Cosmic Distance Duality Relation (CDDR)

• CDDR is a geometric relation asserting that luminosity and angular diameter distances are connected by (1+z)² under metric gravity and photon conservation.
• It underpins tests using supernovae, BAO, galaxy clusters, quasars, and gravitational waves to probe cosmic opacity and potential new physics.
• Empirical and non-parametric analyses consistently find η(z) ≡ 1 across various redshifts, establishing CDDR as a robust consistency check in observational cosmology.

The cosmic distance duality relation (CDDR), also known as the Etherington reciprocity relation, is the statement that the luminosity distance and angular diameter distance to the same source obey $D_L(z)=(1+z)^2D_A(z)$ when spacetime is metric, photons propagate along null geodesics, and photon number is conserved. Because this relation is geometric rather than dynamical, it is independent of Einstein’s equations and of the material content of the Universe, and therefore functions both as a foundational consistency condition in observational cosmology and as a diagnostic for opacity, photon conversion, non-metric gravity, calibration systematics, and redshift-dependent reconstruction bias. Recent analyses using supernovae, baryon acoustic oscillations, cosmic chronometers, galaxy clusters, strong lensing, megamasers, quasars, gamma-ray bursts, and gravitational waves generally find $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$ consistent with unity over the currently accessible redshift range, though the precision and the dominant systematics are strongly probe-dependent (Barua et al., 27 Oct 2025, Avila et al., 9 Sep 2025).

1. Reciprocity theorem and the duality parameter

Etherington’s theorem is usually written in its canonical form

$D_L(z)=(1+z)^2D_A(z),$

or, equivalently,

$\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$

The relation holds if three conditions are satisfied: spacetime is described by a metric theory of gravity, photons follow unique null geodesics, and photon number is conserved. These assumptions recur throughout the modern literature, from early model-dependent opacity analyses to recent model-independent reconstructions (Lv et al., 2016, Kanodia et al., 15 Jul 2025).

The conceptual importance of $\eta(z)$ lies in the fact that it isolates any breakdown of reciprocity from the background expansion model. If $\eta(z)\neq 1$, the departure may arise from genuine new physics, such as photon–axion or photon–dark photon conversion, non-metric gravity, or violations of Lorentz invariance, but it may also arise from astrophysical or instrumental effects, including dust absorption, unaccounted extinction, redshift misestimation, lensing-induced dispersion, or miscalibration (Kanodia et al., 15 Jul 2025). In that sense, CDDR tests are simultaneously probes of fundamental spacetime structure and of cross-survey consistency.

A closely related language is that of effective opacity. If the optical depth between source and observer is $\tau(z)$, then

$$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$

which implies

$\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$

This mapping is widely used in the literature because it converts duality violation into a transparency problem without changing the underlying observational estimator (Lv et al., 2016).

2. Deviation parameterizations and their physical meaning

The simplest CDDR tests treat $\eta(z)$ phenomenologically. One common parameterization writes

$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$0

so that

$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$1

In this formulation, $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$2 recovers the standard relation, and the deviation is directly interpretable as cosmic opacity through $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$3 (Lv et al., 2016).

Much of the recent literature instead works directly with one-parameter deformations of $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$4, most commonly

$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$5

and, in some studies,

$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$6

These parameterizations appear in BAO-plus-supernova analyses, matched-pair cluster-supernova tests, and PAge-based reconstructions, where they are used to capture low-redshift, saturating, logarithmic, or power-law departures from reciprocity (Li et al., 18 Jul 2025, Hu et al., 15 May 2026, Wu, 29 Mar 2026).

Two further points are technically important. First, some analyses fit redshift evolution in a more general form, such as $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$7, especially in strong-lensing applications where broad redshift leverage allows sensitivity to curvature in the functional form (Qin et al., 2021). Second, not all recent work adopts a parametric $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$8. Several studies reconstruct $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$9 non-parametrically with Gaussian Processes or compute it directly at specific redshifts, explicitly avoiding global templates (Avila et al., 9 Sep 2025). The arbitrary-pivot Padé cosmography approach is even more local: it infers $D_L(z)=(1+z)^2D_A(z),$0 at a chosen pivot redshift $D_L(z)=(1+z)^2D_A(z),$1 rather than fitting a global deformation law (Barua et al., 27 Oct 2025).

A recurrent misconception is that all parameterizations have equivalent physical status. They do not. Linear forms are convenient at low redshift but become increasingly artificial as $D_L(z)=(1+z)^2D_A(z),$2 grows; saturating forms such as $D_L(z)=(1+z)^2D_A(z),$3 remain bounded; opacity-inspired power laws inherit a more direct interpretation in terms of photon attenuation. The choice of parameterization can therefore change both the numerical constraint and its physical readability, even when the data are otherwise unchanged.

3. Observational realizations of $D_L(z)=(1+z)^2D_A(z),$4 and $D_L(z)=(1+z)^2D_A(z),$5

In practice, CDDR tests are limited by how $D_L(z)=(1+z)^2D_A(z),$6 and $D_L(z)=(1+z)^2D_A(z),$7 are obtained. Type Ia supernovae remain the dominant luminosity-distance tracer. Their standard relation is

$D_L(z)=(1+z)^2D_A(z),$8

or, equivalently,

$D_L(z)=(1+z)^2D_A(z),$9

with $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$0 the absolute magnitude or an effective calibration parameter after light-curve standardization. Pantheon+, DES Y5, Pantheon, JLA, SNLS, and DES Dovekie are all used in recent CDDR work, but the strong degeneracy between $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$1 and the inferred $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$2 is now recognized as one of the central issues in the field (Li et al., 18 Jul 2025, Zhang et al., 22 Jun 2025).

On the angular-diameter side, BAO provide the cleanest large-scale measurements. Transverse BAO constrain $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$3, with

$\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$4

while isotropic BAO use

$\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$5

DESI DR2 is particularly important because it provides high-precision $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$6, $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$7, and $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$8 measurements over a broad redshift range, including $\eta(z)\equiv \frac{D_L(z)}{(1+z)^2D_A(z)}=1.$9 (Barua et al., 27 Oct 2025).

Cosmic chronometers enter CDDR analyses by supplying direct $\eta(z)$0 measurements,

$\eta(z)$1

which can break degeneracies in BAO-plus-supernova tests and convert isotropic BAO measurements into $\eta(z)$2. They are especially valuable in reconstruction-based studies that attempt to remain agnostic about the late-time background expansion (Barua et al., 27 Oct 2025, Zhang et al., 22 Jun 2025).

Galaxy clusters provide an older but still important angular-diameter route through joint SZE-plus-X-ray analyses. Here a subtlety arises: because the X-ray surface-brightness relation already depends on $\eta(z)$3, a CDDR violation modifies the inferred cluster distance. In the notation of recent cluster analyses,

$\eta(z)$4

so cluster-based CDDR tests must account explicitly for this dependence rather than treating $\eta(z)$5 as a direct geometric observable (Avila et al., 9 Sep 2025). A related same-object method compares gas mass fractions from SZE and X-ray observations, exploiting the relation $\eta(z)$6 (Holanda et al., 2012).

Other probes extend the accessible redshift domain or reduce calibration dependence. Water megamasers provide calibration-free low-redshift geometric distances (Kanodia et al., 15 Jul 2025). Strong-lensing combinations of Einstein radii, velocity dispersions, and time delays supply angular-diameter information with different systematics from clusters or BAO (Rana et al., 2017). Quasars and H II galaxies have been used as high-redshift luminosity and angular-size tracers in machine-learning or GP reconstructions (Liu et al., 2021). Gravitational-wave standard sirens, and especially strongly lensed gravitational waves, are attractive because the GW luminosity distance is unaffected by electromagnetic opacity (Huang et al., 2024).

4. Model-independent and semi-local methodologies

The recent methodological shift in CDDR studies has been away from global background fits and toward local, non-parametric, or calibration-aware constructions. A representative example is the arbitrary-pivot Padé $\eta(z)$7-(2,1) cosmography introduced for DESI-plus-supernova-plus-chronometer analyses. Defining $\eta(z)$8, the basic rational approximation is

$\eta(z)$9

with coefficients matched to Taylor expansions of $\eta(z)\neq 1$0 and $\eta(z)\neq 1$1 about an arbitrary pivot $\eta(z)\neq 1$2. The practical consequence is that $\eta(z)\neq 1$3, $\eta(z)\neq 1$4, $\eta(z)\neq 1$5, and $\eta(z)\neq 1$6 are fitted directly at the redshift of interest, avoiding extrapolation from $\eta(z)\neq 1$7 (Barua et al., 27 Oct 2025).

Non-parametric reconstruction is the other major strategy. Gaussian Processes have been used to reconstruct $\eta(z)\neq 1$8, $\eta(z)\neq 1$9, or $\tau(z)$0 directly from heterogeneous data, including BAO, clusters, supernovae, and quasars. Recent high-redshift work uses GP on binned supernova-plus-quasar luminosity distances and on BAO or cluster angular distances, obtaining $\tau(z)$1 without assuming a specific deformation law (Avila et al., 9 Sep 2025). GP also underlies one branch of the DESI DR2 transparency analysis and the 2026 PAge comparison study (Zhang et al., 22 Jun 2025, Wu, 29 Mar 2026).

Artificial neural networks play a complementary role as redshift-matching tools rather than physical models. In one BAO-plus-supernova study, REFANN with a single hidden layer of 4096 neurons is used to reconstruct $\tau(z)$2 or $\tau(z)$3 at BAO redshift nodes, enabling direct formation of $\tau(z)$4 even when supernova and BAO redshifts do not coincide exactly (Li et al., 18 Jul 2025). A related approach compresses Pantheon into 36 control points and interpolates in $\tau(z)$5, again to generate redshift-matched luminosity distances with propagated covariance (Xu et al., 2022).

Matched-pair methods remain attractive when direct object-by-object comparisons are possible. Recent cluster-supernova work pairs each cluster with a Pantheon+ supernova at nearly identical redshift and jointly fits CDDR violation and possible SN absolute-magnitude evolution (Hu et al., 15 May 2026). Strong-lensing studies have proposed “distance-deviation consistency” rather than a fixed $\tau(z)$6 threshold, so that the allowed redshift mismatch shrinks or grows to keep the fractional distance mismatch constant across redshift (Zhou et al., 2020).

Spatial curvature is a persistent methodological fault line. Many analyses impose $\tau(z)$7, so that $\tau(z)$8, but curvature-aware tests based on the FLRW distance sum rule now exist. These reconstruct a dimensionless comoving distance $\tau(z)$9 from strong lensing and use it together with supernova luminosity distances to test CDDR in both flat and non-flat spacetime (Gahlaut, 25 Jan 2025). Related strong-lensing-plus-quasar work found that CDDR consistency is optimized near the Planck curvature value and argued that large departures from flatness are disfavored in this context (Qin et al., 2021).

5. Empirical status of the relation

The empirical picture is one of broad consistency with $$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$0, but with sharp differences in precision, calibration sensitivity, and redshift reach.

A concise cross-section of recent results is given below.

Study	Data and method	Headline result
(Lv et al., 2016)	SNIa + GRB + $$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$1 + BAO + CMB; model-dependent $$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$2	$$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$3
(Kanodia et al., 15 Jul 2025)	Megamasers + Pantheon+; direct low-$$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$4 test	$$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$5
(Barua et al., 27 Oct 2025)	DESI DR2 + CC + Pantheon+/DES Y5; arbitrary-pivot Padé	No significant deviations; sub-percent precision over most $$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$6
(Avila et al., 9 Sep 2025)	GP reconstruction with BAO, clusters, SNe, quasars	No significant $$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$7 deviation up to $$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$8
(Hu et al., 15 May 2026)	Cluster–SN matched pairs	$$D_{L,\mathrm{obs}}(z)=D_{L,\mathrm{true}}(z)\exp\!\left[\frac{\tau(z)}{2}\right],$$9
(Wu, 29 Mar 2026)	PAge + GP with recent SN/BAO/CC/GRB	No evidence for violation at $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$0

Model-dependent opacity work already indicated that supernovae alone constrain deviation poorly, but that adding $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$1, BAO, and CMB distance information yields $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$2, consistent with zero (Lv et al., 2016). More recent low-redshift, calibration-free megamaser tests reach $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$3, again fully consistent with unity but limited by current megamaser distance precision (Kanodia et al., 15 Jul 2025).

The most precise recent model-independent constraints over $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$4 come from the arbitrary-pivot Padé analysis combining DESI DR2 BAO, cosmic chronometers, and Pantheon+ or DES Y5 supernovae. That study reports no significant deviations overall and achieves sub-percent precision across most pivots. It is also a useful cautionary case: around $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$5–0.8, some dataset combinations show formally high-significance departures because the statistical errors are extremely small, even though the central values remain within $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$6 of unity (Barua et al., 27 Oct 2025). This demonstrates that “multi-$\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$7” and “large physical violation” are not interchangeable statements in CDDR work.

At higher redshift, the picture is mixed but still mostly conservative. A fully non-parametric GP reconstruction using BAO, galaxy clusters, Pantheon+, and quasars finds no significant departure from CDDR at the $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$8 level up to $\eta(z)=\exp\!\left[\frac{\tau(z)}{2}\right].$9 (Avila et al., 9 Sep 2025). By contrast, a two-point DESI-based diagnostic that cancels both $\eta(z)$0 and $\eta(z)$1 analytically reports no significant low-redshift deviation but does find $\eta(z)$2 evidence at $\eta(z)$3 and $\eta(z)$4, with the authors explicitly cautioning that multiple-comparisons effects are not corrected and may reduce the nominal significance (Wang et al., 15 Jun 2025).

Matched cluster-supernova analyses remain much less precise than BAO-plus-supernova constructions but are valuable because they expose different systematics. A recent Pantheon+-plus-cluster test, including possible SN absolute-magnitude evolution $\eta(z)$5, found

$\eta(z)$6

with no statistically significant evidence for either CDDR violation or SN evolution (Hu et al., 15 May 2026).

6. Calibration dependence, systematic limits, and future directions

The dominant controversy in current CDDR work is not whether the formalism is correct, but how strongly the inferred $\eta(z)$7 depends on calibration choices. BAO measure distances in units of the sound horizon $\eta(z)$8, while supernovae measure them relative to the absolute magnitude $\eta(z)$9. Under CDDR, BAO-plus-supernova tests often constrain only combinations such as $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$00, so inconsistent choices of early- and late-Universe calibrations can mimic reciprocity violation (Kanodia et al., 15 Jul 2025). This is now one of the clearest lessons of the recent literature.

The calibration issue appears in several distinct forms. BAO-plus-supernova analyses with PantheonPlus show a strong negative correlation between the CDDR deviation parameter and $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$01; when $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$02 is free, no evidence for violation is found, whereas fixing $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$03 to one calibration prior can produce an apparent $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$04 deviation and fixing it to another can remove it (Li et al., 18 Jul 2025). A broader DESI DR2 transparency study reaches a similar conclusion: direct reconstruction is consistent with CDDR for one $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$05 prior but shows a notable departure when the SH0ES prior is imposed, which the authors interpret as systematic sensitivity to the Hubble-tension calibration rather than robust evidence against Etherington reciprocity (Zhang et al., 22 Jun 2025). A 2026 PAge-plus-GP analysis makes the point even more sharply: imposing both SH0ES $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$06 and Planck $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$07 simultaneously can generate spurious $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$08–$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$09 CDDR violation, while self-consistent calibration choices leave $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$10 consistent with unity (Wu, 29 Mar 2026).

Probe-specific systematics are equally important. Cluster tests inherit hydrostatic-equilibrium assumptions, morphology dependence, SZE and X-ray calibration errors, and the fact that cluster-derived distances may already be $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$11-dependent (Holanda et al., 2012, Avila et al., 9 Sep 2025). Strong-lensing tests depend on mass-profile assumptions, velocity-dispersion calibration, external convergence, anisotropy, and aperture corrections (Rana et al., 2017). Reconstruction methods add their own regularization structure: GP kernel choice, ANN architecture, control-point placement, and redshift binning all change the smoothness of the inferred $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$12 or $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$13.

The cleanest route around opacity systematics is gravitational-wave standard sirens. Strongly lensed GW signals allow $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$14 to be inserted directly into the waveform through the amplitude scaling $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$15, while the lens geometry carries angular-diameter information. A TianQin forecast found that a single strongly lensed GW event can measure $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$16 at the $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$17–$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$18 level for $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$19 and $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$20–$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$21 for $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$22, depending on lens model and source configuration (Huang et al., 2024). A later Taiji-plus-LISA network study projected that combining 10 simulated events could shrink the half-width of the $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$23 credible interval to approximately $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$24 for $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$25 and $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$26 for $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$27, while remaining consistent with $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$28 (Yuan et al., 24 Mar 2026). These are forecasts rather than measurements, but they indicate that future opacity-free tests may surpass electromagnetic constraints by orders of magnitude if lens modeling and cosmology dependence are controlled.

The present state of the field therefore supports three conclusions. First, CDDR remains empirically intact across most tested redshifts, typically within $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$29–$\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$30. Second, apparent violations are often calibration-sensitive and must be interpreted jointly with $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$31, $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$32, and, in some analyses, $\eta(z)\equiv D_L(z)/[(1+z)^2D_A(z)]$33. Third, future progress will come less from new parameterizations than from better cross-calibration, higher-redshift distance tracers, improved chronometer and BAO systematics, and genuinely opacity-free measurements from gravitational waves.