Number Count Dipole in Cosmology

• The number count dipole is the ℓ=1 anisotropy in extragalactic source counts, reflecting both observer motion and contributions from large-scale structure.
• Empirical measurements from radio and quasar surveys often show an amplitude exceeding kinematic expectations, highlighting potential systematic effects.
• Various methods and physical models, including modifications to gravity and intrinsic CMB contributions, are explored to resolve the observed dipole anomaly.

The number count dipole is the $\ell=1$ anisotropy in the angular distribution of extragalactic sources on the sky, inferred from source counts rather than from temperature or intensity. In cosmology it is primarily used to test whether the matter rest frame traced by galaxies, quasars, or radio sources is consistent with the rest frame defined by the cosmic microwave background (CMB). In the standard kinematic interpretation, the dipole is generated by the observer’s motion through an otherwise statistically isotropic source distribution, but the measured signal can also contain contributions from large-scale structure, survey geometry, source evolution, and estimator-dependent systematics (Wagenveld et al., 4 Mar 2025, Dam et al., 2022).

1. Definition and kinematic framework

A number count dipole is usually written as a dipolar modulation of the projected counts,

$$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$

or, in harmonic language, through the dipole power relation

$C_1=\frac{4\pi}{9}D^2,$

with $D = |\mathbf d|$ (Tiwari et al., 2015). For flux-limited catalogues, the standard Ellis–Baldwin form of the kinematic dipole is

$$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$

when the source spectrum is parameterized as $S(\nu)\propto \nu^{-\alpha}$ (Wagenveld et al., 4 Mar 2025). Other papers write

$\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$

when the spectral index is defined instead by $S_\nu\propto \nu^\alpha$ (Böhme et al., 20 Sep 2025). This difference reflects a sign convention for $\alpha$, not a disagreement about the underlying kinematic mechanism.

The physical origin of the coefficient is standard across the literature. The term $2$ comes from aberration, while the source-count term depends on how many objects are shifted across the survey flux threshold by Doppler boosting. In several analyses the observed dipole is decomposed as

$$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$0

where $$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$1 is the contribution from large-scale structure (Moffat, 12 Jan 2026). A recurrent misconception is to treat the observed number count dipole as a direct velocity measurement alone; the cited decomposition makes clear that it is instead an observable sensitive to both motion and structure.

2. Empirical measurements and the amplitude anomaly

Wide-area surveys have produced a long sequence of dipole measurements, especially in radio continuum and quasar catalogues. A central empirical pattern is that the direction of the measured dipole is often broadly consistent with the CMB dipole direction, while the amplitude is frequently larger than the kinematic expectation inferred from the CMB.

Survey or tracer	Reported result	Brief note
NVSS, $$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$2 mJy	$$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$3	Roughly $$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$4 times the CMB speed; direction approximately agrees with the CMB (Tiwari et al., 2013)
NVSUMSS	$$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$5 hemispherical; $$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$6 3D estimator	About $$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$7 times larger than the CMB value; significance up to $$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$8 (Colin et al., 2017)
CatWISE quasars	$$\frac{\Delta N}{N} = \mathbf d \cdot \hat{\mathbf n},$$9	Direction consistent with the CMB; amplitude anomaly quoted at $C_1=\frac{4\pi}{9}D^2,$0 (Dam et al., 2022)
NVSS + RACS-low + LoTSS-DR2	$C_1=\frac{4\pi}{9}D^2,$1	Excess over the kinematic expectation at $C_1=\frac{4\pi}{9}D^2,$2 (Böhme et al., 20 Sep 2025)

The CMB benchmark used in these comparisons is the Solar System peculiar velocity inferred from the CMB dipole, quoted as

$C_1=\frac{4\pi}{9}D^2,$3

in the radio-dipole literature (Moffat, 12 Jan 2026). If the radio excess were interpreted as purely kinematic, the implied effective velocity would be roughly

$C_1=\frac{4\pi}{9}D^2,$4

which several papers describe as difficult to reconcile with standard $C_1=\frac{4\pi}{9}D^2,$5CDM bulk flows (Moffat, 12 Jan 2026).

Not all dipole observables point in the same direction. A blind tomographic analysis of BOSS and eBOSS spectroscopic data over $C_1=\frac{4\pi}{9}D^2,$6 measured an eBOSS redshift dipole corresponding to

$C_1=\frac{4\pi}{9}D^2,$7

which is in $C_1=\frac{4\pi}{9}D^2,$8 agreement with the CMB dipole and in $C_1=\frac{4\pi}{9}D^2,$9-to-$D = |\mathbf d|$0 tension with previous number count studies (Ferreira et al., 2024). This contrast is central to the contemporary debate: the anomaly is robust in some catalogues and methods, but not universal across all tracers and observables.

3. Structure contribution, scale dependence, and redshift dependence

The intrinsic structure contribution is survey-dependent. A realistic $D = |\mathbf d|$1CDM analysis of the NVSS dipole found that, after removing the solar-motion contribution, about 70% of the residual dipole arises from structures within

$D = |\mathbf d|$2

and that the signal is mainly sensitive to scales around

$D = |\mathbf d|$3

The predicted amplitude depends strongly on the radio-galaxy bias. For $D = |\mathbf d|$4 mJy, the standard bias model is consistent with the observed dipole at $D = |\mathbf d|$5, whereas flux thresholds above $D = |\mathbf d|$6 mJy produce tensions close to $D = |\mathbf d|$7; adopting a constant high bias $D = |\mathbf d|$8 reduces the discrepancy to about $D = |\mathbf d|$9 (Tiwari et al., 2015).

Tomographic analyses strengthen the scale argument. In the WISE $$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$0 SuperCOSMOS catalogue, the dipole amplitude decreases with depth, and the strongest anomaly occurs in the lowest redshift shell $$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$1. Above $$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$2, most bins are broadly consistent with $$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$3CDM mock catalogues, especially for the Fiducial sample, while the lowest-redshift shell remains persistently anomalous in both the Fiducial and SVM selections (Bengaly et al., 2017). This supports the view that nearby structure and low-redshift systematics remain difficult to disentangle.

At the same time, the radio-dipole literature emphasizes that the observable is not simply a local-structure statistic. In the modified-gravity interpretation, the radio number-count dipole is described as an infrared-dominated projection of matter and velocity fields, sensitive to ultra-large wavelengths from hundreds of Mpc to gigaparsec scales, with

$$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$4

as a typical standard-model expectation (Moffat, 12 Jan 2026). By contrast, the eBOSS redshift-dipole analysis estimated the intrinsic clustering dipole for its spectroscopic samples to be negligible, less than about $$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$5 of the expected kinematic effect (Ferreira et al., 2024). This suggests that “number count dipole” is not a single-scale observable: the effective redshift kernel, source population, and estimator determine whether local structure, ultra-large-scale modes, or kinematics dominate.

4. Estimation, masking, and statistical systematics

Dipole estimation has been carried out with hemispherical asymmetry measures, 3D linear estimators, spherical-harmonic methods, tomographic least-squares fits, and pixel-based Bayesian count models. The NVSUMSS analysis used both hemispherical number-count asymmetry and a 3D linear estimator; WISE $$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$6 SuperCOSMOS used the delta-map method; CatWISE analyses used a Poisson likelihood for pixel counts with explicit nuisance modeling of ecliptic-dependent modulation (Colin et al., 2017, Bengaly et al., 2017, Dam et al., 2022).

Survey realism matters quantitatively. In CatWISE, a Bayesian reanalysis found that the Galactic plane mask causes a considerable loss of dipole signal through leakage of power into higher multipoles, which exacerbates rather than alleviates the amplitude discrepancy (Dam et al., 2022). In radio surveys, many sources are multi-component objects, so counts-in-cells are often overdispersed relative to a Poisson model. A Bayesian estimator based on the negative binomial distribution was introduced precisely to capture this effect, with the observed source count in cell $$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$7 written as

$$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$8

Using NVSS, RACS-low, and LoTSS-DR2, this treatment still yielded

$$\mathcal D_{\rm kin} = [2+x(1+\alpha)]\beta, \qquad \beta \equiv \frac{v}{c},$$9

but with larger and arguably more realistic uncertainties than under a Poisson assumption (Böhme et al., 20 Sep 2025).

Methodological comparisons have also shown that estimator properties depend strongly on sky coverage. A comparative study of healpy.fit_dipole and Bayesian inference found that the estimator is unbiased in Cartesian components regardless of noise level and sky coverage, but develops large variance at low sky coverage; Bayesian analysis in the same regime is often inconclusive, which the authors present as a safeguard against overinterpretation (Mittal et al., 25 Sep 2025). A separate CatWISE analysis replaced the standard mean-$S(\nu)\propto \nu^{-\alpha}$0, power-law-$S(\nu)\propto \nu^{-\alpha}$1 procedure by a source-by-source forward simulation using each source’s own spectral index, finding a velocity around

$S(\nu)\propto \nu^{-\alpha}$2

instead of the standard

$S(\nu)\propto \nu^{-\alpha}$3

a shift of approximately one sigma (Panwar et al., 7 May 2025). Method dependence is therefore not incidental; it is part of the empirical status of the anomaly.

5. Physical interpretations and proposed resolutions

One approach is phenomenological decomposition. A joint fit to NVSS, RACS-low, and CatWISE modeled the catalogue-dependent dipole as

$S(\nu)\propto \nu^{-\alpha}$4

Under a purely kinematic hypothesis, the combined fit gave

$S(\nu)\propto \nu^{-\alpha}$5

equivalent to

$S(\nu)\propto \nu^{-\alpha}$6

with a common direction offset from the CMB dipole by

$S(\nu)\propto \nu^{-\alpha}$7

When the kinematic component was fixed to the CMB expectation and a residual component was allowed, the inferred residual dipole was

$S(\nu)\propto \nu^{-\alpha}$8

offset from the CMB dipole direction by

$S(\nu)\propto \nu^{-\alpha}$9

(Wagenveld et al., 4 Mar 2025). In this framework, the anomaly is recast as evidence for a non-kinematic contribution shared across catalogues.

A second class of explanations moves the discrepancy to the CMB side rather than the matter side. A detailed analysis of superhorizon perturbations concluded that neither adiabatic nor isocurvature superhorizon modes can generate an intrinsic galaxy number-count dipole at observable redshift, although a superhorizon isocurvature mode can induce a matter–radiation relative velocity and thereby modify the CMB dipole (Domènech et al., 2022). A QCD axion model then developed this logic further: the CMB dipole is written as

$\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$0

with the intrinsic term generated by a super-horizon isocurvature perturbation. In that scenario the galaxy number-count dipole remains very small, of order

$\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$1

while the CMB dipole can receive an intrinsic contribution large enough to address the mismatch (Han, 2022). These proposals do not explain the anomaly as an intrinsic number-count dipole; they reinterpret the CMB dipole as not being purely kinematic.

A third line of work modifies the late-time growth of structure. The STVG-MOG proposal keeps early-universe and small-scale behavior close to standard cosmology but introduces a scale-dependent effective coupling

$\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$2

In this model, gravity becomes effectively Newtonian on small scales and enhanced on ultra-large scales. Because the velocity field contains an explicit $\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$3 weighting,

$\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$4

even modest large-scale growth enhancement can strongly amplify coherent bulk flows and dipole anisotropies. The dipole enhancement is summarized schematically by

$\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$5

and for $\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$6 the model is argued to permit a factor-of-few enhancement, potentially enough to turn a $\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$7CDM-like dipole into the observed $\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$8–$\mathbf d = (2+x[1-\alpha])\frac{\mathbf v}{c}$9 excess (Moffat, 12 Jan 2026). This interpretation treats the radio dipole anomaly as a physical late-time structure signal rather than as an anomalous Solar System velocity.

6. Extensions, related observables, and future probes

The number count dipole is no longer restricted to broadband continuum catalogues. For line-emitting galaxies, the cumulative number density

$S_\nu\propto \nu^\alpha$0

has a dipole

$S_\nu\propto \nu^\alpha$1

Here the amplitude depends not only on $S_\nu\propto \nu^\alpha$2 but also on the frequency evolution of the monopole counts, so different observing frequencies provide independent redshift slices of the same kinematic signal. On this basis, SPHEREx was identified as the optimal near-term survey for a first detection of the spectral line galaxy number-count dipole, while line-intensity mapping surveys were also found promising (Ahn, 16 Jan 2025).

Several related observables probe similar physics but are not identical to the monopole number-count dipole. In the galaxy bispectrum, the Newtonian single-tracer signal has no dipole, whereas relativistic light-cone corrections generate odd multipoles, including a bispectrum dipole that can exceed $S_\nu\propto \nu^\alpha$3 of the monopole on sufficiently large scales (Clarkson et al., 2018). In the two-point function of number-count fluctuations, the Doppler imprint on the covariance matrix can be described with bipolar spherical harmonics; forecasts indicate that radio continuum surveys with SKA Phase 2 should detect this effect at $S_\nu\propto \nu^\alpha$4, while SKA Phase 1 allows only marginal detection in optimistic cases (Pant et al., 2018).

Independent dipole probes are also being developed outside electromagnetic number counts. For binary black hole detections with third-generation gravitational-wave observatories, the number-count dipole simplifies to

$S_\nu\propto \nu^\alpha$5

when threshold effects are negligible, making the signal essentially independent of astrophysical population details. Forecasts indicate that a dipole as large as the radio-galaxy value would be detectable at $S_\nu\propto \nu^\alpha$6 with about $S_\nu\propto \nu^\alpha$7 events, while a CMB-sized dipole would require at least $S_\nu\propto \nu^\alpha$8 detections (Mastrogiovanni et al., 2022). Strongly lensed gravitational waves combined with galaxy surveys provide a second route: in an optimistic 10-year scenario, the combined constraint was forecast as

$S_\nu\propto \nu^\alpha$9

provided systematic uncertainties are mitigated (Chen et al., 19 May 2026). In parallel, the modified-gravity interpretation explicitly identifies future SKA measurements as a potential discriminator between standard and scale-enhanced structure-growth pictures (Moffat, 12 Jan 2026).

Taken together, the literature presents the number count dipole as both a kinematic observable and a precision test of large-scale isotropy. Its current significance lies less in the existence of a dipole itself—which is expected—and more in the unsettled question of whether the excess amplitude reported in radio and quasar catalogues is due to local structure, survey methodology, a residual non-kinematic anisotropy, an intrinsic CMB component, or new late-time gravitational physics.