---
title: Thermodynamic Distance to Equipartition
url: https://www.emergentmind.com/topics/thermodynamic-distance-to-equipartition
type: topic
---

# Thermodynamic Distance to Equipartition

Thermodynamic distance to equipartition denotes a family of quantities that measure deviation from an equipartition reference state, but the reference itself is system-dependent. In driven granular gases it is the difference between large-scale and particle-scale transverse kinetic temperatures, \(D=T_b-T_g\); in macroscopic oscillators under heat flux it is expressed through departures of modal effective temperatures from a physical temperature, through RMS mode deviations, or through Gaussian Kullback–Leibler divergences; in nonequilibrium statistical mechanics it is the Kullback–Leibler divergence to the uniform distribution, \(D_{\mathrm{eq}}(p)=\ln N-S(p)\); in open quantum quadratic systems it is the deviation of per-degree energies and covariances from their classical \(k_BT/2\) limits; and in discrete, spatial, or holographic settings it appears as finite-difference residuals, local-energy heterogeneity, or degrees-of-freedom mismatch [2002.05562] [1305.5084] [2509.09041] [2309.14582] [2407.07772] [2112.06359].

## 1. Equipartition as a reference concept

Equipartition is not a single statement across all branches of statistical physics. In equilibrium fluids and harmonic systems it usually means that each quadratic degree of freedom contributes \(k_BT/2\) to the mean energy. In transverse hydrodynamic spectra it means that the static transverse velocity spectrum is flat in wave number \(k\), so the kinetic energy per transverse mode does not depend on \(k\). In information-theoretic formulations it means the uniform distribution over \(N\) states. In specific-heat formulations it is the high-temperature plateau \(C(T)\to C_{\mathrm{eq}}\). In holographic cosmology it is realized through equality between surface and bulk degrees of freedom in a de Sitter limit [2002.05562] [2509.09041] [2001.03007] [2112.06359].

| Context | Equipartition reference | Distance quantity |
|---|---|---|
| Driven granular gas | Flat \(J_t(k,0)\), hence \(T_b=T_g\) | \(D=T_b-T_g\) |
| Oscillatory nonequilibrium solids | \(T_{\mathrm{eff},j}=T\) or \(T^{\mathrm{fluc}}=T_{\mathrm{ref}}\) | \(D_j\), \(D_{\mathrm{RMS}}\), \(D^{(\mathrm{ref})}\) |
| Distributional nonequilibrium thermodynamics | Uniform law \(u_i=1/N\) | \(D_{\mathrm{eq}}=D_{\mathrm{KL}}(p\Vert u)\) |
| Open quantum quadratic systems | \(E_i=k_BT/2\) per quadratic degree | \(\Delta E_i\), \(r_i\), \(\Delta P_u\), \(\Delta Q\) |
| Specific heat crossover | \(C(T)=C_{\mathrm{eq}}\) | \(\Delta C(T)\), \(\delta(T)\) |

This range of usages indicates that thermodynamic distance to equipartition is a relational notion: it always compares an observed state to a chosen equipartition baseline rather than naming a unique universal scalar. A plausible implication is that comparisons across subfields require first specifying the degrees of freedom, the reference ensemble, and the observable used to define equipartition.

## 2. Spectral and mode-resolved distances in nonequilibrium experiments

In the driven granular-gas experiment of magnetized particles confined to a thin layer, equipartition is defined in the transverse sector by a \(k\)-independent static transverse velocity spectrum. The relevant quantities are
\[
T_b \equiv J_t(k\to 0,0),\qquad
T_g \equiv J_t(k\simeq 2\pi/a,0),\qquad
T_e=\frac{E_k}{m},
\]
and the thermodynamic distance to equipartition is
\[
D \equiv T_b-T_g.
\]
The static transverse structure factor is fitted by
\[
J_t(k,0)=T_g+\frac{T_b-T_g}{1+\xi^2k^2},\qquad \xi=\sqrt{\frac{\nu}{\gamma_b}},
\]
while the decay rate obeys
\[
\tau^{-1}(k)=\nu k^2+\gamma_b.
\]
As the magnetic control parameter suppresses the collision rate according to
\[
f_c(\varepsilon)\simeq 28~\mathrm{s}^{-1}\exp(-0.25\,\varepsilon),
\]
the measured \(D\) decreases essentially monotonically and becomes negligible as \(f_c\to0\); around \(\varepsilon\approx 13.5\), \(D\approx0\), the velocity statistics become nearly Gaussian, and the transverse spectrum becomes nearly flat. For \(\varepsilon>13.5\), however, \(T_g>T_b\), signaling a regime change and the limit of validity for the hydrodynamic description [2002.05562].

In macroscopic oscillators subject to steady heat flux, the relevant departure from equipartition is modal rather than spectral. For a harmonic normal mode,
\[
T_{\mathrm{eff},j}\equiv \frac{E_j}{k_B},
\]
and a mode-level relative distance is defined by
\[
D_j\equiv \frac{|T_{\mathrm{eff},j}-T|}{T}.
\]
The same work also introduces a system-level RMS measure,
\[
D_{\mathrm{RMS}}\equiv \sqrt{\frac{1}{N}\sum_{j=1}^N\left(\frac{T_{\mathrm{eff},j}-T}{T}\right)^2},
\]
and, for Gaussian marginals, an information-theoretic distance
\[
D_{\mathrm{KL}}=\frac12\left(\frac{\sigma_{\mathrm{EQ}}^2}{\sigma_{\mathrm{NEQ}}^2}-1-\ln\frac{\sigma_{\mathrm{EQ}}^2}{\sigma_{\mathrm{NEQ}}^2}\right).
\]
Experimentally, a relative temperature difference of about \(4\%\) produces a longitudinal-mode effective temperature about \(20\%\) above equilibrium and a transverse-mode effective temperature about \(3.5\) times the equilibrium value. The proposed mechanism is the emergence of flux-mediated correlations such as \(\langle x_i v_k\rangle\neq0\), which are absent at equilibrium and couple the heat flux directly to modal variance enhancement [1305.5084].

A closely related but mechanistically distinct formulation arises in the heated micro-cantilever experiment. There, mode-resolved fluctuation temperatures are defined by
\[
\frac12 k_n\langle \delta_n^2\rangle=\frac12 k_B T_n^{\mathrm{fluc}},\qquad
\frac12 \kappa_m\langle \theta_m^2\rangle=\frac12 k_B T_m^{\mathrm{fluc}},
\]
with extended equipartition
\[
T_{n,m}^{\mathrm{fluc}}=\int_0^1 dx\,T(x)\,w_{n,m}^{\mathrm{diss}}(x).
\]
The paper then proposes mode-resolved distances
\[
\Delta_n^{(\mathrm{ref})}=T_n^{\mathrm{fluc}}-T_{\mathrm{ref}},\qquad
\rho_n^{(\mathrm{ref})}=\frac{T_n^{\mathrm{fluc}}}{T_{\mathrm{ref}}}-1,
\]
and an aggregate distance
\[
D^{(\mathrm{ref})}=\left[\sum_n w_n\left(\frac{T_n^{\mathrm{fluc}}-T_{\mathrm{ref}}}{T_{\mathrm{ref}}}\right)^2\right]^{1/2}.
\]
Because dissipation is localized at the clamp, \(T^{\mathrm{fluc}}_{n,m}\approx T^{\mathrm{amb}}\) even when the tip reaches about \(1000~\mathrm{K}\), so the distance to \(T^{\mathrm{avg}}\) is large whereas the distance to the base temperature is small [2002.04488].

## 3. Information-theoretic and variational formulations

The most explicit abstract definition of thermodynamic distance to equipartition is the nonequilibrium construction based on the effective number of accessible states,
\[
\Omega_{\mathrm{eff}}(p)=\prod_{i=1}^N p_i^{-p_i}=e^{S(p)},
\qquad
S(p)=-\sum_{i=1}^N p_i\ln p_i.
\]
With the uniform reference \(u_i=1/N\), the distance is
\[
D_{\mathrm{eq}}(p)=D_{\mathrm{KL}}(p\Vert u)
=\sum_{i=1}^N p_i\ln\frac{p_i}{1/N}
=\ln N-S(p)
=\ln\frac{N}{\Omega_{\mathrm{eff}}(p)}.
\]
It vanishes at equipartition, reaches \(\ln N\) at complete localization, and decreases monotonically under doubly stochastic relaxation. The same formulation supplies a statistical distinguishability bound,
\[
\|p-u\|_1\le \sqrt{2\,D_{\mathrm{KL}}(p\Vert u)},
\]
and a near-equipartition expansion
\[
D_{\mathrm{eq}}(p)\simeq \frac12\,\chi^2(p\Vert u).
\]
For a fixed reference \(q\), the divergence
\[
D(p\Vert q)=\sum_i p_i\ln\frac{p_i}{q_i}
\]
splits into an entropy deficit plus a reference-weight coupling, and for a canonical reference \(\pi_i=e^{-\beta E_i}/Z\) becomes
\[
D(p\Vert \pi)=\beta\big(F[p]-F[\pi]\big).
\]
The same paper identifies the nonadiabatic entropy-production rate with the decay of the divergence,
\[
\frac{d}{dt}D_\pi(p(t))=-\dot e_p^{\mathrm{na}}(t)\le0,
\]
so the distance to equipartition becomes a Lyapunov functional whenever the reference is fixed [2509.09041].

A dynamically grounded but less abstract notion appears in the dynamical Lorentz gas. There, the moving particle equilibrates to a Maxwell–Boltzmann distribution at a temperature determined by generalized equipartition,
\[
\frac12\langle p_i^2\rangle_T=\left\langle \frac{P^2}{2M}\right\rangle_\eta,\qquad
k_B T=\frac{\overline{(\eta')^2P^2}}{M\,\overline{(\eta')^2}}.
\]
Distance to equipartition is then defined operationally as a distance between the instantaneous momentum density \(f_t(p)\) and the Maxwell–Boltzmann law \(f_{\mathrm{MB}}(p;T)\), for example through
\[
D_{\mathrm{KL}}(f_t\Vert f_{\mathrm{MB}}),\qquad
\|f_t-f_{\mathrm{MB}}\|_1,
\]
or a Wasserstein distance. The paper links the approach to equipartition to a Fokker–Planck dynamics in a scaled variable \(\xi\), for which the KL divergence to the stationary law decreases monotonically with collision number [1009.3706].

These formulations differ from mode- or spectrum-based distances because the reference object is a full probability distribution rather than a temperature surrogate. This suggests that information-theoretic distances are the natural setting when equipartition is interpreted as maximal configurational spread or as canonical stationarity rather than as equality of selected modal energies.

## 4. Quantum, discrete, and generalized equilibrium formulations

For arbitrary quadratic systems of multimode Brownian oscillators coupled to multiple reservoirs at the same temperature, the generalized quantum equipartition theorem takes the form
\[
\frac12\left\langle \left\{X_i,\frac{\partial H_{\mathrm{BO}}}{\partial X_j}\right\}\right\rangle
=
\int_0^\infty d\omega\, P_{ij}(\omega)\,\frac{\hbar\omega}{2}\coth\!\left(\frac{\beta\hbar\omega}{2}\right),
\]
with normalized, positive spectral weights \(P_{ii}(\omega)\). The per-degree energy is
\[
E_i(T)=\int_0^\infty d\omega\,P_{ii}(\omega)\,\frac{\hbar\omega}{4}\coth\!\left(\frac{\beta\hbar\omega}{2}\right),
\]
and the energy-based distance to classical equipartition is
\[
\Delta E_i(T)=E_i(T)-\frac{k_B T}{2},\qquad
r_i(T)=\frac{E_i(T)}{k_B T/2}-1.
\]
The same work defines covariance distances,
\[
\Delta P_u(T)=\langle \hat P_u^2\rangle-\frac{k_B T}{\Omega_u},\qquad
\Delta Q(T)=\langle \hat Q\hat Q\rangle-k_B T\left[V-\sum_\alpha \tilde\phi_\alpha(0)\right]^{-1}.
\]
All these distances vanish in the high-temperature limit and remain finite at low temperature because the zero-point contribution
\[
E_i(0)=\int_0^\infty d\omega\,P_{ii}(\omega)\,\frac{\hbar\omega}{4}
\]
does not disappear [2309.14582].

In discrete systems, the relevant structure is a finite-difference version of the conjugate variables theorem. For a discrete degree of freedom \(i\),
\[
\big\langle \Delta_i^- \omega_i \big\rangle
=
\big\langle \omega_i \left[1-e^{-\beta\,\Delta_i^+H(s)}\right]\big\rangle
+\text{(boundary}_i\text{)}.
\]
If \(\omega_i\) is chosen so that \(\Delta_i^-\omega_i=1\) in the interior and the boundary term vanishes, one obtains an equipartition-like identity
\[
1=
\big\langle \omega_i(s)\left[1-e^{-\Delta_i^+H(s)/(k_B T)}\right]\big\rangle.
\]
The paper then defines exact residuals
\[
r_i(T)=\left\langle \omega_i(s)\left[1-e^{-\Delta_i^+H(s)/(k_B T)}\right]\right\rangle-1,
\]
and aggregate distances
\[
D_{\mathrm{L2}}(T)=\left(\sum_i w_i r_i(T)^2\right)^{1/2},\qquad
D_{\mathrm{L1}}(T)=\sum_i w_i |r_i(T)|.
\]
A divergence-based alternative is
\[
D_{\mathrm{KL}}\big(p\Vert p_{\mathrm{eqp}}\big)
=
\sum_{s\in\Omega}p(s)\ln\frac{p(s)}{p_{\mathrm{eqp}}(s)},
\]
where \(p_{\mathrm{eqp}}\) is the canonical distribution at the temperature estimated from the discrete equipartition identity. In the small-step limit, the discrete formula recovers the continuous result
\[
\left\langle x_i\frac{\partial H}{\partial x_i}\right\rangle=k_B T
\]
[2407.07772].

Taken together, the quantum and discrete formulations show that distance to equipartition is not restricted to classical variance-based thermometry. It can be defined spectrally, algebraically, or distributionally, and it remains meaningful even when quantum fluctuations or discrete state space prevent the naive use of \(k_BT/2\) as an exact per-degree benchmark.

## 5. Bounds, ceilings, and crossover measures

One line of work treats the distance to equipartition as a deviation of the dimensionless equipartition factor
\[
\zeta\equiv \beta \bar E=\frac{\bar E}{k_B T}
\]
from a chosen reference. For \(N\) quadratic degrees of freedom, the classical reference is
\[
\zeta_{\mathrm{eq,clas}}=\frac{N}{2},
\]
whereas the proposed equipartition of energy bound is
\[
\zeta\le \frac{\pi^2}{2}\approx 4.9348.
\]
This yields several scalar distances,
\[
D_{\mathrm{abs}}=|\zeta-\zeta_{\mathrm{eq,clas}}|,\qquad
D_{\mathrm{rel}}=\left|\frac{\zeta}{\zeta_{\mathrm{eq,clas}}}-1\right|,\qquad
D_{\mathrm{bound}}=\zeta_{\mathrm{EEB}}-\zeta.
\]
The same framework ties relaxation toward equipartition to the lower bound
\[
\tau_{\mathrm{relax}}\ge \frac{\hbar}{\pi k_B T},
\]
so the admissible rate of change of any such distance is itself thermodynamically constrained [1109.4384].

A thermally complementary formulation uses specific heat instead of modal or distributional observables. There the absolute and relative distances are
\[
\Delta C(T)=C_{\mathrm{eq}}-C(T),\qquad
\delta(T)=1-\frac{C(T)}{C_{\mathrm{eq}}},
\]
and the slope-based quantity
\[
S(T)=-\frac{dC}{dT}
\]
measures how rapidly the system freezes out when cooled. For Debye solids, the equipartition temperature is conventionally \(T_C=\Theta_D\), at which
\[
\frac{C_D(T_C)}{C_{\mathrm{eq}}}\approx 0.952,
\]
while the fastest frozen temperature \(\theta\) is defined by the maximum of \(dC/dT\). Solving the Debye extremum condition gives
\[
\frac{\theta}{\Theta_D}\approx 0.164,
\]
and at that point the per-mode specific heat is about \(0.257\,k_B\). In this usage, distance to equipartition is not a dynamical nonequilibrium observable but a static thermodynamic deficit from the high-temperature plateau [2001.03007].

A different equilibrium variant appears in spatially coupled Gaussian systems. Local interaction energies satisfy
\[
\langle \epsilon_\alpha\rangle=\frac12 k_B T\,P_{\alpha\alpha},\qquad 0\le P_{\alpha\alpha}\le1,
\]
hence the universal bound
\[
\langle \epsilon_\alpha\rangle\le \frac12 k_B T.
\]
The recommended heterogeneity metrics are
\[
D_\infty^{(\mathrm{int})}=\max_\alpha |\langle \epsilon_\alpha\rangle-\langle \epsilon\rangle_{\mathrm{eq}}|,
\qquad
D_2^{(\mathrm{int})}=\left[\sum_\alpha (\langle \epsilon_\alpha\rangle-\langle \epsilon\rangle_{\mathrm{eq}})^2\right]^{1/2},
\]
with
\[
\langle \epsilon\rangle_{\mathrm{eq}}=\frac12 k_B T\,\frac{\mathrm{rank}\,H}{n}.
\]
In one-dimensional short-range systems these distances vanish as \(N\to\infty\), whereas in higher dimensions they remain finite because \(n>N\) and nontrivial spatial patterns persist even at equilibrium [1503.02325].

These bound- and crossover-based formulations broaden the meaning of thermodynamic distance to equipartition beyond “effective temperature differences.” They emphasize ceilings, deficits, and spatial heterogeneity rather than simple modal mismatch.

## 6. Regimes of validity, misconceptions, and broader significance

A persistent misconception is that departure from equipartition can always be represented by a single effective temperature. The cited literature shows otherwise. In the granular gas, \(T_b\), \(T_e\), and \(T_g\) are distinct, with \(T_b>T_e>T_g\) at moderate \(\varepsilon\), and the hydrodynamic fit fails once the spectral slope inverts and crystallization is approached. In the oscillator under heat flux, modal \(T_{\mathrm{eff},j}\) can exceed the hottest physical temperature in the body, so it ceases to be a thermometer for local thermodynamic temperature. In the quantum quadratic problem, low-temperature distances remain nonzero because zero-point energy survives even when the classical reference \(k_BT/2\) tends to zero. In spatially coupled equilibrium systems, normal-mode equipartition does not imply uniform spatial energy density [2002.05562] [1305.5084] [2309.14582] [1503.02325].

The regime of validity of each metric is correspondingly narrow. Granular hydrodynamics works well across the granular-gas phase up to roughly \(\varepsilon\approx10\text{–}17\), but not once strong magnetic repulsion, vertical motion, and hexagonal ordering dominate. The heat-flux modal analysis assumes classical low-frequency oscillators and interprets deviations through flux-mediated correlations. The micro-cantilever formulation presumes local thermal equilibrium, high-\(Q\) modes, and dissipation localization. The discrete equipartition identities require explicit treatment of boundary terms. The information-theoretic distance \(D_{\mathrm{eq}}\) is exact only after a reference distribution has been specified, and its monotonicity requires doubly stochastic relaxation or a fixed stationary reference [2002.05562] [1305.5084] [2002.04488] [2407.07772] [2509.09041].

The concept has also been exported to more exotic domains. In black-hole thermodynamics with non-Gaussian entropies, the standard Bekenstein–Hawking relation
\[
M=\frac12 NT
\]
is replaced by entropy-dependent generalized equipartition laws, such as
\[
M=\frac{\delta}{2}NT
\]
for the Tsallis–Cirto entropy, together with parameter-dependent factors for modified Rényi and Sharma–Mittal entropies. The corresponding “distance” is encoded in the ratio \(R=M/[(1/2)NT]\), while heat-capacity singularities mark phase transitions between different thermodynamic regimes [2002.02435].

In holographic cosmology, the paper does not define a thermodynamic distance explicitly, but it proposes natural candidates. A macroscopic one is the degrees-of-freedom mismatch
\[
D_1^{\mathrm{frac}}=\frac{|N_{\mathrm{sur}}-N_{\mathrm{bulk}}|}{N_{\mathrm{sur}}},
\]
which vanishes in de Sitter equipartition. A microscopic one is the relative fluctuation of horizon energy,
\[
D_2=\frac{\sigma_{\mathcal E_H}}{\langle \mathcal E_H\rangle}
\approx \frac{T_H}{T_P},
\]
with present-epoch magnitude of order \(10^{-62}\). This suggests that equipartition can serve as a unifying reference even when the relevant “degrees of freedom” are horizon bits rather than particles, modes, or lattice variables [2112.06359].

Across these usages, thermodynamic distance to equipartition is best understood not as a single invariant but as a structured comparison between an observed energy distribution and an equipartition benchmark appropriate to the underlying coarse graining. Its value lies precisely in making that benchmark explicit: flat spectra versus sloped spectra, modal thermometry versus physical temperature, uniform distributions versus localized ones, classical \(k_BT/2\) versus quantum spectral averages, or equal surface and bulk degrees of freedom versus their mismatch.

Source: https://www.emergentmind.com/topics/thermodynamic-distance-to-equipartition