---
title: 'Coupled Entropy: Concepts & Applications'
url: https://www.emergentmind.com/topics/coupled-entropy
type: topic
---

# Coupled Entropy: Concepts & Applications

In the cited literature, “coupled entropy” is not a single standardized object. It denotes, in different subfields, the pair-excess entropy that links static structure to transport in strongly coupled liquids, a family of nonadditive entropies built from a nonlinear coupling parameter \(\kappa\), entropy production and information flow in systems strongly coupled to hidden variables or baths, and experimentally inferred entropy changes of coupled quantum dots [1608.08394] [2409.10645] [2506.17229] [2605.04493] [1703.03764] [2011.08187] [2508.09481] [2605.03876]. A recurrent source of confusion is that “coupling” refers to different structures in these works: interparticle correlations, nonlinear statistical coupling, strong system–bath hybridization, or probabilistic coupling of marginals. This suggests that the term must be interpreted relative to the underlying dynamical or probabilistic model.

## 1. Scope of the term

Across the cited works, four uses are technically central. In strongly coupled Yukawa fluids, entropy is an excess-entropy functional of the radial distribution function and is used to organize diffusion and freezing. In nonlinear statistical mechanics, coupled entropy is a generalized entropy built from a coupled logarithm and escort expectations, with maximizing distributions in a coupled stretched-exponential family. In nonequilibrium strong-coupling thermodynamics, entropy is defined through Hamiltonians of mean force, reduced density matrices, or scattering-state currents, and the emphasis is on entropy production, fluctuation relations, and information flow. In mesoscopic experiments, entropy changes are extracted from Maxwell relations and charge sensing in coupled quantum dots [1608.08394] [2605.04493] [1703.03764] [2508.09481].

| Context | Definition used in the literature | Representative quantity |
|---|---|---|
| Strongly coupled liquids | pair-excess entropy \(s_2\) or excess entropy \(s_{\rm ex}\) | \(D^*=0.04\,e^{s_2}\) |
| Nonlinear statistical mechanics | coupled entropy \(H_\kappa\) or \(S_\kappa\) | \(\kappa\)-sum composability |
| Strong system–bath coupling | stochastic or reduced-state entropy | generalized Crooks relation |
| Coupled quantum dots | entropy change from Maxwell relation | \(k_B\ln 2\), \(\ln 3\), \(\ln 4\) features |

A distinct information-theoretic usage concerns minimum-entropy coupling. There, given marginals \(p_1,\ldots,p_m\), the minimum-entropy coupling is the coupling \(X_1,\ldots,X_m\) with smallest joint Shannon entropy,
\[
H^*(\{p_i\}) := \inf_{(X_1,\ldots,X_m)\in \Gamma(\{p_i\})} H(X_1,\ldots,X_m),
\]
and an efficient construction achieves entropy within \(2\) bits of the entropy of the greatest lower bound under majorization [2006.07955]. This is a different use of “coupling” from the strong-correlation and nonadditive-entropy literatures.

## 2. Pair-excess entropy in strongly coupled liquids

For strongly coupled Yukawa liquids, the relevant entropy is the excess entropy relative to an ideal gas at the same \(T\) and density. The total excess entropy per particle is \(s_{\rm ex}\equiv S_{\rm ex}/N\), and in the cited molecular-dynamics study it is obtained by thermodynamic integration of the excess internal energy \(u(\Gamma)\) along the coupling parameter \(\Gamma=Q^2/(4\pi\epsilon_0 a k_BT)\):
\[
s_{\Gamma\leftarrow\Gamma_0}
=u(\Gamma)\,\Gamma-u(\Gamma_0)\,\Gamma_0-\int_{\Gamma_0}^{\Gamma}u(\Gamma')\,d\Gamma' .
\]
The leading structural term is the pair-excess entropy
\[
s_2=-\tfrac12\,n\int\bigl[g(r)\ln g(r)-g(r)+1\bigr]\,d^3r,
\]
with \(g(r)\) the radial distribution function and \(n\) the number density [1608.08394].

In all cases studied, with screening parameter \(\kappa=a/\lambda_D\) from \(1\) to \(4\), the MD data collapse onto
\[
s_2=-\,3.285\;\bigl(\Gamma/\Gamma_m\bigr)^{0.665}
\quad\Longleftrightarrow\quad
s_2=-\,3.285\;\bigl(T_m/T\bigr)^{0.665},
\]
where \(\Gamma_m\) or \(T_m\) is the coupling or temperature at the liquid–solid freezing point. Near melting,
\[
s_2 \approx 0.9\,s_{\rm ex},
\]
so about \(90\%\) of the total excess entropy comes from two-body correlations alone. The physical interpretation given in the paper is that, close to freezing, the liquid develops strong short-range order and nearest-neighbor “cages,” making higher-order many-body contributions relatively unimportant [1608.08394].

The same study identifies a transport scaling in the caged regime. With Enskog collision frequency
\[
\nu_E=4\,\sigma^2\,g(\sigma)\,n\,\sqrt{\pi k_BT/m},
\]
and effective hard-sphere diameter \(\sigma\) taken as the position of the first \(g(r)\) peak, the reduced diffusion coefficient
\[
D^*=\frac{D}{\sigma^2\nu_E}
\]
obeys
\[
D^*=0.04\,e^{s_2}
\]
at sufficiently low \(T\) and large \(\Gamma\), where particles remain trapped in local cages until rearrangement. At higher \(T\), specifically when \(-s_2<1\), cages dissolve rapidly and diffusion deviates from the exponential law [1608.08394].

A later analysis of strongly coupled Yukawa fluids broadened this framework from \(s_2\) to the full excess entropy \(s_{\rm ex}\). In a vibrational paradigm of dense fluids,
\[
s_{\rm ex}= \frac32-\frac32 \left\langle \ln\!\frac{m\Delta^2\omega^2}{2\pi T}\right\rangle,
\]
with \(\Delta=\rho^{-1/3}\), and the mode average is expressed through longitudinal and transverse dispersions within the quasi-localized charge approximation plus an excluded-volume RDF \(g(r)=\Theta(r-Ra)\). The paper reports that this estimate reproduces existing MD data for \(\kappa\approx 0\)–\(4\) without adjustable parameters [2409.10645].

At freezing, direct MD gives
\[
s_{\rm ex}^{\rm fr}(\kappa)\simeq -4.1+0.096\,\kappa
\qquad (\text{valid for }\kappa\lesssim 4),
\]
so \(s_{\rm ex}^{\rm fr}\approx -4\) is nearly constant over a wide range of screening. Over the entire liquid regime, the paper proposes the modified Rosenfeld–Tarazona form
\[
s_{\rm ex}\simeq s_{\rm ex}^{\rm fr}\,(\Gamma/\Gamma_{\rm fr})^{1/2}
\quad\Longleftrightarrow\quad
s_{\rm ex}\simeq s_{\rm ex}^{\rm fr}\,(T_m/T)^{1/2},
\]
which fits all \(\kappa\le 4\) data to within a few percent and gives the correct limits at \(\Gamma=0\) and \(\Gamma=\Gamma_{\rm fr}\). The same work places the Frenkel-line crossover near
\[
s_{\rm ex}\simeq -0.9\pm 0.1,
\]
and reports transport scalings
\[
D_R \simeq 1.1\,\exp[0.9\,s_{\rm ex}], \qquad
\eta_R \simeq 0.13\,\exp[-0.9\,s_{\rm ex}],
\]
with \(D_R\eta_R\simeq 0.14\) [2409.10645].

The liquid-state use of coupled entropy is therefore structural: entropy is a reduced description of many-body order that collapses diffusion, viscosity, and freezing behavior. The papers further state that the same ideas may apply to strongly coupled dusty plasmas, charged colloids, colloidal suspensions, electrolytes, and other systems with soft pairwise interactions, although that extension is conjectural or quasiuniversal rather than exact [1608.08394] [2409.10645].

## 3. Generalized nonadditive entropy and nonlinear statistical coupling

A different line of work uses “coupled entropy” to denote a nonadditive entropy family for systems with nonlinear coupling. An early construction starts from the weighted generalized mean in the probability domain,
\[
P_{\rm avg}^{(r)}=\Bigl(\sum_i p_i^{1+r}\Bigr)^{1/r},
\]
so that the Shannon case corresponds to the weighted geometric mean, \(P_{\rm avg}^{(0)}=e^{-H}\). With \(q=1+\kappa\), Tsallis entropy becomes
\[
S_{1+\kappa}(p)=\frac{1-\sum_i p_i^{1+\kappa}}{\kappa}
=-\ln_\kappa\!\bigl(P_{\rm avg}^{(\kappa)}\bigr),
\]
where
\[
\ln_\kappa(x)=\frac{x^\kappa-1}{\kappa}.
\]
The proposed coupled entropy is the normalized Tsallis entropy divided by one plus the coupling,
\[
E_{\rm coupled}(p)
=-\sum_i p_i^{1+\kappa}\ln_\kappa(p_i)
=\frac{S_{1+\kappa}(p)}{1+\kappa}.
\]
For the generalized Pareto and Student’s \(t\) families, the paper shows that the generalized average uncertainty equals the density at the location plus the scale [1510.06951].

A later formulation introduces a dimension- and shape-dependent parameterization. In “Coupled Entropy: A Goldilocks Generalization?” the entropy is
\[
S_\kappa(p)\equiv \frac{1}{1+d\kappa}\,S_q^{NT}(p)
=\frac{1}{1+d\kappa}\,
\frac{1-\sum_i p_i^q}{(q-1)\sum_i p_i^q},
\]
with
\[
q=1+\frac{\alpha\kappa}{1+d\kappa}.
\]
This places \(S_\kappa\) between ordinary Tsallis entropy and normalized Tsallis entropy, while assigning \(\alpha\) to the near-mode behavior and \(\kappa\) to the asymptotic power-law tail decay. The maximizing distributions form a coupled-exponential family,
\[
p_\kappa(x)=\frac{1}{Z_\kappa(\beta)}\,\exp_\kappa\!\bigl(-\beta\,T(x)\bigr),
\qquad
\exp_\kappa(u)\equiv [1+\kappa u]_+^{1/\kappa},
\]
and the paper argues that the extra \(1/(1+d\kappa)\) factor improves robustness relative to normalized Tsallis entropy. Its composition law is
\[
S_\kappa(A\oplus B)=S_\kappa(A)+S_\kappa(B)+\kappa S_\kappa(A)S_\kappa(B),
\]
so \(\kappa\) directly controls nonadditivity [2506.17229].

A more ambitious axiomatic program appears in “The unique, universal entropy for complex systems.” There the trace-form coupled entropy for a \(d\)-dimensional coupled stretched-exponential density \(f_\kappa\) is
\[
H_\kappa[X;\alpha,d]
=
\int_{x\in \mathbb{R}^d}
f_\kappa(x)^q\,
\ln_\kappa\!\bigl[f_\kappa(x)\bigr]^{-1/(1+d_\alpha\kappa)}\,dx,
\]
with \(d_\alpha\equiv d/\alpha\) and
\[
q\equiv 1+\frac{\kappa}{1+d_\alpha\kappa}.
\]
For the maximizing density,
\[
H_\kappa=d_\alpha+\ln_{\kappa/(1+d_\alpha\kappa)} Z_\kappa,
\]
and the non-trace version
\[
\mathcal H_\kappa=[H_\kappa]^{1/\alpha}
\]
changes the extensive exponent from \(\alpha\) to \(1\). The paper singles out this entropy by Shannon–Khinchin I–III, an informational-scale condition requiring the unique \(x\) where \(d\ln f(x)/d\ln x=-1\), universality-class extensivity across Hanel–Thurner scaling classes, and escort expectations for macroscopic observability. Maximization under
\[
L[f]=H_\kappa[f]+\lambda_0\!\left[1-\int f(x)\,dx\right]
+\lambda_1\!\left[\sigma^\alpha-\int x^\alpha f_{\le q}(x)\,dx\right]
\]
yields
\[
f(x)\propto
\left(1+\kappa\,\frac{x^\alpha}{\alpha\sigma^\alpha}\right)^{-(1+d_\alpha\kappa)/\kappa}
=
\exp_\kappa^{-(1+d_\alpha\kappa)}
\!\left(\frac{x^\alpha}{\alpha\sigma^\alpha}\right).
\]
The same work defines composability by the \(\kappa\)-sum,
\[
H_\kappa(A\cup B)=H_\kappa(A)\oplus_\kappa H_\kappa(B)
\equiv H_\kappa(A)+H_\kappa(B)+\kappa H_\kappa(A)H_\kappa(B),
\]
and presents applications to complexity, a zeroth law of temperature, a coupled free energy, variational inference, active inference, and wireless systems [2605.04493].

“On the uniqueness of the coupled entropy” sharpens this program. It gives a formal definition for a coupled stretched-exponential family \(f(x;\sigma,\kappa,\alpha,d)\), states a uniqueness theorem for the coupled entropy under continuity, maximality, expandability, and “Scale-Shape Equivalence at the Informational Scale,” and proves composability and extensivity lemmas. In the special case \(\gamma=1\), the entropy has the closed form
\[
H_\kappa(f_\kappa;\alpha,d,1)
=\frac{d}{\alpha}+\ln_{\frac{\alpha\kappa}{1+d\kappa}} Z_\kappa(\sigma,\alpha,d),
\]
while composability becomes
\[
H_\kappa(A\cup B)
=
H_\kappa(A)\oplus_{\alpha\kappa}H_\kappa(B|A)
=
H_\kappa(A)+H_\kappa(B|A)+\alpha\kappa\,H_\kappa(A)H_\kappa(B|A).
\]
This paper also states thermodynamic relations with \(T_\kappa=\sigma\) and
\[
F_\kappa=-T_\kappa \ln_\kappa Z_\kappa^{1/(1+\kappa)}.
\]
Within these works, the claim of uniqueness is internal to the stated axioms and parameterization; it is not a cross-field consensus statement [2511.17684].

## 4. Entropy production, hidden variables, and strong coupling to baths

In nonequilibrium thermodynamics, coupled entropy refers not to a new entropy functional on probabilities alone, but to entropy and entropy production when interactions with hidden variables or baths cannot be neglected. For a classical system plus bath with total Hamiltonian
\[
H_{\rm tot}(x,y;\lambda)=H_s(x;\lambda)+H_b(y)+V_{\rm int}(x,y),
\]
the Hamiltonian of mean force is
\[
H^*(x;\lambda)
=-\frac{1}{\beta}\ln\!\int dy\,e^{-\beta H_{\rm tot}(x,y;\lambda)}+{\rm const.}
\]
and defines
\[
Z_s^*(\lambda)=\int dx\,e^{-\beta H^*(x;\lambda)},
\qquad
F^*(\lambda)=-\tfrac1\beta\ln Z_s^*(\lambda).
\]
The corresponding conditional equilibrium bath distribution is
\[
\rho_b^{\rm eq}(y|x;\lambda)
=
\exp\!\bigl[-\beta(H_{\rm tot}(x,y;\lambda)-H^*(x;\lambda))\bigr].
\]
For system–bath states of the form
\[
\sigma(x,y;t)=\rho_s(x;t)\,\rho_b^{\rm eq}(y|x;\lambda_t),
\]
the average entropy production is
\[
\langle \Sigma(t)\rangle
=
D_{\rm KL}\!\bigl[\rho(x,y;t)\,\|\,\sigma(x,y;t)\bigr]
\ge 0,
\]
and the stochastic entropy production satisfies the generalized Crooks relation
\[
\frac{P_{\rm F}(+\Sigma)}{P_{\rm R}(-\Sigma)}=e^{+\Sigma}.
\]
This extends the familiar weak-coupling framework to arbitrary strong coupling and correlated nonequilibrium states [1703.03764].

Partial observation of a coupled system changes the fluctuation-theorem structure. In a harmonically coupled Brownian-particle model, total entropy production obeys the steady-state fluctuation theorem exactly, with scaled cumulant-generating function satisfying Gallavotti–Cohen symmetry. For the observed particle alone,
\[
\Delta S_{\rm tot}^A
\equiv -Q^A/T-\ln\!\left[\frac{P_{ss}^A(v_A(\tau))}{P_{ss}^A(v_A(0))}\right],
\]
and for finite coupling \(k>0\), the corresponding SCGF no longer satisfies \(\mu(\lambda)=\mu(1-\lambda)\). Hence the partial entropy production violates the usual fluctuation theorem. In the weak-coupling limit \(\delta=2mk/\gamma^2\to 0\), or in the harmonically confined variant, the symmetry is recovered and \(f(s)=I(s)-I(-s)\to s\) [1710.11339].

In open quantum systems strongly coupled to baths, one route is to define the thermodynamic entropy by the reduced von Neumann entropy,
\[
S(t)=-{\rm Tr}_S[\hat\rho_S(t)\ln \hat\rho_S(t)],
\qquad
\hat\rho_S(t)={\rm Tr}_{\{B\}}[\hat\rho_{\rm tot}(t)].
\]
Using nonequilibrium Green’s functions, the cited work derives a bath- and energy-resolved Clausius relation
\[
\frac{d}{dt}S_B(t;E)=\beta_B\,\dot Q_B(t;E)+\dot P_B(t;E),
\]
with \(\dot P_B(t;E)\) a positive entropy-production term expressed entirely in system Green’s functions and reservoir self-energies. For multipartite systems, the local relation includes an information-flow term,
\[
\frac{d}{dt}S_p(t;E)
=
\sum_{B_p}\Bigl(\beta_{B_p}\dot Q_{B_p}(t;E)+\dot P_{B_p}(t;E)\Bigr)+\dot I_p(t;E).
\]
The formalism is stated to reduce to expected forms in weak-coupling or steady-state limits and to remain valid for interacting systems via Hubbard NEGF [2011.08187].

A complementary strong-coupling quantum formulation avoids drawing a sharp system–bath boundary inside the hybridized region. In a time-dependent Landauer–Büttiker approach, incoming and outgoing electron modes are related by a full scattering matrix, global von Neumann entropy is conserved by unitary scattering, and the entropy of the scattering region is inferred from lead entropy currents,
\[
I^S_{\rm tot}(t)
=
\int\frac{d\epsilon}{2\pi}\,
{\rm Tr}_c\!\left[
\sigma(\phi^{\rm out}(t,\epsilon))-\sigma(\phi^{\rm in}(\epsilon))
\right].
\]
The resulting inside–outside duality yields
\[
\frac{ds}{dt}
=
-I^S_{\rm tot}
=
\frac{\dot Q}{T}
+
\frac{\dot W^{(2)}}{T},
\]
where \(\dot W^{(2)}\ge 0\) is the dissipated work arising at second order in the adiabatic expansion beyond the quasistatic limit [1709.06043].

A different entropy-exchange construction appears in a binary quasi-classical system with Haldane non-linear statistical correlation. There the inter-correlation contributes a non-additive “exclusion entropy”
\[
\chi = \sum_s n_s\ln\!\left[1-\frac{\gamma_s C_Q}{g_s}\right],
\qquad
C_Q=\sum_q \gamma_q\,\frac{n_q^m}{g_q^{m-1}},
\]
and the paper introduces an entropy reservoir or “entropy-bath” that supplies a bias potential
\[
V_q(E_q)
=
-\frac1\beta
\ln\!\left\{
1-m\langle \zeta_S\rangle
\left(\frac{p_q}{\pi_q}\right)^{m-1}
\right\}.
\]
This entropy-bath accelerated molecular dynamics scheme is then applied to low-temperature vitreous silica, where it broadens sampling of the potential-energy landscape under equilibrium conditions [2505.15474].

## 5. Exactly solvable coupled models and strong-coupling limits

Coupled harmonic models provide explicit realizations of entropy generated by interaction. For two particles with Hamiltonian
\[
H=\frac{p_x^2}{2m}+\frac{p_y^2}{2M}
+\frac12 m\omega_0^2x^2+\frac12 M\Omega^2 y^2+\frac12 k(x-y)^2,
\]
the reduced density matrix of one particle is obtained by tracing over the other in the Euclidean path integral. In the zero-temperature limit, the reduced state is Gaussian and the linear entropy is
\[
S_L=1-{\rm Tr}[\rho_A^2]
=
1-\sqrt{\frac{z_-}{z_+}},
\]
where \(z_\pm\) are the normal-mode frequencies. Increasing the confining potential \(\omega_0\) increases the splitting between \(z_+\) and \(z_-\), so \(S_L\to 1\) in the large-\(\omega_0\) limit, while \(S_L\to 0\) in the decoupling limit \(k\to 0\) [1302.1959].

For two coupled harmonic oscillators in canonical equilibrium with a heat bath, the reduced density matrix again remains Gaussian. The purity
\[
P(\rho)= {\rm Tr}\,\rho^2
\]
is first computed as an explicit function of \(\beta\), the coupling parameter \(\eta\), and a mixing angle \(\theta\), and then the Rényi and von Neumann entropies are expressed purely in terms of \(P\). The Rényi entropy is
\[
S_q=\frac{1}{1-q}\ln {\rm Tr}\,\rho_A^q,
\qquad
{\rm Tr}\,\rho_A^q=
\frac{(2P)^q}{(1+P)^q-(1-P)^q},
\]
and the von Neumann entropy is
\[
S_1=
-\ln\!\frac{2P}{1+P}
-\frac{1-P}{2P}\ln\!\frac{1-P}{1+P}.
\]
In the weak-coupling limit, \(P\to 1\) and the entropies vanish; in the strong-coupling limit, \(P\to 0\) and the entropies diverge; at zero temperature they reduce to ground-state entanglement, and at high temperature they diverge because of thermal mixing [1902.07645].

Strong-coupling limits also appear in lattice and conformal-field-theory settings. In a non-relativistic lattice four-fermion model with infinite fermion mass, a saddle-point analysis of the exact effective potential yields
\[
S_T(T,g\to \infty; a)=0
\]
at each finite temperature on a finite-cutoff lattice. When a topological term is added in a two-dimensional lattice topological quantum field theory, a nontrivial Shannon entropy over topological sectors survives,
\[
S_T^{\rm top}(\theta;V)
=
-\sum_{m\in\mathbb Z}
p_m\ln p_m.
\]
By contrast, in a two-dimensional CFT for one interval,
\[
S_E(\ell)=\frac{c}{3}\ln\frac{\ell}{\epsilon}+{\rm const.},
\]
and the coefficient \(c/3\) is unchanged by different choices of boundary regularization or “centers.” The cited work therefore distinguishes finite-cutoff lattice strong coupling, where local excitations freeze and entropy vanishes, from continuum CFT behavior, where universal entanglement coefficients remain [1609.04550].

## 6. Coupled quantum dots and entropy spectroscopy

Coupled quantum dots provide a direct experimental setting in which entropy changes of a composite system can be measured. In a GaAs/AlGaAs double quantum dot, entropy changes are interpreted with the Boltzmann formula
\[
S=k_B\ln \Omega,
\qquad
\Delta S=k_B\ln(\Omega_f/\Omega_i),
\]
and are extracted through the Maxwell relation
\[
\left(\frac{\partial N}{\partial T}\right)_\mu
=
-\frac{1}{k_B}
\left(\frac{\partial S}{\partial \mu}\right)_T.
\]
A nearby quantum-point-contact charge sensor measures an oscillating detector current when one reservoir is heated periodically. Near a single-dot transition,
\[
I_{\rm det}^{(2\omega)}(V)
\simeq
\left(\frac{\partial I_{\rm det}^{(dc)}}{\partial T}\right)\delta T
=
I_0\,
\cosh^{-2}\!\left[\frac{e\alpha(V-V_{\rm mid})}{2k_B T}\right]
\left(\frac{\Delta S}{2k_B}\right),
\]
so fitting the second-harmonic signal yields \(\Delta S\) directly [2508.09481].

In the weak-coupling regime, the paper reports
\[
\Delta S(N=0\to 1)/k_B
=
0.726\pm 0.012 \;\text{(right dot)},
\qquad
0.750\pm 0.042 \;\text{(left dot)},
\]
with integrated values within \(10\%\) of \(\ln 2\simeq 0.693\). At a weak-coupling triple point, the peak entropy is \(\Delta S_{\rm peak}/k_B\simeq \ln 3\simeq 1.10\), whereas in the stronger-coupling “molecular” regime near the same triple point and at \(k_B T\gtrsim t\),
\[
\Delta S_{\rm peak}/k_B \to \ln 4 \simeq 1.39,
\]
reflecting two twofold-degenerate bonding and antibonding levels. For two-electron transitions out of \((1,0)\), the measured values are
\[
(1,0)\to (1,1):\quad \Delta S/k_B=+0.60\pm 0.05 \simeq \ln 2,
\]
\[
(1,0)\to (2,0):\quad \Delta S/k_B=-0.71\pm 0.07 \simeq -\ln 2.
\]
A rate-equation model further identifies triangular features in the entropy signal as nonequilibrium Pauli-blockade artifacts rather than equilibrium entropy [2508.09481].

Remote entropy measurement in capacitively coupled quantum dots extends this logic from the entropy of an added electron to the entropy of the entire two-dot system. Starting from
\[
\left(\frac{\partial S_{\rm sys}}{\partial N}\right)_T
=
-\left(\frac{\partial \mu}{\partial T}\right)_N,
\]
and controlling \(\epsilon_1\) by a gate voltage, the experiment uses
\[
\left(\frac{\partial S_{\rm sys}}{\partial \epsilon_1}\right)_T
=
-\left(\frac{\partial N_1}{\partial T}\right)_{\epsilon_1}
\]
and
\[
\Delta S_{\rm sys}
=
-\int_{\epsilon_{1,i}}^{\epsilon_{1,f}}
\left(\frac{\partial N_1}{\partial T}\right)_{\epsilon_1}
d\epsilon_1.
\]
The measured charge-sensor current difference satisfies
\[
\Delta I_{\rm CS}(V_g)
\approx
\left(\frac{\partial I_{\rm CS}}{\partial N_1}\right)
\left(\frac{\partial N_1}{\partial T}\right)_{V_g}\Delta T,
\]
giving the working formula
\[
\Delta S_{\rm sys}
=
-\alpha_1 e
\int_{V_{g,i}}^{V_{g,f}}
\frac{\Delta I_{\rm CS}(V_g)}
{(\partial I_{\rm CS}/\partial N_1)\Delta T}\,dV_g.
\]
In the weak dot–reservoir coupling limit, a classical four-state model yields microstate-counting steps such as \(k_B\ln 2\), \(k_B\ln 3\), and transient four-state features. In the strong-coupling regime, where \(\Gamma_i\sim k_B T\) or larger, the paper replaces microstate counting by a two-impurity Anderson model solved with numerical renormalization group, and reports that strong hybridization suppresses the classical entropy plateaus [2605.03876].

Taken together, the coupled-quantum-dot literature uses entropy as a spectroscopic observable. The decisive feature is that one dot can sense the entropy response of the entire coupled pair, not merely its own local degeneracy. This use is conceptually different from both pair-excess entropy in liquids and \(\kappa\)-deformed nonadditive entropies, even though the common theme is that coupling changes how entropy is defined, measured, or interpreted [2508.09481] [2605.03876].

Source: https://www.emergentmind.com/topics/coupled-entropy