Distance Duality Relation (DDR): Fundamentals & Tests

• Distance Duality Relation (DDR) is a fundamental kinematical principle linking luminosity and angular-diameter distances via the (1+z)² scaling.
• It serves as a diagnostic tool for testing photon conservation, cosmic opacity, and possible deviations from standard metric theories.
• Modern analyses employ methods like LOESS, SIMEX, and Gaussian processes on supernova and lensing data to validate DDR within current observational uncertainties.

The distance duality relation (DDR), also called the Etherington relation or reciprocity relation, is the statement that the luminosity distance $D_L$ and angular-diameter distance $D_A$ to the same source at redshift $z$ satisfy

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

It is a kinematical relation rather than a dynamical one: in the standard derivation it follows when spacetime is described by a metric theory, photons propagate along null geodesics, and photon number is conserved. Because it links flux-based and size-based distance measures inferred from distinct classes of observations, DDR is both a null test of standard cosmology and a diagnostic of systematics or new physics (Rana et al., 2015, More et al., 2016).

1. Definition, conventions, and theoretical status

A standard diagnostic is a dimensionless deformation function $\eta(z)$. The recent literature represented here uses two reciprocal conventions: $\eta(z)=\frac{D_L}{D_A(1+z)^2}$ and

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

In both conventions, the standard DDR condition is $\eta(z)=1$. This sign convention matters when comparing reported $\eta_0$ constraints, especially across strong-lensing and supernova analyses that adopt opposite definitions (Rana et al., 2015, Liao et al., 2015).

The significance of a confirmed deviation is correspondingly broad. The observational literature repeatedly interprets $\eta(z)\neq 1$ as a possible sign of photon non-conservation, cosmic opacity, dust extinction, photon-axion conversion, exotic photon propagation, variation of fundamental constants, or a breakdown of the assumptions underlying metric gravity and null geodesics (Rana et al., 2015, Lin et al., 2018, Qi et al., 2024). At the same time, several studies emphasize that DDR tests are often limited by astrophysical modeling and calibration systematics, so an apparent violation is not automatically a detection of new physics (Vavrycuk et al., 2020, Tang et al., 2022).

Theoretical analyses sharpen that distinction. In "Modifications to the Etherington Distance Duality Relation and Observational Limits" (More et al., 2016), the most general linear electrodynamics without birefringence, based on a Lorentzian metric plus dilaton and axion fields, modifies amplitudes and fluxes but leaves the conserved current underlying reciprocity intact, so the standard relation $D_A$0 still holds. A closely related conclusion has been derived for symmetric teleparallel gravity: pure nonmetricity does not break DDR when electromagnetism is minimally coupled and photon number is conserved, whereas a nonminimal interaction $D_A$1 produces a dynamical violation,

$D_A$2

with

$D_A$3

in homogeneous and isotropic backgrounds (D'Agostino, 30 Jun 2026). Within this line of work, the basic dichotomy is between geometric settings that preserve reciprocity and dynamical couplings that alter photon-number transport.

2. Photon conservation, the CMB temperature law, and opacity

DDR is closely tied to the CMB temperature-redshift relation. Under standard assumptions,

$D_A$4

and for continuous blackbody sources one may write

$D_A$5

so that

$D_A$6

This relation was exploited in a non-parametric reconstruction based on CMB temperature measurements up to $D_A$7, which found no evidence of deviation from $D_A$8 within $D_A$9 over $z$0 (Rana et al., 2015). The same analysis used 36 temperature measurements, including multifrequency Sunyaev-Zeldovich observations, Planck thermal SZ measurements, and high-redshift measurements from damped Lyman-$z$1 systems and neutral carbon fine-structure excitation (Rana et al., 2015).

Opacity tests usually rewrite the luminosity distance as

$z$2

or equivalently $z$3, with common phenomenological choices

$z$4

and

$z$5

A scale-free test using JLA supernovae and BAO inferred angular-diameter distances between $z$6 and $z$7 found

$z$8

consistent with transparency and with standard DDR (More et al., 2016).

The interpretation of such opacity constraints is contested. "The failure of testing for cosmic opacity via the distance-duality relation" argues that present DDR-based transparency claims are weakened by three recurring issues: the non-unique interpretation of supernova dimming, the low accuracy and limited redshift range of current $z$9 data, and the physically problematic assumption that opacity is frequency independent across heterogeneous luminosity-distance probes (Vavrycuk et al., 2020). In that analysis, a transparent flat $D_L(z)=(1+z)^2D_A(z).$0CDM model and an opaque Einstein-de Sitter model can fit the same Pantheon supernova data comparably well, so $D_L(z)=(1+z)^2D_A(z).$1 does not by itself prove that the universe is transparent (Vavrycuk et al., 2020). The same paper identifies gravitational-wave standard sirens as a cleaner future route, because gravitational-wave luminosity distances are not affected by electromagnetic opacity (Vavrycuk et al., 2020).

3. Model-independent reconstruction and standard low-redshift probes

A central methodological difficulty in DDR tests is redshift matching: $D_L(z)=(1+z)^2D_A(z).$2 and $D_L(z)=(1+z)^2D_A(z).$3 are typically measured from different objects at different redshifts. An early solution used local regression to estimate $D_L(z)=(1+z)^2D_A(z).$4 from the Union2 supernova Hubble diagram at the exact redshifts of galaxy clusters. In that framework, local regression was less biased than weighted averaging or linear interpolation, and a Euclid-like BAO+SNe forecast was found to reduce the error on the low-redshift slope parameter $D_L(z)=(1+z)^2D_A(z).$5 by about a factor of two when $D_L(z)=(1+z)^2D_A(z).$6 is imposed (Cardone et al., 2012).

A more elaborate non-parametric implementation appears in "Revisiting the distance duality relation using a non-parametric regression method" (Rana et al., 2015). There the reconstruction uses LOESS combined with SIMEX. LOESS supplies a locally weighted regression of $D_L(z)=(1+z)^2D_A(z).$7 with tricube weights and smoothing selected by leave-one-out cross-validation; SIMEX corrects for measurement-error bias by adding controlled noise, fitting the reconstruction as a function of the noise parameter $D_L(z)=(1+z)^2D_A(z).$8, and extrapolating to $D_L(z)=(1+z)^2D_A(z).$9 (Rana et al., 2015). The method was validated on a mock sample of 200 equidistant $\eta(z)$0 points generated from

$\eta(z)$1

with fiducial $\eta(z)$2, using realistic redshift-dependent errors inferred from CMB-temperature data (Rana et al., 2015). Applied to real data, it combined JLA SNe Ia luminosity distances with FRIIb radio-galaxy angular-diameter distances, leaving 12 matched points over $\eta(z)$3, and separately reconstructed $\eta(z)$4 from 36 CMB temperature measurements over $\eta(z)$5. In both cases, the result was no evidence of DDR violation within $\eta(z)$6 (Rana et al., 2015).

Gaussian-process reconstruction has been used in a similar spirit. "New constraints on the distance duality relation from the local data" combines Pantheon supernovae with cluster and BAO angular-diameter distances, reconstructing the supernova magnitude-redshift relation with GaPP and a squared-exponential kernel (Lin et al., 2018). The strongest constraints came from elliptical clusters plus BAO,

$\eta(z)$7

for $\eta(z)$8, and

$\eta(z)$9

for $\eta(z)=\frac{D_L}{D_A(1+z)^2}$0; the spherical-cluster sample was much less constraining because it required substantial intrinsic scatter,

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

which the authors interpreted as evidence that the spherical model is a poor approximation to actual cluster structure (Lin et al., 2018).

Forecast studies extend this model-independent program. "Euclid: Forecast constraints on the cosmic distance duality relation with complementary external probes" reports current parametric constraints of

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

from Pantheon + BAO for the constant-deformation model $\eta(z)=\frac{D_L}{D_A(1+z)^2}$3, and forecasts that Euclid combined with LSST/DESIRE supernovae and DESI BAO can improve current constraints by approximately a factor of six in parametric analyses and by a factor of three in non-parametric Genetic-Algorithm reconstructions (Martinelli et al., 2020). In that framework, Euclid-era parametric data constrain $\eta(z)=\frac{D_L}{D_A(1+z)^2}$4 to the few $\eta(z)=\frac{D_L}{D_A(1+z)^2}$5 level over $\eta(z)=\frac{D_L}{D_A(1+z)^2}$6 (Martinelli et al., 2020).

4. Strong gravitational lensing and time-delay cosmography

Strong gravitational lensing has become a major DDR laboratory because it supplies angular-diameter-distance ratios or, in time-delay systems, absolute distance combinations. An early ratio-based construction used 118 strong lenses from SLACS, BELLS, LSD, and SL2S, together with JLA supernovae, under the flat-FRW identity

$\eta(z)=\frac{D_L}{D_A(1+z)^2}$7

With the singular isothermal sphere model,

$\eta(z)=\frac{D_L}{D_A(1+z)^2}$8

and a first-order violation model $\eta(z)=\frac{D_L}{D_A(1+z)^2}$9, the analysis of 60 matched lensing/SN systems found

$\eta(z)=\frac{D_A(1+z)^2}{D_L}.$0

fully consistent with DDR (Liao et al., 2015).

Time-delay cosmography replaces ratio-based lensing observables by the time-delay distance

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

which depends on three angular-diameter distances. "Testing the Cosmic Distance Duality Relation Using Strong Gravitational Lensing Time Delays and Type Ia Supernovae" combines six H0LiCOW time-delay lenses with Pantheon+ supernovae and reconstructs the supernova magnitude-redshift relation $\eta(z)=\frac{D_A(1+z)^2}{D_L}.$2 using the REFANN neural-network code with one hidden layer of 4096 neurons (Qi et al., 2024). It tests three standard one-parameter deformation laws,

$\eta(z)=\frac{D_A(1+z)^2}{D_L}.$3

and explicitly studies the strong degeneracy between $\eta(z)=\frac{D_A(1+z)^2}{D_L}.$4 and the supernova absolute magnitude $\eta(z)=\frac{D_A(1+z)^2}{D_L}.$5. Whether $\eta(z)=\frac{D_A(1+z)^2}{D_L}.$6 is left free or fixed to the Cepheid-calibrated SH0ES value $\eta(z)=\frac{D_A(1+z)^2}{D_L}.$7 mag, all three parameterizations remain consistent with $\eta(z)=\frac{D_A(1+z)^2}{D_L}.$8 within $\eta(z)=\frac{D_A(1+z)^2}{D_L}.$9 (Qi et al., 2024).

A related analysis, "The cosmic distance duality relation in light of the time-delayed strong gravitational lensing," uses four H0LiCOW systems with both $\eta(z)=1$0 and $\eta(z)=1$1 measurements and reconstructs $\eta(z)=1$2 from Pantheon+ with Gaussian processes (Tang et al., 2024). It samples the lensing and supernova posteriors directly and fits three models,

$\eta(z)=1$3

finding

$\eta(z)=1$4

respectively, again all consistent with DDR at $\eta(z)=1$5 (Tang et al., 2024). Monte Carlo forecasts in the same study suggest that 100 LSST-quality time-delay systems could push the precision to the $\eta(z)=1$6 level (Tang et al., 2024).

The main caution is lens-model dependence. "Deep learning method in testing the cosmic distance duality relation" reconstructs the supernova Hubble diagram with an LSTM-based deep-learning model and combines it with 161 strong-lensing systems out to $\eta(z)=1$7 (Tang et al., 2022). The inferred DDR result depends strongly on the assumed lens mass profile. For the full sample, the paper reports

$\eta(z)=1$8

in the SIS model and

$\eta(z)=1$9

in the extended power-law model, both interpreted there as high-significance apparent violations, whereas the power-law model gives

$\eta_0$0

fully consistent with DDR (Tang et al., 2022). The substantive point is not a consensus detection of violation, but the sensitivity of SGL-based DDR constraints to astrophysical lens modeling.

5. High-redshift and multi-messenger extensions

DDR tests have progressively moved beyond the supernova–cluster redshift range. One extension uses compact radio quasars as standard rulers and gravitational-wave standard sirens as opacity-free luminosity-distance indicators. "Testing the Etherington's distance duality relation at higher redshifts: the combination of radio quasars and gravitational waves" combines 120 intermediate-luminosity quasars over $\eta_0$1 with simulated Einstein Telescope sirens (Qi et al., 2019). For the three common parameterizations

$\eta_0$2

it finds current-level uncertainties of order $\eta_0$3, for example

$\eta_0$4

and forecasts $\eta_0$5-level precision with 500 simulated quasars over $\eta_0$6 plus Einstein Telescope data (Qi et al., 2019).

An even cleaner multi-messenger proposal uses strongly lensed gravitational waves, where the same source can provide both $\eta_0$7 and $\eta_0$8 along the same line of sight. In "Strongly lensed gravitational waves as the probes to test the cosmic distance duality relation," the source angular-diameter distance is reconstructed from lensing observables and time delays, while the luminosity distance is measured from the gravitational-wave amplitude after correcting for lensing magnification (Lin et al., 2020). Under Einstein Telescope assumptions, about 100 strongly lensed GW events can constrain the linear and saturating one-parameter models at the $\eta_0$9 and $\eta(z)\neq 1$0 levels, respectively (Lin et al., 2020).

High-redshift electromagnetic combinations have also been developed. "Testing the Distance Duality Relation with Cosmological Observations at high Redshift using Artificial Neural Network" combines Pantheon+ SNe Ia, Fermi gamma-ray bursts, DESI DR2 BAO, and two galaxy-scale strong-lensing samples, using neural networks to reconstruct luminosity distances over

$\eta(z)\neq 1$1

(Xie et al., 6 Dec 2025). Across both

$\eta(z)\neq 1$2

and

$\eta(z)\neq 1$3

the standard DDR remains consistent with the data at approximately the $\eta(z)\neq 1$4 level; the paper emphasizes that the SGL compilation and the adopted supernova absolute-magnitude prior materially affect the fitted $\eta(z)\neq 1$5 values (Xie et al., 6 Dec 2025).

6. Beyond-standard interpretations, methodological disputes, and current status

DDR has been used as a probe of specific particle-physics mechanisms. "Constraining axionlike particles using the distance-duality relation" simulates three-dimensional ALP-photon mixing in a $\eta(z)\neq 1$6CDM universe with a primordial stochastic magnetic field and interprets the observed DDR scatter as an upper bound on the coupling (Tiwari, 2016). Under the explicit assumption that the observed DDR scatter is fully attributable to ALP-photon mixing, the paper obtains

$\eta(z)\neq 1$7

for

$\eta(z)\neq 1$8

thereby treating DDR as a constraint on photon depletion or enhancement rather than only as a consistency relation (Tiwari, 2016).

Other departures are directional or phenomenological rather than microphysical. "Testing the anisotropy of the Universe with the distance duality relation" introduces a dipolar anisotropic parametrization,

$\eta(z)\neq 1$9

and compares matched supernovae with strong-lensing and cluster data (Li et al., 2017). Union2.1-based analyses remained compatible with DDR, while JLA-based fits showed only mild $D_A$00 hints of anisotropic violation; the same study concluded that no strong evidence exists because of current data uncertainty (Li et al., 2017). An unconventional radio-source study argued that, under the assumption of constant luminosity density, two ultracompact-radio-source samples yield a ratio $D_A$01 more consistent with $D_A$02 than with $D_A$03, whereas enforcing the standard expanding-universe DDR requires $D_A$04; that conclusion is explicitly tied to the adopted source-evolution model and illustrates how strongly DDR inferences can depend on astrophysical calibration (Li, 2023).

Recent tension-driven work treats DDR violation as an effective recalibration between the supernova and BAO sectors. "Implications of distance duality violation for the $D_A$05 tension and evolving dark energy" finds that two toy models can reconcile SH0ES-calibrated supernovae with Planck-calibrated BAO: a constant offset,

$D_A$06

and a low-redshift power-law modification,

$D_A$07

restricted to $D_A$08, together with a constant phantom equation of state $D_A$09 (Teixeira et al., 14 Apr 2025). "Redshift-dependent Distance Duality Violation in Resolving Multidimensional Cosmic Tensions" reaches a related conclusion: a constant DDR offset can move the supernova normalization but leaves the Pantheon-inferred $D_A$10 essentially unchanged, whereas a time-varying $D_A$11 lowers the supernova-inferred $D_A$12, improves the global fit by roughly $D_A$13 without SH0ES, and performs best when combined with evolving dark energy (Zhou et al., 4 Nov 2025). These analyses are explicitly phenomenological and do not claim a detection of DDR violation; they treat $D_A$14 as a late-time recalibration degree of freedom.

Methodological choice can dominate the reported significance. "Probing the Distance Duality Relation with Machine Learning and Recent Data" compares a one-parameter fit

$D_A$15

with a model-independent Genetic-Algorithm reconstruction using DESI DR1, Pantheon+, SH0ES, and DES-SN5YR (Keil et al., 2 Apr 2025). In the parametrized analysis, the uncalibrated DESI + Pantheon+ combination gives an approximately $D_A$16 DDR anomaly, and adding both BBN and SH0ES drives the apparent significance to $D_A$17; in the GA analysis, the uncalibrated case shows no significant deviation, and the calibrated case remains only at about $D_A$18 (Keil et al., 2 Apr 2025). In conjunction with the opacity critique of (Vavrycuk et al., 2020) and the lens-model sensitivity found in (Tang et al., 2022), this establishes the present status of the field: the standard DDR is broadly consistent with current observations, but the strength of any claimed anomaly is highly sensitive to probe selection, calibration, parametrization, and astrophysical modeling.