Preprint
Article

This version is not peer-reviewed.

Entropic Slaving and Collective Gap Renormalization in Finite SU(2) Systems: Structural Analogies to Quantum Absorption Refrigerators

Submitted:

13 November 2026

Posted:

16 November 2026

You are already at the latest version

Abstract

Autonomous quantum thermal machines rely on strict energy-gap resonance conditions that are vulnerable to parameter drift and cross-talk. In this work, we investigate whether information-geometric metric rank collapse in a closed, interacting quantum system provides a structural analogue to these resonance requirements. Analyzing a finite ensemble of \( N = 4 \) fermions governed by a collective \( SU(2) \) quasi-spin Hamiltonian \( \hat{H}(V) = \hat{J}_z - V\hat{X} \), where \( \hat{X} \) is a collective two-body pairing operator, we show that for couplings \( V \gtrsim 0.7 \) and sufficiently low temperatures, the canonical state concentrates onto the two dominant Dicke levels \( \vert{}2, -1\rangle \) and \( \vert{}2, 0\rangle \). Because the nine distinct \( (J, M) \)levels are non-collinear in the \( (M, X) \) plane, the quantum Fisher information metric determinant satisfies \( \det g > 0 \) for all finite inverse temperatures \( \beta \) and couplings \( V \). Consequently, rank collapse is not a sharp phase transition bounded by a finite thermal window, but an asymptotic, tolerance-dependent regime in which the coordinate-invariant normalized determinant \( \Delta _{\text{nat}} \) decays exponentially without bound as \( \beta \) increases. Within this low-temperature limit, the state reduces to a one-parameter Bernoulli family governed by the occupation ratio \( r = e^{-\beta(1-V)} \), rendering \( (\beta, V) \) statistically redundant along iso-state level sets \( V(\beta; r) = 1 + \ln(r)/\beta\ \). Across the dominant doublet, both \( M \) and \( X \) increase by one unit \( (\Delta M = +1, \Delta X = +1) \), shifting the energy gap to \( \Delta E = 1 - V \), which vanishes at the level crossing \( V_c = 1 \). While this vanishing effective gap parallels the level-matching condition \( \delta = 0 \) of a three-qubit quantum absorption refrigerator, we demonstrate that equilibrium metric collapse represents local statistical redundancy rather than a dynamical stabilization mechanism, and realizing steady-state refrigeration requires partitioning the system across multiple thermal reservoirs within a non-equilibrium open-system framework.

Keywords: 
;  ;  ;  

1. Introduction

Controlled quantum technologies are often described as a “second quantum revolution”: hardware that does not merely observe quantum effects but exploits superposition and entanglement to perform tasks beyond classical reach [1]. Within this program, quantum thermodynamics has increasingly turned to autonomous quantum machines [2]. Unlike devices driven by time-dependent external fields, they evolve under a constant Hamiltonian and draw the free energy they need from thermal gradients, in order to keep time, extract work or cool [3,8]. A recurrent practical difficulty is parameter matching. Such machines work well only if their level splittings satisfy resonance conditions imposed by energy conservation, and parameter drift or cross-talk (unwanted coupling or interference between neighbouring qubits) can detune them, degrading cooling power and efficiency [2,3].
A separate line of work uses information geometry to characterize finite many-body systems in thermal equilibrium. For a family of canonical states ρ ^ ( β , V ) = e − β H ^ ( V ) / Z , the quantum Fisher information metric g on the control plane ( β , V ) measures how distinguishable neighbouring states are. When [ H ^ ( V ) , H ^ ( V ′ ) ] = 0 it reduces to the classical covariance matrix of the score fields, and det g vanishes when the two scores are linearly dependent on the thermally populated states, so that the two parameters become locally redundant. Ref. [10] reported that in a finite SU(2) spin-flip fermionic model with N = 4 particles det g becomes extremely small over an extended region of the ( β , V ) plane, in which only two energy levels carry appreciable weight. This was termed quantum entropic slaving, borrowing the word “slaving” from the slaving principle of synergetics [13].
In this paper we examine what this phenomenon does and does not say about the resonance requirements of autonomous thermal machines. We keep the analogy strictly structural: no refrigeration is claimed.

1.1. The Three-Qubit Absorption Refrigerator

The smallest autonomous absorption refrigerator [7] consists of three qubits α ∈ { H , C , A } , each weakly coupled to its own thermal reservoir: a hot bath ( β H ), the cold bath that is to be cooled ( β C ), and an ambient bath ( β A ) that receives the waste heat, with β H < β A < β C . With
H ^ 0 = ∑ α Δ α | 1 〉 〈 1 | α , H ^ 3 B = g | 1 H 1 C 0 A 〉 〈 0 H 0 C 1 A | + h . c . ,
the joint transition | 0 H 0 C 1 A 〉 ↔ | 1 H 1 C 0 A 〉 raises the hot and cold qubits and lowers the ambient one. Heat is thus absorbed from the hot and cold baths and released into the ambient one, which cools the cold bath. Because the transition changes the state of all three qubits at once, the minimal model requires a genuine three-body coupling. The coupling is effective only if the two states it connects are degenerate,
δ ≡ Δ A − Δ H − Δ C = 0 ,
and the steady-state heat current is suppressed, in a Lorentzian fashion, as | δ | grows. Equation (2) is the fine-tuning that drift and cross-talk threaten.

1.2. The SU(2) Model and What Is (and Is Not) Claimed

Our system is a closed ensemble of N fermions described by the collective SU(2) operators J ^ z , J ^ ± . The Hamiltonian is
H ^ = J ^ z − V X ^ , X ^ = J ^ 2 − J ^ z 2 − N 2 = 1 2 J ^ + J ^ − + J ^ − J ^ + − N 2 .
Since J ^ ± and J ^ z are sums of single-site operators, X ^ contains only products of two sites. It is a collective two-body pairing operator, not a three-body interaction. H ^ is diagonal in the Dicke basis | J , M 〉 , with
E ( J , M ) = M − V J ( J + 1 ) − M 2 − N 2 .
For N = 4 there are nine distinct ( J , M ) levels. They carry multiplicities Y J = 50 , 45 , 5 for J = 0 , 1 , 2 , so the Hilbert space has dimension 210.
Because the family ρ ^ ( β , V ) is commuting, the Fisher metric is
g = Var ( H ^ ) − β Cov ( H ^ , X ^ ) − β Cov ( H ^ , X ^ ) β 2 Var ( X ^ ) , det g = β 2 Var ( H ^ ) Var ( X ^ ) − Cov ( H ^ , X ^ ) 2 ≥ 0 ,
which vanishes if and only if H ^ and X ^ are linearly dependent on the populated levels. We also use the normalized determinant Δ = det g / ( g β β g V V ) ∈ [ 0 , 1 ] .
For couplings V of order unity and sufficiently low temperature, the two dominant levels are | a 〉 = | J = 2 , M = − 1 〉 and | b 〉 = | J = 2 , M = 0 〉 , with X a = 3 and X b = 4 . Across this transition Δ M = + 1 and Δ X = + 1 , so
Δ E = E b − E a = Δ M − V Δ X = 1 − V .
The interaction therefore acts as a collective, mean-field-like shift of the one-body level spacing, and the two levels cross at V c = 1 . In this regime the state is a one-parameter Bernoulli family, P b / P a = r = e − β ( 1 − V ) , which depends on ( β , V ) only through β ( 1 − V ) . Pairs ( β , V ) on a curve V = 1 + ln r / β therefore give the same state, and Δ → 0 as the two-level limit is approached.
Three points limit what can be concluded from this.
1.
This is a statement about statistical redundancy of the two control parameters, not a dynamical mechanism. V remains an external parameter, temperature does not renormalize it, and the curves above are level sets of r rather than a renormalization-group flow with a fixed point.
2.
For N = 4 , det g and Δ are quite small ( < 10 − 5 ) but never exactly zero. The collapse is asymptotic in the few-level limit, and the boundaries of any “window” depend on the tolerance used for Δ and on numerical precision. At the crossing V = V c the two levels are degenerate, H ^ is constant on them, and Δ is not small.
3.
The analogy with the refrigerator is limited to level matching. At V = V c the two dominant levels are degenerate, just as Eq. (2) makes the two levels coupled by H ^ 3 B degenerate. But our H ^ has no off-diagonal coupling between | a 〉 and | b 〉 , the system is in equilibrium with a single bath, and at degeneracy the populations are equal and no heat current flows. A refrigerator requires partitioning into three subsystems with separate reservoirs and a non-equilibrium master-equation treatment.
We therefore do not claim that the metric collapse stabilizes a resonance against parameter drift.

1.3. What Novelties Will Be Encountered in This Work?

While prior studies identified the empirical signatures of quantum entropic slaving in finite S U ( 2 ) systems [10], the present work establishes its exact analytical foundations, numerical boundaries, and thermodynamic limitations. Specifically, the reader will encounter four main novelties:
1.
Exact geometric criterion for rank collapse and proof of strict positivity. By expressing the Fisher metric determinant as a manifestly non-negative sum over level triples weighted by squared triangle areas in the ( M , X ) plane, we prove that det g > 0 strictly holds for all finite inverse temperatures 0 < β < ∞ and couplings V in the N = 4 model. Because the nine distinct ( J , M ) levels are non-collinear, rank collapse is shown to be an asymptotic, tolerance-dependent approach to a one-parameter Bernoulli family rather than a sharp phase transition bounded by a finite thermal window.
2.
Coordinate-invariant normalization and resolution of numerical artifacts. We demonstrate that the standard normalized determinant Δ in ( β , V ) coordinates suffers from an artificial breakdown ( Δ ≃ 1 ) at the level crossing V c = 1 solely because g β β ∝ ( Δ E ) 2 vanishes there. Introducing the natural-parameter normalized determinant Δ nat removes this coordinate singularity and reveals an unbounded, monotonic exponential decay at low temperatures ( Δ nat ≃ 4 e − 2 ( 2 V − 1 ) β for 0.7 ≲ V < 1 and Δ nat ≃ 4 e − 2 β for V ≳ 1.2 ), with no collapse occurring at the three-level degeneracies V = 1 / 2 and V = 1 / 3 . Moreover, high-precision arithmetic reveals that apparent upper cutoffs of the collapse region near Δ nat ≲ 10 − 12 are spurious artifacts of catastrophic cancellation in double-precision floating-point evaluations.
3.
Clarification of statistical redundancy versus dynamical stabilization. Decomposing the interaction X ^ = J ^ 2 − J ^ z 2 − N / 2 into single-site fermionic operators confirms that it is strictly a collective two-body pairing operator rather than a three-body interaction. Across the dominant doublet | 2 , − 1 〉 → | 2 , 0 〉 , the interaction renormalizes the energy spacing to Δ E = 1 − V , vanishing at V c = 1 in structural analogy to the resonance condition δ = 0 of an autonomous three-qubit absorption refrigerator. Crucially, however, we show that the iso-state curves V ( β ; r ) = 1 + ln ( r ) / β reflect local statistical parameter redundancy in a single-bath canonical ensemble rather than a renormalization-group flow or dynamical self-stabilization mechanism.
4.
Non-equilibrium tripartite GKSL extension. Because equilibrium degeneracy yields equal doublet populations and zero net energy transport, we formulate an open-system tripartite generalization governed by a Gorini–Kossakowski–Sudarshan–Lindblad (GKSL) master equation with independent hot, cold, and ambient thermal reservoirs ( β H < β A < β C ). This framework defines the exact operational test—evaluating the full quantum Fisher information metric g μ ν on the non-equilibrium steady state ρ ^ NESS —needed to determine whether non-equilibrium rank collapse can actively shield steady-state cooling power Q ˙ C and the Coefficient of Performance against external detuning.
Organization
The paper is organized as follows. Section 2 presents the model and its spectrum. Section 3 derives the two-level (Bernoulli) reduction and the gap (6). Section 4 analyses the Fisher metric and its null directions. Section 5 characterizes numerically how Δ approaches zero. Section 6 formulates the open tripartite problem, with its master equation and heat currents, as the natural next step, and Section 7 summarizes our conclusions.

2. Two-Level Model

2.1. Collective SU(2) Algebra and Hamiltonian

We consider N fermions occupying two single-particle levels, with a collective quasi-spin realization of SU(2). Writing c ^ 1 m † , c ^ 1 m and c ^ 0 m † , c ^ 0 m for the creation and annihilation operators of the upper and lower level, with m labelling the degenerate states within each level, the collective generators are
J ^ z = 1 2 ∑ m n ^ 1 m − n ^ 0 m , J ^ + = ∑ m c ^ 1 m † c ^ 0 m , J ^ − = J ^ + † ,
and satisfy
[ J ^ + , J ^ − ] = 2 J ^ z , [ J ^ z , J ^ ± ] = ± J ^ ± .
The Casimir operator is J ^ 2 = 1 2 ( J ^ + J ^ − + J ^ − J ^ + ) + J ^ z 2 , with eigenvalues J ( J + 1 ) .
The unperturbed Hamiltonian is H ^ 0 = J ^ z , with eigenvalues M (single-particle energies ± 1 2 ). The interaction is generated by the quadratic operator
X ^ = J ^ 2 − J ^ z 2 − N 2 = 1 2 J ^ + J ^ − + J ^ − J ^ + − N 2 ,
and the full Hamiltonian is
H ^ ( V ) = J ^ z − V X ^ , ∂ H ^ ∂ V = − X ^ .
Each term in Eq. (9) is a product of two collective operators of the form J ^ ± , i.e. a quartic expression in the fermion operators. X ^ is therefore a collective two-body pairing operator, and it contains no three-body terms. Both J ^ z and X ^ are functions of J ^ 2 and J ^ z , so [ J ^ z , X ^ ] = [ J ^ 2 , X ^ ] = 0 and H ^ ( V ) is diagonal in the Dicke basis | J , M 〉 for all V. In particular [ H ^ ( V ) , H ^ ( V ′ ) ] = 0 , so the canonical family ρ ^ ( β , V ) is commuting.

2.2. Spectrum and Multiplicities

On | J , M 〉 , with M = − J , … , J and J = N / 2 , N / 2 − 1 , … down to 0 or 1 / 2 ,
X ( J , M ) = J ( J + 1 ) − M 2 − N 2 , E ( J , M ) = M − V X ( J , M ) .
For the maximal multiplet J = N / 2 this gives X = N 2 / 4 − M 2 ≥ 0 . The multiplet of total spin J occurs Y J times, where, with Ω = N / 2 [10,14],
Y J = ( 2 Ω + 2 ) ! ( 2 Ω ) ! ( 2 J + 1 ) ( 2 Ω + 1 ) ( Ω + J + 2 ) ! ( Ω + J + 1 ) ! ( Ω − J + 1 ) ! ( Ω − J ) ! .
The multiplicity Y J does not depend on M. For N = 4 ( Ω = 2 ) one finds Y 2 = 5 , Y 1 = 45 , Y 0 = 50 , so the Hilbert space has dimension ∑ J Y J ( 2 J + 1 ) = 25 + 135 + 50 = 210 . These 210 states form nine distinct ( J , M ) levels, which are listed in Table 1. Throughout, “nine levels” refers to the distinct ( J , M ) labels, and every level carries its multiplicity Y J .

2.3. Canonical Ensemble

The thermal state is ρ ^ ( β , V ) = e − β H ^ ( V ) / Z , and each level ( J , M ) has probability
P ( J , M ) = Y J e − β E ( J , M ) Z ( β , V ) , Z ( β , V ) = ∑ J Y J ∑ M = − J J e − β E ( J , M ) .
Mean energy, entropy and the Fisher metric follow from this distribution (Section 4).

2.4. Structural Properties Used Below

1.
Gap between two levels. For any two levels a and b,
Δ E = E b − E a = Δ M − V Δ X ,
so the interaction shifts a transition energy by − V Δ X .
2.
V-independent states. For J = N / 2 and M = ± N / 2 , X = 0 , so these energies are ± N / 2 for every V.
3.
Reflection M → − M . X ( J , M ) is even in M, so E ( J , − M ) = E ( J , M ) − 2 M for every V. For M > 0 the level ( J , − M ) always lies below ( J , M ) , by 2 M , with the same multiplicity. Hence a level with M > 0 is never populated in preference to its partner ( J , − M ) . In particular | 2 , 1 〉 cannot be one of two dominant levels at low temperature, since | 2 , − 1 〉 is lower by exactly 2 for all V.
4.
Level crossings for N = 4 . From Table 1, the lowest level for V > − 1 is ( 2 , − 2 ) for V < 1 / 3 , ( 2 , − 1 ) for 1 / 3 < V < 1 , and ( 2 , 0 ) for V > 1 . The crossing between ( 2 , − 1 ) and ( 2 , 0 ) at V = 1 has Δ M = + 1 , Δ X = + 1 and Δ E = 1 − V . For V < − 1 the lowest level is ( 0 , 0 ) , with J = 0 . At V = − 1 the three levels ( 2 , − 2 ) , ( 1 , − 1 ) and ( 0 , 0 ) are degenerate at E = − 2 . The transition ( 2 , − 2 ) → ( 1 , − 1 ) has Δ M = + 1 , Δ X = − 1 and Δ E = 1 + V .
Multiplets with J = N / 2 contain no Cooper pairs, while smaller J corresponds to a larger number of pairs, with J = 1 , 0 the fully paired multiplets for N = 4 . In particular, none of the J = 2 levels in Table 1 is a paired level. Which levels dominate the canonical ensemble, and how this leads to a two-level (Bernoulli) description, is the subject of Section 3.

3. Two-Level Reduction at Low Temperature

3.1. Dominant Levels and Truncation Weight

For N = 4 and V > 0 the ground level follows from Table 1: ( 2 , − 2 ) for V < 1 / 3 , ( 2 , − 1 ) for 1 / 3 < V < 1 , and ( 2 , 0 ) for V > 1 . We therefore consider the pair
| a 〉 = | J = 2 , M = − 1 〉 , | b 〉 = | J = 2 , M = 0 〉 ,
whose properties are collected in Table 2. Both levels have the same multiplicity Y 2 = 5 . The level | 2 , 1 〉 is excluded: by the reflection property of Section 2.4 it lies 2 above | 2 , − 1 〉 for every V.
The pair carries the whole ensemble only if the remaining levels are thermally negligible. Let Δ 3 ( V ) be the energy of the lowest level outside the pair, measured from the lower level of the pair. From Table 1,
Δ 3 ( V ) = 3 V − 1 , 1 / 3 < V ≤ 1 , 1 + V , V ≥ 1 ,
where the relevant outside level is | 2 , − 2 〉 for V ≤ 1 and | 2 , 1 〉 for V ≥ 1 . The truncation weight
η ( β , V ) = 1 − P a − P b = ∑ c ∉ { a , b } P c
decays as e − β Δ 3 up to multiplicity prefactors, and the two-level description is accurate when η ≪ 1 , i.e. when β Δ 3 ≫ 1 . Near V = 1 / 3 and V ≈ 1 / 2 the level | 2 , − 2 〉 competes with the pair ( Δ 3 → 0 and 0.5 ), and a three-level description is needed there.

3.2. The Bernoulli Family

When η → 0 , the state is determined by the occupation ratio
r = P b P a = Y b Y a e − β Δ E = e − β ( 1 − V ) , P a = 1 1 + r , P b = r 1 + r .
The pair of control parameters ( β , V ) enters only through ℓ ≡ ln r = − β ( 1 − V ) , so the family is a one-parameter Bernoulli family.

3.3. Fisher Metric of the Bernoulli Family

The score fields are L β = − ( E − 〈 E 〉 ) and L V = β ( X − 〈 X 〉 ) . On two levels, with p = P b , any function f satisfies f − 〈 f 〉 = Δ f u , where u = − p on a and u = 1 − p on b, and 〈 u 2 〉 = p ( 1 − p ) . Hence
L β = − Δ E u , L V = β Δ X u , L V = χ L β , χ = − β Δ X Δ E = − β 1 − V ,
and the metric is
g = p ( 1 − p ) ( 1 − V ) 2 − β ( 1 − V ) − β ( 1 − V ) β 2 = p ( 1 − p ) ∇ ℓ ∇ ℓ T , ∇ ℓ = − ( 1 − V ) , β .
Therefore det g = 0 exactly for the two-level state, with rank one. The null vector is n = ( δ β , δ V ) = ( β , 1 − V ) , which is tangent to the level sets of ℓ since ∇ ℓ · n = 0 . At fixed r these level sets are the curves
V = 1 + ln r β .
Moving along such a curve leaves the state unchanged to first order. This is the statistical redundancy of β and V. For r = 1 the curve is the vertical line V = V c = 1 , where Δ E = 0 , the two levels are degenerate, and the two-level state does not depend on β at all.

3.4. Corrections to the Two-Level Limit

For any commuting family with H ^ = J ^ z − V X ^ , the covariance identity for two random variables gives the exact result
det g = β 2 ∑ i < j < k P i P j P k D i j k 2 , D i j k = Δ M j i Δ X k i − Δ M k i Δ X j i ,
where the sum runs over triples of distinct levels and Δ M j i = M j − M i , etc. D i j k is twice the area of the triangle formed by the three levels in the ( M , X ) plane. It does not depend on V or β , which enter only through the probabilities. Two consequences follow.
1.
With at most two populated levels there are no triples, so det g = 0 exactly. For more levels, det g = 0 if and only if all populated levels are collinear in the ( M , X ) plane.
2.
With the dominant pair ( a , b ) and a small remainder, the leading term is
det g ≃ β 2 P a P b ∑ c ∉ { a , b } P c D a b c 2 , D a b c = ( X c − M c ) − 4 ,
since Δ M b a = Δ X b a = 1 and X a − M a = 4 . The main outside levels have D a b c = − 2 for ( 2 , − 2 ) and ( 2 , 1 ) , − 4 for ( 1 , − 1 ) and − 6 for ( 0 , 0 ) . Hence det g = O ( η ) , and it is never exactly zero for N = 4 because the outside levels are never collinear with the pair. For example, at V = 0.8 and β = 10 Eq. (23) gives 3.0835 × 10 − 5 , equal to the exact det g to the digits shown.
The normalized determinant of Section 5 is, to the same order,
Δ = det g g β β g V V ≃ ∑ c P c D a b c 2 P a P b ( Δ E ) 2 ( Δ X ) 2 ( | Δ E | ≫ η | D | ) .
At V = 0.8 , β = 10 this gives 6.9 × 10 − 4 , against the exact 7.0 × 10 − 4 . Three conclusions follow.
  • Δ → 0 asymptotically, as β Δ 3 → ∞ . It decays like e − β Δ 3 and has no upper edge. For Δ E ≠ 0 , increasing β eventually drives one of the two levels to zero weight. Then g itself vanishes and Δ keeps decreasing. In particular, there is no regime at large β in which the rank collapse is lifted. Any “window” boundaries depend on the tolerance chosen for Δ and on the numerical precision.
  • Near the crossing, | Δ E | ≲ η | D | , the estimate (24) fails. There g β β ∝ ( Δ E ) 2 is itself of the order of the contribution of the outside levels, so Δ is not small ( Δ ≃ 1 at V = V c = 1 ). This is a property of the normalization in ( β , V ) coordinates. det g itself remains O ( η ) at the crossing, and the normalized determinant Δ nat of Section 4, built from J ^ z and X ^ , stays small there.
  • The reduction is a statement about statistical redundancy at low temperature. It is not a dynamical mechanism. V remains an external parameter, and the curves (21) are level sets of r, with no renormalization-group flow or fixed point.

4. Fisher Metric, Rank Collapse and Statistical Redundancy

4.1. The Metric of the Commuting Family

Since [ ρ ^ ( β , V ) , ρ ^ ( β ′ , V ′ ) ] = 0 , the symmetric-logarithmic-derivative operators reduce to the classical scores, and g a b = Cov ( L a , L b ) . With ∂ H ^ / ∂ V = − X ^ ,
L β = − H ^ − 〈 H ^ 〉 , L V = β X ^ − 〈 X ^ 〉 ,
and
g = Var ( H ^ ) − β Cov ( H ^ , X ^ ) − β Cov ( H ^ , X ^ ) β 2 Var ( X ^ ) , det g = β 2 Var ( H ^ ) Var ( X ^ ) − Cov ( H ^ , X ^ ) 2 .
In terms of the natural parameters θ = ( β , λ ) with λ = β V , one has ln ρ ^ = − β J ^ z + λ X ^ − ln Z , an exponential family with sufficient statistics ( J ^ z , X ^ ) . The metric in θ is the Hessian of ln Z , namely
g ( θ ) = Var ( J ^ z ) − Cov ( J ^ z , X ^ ) − Cov ( J ^ z , X ^ ) Var ( X ^ ) .
Using the Jacobian ∂ θ / ∂ ( β , V ) = 1 0 V β one obtains
det g = β 2 det g ( θ ) = β 2 Var ( J ^ z ) Var ( X ^ ) − Cov ( J ^ z , X ^ ) 2 .
Thus the collapse condition involves only J ^ z and X ^ and does not depend explicitly on V. A change of control coordinates multiplies det g by the square of a Jacobian: for example det g T V = det g β V / T 4 . The zero set is therefore the same in ( β , V ) and ( T , V ) .

4.2. Exact Criteria for Rank Collapse

By the identity (22) of Section 3.4,
det g = β 2 ∑ i < j < k P i P j P k D i j k 2 ,
so det g = 0 if and only if all populated levels are collinear in the ( M , X ) plane. Two consequences follow.
1.
With exactly two populated levels, det g = 0 exactly. This is the Bernoulli family of Section 3.
2.
For N = 4 and any finite β > 0 and V, all nine ( J , M ) levels have strictly positive weight, and they are not collinear. For example, ( M , X ) = ( − 2 , 0 ) , ( − 1 , 3 ) and ( 0 , 4 ) are the points of | 2 , − 2 〉 , | 2 , − 1 〉 and | 2 , 0 〉 , which are not collinear. Hence
det g > 0 for all 0 < β < ∞ and all finite V .
Equation (30) implies that det g cannot vanish identically on any region of the ( β , V ) plane for this model. It vanishes only in limits: det g → 0 as β → 0 (like β 2 , see below) and as β → ∞ (exponentially). In between, it can be exponentially small, as in the two-level regime. This is the sense in which the collapse is to be understood.

4.3. Scores, Redundancy and the Slaving Coefficient

Regress L V on L β :
L V = χ L β + R , χ = g β V g β β , Var ( R ) = det g g β β .
The normalized determinant
Δ = det g g β β g V V = 1 − ρ 2 = Var ( R ) Var ( L V ) , ρ = g β V g β β g V V ,
is the fraction of the variance of L V that is not explained by L β . At a rank-one point, R = 0 on the support, L V = χ L β , and the null direction n = ( δ β , δ V ) , defined by n β L β + n V L V = 0 , satisfies
d V d β = − 1 χ = − g β β g β V .
For the two-level state, χ = − β / ( 1 − V ) and the slope is ( 1 − V ) / β , in agreement with the curves (21). Along such a curve a change of V is compensated by a change of β to first order, so that β and V are statistically redundant. This is a local statement about distinguishability of states. It does not mean that V is a function of β in the physical system, which is the content of the next subsection.

4.4. Which Normalization to Use

The magnitude of Δ depends on the choice of coordinates, even though the zero set of det g does not. In the natural coordinates θ = ( β , λ ) the corresponding normalized determinant is
Δ nat = det g ( θ ) Var ( J ^ z ) Var ( X ^ ) = 1 − ρ nat 2 , ρ nat = Cov ( J ^ z , X ^ ) Var ( J ^ z ) Var ( X ^ ) ,
and
Δ Δ nat = Var ( J ^ z ) Var ( H ^ ) .
Hence Δ is small whenever the interaction energy V X ^ dominates the fluctuations of H ^ , even without any redundancy between J ^ z and X ^ . We therefore report both Δ and Δ nat , and base statements about collinearity on Δ nat or on det g .
Two limits for N = 4 illustrate this.
  • High temperature. At β → 0 all levels have weight proportional to Y J . Then Var ( J ^ z ) = 2 / 3 , Var ( X ^ ) = 110 / 63 and Cov ( J ^ z , X ^ ) = 0 , by the symmetry M → − M . Therefore
    det g ≃ 220 189 β 2 , Δ nat → 1 , Δ → 21 21 + 55 V 2 .
    The determinant is small only because the factor β 2 reflects that the state tends to the uniform distribution, which no longer depends on the parameters. The parameters themselves are not redundant. The values of Δ in Eq. (36) (for example 0.28 at V = 1 and 0.09 at V = 2 ) are small only through the factor (35).
  • Low temperature. In the two-level regime of Section 3,
    Δ nat ≃ ∑ c ∉ { a , b } P c D a b c 2 P a P b = O ( η ) ,
    which has no singularity at the crossing V = V c because Δ M b a = Δ X b a = 1 . For example, at V = 1 and β = 10 , Δ nat ≃ 3 × 10 − 8 while Δ ≃ 1 . At the crossing the apparent non-collapse of Δ comes from g β β ∝ ( Δ E ) 2 vanishing, not from a loss of redundancy.
Rank collapse in the ( β , V ) plane is therefore a low-temperature, asymptotic property. It is tied to the near-collinearity of the populated levels in the ( M , X ) plane, and it is controlled by the truncation weight η ∼ e − β Δ 3 .

4.5. What the Collapse Does and Does Not Imply

The null direction defines, in the two-level regime, curves in the ( β , V ) plane along which the state is constant to first order. These curves are level sets of the occupation ratio r. They do not describe a dynamical flow: V remains an external control parameter, temperature does not renormalize it, and no scale-invariant fixed point is involved. The point V c = 1 is simply the level crossing at which Δ E = 0 . The geometric content is therefore the following: in the regime where the populated levels are nearly collinear in the ( M , X ) plane, the two control parameters can no longer be resolved independently from measurements of the state.

5. Numerical Characterization of the Low-Temperature Collapse

5.1. Method and Numerical Precision

The weights P ( J , M ) = Y J e − β E ( J , M ) / Z are known in closed form, so all quantities in Section 4 follow from sums over the nine levels. The only numerical difficulty is cancellation. The quantity det g ( θ ) = Var ( J ^ z ) Var ( X ^ ) − Cov ( J ^ z , X ^ ) 2 is a difference of two nearly equal products when the state is close to rank one, so in double arithmetic its relative error is of order 10 − 16 / Δ nat . Table 3 compares a naive double-precision evaluation of Δ nat with the exact value at V = 0.8 : the two agree down to Δ nat ∼ 10 − 10 ( β = 20 ) and disagree completely, even in sign, once Δ nat ≲ 10 − 12 ( β ≳ 25 ).
All results below use either high-precision arithmetic or the manifestly non-negative sum (29). Plots of det g or Δ nat at large β in double precision should not be trusted, and any apparent edge of a collapse region near Δ nat ∼ 10 − 12 is a numerical artifact.

5.2. Behaviour as a Function of β

Table 4 shows det g , Δ nat , Δ and the truncation weight η at V = 0.8 .
All quantities decrease monotonically beyond β ≈ 4 and none of them becomes exactly zero, in agreement with Eq. (30). In particular there is no value of β beyond which the collapse is lifted. Beyond β ≈ 6 the metric is dominated by a single level and det g falls faster than Δ nat .
The rate of decay follows from the two-level analysis of Section 3. Writing E hi for the higher of E a , E b and E c for the outside levels,
Δ nat ≃ 1 + e − β | Δ E | ∑ c ∉ { a , b } Y c Y 2 D a b c 2 e − β ( E c − E hi ) ,
so Δ nat decays at the rate E c − E hi = Δ 3 − | Δ E | , the distance of the lowest outside level from the upper member of the pair. For N = 4 this gives
Δ nat ≃ 4 e − 2 ( 2 V − 1 ) β , 0.7 ≲ V < 1 ( c = | 2 , − 2 〉 ) , 4 e − 2 β , V ≳ 1.2 ( c = | 2 , 1 〉 ) ,
up to the factor ( 1 + e − β | Δ E | ) , which equals 2 at the crossing V = 1 . At V = 0.8 and β = 10 , Eq. (38) gives 2.8 × 10 − 5 , equal to the exact value. For V ≳ 1.2 the rate is V-independent, because the partner | 2 , 1 〉 of | 2 , − 1 〉 lies exactly 2 above it for every V (Section 2.4).

5.3. Tolerance-Dependent Collapse Region

Since Δ nat never reaches zero, any region in which the collapse holds must be defined by a tolerance. Table 5 gives, for several V and tolerances ϵ , the smallest β * beyond which Δ nat < ϵ (scanning in steps of 0.05 ).
Three features stand out.
  • The lower boundary of the collapse region moves from β * ≈ 3 to β * ≈ 36 as the tolerance is tightened from 10 − 2 to 10 − 12 . There is no tolerance-independent window. For V ≳ 1.2 , Eq. (39) gives β * ≃ 1 2 ln ( 4 / ϵ ) , in agreement with the table.
  • There is no upper boundary: Δ nat keeps decreasing for all larger β (Table 4).
  • At V = 1 / 2 no collapse occurs at any β . There | 2 , 0 〉 and | 2 , − 2 〉 are degenerate, so the low-temperature ensemble consists of three levels that are not collinear in the ( M , X ) plane, and Δ nat tends to a non-zero constant. The same holds near V = 1 / 3 , where | 2 , − 2 〉 and | 2 , − 1 〉 cross. The two-level reduction requires Δ 3 ( V ) of order one or larger, i.e. V above roughly 0.7 .

5.4. Entropy of the Two-Level State

The von Neumann entropy of ρ ^ is S = − ∑ J , M P ( J , M ) ln P ( J , M ) Y J = H ( P ) + ∑ J , M P ( J , M ) ln Y J , where H ( P ) is the Shannon entropy of the level probabilities. In the two-level regime, with Y a = Y b = 5 ,
S ≃ h ( r ) + ln 5 , h ( r ) = ln ( 1 + r ) − r ln r 1 + r , r = e − β ( 1 − V ) .
Numerically, S = 1.97478 (exact) against 1.97477 (Eq. (40)) at ( β , V ) = ( 10 , 0.8 ) , and S = ln 10 = 2.30259 at ( 10 , 1.0 ) . The binary entropy h ( r ) is maximal, h = ln 2 , at r = 1 , that is, at the crossing V = V c = 1 where the two levels are equally populated. This is a property of the state at the degeneracy and not a mechanism: V is not an entropy-maximizing variable of the system, and the maximum of h ( r ) does not select V = V c .

5.5. Summary of the Numerical Findings

1.
det g > 0 everywhere for 0 < β < ∞ . It is exponentially small in the low-temperature two-level regime and falls like β 2 at high temperature.
2.
In the two-level regime Δ nat ≃ 4 e − 2 ( 2 V − 1 ) β for 0.7 ≲ V < 1 and ≃ 4 e − 2 β for V ≳ 1.2 . Both decay monotonically with no upper edge.
3.
Any “collapse region” depends on the tolerance chosen for Δ nat , and apparent sharp edges at Δ nat ≲ 10 − 12 in double precision are numerical artifacts.
4.
The collapse is absent at V = 1 / 2 and near V = 1 / 3 , where three non-collinear levels compete.

6. Open Tripartite Extension: Master Equation and Heat Currents

As established in Section 3, rank collapse in a closed, single-bath canonical ensemble reflects local statistical redundancy rather than a dynamical stabilization mechanism. At equilibrium degeneracy ( V = V c = 1 ), the dominant levels equalize in population ( r = 1 ) and no net energy flows through the system. To bridge this structural analogy to an operational quantum absorption refrigerator, the closed collective architecture must be partitioned into an open, non-equilibrium tripartite system coupled to separate thermal reservoirs.

6.1. Tripartite Partition and Total Hamiltonian

Following the minimal autonomous refrigerator paradigm of Section 1, we partition the finite fermionic ensemble (or an effective three-mode subspace thereof) into three distinct subsystems α ∈ { H , C , A } , corresponding to the hot, cold, and ambient degrees of freedom. The total system Hamiltonian takes the form
H ^ sys ( V ) = H ^ 0 + H ^ int ( V ) , H ^ 0 = ∑ α ∈ { H , C , A } H ^ α ,
where H ^ α governs the bare local level structure of subsystem α with characteristic energy splitting Δ α , and H ^ int ( V ) mediates the inter-subsystem coupling parameterized by V. Whereas the closed collective model relies on a two-body pairing operator X ^ diagonal in the Dicke basis, autonomous cooling requires H ^ int ( V ) to induce coherent, energy-exchanging transitions between the cold-plus-hot excitations and the ambient mode. In the resonant manifold, this interaction drives transitions of the form 0 H 0 C 1 A ↔ 1 H 1 C 0 A , where the effective detuning δ ( V ) plays the role of the renormalized gap Δ E ( V ) in Eq. (6).

6.2. GKSL Master Equation and Non-Equilibrium Steady State

To drive a continuous thermodynamic cycle, each subsystem α is weakly coupled to its own independent bosonic or fermionic thermal reservoir at inverse temperature β α , ordered such that β H < β A < β C . Under the standard Born–Markov and secular approximations, the reduced density matrix ρ ^ ( t ) of the tripartite system evolves according to the Gorini–Kossakowski–Sudarshan–Lindblad (GKSL) master equation:
d ρ ^ d t = − i H ^ sys ( V ) , ρ ^ + ∑ α ∈ { H , C , A } D α [ ρ ^ ] ,
where we set ℏ = 1 . Each local dissipator D α acts exclusively on subsystem α and is given by
D α [ ρ ^ ] = γ α + L ^ α † ρ ^ L ^ α − 1 2 L ^ α L ^ α † , ρ ^ + γ α − L ^ α ρ ^ L ^ α † − 1 2 L ^ α † L ^ α , ρ ^ ,
where L ^ α ( L ^ α † ) denotes the lowering (raising) jump operator associated with the local transition of energy Δ α . Thermodynamic consistency with the respective reservoirs is enforced by the local detailed-balance condition on the transition rates:
γ α + γ α − = e − β α Δ α , α ∈ { H , C , A } .
At long times ( t → ∞ ), the system relaxes to a unique non-equilibrium steady state (NESS) defined by L [ ρ ^ NESS ] = 0 , where L denotes the full Liouvillian superoperator on the right-hand side of Eq. (42). Unlike the equilibrium canonical state ρ ^ ( β , V ) , ρ ^ NESS ( β H , β A , β C , V ) sustains stationary off-diagonal coherences in the bare eigenbasis of H ^ 0 whenever H ^ int ( V ) does not commute with H ^ 0 .

6.3. Steady-State Heat Currents and Non-Equilibrium Metric Collapse

In the steady state, the average energy of the tripartite system is constant, Tr H ^ sys d ρ ^ NESS / d t = 0 . Defining the stationary heat current flowing from reservoir α into the system as
Q ˙ α = Tr H ^ sys D α [ ρ ^ NESS ] ,
the First Law of Thermodynamics imposes energy conservation across the three baths,
Q ˙ H + Q ˙ C + Q ˙ A = 0 ,
while the Second Law requires non-negative entropy production, Σ = − ∑ α β α Q ˙ α ≥ 0 . The device operates as an autonomous absorption refrigerator whenever heat is continuously extracted from the cold reservoir ( Q ˙ C > 0 ) and driven by the hot reservoir ( Q ˙ H > 0 ), with waste heat dumped into the ambient bath ( Q ˙ A < 0 ). Its thermodynamic efficiency is quantified by the Coefficient of Performance (COP),
COP = Q ˙ C Q ˙ H .
This open-system formulation provides the exact setting needed to test whether information-geometric rank collapse can actively protect autonomous cooling against detuning. Specifically, one evaluates the full quantum Fisher information metric g μ ν on the non-equilibrium manifold parameterized by θ = ( β H , β A , β C , V ) via the symmetric logarithmic derivatives L ^ μ defined implicitly by
∂ μ ρ ^ NESS = 1 2 L ^ μ , ρ ^ NESS , g μ ν = 1 2 Tr ρ ^ NESS L ^ μ , L ^ ν .
Because [ ρ ^ NESS ( θ ) , ρ ^ NESS ( θ ′ ) ] ≠ 0 in the presence of stationary coherences, g μ ν genuinely incorporates quantum coherence contributions beyond the classical Fisher covariance matrix. Identifying parameter regimes where det g μ ν → 0 on ρ ^ NESS —and determining whether the resulting null directions coincide with level sets of constant cooling power Q ˙ C and constant COP —constitutes the direct operational test of entropic slaving in autonomous quantum thermal machines.

7. Conclusions

In this work, we have rigorously examined the information-geometric properties of a strictly finite, interacting fermionic system governed by a collective S U ( 2 ) quasi-spin algebra with N = 4 particles ( Ω = 2 , spanning 210 microstates across nine distinct ( J , M ) energy levels), and evaluated the precise structural boundaries of its analogy to autonomous three-qubit quantum absorption refrigerators.
Our analysis establishes four central physical and geometric results:
1.
Asymptotic Two-Level Reduction and Statistical Redundancy: For bare couplings V ≳ 0.7 and sufficiently low temperatures ( β Δ 3 ≫ 1 ), the canonical ensemble concentrates onto the dominant doublet | a 〉 = | 2 , − 1 〉 and | b 〉 = | 2 , 0 〉 within the maximal collectivity multiplet J = 2 (each carrying multiplicity Y 2 = 5 ), while | 2 , 1 〉 remains thermally excluded due to the exact spectral reflection symmetry E ( J , − M ) = E ( J , M ) − 2 M . In this low-temperature limit, the state reduces to a one-parameter Bernoulli family governed solely by the occupation ratio r = P b / P a = e − β ( 1 − V ) . Consequently, the quantum Fisher information metric g approaches rank one ( det g → 0 ), and the control parameters ( β , V ) become statistically redundant along the constant-state level sets V ( β ; r ) = 1 + ln ( r ) / β .
2.
Absence of Sharp Boundaries and Numerical Precision Requirements: Because the nine ( J , M ) levels are non-collinear in the ( M , X ) plane, the exact metric determinant satisfies det g > 0 for every finite inverse temperature 0 < β < ∞ and finite V. Rank collapse is therefore an asymptotic, tolerance-dependent regime rather than a sharp phase transition bounded by a fixed upper temperature cutoff. Specifically, the coordinate-invariant normalized determinant Δ nat decays exponentially without bound as Δ nat ≃ 4 e − 2 ( 2 V − 1 ) β for 0.7 ≲ V < 1 and Δ nat ≃ 4 e − 2 β for V ≳ 1.2 , whereas at V = 1 / 2 and V = 1 / 3 three-level degeneracies prevent collapse altogether. Furthermore, high-precision arithmetic (or the manifestly positive triple-sum representation of det g ) is essential at low temperatures, as catastrophic cancellation in standard double precision produces spurious sign changes and artificial window edges once Δ nat ≲ 10 − 12 .
3.
Algebraic Order of the Interaction and Meaning of the Crossing Point: Decomposing the collective operator X ^ = J ^ 2 − J ^ z 2 − N / 2 into single-site fermionic modes confirms that it is strictly a collective two-body pairing operator rather than a three-body interaction. Across the dominant doublet | 2 , − 1 〉 → | 2 , 0 〉 , both M and X increase by one unit ( Δ M = + 1 , Δ X = + 1 ), shifting the bare single-particle gap from Δ M = 1 to Δ E = 1 − V . At the level crossing V c = 1 , the renormalized gap vanishes ( Δ E = 0 ), the populations equalize ( r = 1 ), and the von Neumann entropy reaches its local doublet ceiling S = ln 10 . However, we emphasize that V remains an external mechanical control parameter rather than a dynamically renormalized coupling, and the curves V = 1 + ln ( r ) / β represent iso-state manifolds rather than a renormalization-group flow toward an attractor.
4.
Scope of the Structural Analogy to Absorption Refrigerators: The vanishing of the effective two-level spacing ( Δ E = 1 − V = 0 at V c = 1 ) provides a clean algebraic parallel to the energy-conservation resonance condition δ ≡ Δ A − Δ H − Δ C = 0 required for coherent three-body exchange in an autonomous three-qubit quantum absorption refrigerator. Nevertheless, metric rank collapse in a closed, single-bath canonical ensemble does not by itself stabilize a thermal machine against parameter drift or generate refrigeration: at equilibrium degeneracy, populations are equal and no steady-state heat current flows.
Translating these geometric insights into an operational, self-regulating quantum thermal machine requires moving beyond the closed single-bath canonical framework. Future work will focus on partitioning the collective architecture into three distinct subsystems coupled weakly to independent hot, cold, and ambient thermal reservoirs ( β H < β A < β C ) under a Gorini–Kossakowski–Sudarshan–Lindblad (GKSL) master-equation description. Computing the quantum Fisher information metric directly on the resulting non-equilibrium steady state ρ ^ NESS ( β H , β A , β C , V ) will determine whether non-equilibrium rank collapse can emerge in open multi-bath systems and actively shield steady-state cooling currents and the Coefficient of Performance against external detuning.

References

  1. Acín, S.; Bloch, I.; Bruss, H.; Calarco, T.; Eisert, J.; Fan, J.; Gerdts, M. V.; Glätzle, A.; Halfmann, T. V.; Harris, P. J.; et al. The quantum technologies roadmap: a European community. New J. Phys. 2018, vol. 20(no. 8), 080201. [Google Scholar] [CrossRef]
  2. Marín Guzmán, J. A.; Erker, P.; Gasparinetti, S.; Huber, M.; Halpern, N. Yunger. Key issues review: useful autonomous quantum machines. Rep. Prog. Phys. 2024, 87, 122001. [Google Scholar] [CrossRef] [PubMed]
  3. Mitchison, M. T. Quantum thermal absorption machines: refrigerators, engines and clocks. Contemp. Phys. 2019, 60, 164. [Google Scholar] [CrossRef]
  4. Erker, P.; Mitchison, M. T.; Silva, R.; Woods, M. P.; Brunner, N.; Huber, M. Autonomous quantum clocks: does thermodynamics limit our ability to measure time? Phys. Rev. X 2017, 7, 031022. [Google Scholar] [CrossRef]
  5. Schwarzhans, E.; Lock, M. P.; Erker, P.; Friis, N.; Huber, M. Autonomous temporal probability concentration: clockworks and the second law of thermodynamics. Phys. Rev. X 2021, 11, 011006. [Google Scholar] [CrossRef]
  6. Kosloff, R.; Levy, A. Quantum heat engines and refrigerators: continuous devices. Annu. Rev. Phys. Chem. 2014, 65, 365. [Google Scholar] [CrossRef] [PubMed]
  7. Linden, N.; Popescu, S.; Skrzypczyk, P. How small can thermal machines be? The smallest possible refrigerator. Phys. Rev. Lett. 2010, 105, 130401. [Google Scholar] [CrossRef] [PubMed]
  8. Aamir, M. A.; Suria, P. J.; Marín Guzmán, J. A.; Castillo-Moreno, C.; Epstein, J. M.; Halpern, N. Yunger; Gasparinetti, S. Thermally driven quantum refrigerator autonomously resets superconducting qubit. arXiv 2023, arXiv:2305.16710. [Google Scholar]
  9. Correa, L. A.; Palao, J. P.; Alonso, D.; Adesso, G. Quantum-enhanced absorption refrigerators. Sci. Rep. 2014, 4, 3949. [Google Scholar] [CrossRef] [PubMed]
  10. Plastino, A.; Vampa, V. Temperature-Induced Order in Quantum Many-Body Systems: Entropic Slaving. Preprints.org 2026. [Google Scholar] [CrossRef]
  11. K. Huang, Statistical Mechanics; John Wiley & Sons, New York, 1987.
  12. Goldenfeld, N. Lectures on Phase Transitions and the Renormalization Group; CRC Press: Boca Raton, 2018. [Google Scholar]
  13. Haken, H. Synergetics: An Introduction: Nonequilibrium Phase Transitions and Self-Organization in Physics, Chemistry, and Biology, 3rd ed.; Springer-Verlag, Berlin, 1983.
  14. Rossignoli, R.; Plastino, A. Thermal effects and the interplay between pairing and shape deformations. Phys. Rev. C 1985, 32, 1040. [Google Scholar] [CrossRef] [PubMed]
Table 1. The nine distinct ( J , M ) levels of the N = 4 model, with multiplicity Y J , interaction eigenvalue X and energy E = M − V X .
Table 1. The nine distinct ( J , M ) levels of the N = 4 model, with multiplicity Y J , interaction eigenvalue X and energy E = M − V X .
J M Y J X ( J , M ) E ( J , M )
2 − 2 5 0 − 2
2 − 1 5 3 − 1 − 3 V
2 0 5 4 − 4 V
2 1 5 3 1 − 3 V
2 2 5 0 2
1 − 1 45 − 1 − 1 + V
1 0 45 0 0
1 1 45 − 1 1 + V
0 0 50 − 2 2 V
Table 2. Properties of the dominant pair for N = 4 and V > 1 / 3 .
Table 2. Properties of the dominant pair for N = 4 and V > 1 / 3 .
| a 〉 = | 2 , − 1 〉 | b 〉 = | 2 , 0 〉
Multiplicity Y J 5 5
M − 1 0
X ( J , M ) 3 4
E ( J , M ) − 1 − 3 V − 4 V
Δ M = + 1 , Δ X = + 1 , Δ E = E b − E a = 1 − V
Table 3. Δ nat at V = 0.8 : naive double-precision evaluation of the covariances versus exact (120-digit) values.
Table 3. Δ nat at V = 0.8 : naive double-precision evaluation of the covariances versus exact (120-digit) values.
β double precision exact
10 2.797 × 10 − 5 2.797 × 10 − 5
20 1.537 × 10 − 10 1.538 × 10 − 10
30 − 4.95 × 10 − 13 9.30 × 10 − 16
40 − 8.6 × 10 − 12 5.7 × 10 − 21
60 2.2 × 10 − 10 2.2 × 10 − 31
Table 4. Exact N = 4 results at V = 0.8 .
Table 4. Exact N = 4 results at V = 0.8 .
β det g Δ nat Δ η
2 5.0 × 10 − 1 4.7 × 10 − 1 6.0 × 10 − 1 5.9 × 10 − 2
4 4.0 × 10 − 2 4.7 × 10 − 2 7.2 × 10 − 1 2.8 × 10 − 3
6 4.6 × 10 − 3 4.0 × 10 − 3 9.4 × 10 − 2 1.8 × 10 − 4
10 3.1 × 10 − 5 2.8 × 10 − 5 7.0 × 10 − 4 7.3 × 10 − 7
20 1.9 × 10 − 11 1.5 × 10 − 10 3.8 × 10 − 9 6.8 × 10 − 13
40 1.0 × 10 − 24 5.7 × 10 − 21 1.4 × 10 − 19 4.8 × 10 − 25
80 6.6 × 10 − 52 8.1 × 10 − 42 2.0 × 10 − 40 2.3 × 10 − 49
Table 5. β * ( V , ϵ ) for N = 4 : smallest β with Δ nat < ϵ .
Table 5. β * ( V , ϵ ) for N = 4 : smallest β with Δ nat < ϵ .
V ϵ = 10 − 2 10 − 4 10 − 8 10 − 12
0.5 — — — —
0.7 7.6 13.3 24.8 36.3
0.8 5.3 9.0 16.6 24.2
1.0 3.7 6.0 10.6 15.3
1.2 3.3 5.5 10.0 14.6
1.5 3.1 5.4 10.0 14.6
2.0 3.1 5.4 10.0 14.6
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.