ZTF Dipole Investigation of Cosmic Expansion

• The paper presents a robust framework using ZTF Type Ia supernovae to detect a velocity dipole in cosmic expansion and test isotropy.
• It compares three dipole injection schemes—d_L, z, and m_B—with the m_B method proving both efficient and unbiased in recovering ΔH₀ and its sky direction.
• Rigorous statistical and systematic error modeling confirms the pipeline’s potential for high-precision constraints on H₀ anisotropy.

The ZTF Dipole Investigation encompasses the methodological and statistical framework for testing cosmic expansion isotropy through Type Ia supernovae (SNe Ia) observed by the Zwicky Transient Facility (ZTF). It is motivated by growing interest in potential large-scale anisotropies that would violate the cosmological principle, with the specific focus on directional variations in the Hubble constant, $H_0$. Using ZTF’s wide sky coverage and realistic SN simulations, the approach develops, injects, and detects a velocity dipole in the expansion rate, quantifying both amplitude and sky direction, while rigorously controlling for systematics linked to cadence, calibration, and peculiar velocities (Barjou-Delayre et al., 16 Jan 2026).

1. Theoretical Framework for Hubble Flow Anisotropy

In this framework, the local Hubble constant is modeled as a directional quantity:

$H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$

where $H_0$ is the isotropic (monopole) value, $ΔH_0$ the dipole amplitude, $n̂$ the supernova’s sky direction, and $p̂$ the dipole’s maximum. In equatorial coordinates $(α, δ)$,

$$\cos Δθ = \sin δ \sin δ_0 + \cos δ \cos δ_0 \cos(α - α_0)$$

so that

$H_0'(α, δ) = H_0 + ΔH_0 \cos Δθ.$

The luminosity distance in flat $\Lambda$CDM incorporating this anisotropy is:

$H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$0

with the corresponding distance modulus:

$H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$1

This dipole parameterization provides a rigorous test of the cosmological principle at scales probed by ZTF SNe Ia.

2. Dipole Injection in ZTF Supernova Simulations

Mock ZTF catalogs are synthesized using the skysurvey package and SALT2.4 light-curve model, emulating the cadence, depth, and spectroscopic selection of ZTF’s DR2.5. Two main simulation sets are created: "unclustered" (SNe Ia placed isotropically, no large-scale structure) and "clustered" (positions and peculiar velocities from the Uchuu N-body halo catalog, with 27 sub-boxes to z_max ≈ 0.11 and random Gaussian velocities with σ=250 km/s beyond).

Three dipole injection schemes are compared:

• d_L–method: Recalculates $H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$2 using $H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$3 in the integral (computationally exact, but impractical for multiple directions).
• z–method: Replaces $H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$4 with $H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$5 from cz=H₀d (simple but produces a ≥6% bias even at $H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$6).
• m_B–method: Shifts the simulated apparent magnitude $H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$7 post hoc:

$H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$8

This is algebraically equivalent to shifting $H_0'(n̂) = H_0 + ΔH_0 (n̂·p̂)$9 and remains unbiased at <0.5% while being highly efficient.

Subsequent analyses adopt the m_B–method with a reference injected $H_0$0 km s⁻¹ Mpc⁻¹ and various dipole orientations.

3. Dipole Recovery: Fitting Procedures and Likelihood Construction

Recovered distances are standardized via the Tripp relation:

$H_0$1

where $H_0$2 are SALT2 parameters, and $H_0$3 are nuisance parameters.

The dipole recovery likelihood is:

$H_0$4

with $H_0$5.

To resolve the well-known $H_0$6 degeneracy, the minimization is performed in four structured steps:

1. Fit $H_0$7 with no dipole.
2. Coarse scan over 12 Healpix directions, fitting only $H_0$8 to localize $H_0$9.
3. Full parameter fit $ΔH_0$0 from previous results.
4. Final refit of $ΔH_0$1 to mitigate joint-fit bias.

This approach has minimal dependence on the assumed $ΔH_0$2 and robustly converges to the injected $ΔH_0$3 and dipole sky position.

4. Error Modeling: Statistical and Systematic Budget

Statistical uncertainty is quantified by the median absolute deviation (MAD) from multiple independent realizations. The chief systematic arises from residuals in $ΔH_0$4 recovery dependent on the dipole direction—attributed primarily to imprints from observing cadence. Errors are combined in quadrature:

$ΔH_0$5

where $ΔH_0$6 is empirically modeled across varying $ΔH_0$7.

For the clustered ZTF-like configuration with $ΔH_0$8 km s⁻¹ Mpc⁻¹, the achieved precisions are:

• $ΔH_0$9 km s⁻¹ Mpc⁻¹
• Directional precisions: $n̂$0 (right ascension), $n̂$1 (declination)

With no injected dipole, $n̂$2 km s⁻¹ Mpc⁻¹ and $n̂$3 km s⁻¹ Mpc⁻¹, giving a combined $n̂$4 km s⁻¹ Mpc⁻¹, ensuring no false >1$n̂$5 signals in null cases.

5. Robustness Tests and Sensitivity Analysis

Multiple robustness tests validate the stability of the pipeline:

• Monopole independence: $n̂$6 values of 67, 70, 73 km s⁻¹ Mpc⁻¹ produce identical $n̂$7 detection and dipole direction.
• Amplitude recovery: $n̂$8 of 1, 2, 3 km s⁻¹ Mpc⁻¹ are unbiasedly recovered; sensitivity, $n̂$9, is %%%%50$ΔH_0$51%%%% even for $p̂$2.
• Sky coverage: Restriction to ZTF’s footprint, MW masking, completeness modeling, and real observing logs do not bias $p̂$3 or direction; only cadence imprints induce mild latitude-dependent scatter well within $p̂$4.
• Large-scale structure: N-body derived peculiar velocities add scatter but do not bias dipole recovery.
• Volume-limited samples: A $p̂$5 subsample ($p̂$61,000 SNe) yields the same dipole detection but with 25–30% larger errors, as expected from reduced statistics.

This suite of tests affirms internal consistency, lack of bias from prior parameter choices, and control of major known systematics.

6. Prospects and Recommendations for ZTF Dipole Searches

For imminent ZTF analyses, the recommended pipeline begins with a bias-controlled, volume-limited SN Ia sample ($p̂$7), applies SALT2 fitting, and adopts the four-step dipole fitting procedure enhanced by empirical error modeling:

• Final statistical and systematic errors are summed in quadrature ($p̂$80.16 km s⁻¹ Mpc⁻¹ and $p̂$90.25–0.40 km s⁻¹ Mpc⁻¹, respectively).
• The expected DR2.5-level precision is $(α, δ)$00.3 km s⁻¹ Mpc⁻¹ in $(α, δ)$1 and a few degrees in direction.
• Chief remaining challenges: precise $(α, δ)$2 calibration, peculiar velocity corrections for $(α, δ)$3, and sky inhomogeneities from cadence.

Forthcoming ZTF releases, with larger and more complete samples, and LSST-era surveys, should feasibly reach $(α, δ)$4 sensitivity $(α, δ)$50.1 km s⁻¹ Mpc⁻¹ and sky localization to sub-degree scales. This suggests the methodology is scalable and adaptable to future wide-field SN samples. The machinery developed is robust against known systematics and is poised for application to real data, offering the potential for high-significance detection or stringent constraints on cosmic expansion anisotropy (Barjou-Delayre et al., 16 Jan 2026).