Entropy Density: Intensive Entropic Constructs

• Entropy Density is an intensive entropic quantity defined variably across contexts by normalizing entropy over natural supports such as volume, area, mode space, or qubit count.
• It appears in disciplines like relativistic thermodynamics, black-hole physics, and lattice gauge theory to relate geometric aspects and state variables with thermal characteristics.
• Its formulation employs local balance laws, variational functionals, and information-theoretic methods to bridge microscopic dynamics with macroscopic observables.

Entropy density denotes an intensive entropic quantity, but the object made intensive varies sharply across fields. In relativistic thermodynamics it can mean a local quantity such as $\sigma \equiv Ts = \rho+p$, introduced so that thermal content can be written directly in geometric form without first postulating a horizon or temperature (Yang, 2011). In black-hole thermodynamics it can mean entropy per unit horizon area, $\sigma_S \equiv S/A$ (Cai et al., 2021). In lattice gauge theory it usually means the volumetric thermodynamic density $s$, often reported as $s/T^3$ (Loan et al., 2023). Other literatures use the phrase for mode-resolved radiative entropy densities (Narayanaswamy et al., 2012), radial Shannon-integrand constructions derived from normalized nuclear densities (Ma et al., 2024), Rényi-2 entropy per qubit in noisy quantum circuits (Demarty et al., 2024), or asymptotic entropy per lattice area in combinatorial counting problems (Knauf et al., 2014). Taken together, these usages suggest that “entropy density” is not a single invariant concept but a family of intensive entropic constructions tied to the natural support of the problem—volume, area, mode space, particle number, qubit number, or lattice area.

1. Semantic range and principal definitions

A common misconception is that entropy density must always mean the local thermodynamic state variable $s(\mathbf{x},t)$ in the continuum-mechanics sense. The literature does not support that restriction. Several non-equivalent meanings coexist, and many of them are exact within their own frameworks.

Context	Representative quantity	Meaning
Spacetime thermodynamics	$\sigma=Ts=\rho+p$	Local thermal entropy density of matter/spacetime
Black-hole thermodynamics	$\sigma_S=S/A$	Entropy per unit horizon area
Lattice QCD	$s/T^3$	Normalized equilibrium entropy density
Near-field radiation	$s_h^{(j)}(\mu,z)$	Mode-resolved entropy density in a vacuum gap
Nuclear information theory	$s_r(r)=-p_W(r)\ln p_W(r)$	Radial Shannon-entropy integrand
Quantum circuits	$\sigma_S \equiv S/A$0	Rényi-2 entropy per qubit
Lattice triangulations	$\sigma_S \equiv S/A$1	Entropy per lattice area

These definitions are not merely notational variants. They differ in ontology. Some are local state functions, some are geometric ratios, some are pathwise irreversibility measures integrated over space, some are global functionals of density fields, and some are asymptotic growth constants. The distinction is explicit in several papers. For example, the inhomogeneous-fluid work defines a total entropy functional $\sigma_S \equiv S/A$2 rather than a pointwise field $\sigma_S \equiv S/A$3 (Schmidt, 26 Jun 2026), and the nuclear-density work states that its “entropy-density-like” object is only the Shannon integrand of a normalized radial probability distribution, not a thermodynamic entropy density (Ma et al., 2024).

2. Local balance laws and continuum thermodynamics

In kinetic theory, entropy density is often defined operationally by the existence of a local balance equation. Kadanoff formulates kinetic entropy through fields $\sigma_S \equiv S/A$4, $\sigma_S \equiv S/A$5, and $\sigma_S \equiv S/A$6 satisfying

$\sigma_S \equiv S/A$7

For the Boltzmann equation this gives

$\sigma_S \equiv S/A$8

while for Landau quasiparticles it yields

$\sigma_S \equiv S/A$9

with the standard fermionic form $s$0 for $s$1 (Kadanoff, 2014). The same paper also shows that this construction fails generically outside local thermodynamic equilibrium in nonequilibrium Green-function theory: the obstacle is not negativity of the collision term, but the impossibility of writing the dynamics as a pure local entropy-balance law (Kadanoff, 2014).

Granular fluids modify the classical structure further. In the Boltzmann description of inelastic granular gases, the entropy balance contains not only entropy density and entropy flux but also an entropy density rate or entropy supply term,

$s$2

with

$s$3

Near equilibrium the paper derives

$s$4

where $s$5 is the internal-energy density rate induced by inelastic collisions (Kremer, 2010). Entropy density remains a local state variable, but the balance law is no longer that of a simple fluid.

Near-field thermal radiation provides a different local continuum notion. There the entropy density is constructed by bosonic state counting over electromagnetic modes in the vacuum gap between two half-spaces. For photons from side $s$6, polarization $s$7, and mode-space coordinate $s$8, the mode-resolved entropy density is

$s$9

The total entropy density is obtained by integrating this over propagating and evanescent modes, and entropy flux is then formed with a mode-dependent transmission velocity rather than a universal factor $s/T^3$0 (Narayanaswamy et al., 2012). Here “density” refers simultaneously to physical space and mode space.

3. Geometrization, horizons, and cosmological expansion

One prominent use of the term is geometric. In “The thermal entropy density of spacetime,” thermal entropy density is defined locally from the first law for a perfect fluid in a small volume,

$s/T^3$1

which implies

$s/T^3$2

The paper then defines

$s/T^3$3

and identifies $s/T^3$4 as the thermal entropy density (Yang, 2011). Its central claim is methodological: $s/T^3$5 can be introduced in arbitrary spacetime without first assuming a horizon, a Hawking temperature, an Unruh temperature, or a special symmetry (Yang, 2011).

Combining Einstein’s equations with a $s/T^3$6 split, the same paper rewrites $s/T^3$7 purely geometrically: $s/T^3$8 and, after eliminating the four-dimensional Ricci scalar,

$s/T^3$9

This is the sense in which the paper says that thermal entropy density can be “geometrized” (Yang, 2011). The restricted setting is explicit: perfect fluids in Einstein gravity with cosmological constant, without viscosity, heat flux, anisotropic stress, or chemical-potential terms (Yang, 2011).

A second geometric usage appears in black-hole thermodynamics, where entropy density becomes surface density. The black-hole paper defines

$s(\mathbf{x},t)$0

so that in units $s(\mathbf{x},t)$1, the Bekenstein–Hawking law $s(\mathbf{x},t)$2 gives the baseline value $s(\mathbf{x},t)$3 (Cai et al., 2021). It further introduces the entropy per black-hole molecule,

$s(\mathbf{x},t)$4

and, with $s(\mathbf{x},t)$5, derives the proportionality

$s(\mathbf{x},t)$6

This supports a classification into strong, weak, and standard Bekenstein–Hawking black holes according as $s(\mathbf{x},t)$7 is greater than, less than, or equal to $s(\mathbf{x},t)$8 (Cai et al., 2021).

Cosmological FLRW work uses yet another volumetric notion. There the entropy density is defined by

$s(\mathbf{x},t)$9

and energy-momentum conservation implies

$\sigma=Ts=\rho+p$0

The same paper derives

$\sigma=Ts=\rho+p$1

and, in a Hubble-horizon thermodynamic construction, obtains

$\sigma=Ts=\rho+p$2

in natural units (Liu et al., 2010). This is again a local density, but now one tied to cosmological expansion rather than to local gravitational geometry or horizon area.

4. Density functionals, density fields, and fluctuation entropies

In several many-body theories, entropy density is replaced by entropy functionals of reduced density variables. “Entropy density functional theory for inhomogeneous fluids” is explicit on this point: its central object is a global entropy functional

$\sigma=Ts=\rho+p$3

where $\sigma=Ts=\rho+p$4 is the one-body density and $\sigma=Ts=\rho+p$5 is the global interparticle distance distribution (Schmidt, 26 Jun 2026). The excess part

$\sigma=Ts=\rho+p$6

is called universal because, for equilibrium systems with pairwise interactions, it is independent of the particular external field and pair potential; the latter enters only through the explicit energy term $\sigma=Ts=\rho+p$7 (Schmidt, 26 Jun 2026). This is not a theory of a local scalar $\sigma=Ts=\rho+p$8, but of an intrinsic total entropy functional whose variational derivatives define entropic direct correlation functionals (Schmidt, 26 Jun 2026).

A related inferential reformulation appears in entropic density functional theory for quantum fluids. There the expected density field

$\sigma=Ts=\rho+p$9

is imposed as information, and maximum entropy produces a density-parametrized trial state

$\sigma_S=S/A$0

The intrinsic functional

$\sigma_S=S/A$1

then emerges from entropy maximization, and the usual finite-temperature DFT Euler equation follows from stationarity with respect to $\sigma_S=S/A$2 (Yousefi et al., 2022). Here density is the constraint, entropy is the ranking principle, and the resulting object is again a functional rather than a local density field.

Field-theoretic nonequilibrium work pushes the shift further from state entropy to irreversibility. In stochastic density field theories, the central quantity is the trajectory-level entropy production

$\sigma_S=S/A$3

not a local equilibrium entropy density (Brossollet et al., 20 Jul 2025). For Dean’s exact equation and its coarse-grained generalizations, the paper derives closed expressions for $\sigma_S=S/A$4 and for the steady-state entropy production rate $\sigma_S=S/A$5, including cases with density-dependent diffusivity and active matter (Brossollet et al., 20 Jul 2025). A spatial integrand can then be read as a local entropy-production density, but the primary object is irreversibility of density trajectories.

Biomolecular solution theory provides an intermediate case. There the conformational entropy of a biomolecule is derived from a Gaussian distribution of atomic displacements, and the solvent contribution is defined at fixed pair positions,

$\sigma_S=S/A$6

A one-position quantity

$\sigma_S=S/A$7

is then used as a spatial decomposition (Hirata, 2023). The paper states that this should not be read as a strict local thermodynamic state function; it is a fluctuation-based spatial decomposition of entropy associated with solvent density correlations (Hirata, 2023).

5. Information-theoretic and computational meanings

In information-theoretic contexts, entropy density often means an entropy contribution per unit of representation rather than per unit physical volume. Nuclear-density analysis is explicit: starting from a normalized radial probability density

$\sigma_S=S/A$8

the total information entropy is

$\sigma_S=S/A$9

and the closest thing to an entropy density is the radial integrand

$s/T^3$0

The paper emphasizes that this is not thermodynamic entropy density in $s/T^3$1, but a Shannon-information measure of the spread, diffuseness, halo structure, neutron skin, or clustering of one-body nuclear densities (Ma et al., 2024).

Near-term quantum computing uses an even more abstract normalization. There entropy density is defined as the Rényi-2 entropy per qubit of the full output state: $s/T^3$2 For the maximally mixed $s/T^3$3-qubit state, this density equals $s/T^3$4. The paper models entropy accumulation under noisy hardware-efficient circuits with a global depolarizing ansatz and uses the resulting entropy density to bound circuit-depth thresholds beyond which quantum advantage is unattainable (Demarty et al., 2024). The normalization is by qubit number, not by spatial volume or subsystem area.

High-rate scalar quantization uses “entropy density” in yet another sense. Under Rényi-entropy-constrained quantization, the asymptotic entropy contribution of an interval is governed by a measure with density proportional to

$s/T^3$5

The corresponding normalized measure

$s/T^3$6

is the paper’s entropy density, and it generalizes classical point density from fixed-rate quantization to Rényi-entropy-constrained quantization (Kreitmeier et al., 2011). This is not a thermodynamic quantity at all; it is a local allocation law for coding resources.

The density-estimation approach to nonequilibrium Ising ensembles is a useful counterexample. It estimates the global configurational entropy

$s/T^3$7

from a learned density model $s/T^3$8, but does not explicitly define an entropy density (Gelman et al., 2024). A plausible implication is that one could normalize by system size to obtain an entropy per site, but the paper itself does not carry out that step (Gelman et al., 2024).

6. Intensive entropies in many-body matter and discrete structures

In equilibrium many-body physics, the most conventional usage remains thermodynamic entropy per volume. In SU(3) lattice QCD, the entropy density is computed directly from the expectation value of the improved energy-momentum tensor with shifted temporal boundary conditions and reported as $s/T^3$9. Over $s_h^{(j)}(\mu,z)$0, the continuum values rise rapidly just above $s_h^{(j)}(\mu,z)$1 and then more slowly, reaching $s_h^{(j)}(\mu,z)$2 at $s_h^{(j)}(\mu,z)$3; at about $s_h^{(j)}(\mu,z)$4, the entropy density is still roughly $s_h^{(j)}(\mu,z)$5 below the Stefan–Boltzmann limit (Loan et al., 2023). Here entropy density is the standard equation-of-state quantity.

A related but conceptually different intensive use appears in molecular-fluid thermodynamics. “Connecting Entropy Scaling and Density Scaling” is careful that its central object is not $s_h^{(j)}(\mu,z)$6, but the residual molar entropy

$s_h^{(j)}(\mu,z)$7

The paper argues that this residual entropy is “mostly synonymous” with the variable used in density scaling for the systems studied (Bell et al., 2022). This is not entropy density in the literal volumetric sense, but an interaction-induced entropy deficit at fixed $s_h^{(j)}(\mu,z)$8 and $s_h^{(j)}(\mu,z)$9.

Multiple-occupancy cell models introduce yet another normalization. The cell-fluid paper distinguishes entropy per particle,

$s_r(r)=-p_W(r)\ln p_W(r)$0

from entropy per cell,

$s_r(r)=-p_W(r)\ln p_W(r)$1

with $s_r(r)=-p_W(r)\ln p_W(r)$2 the average occupancy per cell (Romanik et al., 26 Oct 2025). At low temperature, both $s_r(r)=-p_W(r)\ln p_W(r)$3 and $s_r(r)=-p_W(r)\ln p_W(r)$4 develop minima near integer-valued densities, which the paper interprets as a generic feature of multiple-occupancy models (Romanik et al., 26 Oct 2025).

Discrete combinatorics uses the phrase in an asymptotic counting sense. For unimodular lattice triangulations of the integer rectangle $s_r(r)=-p_W(r)\ln p_W(r)$5, the entropy density or capacity is

$s_r(r)=-p_W(r)\ln p_W(r)$6

where $s_r(r)=-p_W(r)\ln p_W(r)$7 is the number of triangulations. Wang–Landau density-of-states estimation yields the square-lattice asymptotic value

$s_r(r)=-p_W(r)\ln p_W(r)$8

so that $s_r(r)=-p_W(r)\ln p_W(r)$9 up to subexponential corrections (Knauf et al., 2014). “Density” here means per unit lattice area in an exponential counting law.

Heavy-ion phenomenology illustrates how terminological drift can occur. In the heavy-ion paper, the formal local entropy density is $\sigma_S \equiv S/A$00, but the reported numerical values such as $\sigma_S \equiv S/A$01 for central Pb+Pb at $\sigma_S \equiv S/A$02 TeV and $\sigma_S \equiv S/A$03 for $\sigma_S \equiv S/A$04 TeV are values of the specific entropy $\sigma_S \equiv S/A$05, not volumetric entropy densities (Lukács et al., 2011). That paper is therefore a reminder that the phrase “entropy density” may sometimes be used loosely for entropy normalized by conserved particle number.

Across these literatures, entropy density functions less as a single concept than as a normalization principle. The normalized carrier may be a spatial volume, a horizon area, a qubit register, a lattice area, a cell, a conditional interval in source space, or a reduced structural descriptor. The technical content therefore lies not in the word “density” alone, but in the choice of support and in the dynamical or variational structure that makes the corresponding entropy intensive.