Unified Dark Fluid (UDF) Models

• Unified dark fluid (UDF) models represent a single dark sector fluid that behaves like dust during structure formation and transitions to a dark-energy-like phase in later epochs.
• They encompass various parameterizations—such as constant-sound-speed and fast-transition formulations—that interpolate between matter-like and dark-energy-like behaviors.
• The models focus on effective sound speed and clustering properties to break dark degeneracy, with observational tests targeting large-scale structure, perturbations, and nonlinear dynamics.

Unified dark fluid (UDF) denotes a class of cosmological models in which the entire dark sector is represented by a single effective component rather than by separate cold dark matter and dark energy. The central motivation is the dark degeneracy: background observables, and to some extent linear perturbations, constrain the total gravitational effect of the dark sector more directly than any unique decomposition into dark matter and dark energy. In viable UDF constructions, the unified component behaves approximately as dust at early times, thereby supporting structure formation, and evolves toward a negative-pressure phase at late times, thereby driving accelerated expansion (Xu et al., 2011).

1. Conceptual basis and scope

The defining idea of a UDF is not a particular equation of state but a unification principle. In its most common form, the dark sector is modeled as a single barotropic or effectively non-adiabatic fluid that couples gravitationally to baryons and radiation and is engineered to interpolate between matter-like and dark-energy-like regimes. In the constant-adiabatic-sound-speed formulation of Xu et al., this is encoded by a linear relation $p_d=\alpha \rho_d-A$, while in other constructions the interpolation is imposed directly through $w(a)$, $P(z)$, or a late-time expansion ansatz (Xu et al., 2011).

A common misconception is to identify UDF models with generalized Chaplygin gas alone. The literature is broader. It includes constant-sound-speed barotropes, fast-transition $w(a)$ models, pressure-parametrized fluids, PAge-like late-time parameterizations, viscous and dissipative fluids, null-sound-speed effective fluids, and even unified dark radiation–dark matter models (Xu, 2012, Wang et al., 2024, Geng et al., 2013). The generalized Chaplygin gas is therefore one special case inside a wider UDF program, not the program itself.

Another misconception is that “unified” implies a fundamental single substance. Several works explicitly treat the UDF as an effective description. A barotropic fluid with $c_s^2=0$ is exactly degenerate with $\Lambda$CDM at background and linear level, and can be interpreted as a coarse-grained combination of a clustering piece and a vacuum-like piece, even though it is written as one fluid (Aviles et al., 2014). Conversely, other papers derive UDF behavior from scalar-field, superfluid, or modified-gravity constructions, indicating that unification may be phenomenological or microphysical depending on the model class (Ferreira et al., 2018, Elkhateeb, 2023).

The unification principle also extends beyond the standard dark matter–dark energy split. One variant describes a single fluid with

$$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$

which behaves as dark radiation at early times and dark matter at late times, and can be decomposed into interacting dark-radiation and dark-matter sectors (Geng et al., 2013). This makes clear that “UDF” is best understood as a modeling strategy for the dark sector, rather than as one fixed phenomenological ansatz.

2. Background dynamics and representative parameterizations

At background level, most UDF models are built to reproduce the same qualitative history: $w\simeq 0$ when the Universe is matter dominated and $w< -1/3$ at late times. The main differences lie in how the interpolation is parameterized and in whether the model remains barotropic, admits entropy perturbations, or enforces $c_{s,\rm eff}^2=0$.

For the constant-adiabatic-sound-speed UDF of Xu et al., the defining relations are

$w(a)$0

$w(a)$1

with $w(a)$2 and $w(a)$3. The corresponding

$w(a)$4

tends to $w(a)$5 at early times and to $w(a)$6 at late times. In the limit $w(a)$7, the model reduces exactly to a $w(a)$8CDM-like background with effective $w(a)$9 and $P(z)$0 (Xu et al., 2011).

A different family imposes a fast transition directly in the equation of state,

$P(z)$1

so that $P(z)$2 for $P(z)$3 and $P(z)$4 for $P(z)$5. Here $P(z)$6 fixes the transition epoch and $P(z)$7 its width; smaller $P(z)$8 means a sharper transition (Yang et al., 2013).

Other parameterized UDFs are designed to include $P(z)$9CDM as a special case. One example takes

$w(a)$0

which reduces to generalized Chaplygin gas for a specific parameter mapping and reproduces the exact $w(a)$1CDM background when $w(a)$2, $w(a)$3, and $w(a)$4 (Xu, 2012). Another uses a pressure parameterization

$w(a)$5

leading to

$w(a)$6

so that the unified density explicitly separates into a dust-like term, a constant term, and a controlled deviation from $w(a)$7CDM (Wang et al., 2017).

A further one-parameter construction uses

$w(a)$8

with exact density evolution

$w(a)$9

This model behaves as dust at early times and asymptotically approaches $c_s^2=0$0 at late times, with $c_s^2=0$1 controlling the transition (Yang et al., 2019).

More recent work introduces a PAge-like UDF in which the late-time expansion is encoded by $c_s^2=0$2 rather than by a fluid $c_s^2=0$3. In that formulation,

$c_s^2=0$4

and the unified density is defined by subtracting baryons from the PAge background. The resulting effective equation of state is reconstructed through the continuity equation rather than imposed directly (Wang et al., 2024).

Model family	Defining relation	Limiting behavior
Constant $c_s^2=0$5 UDF (Xu et al., 2011)	$c_s^2=0$6	$c_s^2=0$7 early, $c_s^2=0$8 late
Fast-transition UDF (Yang et al., 2013)	$c_s^2=0$9	Matter-like early, dark-energy-like late
Generalized-parametric UDF (Xu, 2012)	$\Lambda$0	Includes gCg and $\Lambda$1CDM limits
Pressure-parametrized UDF (Wang et al., 2017)	$\Lambda$2	$\Lambda$3-like constant pressure plus deviations
Sinc UDF (Yang et al., 2019)	$\Lambda$4	Dust early, $\Lambda$5 late
PAge-like UDF (Wang et al., 2024)	$\Lambda$6 with $\Lambda$7	Dust-like clustering with late acceleration

Taken together, these parameterizations show that UDF cosmology is defined less by a unique background law than by a common interpolation structure and by its treatment of perturbations.

3. Linear perturbations, sound speed, and clustering

The perturbative sector is the decisive part of UDF phenomenology. At background level many UDFs can closely mimic $\Lambda$8CDM; the distinction typically appears in the effective sound speed, the Jeans scale, and the evolution of gravitational potentials.

For the constant-adiabatic-sound-speed model, the synchronous-gauge perturbations satisfy

$\Lambda$9

$$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$0

with $$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$1 and $$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$2. A nonzero $$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$3 introduces pressure support and a finite sound horizon, suppressing clustering for modes above the sound-horizon scale and enhancing the late Integrated Sachs–Wolfe contribution through more rapid potential decay (Xu et al., 2011).

Once entropy perturbations are allowed, adiabatic and effective sound speeds decouple. In the entropic formulation,

$$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$4

while the rest-frame effective sound speed is defined through

$$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$5

The corresponding synchronous-gauge system contains the entropy term

$$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$6

and permits $$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$7 to remain small even when $$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$8 differs from zero or becomes negative (Xu, 2012). This is the standard route by which UDF models avoid the perturbative pathologies of purely adiabatic negative-pressure fluids.

The fast-transition UDF with entropic perturbations makes the same point in a more explicit Jeans-scale language. Its effective Jeans wavenumber is written as

$$\rho_{\rm dark}=(Aa^{-4(1+\alpha)}+Ba^{-3(1+\alpha)})^{1/(1+\alpha)},$$9

so that viability requires either a transient boost in the absolute factor around the transition or, more robustly, a very small $w\simeq 0$0 (Yang et al., 2013).

Several later UDF constructions impose the strongest possible condition: $w\simeq 0$1 This appears in the Newtonian/relativistic barotropic analysis with constant pressure $w\simeq 0$2, in the PAge-like UDF, and in the null-sound-speed UDF matched to DESI-motivated CPL backgrounds (Aviles et al., 2014, Wang et al., 2024, Kou et al., 19 Sep 2025). In that regime, the UDF clusters like CDM on sub-horizon scales, the Jeans pressure-support term vanishes, and the linear growth is either exactly degenerate with $w\simeq 0$3CDM or differs only through the chosen background history.

The observational consequence is severe. In the constant-sound-speed model, even modest $w\simeq 0$4 suppresses small-scale power, and matching the SDSS DR7 matter power spectrum requires

$w\simeq 0$5

substantially tighter than the CMB+SN+BAO constraint alone (Xu et al., 2011). This is why many viable UDF models either sit extremely close to $w\simeq 0$6CDM in the adiabatic sector or introduce entropy perturbations so that the physically relevant clustering sound speed remains near zero.

A useful corrective to a common misconception follows from this literature: the key observational constraint is not whether the background can interpolate between $w\simeq 0$7 and $w\simeq 0$8, but whether the fluid can do so while keeping its effective clustering sound speed sufficiently small. This is why the most successful recent UDF implementations are explicitly low-sound-speed models (Wang et al., 2024, Kou et al., 19 Sep 2025, su et al., 1 Apr 2025).

4. Nonlinear structure, spherical collapse, and thermodynamic critiques

Nonlinear collapse has mainly been studied in spherical top-hat form. For the constant-adiabatic-sound-speed UDF, the overdensity equations for each component $w\simeq 0$9 are written as

$w< -1/3$0

$w< -1/3$1

with $w< -1/3$2 for the UDF (Xu, 2013). Within that top-hat approximation, larger $w< -1/3$3 and larger $w< -1/3$4 make turnaround and collapse earlier.

A viscous extension replaces the effective pressure by

$w< -1/3$5

with background evolution

$w< -1/3$6

and effective sound speed $w< -1/3$7. In that spherical-collapse framework, larger $w< -1/3$8 and smaller $w< -1/3$9 produce earlier and faster collapse, while for $c_{s,\rm eff}^2=0$0 the nonlinear collapse curves are reported to be almost indistinguishable from $c_{s,\rm eff}^2=0$1CDM (Li et al., 2014).

The null-sound-speed UDF constructed as an alternative to phantom dark energy was also carried into the nonlinear regime. Using a local-Fermi-patch treatment and a top-hat radius equation,

$c_{s,\rm eff}^2=0$2

the model yields collapse thresholds and halo abundances very close to standard scenarios, with only a slight delay relative to smooth-DE CPL backgrounds because clustering dark energy can develop negative perturbations when $c_{s,\rm eff}^2=0$3 in the matched split description (Kou et al., 19 Sep 2025).

These nonlinear results should not be conflated with linear-scale clustering constraints. This suggests a useful distinction: top-hat studies suppress spatial gradients by construction, whereas linear Boltzmann analyses explicitly retain pressure support and Jeans suppression. The statement that larger $c_{s,\rm eff}^2=0$4 can accelerate collapse in the top-hat approximation therefore does not negate the linear result that nonzero $c_{s,\rm eff}^2=0$5 suppresses small-scale power (Xu, 2013, Xu et al., 2011).

Thermodynamic analyses introduce a separate critique. For the fast-transition UDF, the entropy of the apparent horizon satisfies the second law, but the first and second derivatives of the entropy display sharp oscillations around the transition epoch, raising doubts about the soundness of very abrupt UDF transitions (Radicella et al., 2014). This criticism is not a direct observational exclusion, but it has become part of the theoretical discussion surrounding fast-transition unified fluids.

5. Observational constraints and model comparison

The earliest detailed global constraint on a clustering UDF is the constant-sound-speed analysis of Xu et al. Using WMAP7, Union2 SN Ia, and BAO, the best fit is statistically indistinguishable from $c_{s,\rm eff}^2=0$6CDM, with

$c_{s,\rm eff}^2=0$7

versus $c_{s,\rm eff}^2=0$8 for $c_{s,\rm eff}^2=0$9CDM. The reported constraints are

$w(a)$00

$w(a)$01

with $w(a)$02. Once the SDSS DR7 power spectrum is included illustratively, the much stronger requirement $w(a)$03 emerges (Xu et al., 2011).

Allowing entropy perturbations shifts the emphasis from $w(a)$04 to $w(a)$05. For the constant-$w(a)$06 UDF with entropic perturbations, the combined WMAP7+BAO+Union2.1 analysis gives

$w(a)$07

$w(a)$08

$w(a)$09

again favoring CDM-like clustering and a background nearly degenerate with $w(a)$10CDM (Xu, 2012).

For the fast-transition UDF with entropic perturbations, the full WMAP7+BAO+SNe analysis yields

$w(a)$11

$w(a)$12

and a model-selection result

$w(a)$13

so the model is statistically equivalent to $w(a)$14CDM, but the data do not favor a fast transition (Yang et al., 2013).

Later phenomenological UDFs have been tested with more recent data and with differing conclusions about $w(a)$15. The one-parameter sinc model, constrained with Planck 2015, Pantheon, and cosmic chronometers, yields $w(a)$16 values in the $w(a)$17 range for CMB-inclusive fits and therefore alleviates the $w(a)$18 tension, but Bayesian evidence still favors $w(a)$19CDM over the UDF for all dataset combinations considered (Yang et al., 2019). By contrast, the pressure-parametrized UDF of 2017 gives

$w(a)$20

supporting the Planck global determination at $w(a)$21 rather than the higher local value (Wang et al., 2017).

Dissipative variants have also been constrained at background level. The 2021 dissipative UDF study reports

$w(a)$22

for the dissipative model and

$w(a)$23

for its non-dissipative counterpart. Although the dissipative model has the lower $w(a)$24, the non-dissipative UDF has the minimum AIC, with $w(a)$25 and $w(a)$26 relative to that best model (Elkhateeb et al., 2021).

The Planck-2018-era PAge-like UDF is more conservative. For CMB+BAO+SN+CC it gives

$w(a)$27

with $w(a)$28 and a baryon power spectrum differing by less than $w(a)$29 from $w(a)$30CDM at $w(a)$31. However, Bayesian evidence yields

$w(a)$32

for PUDF versus $w(a)$33CDM, indicating strong preference for $w(a)$34CDM in that analysis (Wang et al., 2024).

A 2025 follow-up re-examined this PAge-like model, corrected a Boltzmann-code inconsistency, and found that the primary CMB TT and EE spectra can agree with $w(a)$35CDM at the level

$w(a)$36

The updated comparison gives

$w(a)$37

so the preference for $w(a)$38CDM remains, but is milder than in the original 2024 analysis (su et al., 1 Apr 2025).

The most recent null-sound-speed UDF, matched to DESI DR2, Planck, and DES Y5 through a CPL-equivalent background, reaches perhaps the clearest statement of present observational status. Its best-fit $w(a)$39 values differ from CPL by less than $w(a)$40 for current data combinations, while a stage-IV linear forecast gives

$w(a)$41

This implies that even future linear probes may have only limited power to distinguish a clustering UDF from a split CPL dark sector when both share the same background expansion (Kou et al., 19 Sep 2025).

6. Microphysical realizations, modified gravity, and current status

UDF phenomenology has been embedded in several broader theoretical frameworks. One class uses scalar fields. The constant-sound-speed UDF can be mapped to a k-essence-type scalar field, and the pressure-parametrized UDF has been reconstructed with both quintessence and phantom scalar fields, with $w(a)$42 corresponding to the quintessence branch and $w(a)$43 to the phantom branch (Wang et al., 2017, Xu et al., 2011).

Another class places UDF behavior inside modified gravity. In generalized Brans–Dicke theory, the unified fluid is often written as an affine or wet-dark-fluid equation of state,

$w(a)$44

with

$w(a)$45

Within anisotropic backgrounds, the evolving Brans–Dicke scalar makes accelerated expansion possible even when linear relations among directional Hubble rates would forbid it in GR (Tripathy et al., 2014). A related 2020 study combines a linear UDF equation of state with a hybrid scale factor in generalized Brans–Dicke theory and derives explicit predictions for $w(a)$46, but also notes that the present rates are larger than the most stringent local bounds (Tripathy et al., 2020).

A more formal reconstruction program starts from a UDF background and derives an $w(a)$47 action that reproduces it. In the 2023 reconstruction of Elkhateeb’s UDF,

$w(a)$48

the resulting $w(a)$49 function satisfies the usual viability conditions $w(a)$50 and $w(a)$51 on the viable branch and can simultaneously support late-time de Sitter behavior and an inflationary regime with $w(a)$52 and $w(a)$53 compatible with Planck 2018 for $w(a)$54 (Elkhateeb, 2023).

A distinct microphysical proposal is the unified superfluid dark sector, in which two non-relativistic dark-matter superfluids interact through a Josephson/Rabi-type cosine potential. In the coarse-grained description,

$w(a)$55

so the system behaves as a dust component plus an emergent dark-energy-like term. Because the component sound speeds remain small, the model can track $w(a)$56CDM in expansion while predicting late-time growth-rate deviations at the $w(a)$57 level for the illustrative parameter set studied (Ferreira et al., 2018).

The theoretical status of UDF models is therefore mixed rather than uniform. On the one hand, they provide a clean language for dark degeneracy, admit multiple effective and microphysical realizations, and can fit present background and even linear-perturbation data well when $w(a)$58 is sufficiently small. On the other hand, current observations repeatedly drive them toward regimes that are either exactly or nearly $w(a)$59CDM-like: $w(a)$60, $w(a)$61, slow rather than sharp transitions, or explicit null-sound-speed prescriptions (Xu et al., 2011, Yang et al., 2013, Wang et al., 2024).

A final misconception is therefore worth dispelling. UDF is not primarily an alternative late-time background fit; many such fits are already degenerate with $w(a)$62CDM. Its real significance is methodological. It provides a framework in which the split of the dark sector is treated as contingent, and it identifies the effective sound speed, entropy sector, and nonlinear clustering as the principal observables that can break that degeneracy. This suggests that the decisive tests of UDF ideas will come not from background distances alone, but from full-shape large-scale structure, redshift-space distortions, weak lensing, CMB lensing, ISW-sensitive cross-correlations, and nonlinear halo statistics (Kou et al., 19 Sep 2025, su et al., 1 Apr 2025).