance. For sufficiently large axion-photon cou- −15 −1 large). Proposed spectral distortion experiments such as plings (gaγγ & 10 GeV ), these clouds may undergo PIXIE [86], PRISM [83, 84], or the proposals for Voy- resonant decay into low energy monochromatic photons age 2050 [82, 87, 88] have angular resolutions as low as that propagate into the IGM. In this work we have inves- O(5) arc-seconds, and have estimated sensitivities near tigated the observable signatures that may arise in the −7 the level of yc ∼ 10 ; consequently Fig.4 shows that CMB from the superradiant production and subsequent photon superradiance has the potential to induce strong resonant decay of axions. signals observable in these experiments. The photons produced in such processes must have The ionization induced by superradiant events that are very low energies, no more than a few orders of mag- triggered prior to reionization can also leave an imprint nitude above the plasma frequency of the surrounding 8

FIG. 4. Relative contribution to Compton y−parameter (left), defined as δyc ≡ yc − yc, ΛCDM, and optical depth (right) as a function of angular scale. The 3σ upper limit on the optical depth from Planck, based on the sky-average value, is shown as a reference. For comparison, the angular resolution of PRISM [83, 84] spans between 10−1 and 17 arc-minutes, depending on the frequency. medium. Consequently, these photons are efficiently ab- dramatically enhance the observability of this signature, sorbed via inverse bremsstrahlung scattering off ambient however the uncertainty in the distribution of black holes electrons and ions, with typical mean free paths capable at high redshifts is large – we leave a more thorough of being much below the parsec scale. The medium is im- investigation of the effects of black hole populations to mediately heated to ultra-high temperatures, producing future work. Finally, we emphasize that if a putative de- a huge pressure gradient that can drive shockwaves to tection of either feature be observed, it may be possible Mpc scales. This process heats and ionizes the IGM, and to gain direct insight into the axion mass, the distribu- can leave an imprint on both the optical depth and y-type tions of highly spinning black holes, and the properties spectral distortions. For a single event, this is expected to of the IGM. Resonant decay of axion superradiance thus produce an anisotropic feature on arc-minute scales, be- offers an interesting and novel probe with importance to low the observable limit but potentially detectable with particle physics, astrophysics, and cosmology alike. −13 −1 future experiments. For couplings gaγγ & 10 GeV , spectral distortions may also arise from the conversion of Acknowledgments: The authors thank Vivian Poulin, CMB photons to axions, potentially providing a comple- Jens Chluba, Thomas Schwetz, Vitor Cardoso, and Paolo mentary probe [95–97]. Pani for their comments. SJW acknowledges support Accounting for realistic populations of black holes may under the Juan de la Cierva Formaci´onFellowship.

[1] R. Peccei and H. R. Quinn, Phys. Rev. Lett. 38, 1440 (1974). (1977). [14] T. Damour, N. Deruelle, and R. Ruffini, Lett. Nuovo [2] R. Peccei and H. R. Quinn, Phys. Rev. D 16, 1791 (1977). Cim. 15, 257 (1976). [3] S. Weinberg, Phys. Rev. Lett. 40, 223 (1978). [15] S. Detweiler, Physical Review D 22, 2323 (1980). [4] F. Wilczek, Phys. Rev. Lett. 40, 279 (1978). [16] V. Cardoso, O. J. Dias, J. P. Lemos, and S. Yoshida, [5] J. Preskill, M. B. Wise, and F. Wilczek, Phys. Lett. B Phys. Rev. D 70, 044039 (2004), [Erratum: Phys.Rev.D 120, 127 (1983). 70, 049903 (2004)], arXiv:hep-th/0404096. [6] M. Dine and W. Fischler, Phys. Lett. B 120, 137 (1983). [17] R. Brito, V. Cardoso, and P. Pani, Superradiance, Vol. [7] L. Abbott and P. Sikivie, Phys. Lett. B 120, 133 (1983). 906 (Springer, 2015) pp. 1501–06570, arXiv:1501.06570 [8] R. L. Davis, Phys. Lett. B 180, 225 (1986). [gr-qc]. [9] A. Arvanitaki, S. Dimopoulos, S. Dubovsky, N. Kaloper, [18] W. E. East, F. M. Ramazano˘ glu, and F. Pretorius, and J. March-Russell, Phys. Rev. D 81, 123530 (2010), Phys. Rev. D 89, 061503 (2014), arXiv:1312.4529 [gr-qc]. arXiv:0905.4720 [hep-th]. [19] R. Brito, V. Cardoso, and P. Pani, Class. Quant. Grav. [10] D. J. E. Marsh, Phys. Rept. 643, 1 (2016), 32, 134001 (2015), arXiv:1411.0686 [gr-qc]. arXiv:1510.07633 [astro-ph.CO]. [20] A. Arvanitaki and S. Dubovsky, Phys. Rev. D 83, 044026 [11] Y. B. Zel’Dovich, JETPL 14, 180 (1971). (2011), arXiv:1004.3558 [hep-th]. [12] W. H. Press and S. A. Teukolsky, Nature 238, 211 (1972). [21] A. Arvanitaki, M. Baryakhtar, and X. Huang, Phys. Rev. [13] S. Teukolsky and W. Press, Astrophys. J. 193, 443 D 91, 084011 (2015), arXiv:1411.2263 [hep-ph]. 9

[22] A. Arvanitaki, M. Baryakhtar, S. Dimopoulos, 102, 89 (2018), arXiv:1801.08127 [hep-ph]. S. Dubovsky, and R. Lasenby, Phys. Rev. D 95, [52] K. S. Thorne, Astrophys. J. 191, 507 (1974). 043001 (2017), arXiv:1604.03958 [hep-ph]. [53] M. Campanelli, C. O. Lousto, Y. Zlochower, and [23] M. Baryakhtar, R. Lasenby, and M. Teo, Phys. Rev. D D. Merritt, Phys. Rev. Lett. 98, 231102 (2007), 96, 035019 (2017), arXiv:1704.05081 [hep-ph]. arXiv:gr-qc/0702133. [24] R. Brito, S. Ghosh, E. Barausse, E. Berti, V. Cardoso, [54] B. Bruegmann, J. A. Gonzalez, M. Hannam, S. Husa, I. Dvorkin, A. Klein, and P. Pani, Phys. Rev. Lett. 119, and U. Sperhake, Phys. Rev. D 77, 124047 (2008), 131101 (2017), arXiv:1706.05097 [gr-qc]. arXiv:0707.0135 [gr-qc]. [25] R. Brito, S. Ghosh, E. Barausse, E. Berti, V. Cardoso, [55] J. Healy, F. Herrmann, I. Hinder, D. M. Shoemaker, I. Dvorkin, A. Klein, and P. Pani, Phys. Rev. D 96, P. Laguna, and R. A. Matzner, Phys. Rev. Lett. 102, 064050 (2017), arXiv:1706.06311 [gr-qc]. 041101 (2009), arXiv:0807.3292 [gr-qc]. [26] J. G. Rosa and T. W. Kephart, Phys. Rev. Lett. 120, [56] C. O. Lousto and Y. Zlochower, Phys. Rev. D 83, 024003 231102 (2018), arXiv:1709.06581 [gr-qc]. (2011), arXiv:1011.0593 [gr-qc]. [27] S. J. Zhu, M. Baryakhtar, M. A. Papa, D. Tsuna, [57] C. O. Lousto and Y. Zlochower, Phys. Rev. Lett. 107, N. Kawanaka, and H.-B. Eggenstein, (2020), 231102 (2011), arXiv:1108.2009 [gr-qc]. arXiv:2003.03359 [gr-qc]. [58] D. Gerosa and A. Sesana, Mon. Not. Roy. Astron. Soc. [28] H. Davoudiasl and P. B. Denton, Phys. Rev. Lett. 123, 446, 38 (2015), arXiv:1405.2072 [astro-ph.GA]. 021102 (2019), arXiv:1904.09242 [astro-ph.CO]. [59] D. Gerosa, F. Hebert, and L. C. Stein, Phys. Rev. D 97, [29] H. Fukuda and K. Nakayama, JHEP 01, 128 (2020), 104049 (2018), arXiv:1802.04276 [gr-qc]. arXiv:1910.06308 [hep-ph]. [60] U. Sperhake, R. Rosca-Mead, D. Gerosa, and E. Berti, [30] A. Mathur, S. Rajendran, and E. H. Tanin, (2020), Phys. Rev. D 101, 024044 (2020), arXiv:1910.01598 arXiv:2004.12326 [hep-ph]. [gr-qc]. [31] D. Blas and S. J. Witte, (2020), arXiv:2009.10075 [hep- [61] U. Sperhake, E. Berti, V. Cardoso, F. Pretorius, -ph]. and N. Yunes, Phys. Rev. D 83, 024037 (2011), [32] S. Sen, Phys. Rev. D 98, 103012 (2018), arXiv:1805.06471 arXiv:1011.3281 [gr-qc]. [hep-ph]. [62] P. Kaw and J. Dawson, The Physics of Fluids 13, 472 [33] M. Boskovic, R. Brito, V. Cardoso, T. Ikeda, (1970). and H. Witek, Phys. Rev. D 99, 035006 (2019), [63] C. Max and F. Perkins, Physical Review Letters 27, 1342 arXiv:1811.04945 [gr-qc]. (1971). [34] T. Ikeda, R. Brito, and V. Cardoso, Phys. Rev. Lett. [64] V. Cardoso, W.-d. Guo, C. F. Macedo, and P. Pani, 122, 081101 (2019), arXiv:1811.04950 [gr-qc]. (2020), arXiv:2009.07287 [gr-qc]. [35] T. Zouros and D. Eardley, Annals Phys. 118, 139 (1979). [65] V. Anastassopoulos et al. (CAST), Nature Phys. 13, 584 [36] S. L. Detweiler, Phys. Rev. D 22, 2323 (1980). (2017), arXiv:1705.02290 [hep-ex]. [37] H. Witek, V. Cardoso, A. Ishibashi, and U. Sperhake, [66] A. Payez, C. Evoli, T. Fischer, M. Giannotti, A. Mirizzi, Phys. Rev. D 87, 043513 (2013), arXiv:1212.0551 [gr-qc]. and A. Ringwald, JCAP 02, 006 (2015), arXiv:1410.3747 [38] H. Kodama and H. Yoshino, Int. J. Mod. Phys. Conf. [astro-ph.HE]. Ser. 7, 84 (2012), arXiv:1108.1365 [hep-th]. [67] C. Hagmann et al. (ADMX), Phys. Rev. Lett. 80, 2043 [39] H. Yoshino and H. Kodama, Prog. Theor. Phys. 128, 153 (1998), arXiv:astro-ph/9801286. (2012), arXiv:1203.5070 [gr-qc]. [68] S. J. Asztalos et al. (ADMX), Phys. Rev. D 64, 092003 [40] H. Yoshino and H. Kodama, PTEP 2015, 061E01 (2015), (2001). arXiv:1407.2030 [gr-qc]. [69] N. Du et al. (ADMX), Phys. Rev. Lett. 120, 151301 [41] I. Tkachev, Phys. Lett. B 191, 41 (1987). (2018), arXiv:1804.05750 [hep-ex]. [42] I. Tkachev, JETP Lett. 101, 1 (2015), arXiv:1411.3900 [70] L. Di Luzio, F. Mescia, and E. Nardi, Phys. Rev. Lett. [astro-ph.HE]. 118, 031801 (2017), arXiv:1610.07593 [hep-ph]. [43] A. Arza, Eur. Phys. J. C 79, 250 (2019), [71] B. T. Draine, Physics of the interstellar and intergalactic arXiv:1810.03722 [hep-ph]. medium (Princeton University Press, 2010). [44] M. P. Hertzberg and E. D. Schiappacasse, JCAP 11, 004 [72] M. Tegmark, J. Silk, and A. E. Evrard, Astrophys. J. (2018), arXiv:1805.00430 [hep-ph]. 417, 54 (1993), arXiv:astro-ph/9306021. [45] G. Sigl and P. Trivedi, (2019), arXiv:1907.04849 [as- [73] C. F. McKee and J. P. Ostriker, The Astrophysical Jour- tro-ph.CO]. nal 218, 148 (1977). [46] G. Alonso-Alvarez,´ R. S. Gupta, J. Jaeckel, and M. Span- [74] R. Weaver, R. McCray, J. Castor, P. Shapiro, and nowsky, JCAP 03, 052 (2020), arXiv:1911.07885 [hep- R. Moore, The Astrophysical Journal 218, 377 (1977). -ph]. [75] R. McCray and T. P. Snow Jr, Annual Review of Astron- [47] A. Arza, T. Schwetz, and E. Todarello, (2020), omy and Astrophysics 17, 213 (1979). arXiv:2004.01669 [hep-ph]. [76] F. Bruhweiler, T. Gull, M. Kafatos, and S. Sofia, The [48] D. Levkov, A. Panin, and I. Tkachev, Phys. Rev. D 102, Astrophysical Journal 238, L27 (1980). 023501 (2020), arXiv:2004.05179 [astro-ph.CO]. [77] R. McCray and M. Kafatos, (1987). [49] A. Caputo, C. P. Garay, and S. J. Witte, Phys. Rev. [78] J. P. Ostriker and C. F. McKee, Rev. Mod. Phys. 60, 1 D 98, 083024 (2018), [Erratum: Phys.Rev.D 99, 089901 (1988). (2019)], arXiv:1805.08780 [astro-ph.CO]. [79] R. A. Sunyaev and R. Khatri, Int. J. Mod. Phys. D 22, [50] A. Caputo, M. Regis, M. Taoso, and S. J. Witte, JCAP 1330014 (2013), arXiv:1302.6553 [astro-ph.CO]. 03, 027 (2019), arXiv:1811.08436 [hep-ph]. [80] Y. B. Zeldovich and R. A. Sunyaev, Astrophys. Space [51] I. G. Irastorza and J. Redondo, Prog. Part. Nucl. Phys. Sci. 4, 301 (1969). [81] W. Hu, D. Scott, and J. Silk, Astrophys. J. Lett. 430, 10

L5 (1994), arXiv:astro-ph/9402045. M. McQuinn, Astrophys. J. 630, 657 (2005), [82] J. Chluba et al., (2019), arXiv:1909.01593 [astro-ph.CO]. arXiv:astro-ph/0503166. [83] P. Andre et al. (PRISM), (2013), arXiv:1306.2259 [as- [91] C. Dvorkin and K. M. Smith, Phys. Rev. D 79, 043003 tro-ph.CO]. (2009), arXiv:0812.1566 [astro-ph]. [84] P. Andre et al. (PRISM), JCAP 02, 006 (2014), [92] T. Namikawa, Phys. Rev. D 97, 063505 (2018), arXiv:1310.1554 [astro-ph.CO]. arXiv:1711.00058 [astro-ph.CO]. [85] D. J. Fixsen, E. S. Cheng, J. M. Gales, J. C. Mather, [93] A. Roy, A. Lapi, D. Spergel, and C. Baccigalupi, JCAP R. A. Shafer, and E. L. Wright, Astrophys. J. 473, 576 05, 014 (2018), arXiv:1801.02393 [astro-ph.CO]. (1996), arXiv:astro-ph/9605054 [astro-ph]. [94] N. Aghanim et al. (Planck), Astron. Astrophys. 641, A6 [86] A. Kogut et al., JCAP 1107, 025 (2011), arXiv:1105.2044 (2020), arXiv:1807.06209 [astro-ph.CO]. [astro-ph.CO]. [95] S. Mukherjee, R. Khatri, and B. D. Wandelt, JCAP 04, [87] K. Basu et al., (2019), arXiv:1909.01592 [astro-ph.CO]. 045 (2018), arXiv:1801.09701 [astro-ph.CO]. [88] J. Delabrouille et al., (2019), arXiv:1909.01591 [as- [96] S. Mukherjee, R. Khatri, and B. D. Wandelt, JCAP 06, tro-ph.CO]. 031 (2019), arXiv:1811.11177 [astro-ph.CO]. [89] A. Gruzinov and W. Hu, Astrophys. J. 508, 435 (1998), [97] S. Mukherjee, D. N. Spergel, R. Khatri, and B. D. arXiv:astro-ph/9803188. Wandelt, JCAP 02, 032 (2020), arXiv:1908.07534 [as- [90] O. Zahn, M. Zaldarriaga, L. Hernquist, and tro-ph.CO].