betaFeedback

WorksPhysics and Astronomy

Pulsar Timing Correlations Induced by Stochastic Spin-2 Ultralight Dark Matter

Abstract

Ultralight dark matter (ULDM) can generate time-dependent gravitational perturbations, providing a potential signal for pulsar timing arrays (PTAs). In Galactic halos, the velocity dispersion of ULDM gives rise to finite spatial coherence, which influences the correlations between pulsar timing signals. In this work, we derive the timing-residual cross-correlations induced by rapid metric oscillations of spin-2 ULDM, allowing for general polarization statistics. Under equipartition among the five polarization modes, the intrinsic angular kernel contains monopole and quadrupole components and exhibits an angular profile closely resembling that of spin-1 ULDM, suggesting a potential degeneracy between the two ULDM candidates in PTA observations.

Introduction

Astrophysical and cosmological observations provide compelling evidence for dark matter, which accounts for most of the matter in the Universe [1], although its microscopic nature remains unknown [2]. The standard Λ\LambdaCDM model, in which cold dark matter (CDM) plays a central role, successfully describes the cosmic microwave background and the large-scale structure of the Universe [1, 3]. Despite this success, long-standing challenges associated with the CDM paradigm remain on small scales, including the core–cusp and missing-satellites problems [4]. These challenges motivate alternatives to CDM, including ULDM, whose macroscopic de Broglie wavelength can make wave effects relevant to structure formation on astrophysical scales [5, 6], while its behavior can remain approximately CDM-like on sufficiently large scales [7]. The large occupation number of ULDM in Galactic halos justifies a classical-field description [6, 7, 8]. The wave-like nature of ULDM can produce observable effects over a wide range of frequencies, motivating searches using pulsar timing [9], laser interferometry [10, 11, 12], and astrometry [13, 14]. ULDM can also be probed through its modulation of astrophysical gravitational-wave (GW) signals [15, 16].

Among these observational approaches, pulsar timing arrays (PTAs), which monitor the pulse arrival times of millisecond pulsars [17, 18], can probe ULDM with masses around 10−2410^{-24}–10−2210^{-22} eV [19, 20, 21]. The ULDM stress-energy tensor can source time-dependent metric perturbations, which induce fractional frequency shifts in propagating radio pulses and produce observable timing residuals [9, 22, 23, 24]. This gravitational effect has been investigated for ULDM with different spins [9, 24, 25, 26]. In particular, spin-2 ULDM can arise in ghost-free bimetric gravity, which contains a massless spin-2 mode mediating ordinary gravitational interactions and a massive spin-2 mode that can serve as a dark matter candidate [27, 28, 29, 30]. The massive mode has five physical polarization degrees of freedom, potentially giving rise to a rich geometric structure in its gravitational response. Indeed, previous studies have shown that the angular dependence of the timing signals can differ among spin-0 (scalar), spin-1 (vector), and spin-2 (tensor) ULDM, reflecting their different polarization structures [24, 25, 31, 32]. In addition to gravitational effects, ULDM can produce timing signals through direct couplings to Standard Model fields [33, 34, 35]. Such couplings have also been considered in studies of angular correlations induced by spin-2 ULDM [36].

Many earlier studies adopted simplified descriptions of the ULDM field, assuming monochromatic oscillations and considering two limiting cases of spatial coherence. In the fully correlated limit, the Earth and pulsars share the same ULDM field realization, whereas in the fully uncorrelated limit, their local fields are statistically independent. However, in a virialized Galactic halo, ULDM is more generally described as a superposition of nonrelativistic waves with different velocities and phases [6, 37, 38, 39]. The interference between these waves leads to a finite spatial coherence length ℓcoh∼(mv0)−1\ell_{\mathrm{coh}} \sim(m v_0)^{-1}, where mm is the ULDM mass and v0v_0 characterizes the velocity dispersion of dark matter in the Galactic halo [6, 38, 39]. For part of the ULDM mass range accessible to PTAs, this coherence length can be comparable to Earth–pulsar distances, so the ULDM fields at the Earth and pulsar locations are generally partially correlated. A stochastic-field treatment that retains finite spatial correlations is therefore needed to characterize the timing signals more completely.

Recent studies have investigated the timing correlations induced by stochastic spin-0 and spin-1 ULDM, accounting for the velocity distribution and finite spatial coherence [37, 38, 39, 40]. Because the stress-energy tensor is quadratic in the ULDM field, products of field modes generate sum- and difference-frequency contributions, corresponding to the conventional fast mode with angular frequency ωF≃2m\omega_{\mathrm{F}} \simeq2m and the more recently identified slow mode with characteristic frequency ωS∼mv2\omega_{\mathrm{S}} \sim mv^2, respectively. The slow mode probes higher ULDM masses around 10−1810^{-18}–10−16 eV10^{-16}\,\mathrm{eV} [37, 39]. For spin-1 ULDM, existing stochastic calculations have adopted polarization equipartition among its three physical modes [39].

For spin-2 ULDM, previous calculations of its gravitational timing response have largely relied on coherent-field descriptions [24]. In this work, we extend the analysis to a stochastic spin-2 field. We focus on the fast mode and derive the corresponding metric perturbations and timing-residual cross-correlations, accounting for finite spatial coherence and general polarization statistics among the five physical modes. Under five-polarization equipartition, we obtain an intrinsic angular kernel consisting of monopole and quadrupole components. Comparing this kernel with that of spin-1 ULDM under three-polarization equipartition, we find remarkably similar angular profiles, suggesting a potential degeneracy between the two spin sectors in PTA observations.

The remainder of this paper is organized as follows. § 2 develops the field model and derives the corresponding metric perturbations, while § 2.2 derives the timing-residual cross-correlations. § 2.3 discusses the angular correlations and compares the spin-2 and spin-1 cases. § 2.4 summarizes our findings.

We use natural units c=ℏ=1c = \hbar= 1 and the metric signature (−,+,+,+)(-,+,+,+). Greek indices μ,ν,…\mu,\nu,\ldots denote spacetime components, while Latin indices i,j,…i,j,\ldots denote spatial components. Repeated indices are summed over, and Newton’s constant GG is kept explicit.

Stochastic Spin-2 ULDM Field and Fast-Mode Gravitational Response

In this section, we begin by modeling spin-2 ULDM as a stochastic superposition of nonrelativistic field modes and then derive the gravitational response sourced by the sum-frequency components of its energy-momentum tensor.

Stochastic Spin-2 Field

We describe the massive spin-2 field MμνM_{\mu\nu} propagating on an effective background metric gμνg_{\mu\nu} using the quadratic Fierz–Pauli action [27, 28, 41],

SFP(2)=∫d4x−g LFP(2).(1)S_{\mathrm{FP}}^{(2)} = \int d^4x \sqrt{-g}\,\mathcal{L}_{\mathrm{FP}}^{(2)}. \tag*{(1)}

with

LFP(2)=−18∇λMμν∇λMμν+14∇λMμν∇νMμλ−14∇μMμν∇νM+18∇λM∇λM−18m2(MμνMμν−M2),(2)\mathcal{L}_{\mathrm{FP}}^{(2)} = -\frac{1}{8}\nabla_\lambda M_{\mu\nu}\nabla^\lambda M^{\mu\nu} + \frac{1}{4}\nabla_\lambda M_{\mu\nu}\nabla^\nu M^{\mu\lambda} - \frac{1}{4}\nabla_\mu M^{\mu\nu}\nabla_\nu M + \frac{1}{8}\nabla_\lambda M\nabla^\lambda M - \frac{1}{8}m^2\left(M_{\mu\nu}M^{\mu\nu}-M^2\right), \tag*{(2)}

where M≡MμμM \equiv M^\mu{}_\mu. On Galactic scales, we neglect cosmological expansion and work on a flat background gμν≃ημνg_{\mu\nu} \simeq\eta_{\mu\nu}. Varying the Fierz–Pauli action with respect to MμνM_{\mu\nu} gives

(□−m2)Mμν−∂μ∂ρMνρ−∂ν∂ρMμρ+∂μ∂νM+ημν(∂ρ∂σMρσ−□M+m2M)=0,(3)(\Box-m^2)M_{\mu\nu}-\partial_\mu\partial^\rho M_{\nu\rho}-\partial_\nu\partial^\rho M_{\mu\rho}+\partial_\mu\partial_\nu M+\eta_{\mu\nu}\left(\partial^\rho\partial^\sigma M_{\rho\sigma}-\Box M+m^2M\right)=0, \tag*{(3)}

Taking the divergence and trace of this equation yields, for m≠0m \ne0,

M=0,∂μMμν=0,(□−m2)Mμν=0.(4)M=0,\qquad\partial^\mu M_{\mu\nu}=0,\qquad(\Box-m^2)M_{\mu\nu}=0. \tag*{(4)}

For a Fourier mode with wave vector k\mathbf{k}, the dispersion relation is ωk=m2+k2\omega_k=\sqrt{m^2+k^2}. In a virialized Galactic halo, ∣k∣/m∼v∼v0∼10−3|\mathbf{k}|/m\sim v\sim v_0\sim10^{-3}, so ωk=m+k2/(2m)+O(mv4)=m+O(mv2)\omega_k=m+k^2/(2m)+\mathcal{O}(mv^4)=m+\mathcal{O}(mv^2) [6, 38, 39]. Transversality gives M0i=O(vMij)M_{0i}=\mathcal{O}(vM_{ij}) and M00=O(v2Mij)M_{00}=\mathcal{O}(v^2M_{ij}); together with M=0M=0, this implies Mii=O(v2Mij)M_{ii}=\mathcal{O}(v^2M_{ij}). To leading nonrelativistic order, the five degrees of freedom are thus described by a spatial symmetric trace-free (STF) tensor MijM_{ij} [24, 30]. We expand it as

Mij(t,x)=∑A=15∫d3k(2π)3MA(k)EijAe−iωkt+ik⋅x+c.c.,(5)M_{ij}(t,\mathbf{x})=\sum_{A=1}^{5}\int\frac{d^3k}{(2\pi)^3}M_A(\mathbf{k})E^A_{ij}e^{-i\omega_k t+i\mathbf{k}\cdot\mathbf{x}}+\mathrm{c.c.}, \tag*{(5)}

where EijAE^A_{ij} form a fixed orthonormal STF basis and MA(k)M_A(\mathbf{k}) are complex random mode amplitudes.

We assume invariance of their joint distribution under a common phase rotation MA(k)→eiϕMA(k)M_A(\mathbf{k}) \to e^{i\phi}M_A(\mathbf{k}). This implies ⟨MA(k)⟩=0\langle M_A(\mathbf{k})\rangle= 0 and ⟨MA(k)MB(k′)⟩=0\langle M_A(\mathbf{k})M_B(\mathbf{k}')\rangle= 0, where ⟨⋯ ⟩\langle\cdots\rangle denotes an ensemble average over stochastic field realizations [38, 39]. Statistical homogeneity gives the conjugated two-point function [42, 43]

⟨MA(k)MB∗(k′)⟩=(2π)3δ(3)(k−k′)PAB(k).(6)\langle M_A(\mathbf{k})M_B^*(\mathbf{k}')\rangle= (2\pi)^3\delta^{(3)}(\mathbf{k}-\mathbf{k}')P_{AB}(\mathbf{k}). \tag*{(6)}

To parameterize the momentum and polarization statistics independently, we adopt the separable model

PAB(k)=PMfk(k)WAB,(7)P_{AB}(\mathbf{k}) = P_M f_{\mathbf{k}}(\mathbf{k})W_{AB}, \tag*{(7)}

where PMP_M fixes the overall power, fk(k)f_{\mathbf{k}}(\mathbf{k}) describes the momentum distribution, and the Hermitian, positive-semidefinite matrix WABW_{AB} describes the relative powers and cross-correlations of the five polarizations. We normalize them by

∫d3k(2π)3fk(k)=1,Tr⁡W=1.(8)\int\frac{\mathrm{d}^3 k}{(2\pi)^3}f_{\mathbf{k}}(\mathbf{k}) = 1,\qquad\operatorname{Tr} W = 1. \tag*{(8)}

The diagonal entries of WW specify the polarization powers, while its off-diagonal entries describe the correlations between different polarizations. The polarization-equipartition case corresponds to WAB=δAB/5W_{AB} = \delta_{AB}/5.

Energy-Momentum Tensor and Frequency Components

The field sources gravity through its energy-momentum tensor, defined at quadratic order by the Hilbert prescription Tμν=−(2/−g)δSFP(2)/δgμνT_{\mu\nu} = -(2/\sqrt{-g})\delta S_{\mathrm{FP}}^{(2)}/\delta g^{\mu\nu} [24]. Using the free-field constraints and the nonrelativistic scaling ∂t∼m\partial_t \sim m, ∂i∼mv\partial_i \sim mv, the leading energy density is

T00=18(M˙ijM˙ij+∂kMij∂kMij+m2MijMij).(9)T_{00} = \frac{1}{8}\left(\dot{M}_{ij}\dot{M}_{ij}+\partial_kM_{ij}\partial_kM_{ij}+m^2M_{ij}M_{ij}\right). \tag*{(9)}

Averaging over the rapid oscillations and the stochastic ensemble gives the mean energy density,

ρˉ≡⟨T00⟩=12m2PM[1+O(v2)],(10)\bar{\rho} \equiv\langle T_{00}\rangle= \frac{1}{2}m^2P_M[1+\mathcal{O}(v^2)], \tag*{(10)}

where the overbar denotes the oscillation average. The leading spatial stress is

Tij=18δij(M˙kℓM˙kℓ−m2MkℓMkℓ)+12(m2MiℓMjℓ−M˙iℓM˙jℓ)+O(ρˉv2).(11)T_{ij} = \frac{1}{8}\delta_{ij}\left(\dot{M}_{k\ell}\dot{M}_{k\ell}-m^2M_{k\ell}M_{k\ell}\right)+\frac{1}{2}\left(m^2M_{i\ell}M_{j\ell}-\dot{M}_{i\ell}\dot{M}_{j\ell}\right)+\mathcal{O}(\bar{\rho}v^2). \tag*{(11)}

The momentum density satisfies T0i=O(ρˉv)T_{0i}=\mathcal{O}(\bar{\rho}v).

Since TμνT_{\mu\nu} is quadratic in MijM_{ij}, pairs of modes produce sum- and difference-frequency components. The former have ωk+ωk′=2m+O(mv2)\omega_{\mathbf{k}}+\omega_{\mathbf{k}'}=2m+\mathcal{O}(mv^2) (fast mode), while the latter have ωk−ωk′=O(mv2)\omega_{\mathbf{k}}-\omega_{\mathbf{k}'}=\mathcal{O}(mv^2) (slow mode) [9, 37, 39]. At leading nonrelativistic order, the sum-frequency terms survive in the oscillatory spatial stress, whereas the difference-frequency terms contribute to density fluctuations.

Substituting Eq. (5) into Eq. (11) and retaining the sum-frequency terms yields

TijF(t,x)=m2∑A,B∫d3k d3k′(2π)6MA(k)MB(k′)[12(EiℓAEjℓB+EjℓAEiℓB)−16δijδAB]e−i(ωk+ωk′)t+i(k+k′)⋅x+c.c.+O(v2ρˉ).(12)T_{ij}^{F}(t,\mathbf{x}) = m^2\sum_{A,B}\int\frac{\mathrm{d}^3k\,\mathrm{d}^3k'}{(2\pi)^6}M_A(\mathbf{k})M_B(\mathbf{k}')\left[\frac{1}{2}\left(E_{i\ell}^{A}E_{j\ell}^{B}+E_{j\ell}^{A}E_{i\ell}^{B}\right)-\frac{1}{6}\delta_{ij}\delta^{AB}\right]e^{-i(\omega_{\mathbf{k}}+\omega_{\mathbf{k}'})t+i(\mathbf{k}+\mathbf{k}')\cdot\mathbf{x}}+\mathrm{c.c.}+\mathcal{O}(v^2\bar{\rho}). \tag*{(12)}

The fast-mode stress therefore retains the polarization-dependent tensor structure required for the gravitational response.

The slow mode, corresponding to density fluctuations, produces a leading scalar gravitational response analogous to that of spin-0 and spin-1 ULDM [37, 39]. With a characteristic frequency ωS∼mv2\omega_{\mathrm{S}}\sim mv^2, it allows PTAs to probe higher ULDM masses, typically around 10−1810^{-18}–10−16 eV10^{-16}\,\mathrm{eV}. In the following, we focus on the fast-mode gravitational response.

Fast-Mode Metric Perturbations

We perturb the flat metric as gμν=ημν+hμνg_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}. The ULDM stress tensor sources the metric through the linearized Einstein equations,

δGμν[h]=8πGTμν.(13)\delta G_{\mu\nu}[h]=8\pi GT_{\mu\nu}. \tag*{(13)}

For the fast branch, we use synchronous gauge, h00=h0i=0h_{00}=h_{0i}=0, and decompose the spatial perturbation as hij=2sij−2Ψδijh_{ij}=2s_{ij}-2\Psi\delta_{ij}, where sii=0s_{ii}=0 [24, 39]. Using ∂i/∂t=O(v)\partial_i/\partial_t=O(v), the spatial Einstein tensor at leading order reads

δGij≃s¨ij+2δijΨ¨.(14)\delta G_{ij}\simeq\ddot{s}_{ij}+2\delta_{ij}\ddot{\Psi}. \tag*{(14)}

Decomposing the fast-mode stress into its trace and traceless parts, TijF=δijTkkF/3+πijFT_{ij}^{\mathrm{F}}=\delta_{ij}T_{kk}^{\mathrm{F}}/3+\pi_{ij}^{\mathrm{F}} with πiiF=0\pi_{ii}^{\mathrm{F}}=0, gives

Ψ¨F=4πG3TkkF,s¨ijF=8πGπijF.(15)\ddot{\Psi}^{\mathrm{F}}=\frac{4\pi G}{3}T_{kk}^{\mathrm{F}},\qquad\ddot{s}_{ij}^{\mathrm{F}}=8\pi G\pi_{ij}^{\mathrm{F}}. \tag*{(15)}

For the sourced fast-mode response, ∂t2≃−4m2\partial_t^2\simeq-4m^2 at leading order. Substituting Eq. (12) yields

hijF(t,x)≃−4πG∑A,B∫d3k d3k′(2π)6MA(k)MB(k′)RijABe−i(ωk+ωk′)t+i(k+k′)⋅x+c.c.,(16)h_{ij}^{\mathrm{F}}(t,\boldsymbol{x})\simeq-4\pi G\sum_{A,B}\int\frac{\mathrm{d}^3k\,\mathrm{d}^3k'}{(2\pi)^6}M_A(\boldsymbol{k})M_B(\boldsymbol{k}')\mathcal{R}_{ij}^{AB}e^{-i(\omega_k+\omega_{k'})t+i(\boldsymbol{k}+\boldsymbol{k}')\cdot\boldsymbol{x}}+\mathrm{c.c.}, \tag*{(16)}

where

RijAB=12(EiℓAEjℓB+EjℓAEiℓB)−38δijδAB.(17)\mathcal{R}_{ij}^{AB}=\frac{1}{2}\left(E_{i\ell}^{A}E_{j\ell}^{B}+E_{j\ell}^{A}E_{i\ell}^{B}\right)-\frac{3}{8}\delta_{ij}\delta^{AB}. \tag*{(17)}

At this order, RijAB\mathcal{R}_{ij}^{AB} contains the polarization geometry and is independent of the mode momenta.

For the subsequent timing analysis, it is useful to separate the carrier frequency 2m2m from the slowly varying envelope,

hijF(t,x)=Hij(t,x)e−i2mt+Hij∗(t,x)ei2mt,(18)h_{ij}^{\mathrm{F}}(t,\boldsymbol{x})=H_{ij}(t,\boldsymbol{x})e^{-i2mt}+H_{ij}^{*}(t,\boldsymbol{x})e^{i2mt}, \tag*{(18)}

with

Hij(t,x)=−4πG∑A,B∫d3k d3k′(2π)6MA(k)MB(k′)RijABe−iδωkk′t+i(k+k′)⋅x,(19)H_{ij}(t,\boldsymbol{x})=-4\pi G\sum_{A,B}\int\frac{\mathrm{d}^3k\,\mathrm{d}^3k'}{(2\pi)^6}M_A(\boldsymbol{k})M_B(\boldsymbol{k}')\mathcal{R}_{ij}^{AB}e^{-i\delta\omega_{\boldsymbol{k}\boldsymbol{k}'}t+i(\boldsymbol{k}+\boldsymbol{k}')\cdot\boldsymbol{x}}, \tag*{(19)}

where δωkk′≡ωk+ωk′−2m=O(mv2)\delta\omega_{\boldsymbol{k}\boldsymbol{k}'}\equiv\omega_k+\omega_{k'}-2m=O(mv^2). The envelope varies on the kinetic-frequency scale O(mv2)O(mv^2) but remains part of the fast branch, distinct from the slow mode sourced by difference-frequency products.

Pulsar Timing Response and Two-Point Statistics

For pulsar aa, let dad_a be its distance from the Earth and n^a\hat{\boldsymbol{n}}_a the unit vector pointing toward it. With the Earth at x=0\boldsymbol{x}=0, the pulsar is located at xa=dan^a\boldsymbol{x}_a=d_a\hat{\boldsymbol{n}}_a. At leading order in the nonrelativistic expansion, the fractional frequency shift of the radio pulses is [22, 24, 25, 39]

zaF(t)≃12n^ain^aj[hijF(t,0)−hijF(t−da,xa)],(20)z_a^{\mathrm{F}}(t)\simeq\frac{1}{2}\hat{n}_a^i\hat{n}_a^j\left[h_{ij}^{\mathrm{F}}(t,0)-h_{ij}^{\mathrm{F}}(t-d_a,\boldsymbol{x}_a)\right], \tag*{(20)}

where the two terms are the Earth and pulsar contributions. The timing residual is raF(t)=∫tdt1 zaF(t1)r_a^{\mathrm{F}}(t)=\int^t\mathrm{d}t_1\,z_a^{\mathrm{F}}(t_1), with the integration constant absorbed into the timing model. The time integration of the fast oscillation introduces a factor 1/(2m)1/(2m), up to relative O(v2)O(v^2) corrections. We now calculate the timing-residual cross-correlation between pulsars aa and bb, ⟨raF(t)rbF(t′)⟩\langle r_a^{\mathrm{F}}(t)r_b^{\mathrm{F}}(t')\rangle.

The fast-mode metric in Eq. (19) is quadratic in the stochastic field amplitudes, so its two-point function involves their four-point statistics. We assume Gaussian mode amplitudes and apply Wick’s theorem [38, 39]:

⟨MA(k)MB(k′)MC∗(p)MD∗(p′)⟩=⟨MA(k)MC∗(p)⟩⟨MB(k′)MD∗(p′)⟩+⟨MA(k)MD∗(p′)⟩⟨MB(k′)MC∗(p)⟩.(21)\left\langle M_A(\boldsymbol{k})M_B(\boldsymbol{k}')M_C^*(\boldsymbol{p})M_D^*(\boldsymbol{p}')\right\rangle=\left\langle M_A(\boldsymbol{k})M_C^*(\boldsymbol{p})\right\rangle\left\langle M_B(\boldsymbol{k}')M_D^*(\boldsymbol{p}')\right\rangle+\left\langle M_A(\boldsymbol{k})M_D^*(\boldsymbol{p}')\right\rangle\left\langle M_B(\boldsymbol{k}')M_C^*(\boldsymbol{p})\right\rangle. \tag*{(21)}

The pairing of two mode covariances in Eq. (6), together with the separable spectrum in Eq. (7), then gives the factorized envelope correlation

⟨Hij(t,x)Hkl∗(t′,x′)⟩=AHΓcoh(Δt,r)Kij,kl[W],(22)\left\langle H_{ij}(t,\boldsymbol{x})H_{kl}^*(t',\boldsymbol{x}')\right\rangle=A_H\Gamma_{\mathrm{coh}}(\Delta t,r)\mathcal{K}_{ij,kl}[W], \tag*{(22)}

where Δt≡t−t′\Delta t\equiv t-t', r≡x−x′r\equiv\boldsymbol{x}-\boldsymbol{x}', and

AH=(4πG)2ρM2=64π2G2ρˉ 2m4,(23)A_H=(4\pi G)^2\rho_M^2=\frac{64\pi^2G^2\bar{\rho}^{\,2}}{m^4}, \tag*{(23)}
Γcoh(Δt,r)=[∫d3k(2π)3fk(k)e−i(ωk−m)Δt+ik⋅r]2(24)\Gamma_{\mathrm{coh}}(\Delta t,\mathbf{r})=\left[\int\frac{\mathrm{d}^{3}k}{(2\pi)^{3}}f_k(\mathbf{k})e^{-i(\omega_k-m)\Delta t+i\mathbf{k}\cdot\mathbf{r}}\right]^2 \tag*{(24)}
Kij,kl[W]=∑A,B,C,DRijABRklCD(WACWBD+WADWBC).(25)\mathcal{K}_{ij,kl}[W]=\sum_{A,B,C,D}\mathcal{R}_{ij}^{AB}\mathcal{R}_{kl}^{CD}\left(W_{AC}W_{BD}+W_{AD}W_{BC}\right). \tag*{(25)}

Thus, Γcoh\Gamma_{\mathrm{coh}} describes the spacetime coherence of the fast-mode envelope, whereas Kij,kl[W]\mathcal{K}_{ij,kl}[W] contains the polarization statistics and response geometry. Their factorization follows from the separable momentum–polarization distribution and the momentum-independent response tensor at leading nonrelativistic order.

Using k=mv\mathbf{k}=m\mathbf{v} and ωk−m≃mv2/2\omega_k-m\simeq m v^2/2, we express the coherence kernel in terms of the normalized velocity distribution fv(v)=m3fk(mv)/(2π)3f_v(\mathbf{v})=m^3f_k(m\mathbf{v})/(2\pi)^3:

Γcoh(Δt,r)=[∫d3v fv(v)e−imv2Δt/2+imv⋅r]2.(26)\Gamma_{\mathrm{coh}}(\Delta t,\mathbf{r})=\left[\int\mathrm{d}^{3}v\,f_v(\mathbf{v})e^{-imv^2\Delta t/2+im\mathbf{v}\cdot\mathbf{r}}\right]^2. \tag*{(26)}

For a shifted Maxwellian velocity distribution,

fv(v)=1(2πv02)3/2exp⁡[−∣v−vˉ∣22v02],(27)f_v(\mathbf{v})=\frac{1}{(2\pi v_0^2)^{3/2}}\exp\left[-\frac{|\mathbf{v}-\bar{\mathbf{v}}|^2}{2v_0^2}\right], \tag*{(27)}

where vˉ\bar{\mathbf{v}} is the mean velocity and v0v_0 is the standard deviation of each Cartesian velocity component, following Refs. [38, 39]. The resulting coherence kernel is

Γcoh(Δt,r)=e−im∣vˉ∣2Δt+2imvˉ⋅r(1+iΔt/τcoh)3×exp⁡[−∣r−vˉΔt∣2/ℓcoh21+iΔt/τcoh],(28)\Gamma_{\mathrm{coh}}(\Delta t,\mathbf{r})=\frac{e^{-im|\bar{\mathbf{v}}|^2\Delta t+2im\bar{\mathbf{v}}\cdot\mathbf{r}}}{(1+i\Delta t/\tau_{\mathrm{coh}})^3}\times\exp\left[-\frac{|\mathbf{r}-\bar{\mathbf{v}}\Delta t|^2/\ell_{\mathrm{coh}}^2}{1+i\Delta t/\tau_{\mathrm{coh}}}\right], \tag*{(28)}

where ℓcoh=1/(mv0)\ell_{\mathrm{coh}}=1/(mv_0) and τcoh=1/(mv02)\tau_{\mathrm{coh}}=1/(mv_0^2). For the fast mode, the relevant time separations satisfy ∣Δt∣≪τcoh|\Delta t|\ll\tau_{\mathrm{coh}}, so the coherence kernel reduces to

Γcoh(Δt,r)≃exp⁡(−r2ℓcoh2+2imvˉ⋅r).(29)\Gamma_{\mathrm{coh}}(\Delta t,\mathbf{r})\simeq\exp\left(-\frac{r^2}{\ell_{\mathrm{coh}}^2}+2im\bar{\mathbf{v}}\cdot\mathbf{r}\right). \tag*{(29)}

Consequently, Re⁡[e−i2mΔtΓcoh]≃e−r2/ℓcoh2cos⁡[2m(Δt−vˉ⋅r)]\operatorname{Re}[e^{-i2m\Delta t}\Gamma_{\mathrm{coh}}]\simeq e^{-r^2/\ell_{\mathrm{coh}}^2}\cos[2m(\Delta t-\bar{\mathbf{v}}\cdot\mathbf{r})], in agreement with the spacetime dependence of the fast-mode correlations in Ref. [39]. In the zero-mean-velocity limit used in Ref. [38], the kernel reduces to the Gaussian spatial suppression e−r2/ℓcoh2e^{-r^2/\ell_{\mathrm{coh}}^2}.

Since ⟨HijHkl⟩=0\langle H_{ij}H_{kl}\rangle=0 under the phase-invariant field ensemble, the real fast-mode metric satisfies

⟨hijf(t,x)hklf(t′,x′)⟩=2AHRe⁡[e−i2mΔtΓcoh(Δt,r)Kij,kl[W]].(30)\left\langle h_{ij}^{\mathrm{f}}(t,\mathbf{x})h_{kl}^{\mathrm{f}}(t',\mathbf{x}')\right\rangle=2A_H\operatorname{Re}\left[e^{-i2m\Delta t}\Gamma_{\mathrm{coh}}(\Delta t,\mathbf{r})\mathcal{K}_{ij,kl}[W]\right]. \tag*{(30)}

Projecting this result onto the pulsar lines of sight, we define

Gab[W]≡n^ain^ajn^bkn^blKij,kl[W].(31)\mathcal{G}_{ab}[W]\equiv\hat{n}_a^i\hat{n}_a^j\hat{n}_b^k\hat{n}_b^l\mathcal{K}_{ij,kl}[W]. \tag*{(31)}

Using Eq. (20) and integrating the two redshifts over time gives the fast-mode timing-residual cross-correlation Cab(t,t′)≡⟨raf(t)rbf(t′)⟩C_{ab}(t,t')\equiv\langle r_a^{\mathrm{f}}(t)r_b^{\mathrm{f}}(t')\rangle:

Cab(t,t′)≃8π2G2ρˉ 2m6Re⁡{Gab[W]e−i2mΔt[Γcoh(Δt,0)−e−i2mdbΓcoh(Δt+db,−xb)−ei2mdaΓcoh(Δt−da,xa)+ei2m(da−db)Γcoh(Δt−da+db,xa−xb)]}.(32)\begin{aligned} C_{ab}(t,t')\simeq{}&\frac{8\pi^2G^2\bar{\rho}^{\,2}}{m^6}\operatorname{Re}\Bigl\{\mathcal{G}_{ab}[W]e^{-i2m\Delta t}\Bigl[\Gamma_{\mathrm{coh}}(\Delta t,\mathbf{0})-e^{-i2md_b}\Gamma_{\mathrm{coh}}(\Delta t+d_b,-\mathbf{x}_b)\\ &\qquad-e^{i2md_a}\Gamma_{\mathrm{coh}}(\Delta t-d_a,\mathbf{x}_a)+e^{i2m(d_a-d_b)}\Gamma_{\mathrm{coh}}(\Delta t-d_a+d_b,\mathbf{x}_a-\mathbf{x}_b)\Bigr]\Bigr\}. \tag*{(32)} \end{aligned}

The four terms in brackets are the Earth–Earth, Earth–pulsar, pulsar–Earth, and pulsar–pulsar contributions. Within our leading-order separable model, they share the same polarization-geometric factor Gab[W]\mathcal{G}_{ab}[W], while the endpoint phases and coherence kernel retain the dependence on pulsar distances and positions. The characteristic residual amplitude scales as rrms∼Gρˉ/m3r_{\mathrm{rms}}\sim G\bar{\rho}/m^3 [24, 39].

Equation (32) applies to the general polarization matrix WW introduced in § 2. We examine the polarization-equipartition case and its intrinsic angular kernel, including the comparison with spin-1 ULDM, in § 2.3.

Discussion

As a natural benchmark, we consider rotationally invariant polarization statistics for spin-2 ULDM. Within our separable model, rotational invariance requires the polarization covariance matrix to be proportional to the identity, corresponding to polarization equipartition among the five STF components. Such an assumption is also physically motivated by the nonlinear gravitational evolution and subsequent virialization of Galactic dark matter halos, which may erase the memory of the initial polarization distribution, as discussed for spin-1 ULDM [39]. With Tr⁡W=1\operatorname{Tr} W=1, we have

WAB=15δAB.(33)W_{AB}=\frac{1}{5}\delta_{AB}. \tag*{(33)}

Substituting this into Eq. (25) gives

Kij,kleq=αδijδkl+β(δikδjl+δilδjk),α=−1077200,β=7300.(34)\mathcal{K}^{\mathrm{eq}}_{ij,kl}=\alpha\delta_{ij}\delta_{kl}+\beta(\delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk}),\qquad\alpha=-\frac{107}{7200},\qquad\beta=\frac{7}{300}. \tag*{(34)}

The coefficients follow from the STF completeness relation, as detailed in Appendix § 2.1. With μab≡n^a⋅n^b=cos⁡ξab\mu_{ab}\equiv\hat{\mathbf{n}}_a\cdot\hat{\mathbf{n}}_b=\cos\xi_{ab}, the intrinsic polarization-geometric kernel defined in Eq. (31) becomes

Gab=−1077200+7150μab2=11440P0(μab)+7225P2(μab).(35)\mathcal{G}_{ab}=-\frac{107}{7200}+\frac{7}{150}\mu_{ab}^{2}=\frac{1}{1440}P_0(\mu_{ab})+\frac{7}{225}P_2(\mu_{ab}). \tag*{(35)}

where P0(μ)=1P_0(\mu)=1 and P2(μ)=(3μ2−1)/2P_2(\mu)=(3\mu^2-1)/2 are the monopole and quadrupole Legendre polynomials, respectively. Thus, under five-polarization equipartition, the intrinsic spin-2 angular kernel contains only monopole and quadrupole components, with the latter dominant.

To compare the intrinsic angular kernels of different spin sectors at fixed ULDM mass and mean density, we factor out the common dimensional amplitude π2G2ρˉ2/m6\pi^2G^2\bar{\rho}^2/m^6 from the fast-mode timing-residual covariance. The resulting dimensionless kernels for spin-0, spin-1, and spin-2 ULDM under polarization equipartition are

G^ab(0)=14,G^ab(1)=16μab2−536,G^ab(2)=336μab2−107900.(36)\widehat{\mathcal{G}}_{ab}^{(0)}=\frac{1}{4},\qquad\widehat{\mathcal{G}}_{ab}^{(1)}=\frac{16\mu_{ab}^{2}-5}{36},\qquad\widehat{\mathcal{G}}_{ab}^{(2)}=\frac{336\mu_{ab}^{2}-107}{900}. \tag*{(36)}

The spin-0 and spin-1 results are taken from Ref. [38] and [39], respectively, while the spin-2 result follows from Eq. (35).

As illustrated in Figure 1, the spin-0 kernel is independent of angular separation, whereas the spin-1 and spin-2 kernels exhibit similar quadrupole-dominated angular dependence. In particular, they satisfy

Figure 1. Dimensionless intrinsic angular kernels for spin-0 (green), spin-1 (orange), and spin-2 (blue) ULDM under polarization equipartition. The kernels are normalized by the common dimensional factor π2Gρ2ρˉ2/m6\pi^2 G_\rho^2 \bar{\rho}^2/m^6.

G^ab(2)=107125G^ab(1)−81125μab2.(37)\widehat{\mathcal{G}}_{ab}^{(2)}=\frac{107}{125}\widehat{\mathcal{G}}_{ab}^{(1)}-\frac{8}{1125}\mu_{ab}^{2}. \tag*{(37)}

The small correction to the rescaled spin-1 kernel highlights the close similarity of their angular profiles, despite their different overall amplitudes.

These results suggest that, if polarization equipartition is established during the evolution of Galactic dark matter halos, the similarity of these angular profiles may make spin-1 and spin-2 ULDM difficult to distinguish using PTA timing correlations, particularly since differences in their overall amplitudes can be absorbed into the inferred ULDM density.

For completeness, the slow-mode density-fluctuation correlations depend on the polarization matrix through Tr⁡(W2)\operatorname{Tr}(W^2) under the Gaussian and separable statistical model adopted here. Under polarization equipartition, the resulting slow-mode timing-residual correlations have relative amplitudes 11, 1/31/3, and 1/51/5 for spin-0, spin-1, and spin-2 ULDM, respectively, when compared at fixed mean density, mass, and velocity distribution. A short derivation is given in Appendix § 2.1.1.

Conclusion

We have investigated the stochastic gravitational timing signal induced by the fast mode of spin-2 ULDM. Modeling the nonrelativistic field as a stochastic superposition of five physical polarizations, we derived the metric perturbations and pulsar timing-residual correlations under Gaussian field statistics and a separable separable momentum-polarization power spectrum. The resulting covariance incorporates both Earth and pulsar terms, with the polarization geometry factorized from the spacetime coherence of the field. This framework extends previous coherent-field treatments by accounting for stochastic field statistics and finite spatial coherence.

Under five-polarization equipartition, the intrinsic angular kernel contains monopole and quadrupole components, with the quadrupole dominant. Comparing the spin-0, spin-1, and spin-2 angular kernels at fixed ULDM mass and mean density, we find that spin-1 and spin-2 ULDM exhibit remarkably similar angular profiles, despite their different overall amplitudes. This similarity is not a generic consequence of rotational invariance, which does not fix the relative weights of different multipoles. If polarization equipartition is established during the evolution of Galactic dark matter halos, these results suggest a potential degeneracy between the two ULDM candidates in PTA observations, as their amplitude differences can be absorbed into the inferred ULDM density.

For completeness, we have also examined the slow mode, whose leading gravitational response arises from density fluctuations. Under the Gaussian and separable statistical assumptions adopted here, its density correlations depend on the polarization covariance through Tr⁡(W2)\operatorname{Tr}(W^2). Under polarization equipartition, this yields relative slow-mode timing-correlation amplitudes of 1, 1/3, and 1/5 for spin-0, spin-1, and spin-2 ULDM, respectively.

Acknowledgments

YMW is supported by the National Natural Science Foundation of China (Grant No. 12505086). CL is supported by the National Natural Science Foundation of China (Grant No. 12405074).

References

A Spin-2 Polarization Algebra

For the five-polarization equipartition model in Sec. IV, the STF basis obeys the completeness relation

∑A=15EijAEklA=12(δikδjl+δilδjk)−13δijδkl.(38)\sum_{A=1}^{5} E^{A}_{ij}E^{A}_{kl}=\frac{1}{2}\left(\delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk}\right)-\frac{1}{3}\delta_{ij}\delta_{kl}. \tag*{(38)}

Using the response tensor RijAB\mathcal{R}^{AB}_{ij} in Eq. (17), Eq. (25) reduces to Kij,kleq=(2/25)∑A,BRijABRklAB\mathcal{K}^{\mathrm{eq}}_{ij,kl}=(2/25)\sum_{A,B}\mathcal{R}^{AB}_{ij}\mathcal{R}^{AB}_{kl}. The two independent contractions are

Kii,kkeq=1160,Kij,ijeq=113480.(39)\mathcal{K}^{\mathrm{eq}}_{ii,kk}=\frac{1}{160},\qquad\mathcal{K}^{\mathrm{eq}}_{ij,ij}=\frac{113}{480}. \tag*{(39)}

For the isotropic tensor decomposition Kij,kleq=αδijδkl+β(δikδjl+δilδjk)\mathcal{K}^{\mathrm{eq}}_{ij,kl}=\alpha\delta_{ij}\delta_{kl}+\beta(\delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk}), these contractions give 9α+6β=1/1609\alpha+6\beta=1/160 and 3α+12β=113/4803\alpha+12\beta=113/480, hence

α=−1077200,β=7300.(40)\alpha=-\frac{107}{7200},\qquad\beta=\frac{7}{300}. \tag*{(40)}

These coefficients determine the intrinsic spin-2 angular kernel in Eq. (35).

B Slow-Mode Density Correlations

The slow mode arises from difference-frequency products of the nonrelativistic spin-2 field, which generate density fluctuations and source a leading scalar gravitational response, as in spin-0 and spin-1 ULDM [37, 39]. Defining δρS=T00S−ρˉ\delta\rho^{\mathrm{S}}=T^{\mathrm{S}}_{00}-\bar{\rho}, we obtain

δρS(t,x)=m22∑A∫d3k d3k′(2π)6MA(k)MA∗(k′)e−i(ωk−ωk′)t+i(k−k′)⋅x−ρˉ.(41)\delta\rho^{\mathrm{S}}(t,\boldsymbol{x})=\frac{m^{2}}{2}\sum_{A}\int\frac{\mathrm{d}^{3}k\,\mathrm{d}^{3}k'}{(2\pi)^{6}}M_{A}(\boldsymbol{k})M_{A}^{*}(\boldsymbol{k}')e^{-i(\omega_{k}-\omega_{k'})t+i(\boldsymbol{k}-\boldsymbol{k}')\cdot\boldsymbol{x}}-\bar{\rho}. \tag*{(41)}

For the Gaussian field statistics and separable momentum–polarization spectrum introduced in Sec. II, Wick’s theorem gives the connected density correlation

⟨δρS(t,x)δρS(t′,x′)⟩c=ρˉ2Tr⁡(W2)FS(Δt,r),(42)\left\langle\delta\rho^{\mathrm{S}}(t,\boldsymbol{x})\delta\rho^{\mathrm{S}}(t',\boldsymbol{x}')\right\rangle_{\mathrm{c}}=\bar{\rho}^{2}\operatorname{Tr}(W^{2})F_{\mathrm{S}}(\Delta t,r), \tag*{(42)}

where Δt=t−t′\Delta t=t-t', r=x−x′r=\boldsymbol{x}-\boldsymbol{x}', and

FS(Δt,r)=∣∫d3k(2π)3fk(k)e−i(ωk−m)Δt+ik⋅r∣2.(43)F_{\mathrm{S}}(\Delta t,r)=\left|\int\frac{\mathrm{d}^{3}k}{(2\pi)^{3}}f_{k}(\boldsymbol{k})e^{-i(\omega_{k}-m)\Delta t+i\boldsymbol{k}\cdot\boldsymbol{r}}\right|^{2}. \tag*{(43)}

Thus, the polarization dependence enters through Tr⁡(W2)\operatorname{Tr}(W^{2}), while the spacetime dependence is determined by the momentum distribution.

To obtain the gravitational response, we use the Newtonian-gauge metric

ds2=−(1+2ΦS)dt2+(1−2ΨS)δijdxidxj,(44)\mathrm{d}s^{2}=-(1+2\Phi^{\mathrm{S}})\mathrm{d}t^{2}+(1-2\Psi^{\mathrm{S}})\delta_{ij}\mathrm{d}x^{i}\mathrm{d}x^{j}, \tag*{(44)}

where ΦS\Phi^{\mathrm{S}} and ΨS\Psi^{\mathrm{S}} are the scalar metric potentials induced by the slow-mode density fluctuations. At leading nonrelativistic order, the linearized Einstein equations give

∇2ΨS=4πG δρS,ΦS≃ΨS.(45)\nabla^{2}\Psi^{\mathrm{S}}=4\pi G\,\delta\rho^{\mathrm{S}},\qquad\Phi^{\mathrm{S}}\simeq\Psi^{\mathrm{S}}. \tag*{(45)}

The gravitational acceleration of a nonrelativistic body is v˙i≃−∂iΦS\dot{v}^{i}\simeq-\partial_{i}\Phi^{\mathrm{S}}. For the slow mode, the leading pulsar redshift is dominated by the relative motion of the Earth and pulsar induced by this acceleration [37, 39]. Using the pulsar geometry defined in Sec. III, we obtain

zaS(t)≃n^ai∫t0tdt1[∂iΦS(t1,0)−∂iΦS(t1−da,xa)].(46)z^{\mathrm{S}}_{a}(t)\simeq\hat{n}^{i}_{a}\int_{t_{0}}^{t}\mathrm{d}t_{1}\left[\partial_{i}\Phi^{\mathrm{S}}(t_{1},0)-\partial_{i}\Phi^{\mathrm{S}}(t_{1}-d_{a},\boldsymbol{x}_{a})\right]. \tag*{(46)}

up to constant velocity contributions absorbed by the timing model. The corresponding timing residual is raS(t)=∫tdt1 zaS(t1)r_a^{\mathrm{S}}(t) = \int^{t} \mathrm{d}t_1\,z_a^{\mathrm{S}}(t_1).

Since the potential, redshift, and timing residual are linear functionals of δρS\delta\rho^{\mathrm{S}} at this order, their connected two-point functions inherit the polarization factor Tr⁡(W2)\operatorname{Tr}(W^2). For NN polarization modes under equipartition, WAB=δAB/NW_{AB} = \delta_{AB}/N, giving Tr⁡(W2)=1/N\operatorname{Tr}(W^2) = 1/N. Consequently, at fixed ULDM mass, mean density, and velocity distribution, and with the same leading-order timing-response prescription,

Cab,spin−0S:Cab,spin−1S:Cab,spin−2S=1:13:15.(47)C^{\mathrm{S}}_{ab,\mathrm{spin-0}} : C^{\mathrm{S}}_{ab,\mathrm{spin-1}} : C^{\mathrm{S}}_{ab,\mathrm{spin-2}} = 1 : \frac{1}{3} : \frac{1}{5}. \tag*{(47)}

For general polarization statistics, the corresponding factor is Tr⁡(W2)\operatorname{Tr}(W^2) rather than 1/51/5.

References

  1. [1]N. Aghanim et al. (Planck), Planck 2018 results. VI. cosmological parameters, Astron. Astrophys. 641, A6 (2020), [Erratum: Astron. Astrophys. 652, C4 (2021)], arXiv:1807.06209 [astro-ph.CO].DOI
  2. [2]G. Bertone, D. Hooper, and J. Silk, Particle dark matter: Evidence, candidates and constraints, Phys. Rept. 405, 279 (2005), arXiv:hep-ph/0404175.DOI
  3. [3]V. Springel, C. S. Frenk, and S. D. M. White, The large-scale structure of the universe, Nature 440, 1137 (2006), arXiv:astro-ph/0604561 [astro-ph].DOI
  4. [4]J. S. Bullock and M. Boylan-Kolchin, Small-scale challenges to the ΛCDM paradigm, Annual Review of Astronomy and Astrophysics 55, 343 (2017), arXiv:1707.02456 [astro-ph.CO].
  5. [5]W. Hu, R. Barkana, and A. Gruzinov, Fuzzy cold dark matter: The wave properties of ultralight particles, Phys. Rev. Lett. 85, 1158 (2000), arXiv:astro-ph/0003365.
  6. [6]L. Hui, J. P. Ostriker, S. Tremaine, and E. Witten, Ultralight scalars as cosmological dark matter, Phys. Rev. D 95, 043541 (2017), arXiv:1610.08297 [astro-ph.CO].DOI
  7. [7]E. G. M. Ferreira, Ultra-light dark matter, Astron. Astrophys. Rev. 29, 7 (2021), arXiv:2005.03254 [astro-ph.CO].DOI
  8. [8]D. Y. Cheong, N. L. Rodd, and L.-T. Wang, Quantum description of wave dark matter, Phys. Rev. D 111, 015028 (2025), arXiv:2408.04696 [hep-ph].DOI
  9. [9]A. Khmelnitsky and V. Rubakov, Pulsar timing signal from ultralight scalar dark matter, JCAP 2014 (02), 019, arXiv:1309.5888 [astro-ph.CO].DOI
  10. [10]A. Aoki and J. Soda, Detecting ultralight axion dark matter wind with laser interferometers, Int. J. Mod. Phys. D 26, 1750063 (2017), arXiv:1608.05933.DOI
  11. [11]J.-C. Yu, Y. Cao, Y. Tang, and Y.-L. Wu, Detecting ultralight dark matter gravitationally with laser interferometers in space, Phys. Rev. D 110, 023025 (2024), arXiv:2404.04333 [hep-ph].DOI
  12. [12]J.-R. Zhang, J. Chen, H.-S. Jiao, R.-G. Cai, and Y.-L. Zhang, Probing spin-2 ultralight dark matter with space-based gravitational wave detectors in the MHz regime, Phys. Rev. D 112, 064030 (2025), arXiv:2501.11071 [gr-qc].DOI
  13. [13]J. A. Dror and S. Verner, New method for the astrometric direct detection of ultralight dark matter, Phys. Rev. Lett. 134, 111003 (2025), arXiv:2406.03526 [hep-ph].
  14. [14]H. Kim, Astrometric search for ultralight dark matter, Phys. Rev. D 110, 083031 (2024), arXiv:2406.03539.arxiv.org/abs/2406.03539
  15. [15]P. Brax, C. Burrage, J. A. R. Cembranos, and P. Valeḡas, Detecting dark matter oscillations with gravitational waveforms, Phys. Rev. D 110, 083515 (2024), arXiv:2402.04819.DOI
  16. [16]D. Blas, S. Gasparotto, and R. Vicente, Searching for ultralight dark matter through frequency modulation of gravitational waves, Phys. Rev. D 111, 042008 (2025), arXiv:2410.07330.arxiv.org/abs/2410.07330
  17. [17]R. S. Foster and D. C. Backer, Constructing a pulsar timing array, Astrophys. J. 361, 300 (1990).
  18. [18]S. Burke-Spolaor, S. R. Taylor, M. Charisi, T. Dolch, J. S. Hazboun, A. M. Holgado, L. Z. Kelley, T. J. W. Lazio, D. R. Madison, N. McMann, C. M. F. Mingarelli, A. Rasskazov, X. Siemens, J. J. Simon, and T. L. Smith, The astrophysics of nanohertz gravitational waves, Astron. Astrophys. Rev. 27, 5 (2019), arXiv:1811.08826 [astro-ph.HE].
  19. [19]N. K. Porayko, X. Zhu, Y. Levin, L. Hui, G. Hobbs, A. Grudskaya, K. Postnov, et al., Parkes pulsar timing array constraints on ultralight scalar-field dark matter, Phys. Rev. D 98, 102002 (2018), arXiv:1810.03227 [astro-ph.CO].DOI
  20. [20]Y.-M. Wu, Z.-C. Chen, Q.-G. Huang, X. Zhu, N. D. R. Bhat, Y. Feng, G. Hobbs, R. N. Manchester, C. J. Russell, and R. M. Shannon (PPTA), Constraining ultralight vector dark matter with the Parkes Pulsar Timing Array second data release, Phys. Rev. D 106, L081101 (2022), arXiv:2210.03880 [astro-ph.CO].DOI
  21. [21]C. Smarra et al. (European Pulsar Timing Array), Second data release from the European Pulsar Timing Array: Challenging the ultralight dark matter paradigm, Phys. Rev. Lett. 131, 171001 (2023), arXiv:2306.16228 [astro-ph.HE].DOI
  22. [22]S. L. Detweiler, Pulsar timing measurements and the search for gravitational waves, Astrophys. J. 234, 1100 (1979).DOI
  23. [23]S. Sun, X.-Y. Yang, and Y.-L. Zhang, Pulsar timing residual induced by wideband ultralight dark matter with spin 0, 1, 2, Phys. Rev. D 106, 066006 (2022), arXiv:2112.15593 [astro-ph.CO].DOI
  24. [24]Y.-M. Wu, Z.-C. Chen, and Q.-G. Huang, Pulsar timing residual induced by ultralight tensor dark matter, JCAP 2023 (09), 021, arXiv:2305.08091 [hep-ph].DOI
  25. [25]K. Nomura, A. Ito, and J. Soda, Pulsar timing residual induced by ultralight vector dark matter, Eur. Phys. J. C 80, 419 (2020), arXiv:1912.10210 [gr-qc].DOI
  26. [26]C. Ünal, F. R. Urban, and E. D. Kovetz, Probing ultralight scalar, vector and tensor dark matter with pulsar timing arrays, Phys. Lett. B 855, 138830 (2024), arXiv:2209.02741.DOI
  27. [27]E. Babichev, L. Marzola, M. Raidal, A. Schmidt-May, F. Urban, H. Veermäe, and M. von Strauss, Bigravitational origin of dark matter, Phys. Rev. D 94, 084055 (2016), arXiv:1604.08564 [hep-ph].DOI
  28. [28]E. Babichev, L. Marzola, M. Raidal, A. Schmidt-May, F. Urban, H. Veermäe, and M. von Strauss, Heavy spin-2 Dark Matter, JCAP 09, 016, arXiv:1607.03497 [hep-th].DOI
  29. [29]S. F. Hassan and R. A. Rosen, Bimetric Gravity from Ghost-free Massive Gravity, JHEP 02, 126, arXiv:1109.3515 [hep-th].
  30. [30]L. Marzola, M. Raidal, and F. R. Urban, Oscillating spin-2 dark matter, Phys. Rev. D 97, 024010 (2018), arXiv:1708.04253 [hep-ph].DOI
  31. [31]H. Omiya, K. Nomura, and J. Soda, Hellings-Downs curve deformed by ultralight vector dark matter, Phys. Rev. D 108, 104006 (2023), arXiv:2307.12624 [astro-ph.CO].arxiv.org/abs/2307.12624
  32. [32]K. Nomura, H. Omiya, and T. Tanaka, Signature of polarized ultralight vector dark matter in pulsar timing arrays, Phys. Rev. D 112, 123524 (2025), arXiv:2508.07358 [astro-ph.CO].DOI
  33. [33]C. Smarra, A. Kuntz, E. Barausse, B. Goncharov, D. López Nacir, D. Blas, L. Shao, J. Antoniadis, D. J. Champion, I. Cognard, L. Guillemot, H. Hu, M. Keith, M. Kramer, K. Liu, D. Perrodin, S. A. Sanidas, and G. Theureau, Constraints on conformal ultralight dark matter couplings from the European Pulsar Timing Array, Phys. Rev. D 110, 043033 (2024), arXiv:2405.01633 [astro-ph.HE].DOI
  34. [34]Y.-M. Wu and Q.-G. Huang, Constraining ultralight scalar dark matter couplings with the European Pulsar Timing Array second data release, Phys. Rev. D 111, 063032 (2025), arXiv:2411.02915 [astro-ph.CO].arxiv.org/abs/2411.02915
  35. [35]Z.-Q. Xia, T.-P. Tang, X. Huang, Q. Yuan, and Y.-Z. Fan, Constraining ultralight dark matter using the Fermi-LAT pulsar timing array, Phys. Rev. D 107, L121302 (2023), arXiv:2303.17545 [astro-ph.HE].DOI
  36. [36]R.-G. Cai, J.-R. Zhang, and Y.-L. Zhang, Angular correlation and deformed Hellings–Downs curve from spin-2 ultralight dark matter, Phys. Rev. D 110, 044052 (2024), arXiv:2402.03984 [gr-qc].DOI
  37. [37]H. Kim and A. Mitridate, Stochastic ultralight dark matter fluctuations in pulsar timing arrays, Phys. Rev. D 109, 055017 (2024), arXiv:2312.12225 [hep-ph].arxiv.org/abs/2312.12225
  38. [38]K. K. Boddy, J. A. Dror, and A. Lam, Ultralight dark matter statistics for pulsar timing detection, Phys. Rev. Lett. 135, 101001 (2025), arXiv:2502.15874.DOI
  39. [39]J. A. Dror and Q. Wei, Pulsar timing detection of ultralight vector dark matter, Phys. Rev. D 112, 075024 (2025), arXiv:2505.22719.DOI
  40. [40]J. W. Foster, K. K. Boddy, J. A. Dror, V. S. H. Lee, A. Mitridate, T. L. Smith, K. Taylor, and T. Trickle, Correlated signals of ultralight scalar dark matter in pulsar timing (2026), arXiv:2607.24912 [astro-ph.CO].arxiv.org/abs/2607.24912
  41. [41]M. Fierz and W. Pauli, On relativistic wave equations for particles of arbitrary spin in an electromagnetic field, Proc. Roy. Soc. Lond. A 173, 211 (1939).DOI
  42. [42]M. A. Alonso and E. Wolf, The cross-spectral density matrix of a planar, electromagnetic stochastic source as a correlation matrix, Opt. Commun. 281, 2393 (2008).
  43. [43]T. Kurita and M. Takada, Analysis method for 3d power spectrum of projected tensor fields with fast estimator and window convolution modeling: An application to intrinsic alignments, Phys. Rev. D 105, 123501 (2022), arXiv:2202.11839 [astro-ph.CO].DOI

Paper details

vlnc:W_bnzb5kqu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents