Introduction
The ergotropy quantifies the maximum work extractable from a quantum state through cyclic unitary transformations and provides a central quantity in quantum thermodynamics [1 ] . For composite systems, the extractable work depends on both the correlations among subsystems and the class of operations allowed for its extraction. These aspects have been investigated in connection with work stored in correlations, collective work extraction through entangling operations, and the difference between local and global ergotropy [2 , 3 , 4 , 5 , 6 , 7 ] . Related studies have further examined the role of correlations in quantum batteries, including their effects on extractable work and ergotropy transport [8 , 9 , 10 ] . The contribution of quantum coherence to work extraction has also been studied in both closed and open quantum systems [11 , 12 ] . For bosonic systems described by quadratic Hamiltonians, restricting the allowed transformations to Gaussian unitaries leads to Gaussian ergotropy, for which analytical expressions have been derived in terms of the covariance matrix and its symplectic eigenvalues [13 , 14 ] . More recently, ergotropic properties have also been employed to investigate entanglement in continuous-variable Gaussian systems [15 ] . These developments provide a thermodynamic framework for studying the role of correlations in the work extraction of continuous-variable quantum systems.
The Unruh effect establishes a fundamental connection between quantum field fluctuations and thermal physics, whereby a uniformly accelerated detector in the Minkowski vacuum responds as if immersed in a thermal bath at temperature T U = a / ( 2 π ) T_U = a/(2\pi) T U = a / ( 2 π ) [16 , 17 , 18 ] . Together with Hawking radiation and the Gibbons–Hawking effect [19 , 20 ] , it has motivated extensive studies of particle-detector response and thermalization under acceleration and in curved spacetimes, including finite-time interactions and spatially extended detectors [21 , 22 , 23 , 24 , 25 , 26 ] . In parallel, relativistic quantum-information research has explored observer-dependent entanglement and quantum communication [27 , 28 , 29 ] , continuous-variable and multipartite entanglement, and entanglement harvesting [30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 ] , as well as quantum coherence [43 , 44 , 45 , 46 , 47 , 48 ] , Bell nonlocality [49 , 50 , 51 , 52 , 53 , 54 ] , entropic uncertainty relations [55 , 56 , 57 ] , quantum steering [58 , 59 , 60 , 61 , 62 ] , and quantum discord [37 , 63 , 64 , 65 , 66 , 67 , 68 ] . More recently, quantum work extraction has emerged as a complementary perspective in relativistic quantum thermodynamics, with investigations of accelerated Unruh–DeWitt quantum batteries [69 ] , coherence-assisted ergotropy as a probe of Unruh thermality [70 ] , relativistic quantum Otto cycles [71 ] , and ergotropy in de Sitter and anti-de Sitter spacetimes [72 ] . Collectively, these investigations encompass discrete- and continuous-variable systems, diverse quantum correlations, and a variety of accelerated and curved-spacetime backgrounds, broadening our understanding of quantum information and thermodynamic processes in relativistic settings.
For spatially separated detectors, the field correlations sampled along distinct worldlines introduce an additional dependence on their relative geometry. While the local response is determined by the field correlation function evaluated along an individual trajectory, the collective dynamics also involves the cross Wightman function connecting the two detector trajectories. Field-mediated entanglement generation and harvesting, acceleration-dependent inter-detector interactions, and spatially dependent response spectra have been investigated extensively in this context [31 , 32 , 33 , 65 , 73 , 74 , 75 , 76 ] . Related investigations have also examined the influence of spacetime curvature and horizon structure on correlations extracted by multiple detectors [34 , 35 , 36 , 37 ] . These studies highlight the importance of inter-detector field correlations in determining the collective behavior of spatially separated quantum systems.
Harmonic-oscillator detectors offer a natural framework for investigating how field correlations affect the collective dynamics of spatially separated quantum systems. Their linear coupling to a free scalar field leads to a quadratic system–environment Hamiltonian, allowing the detector dynamics to be treated within the covariance-matrix formalism beyond perturbation theory [77 ] . This approach is closely related to quantum Brownian motion, where the environmental degrees of freedom can be integrated out to obtain the effective dissipation and fluctuation kernels governing the system dynamics [78 , 79 , 80 , 81 , 82 ] . Extensions to multiple coupled oscillators interacting with a common environment have further established the role of bath-mediated correlations in their dynamics and decoherence [83 , 84 , 85 , 86 ] . The resulting covariance-matrix description naturally incorporates both local fluctuations and inter-detector correlations, providing a suitable basis for studying their thermodynamic consequences.
In this work, we investigate whether Gaussian ergotropy can serve as a quantitative probe of spatial field correlations by distinguishing environments with identical local spectra but different spatial cross spectra. We consider two transversely separated harmonic-oscillator detectors with equal proper acceleration, linearly coupled to a common massless real scalar field and directly coupled through a two-mode-squeezing interaction. We compare the Minkowski vacuum sampled by uniformly accelerated detectors with an inertial thermal field at the same Unruh temperature, and also consider independent local thermal baths as an uncorrelated reference. Starting from the corresponding field correlation functions, we construct the oscillator-bath models and derive the equilibrium covariance matrices using the Green-function formalism. In the weak-coupling regime, we establish an analytical relation between the equilibrium Gaussian ergotropy and a response-weighted spatial correlation of the field. We identify a finite-distance blind point at which the single-setting work responses of the two environments coincide despite their different spatial correlations. To resolve this ambiguity, we propose a two-setting protocol with opposite direct squeezing couplings, λ = ± g \lambda= \pm g λ = ± g , and derive an explicit inversion relation that reconstructs the response-weighted spatial correlation from the corresponding ergotropies, eliminating the common local work scale. We further show that the additional work accessible through joint Gaussian operations is proportional to the square of this correlation when the direct coupling is switched off, and can remain nonzero even for separable detector states. We also investigate finite-coupling effects using the full, unexpanded Green functions, quantifying their corrections to the analytical predictions and the reconstructed correlations. Finally, we study the finite-time evolution of Gaussian ergotropy and characterize its oscillatory decay under the accelerated-vacuum environment.
The paper is organized as follows. Sections II–IV introduce the oscillator-bath model and develop the covariance-matrix dynamics and equilibrium Gaussian ergotropy. Sections V–VI establish the relation between spatial correlations and work extraction and present the two-setting reconstruction. Sections VII and VIII discuss the additional global work and finite-time dynamics, respectively. Finally, Sec. IX presents our conclusions.
Oscillator-bath representation of the accelerated vacuum and the matched thermal field
We consider a massless real scalar field in Minkowski spacetime, where the metric line element is given by
d s 2 = − d t 2 + d x 2 + d y 2 + d z 2 . (1) \mathrm{d}s^{2}=-\mathrm{d}t^{2}+\mathrm{d}x^{2}+\mathrm{d}y^{2}+\mathrm{d}z^{2}. \tag*{(1)} d s 2 = − d t 2 + d x 2 + d y 2 + d z 2 . (1)
Suppose two detectors, Alice and Bob, moving with uniform proper acceleration along the x x x direction. They are separated by a distance d d d in the transverse y y y direction, with the z z z coordinate fixed at zero.
The worldlines of the two detectors are given by
x A = ( sinh ( a τ ) a , cosh ( a τ ) a , 0 , 0 ) , (2) x_{A}=\left(\frac{\sinh(a\tau)}{a},\frac{\cosh(a\tau)}{a},0,0\right), \tag*{(2)} x A = ( a sinh ( a τ ) , a cosh ( a τ ) , 0 , 0 ) , (2)
x B = ( sinh ( a τ ) a , cosh ( a τ ) a , d , 0 ) , (3) x_{B}=\left(\frac{\sinh(a\tau)}{a},\frac{\cosh(a\tau)}{a},d,0\right), \tag*{(3)} x B = ( a sinh ( a τ ) , a cosh ( a τ ) , d , 0 ) , (3)
where both detectors possess the same proper acceleration, a A = a B = a a_{A}=a_{B}=a a A = a B = a , and share a common Rindler time parameter τ \tau τ .
The massless real scalar field obeys the massless Klein–Gordon equation:
□ ϕ ^ ( x ) = 0 , (4) \Box\hat{\phi}(x)=0, \tag*{(4)} □ ϕ ^ ( x ) = 0 , (4)
where □ = − ∂ t 2 + ∂ x 2 + ∂ y 2 + ∂ z 2 \Box=-\partial_{t}^{2}+\partial_{x}^{2}+\partial_{y}^{2}+\partial_{z}^{2} □ = − ∂ t 2 + ∂ x 2 + ∂ y 2 + ∂ z 2 is the d’Alembertian operator.
The field operator ϕ ^ ( x ) \hat{\phi}(x) ϕ ^ ( x ) is expanded in plane waves [87 ] :
ϕ ^ ( x ) = ∫ d 3 k ( 2 π ) 3 1 2 ∣ k ∣ [ a ^ k e − i ∣ k ∣ t + i k ⋅ x + a ^ k † e i ∣ k ∣ t − i k ⋅ x ] . (5) \hat{\phi}(x)=\int\frac{\mathrm{d}^{3}\mathbf{k}}{(2\pi)^{3}}\frac{1}{\sqrt{2\lvert\mathbf{k}\rvert}}\left[\hat{a}_{\mathbf{k}}e^{-i\lvert\mathbf{k}\rvert t+i\mathbf{k}\cdot\mathbf{x}}+\hat{a}_{\mathbf{k}}^{\dagger}e^{i\lvert\mathbf{k}\rvert t-i\mathbf{k}\cdot\mathbf{x}}\right]. \tag*{(5)} ϕ ^ ( x ) = ∫ ( 2 π ) 3 d 3 k 2 ∣ k ∣ 1 [ a ^ k e − i ∣ k ∣ t + i k ⋅ x + a ^ k † e i ∣ k ∣ t − i k ⋅ x ] . (5)
where k \mathbf{k} k is the wave vector, ∣ k ∣ \lvert\mathbf{k}\rvert ∣ k ∣ is the wave number, and a ^ k \hat{a}_{\mathbf{k}} a ^ k and a ^ k † \hat{a}_{\mathbf{k}}^{\dagger} a ^ k † are the annihilation and creation operators for mode k \mathbf{k} k , respectively.
The Minkowski vacuum satisfies
a ^ k ∣ 0 ⟩ M = 0 , ∀ k , (6) \hat{a}_{\mathbf{k}}\lvert0\rangle_{M}=0,\qquad\forall\mathbf{k}, \tag*{(6)} a ^ k ∣ 0 ⟩ M = 0 , ∀ k , (6)
where $\left[\hat{a}_{\mathbf{k}},\hat{a}_{\mathbf{k}'}^{\dagger}\right]=(2\pi)^{3}\delta^{3}(\mathbf{k}-\mathbf{k}'),\left[\hat{a}_{\mathbf{k}},\hat{a}_{\mathbf{k}'}\right]=0.
The two-point correlation function of the quantum field, i.e. the Wightman function, is defined as [17 , 88 , 89 , 90 ]
W i j ( x , x ′ ) = ⟨ 0 M ∣ ϕ ^ ( x ) ϕ ^ ( x ′ ) ∣ 0 M ⟩ , i , j ∈ { A , B } , (7) W_{ij}(x,x')=\langle0_M|\hat{\phi}(x)\hat{\phi}(x')|0_M\rangle,\quad i,j\in\{A,B\}, \tag*{(7)} W ij ( x , x ′ ) = ⟨ 0 M ∣ ϕ ^ ( x ) ϕ ^ ( x ′ ) ∣ 0 M ⟩ , i , j ∈ { A , B } , (7)
where the diagonal elements describe the autocorrelations of individual detectors, while the off-diagonal elements describe the cross-correlations between the two detectors.
Equation (7) is given explicitly by
W i j ( x , x ′ ) = ∫ d 3 k ( 2 π ) 3 2 ∣ k ∣ e − i ∣ k ∣ Δ t + i k ⋅ Δ x . (8) W_{ij}(x,x')=\int\frac{\mathrm{d}^3k}{(2\pi)^3 2|\mathbf{k}|}e^{-i|\mathbf{k}|\Delta t+i\mathbf{k}\cdot\Delta\mathbf{x}}. \tag*{(8)} W ij ( x , x ′ ) = ∫ ( 2 π ) 3 2∣ k ∣ d 3 k e − i ∣ k ∣Δ t + i k ⋅ Δ x . (8)
We first evaluate the solid-angle integral by exploiting the spatial spherical symmetry. Let r = ∣ Δ x ∣ r=|\Delta\mathbf{x}| r = ∣Δ x ∣ denote the spatial separation between the two spacetime points, k = ∣ k ∣ k=|\mathbf{k}| k = ∣ k ∣ the magnitude of the wave vector, and θ \theta θ the angle between k \mathbf{k} k and Δ x \Delta\mathbf{x} Δ x , so that k ⋅ Δ x = k r cos θ \mathbf{k}\cdot\Delta\mathbf{x}=kr\cos\theta k ⋅ Δ x = k r cos θ . The angular integral yields
∫ d Ω e i k r cos θ = 4 π sin ( k r ) k r . (9) \int\mathrm{d}\Omega e^{ikr\cos\theta}=\frac{4\pi\sin(kr)}{kr}. \tag*{(9)} ∫ d Ω e ik r c o s θ = k r 4 π sin ( k r ) . (9)
Substituting this result into the Wightman function of the Minkowski vacuum and decomposing the three-dimensional wave-vector integral into radial and angular parts, we obtain after successive simplifications
W i j ( x , x ′ ) = 1 8 π 2 i r ∫ 0 ∞ d k [ e − i k ( Δ t − r ) − e − i k ( Δ t + r ) ] . (10) W_{ij}(x,x')=\frac{1}{8\pi^2ir}\int_0^\infty\mathrm{d}k\left[e^{-ik(\Delta t-r)}-e^{-ik(\Delta t+r)}\right]. \tag*{(10)} W ij ( x , x ′ ) = 8 π 2 i r 1 ∫ 0 ∞ d k [ e − ik ( Δ t − r ) − e − ik ( Δ t + r ) ] . (10)
The two integrals derived above are oscillatory integrals over the positive half-line, and they do not converge in the ordinary Riemann sense. We adopt the standard i ϵ i\epsilon i ϵ prescription by analytically continuing the time interval as Δ t → Δ t − i ϵ \Delta t\to\Delta t-i\epsilon Δ t → Δ t − i ϵ with ϵ → 0 + \epsilon\to0^+ ϵ → 0 + , which defines the integrals in the distributional sense. The integral identity reads
∫ 0 ∞ e − i k u d k = π δ ( u ) − i PV 1 u , (11) \int_0^\infty e^{-iku}\mathrm{d}k=\pi\delta(u)-i\operatorname{PV}\frac{1}{u}, \tag*{(11)} ∫ 0 ∞ e − ik u d k = π δ ( u ) − i PV u 1 , (11)
where PV denotes the Cauchy principal value.
Substituting this expression into the one-dimensional Wightman function, we obtain
W i j ( x , x ′ ) = − 1 4 π 2 1 ( Δ t − i ϵ ) 2 − ∣ Δ x ∣ 2 . (12) W_{ij}(x,x')=-\frac{1}{4\pi^2}\frac{1}{(\Delta t-i\epsilon)^2-|\Delta\mathbf{x}|^2}. \tag*{(12)} W ij ( x , x ′ ) = − 4 π 2 1 ( Δ t − i ϵ ) 2 − ∣Δ x ∣ 2 1 . (12)
We let Δ τ = τ − τ ′ \Delta\tau=\tau-\tau' Δ τ = τ − τ ′ . Using Eqs. (2) and (3) , we obtain
Δ t = 2 a cosh a ( τ + τ ′ ) 2 sinh a Δ τ 2 , \Delta t=\frac{2}{a}\cosh\frac{a(\tau+\tau')}{2}\sinh\frac{a\Delta\tau}{2}, Δ t = a 2 cosh 2 a ( τ + τ ′ ) sinh 2 a Δ τ ,
Δ x = 2 a sinh a ( τ + τ ′ ) 2 sinh a Δ τ 2 . (13) \Delta x=\frac{2}{a}\sinh\frac{a(\tau+\tau')}{2}\sinh\frac{a\Delta\tau}{2}. \tag*{(13)} Δ x = a 2 sinh 2 a ( τ + τ ′ ) sinh 2 a Δ τ . (13)
Therefore,
Δ t 2 − Δ x 2 = 4 a 2 sinh 2 a Δ τ 2 . (14) \Delta t^2-\Delta x^2=\frac{4}{a^2}\sinh^2\frac{a\Delta\tau}{2}. \tag*{(14)} Δ t 2 − Δ x 2 = a 2 4 sinh 2 2 a Δ τ . (14)
Substituting Eq. (14) into Eq. (12) , we arrive at the single-detector Wightman function:
W A A ( Δ τ ) = W B B ( Δ τ ) = − a 2 16 π 2 1 sinh 2 [ a 2 ( Δ τ − i 0 + ) ] . (15) W_{AA}(\Delta\tau)=W_{BB}(\Delta\tau)=-\frac{a^2}{16\pi^2}\frac{1}{\sinh^2\left[\frac{a}{2}(\Delta\tau-i0^+)\right]}. \tag*{(15)} W AA ( Δ τ ) = W B B ( Δ τ ) = − 16 π 2 a 2 sinh 2 [ 2 a ( Δ τ − i 0 + ) ] 1 . (15)
It is worth noting that the resulting Wightman function depends exclusively on the proper time interval Δ τ \Delta\tau Δ τ .
We define the Fourier transform of the two-point function as
W ~ i j ( Ω ) = ∫ − ∞ ∞ d Δ τ e i Ω Δ τ W i j ( Δ τ ) , (16) \widetilde{W}_{ij}(\Omega)=\int_{-\infty}^{\infty}\mathrm{d}\Delta\tau\,e^{i\Omega\Delta\tau}W_{ij}(\Delta\tau), \tag*{(16)} W ij ( Ω ) = ∫ − ∞ ∞ d Δ τ e i ΩΔ τ W ij ( Δ τ ) , (16)
where Ω \Omega Ω is the frequency conjugate to the detector’s proper time.
Taking the Fourier transform of Eq. (15) , we obtain (For details, see Appendix A):
W ~ A A ( Ω ) = W ~ B B ( Ω ) = Ω 2 π 1 1 − e − 2 π Ω / a , Ω > 0. (17) \widetilde{W}_{AA}(\Omega)=\widetilde{W}_{BB}(\Omega)=\frac{\Omega}{2\pi}\frac{1}{1-e^{-2\pi\Omega/a}},\qquad\Omega>0. \tag*{(17)} W AA ( Ω ) = W B B ( Ω ) = 2 π Ω 1 − e − 2 π Ω/ a 1 , Ω > 0. (17)
The mean occupation number of quanta with frequency Ω \Omega Ω at temperature T T T obeys the Bose–Einstein distribution:
n ( Ω ) = 1 e Ω / T − 1 . (18) n(\Omega)=\frac{1}{e^{\Omega/T}-1}. \tag*{(18)} n ( Ω ) = e Ω/ T − 1 1 . (18)
A simple algebraic identity for the Bose distribution can be obtained by multiplying its numerator and denominator by e − Ω / T e^{-\Omega/T} e − Ω/ T :
n ( Ω ) + 1 = 1 1 − e − Ω / T . (19) n(\Omega)+1=\frac{1}{1-e^{-\Omega/T}}. \tag*{(19)} n ( Ω ) + 1 = 1 − e − Ω/ T 1 . (19)
By directly comparing Eq. (17) with the above identity, we find that the factor 1 / ( 1 − e − 2 π Ω / a ) 1/(1-e^{-2\pi\Omega/a}) 1/ ( 1 − e − 2 π Ω/ a ) in the Wightman spectrum takes precisely the form n ( Ω ) + 1 n(\Omega)+1 n ( Ω ) + 1 . The Unruh temperature can thus be directly read off [16 , 18 ]
T U = a 2 π , (20) T_U=\frac{a}{2\pi}, \tag*{(20)} T U = 2 π a , (20)
where T U T_U T U is the standard Unruh temperature for a uniformly accelerated observer with proper acceleration a a a .
Introducing the thermal occupation number evaluated at the Unruh temperature
n U ( Ω ) = 1 e 2 π Ω / a − 1 , (21) n_U(\Omega)=\frac{1}{e^{2\pi\Omega/a}-1}, \tag*{(21)} n U ( Ω ) = e 2 π Ω/ a − 1 1 , (21)
the positive-frequency Wightman function can be recast into the physically transparent form
W ~ A A ( Ω ) = W ~ B B ( Ω ) = Ω 2 π [ n U ( Ω ) + 1 ] , Ω > 0. (22) \widetilde{W}_{AA}(\Omega)=\widetilde{W}_{BB}(\Omega)=\frac{\Omega}{2\pi}\left[n_U(\Omega)+1\right],\qquad\Omega>0. \tag*{(22)} W AA ( Ω ) = W B B ( Ω ) = 2 π Ω [ n U ( Ω ) + 1 ] , Ω > 0. (22)
For Ω > 0 \Omega>0 Ω > 0 , we readily obtain:
W ~ A A ( − Ω ) = W ~ B B ( − Ω ) = Ω 2 π n U ( Ω ) . (23) \widetilde{W}_{AA}(-\Omega)=\widetilde{W}_{BB}(-\Omega)=\frac{\Omega}{2\pi}n_U(\Omega). \tag*{(23)} W AA ( − Ω ) = W B B ( − Ω ) = 2 π Ω n U ( Ω ) . (23)
This satisfies the detailed balance condition for thermal equilibrium, i.e., the Kubo–Martin–Schwinger (KMS) [90 , 91 , 92 , 93 ] condition:
W ~ A A ( − Ω ) W ~ A A ( Ω ) = W ~ B B ( − Ω ) W ~ B B ( Ω ) = e − β U Ω . (24) \frac{\widetilde{W}_{AA}(-\Omega)}{\widetilde{W}_{AA}(\Omega)}=\frac{\widetilde{W}_{BB}(-\Omega)}{\widetilde{W}_{BB}(\Omega)}=e^{-\beta_U\Omega}. \tag*{(24)} W AA ( Ω ) W AA ( − Ω ) = W B B ( Ω ) W B B ( − Ω ) = e − β U Ω . (24)
Since the two detectors undergo identical motion in the t t t -x x x plane, the squared spatial separation simply acquires an extra transverse term ∣ Δ x ∣ 2 = Δ x 2 + d 2 \lvert\Delta x\rvert^2=\Delta x^2+d^2 ∣ Δ x ∣ 2 = Δ x 2 + d 2 . The Wightman function between the two distinct detectors is then given by:
W A B ( Δ τ ) = W B A ( Δ τ ) = − a 2 16 π 2 1 sinh 2 [ a 2 ( Δ τ − i 0 + ) ] − ( a d 2 ) 2 . (25) W_{AB}(\Delta\tau)=W_{BA}(\Delta\tau)=-\frac{a^2}{16\pi^2}\frac{1}{\sinh^2\left[\frac{a}{2}(\Delta\tau-i0^+)\right]-\left(\frac{ad}{2}\right)^2}. \tag*{(25)} W A B ( Δ τ ) = W B A ( Δ τ ) = − 16 π 2 a 2 sinh 2 [ 2 a ( Δ τ − i 0 + ) ] − ( 2 a d ) 2 1 . (25)
Taking the Fourier transform of Eq. (25) , we obtain (For details, see Appendix A):
W ~ A B ( Ω ) = W ~ B A ( Ω ) = 1 2 π d 1 + ( a d 2 ) 2 sin [ 2 Ω a arsinh a d 2 ] 1 − e − 2 π Ω / a . (26) \widetilde{W}_{AB}(\Omega)=\widetilde{W}_{BA}(\Omega)=\frac{1}{2\pi d\sqrt{1+\left(\frac{ad}{2}\right)^2}}\frac{\sin\left[\frac{2\Omega}{a}\operatorname{arsinh}\frac{ad}{2}\right]}{1-e^{-2\pi\Omega/a}}. \tag*{(26)} W A B ( Ω ) = W B A ( Ω ) = 2 π d 1 + ( 2 a d ) 2 1 1 − e − 2 π Ω/ a sin [ a 2Ω arsinh 2 a d ] . (26)
We define the geometric correlation factor γ E \gamma_E γ E as:
γ E ( Ω , a , d ) ≡ W ~ A B ( Ω ) W ~ A A ( Ω ) = sin [ 2 Ω a arsinh a d 2 ] Ω d 1 + ( a d 2 ) 2 . (27) \gamma_E(\Omega,a,d)\equiv\frac{\widetilde{W}_{AB}(\Omega)}{\widetilde{W}_{AA}(\Omega)}=\frac{\sin\left[\frac{2\Omega}{a}\operatorname{arsinh}\frac{ad}{2}\right]}{\Omega d\sqrt{1+\left(\frac{ad}{2}\right)^2}}. \tag*{(27)} γ E ( Ω , a , d ) ≡ W AA ( Ω ) W A B ( Ω ) = Ω d 1 + ( 2 a d ) 2 sin [ a 2Ω arsinh 2 a d ] . (27)
The geometric correlation factor γ E \gamma_E γ E satisfies the following three fundamental properties (The proof is given in Appendix A):
Property 1 (Coincident-point limit). In the limit of vanishing detector separation d → 0 d\to0 d → 0 , the correlation factor approaches unity:
γ E → 1 as d → 0. (28) \gamma_E\to1\quad\text{as }d\to0. \tag*{(28)} γ E → 1 as d → 0. (28)
Property 2 (Inertial limit). In the limit a → 0 a\to0 a → 0 , the result reduces to
γ E → sin Ω d Ω d as a → 0. (29) \gamma_E\to\frac{\sin\Omega d}{\Omega d}\quad\text{as }a\to0. \tag*{(29)} γ E → Ω d sin Ω d as a → 0. (29)
Property 3 (Boundedness). The magnitude of the correlation factor is uniformly bounded by unity:
∣ γ E ∣ ≤ 1. (30) \lvert\gamma_E\rvert\le1. \tag*{(30)} ∣ γ E ∣ ≤ 1. (30)
The field we consider here is a Gaussian scalar field, for which all field expectation values are fully determined by the Wightman function defined in Eq. (12) . We arrange the same-detector and cross-detector Wightman functions into the following matrix form:
W ~ U ( Ω ) = ( W ~ A A ( Ω ) W ~ A B ( Ω ) W ~ B A ( Ω ) W ~ B B ( Ω ) ) = Ω 2 π [ n U ( Ω ) + 1 ] ( 1 γ E ( Ω ) γ E ( Ω ) 1 ) . (31) \widetilde{W}_U(\Omega)=
\begin{pmatrix}
\widetilde{W}_{AA}(\Omega) & \widetilde{W}_{AB}(\Omega)\\
\widetilde{W}_{BA}(\Omega) & \widetilde{W}_{BB}(\Omega)
\end{pmatrix}
=\frac{\Omega}{2\pi}\left[n_U(\Omega)+1\right]
\begin{pmatrix}
1 & \gamma_E(\Omega)\\
\gamma_E(\Omega) & 1
\end{pmatrix}. \tag*{(31)} W U ( Ω ) = ( W AA ( Ω ) W B A ( Ω ) W A B ( Ω ) W B B ( Ω ) ) = 2 π Ω [ n U ( Ω ) + 1 ] ( 1 γ E ( Ω ) γ E ( Ω ) 1 ) . (31)
W ~ U ( − Ω ) = ( W ~ A A ( − Ω ) W ~ A B ( − Ω ) W ~ B A ( − Ω ) W ~ B B ( − Ω ) ) = Ω 2 π n U ( Ω ) ( 1 γ E ( Ω ) γ E ( Ω ) 1 ) . (32) \widetilde{W}_U(-\Omega)=
\begin{pmatrix}
\widetilde{W}_{AA}(-\Omega) & \widetilde{W}_{AB}(-\Omega)\\
\widetilde{W}_{BA}(-\Omega) & \widetilde{W}_{BB}(-\Omega)
\end{pmatrix}
=\frac{\Omega}{2\pi}n_U(\Omega)
\begin{pmatrix}
1 & \gamma_E(\Omega)\\
\gamma_E(\Omega) & 1
\end{pmatrix}. \tag*{(32)} W U ( − Ω ) = ( W AA ( − Ω ) W B A ( − Ω ) W A B ( − Ω ) W B B ( − Ω ) ) = 2 π Ω n U ( Ω ) ( 1 γ E ( Ω ) γ E ( Ω ) 1 ) . (32)
This also satisfies the detailed balance (KMS) condition for thermal equilibrium:
W ~ U ( − Ω ) = e − β U Ω W ~ U T ( Ω ) . (33) \widetilde{W}_U(-\Omega)=e^{-\beta_U\Omega}\widetilde{W}_U^T(\Omega). \tag*{(33)} W U ( − Ω ) = e − β U Ω W U T ( Ω ) . (33)
The transpose exchanges the two detector indices; the spectral matrix is symmetric for the present trajectories. We can also obtain its vacuum correlation matrix:
W ~ U v a c ( Ω ) = Ω 2 π ( 1 γ U ( Ω ) γ U ( Ω ) 1 ) . (34) \widetilde{W}^{\mathrm{vac}}_U(\Omega)=\frac{\Omega}{2\pi}
\begin{pmatrix}
1 & \gamma_U(\Omega)\\
\gamma_U(\Omega) & 1
\end{pmatrix}. \tag*{(34)} W U vac ( Ω ) = 2 π Ω ( 1 γ U ( Ω ) γ U ( Ω ) 1 ) . (34)
We next construct a matched inertial thermal bath by considering two inertial detectors at rest in the same inertial frame, separated by a fixed distance d d d , with the massless scalar field prepared in a thermal state at temperature T U = a / ( 2 π ) T_U=a/(2\pi) T U = a / ( 2 π ) . The corresponding thermal occupation number is given by
n T ( Ω ) = n U ( Ω ) . (35) n_T(\Omega)=n_U(\Omega). \tag*{(35)} n T ( Ω ) = n U ( Ω ) . (35)
The geometric correlation factor is given by
γ T ( Ω , d ) = lim a → 0 γ E ( Ω , a , d ) = sin ( Ω d ) Ω d . (36) \gamma_T(\Omega,d)=\lim_{a\to0}\gamma_E(\Omega,a,d)=\frac{\sin(\Omega d)}{\Omega d}. \tag*{(36)} γ T ( Ω , d ) = a → 0 lim γ E ( Ω , a , d ) = Ω d sin ( Ω d ) . (36)
The corresponding spectral matrix reads
W ~ T ( Ω ) = Ω 2 π [ n U ( Ω ) + 1 ] ( 1 γ T ( Ω , d ) γ T ( Ω , d ) 1 ) . (37) \widetilde{W}_T(\Omega)=\frac{\Omega}{2\pi}\left[n_U(\Omega)+1\right]
\begin{pmatrix}
1 & \gamma_T(\Omega,d)\\
\gamma_T(\Omega,d) & 1
\end{pmatrix}. \tag*{(37)} W T ( Ω ) = 2 π Ω [ n U ( Ω ) + 1 ] ( 1 γ T ( Ω , d ) γ T ( Ω , d ) 1 ) . (37)
The local spectral contribution coincides exactly with the corresponding term in Eq. (31) . Consequently, the two environments have the same temperature and identical local spectra, whereas their cross spectra are generally different.
Finally, we introduce an uncorrelated reference thermal bath, consisting of two independent local thermal reservoirs, each coupled to one detector. Both reservoirs are maintained at the same temperature T U T_U T U , so that their thermal occupation numbers coincide with those of the previous two environments. Since the two reservoirs are statistically independent, the cross-correlation between the corresponding bath forces vanishes. Consequently, the geometric correlation factor is given by
γ C = 0. (38) \gamma_C=0. \tag*{(38)} γ C = 0. (38)
The corresponding spectral matrix is therefore given by
W ~ C ( Ω ) = Ω 2 π [ n U ( Ω ) + 1 ] ( 1 0 0 1 ) . (39) \widetilde{W}_C(\Omega)=\frac{\Omega}{2\pi}\left[n_U(\Omega)+1\right]
\begin{pmatrix}
1 & 0\\
0 & 1
\end{pmatrix}. \tag*{(39)} W C ( Ω ) = 2 π Ω [ n U ( Ω ) + 1 ] ( 1 0 0 1 ) . (39)
In what follows, we use E ∈ { U , T , C } E\in\{U,T,C\} E ∈ { U , T , C } to label the three environments.
We now model each detector as a harmonic oscillator with natural frequency ω \omega ω . When the two detectors are directly coupled, the total system Hamiltonian is given by (Appendix A)
H S = ω 2 ( q A 2 + p A 2 ) + ω 2 ( q B 2 + p B 2 ) + λ ( q A q B − p A p B ) , (40) \mathcal{H}_{S}=\frac{\omega}{2}\left(q_{A}^{2}+p_{A}^{2}\right)+\frac{\omega}{2}\left(q_{B}^{2}+p_{B}^{2}\right)+\lambda\left(q_{A}q_{B}-p_{A}p_{B}\right), \tag*{(40)} H S = 2 ω ( q A 2 + p A 2 ) + 2 ω ( q B 2 + p B 2 ) + λ ( q A q B − p A p B ) , (40)
where λ \lambda λ parametrizes the strength of the direct bilinear coupling between the two detectors. q i q_i q i and p i p_i p i are the canonical coordinate and canonical momentum of the system, respectively, with i ∈ { A , B } i \in\{A,B\} i ∈ { A , B } .
From Eq. (5) , the Hamiltonian of the free field is given by
H f i e l d = ∫ d 3 k ( 2 π ) 3 ω k a k † a k . (41) \mathcal{H}_{\mathrm{field}}=\int\frac{\mathrm{d}^{3}k}{(2\pi)^{3}}\,\omega_k\,a_k^\dagger a_k. \tag*{(41)} H field = ∫ ( 2 π ) 3 d 3 k ω k a k † a k . (41)
The interaction Hamiltonian between the detectors and the field is given by [94 ]
H i n t ( τ ) = κ χ ( τ ) [ q A ( τ ) ϕ ^ ( x A ( τ ) ) + q B ( τ ) ϕ ^ ( x B ( τ ) ) ] , (42) \mathcal{H}_{\mathrm{int}}(\tau)=\kappa\chi(\tau)\left[q_A(\tau)\hat{\phi}\left(x_A(\tau)\right)+q_B(\tau)\hat{\phi}\left(x_B(\tau)\right)\right], \tag*{(42)} H int ( τ ) = κ χ ( τ ) [ q A ( τ ) ϕ ^ ( x A ( τ ) ) + q B ( τ ) ϕ ^ ( x B ( τ ) ) ] , (42)
where κ \kappa κ is the detector-field coupling constant and χ ( τ ) \chi(\tau) χ ( τ ) is the proper-time switching modulation function (Appendix A).
According to the Feynman–Vernon influence-functional theory [78 ] , for a Gaussian field linearly coupled to Gaussian detectors, Wick’s theorem [95 ] and the Gaussian influence-functional construction [78 , 79 , 80 , 81 ] ensure that the reduced detector dynamics is entirely encoded in the two-point Wightman function [88 ] along the detector worldlines; the resulting Gaussian detector dynamics can be expressed using the standard covariance-matrix formalism [96 ] . A Gaussian environment with a prescribed two-point spectrum can equivalently be represented by a collection of harmonic oscillators through spectral-density matching [79 , 80 , 97 ] ; consequently, the explicit three-dimensional field may be replaced by an effective one-dimensional harmonic-oscillator bath without changing the detector dynamics.
For each continuous frequency Ω ∈ [ 0 , ∞ ) \Omega\in[0,\infty) Ω ∈ [ 0 , ∞ ) , we introduce two bath oscillators at every frequency, with their canonical coordinates and momenta denoted by Q Ω = ( Q Ω , + , Q Ω , − ) T \mathbf{Q}_{\Omega}=(Q_{\Omega,+},Q_{\Omega,-})^{\mathrm{T}} Q Ω = ( Q Ω , + , Q Ω , − ) T , P Ω = ( P Ω , + , P Ω , − ) T \mathbf{P}_{\Omega}=(P_{\Omega,+},P_{\Omega,-})^{\mathrm{T}} P Ω = ( P Ω , + , P Ω , − ) T . We denote the coordinate vector of the two-detector system as q = ( q A , q B ) T \mathbf{q}=(q_A,q_B)^{\mathrm{T}} q = ( q A , q B ) T . The canonical commutation relations read:
[ Q Ω , P Ω ′ T ] = i I 2 δ ( Ω − Ω ′ ) , [ Q Ω , Q Ω ′ T ] = [ P Ω , P Ω ′ T ] = 0 , [\mathbf{Q}_{\Omega},\mathbf{P}_{\Omega'}^{\mathrm{T}}]=iI_2\delta(\Omega-\Omega'),\qquad[\mathbf{Q}_{\Omega},\mathbf{Q}_{\Omega'}^{\mathrm{T}}]=[\mathbf{P}_{\Omega},\mathbf{P}_{\Omega'}^{\mathrm{T}}]=0, [ Q Ω , P Ω ′ T ] = i I 2 δ ( Ω − Ω ′ ) , [ Q Ω , Q Ω ′ T ] = [ P Ω , P Ω ′ T ] = 0 ,
where I 2 I_2 I 2 denotes the unity matrix in 2 × 2 2\times2 2 × 2 space.
The bath Hamiltonian and the system-bath coupling Hamiltonian can then be written as:
H b a t h = ∫ 0 ∞ d Ω Ω 2 ∑ ζ = ± ( Q Ω , ζ 2 + P Ω , ζ 2 ) , (43) \mathcal{H}_{\mathrm{bath}}=\int_{0}^{\infty}\mathrm{d}\Omega\,\frac{\Omega}{2}\sum_{\zeta=\pm}\left(Q_{\Omega,\zeta}^{2}+P_{\Omega,\zeta}^{2}\right), \tag*{(43)} H bath = ∫ 0 ∞ d Ω 2 Ω ζ = ± ∑ ( Q Ω , ζ 2 + P Ω , ζ 2 ) , (43)
H S B ( τ ) = χ ( τ ) ∫ 0 ∞ d Ω q T h ( Ω ) Q Ω , (44) \mathcal{H}_{SB}(\tau)=\chi(\tau)\int_{0}^{\infty}\mathrm{d}\Omega\,\mathbf{q}^{\mathrm{T}}\mathbf{h}(\Omega)\mathbf{Q}_{\Omega}, \tag*{(44)} H S B ( τ ) = χ ( τ ) ∫ 0 ∞ d Ω q T h ( Ω ) Q Ω , (44)
where h ( Ω ) \mathbf{h}(\Omega) h ( Ω ) is a 2 × 2 2\times2 2 × 2 continuous coupling matrix: its rows correspond to detectors A A A and B B B , and its columns correspond to bath channels + + + and − - − .
We define the spectral density matrix as:
J E ( Ω ) = π 2 h ( Ω ) h ( Ω ) T . (45) \mathbf{J}_{E}(\Omega)=\frac{\pi}{2}\mathbf{h}(\Omega)\mathbf{h}(\Omega)^{\mathrm{T}}. \tag*{(45)} J E ( Ω ) = 2 π h ( Ω ) h ( Ω ) T . (45)
The two-time correlation function of the collective bath coordinate
B E ( τ ) = ∫ d Ω h ( Ω ) Q Ω ( τ ) (46) \mathbf{B}_{E}(\tau)=\int\mathrm{d}\Omega\,\mathbf{h}(\Omega)\mathbf{Q}_{\Omega}(\tau) \tag*{(46)} B E ( τ ) = ∫ d Ω h ( Ω ) Q Ω ( τ ) (46)
is computed from Eq. (7) as
C b a t h E ( Δ τ ) = ∫ 0 ∞ d Ω π J ( Ω ) [ ( n U ( Ω ) + 1 ) e − i Ω Δ τ + n U ( Ω ) e i Ω Δ τ ] . (47) C_{\mathrm{bath}}^{E}(\Delta\tau)=\int_{0}^{\infty}\frac{\mathrm{d}\Omega}{\pi}\mathbf{J}(\Omega)\left[(n_{U}(\Omega)+1)e^{-i\Omega\Delta\tau}+n_{U}(\Omega)e^{i\Omega\Delta\tau}\right]. \tag*{(47)} C bath E ( Δ τ ) = ∫ 0 ∞ π d Ω J ( Ω ) [ ( n U ( Ω ) + 1 ) e − i ΩΔ τ + n U ( Ω ) e i ΩΔ τ ] . (47)
Since W ~ A A \widetilde{W}_{AA} W AA grows linearly at high frequencies (the intrinsic UV divergence of point-like detectors), we introduce a phenomenological regularization. The regularized field correlation function is then given by
C f i e l d E ( Δ τ ) = κ 2 ∫ 0 ∞ d Ω 2 π W ~ v a c ( Ω ) e − Ω / Ω c [ ( n U ( Ω ) + 1 ) e − i Ω Δ τ + n U ( Ω ) e i Ω Δ τ ] , (48) C_{\mathrm{field}}^{E}(\Delta\tau)=\kappa^{2}\int_{0}^{\infty}\frac{\mathrm{d}\Omega}{2\pi}\widetilde{W}^{\mathrm{vac}}(\Omega)e^{-\Omega/\Omega_{c}}\left[(n_{U}(\Omega)+1)e^{-i\Omega\Delta\tau}+n_{U}(\Omega)e^{i\Omega\Delta\tau}\right], \tag*{(48)} C field E ( Δ τ ) = κ 2 ∫ 0 ∞ 2 π d Ω W vac ( Ω ) e − Ω/ Ω c [ ( n U ( Ω ) + 1 ) e − i ΩΔ τ + n U ( Ω ) e i ΩΔ τ ] , (48)
where Ω c \Omega_c Ω c is the UV cutoff frequency.
To ensure that our effective thermal bath produces the same physical effect as the quantum field, we set
C b a t h E = C f i e l d E , (49) C_{\mathrm{bath}}^{E}=C_{\mathrm{field}}^{E}, \tag*{(49)} C bath E = C field E , (49)
which yields:
J E ( Ω ) = κ 2 2 W ~ v a c ( Ω ) e − Ω / Ω c . (50) \mathbf{J}_{E}(\Omega)=\frac{\kappa^{2}}{2}\widetilde{W}^{\mathrm{vac}}(\Omega)e^{-\Omega/\Omega_{c}}. \tag*{(50)} J E ( Ω ) = 2 κ 2 W vac ( Ω ) e − Ω/ Ω c . (50)
Further substituting into Eq. (34) yields:
J E ( Ω ) = J 0 ( Ω ) ( 1 γ E ( Ω ) γ E ( Ω ) 1 ) , (51) \mathbf{J}_{E}(\Omega)=J_{0}(\Omega)\begin{pmatrix}1 & \gamma_{E}(\Omega) \\ \gamma_{E}(\Omega) & 1\end{pmatrix}, \tag*{(51)} J E ( Ω ) = J 0 ( Ω ) ( 1 γ E ( Ω ) γ E ( Ω ) 1 ) , (51)
where J 0 ( Ω ) = κ 2 4 π Ω e − Ω / Ω c J_{0}(\Omega)=\frac{\kappa^{2}}{4\pi}\Omega e^{-\Omega/\Omega_{c}} J 0 ( Ω ) = 4 π κ 2 Ω e − Ω/ Ω c . We have det J E ( Ω ) = J 0 2 ( Ω ) ( 1 − γ E 2 ( Ω ) ) ≥ 0 \det\mathbf{J}_{E}(\Omega)=J_{0}^{2}(\Omega)(1-\gamma_{E}^{2}(\Omega))\geq0 det J E ( Ω ) = J 0 2 ( Ω ) ( 1 − γ E 2 ( Ω )) ≥ 0 , which is guaranteed by ∣ γ E ( Ω ) ∣ ≤ 1 |\gamma_{E}(\Omega)|\leq1 ∣ γ E ( Ω ) ∣ ≤ 1 . Together with the non-negativity of its diagonal entries, we conclude that
J E ( Ω ) ⪰ 0. \mathbf{J}_{E}(\Omega)\succeq0. J E ( Ω ) ⪰ 0.
From Eq. (45) , we readily obtain:
h ( Ω ) = 1 2 ( h 1 ( Ω ) h 2 ( Ω ) h 1 ( Ω ) − h 2 ( Ω ) ) , (52) \mathbf{h}(\Omega)=\frac{1}{\sqrt{2}}\begin{pmatrix}h_{1}(\Omega)&h_{2}(\Omega)\\h_{1}(\Omega)&-h_{2}(\Omega)\end{pmatrix}, \tag*{(52)} h ( Ω ) = 2 1 ( h 1 ( Ω ) h 1 ( Ω ) h 2 ( Ω ) − h 2 ( Ω ) ) , (52)
where
h 1 ( Ω ) 2 = 2 π J 0 ( Ω ) ( 1 + γ E ( Ω ) ) , h_{1}(\Omega)^{2}=\frac{2}{\pi}J_{0}(\Omega)(1+\gamma_{E}(\Omega)), h 1 ( Ω ) 2 = π 2 J 0 ( Ω ) ( 1 + γ E ( Ω )) ,
e n d s m i d = 1 {ends_mid=1} e n d s m i d = 1
h 2 ( Ω ) 2 = 2 π J 0 ( Ω ) ( 1 − γ E ( Ω ) ) . h_2(\Omega)^2 = \frac{2}{\pi}J_0(\Omega)\left(1-\gamma_E(\Omega)\right). h 2 ( Ω ) 2 = π 2 J 0 ( Ω ) ( 1 − γ E ( Ω ) ) .
To discretize the bath, we introduce a set of positive-frequency nodes { Ω j } j = 1 N \{\Omega_j\}_{j=1}^{N} { Ω j } j = 1 N together with the corresponding positive quadrature weights { w j } j = 1 N \{w_j\}_{j=1}^{N} { w j } j = 1 N , such that
∫ 0 ∞ f ( Ω ) d Ω ≃ ∑ j = 1 N w j f ( Ω j ) . (53) \int_0^\infty f(\Omega)\,\mathrm{d}\Omega\simeq\sum_{j=1}^{N} w_j f(\Omega_j). \tag*{(53)} ∫ 0 ∞ f ( Ω ) d Ω ≃ j = 1 ∑ N w j f ( Ω j ) . (53)
The dimensionless discrete canonical variables are then defined by
Q j , ζ = w j Q Ω j , ζ , P j , ζ = w j P Ω j , ζ , ζ ∈ { + , − } . (54) Q_{j,\zeta}=\sqrt{w_j}\,Q_{\Omega_j,\zeta},\qquad P_{j,\zeta}=\sqrt{w_j}\,P_{\Omega_j,\zeta},\qquad\zeta\in\{+,-\}. \tag*{(54)} Q j , ζ = w j Q Ω j , ζ , P j , ζ = w j P Ω j , ζ , ζ ∈ { + , − } . (54)
At each frequency node, we thus obtain two bath oscillators corresponding to the symmetric and antisymmetric channels. They satisfy the canonical commutation relations
[ Q i , ζ , P j , ζ ′ ] = i δ i j δ ζ ζ ′ , ζ , ζ ′ ∈ { + , − } . (55) [Q_{i,\zeta},P_{j,\zeta'}]=\mathrm{i}\delta_{ij}\delta_{\zeta\zeta'},\qquad\zeta,\zeta'\in\{+,-\}. \tag*{(55)} [ Q i , ζ , P j , ζ ′ ] = i δ ij δ ζ ζ ′ , ζ , ζ ′ ∈ { + , − } . (55)
The discrete coupling matrix is defined as
g j = w j h ( Ω j ) . (56) \mathbf{g}_j=\sqrt{w_j}\,\mathbf{h}(\Omega_j). \tag*{(56)} g j = w j h ( Ω j ) . (56)
The discretized thermal bath Hamiltonian is then given by
H B = ∑ j = 1 N Ω j 2 ( Q j , + 2 + P j , + 2 + Q j , − 2 + P j , − 2 ) . (57) \mathcal{H}_B=\sum_{j=1}^{N}\frac{\Omega_j}{2}\left(Q_{j,+}^2+P_{j,+}^2+Q_{j,-}^2+P_{j,-}^2\right). \tag*{(57)} H B = j = 1 ∑ N 2 Ω j ( Q j , + 2 + P j , + 2 + Q j , − 2 + P j , − 2 ) . (57)
The system-bath coupling Hamiltonian reads
H S B ( τ ) = χ ( τ ) ∑ j = 1 N [ g j , + 2 ( q A + q B ) Q j , + + g j , − 2 ( q A − q B ) Q j , − ] . (58) \mathcal{H}_{SB}(\tau)=\chi(\tau)\sum_{j=1}^{N}\left[\frac{g_{j,+}}{\sqrt{2}}(q_A+q_B)Q_{j,+}+\frac{g_{j,-}}{\sqrt{2}}(q_A-q_B)Q_{j,-}\right]. \tag*{(58)} H S B ( τ ) = χ ( τ ) j = 1 ∑ N [ 2 g j , + ( q A + q B ) Q j , + + 2 g j , − ( q A − q B ) Q j , − ] . (58)
Here the discrete coupling strengths are fixed by spectral matching,
g j , + 2 = 2 w j π J 0 ( Ω j ) ( 1 + γ j ) , g j , − 2 = 2 w j π J 0 ( Ω j ) ( 1 − γ j ) , (59) g_{j,+}^2=\frac{2w_j}{\pi}J_0(\Omega_j)(1+\gamma_j),\qquad g_{j,-}^2=\frac{2w_j}{\pi}J_0(\Omega_j)(1-\gamma_j), \tag*{(59)} g j , + 2 = π 2 w j J 0 ( Ω j ) ( 1 + γ j ) , g j , − 2 = π 2 w j J 0 ( Ω j ) ( 1 − γ j ) , (59)
where γ j ≡ γ E ( Ω j ) \gamma_j\equiv\gamma_E(\Omega_j) γ j ≡ γ E ( Ω j ) .
We arrange the discrete coupling strengths into a 2 × 2 2\times2 2 × 2 matrix (rows correspond to detectors A A A , B B B ; columns correspond to bath channels 1, 2):
g j = 1 2 ( g j , + g j , − g j , + − g j , − ) , (60) \mathbf{g}_j=\frac{1}{\sqrt{2}}
\begin{pmatrix}
g_{j,+} & g_{j,-}\\
g_{j,+} & -g_{j,-}
\end{pmatrix}, \tag*{(60)} g j = 2 1 ( g j , + g j , + g j , − − g j , − ) , (60)
which gives the matrix spectral matching condition
1 2 g j g j T = w j π J E ( Ω j ) . (61) \frac{1}{2}\mathbf{g}_j\mathbf{g}_j^{T}=\frac{w_j}{\pi}\mathbf{J}_E(\Omega_j). \tag*{(61)} 2 1 g j g j T = π w j J E ( Ω j ) . (61)
Figure 1. Schematic of the two-detector setup in the accelerated-vacuum (U U U ) and matched inertial-thermal (T T T ) environments. A single setting may exhibit a blind point, whereas two opposite settings (± λ \pm\lambda ± λ ) enable the reconstruction of spatial correlations.
Completing the bath squares produces a negative static detector interaction. We choose the following counterterm to compensate that interaction and keep the calibrated static detector Hamiltonian equal to H S \mathcal{H}_S H S :
H C T ( τ ) = χ 2 ( τ ) 2 [ δ ω A ( q A 2 + q B 2 ) + 2 δ ω A B q A q B ] . (62) \mathcal{H}_{CT}(\tau)=\frac{\chi^2(\tau)}{2}\left[\delta\omega_A(q_A^2+q_B^2)+2\delta\omega_{AB}q_Aq_B\right]. \tag*{(62)} H C T ( τ ) = 2 χ 2 ( τ ) [ δ ω A ( q A 2 + q B 2 ) + 2 δ ω A B q A q B ] . (62)
with the static counterterm coefficients given directly by the quadrature,
δ ω A = ∑ j = 1 N 2 w j π J 0 ( Ω j ) Ω j = κ 2 2 π 2 ∑ j = 1 N w j e − Ω j / Ω c , (63) \delta\omega_A=\sum_{j=1}^{N}\frac{2w_j}{\pi}\frac{J_0(\Omega_j)}{\Omega_j}=\frac{\kappa^2}{2\pi^2}\sum_{j=1}^{N}w_j e^{-\Omega_j/\Omega_c}, \tag*{(63)} δ ω A = j = 1 ∑ N π 2 w j Ω j J 0 ( Ω j ) = 2 π 2 κ 2 j = 1 ∑ N w j e − Ω j / Ω c , (63)
δ ω A B = ∑ j = 1 N 2 w j π J 0 ( Ω j ) γ j Ω j = κ 2 2 π 2 ∑ j = 1 N w j e − Ω j / Ω c γ j . (64) \delta\omega_{AB}=\sum_{j=1}^{N}\frac{2w_j}{\pi}\frac{J_0(\Omega_j)\gamma_j}{\Omega_j}=\frac{\kappa^2}{2\pi^2}\sum_{j=1}^{N}w_j e^{-\Omega_j/\Omega_c}\gamma_j. \tag*{(64)} δ ω A B = j = 1 ∑ N π 2 w j Ω j J 0 ( Ω j ) γ j = 2 π 2 κ 2 j = 1 ∑ N w j e − Ω j / Ω c γ j . (64)
The total Hamiltonian is then given by
H t o t ( τ ) = H S + H B + H S B ( τ ) + H C T ( τ ) , (65) \mathcal{H}_{\mathrm{tot}}(\tau)=\mathcal{H}_S+\mathcal{H}_B+\mathcal{H}_{SB}(\tau)+\mathcal{H}_{CT}(\tau), \tag*{(65)} H tot ( τ ) = H S + H B + H S B ( τ ) + H C T ( τ ) , (65)
where the total number of oscillators is given by
n = 2 + 2 N . (66) n=2+2N. \tag*{(66)} n = 2 + 2 N . (66)
the time evolution of the covariance matrix
In this section, we evolve the state using the total Hamiltonian introduced in Sec. II, which is equivalent to propagating the corresponding covariance matrix. We first define the total coordinate vector
R = ( q A , q B , Q 1 , + , Q 1 , − , … , Q N , + , Q N , − , p A , p B , P 1 , + , P 1 , − , … , P N , + , P N , − ) T . (67) \mathbf{R}=\left(q_A,q_B,Q_{1,+},Q_{1,-},\ldots,Q_{N,+},Q_{N,-},p_A,p_B,P_{1,+},P_{1,-},\ldots,P_{N,+},P_{N,-}\right)^{T}. \tag*{(67)} R = ( q A , q B , Q 1 , + , Q 1 , − , … , Q N , + , Q N , − , p A , p B , P 1 , + , P 1 , − , … , P N , + , P N , − ) T . (67)
Since all four Hamiltonians are quadratic forms of the canonical variables, the total Hamiltonian can be written compactly as (Appendix A)
H t o t ( τ ) = 1 2 R T H ( τ ) R . (68) \mathcal{H}_{\mathrm{tot}}(\tau)=\frac{1}{2}\mathbf{R}^{T}H(\tau)\mathbf{R}. \tag*{(68)} H tot ( τ ) = 2 1 R T H ( τ ) R . (68)
Here, H ( τ ) H(\tau) H ( τ ) is a 2 n × 2 n 2n \times2n 2 n × 2 n real symmetric matrix, referred to as the Hamiltonian matrix (quadratic form matrix). It takes the block-diagonal form
H ( τ ) = H q q ( τ ) ⊕ H p p , (69) H(\tau)=H_{qq}(\tau)\oplus H_{pp}, \tag*{(69)} H ( τ ) = H q q ( τ ) ⊕ H pp , (69)
with
H q q ( τ ) = ( H q S + χ ( τ ) 2 H C T χ ( τ ) C χ ( τ ) C T D ) , H p p = ( H p S 0 0 D ) . (70) H_{qq}(\tau)=
\begin{pmatrix}
H_{q}^{S}+\chi(\tau)^{2}H_{CT} & \chi(\tau)C \\
\chi(\tau)C^{T} & D
\end{pmatrix},
\qquad
H_{pp}=
\begin{pmatrix}
H_{p}^{S} & 0 \\
0 & D
\end{pmatrix}. \tag*{(70)} H q q ( τ ) = ( H q S + χ ( τ ) 2 H C T χ ( τ ) C T χ ( τ ) C D ) , H pp = ( H p S 0 0 D ) . (70)
The individual submatrices are defined as follows.
(1) Detector coordinate and momentum blocks (each of size 2 × 2 2 \times2 2 × 2 ):
H q S = ( ω λ λ ω ) , H p S = ( ω − λ − λ ω ) . (71) H_{q}^{S}=
\begin{pmatrix}
\omega& \lambda\\
\lambda& \omega
\end{pmatrix},
\qquad
H_{p}^{S}=
\begin{pmatrix}
\omega& -\lambda\\
-\lambda& \omega
\end{pmatrix}. \tag*{(71)} H q S = ( ω λ λ ω ) , H p S = ( ω − λ − λ ω ) . (71)
(2) Counterterm block (2 × 2 2 \times2 2 × 2 , appearing only in the coordinate block):
H C T = ( δ ω A δ ω A B δ ω A B δ ω A ) . (72) H_{CT}=
\begin{pmatrix}
\delta\omega_{A} & \delta\omega_{AB} \\
\delta\omega_{AB} & \delta\omega_{A}
\end{pmatrix}. \tag*{(72)} H C T = ( δ ω A δ ω A B δ ω A B δ ω A ) . (72)
(3) Bath frequency block D D D (2 N × 2 N 2N \times2N 2 N × 2 N diagonal matrix): each frequency node hosts two channels, thus
D = diag ( Ω 1 , Ω 1 , Ω 2 , Ω 2 , … , Ω N , Ω N ) . (73) D=\operatorname{diag}(\Omega_{1},\Omega_{1},\Omega_{2},\Omega_{2},\ldots,\Omega_{N},\Omega_{N}). \tag*{(73)} D = diag ( Ω 1 , Ω 1 , Ω 2 , Ω 2 , … , Ω N , Ω N ) . (73)
(4) System-bath coupling block C C C (2 × 2 N 2 \times2N 2 × 2 N ): formed by horizontally concatenating 2 × 2 2 \times2 2 × 2 column blocks for each frequency node,
C = ( g 1 g 2 ⋯ g N ) . (74) C=\begin{pmatrix}\mathbf{g}_{1}&\mathbf{g}_{2}&\cdots&\mathbf{g}_{N}\end{pmatrix}. \tag*{(74)} C = ( g 1 g 2 ⋯ g N ) . (74)
When the switching function satisfies χ ( 0 ) = 0 \chi(0)=0 χ ( 0 ) = 0 , the system and bath are fully decoupled at initial time, and the initial state is a product state, whose covariance matrix reads
σ ( 0 ) = ( σ q S 0 0 0 0 d B 0 0 0 0 σ p S 0 0 0 0 d B ) . (75) \sigma(0)=
\begin{pmatrix}
\sigma_{q}^{S} & 0 & 0 & 0 \\
0 & d_{B} & 0 & 0 \\
0 & 0 & \sigma_{p}^{S} & 0 \\
0 & 0 & 0 & d_{B}
\end{pmatrix}. \tag*{(75)} σ ( 0 ) = σ q S 0 0 0 0 d B 0 0 0 0 σ p S 0 0 0 0 d B . (75)
where σ q S \sigma_{q}^{S} σ q S and σ p S \sigma_{p}^{S} σ p S denote the coordinate and momentum blocks, respectively, of the detector ground-state covariance matrix σ S ( 0 ) \sigma_{S}(0) σ S ( 0 ) .
The initial state of the detectors is taken as the ground state of H S H_{S} H S , whose covariance matrix is [96 ]
σ S ( 0 ) = 1 2 ν S ( ω − λ 0 0 − λ ω 0 0 0 0 ω λ 0 0 λ ω ) , (76) \sigma_{S}(0)=\frac{1}{2\nu_{S}}
\begin{pmatrix}
\omega& -\lambda& 0 & 0 \\
-\lambda& \omega& 0 & 0 \\
0 & 0 & \omega& \lambda\\
0 & 0 & \lambda& \omega
\end{pmatrix}, \tag*{(76)} σ S ( 0 ) = 2 ν S 1 ω − λ 0 0 − λ ω 0 0 0 0 ω λ 0 0 λ ω , (76)
where
ν S = ω 2 − λ 2 ( ∣ λ ∣ < ω ) . \nu_{S}=\sqrt{\omega^{2}-\lambda^{2}}\qquad(|\lambda|<\omega). ν S = ω 2 − λ 2 ( ∣ λ ∣ < ω ) .
The bath is initially prepared in a thermal state at the Unruh temperature
T U = a 2 π , (77) T_{U}=\frac{a}{2\pi}, \tag*{(77)} T U = 2 π a , (77)
and its covariance matrix takes the block-diagonal form
σ B ( 0 ) = ( d B 0 0 d B ) , d B = diag ( v 1 , v 1 , v 2 , v 2 , … , v N , v N ) , (78) \sigma_{B}(0)=
\begin{pmatrix}
d_{B} & 0 \\
0 & d_{B}
\end{pmatrix},
\qquad
d_{B}=\operatorname{diag}(v_{1},v_{1},v_{2},v_{2},\ldots,v_{N},v_{N}), \tag*{(78)} σ B ( 0 ) = ( d B 0 0 d B ) , d B = diag ( v 1 , v 1 , v 2 , v 2 , … , v N , v N ) , (78)
where
v j = 1 2 coth ( π Ω j a ) . v_{j}=\frac{1}{2}\coth\left(\frac{\pi\Omega_{j}}{a}\right). v j = 2 1 coth ( a π Ω j ) .
For a quadratic Hamiltonian, linearizing the Heisenberg equations of motion yields
R ˙ ( τ ) = A ( τ ) R ( τ ) , (79) \dot{\mathbf{R}}(\tau)=A(\tau)\mathbf{R}(\tau), \tag*{(79)} R ˙ ( τ ) = A ( τ ) R ( τ ) , (79)
where A ( τ ) = Ω c a n H ( τ ) A(\tau)=\Omega_{\mathrm{can}}H(\tau) A ( τ ) = Ω can H ( τ ) . We define the propagator M ( τ ) M(\tau) M ( τ ) via
M ˙ ( τ ) = A ( τ ) M ( τ ) , M ( 0 ) = I 2 n . (80) \dot{M}(\tau)=A(\tau)M(\tau),\qquad M(0)=I_{2n}. \tag*{(80)} M ˙ ( τ ) = A ( τ ) M ( τ ) , M ( 0 ) = I 2 n . (80)
It is straightforward to verify that the propagator is a symplectic matrix, satisfying
M Ω c a n M T = Ω c a n . (81) M\Omega_{\mathrm{can}}M^{T}=\Omega_{\mathrm{can}}. \tag*{(81)} M Ω can M T = Ω can . (81)
It follows that the phase-space vector evolves according to
R ( τ ) = M ( τ ) R ( 0 ) . (82) \mathbf{R}(\tau)=M(\tau)\mathbf{R}(0). \tag*{(82)} R ( τ ) = M ( τ ) R ( 0 ) . (82)
The evolution equation for the total covariance matrix then reads
σ ˙ ( τ ) = A ( τ ) σ ( τ ) + σ ( τ ) A ( τ ) T , (83) \dot{\sigma}(\tau)=A(\tau)\sigma(\tau)+\sigma(\tau)A(\tau)^{T}, \tag*{(83)} σ ˙ ( τ ) = A ( τ ) σ ( τ ) + σ ( τ ) A ( τ ) T , (83)
or equivalently
σ ( τ ) = M ( τ ) σ ( 0 ) M ( τ ) T . (84) \sigma(\tau)=M(\tau)\sigma(0)M(\tau)^{T}. \tag*{(84)} σ ( τ ) = M ( τ ) σ ( 0 ) M ( τ ) T . (84)
Since the detector coordinates correspond to the first two entries and the detector momenta correspond to entries n + 1 n+1 n + 1 and n + 2 n+2 n + 2 of the full phase-space vector, we define a 4 × 2 n 4 \times2n 4 × 2 n projection matrix P S P_{S} P S to extract the reduced covariance matrix of the system at arbitrary time τ \tau τ :
σ S ( τ ) = P S σ ( τ ) P S T . (85) \sigma_{S}(\tau)=P_{S}\sigma(\tau)P_{S}^{T}. \tag*{(85)} σ S ( τ ) = P S σ ( τ ) P S T . (85)
Gaussian ergotropy at infinite proper time
In this section, we derive the analytical expression for the Gaussian ergotropy in the limit of infinite proper time. For convenience, we let ζ = ± \zeta=\pm ζ = ± .
Gaussian ergotropy is the maximal amount of work that cyclic Gaussian Hamiltonian processes can extract from a Gaussian state. It is computed as follows [13 ] :
G = 1 2 Tr [ σ H ] − ∑ k = 1 n S s k ↑ l k ↓ , (86) \mathcal{G}=\frac{1}{2}\operatorname{Tr}[\sigma H]-\sum_{k=1}^{n_S}s_k^{\uparrow}l_k^{\downarrow}, \tag*{(86)} G = 2 1 Tr [ σ H ] − k = 1 ∑ n S s k ↑ l k ↓ , (86)
where n S = 2 n_S=2 n S = 2 is the number of detector modes, and s k s_k s k and l k l_k l k denote the symplectic eigenvalues of the covariance matrix and the quadratic Hamiltonian matrix, respectively, with the upward and downward superscript arrows indicating ascending and descending order.
Next, we investigate the Gaussian ergotropy with respect to the Hamiltonian H S \mathcal{H}_S H S . The minimization allows joint Gaussian unitaries on both detector modes, including collective-mode interchange; it is restricted to Gaussian operations.
First, we introduce the orthogonal transformation
q ζ = q A ± q B 2 , p ζ = p A ± p B 2 . (87) q_\zeta=\frac{q_A\pm q_B}{\sqrt{2}},\qquad p_\zeta=\frac{p_A\pm p_B}{\sqrt{2}}. \tag*{(87)} q ζ = 2 q A ± q B , p ζ = 2 p A ± p B . (87)
The canonical commutation relations are preserved:
[ q ζ , p ζ ] = i , [ q + , p − ] = 0. (88) [q_\zeta,p_\zeta]=i,\qquad[q_+,p_-]=0. \tag*{(88)} [ q ζ , p ζ ] = i , [ q + , p − ] = 0. (88)
Substituting Eq. (87) into Eq. (40) (H S \mathcal{H}_S H S ) yields
H S = ω 2 ( q + 2 + q − 2 + p + 2 + p − 2 ) + λ 2 [ ( q + 2 − q − 2 ) − ( p + 2 − p − 2 ) ] = ω + λ 2 q + 2 + ω − λ 2 p + 2 + ω − λ 2 q − 2 + ω + λ 2 p − 2 = H + + H − , (89) \begin{aligned}
\mathcal{H}_S=\frac{\omega}{2}\left(q_+^2+q_-^2+p_+^2+p_-^2\right)+\frac{\lambda}{2}\left[(q_+^2-q_-^2)-(p_+^2-p_-^2)\right] \\
&=\frac{\omega+\lambda}{2}q_+^2+\frac{\omega-\lambda}{2}p_+^2+\frac{\omega-\lambda}{2}q_-^2+\frac{\omega+\lambda}{2}p_-^2 \\
&=\mathcal{H}_++\mathcal{H}_-, \tag*{(89)}
\end{aligned} H S = 2 ω ( q + 2 + q − 2 + p + 2 + p − 2 ) + 2 λ [ ( q + 2 − q − 2 ) − ( p + 2 − p − 2 ) ] = 2 ω + λ q + 2 + 2 ω − λ p + 2 + 2 ω − λ q − 2 + 2 ω + λ p − 2 = H + + H − , (89)
where
H ζ = α ζ 2 q ζ 2 + β ζ 2 p ζ 2 , α ζ = ω ± λ , β ζ = ω ∓ λ . (90) \mathcal{H}_\zeta=\frac{\alpha_\zeta}{2}q_\zeta^2+\frac{\beta_\zeta}{2}p_\zeta^2,\qquad\alpha_\zeta=\omega\pm\lambda,\qquad\beta_\zeta=\omega\mp\lambda. \tag*{(90)} H ζ = 2 α ζ q ζ 2 + 2 β ζ p ζ 2 , α ζ = ω ± λ , β ζ = ω ∓ λ . (90)
We thus find that the two modes are exactly decoupled. From the Heisenberg equations of motion
q ˙ ζ = i [ H ζ , q ζ ] = β ζ p ζ , p ˙ ζ = i [ H ζ , p ζ ] , (91) \dot{q}_\zeta=i[\mathcal{H}_\zeta,q_\zeta]=\beta_\zeta p_\zeta,\qquad\dot{p}_\zeta=i[\mathcal{H}_\zeta,p_\zeta], \tag*{(91)} q ˙ ζ = i [ H ζ , q ζ ] = β ζ p ζ , p ˙ ζ = i [ H ζ , p ζ ] , (91)
we find that the two modes share the same normal-mode frequency:
ν S = ω 2 − λ 2 . (92) \nu_S=\sqrt{\omega^2-\lambda^2}. \tag*{(92)} ν S = ω 2 − λ 2 . (92)
Diagonalizing the spectral density matrix (51) yields:
J E ζ ( Ω ) = J 0 ( Ω ) [ 1 ± γ E ( Ω ) ] . (93) J_{E\zeta}(\Omega)=J_0(\Omega)[1\pm\gamma_E(\Omega)]. \tag*{(93)} J E ζ ( Ω ) = J 0 ( Ω ) [ 1 ± γ E ( Ω )] . (93)
Under the new transformation, Eq. (62) becomes:
H C T = 1 2 [ ( δ ω A + δ ω A B ) q + 2 + ( δ ω A − δ ω A B ) q − 2 ] = 1 2 ( δ ω + q + 2 + δ ω − q − 2 ) , (94) \begin{aligned}
\mathcal{H}_{\mathrm{CT}}=\frac{1}{2}\left[(\delta\omega_A+\delta\omega_{AB})q_+^2+(\delta\omega_A-\delta\omega_{AB})q_-^2\right] \\
&=\frac{1}{2}\left(\delta\omega_+q_+^2+\delta\omega_-q_-^2\right), \tag*{(94)}
\end{aligned} H CT = 2 1 [ ( δ ω A + δ ω A B ) q + 2 + ( δ ω A − δ ω A B ) q − 2 ] = 2 1 ( δ ω + q + 2 + δ ω − q − 2 ) , (94)
where
δ ω ζ = δ ω A ± δ ω A B . (95) \delta\omega_\zeta=\delta\omega_A\pm\delta\omega_{AB}. \tag*{(95)} δ ω ζ = δ ω A ± δ ω A B . (95)
Now taking the limits N → ∞ N\to\infty N → ∞ , we arrive at the closed-form expression for δ ω ζ \delta\omega_\zeta δ ω ζ in the continuous limit.
δ ω ζ = κ 2 2 π 2 ∫ 0 ∞ d Ω e − Ω / Ω c [ 1 ± γ U ( Ω ) ] = κ 2 2 π 2 [ Ω c ± 1 d 1 + ( a d / 2 ) 2 arctan ( 2 Ω c a arsinh a d 2 ) ] . (96) \begin{aligned}
\delta\omega_\zeta=\frac{\kappa^2}{2\pi^2}\int_0^\infty d\Omega\,e^{-\Omega/\Omega_c}[1\pm\gamma_U(\Omega)] \\
&=\frac{\kappa^2}{2\pi^2}\left[\Omega_c\pm\frac{1}{d\sqrt{1+(ad/2)^2}}\arctan\left(\frac{2\Omega_c}{a}\operatorname{arsinh}\frac{ad}{2}\right)\right]. \tag*{(96)}
\end{aligned} δ ω ζ = 2 π 2 κ 2 ∫ 0 ∞ d Ω e − Ω/ Ω c [ 1 ± γ U ( Ω )] = 2 π 2 κ 2 [ Ω c ± d 1 + ( a d /2 ) 2 1 arctan ( a 2 Ω c arsinh 2 a d ) ] . (96)
The total Hamiltonian can then be decomposed into two sectors, each of which corresponds to a fully quadratic model consisting of one harmonic oscillator coupled to an independent oscillator bath.
H t o t = ∑ ζ = ± H ζ t o t , \mathcal{H}_{\mathrm{tot}}=\sum_{\zeta=\pm}\mathcal{H}^{\mathrm{tot}}_\zeta, H tot = ζ = ± ∑ H ζ tot ,
H ζ t o t = α ζ + δ ω ζ 2 q ζ 2 + β ζ 2 p ζ 2 + ∑ j Ω j 2 ( Q j , ζ 2 + P j , ζ 2 ) + ∑ j g j , ζ q ζ Q j , ζ . (97) \mathcal{H}^{\mathrm{tot}}_\zeta=\frac{\alpha_\zeta+\delta\omega_\zeta}{2}q_\zeta^2+\frac{\beta_\zeta}{2}p_\zeta^2+\sum_j\frac{\Omega_j}{2}(Q_{j,\zeta}^2+P_{j,\zeta}^2)+\sum_j g_{j,\zeta}q_\zeta Q_{j,\zeta}. \tag*{(97)} H ζ tot = 2 α ζ + δ ω ζ q ζ 2 + 2 β ζ p ζ 2 + j ∑ 2 Ω j ( Q j , ζ 2 + P j , ζ 2 ) + j ∑ g j , ζ q ζ Q j , ζ . (97)
Next, we carry out a scaling transformation for the system Hamiltonian H S \mathcal{H}_S H S (89) :
q ζ ′ = ( α ζ β ζ ) 1 / 4 q ζ , p ζ ′ = ( β ζ α ζ ) 1 / 4 p ζ . (98) q'_\zeta=\left(\frac{\alpha_\zeta}{\beta_\zeta}\right)^{1/4}q_\zeta,\qquad p'_\zeta=\left(\frac{\beta_\zeta}{\alpha_\zeta}\right)^{1/4}p_\zeta. \tag*{(98)} q ζ ′ = ( β ζ α ζ ) 1/4 q ζ , p ζ ′ = ( α ζ β ζ ) 1/4 p ζ . (98)
so
H S = ν S 2 ∑ ζ = ± 1 ( q ζ ′ 2 + p ζ ′ 2 ) . (99) \mathcal{H}_S=\frac{\nu_S}{2}\sum_{\zeta=\pm1}\left(q_\zeta^{\prime2}+p_\zeta^{\prime2}\right). \tag*{(99)} H S = 2 ν S ζ = ± 1 ∑ ( q ζ ′2 + p ζ ′2 ) . (99)
It is then straightforward to derive:
⟨ q ζ ′ 2 ⟩ 0 = β ζ 2 ν S , ⟨ p ζ ′ 2 ⟩ 0 = α ζ 2 ν S , \langle q_\zeta^{\prime2}\rangle_0=\frac{\beta_\zeta}{2\nu_S},\qquad\langle p_\zeta^{\prime2}\rangle_0=\frac{\alpha_\zeta}{2\nu_S}, ⟨ q ζ ′2 ⟩ 0 = 2 ν S β ζ , ⟨ p ζ ′2 ⟩ 0 = 2 ν S α ζ ,
⟨ q + q − ⟩ 0 = ⟨ p + p − ⟩ 0 = 0 , \langle q_+q_-\rangle_0=\langle p_+p_-\rangle_0=0, ⟨ q + q − ⟩ 0 = ⟨ p + p − ⟩ 0 = 0 ,
1 2 ⟨ q ζ p ζ + p ζ q ζ ⟩ 0 = 0. (100) \frac{1}{2}\langle q_\zeta p_\zeta+p_\zeta q_\zeta\rangle_0=0. \tag*{(100)} 2 1 ⟨ q ζ p ζ + p ζ q ζ ⟩ 0 = 0. (100)
The canonical commutator gives ⟨ q ζ p ζ ⟩ 0 = i / 2 \langle q_\zeta p_\zeta\rangle_0=i/2 ⟨ q ζ p ζ ⟩ 0 = i /2 ; the covariance uses the symmetrized moment displayed above. The covariance matrix for each sector is given by:
σ ζ = ( ⟨ q ζ ′ 2 ⟩ 0 0 0 ⟨ p ζ ′ 2 ⟩ 0 ) . (101) \sigma_\zeta=
\begin{pmatrix}
\langle q_\zeta^{\prime2}\rangle_0 & 0\\
0 & \langle p_\zeta^{\prime2}\rangle_0
\end{pmatrix}. \tag*{(101)} σ ζ = ( ⟨ q ζ ′2 ⟩ 0 0 0 ⟨ p ζ ′2 ⟩ 0 ) . (101)
We obtain the stationary work from the reduced equilibrium state of the interacting detector–bath Hamiltonian. Local relaxation requires the continuum-reservoir limit before the long-time limit: a finite closed oscillator bath instead has quasiperiodic dynamics.
More precisely, the reduced equilibrium state is formally defined as
ρ S e q = Tr B [ e − H t o t / T U ] Z t o t , (102) \rho^{\mathrm{eq}}_{S}=\frac{\operatorname{Tr}_{B}\left[e^{-H_{\mathrm{tot}}/T_{U}}\right]}{Z_{\mathrm{tot}},
} \tag*{(102)} ρ S eq = Z tot , Tr B [ e − H tot / T U ] (102)
where the total partition function is
Z t o t = Tr S ⊗ B [ e − H t o t / T U ] . (103) Z_{\mathrm{tot}}=\operatorname{Tr}_{S\otimes B}\left[e^{-H_{\mathrm{tot}}/T_{U}}\right]. \tag*{(103)} Z tot = Tr S ⊗ B [ e − H tot / T U ] . (103)
Here, Tr B \operatorname{Tr}_{B} Tr B denotes the partial trace over all bath modes, whereas Tr S ⊗ B \operatorname{Tr}_{S\otimes B} Tr S ⊗ B denotes the full trace over the Hilbert space of the combined detector system and bath. The above continuum expressions are understood as the limits of the corresponding finite-bath equilibrium quantities.
We now calculate the equilibrium covariance matrix by employing the Euclidean action. Setting the interaction switching function to its final value, χ = 1 \chi=1 χ = 1 , the Euclidean action for a single collective sector is given by [80 ]
S E ( ζ ) = ∫ 0 1 / T U d u [ Q ˙ ζ 2 2 β ζ + α ζ 2 q ζ 2 + δ ω ζ 2 q ζ 2 + ∑ j ( Q ˙ j , ζ 2 2 Ω j + Ω j 2 Q j , ζ 2 + g j , ζ q ζ Q j , ζ ) ] , (104) S^{(\zeta)}_{E}=\int_{0}^{1/T_{U}}\mathrm{d}u\left[\frac{\dot{Q}_{\zeta}^{2}}{2\beta_{\zeta}}+\frac{\alpha_{\zeta}}{2}q_{\zeta}^{2}+\frac{\delta\omega_{\zeta}}{2}q_{\zeta}^{2}+\sum_{j}\left(\frac{\dot{Q}_{j,\zeta}^{2}}{2\Omega_{j}}+\frac{\Omega_{j}}{2}Q_{j,\zeta}^{2}+g_{j,\zeta}q_{\zeta}Q_{j,\zeta}\right)\right], \tag*{(104)} S E ( ζ ) = ∫ 0 1/ T U d u [ 2 β ζ Q ˙ ζ 2 + 2 α ζ q ζ 2 + 2 δ ω ζ q ζ 2 + j ∑ ( 2 Ω j Q ˙ j , ζ 2 + 2 Ω j Q j , ζ 2 + g j , ζ q ζ Q j , ζ ) ] , (104)
where u u u denotes imaginary time, and the dot represents differentiation with respect to u u u . The static counterterm is determined by
δ ω ζ = ∑ j g j , ζ 2 Ω j . (105) \delta\omega_{\zeta}=\sum_{j}\frac{g_{j,\zeta}^{2}}{\Omega_{j}}. \tag*{(105)} δ ω ζ = j ∑ Ω j g j , ζ 2 . (105)
The Gaussian integration over the bath coordinates Q j , ζ Q_{j,\zeta} Q j , ζ is determined by the Euclidean equation
( − ∂ u 2 + Ω j 2 ) Q j , ζ = − g j , ζ Ω j q ζ , (106) \left(-\partial_{u}^{2}+\Omega_{j}^{2}\right)Q_{j,\zeta}=-g_{j,\zeta}\Omega_{j}q_{\zeta}, \tag*{(106)} ( − ∂ u 2 + Ω j 2 ) Q j , ζ = − g j , ζ Ω j q ζ , (106)
whose solution in the Matsubara frequency domain reads
Q j , ζ ( i ν n ) = − g j , ζ Ω j ν n 2 + Ω j 2 q ζ ( i ν n ) , (107) Q_{j,\zeta}(i\nu_{n})=-\frac{g_{j,\zeta}\Omega_{j}}{\nu_{n}^{2}+\Omega_{j}^{2}}q_{\zeta}(i\nu_{n}), \tag*{(107)} Q j , ζ ( i ν n ) = − ν n 2 + Ω j 2 g j , ζ Ω j q ζ ( i ν n ) , (107)
where the bosonic Matsubara frequencies are defined as ν n = 2 π n T U \nu_{n}=2\pi nT_{U} ν n = 2 π n T U , with n ∈ Z n\in\mathbb{Z} n ∈ Z . Substituting this result back into Eq. (104) yields the exact inverse Green function:
G ζ − 1 ( i ν n ) = D 0 , ζ ( ν n ) + Σ ζ ( ν n ) , (108) G_{\zeta}^{-1}(i\nu_{n})=D_{0,\zeta}(\nu_{n})+\Sigma_{\zeta}(\nu_{n}), \tag*{(108)} G ζ − 1 ( i ν n ) = D 0 , ζ ( ν n ) + Σ ζ ( ν n ) , (108)
where
D 0 , ζ ( ν n ) = α ζ + ν n 2 β ζ , Σ ζ ( ν n ) = ν n 2 Λ ζ ( ν n ) , Λ ζ ( ν n ) ≡ ∑ j g j , ζ 2 Ω j ( ν n 2 + Ω j 2 ) . (109) \begin{aligned}
D_{0,\zeta}(\nu_{n})&=\alpha_{\zeta}+\frac{\nu_{n}^{2}}{\beta_{\zeta}},\\
\Sigma_{\zeta}(\nu_{n})&=\nu_{n}^{2}\Lambda_{\zeta}(\nu_{n}),\\
\Lambda_{\zeta}(\nu_{n})&\equiv\sum_{j}\frac{g_{j,\zeta}^{2}}{\Omega_{j}(\nu_{n}^{2}+\Omega_{j}^{2})}. \tag*{(109)}
\end{aligned} D 0 , ζ ( ν n ) Σ ζ ( ν n ) Λ ζ ( ν n ) = α ζ + β ζ ν n 2 , = ν n 2 Λ ζ ( ν n ) , ≡ j ∑ Ω j ( ν n 2 + Ω j 2 ) g j , ζ 2 . (109)
The static counterterm compensates the zero-frequency bath-induced displacement. Consequently, the self-energy vanishes at zero frequency, while its frequency-dependent part is retained in Σ ζ \Sigma_{\zeta} Σ ζ .
Starting from Eq. (108) , we carry out a geometric series expansion. Since Σ ζ = O ( κ 2 ) \Sigma_{\zeta}=O(\kappa^{2}) Σ ζ = O ( κ 2 ) , we obtain
G ζ ( i ν n ) = G 0 , ζ ( i ν n ) − ν n 2 Λ ζ ( ν n ) G 0 , ζ 2 ( i ν n ) + O ( κ 4 ) , (110) G_{\zeta}(i\nu_{n})=G_{0,\zeta}(i\nu_{n})-\nu_{n}^{2}\Lambda_{\zeta}(\nu_{n})G_{0,\zeta}^{2}(i\nu_{n})+O(\kappa^{4}), \tag*{(110)} G ζ ( i ν n ) = G 0 , ζ ( i ν n ) − ν n 2 Λ ζ ( ν n ) G 0 , ζ 2 ( i ν n ) + O ( κ 4 ) , (110)
where G 0 , ζ ≡ D 0 , ζ − 1 G_{0,\zeta}\equiv D_{0,\zeta}^{-1} G 0 , ζ ≡ D 0 , ζ − 1 .
The steady-state coordinate variance can be directly obtained from the Green function:
Q ζ e q = T U ∑ n ∈ Z G ζ ( i ν n ) . (111) Q_{\zeta}^{\mathrm{eq}}=T_{U}\sum_{n\in\mathbb{Z}}G_{\zeta}(i\nu_{n}). \tag*{(111)} Q ζ eq = T U n ∈ Z ∑ G ζ ( i ν n ) . (111)
To obtain the momentum covariance, we employ the stationary equation for the symmetrized coordinate-momentum correlator. In the steady state, its time derivative vanishes in the long-time limit:
d d τ ⟨ { q ζ , p ζ } 2 ⟩ e q = 0. (112) \frac{\mathrm{d}}{\mathrm{d}\tau}\left\langle\frac{\{q_{\zeta},p_{\zeta}\}}{2}\right\rangle^{\mathrm{eq}}=0. \tag*{(112)} d τ d ⟨ 2 { q ζ , p ζ } ⟩ eq = 0. (112)
Substituting this relation into the Heisenberg equations of motion,
d q ζ d τ = i [ H t o t , q ζ ] = β ζ p ζ , \frac{\mathrm{d}q_{\zeta}}{\mathrm{d}\tau}=i[H_{\mathrm{tot}},q_{\zeta}]=\beta_{\zeta}p_{\zeta}, d τ d q ζ = i [ H tot , q ζ ] = β ζ p ζ ,
d p ζ d τ = i [ H t o t , p ζ ] = − ( α ζ + δ ω ζ ) q ζ − ∑ j g j , ζ Q j , ζ , (113) \frac{\mathrm{d}p_{\zeta}}{\mathrm{d}\tau}=i[H_{\mathrm{tot}},p_{\zeta}]
=-(\alpha_{\zeta}+\delta\omega_{\zeta})q_{\zeta}-\sum_{j}g_{j,\zeta}Q_{j,\zeta}, \tag*{(113)} d τ d p ζ = i [ H tot , p ζ ] = − ( α ζ + δ ω ζ ) q ζ − j ∑ g j , ζ Q j , ζ , (113)
we immediately obtain the virial identity for the ζ \zeta ζ sector:
β ζ P ζ e q = ( α ζ + δ ω ζ ) Q ζ e q + ∑ j g j , ζ ⟨ q ζ Q j , ζ ⟩ e q . (114) \beta_{\zeta}P_{\zeta}^{\mathrm{eq}}=(\alpha_{\zeta}+\delta\omega_{\zeta})Q_{\zeta}^{\mathrm{eq}}+\sum_{j}g_{j,\zeta}\langle q_{\zeta}Q_{j,\zeta}\rangle^{\mathrm{eq}}. \tag*{(114)} β ζ P ζ eq = ( α ζ + δ ω ζ ) Q ζ eq + j ∑ g j , ζ ⟨ q ζ Q j , ζ ⟩ eq . (114)
The system-bath cross-correlation can be evaluated using Eq. (107) :
⟨ q ζ Q j , ζ ⟩ e q = − T U ∑ n g j , ζ Ω j ν n 2 + Ω j 2 G ζ ( i ν n ) . (115) \langle q_{\zeta}Q_{j,\zeta}\rangle^{\mathrm{eq}}=-T_{U}\sum_{n}\frac{g_{j,\zeta}\Omega_{j}}{\nu_{n}^{2}+\Omega_{j}^{2}}G_{\zeta}(i\nu_{n}). \tag*{(115)} ⟨ q ζ Q j , ζ ⟩ eq = − T U n ∑ ν n 2 + Ω j 2 g j , ζ Ω j G ζ ( i ν n ) . (115)
Using the counterterm relation and the definition of Λ ζ \Lambda_{\zeta} Λ ζ , we find
∑ j g j , ζ ⟨ q ζ Q j , ζ ⟩ e q = − δ ω ζ Q ζ e q + T U ∑ n ν n 2 Λ ζ ( ν n ) G ζ ( i ν n ) . (116) \sum_{j}g_{j,\zeta}\langle q_{\zeta}Q_{j,\zeta}\rangle^{\mathrm{eq}}
=-\delta\omega_{\zeta}Q_{\zeta}^{\mathrm{eq}}+T_{U}\sum_{n}\nu_{n}^{2}\Lambda_{\zeta}(\nu_{n})G_{\zeta}(i\nu_{n}). \tag*{(116)} j ∑ g j , ζ ⟨ q ζ Q j , ζ ⟩ eq = − δ ω ζ Q ζ eq + T U n ∑ ν n 2 Λ ζ ( ν n ) G ζ ( i ν n ) . (116)
Combining this result with the virial identity Eq. (114) , we obtain
P ζ e q = T U β ζ ∑ n [ α ζ + Σ ζ ( ν n ) ] G ζ ( i ν n ) . (117) P_{\zeta}^{\mathrm{eq}}=\frac{T_{U}}{\beta_{\zeta}}\sum_{n}\left[\alpha_{\zeta}+\Sigma_{\zeta}(\nu_{n})\right]G_{\zeta}(i\nu_{n}). \tag*{(117)} P ζ eq = β ζ T U n ∑ [ α ζ + Σ ζ ( ν n ) ] G ζ ( i ν n ) . (117)
Moreover, the equilibrium state is invariant under time reversal, so that the symmetrized coordinate-momentum correlator vanishes:
1 2 ⟨ { q ζ , p ζ } ⟩ e q = 0. (118) \frac{1}{2}\langle\{q_{\zeta},p_{\zeta}\}\rangle^{\mathrm{eq}}=0. \tag*{(118)} 2 1 ⟨{ q ζ , p ζ } ⟩ eq = 0. (118)
We next introduce the uncoupled thermal reference covariances. Using
ν S = α ζ β ζ = ω 2 − λ 2 , \nu_{S}=\sqrt{\alpha_{\zeta}\beta_{\zeta}}=\sqrt{\omega^{2}-\lambda^{2}}, ν S = α ζ β ζ = ω 2 − λ 2 ,
and the Matsubara summation identity [98 ] ,
T U ∑ n G 0 , ζ ( i ν n ) = β ζ 2 ν S coth ( ν S 2 T U ) , (119) T_{U}\sum_{n}G_{0,\zeta}(i\nu_{n})=\frac{\beta_{\zeta}}{2\nu_{S}}\coth\left(\frac{\nu_{S}}{2T_{U}}\right), \tag*{(119)} T U n ∑ G 0 , ζ ( i ν n ) = 2 ν S β ζ coth ( 2 T U ν S ) , (119)
we define
Q 0 , ζ = β ζ c U 2 ν S , P 0 , ζ = α ζ c U 2 ν S , c U = coth ( ν S 2 T U ) . (120) Q_{0,\zeta}=\frac{\beta_{\zeta}c_{U}}{2\nu_{S}},\qquad P_{0,\zeta}=\frac{\alpha_{\zeta}c_{U}}{2\nu_{S}},\qquad c_{U}=\coth\left(\frac{\nu_{S}}{2T_{U}}\right). \tag*{(120)} Q 0 , ζ = 2 ν S β ζ c U , P 0 , ζ = 2 ν S α ζ c U , c U = coth ( 2 T U ν S ) . (120)
These are the uncoupled thermal reference covariances used in the equilibrium expansion, rather than the initial ground-state covariances of the dynamical evolution.
Writing
δ Q ζ ≡ Q ζ e q − Q 0 , ζ , δ P ζ ≡ P ζ e q − P 0 , ζ , \delta Q_{\zeta}\equiv Q_{\zeta}^{\mathrm{eq}}-Q_{0,\zeta},\qquad\delta P_{\zeta}\equiv P_{\zeta}^{\mathrm{eq}}-P_{0,\zeta}, δ Q ζ ≡ Q ζ eq − Q 0 , ζ , δ P ζ ≡ P ζ eq − P 0 , ζ ,
and combining Eqs. (110) and (119) , we obtain the coordinate variance correction:
δ Q ζ = − T U ∑ n ν n 2 Λ ζ ( ν n ) D 0 , ζ ( ν n ) 2 + O ( κ 4 ) . (121) \delta Q_{\zeta}=-T_{U}\sum_{n}\frac{\nu_{n}^{2}\Lambda_{\zeta}(\nu_{n})}{D_{0,\zeta}(\nu_{n})^{2}}+O(\kappa^{4}). \tag*{(121)} δ Q ζ = − T U n ∑ D 0 , ζ ( ν n ) 2 ν n 2 Λ ζ ( ν n ) + O ( κ 4 ) . (121)
Similarly, expanding Eq. (117) to the same order gives the momentum variance correction:
δ P ζ = α ζ β ζ δ Q ζ + T U β ζ ∑ n ν n 2 Λ ζ ( ν n ) D 0 , ζ ( ν n ) + O ( κ 4 ) . (122) \delta P_{\zeta}=\frac{\alpha_{\zeta}}{\beta_{\zeta}}\delta Q_{\zeta}+\frac{T_{U}}{\beta_{\zeta}}\sum_{n}\frac{\nu_{n}^{2}\Lambda_{\zeta}(\nu_{n})}{D_{0,\zeta}(\nu_{n})}+O(\kappa^{4}). \tag*{(122)} δ P ζ = β ζ α ζ δ Q ζ + β ζ T U n ∑ D 0 , ζ ( ν n ) ν n 2 Λ ζ ( ν n ) + O ( κ 4 ) . (122)
We now turn our attention to the Matsubara sums
T U ∑ n ν n 2 Λ ζ ( ν n ) D 0 , ζ ( ν n ) k , k = 1 , 2. (123) T_{U}\sum_{n}\frac{\nu_{n}^{2}\Lambda_{\zeta}(\nu_{n})}{D_{0,\zeta}(\nu_{n})^{k}},\qquad k=1,2. \tag*{(123)} T U n ∑ D 0 , ζ ( ν n ) k ν n 2 Λ ζ ( ν n ) , k = 1 , 2. (123)
Interchanging the order of the bath summation and the Matsubara summation, we obtain
T U ∑ n ν n 2 Λ ζ ( ν n ) D 0 , ζ ( ν n ) k = ∑ j g j , ζ 2 Ω j 2 β ζ k B k ( Ω j ) , (124) T_{U}\sum_{n}\frac{\nu_{n}^{2}\Lambda_{\zeta}(\nu_{n})}{D_{0,\zeta}(\nu_{n})^{k}}=\sum_{j}\frac{g_{j,\zeta}^{2}}{\Omega_{j}^{2}}\beta_{\zeta}^{k}B_{k}(\Omega_{j}), \tag*{(124)} T U n ∑ D 0 , ζ ( ν n ) k ν n 2 Λ ζ ( ν n ) = j ∑ Ω j 2 g j , ζ 2 β ζ k B k ( Ω j ) , (124)
where
B k ( Ω ) ≡ T U ∑ n ∈ Z ν n 2 ( ν n 2 + Ω 2 ) ( ν n 2 + ν S 2 ) k , k = 1 , 2. (125) B_{k}(\Omega)\equiv T_{U}\sum_{n\in\mathbb{Z}}\frac{\nu_{n}^{2}}{(\nu_{n}^{2}+\Omega^{2})(\nu_{n}^{2}+\nu_{S}^{2})^{k}},\qquad k=1,2. \tag*{(125)} B k ( Ω ) ≡ T U n ∈ Z ∑ ( ν n 2 + Ω 2 ) ( ν n 2 + ν S 2 ) k ν n 2 , k = 1 , 2. (125)
Evaluating the Matsubara summation explicitly, we obtain
B 1 ( Ω ) = Ω coth ( Ω 2 T U ) − ν S c U 2 ( Ω 2 − ν S 2 ) . (126) B_{1}(\Omega)=\frac{\Omega\coth\left(\frac{\Omega}{2T_{U}}\right)-\nu_{S}c_{U}}{2(\Omega^{2}-\nu_{S}^{2})}. \tag*{(126)} B 1 ( Ω ) = 2 ( Ω 2 − ν S 2 ) Ω coth ( 2 T U Ω ) − ν S c U . (126)
The second function can be expressed more compactly as
B 2 ( Ω ) = − 1 2 ν S ∂ B 1 ( Ω ) ∂ ν S , (127) B_{2}(\Omega)=-\frac{1}{2\nu_{S}}\frac{\partial B_{1}(\Omega)}{\partial\nu_{S}}, \tag*{(127)} B 2 ( Ω ) = − 2 ν S 1 ∂ ν S ∂ B 1 ( Ω ) , (127)
where the derivative is taken at fixed T U T_{U} T U and Ω \Omega Ω . The apparent singularity at Ω = ν S \Omega=\nu_{S} Ω = ν S is removable. In particular,
B 1 ( ν S ) = c U − ν S 2 T U csch 2 ( ν S 2 T U ) 4 ν S . (128) B_{1}(\nu_{S})=\frac{c_{U}-\frac{\nu_{S}}{2T_{U}}\operatorname{csch}^{2}\left(\frac{\nu_{S}}{2T_{U}}\right)}{4\nu_{S}}. \tag*{(128)} B 1 ( ν S ) = 4 ν S c U − 2 T U ν S csch 2 ( 2 T U ν S ) . (128)
At zero temperature, these functions reduce to
B 1 ( Ω ) = 1 2 ( Ω + ν S ) , B 2 ( Ω ) = 1 4 ν S ( Ω + ν S ) 2 . (129) B_{1}(\Omega)=\frac{1}{2(\Omega+\nu_{S})},\qquad B_{2}(\Omega)=\frac{1}{4\nu_{S}(\Omega+\nu_{S})^{2}}. \tag*{(129)} B 1 ( Ω ) = 2 ( Ω + ν S ) 1 , B 2 ( Ω ) = 4 ν S ( Ω + ν S ) 2 1 . (129)
We now take the continuum-reservoir limit. For the three environments E ∈ { U , T , C } E\in\{U,T,C\} E ∈ { U , T , C } , the spectral-matching relation gives
J E , ζ ( Ω ) = κ 2 Ω e − Ω / Ω c 4 π [ 1 ± γ E ( Ω ) ] , (130) J_{E,\zeta}(\Omega)=\frac{\kappa^{2}\Omega e^{-\Omega/\Omega_{c}}}{4\pi}[1\pm\gamma_{E}(\Omega)], \tag*{(130)} J E , ζ ( Ω ) = 4 π κ 2 Ω e − Ω/ Ω c [ 1 ± γ E ( Ω )] , (130)
where ζ = + 1 \zeta=+1 ζ = + 1 and ζ = − 1 \zeta=-1 ζ = − 1 denote the symmetric and antisymmetric channels, respectively. The discrete couplings satisfy
g j , ζ 2 = 2 ω j π J E , ζ ( Ω j ) . (131) g_{j,\zeta}^{2}=\frac{2\omega_{j}}{\pi}J_{E,\zeta}(\Omega_{j}). \tag*{(131)} g j , ζ 2 = π 2 ω j J E , ζ ( Ω j ) . (131)
Consequently, for a sufficiently dense frequency grid,
∑ j g j , ζ 2 f ( Ω j ) ⟶ κ 2 2 π 2 ∫ 0 ∞ d Ω Ω e − Ω / Ω c [ 1 ± γ E ( Ω ) ] f ( Ω ) . (132) \sum_{j}g_{j,\zeta}^{2}f(\Omega_{j})
\longrightarrow\frac{\kappa^{2}}{2\pi^{2}}\int_{0}^{\infty}\mathrm{d}\Omega\,\Omega e^{-\Omega/\Omega_{c}}[1\pm\gamma_{E}(\Omega)]f(\Omega). \tag*{(132)} j ∑ g j , ζ 2 f ( Ω j ) ⟶ 2 π 2 κ 2 ∫ 0 ∞ d Ω Ω e − Ω/ Ω c [ 1 ± γ E ( Ω )] f ( Ω ) . (132)
Substituting Eq. (132) into Eqs. (121) and (122) , we obtain
δ Q ζ = − κ 2 β ζ 2 2 π 2 ∫ 0 ∞ d Ω e − Ω / Ω c [ 1 ± γ E ( Ω ) ] B 2 ( Ω ) + O ( κ 4 ) , (133) \delta Q_{\zeta}=-\frac{\kappa^{2}\beta_{\zeta}^{2}}{2\pi^{2}}\int_{0}^{\infty}\mathrm{d}\Omega\,e^{-\Omega/\Omega_{c}}[1\pm\gamma_{E}(\Omega)]B_{2}(\Omega)+O(\kappa^{4}), \tag*{(133)} δ Q ζ = − 2 π 2 κ 2 β ζ 2 ∫ 0 ∞ d Ω e − Ω/ Ω c [ 1 ± γ E ( Ω )] B 2 ( Ω ) + O ( κ 4 ) , (133)
δ P ζ = α ζ β ζ δ Q ζ + κ 2 2 π 2 ∫ 0 ∞ d Ω e − Ω / Ω c [ 1 ± γ E ( Ω ) ] B 1 ( Ω ) + O ( κ 4 ) . (134) \delta P_{\zeta}=\frac{\alpha_{\zeta}}{\beta_{\zeta}}\delta Q_{\zeta}+\frac{\kappa^{2}}{2\pi^{2}}\int_{0}^{\infty}\mathrm{d}\Omega\,e^{-\Omega/\Omega_{c}}[1\pm\gamma_{E}(\Omega)]B_{1}(\Omega)+O(\kappa^{4}). \tag*{(134)} δ P ζ = β ζ α ζ δ Q ζ + 2 π 2 κ 2 ∫ 0 ∞ d Ω e − Ω/ Ω c [ 1 ± γ E ( Ω )] B 1 ( Ω ) + O ( κ 4 ) . (134)
Since each collective sector is in a single-mode Gaussian equilibrium state, its Gaussian ergotropy relative to the corresponding sector Hamiltonian H ζ \mathcal{H}_{\zeta} H ζ is
G E , ζ e q = α ζ Q ζ e q + β ζ P ζ e q 2 − ν S Q ζ e q P ζ e q − ( 1 2 ⟨ { q ζ , p ζ } ⟩ e q ) 2 . (135) \mathcal{G}_{E,\zeta}^{\mathrm{eq}}=\frac{\alpha_{\zeta}Q_{\zeta}^{\mathrm{eq}}+\beta_{\zeta}P_{\zeta}^{\mathrm{eq}}}{2}-\nu_{S}\sqrt{Q_{\zeta}^{\mathrm{eq}}P_{\zeta}^{\mathrm{eq}}-\left(\frac{1}{2}\langle\{q_{\zeta},p_{\zeta}\}\rangle^{\mathrm{eq}}\right)^{2}}. \tag*{(135)} G E , ζ eq = 2 α ζ Q ζ eq + β ζ P ζ eq − ν S Q ζ eq P ζ eq − ( 2 1 ⟨{ q ζ , p ζ } ⟩ eq ) 2 . (135)
The total Gaussian ergotropy is obtained by summing over the two collective sectors:
G E e q = ∑ ζ = ± 1 G E , ζ e q . (136) \mathcal{G}_{E}^{\mathrm{eq}}=\sum_{\zeta=\pm1}\mathcal{G}_{E,\zeta}^{\mathrm{eq}}. \tag*{(136)} G E eq = ζ = ± 1 ∑ G E , ζ eq . (136)
Here the extraction is defined with respect to the full detector Hamiltonian H S \mathcal{H}_{S} H S and allows collective Gaussian unitaries on the two detectors.
To obtain the leading weak-coupling behavior, we introduce the normalized covariance corrections
x ζ ≡ δ Q ζ Q 0 , ζ , y ζ ≡ δ P ζ P 0 , ζ . (137) x_{\zeta} \equiv\frac{\delta Q_{\zeta}}{Q_{0,\zeta}}, \qquad y_{\zeta} \equiv\frac{\delta P_{\zeta}}{P_{0,\zeta}}. \tag*{(137)} x ζ ≡ Q 0 , ζ δ Q ζ , y ζ ≡ P 0 , ζ δ P ζ . (137)
Since 1 2 ⟨ { q ζ , p ζ } ⟩ e q = 0 \frac{1}{2}\langle\{q_{\zeta},p_{\zeta}\}\rangle^{\mathrm{eq}}=0 2 1 ⟨{ q ζ , p ζ } ⟩ eq = 0 , the ergotropy of each sector can be rewritten as
G E , ζ e q = ν S c U 4 [ 2 + x ζ + y ζ − 2 ( 1 + x ζ ) ( 1 + y ζ ) ] . (138) \mathcal{G}^{\mathrm{eq}}_{E,\zeta}=\frac{\nu_{S}c_{U}}{4}\left[2+x_{\zeta}+y_{\zeta}-2\sqrt{(1+x_{\zeta})(1+y_{\zeta})}\right]. \tag*{(138)} G E , ζ eq = 4 ν S c U [ 2 + x ζ + y ζ − 2 ( 1 + x ζ ) ( 1 + y ζ ) ] . (138)
Expanding the square root, the terms linear in x ζ x_{\zeta} x ζ and y ζ y_{\zeta} y ζ cancel. Therefore,
G E e q = ν S c U 16 ∑ ζ = ± 1 ( x ζ − y ζ ) 2 + O ( κ 6 ) . (139) \mathcal{G}^{\mathrm{eq}}_{E}=\frac{\nu_{S}c_{U}}{16}\sum_{\zeta=\pm1}(x_{\zeta}-y_{\zeta})^{2}+O(\kappa^{6}). \tag*{(139)} G E eq = 16 ν S c U ζ = ± 1 ∑ ( x ζ − y ζ ) 2 + O ( κ 6 ) . (139)
Moreover, since
P 0 , ζ Q 0 , ζ = α ζ β ζ , \frac{P_{0,\zeta}}{Q_{0,\zeta}}=\frac{\alpha_{\zeta}}{\beta_{\zeta}}, Q 0 , ζ P 0 , ζ = β ζ α ζ ,
the contribution proportional to δ Q ζ \delta Q_{\zeta} δ Q ζ in Eqs. (133) and (134) cancels in the difference x ζ − y ζ x_{\zeta}-y_{\zeta} x ζ − y ζ . We consequently obtain
x ζ − y ζ = − κ 2 ν S π 2 α ζ c U I E , ζ + O ( κ 4 ) , (140) x_{\zeta}-y_{\zeta}=-\frac{\kappa^{2}\nu_{S}}{\pi^{2}\alpha_{\zeta}c_{U}}I_{E,\zeta}+O(\kappa^{4}), \tag*{(140)} x ζ − y ζ = − π 2 α ζ c U κ 2 ν S I E , ζ + O ( κ 4 ) , (140)
where
I E , ζ = ∫ 0 ∞ d Ω e − Ω / Ω c [ 1 ± γ E ( Ω ) ] B 1 ( Ω ) . (141) I_{E,\zeta}=\int_{0}^{\infty}\mathrm{d}\Omega\,e^{-\Omega/\Omega_{c}}\left[1\pm\gamma_{E}(\Omega)\right]B_{1}(\Omega). \tag*{(141)} I E , ζ = ∫ 0 ∞ d Ω e − Ω/ Ω c [ 1 ± γ E ( Ω ) ] B 1 ( Ω ) . (141)
Substituting Eq. (140) into Eq. (139) , we finally obtain
G E e q = κ 4 N E + O ( κ 6 ) , (142) \mathcal{G}^{\mathrm{eq}}_{E}=\kappa^{4}\mathcal{N}_{E}+O(\kappa^{6}), \tag*{(142)} G E eq = κ 4 N E + O ( κ 6 ) , (142)
where
N E = ν S 3 16 π 4 c U ∑ ζ = ± 1 I E , ζ 2 α ζ 2 . (143) \mathcal{N}_{E}=\frac{\nu_{S}^{3}}{16\pi^{4}c_{U}}\sum_{\zeta=\pm1}\frac{I_{E,\zeta}^{2}}{\alpha_{\zeta}^{2}}. \tag*{(143)} N E = 16 π 4 c U ν S 3 ζ = ± 1 ∑ α ζ 2 I E , ζ 2 . (143)
Equivalently,
N E = lim κ → 0 G E e q κ 4 . (144) \mathcal{N}_{E}=\lim_{\kappa\to0}\frac{\mathcal{G}^{\mathrm{eq}}_{E}}{\kappa^{4}}. \tag*{(144)} N E = κ → 0 lim κ 4 G E eq . (144)
Thus, in the weak-coupling regime, the equilibrium Gaussian ergotropy exhibits a quartic dependence on the detector–bath coupling:
G E e q ∝ κ 4 . (145) \mathcal{G}^{\mathrm{eq}}_{E}\propto\kappa^{4}. \tag*{(145)} G E eq ∝ κ 4 . (145)
Gaussian Work as a Probe of Spatial Correlations
The accelerated vacuum and the matched inertial thermal field share the same local spectrum. Their distinction is encoded in the cross spectrum sampled by the two detectors. We now determine how this difference enters the leading-order work coefficient N E \mathcal{N}_{E} N E derived in Sec. IV. Throughout this analysis the static interaction is fully compensated and the extraction reference is the common Hamiltonian H S \mathcal{H}_{S} H S .
For convenience, we write the geometric factors obtained in Sec. II as
L ( a , d ) = 2 a arsinh a d 2 , L(a,d)=\frac{2}{a}\operatorname{arsinh}\frac{ad}{2}, L ( a , d ) = a 2 arsinh 2 a d ,
D ( a , d ) = d 1 + ( a d 2 ) 2 , D(a,d)=d\sqrt{1+\left(\frac{ad}{2}\right)^{2}}, D ( a , d ) = d 1 + ( 2 a d ) 2 ,
γ U ( Ω , a , d ) = sin [ Ω L ( a , d ) ] Ω D ( a , d ) , γ T ( Ω , d ) = sin ( Ω d ) Ω d . (146) \gamma_{U}(\Omega,a,d)=\frac{\sin[\Omega L(a,d)]}{\Omega D(a,d)}, \qquad\gamma_{T}(\Omega,d)=\frac{\sin(\Omega d)}{\Omega d}. \tag*{(146)} γ U ( Ω , a , d ) = Ω D ( a , d ) sin [ Ω L ( a , d )] , γ T ( Ω , d ) = Ω d sin ( Ω d ) . (146)
Defining b ( a , d ) ≡ L ( a , d ) / D ( a , d ) b(a,d)\equiv L(a,d)/D(a,d) b ( a , d ) ≡ L ( a , d ) / D ( a , d ) , we can write
γ U ( Ω , a , d ) = b ( a , d ) sinc [ Ω L ( a , d ) ] , (147) \gamma_{U}(\Omega,a,d)=b(a,d)\operatorname{sinc}[\Omega L(a,d)], \tag*{(147)} γ U ( Ω , a , d ) = b ( a , d ) sinc [ Ω L ( a , d )] , (147)
where sinc ( x ) ≡ sin x / x \operatorname{sinc}(x)\equiv\sin x/x sinc ( x ) ≡ sin x / x , with sinc ( 0 ) = 1 \operatorname{sinc}(0)=1 sinc ( 0 ) = 1 .
We choose the exponential regulator as
r exp ( Ω ) = e − Ω / Ω c . (148) r_{\exp}(\Omega)=e^{-\Omega/\Omega_{c}}. \tag*{(148)} r e x p ( Ω ) = e − Ω/ Ω c . (148)
To examine the dependence on its form, we also introduce a Gaussian regulator:
r G ( Ω ) = e − π Ω 2 / ( 4 Ω c 2 ) . (149) r_{G}(\Omega)=e^{-\pi\Omega^{2}/(4\Omega_{c}^{2})}. \tag*{(149)} r G ( Ω ) = e − π Ω 2 / ( 4 Ω c 2 ) . (149)
Both satisfy ∫ 0 ∞ r ( Ω ) d Ω = Ω c \int_{0}^{\infty}r(\Omega)\mathrm{d}\Omega=\Omega_{c} ∫ 0 ∞ r ( Ω ) d Ω = Ω c . In the following formulas, r r r denotes either specified frequency regulator; setting r = r exp r=r_{\exp} r = r e x p recovers the preceding model.
The low-frequency collective response
The symmetric and antisymmetric channels sample the combinations 1 ± γ E 1\pm\gamma_{E} 1 ± γ E . Their low-frequency behavior therefore exposes how the same local spectrum can produce different collective fluctuations. Put b = L / D b=L/D b = L / D , so that 0 < b < 1 0<b<1 0 < b < 1 for a , d > 0 a,d>0 a , d > 0 . Expanding the sine factors at fixed geometry gives
J U − ( Ω ) = κ 2 r ( Ω ) 4 π [ ( 1 − b ) Ω + b L 2 6 Ω 3 + O ( Ω 5 ) ] , J_{U-}(\Omega)=\frac{\kappa^{2}r(\Omega)}{4\pi}\left[(1-b)\Omega+\frac{bL^{2}}{6}\Omega^{3}+O(\Omega^{5})\right], J U − ( Ω ) = 4 π κ 2 r ( Ω ) [ ( 1 − b ) Ω + 6 b L 2 Ω 3 + O ( Ω 5 ) ] ,
J T − ( Ω ) = κ 2 r ( Ω ) 4 π [ d 2 6 Ω 3 + O ( Ω 5 ) ] . (150) J_{T-}(\Omega)=\frac{\kappa^{2}r(\Omega)}{4\pi}\left[\frac{d^{2}}{6}\Omega^{3}+O(\Omega^{5})\right]. \tag*{(150)} J T − ( Ω ) = 4 π κ 2 r ( Ω ) [ 6 d 2 Ω 3 + O ( Ω 5 ) ] . (150)
For a common regulator continuous at the origin with r ( 0 ) = 1 r(0)=1 r ( 0 ) = 1 , the antisymmetric accelerated channel is therefore Ohmic, while the matched inertial thermal channel is super-Ohmic with exponent three. The symmetric channels are Ohmic in both environments.
The symmetrized force spectrum is S E − ( Ω ) = J E − ( Ω ) coth [ Ω / ( 2 T U ) ] S_{E-}(\Omega)=J_{E-}(\Omega)\coth[\Omega/(2T_U)] S E − ( Ω ) = J E − ( Ω ) coth [ Ω/ ( 2 T U )] . At fixed T U > 0 T_U>0 T U > 0 , Eq. (150) gives
S U − ( Ω ) ⟶ κ 2 T U 2 π ( 1 − b ) , S_{U-}(\Omega)\longrightarrow\frac{\kappa^2 T_U}{2\pi}(1-b), S U − ( Ω ) ⟶ 2 π κ 2 T U ( 1 − b ) ,
S T − ( Ω ) ∼ κ 2 T U d 2 12 π Ω 2 . (151) S_{T-}(\Omega)\sim\frac{\kappa^2 T_U d^2}{12\pi}\Omega^2. \tag*{(151)} S T − ( Ω ) ∼ 12 π κ 2 T U d 2 Ω 2 . (151)
Thus the antisymmetric coordinate samples a finite low-frequency noise floor in U, whereas its noise in T is suppressed quadratically. Gaussian work, however, samples the full frequency-dependent detector response rather than the infrared limit alone.
For small separation at fixed a a a and Ω \Omega Ω , the same expansion takes the form
J U ( Ω ) = κ 2 r ( Ω ) d 2 24 π ( a 2 Ω + Ω 3 ) + O ( d 4 ) , J T ( Ω ) = κ 2 r ( Ω ) d 2 24 π Ω 3 + O ( d 4 ) . (152) \begin{aligned}
J_U(\Omega)&=\frac{\kappa^2 r(\Omega)d^2}{24\pi}(a^2\Omega+\Omega^3)+O(d^4),\\
J_T(\Omega)&=\frac{\kappa^2 r(\Omega)d^2}{24\pi}\Omega^3+O(d^4). \tag*{(152)}
\end{aligned} J U ( Ω ) J T ( Ω ) = 24 π κ 2 r ( Ω ) d 2 ( a 2 Ω + Ω 3 ) + O ( d 4 ) , = 24 π κ 2 r ( Ω ) d 2 Ω 3 + O ( d 4 ) . (152)
At exact coincidence the antisymmetric spectrum vanishes in both environments, and at a = 0 a=0 a = 0 the two spectra agree. The Ohmic and cubic responses refer to fixed a , d > 0 a,d>0 a , d > 0 .
From the weighted correlation to work
To isolate the spatial input from the common local response, we separate the local and cross contributions to the integral in Eq. (141) :
I 0 = ∫ 0 ∞ d Ω r ( Ω ) B 1 ( Ω ) , I c , E = ∫ 0 ∞ d Ω r ( Ω ) γ E ( Ω ) B 1 ( Ω ) , η E = I c , E I 0 , E ∈ { U , T , C } . (153) \begin{aligned}
I_0&=\int_0^\infty\mathrm{d}\Omega\,r(\Omega)B_1(\Omega),\\
I_{c,E}&=\int_0^\infty\mathrm{d}\Omega\,r(\Omega)\gamma_E(\Omega)B_1(\Omega),\\
\eta_E&=\frac{I_{c,E}}{I_0},\qquad E\in\{U,T,C\}. \tag*{(153)}
\end{aligned} I 0 I c , E η E = ∫ 0 ∞ d Ω r ( Ω ) B 1 ( Ω ) , = ∫ 0 ∞ d Ω r ( Ω ) γ E ( Ω ) B 1 ( Ω ) , = I 0 I c , E , E ∈ { U , T , C } . (153)
The weight r ( Ω ) B 1 ( Ω ) r(\Omega)B_1(\Omega) r ( Ω ) B 1 ( Ω ) is fixed by the detector frequency, the magnitude of the direct coupling, the temperature, and the regulator. We hold ( ω , ∣ λ ∣ , T U , r ) (\omega,|\lambda|,T_U,r) ( ω , ∣ λ ∣ , T U , r ) fixed when comparing the spatial correlations; changing the sign of λ \lambda λ leaves ν S \nu_S ν S and this weight unchanged. Thus η E \eta_E η E is one response-weighted spatial correlation.
Since B 1 B_1 B 1 is positive and ∣ γ E ∣ ≤ 1 |\gamma_E|\leq1 ∣ γ E ∣ ≤ 1 , the weighted correlation satisfies ∣ η E ∣ ≤ 1 |\eta_E|\leq1 ∣ η E ∣ ≤ 1 . We have
I E , ζ = I 0 ( 1 ± η E ) , (154) I_{E,\zeta}=I_0(1\pm\eta_E), \tag*{(154)} I E , ζ = I 0 ( 1 ± η E ) , (154)
and η C = 0 \eta_C=0 η C = 0 .
Using the geometric factor, we obtain the following relation between the weighted spatial correlations:
η U ( a , d ) = b ( a , d ) η T ( T U , L ( a , d ) ) . (155) \eta_U(a,d)=b(a,d)\eta_T(T_U,L(a,d)). \tag*{(155)} η U ( a , d ) = b ( a , d ) η T ( T U , L ( a , d )) . (155)
Both sides are evaluated at the same temperature, T U = a / ( 2 π ) T_U=a/(2\pi) T U = a / ( 2 π ) , showing that acceleration modifies the structure of the weighted correlation. This identity separates the geometric effects from the common local thermal response.
Substituting these expressions into Eq. (143) , we obtain
N C = I 0 2 ( ω 2 + λ 2 ) 8 π 4 ν S c U , \mathcal{N}_C=\frac{I_0^2(\omega^2+\lambda^2)}{8\pi^4\nu_S c_U}, N C = 8 π 4 ν S c U I 0 2 ( ω 2 + λ 2 ) ,
R E ≡ N E N C = 1 + η E 2 − s η E , s = 4 ω λ ω 2 + λ 2 . (156) R_E\equiv\frac{\mathcal{N}_E}{\mathcal{N}_C}=1+\eta_E^2-s\eta_E,\qquad s=\frac{4\omega\lambda}{\omega^2+\lambda^2}. \tag*{(156)} R E ≡ N C N E = 1 + η E 2 − s η E , s = ω 2 + λ 2 4 ω λ . (156)
Thus, the cross correlation enters the leading-order work coefficient through a quadratic dependence on the weighted correlation. In particular, for λ = 0 \lambda=0 λ = 0 we find R E = 1 + η E 2 R_E=1+\eta_E^2 R E = 1 + η E 2 . For positive λ \lambda λ and 0 < η E < 1 0<\eta_E<1 0 < η E < 1 , correlations enhance the coefficient relative to C C C when η E > s \eta_E>s η E > s , and suppress it when η E < s \eta_E<s η E < s . The minimum occurs at
η E = s 2 , R min = 1 − s 2 4 = ( ω 2 − λ 2 ) 2 ( ω 2 + λ 2 ) 2 . (157) \eta_E=\frac{s}{2},\qquad R_{\min}=1-\frac{s^2}{4}=\frac{(\omega^2-\lambda^2)^2}{(\omega^2+\lambda^2)^2}. \tag*{(157)} η E = 2 s , R m i n = 1 − 4 s 2 = ( ω 2 + λ 2 ) 2 ( ω 2 − λ 2 ) 2 . (157)
This quadratic dependence establishes how the weighted correlation is encoded in the work and determines what can be inferred from work measurements. We now apply this relation to investigate the distinguishability between the accelerated and inertial environments.
Equilibrium correlation readout and finite-coupling tests
The unique finite-distance blind point
Subtracting the two physical coefficients in Eq. (156) gives
N U − N T N C = ( η U − η T ) ( η U + η T − s ) . (158) \frac{\mathcal{N}_U-\mathcal{N}_T}{\mathcal{N}_C}=(\eta_U-\eta_T)(\eta_U+\eta_T-s). \tag*{(158)} N C N U − N T = ( η U − η T ) ( η U + η T − s ) . (158)
For the exponential and Gaussian regulators in Eq. (149) , the weighted correlations satisfy
0 < η U ( a , d ) < η T ( T U , d ) < 1 , a , d > 0. (159) 0<\eta_U(a,d)<\eta_T(T_U,d)<1,\qquad a,d>0. \tag*{(159)} 0 < η U ( a , d ) < η T ( T U , d ) < 1 , a , d > 0. (159)
Both weighted correlations decrease strictly with separation at fixed a > 0 a>0 a > 0 . Appendix C proves these properties by expressing the cross integrals through a positive, decreasing cosine transform.
Since η U + η T \eta_U+\eta_T η U + η T decreases continuously and monotonically from 2 to 0 as the separation d d d increases, and 0 < s < 2 0<s<2 0 < s < 2 for 0 < λ < ω 0<\lambda<\omega 0 < λ < ω , there exists a unique finite and nonzero separation d 0 > 0 d_0>0 d 0 > 0 , which we refer to as the blind point, at which the two environments cannot be distinguished by a single-setting work measurement. It is determined by
η U ( a , d 0 ) + η T ( T U , d 0 ) = s . (160) \eta_U(a,d_0)+\eta_T(T_U,d_0)=s. \tag*{(160)} η U ( a , d 0 ) + η T ( T U , d 0 ) = s . (160)
At this separation the leading work contrast vanishes although the field correlations remain different. For 0 < d < d 0 0<d<d_0 0 < d < d 0 we have N U < N T \mathcal{N}_U<\mathcal{N}_T N U < N T , whereas for d > d 0 d>d_0 d > d 0 we have N U > N T \mathcal{N}_U > \mathcal{N}_T N U > N T . Each coefficient has one distance minimum, and their positions obey
d min , U < d 0 < d min , T . (161) d_{\min,U} < d_0 < d_{\min,T}. \tag*{(161)} d m i n , U < d 0 < d m i n , T . (161)
For λ ≤ 0 \lambda\le0 λ ≤ 0 , the work ordering is N U < N T \mathcal{N}_U < \mathcal{N}_T N U < N T at every finite a , d > 0 a,d>0 a , d > 0 . At d = 0 d=0 d = 0 the two environments share η U = η T = 1 \eta_U=\eta_T=1 η U = η T = 1 , and their work contrast vanishes. The contrast also tends to zero as d → ∞ d\to\infty d → ∞ and is identically zero at a = 0 a=0 a = 0 . The point d 0 d_0 d 0 is therefore the unique interior zero for positive direct coupling: two different weighted correlations give the same leading work.
Two-setting reconstruction of the weighted correlation
In this subsection, we show how to overcome this blind point. We take λ = + g \lambda=+g λ = + g in the first equilibrium preparation and λ = − g \lambda=-g λ = − g in the second, with 0 < g < ω 0<g<\omega 0 < g < ω and all other parameters fixed. The normalized work difference between the two settings is
A E = N E ( − g ) − N E ( + g ) N E ( − g ) + N E ( + g ) = s g η E 1 + η E , s g = 4 ω g ω 2 + g 2 . (162) \mathcal{A}_E=\frac{\mathcal{N}_E(-g)-\mathcal{N}_E(+g)}{\mathcal{N}_E(-g)+\mathcal{N}_E(+g)}=\frac{s_g\eta_E}{1+\eta_E},
\qquad
s_g=\frac{4\omega g}{\omega^2+g^2}. \tag*{(162)} A E = N E ( − g ) + N E ( + g ) N E ( − g ) − N E ( + g ) = 1 + η E s g η E , s g = ω 2 + g 2 4 ω g . (162)
The common factor containing I 0 2 I_0^2 I 0 2 cancels. Differentiating the last expression gives
d A d η = s g ( 1 − η 2 ) ( 1 + η 2 ) 2 > 0 , 0 < η < 1. (163) \frac{d\mathcal{A}}{d\eta}=\frac{s_g(1-\eta^2)}{(1+\eta^2)^2}>0,
\qquad0<\eta<1. \tag*{(163)} d η d A = ( 1 + η 2 ) 2 s g ( 1 − η 2 ) > 0 , 0 < η < 1. (163)
Combining Eqs. (159) and (163) , we find
0 < A U < A T < s g 2 . (164) 0<\mathcal{A}_U<\mathcal{A}_T<\frac{s_g}{2}. \tag*{(164)} 0 < A U < A T < 2 s g . (164)
Thus the two settings remove the finite-distance blind point of the leading coefficient. Solving Eq. (162) for the root in 0 < η < 1 0<\eta<1 0 < η < 1 , we obtain
η E = 2 A E s g + s g 2 − 4 A E 2 . (165) \eta_E=\frac{2\mathcal{A}_E}{s_g+\sqrt{s_g^2-4\mathcal{A}_E^2}}. \tag*{(165)} η E = s g + s g 2 − 4 A E 2 2 A E . (165)
The common factor I 0 2 I_0^2 I 0 2 cancels in the normalized work asymmetry, allowing the weighted spatial correlation to be determined without separately evaluating the local spectral response integral.
Matched calculations and full Gibbs tests
In practice, the coupling strength κ \kappa κ cannot be taken arbitrarily close to zero and must remain finite. This raises the question of whether the preceding theoretical relations can still distinguish the two environments at finite coupling. To address this question, we retain the full interacting Green function without performing a weak-coupling expansion, obtain the covariance matrix of the full interacting Gibbs state, and calculate the Gaussian ergotropy directly from this covariance matrix. We first examine both the single-setting and two-setting readout schemes at the blind-point separation d 0 d_0 d 0 predicted by the weak-coupling theory. At this separation, we examine whether the full Gibbs calculation can still distinguish the two environments and whether the two-setting scheme improves the distinction. We further substitute the finite-coupling work values into the inversion formula derived in the weak-coupling regime and compare the reconstructed weighted correlations η ^ E \hat{\eta}_E η ^ E with the theoretical values η E \eta_E η E obtained directly from their defining integrals. These comparisons allow us to assess the applicability of the preceding weak-coupling relations at finite coupling.
In the following calculations, we adopt the exponential regulator and set ω = 1 \omega=1 ω = 1 , a = 2 π a=2\pi a = 2 π , T U = 1 T_U=1 T U = 1 , Ω c = 5 \Omega_c=5 Ω c = 5 , and g = 0.1 g=0.1 g = 0.1 . All Gaussian ergotropies are evaluated with respect to the same detector Hamiltonian H S \mathcal{H}_S H S .
Figure 2 displays the complete chain from the weighted correlation to work and back. For a single interaction setting, the two work coefficients intersect at the blind-point separation d 0 d_0 d 0 , while their respective minima occur at d min , U d_{\min,U} d m i n , U and d min , T d_{\min,T} d m i n , T . Numerically, we obtain
Figure 2. Dependence of the weighted correlation and Gaussian ergotropy on the detector separation d d d in the accelerated vacuum (U) and the inertial thermal field (T), with fixed parameters ω = 1 \omega= 1 ω = 1 , a = 2 π a = 2\pi a = 2 π , and Ω c = 5 \Omega_c = 5 Ω c = 5 . (a) Theoretical weighted correlation η E \eta_E η E . (b) Normalized Gaussian ergotropy for λ = 0.1 \lambda= 0.1 λ = 0.1 . (c) A E A_E A E for λ = ± 0.1 \lambda= \pm0.1 λ = ± 0.1 . (d) Weighted correlation η ^ E \widehat{\eta}_E η E reconstructed from the work asymmetry. The solid and dashed curves represent the weak-coupling results, while the open circles show the finite-coupling results at κ 2 = 0.05 \kappa^2 = 0.05 κ 2 = 0.05 , obtained using the full interacting Green function without a weak-coupling expansion. The vertical dotted lines mark d min , U d_{\min,U} d m i n , U , d 0 d_0 d 0 , and d min , T d_{\min,T} d m i n , T .
d 0 = 1.1237119953 , d min , U = 0.6396353494 , d min , T = 1.9104614104. (166) d_0=1.1237119953,
\qquad
d_{\min,U}=0.6396353494,
\qquad
d_{\min,T}=1.9104614104. \tag*{(166)} d 0 = 1.1237119953 , d m i n , U = 0.6396353494 , d m i n , T = 1.9104614104. (166)
These values satisfy the analytical ordering in Eq. (161) . At d 0 d_0 d 0 , the leading correlations are η U = 0.0770176168 \eta_U=0.0770176168 η U = 0.0770176168 and η T = 0.3190219871 \eta_T=0.3190219871 η T = 0.3190219871 : the equality of the single-setting work occurs between distinct spatial correlations.
The solid and dashed curves in Figure 2 are obtained from the leading-order weak-coupling expressions. In contrast, the open circles are calculated at finite coupling κ 2 = 0.05 \kappa^2=0.05 κ 2 = 0.05 , retaining the full interacting Green function without performing a weak-coupling expansion. At the blind-point separation d 0 d_0 d 0 predicted by the weak-coupling analytical expressions, the Gaussian ergotropies obtained from the full interacting Gibbs covariance are
G U ( + g ) = 1.47351849 × 10 − 7 , \mathcal{G}_U(+g)=1.47351849\times10^{-7}, G U ( + g ) = 1.47351849 × 1 0 − 7 ,
G T ( + g ) = 1.47344610 × 10 − 7 , \mathcal{G}_T(+g)=1.47344610\times10^{-7}, G T ( + g ) = 1.47344610 × 1 0 − 7 ,
G U ( − g ) = 1.56559763 × 10 − 7 , \mathcal{G}_U(-g)=1.56559763\times10^{-7}, G U ( − g ) = 1.56559763 × 1 0 − 7 ,
G T ( − g ) = 1.85483574 × 10 − 7 . (167) \mathcal{G}_T(-g)=1.85483574\times10^{-7}. \tag*{(167)} G T ( − g ) = 1.85483574 × 1 0 − 7 . (167)
At the blind point predicted by the weak-coupling theory, we find that G U ( + g ) \mathcal{G}_U(+g) G U ( + g ) and G T ( + g ) \mathcal{G}_T(+g) G T ( + g ) are nearly identical. Using both interaction settings, we obtain A U , W = 0.03029800 \mathcal{A}_{U,W}=0.03029800 A U , W = 0.03029800 and A T , W = 0.11459055 \mathcal{A}_{T,W}=0.11459055 A T , W = 0.11459055 . This demonstrates that the two-setting measurement retains a finite ability to distinguish the two environments precisely where the single-setting contrast vanishes.
Substituting the work values in Eq. (167) into Eq. (165) , we reconstruct η ^ U = 0.07695551 \hat{\eta}_U=0.07695551 η ^ U = 0.07695551 and η ^ T = \hat{\eta}_T= η ^ T = 0.31873607. These values are in close agreement with the theoretical weighted correlations, η U = 0.07701762 \eta_U = 0.07701762 η U = 0.07701762 and η T = 0.31902199 \eta_T = 0.31902199 η T = 0.31902199 , indicating that the two-setting reconstruction scheme retains its ability to distinguish the two environments at finite coupling.
Figure 3 (a) shows the equilibrium Gaussian ergotropy normalized by its leading-order weak-coupling contribution κ 4 N E \kappa^4 \mathcal{N}_E κ 4 N E . For both environments, the normalized ergotropy decreases monotonically with increasing κ 2 \kappa^2 κ 2 and remains below unity over the displayed range. This indicates that the full Gibbs ergotropy is lower than the leading-order prediction at finite coupling. The decrease is more pronounced for the accelerated vacuum than for the inertial thermal field, revealing a stronger finite-coupling correction in the former. As κ 2 \kappa^2 κ 2 decreases, both curves approach unity, and the deviation from the leading-order prediction gradually disappears. Figure 3 (b) shows the deviations of the reconstructed weighted correlations from their theoretical values. The reconstruction errors decrease as the detector–field coupling becomes weaker, exhibiting an approximately linear dependence on κ 2 \kappa^2 κ 2 . The error for the inertial thermal field remains larger than that for the accelerated vacuum, indicating different finite-coupling corrections in the two environments. Both errors vanish in the weak-coupling limit, consistent with the O ( κ 2 ) O(\kappa^2) O ( κ 2 ) correction predicted by the leading-order inversion formula. Taken together, these results demonstrate that the weak-coupling theory provides a reliable description of the full Gibbs calculation and that the two-setting scheme remains effective for reconstructing the weighted spatial correlations at finite coupling.
Figure 3. Dependence of the equilibrium Gaussian ergotropy and the correlation reconstruction error on the detector–field coupling strength κ 2 \kappa^2 κ 2 in the accelerated vacuum (U) and the inertial thermal field (T), with fixed parameters ω = 1 \omega= 1 ω = 1 , d = 1 d = 1 d = 1 , a = 2 π a = 2\pi a = 2 π , and Ω c = 5 \Omega_c = 5 Ω c = 5 . (a) Equilibrium Gaussian ergotropy normalized by its leading-order contribution κ 4 N E \kappa^4\mathcal{N}_E κ 4 N E , for λ = 0.5 \lambda= 0.5 λ = 0.5 ; the blue dashed line indicates the weak-coupling limit. (b) Absolute reconstruction error ∣ η E − η E ∣ \lvert\eta_E-\eta_E\rvert ∣ η E − η E ∣ obtained from the two settings λ = ± 0.1 \lambda= \pm0.1 λ = ± 0.1 . The solid curves are obtained using the full interacting Green function without a weak-coupling expansion.
Spatial correlations and the additional global work
The equilibrium work readout established above relies on joint Gaussian operations on the two detectors. This raises a further question: how much of the extracted work can already be obtained through independent local operations on the two detectors, and how much constitutes an additional contribution accessible only through joint operations? To distinguish these two contributions, we set λ = 0 \lambda= 0 λ = 0 , so that the extraction Hamiltonian becomes additive,
H S = ω 2 ( q A 2 + p A 2 + q B 2 + p B 2 ) = ω 2 ( q + 2 + p + 2 + q − 2 + p − 2 ) . (168) \mathcal{H}_S = \frac{\omega}{2}\left(q_A^2+p_A^2+q_B^2+p_B^2\right) = \frac{\omega}{2}\left(q_+^2+p_+^2+q_-^2+p_-^2\right). \tag*{(168)} H S = 2 ω ( q A 2 + p A 2 + q B 2 + p B 2 ) = 2 ω ( q + 2 + p + 2 + q − 2 + p − 2 ) . (168)
At equilibrium, we define the second moments of the collective coordinates and momenta as
Q ζ e q = ⟨ q ζ 2 ⟩ e q , P ζ e q = ⟨ p ζ 2 ⟩ e q , ζ = ± . (169) Q_\zeta^{\mathrm{eq}}=\langle q_\zeta^2\rangle_{\mathrm{eq}},\qquad P_\zeta^{\mathrm{eq}}=\langle p_\zeta^2\rangle_{\mathrm{eq}},\qquad\zeta=\pm. \tag*{(169)} Q ζ eq = ⟨ q ζ 2 ⟩ eq , P ζ eq = ⟨ p ζ 2 ⟩ eq , ζ = ± . (169)
The corresponding collective covariance matrices are
σ ζ e q = ( Q ζ e q 0 0 P ζ e q ) . (170) \sigma_\zeta^{\mathrm{eq}}=
\begin{pmatrix}
Q_\zeta^{\mathrm{eq}} & 0\\
0 & P_\zeta^{\mathrm{eq}}
\end{pmatrix}. \tag*{(170)} σ ζ eq = ( Q ζ eq 0 0 P ζ eq ) . (170)
Transforming back to the detector basis, the equilibrium covariance matrix takes the form
σ A B e q = ( σ ˉ e q σ c e q σ c e q σ ˉ e q ) , (171) \sigma_{AB}^{\mathrm{eq}}=
\begin{pmatrix}
\bar{\sigma}^{\mathrm{eq}} & \sigma_c^{\mathrm{eq}}\\
\sigma_c^{\mathrm{eq}} & \bar{\sigma}^{\mathrm{eq}}
\end{pmatrix}, \tag*{(171)} σ A B eq = ( σ ˉ eq σ c eq σ c eq σ ˉ eq ) , (171)
where
σ ˉ e q = σ + e q + σ − e q 2 , σ c e q = σ + e q − σ − e q 2 . (172) \bar{\sigma}^{\mathrm{eq}}=\frac{\sigma_+^{\mathrm{eq}}+\sigma_-^{\mathrm{eq}}}{2},\qquad
\sigma_c^{\mathrm{eq}}=\frac{\sigma_+^{\mathrm{eq}}-\sigma_-^{\mathrm{eq}}}{2}. \tag*{(172)} σ ˉ eq = 2 σ + eq + σ − eq , σ c eq = 2 σ + eq − σ − eq . (172)
Here, σ ˉ e q \bar{\sigma}^{\mathrm{eq}} σ ˉ eq is the covariance matrix of either local detector state, while σ c e q \sigma_c^{\mathrm{eq}} σ c eq describes the inter-detector covariance. The equality of the two local covariance matrices follows from the symmetry between the detectors.
We first consider local work extraction, for which the allowed Gaussian unitary operations are restricted to independent transformations on the two detectors. The symplectic eigenvalue of either local covariance matrix is
ν ˉ e q = det σ ˉ e q . (173) \bar{\nu}^{\mathrm{eq}}=\sqrt{\det\bar{\sigma}^{\mathrm{eq}}}. \tag*{(173)} ν ˉ eq = det σ ˉ eq . (173)
For a single detector with the extraction Hamiltonian H A = ω ( q A 2 + p A 2 ) / 2 \mathcal{H}_A=\omega(q_A^2+p_A^2)/2 H A = ω ( q A 2 + p A 2 ) /2 , the minimum energy attainable under Gaussian unitary operations is ω ν ˉ e q \omega\bar{\nu}^{\mathrm{eq}} ω ν ˉ eq . The same result holds for detector B. Consequently, the total local passive energy is
E p a s l o c , e q = 2 ω ν ˉ e q = 2 ω det σ ˉ e q . (174) E_{\mathrm{pas}}^{\mathrm{loc,eq}}=2\omega\bar{\nu}^{\mathrm{eq}}=2\omega\sqrt{\det\bar{\sigma}^{\mathrm{eq}}}. \tag*{(174)} E pas loc , eq = 2 ω ν ˉ eq = 2 ω det σ ˉ eq . (174)
The initial energy of the equilibrium detector state, evaluated with respect to H S \mathcal{H}_S H S , is
E S e q = ω 2 ( Q + e q + P + e q + Q − e q + P − e q ) . (175) E_S^{\mathrm{eq}}=\frac{\omega}{2}\left(Q_+^{\mathrm{eq}}+P_+^{\mathrm{eq}}+Q_-^{\mathrm{eq}}+P_-^{\mathrm{eq}}\right). \tag*{(175)} E S eq = 2 ω ( Q + eq + P + eq + Q − eq + P − eq ) . (175)
It follows that the local Gaussian ergotropy is
G l o c e q = E S e q − 2 ω det σ ˉ e q . (176) \mathcal{G}_{\mathrm{loc}}^{\mathrm{eq}}=E_S^{\mathrm{eq}}-2\omega\sqrt{\det\bar{\sigma}^{\mathrm{eq}}}. \tag*{(176)} G loc eq = E S eq − 2 ω det σ ˉ eq . (176)
We now turn to global work extraction, where joint Gaussian unitary operations are allowed on the entire two-detector system. In the collective-mode representation, the symplectic eigenvalues of the equilibrium covariance matrix are
ν ± e q = det σ ± e q = Q ± e q P ± e q . (177) \nu_\pm^{\mathrm{eq}}=\sqrt{\det\sigma_\pm^{\mathrm{eq}}}=\sqrt{Q_\pm^{\mathrm{eq}}P_\pm^{\mathrm{eq}}}. \tag*{(177)} ν ± eq = det σ ± eq = Q ± eq P ± eq . (177)
Since both collective modes have the same frequency ω \omega ω , the global passive energy is given by the sum of their minimum energies,
E p a s g l o b , e q = ω ( ν + e q + ν − e q ) = ω ( det σ + e q + det σ − e q ) . (178) E_{\mathrm{pas}}^{\mathrm{glob,eq}}=\omega\left(\nu_+^{\mathrm{eq}}+\nu_-^{\mathrm{eq}}\right)=\omega\left(\sqrt{\det\sigma_+^{\mathrm{eq}}}+\sqrt{\det\sigma_-^{\mathrm{eq}}}\right). \tag*{(178)} E pas glob , eq = ω ( ν + eq + ν − eq ) = ω ( det σ + eq + det σ − eq ) . (178)
The corresponding global Gaussian ergotropy is therefore
G g l o b e q = E S e q − ω ( det σ + e q + det σ − e q ) . (179) \mathcal{G}_{\mathrm{glob}}^{\mathrm{eq}}=E_S^{\mathrm{eq}}-\omega\left(\sqrt{\det\sigma_+^{\mathrm{eq}}}+\sqrt{\det\sigma_-^{\mathrm{eq}}}\right). \tag*{(179)} G glob eq = E S eq − ω ( det σ + eq + det σ − eq ) . (179)
Comparing the two extraction protocols, we define the additional work accessible through joint operations as
Δ G c o r r e q = G g l o b e q − G l o c e q . (180) \Delta\mathcal{G}_{\mathrm{corr}}^{\mathrm{eq}}=\mathcal{G}_{\mathrm{glob}}^{\mathrm{eq}}-\mathcal{G}_{\mathrm{loc}}^{\mathrm{eq}}. \tag*{(180)} Δ G corr eq = G glob eq − G loc eq . (180)
Substituting the above expressions, the common initial energy E S e q E_S^{\mathrm{eq}} E S eq cancels, yielding
Δ G c o r r e q = ω [ 2 det σ ˉ e q − det σ + e q − det σ − e q ] . (181) \Delta\mathcal{G}_{\mathrm{corr}}^{\mathrm{eq}} = \omega\left[2\sqrt{\det\bar{\sigma}^{\mathrm{eq}}} - \sqrt{\det\sigma_{+}^{\mathrm{eq}}} - \sqrt{\det\sigma_{-}^{\mathrm{eq}}}\right]. \tag*{(181)} Δ G corr eq = ω [ 2 det σ ˉ eq − det σ + eq − det σ − eq ] . (181)
The non-negativity of this quantity can be established explicitly. Define
A e q = ( Q + e q + Q − e q ) ( P + e q + P − e q ) , B e q = Q + e q P + e q + Q − e q P − e q . (182) \begin{aligned}
A_{\mathrm{eq}} &= \sqrt{\left(Q_{+}^{\mathrm{eq}} + Q_{-}^{\mathrm{eq}}\right)\left(P_{+}^{\mathrm{eq}} + P_{-}^{\mathrm{eq}}\right)},\\
B_{\mathrm{eq}} &= \sqrt{Q_{+}^{\mathrm{eq}}P_{+}^{\mathrm{eq}}} + \sqrt{Q_{-}^{\mathrm{eq}}P_{-}^{\mathrm{eq}}}. \tag*{(182)}
\end{aligned} A eq B eq = ( Q + eq + Q − eq ) ( P + eq + P − eq ) , = Q + eq P + eq + Q − eq P − eq . (182)
Equation (181) then gives
Δ G c o r r e q ω = A e q − B e q . (183) \frac{\Delta\mathcal{G}_{\mathrm{corr}}^{\mathrm{eq}}}{\omega} = A_{\mathrm{eq}} - B_{\mathrm{eq}}. \tag*{(183)} ω Δ G corr eq = A eq − B eq . (183)
Using
A e q 2 − B e q 2 = Q + e q P − e q + Q − e q P + e q − 2 Q + e q P + e q Q − e q P − e q = ( Q + e q P − e q − Q − e q P + e q ) 2 , (184) \begin{aligned}
A_{\mathrm{eq}}^2 - B_{\mathrm{eq}}^2 &= Q_{+}^{\mathrm{eq}}P_{-}^{\mathrm{eq}} + Q_{-}^{\mathrm{eq}}P_{+}^{\mathrm{eq}} \\
&\quad- 2\sqrt{Q_{+}^{\mathrm{eq}}P_{+}^{\mathrm{eq}}Q_{-}^{\mathrm{eq}}P_{-}^{\mathrm{eq}}}\\
&= \left(\sqrt{Q_{+}^{\mathrm{eq}}P_{-}^{\mathrm{eq}}} - \sqrt{Q_{-}^{\mathrm{eq}}P_{+}^{\mathrm{eq}}}\right)^2, \tag*{(184)}
\end{aligned} A eq 2 − B eq 2 = Q + eq P − eq + Q − eq P + eq − 2 Q + eq P + eq Q − eq P − eq = ( Q + eq P − eq − Q − eq P + eq ) 2 , (184)
we obtain
Δ G c o r r e q ω = ( Q + e q P − e q − Q − e q P + e q ) 2 A e q + B e q ≥ 0. (185) \frac{\Delta\mathcal{G}_{\mathrm{corr}}^{\mathrm{eq}}}{\omega} = \frac{\left(\sqrt{Q_{+}^{\mathrm{eq}}P_{-}^{\mathrm{eq}}} - \sqrt{Q_{-}^{\mathrm{eq}}P_{+}^{\mathrm{eq}}}\right)^2}{A_{\mathrm{eq}} + B_{\mathrm{eq}}} \ge0. \tag*{(185)} ω Δ G corr eq = A eq + B eq ( Q + eq P − eq − Q − eq P + eq ) 2 ≥ 0. (185)
Thus, joint Gaussian operations can never yield less extractable work than independent local operations. Equality holds precisely when
Q + e q P − e q = Q − e q P + e q , (186) Q^{\mathrm{eq}}_+P^{\mathrm{eq}}_- = Q^{\mathrm{eq}}_-P^{\mathrm{eq}}_+, \tag*{(186)} Q + eq P − eq = Q − eq P + eq , (186)
or, equivalently,
Q + e q P + e q = Q − e q P − e q . (187) \frac{Q^{\mathrm{eq}}_+}{P^{\mathrm{eq}}_+}=\frac{Q^{\mathrm{eq}}_-}{P^{\mathrm{eq}}_-}. \tag*{(187)} P + eq Q + eq = P − eq Q − eq . (187)
This condition means that the two collective covariance matrices are proportional. Hence, a difference in the overall fluctuation amplitudes of the two collective modes is not sufficient to produce additional global work; the two modes must also exhibit different ratios of position and momentum fluctuations.
To establish the connection between the additional global work and the spatial correlations induced by the common environment, we consider the weak-coupling expansion. At zero detector–field coupling, the two collective modes reduce to identical thermal states with Q ± e q = P ± e q = c U / 2 Q^{\mathrm{eq}}_\pm=P^{\mathrm{eq}}_\pm=c_U/2 Q ± eq = P ± eq = c U /2 , where c U = coth ( ω / 2 T U ) c_U=\coth(\omega/2T_U) c U = coth ( ω /2 T U ) . For weak but finite coupling, the equilibrium variances therefore admit the expansions
Q ζ e q = c U 2 + κ 2 q ζ ( 1 ) + O ( κ 4 ) , P ζ e q = c U 2 + κ 2 p ζ ( 1 ) + O ( κ 4 ) , ζ = ± , (188) \begin{aligned}
Q^{\mathrm{eq}}_\zeta&= \frac{c_U}{2}+\kappa^2 q^{(1)}_\zeta+O(\kappa^4),\\
P^{\mathrm{eq}}_\zeta&= \frac{c_U}{2}+\kappa^2 p^{(1)}_\zeta+O(\kappa^4), \qquad\zeta=\pm, \tag*{(188)}
\end{aligned} Q ζ eq P ζ eq = 2 c U + κ 2 q ζ ( 1 ) + O ( κ 4 ) , = 2 c U + κ 2 p ζ ( 1 ) + O ( κ 4 ) , ζ = ± , (188)
where the first-order Taylor coefficients are decomposed into their average and half-difference as
q 0 = q + ( 1 ) + q − ( 1 ) 2 , q c = q + ( 1 ) − q − ( 1 ) 2 , p 0 = p + ( 1 ) + p − ( 1 ) 2 , p c = p + ( 1 ) − p − ( 1 ) 2 . (189) \begin{aligned}
q_0&=\frac{q^{(1)}_+ + q^{(1)}_-}{2}, \qquad q_c=\frac{q^{(1)}_+ - q^{(1)}_-}{2},\\
p_0&=\frac{p^{(1)}_+ + p^{(1)}_-}{2}, \qquad p_c=\frac{p^{(1)}_+ - p^{(1)}_-}{2}. \tag*{(189)}
\end{aligned} q 0 p 0 = 2 q + ( 1 ) + q − ( 1 ) , q c = 2 q + ( 1 ) − q − ( 1 ) , = 2 p + ( 1 ) + p − ( 1 ) , p c = 2 p + ( 1 ) − p − ( 1 ) . (189)
Thus, q 0 q_0 q 0 and p 0 p_0 p 0 represent the average first-order corrections of the two collective modes, while q c q_c q c and p c p_c p c quantify half the differences between their respective corrections. Accordingly, the average corrections affect both modes equally, whereas the difference corrections enter with opposite signs for the + + + and − - − modes.
For either collective mode, the corresponding single-mode Gaussian ergotropy is
G ζ e q = ω [ Q ζ e q + P ζ e q 2 − Q ζ e q P ζ e q ] = ω κ 4 4 c U ( p ζ ( 1 ) − q ζ ( 1 ) ) 2 + O ( κ 6 ) . (190) \begin{aligned}
\mathcal{G}^{\mathrm{eq}}_\zeta
&=\omega\left[\frac{Q^{\mathrm{eq}}_\zeta+P^{\mathrm{eq}}_\zeta}{2}-\sqrt{Q^{\mathrm{eq}}_\zeta P^{\mathrm{eq}}_\zeta}\right]\\
&=\frac{\omega\kappa^4}{4c_U}\left(p^{(1)}_\zeta-q^{(1)}_\zeta\right)^2+O(\kappa^6). \tag*{(190)}
\end{aligned} G ζ eq = ω [ 2 Q ζ eq + P ζ eq − Q ζ eq P ζ eq ] = 4 c U ω κ 4 ( p ζ ( 1 ) − q ζ ( 1 ) ) 2 + O ( κ 6 ) . (190)
The term linear in κ 2 \kappa^2 κ 2 vanishes because the first derivatives of the work function with respect to Q Q Q and P P P vanish at the reference thermal covariance Q = P = c U / 2 Q=P=c_U/2 Q = P = c U /2 . Consequently, higher-order corrections to the covariance matrices do not affect the leading O ( κ 4 ) O(\kappa^4) O ( κ 4 ) work.
Summing the contributions from the two collective modes, the global Gaussian ergotropy becomes
G g l o b e q = ∑ ζ = ± G ζ e q = ω κ 4 2 c U [ ( p 0 − q 0 ) 2 + ( p c − q c ) 2 ] + O ( κ 6 ) . (191) \begin{aligned}
\mathcal{G}^{\mathrm{eq}}_{\mathrm{glob}}&=\sum_{\zeta=\pm}\mathcal{G}^{\mathrm{eq}}_\zeta\\
&=\frac{\omega\kappa^4}{2c_U}\left[(p_0-q_0)^2+(p_c-q_c)^2\right]+O(\kappa^6). \tag*{(191)}
\end{aligned} G glob eq = ζ = ± ∑ G ζ eq = 2 c U ω κ 4 [ ( p 0 − q 0 ) 2 + ( p c − q c ) 2 ] + O ( κ 6 ) . (191)
For local extraction, the two detectors have identical marginal variances,
Q ˉ e q = Q + e q + Q − e q 2 = c U 2 + κ 2 q 0 + O ( κ 4 ) , \bar{Q}^{\mathrm{eq}} = \frac{Q_{+}^{\mathrm{eq}} + Q_{-}^{\mathrm{eq}}}{2} = \frac{c_U}{2} + \kappa^2 q_0 + O(\kappa^4), Q ˉ eq = 2 Q + eq + Q − eq = 2 c U + κ 2 q 0 + O ( κ 4 ) ,
P ˉ e q = P + e q + P − e q 2 = c U 2 + κ 2 p 0 + O ( κ 4 ) . (192) \bar{P}^{\mathrm{eq}} = \frac{P_{+}^{\mathrm{eq}} + P_{-}^{\mathrm{eq}}}{2} = \frac{c_U}{2} + \kappa^2 p_0 + O(\kappa^4). \tag*{(192)} P ˉ eq = 2 P + eq + P − eq = 2 c U + κ 2 p 0 + O ( κ 4 ) . (192)
The mode-dependent corrections cancel in the local marginals. Since the two detectors contribute equally, their total local Gaussian ergotropy is
G l o c e q = 2 ω [ Q ˉ e q + P ˉ e q 2 − Q ˉ e q P ˉ e q ] = ω κ 4 2 c U ( p 0 − q 0 ) 2 + O ( κ 6 ) . (193) \begin{aligned}
\mathcal{G}_{\mathrm{loc}}^{\mathrm{eq}} = 2\omega\left[\frac{\bar{Q}^{\mathrm{eq}}+\bar{P}^{\mathrm{eq}}}{2} - \sqrt{\bar{Q}^{\mathrm{eq}}\bar{P}^{\mathrm{eq}}}\right] \\
&= \frac{\omega\kappa^4}{2c_U}(p_0-q_0)^2 + O(\kappa^6). \tag*{(193)}
\end{aligned} G loc eq = 2 ω [ 2 Q ˉ eq + P ˉ eq − Q ˉ eq P ˉ eq ] = 2 c U ω κ 4 ( p 0 − q 0 ) 2 + O ( κ 6 ) . (193)
Subtracting Eq. (193) from Eq. (191) , we obtain
Δ G c o r r e q = ω κ 4 2 c U ( p c − q c ) 2 + O ( κ 6 ) . (194) \Delta\mathcal{G}_{\mathrm{corr}}^{\mathrm{eq}} = \frac{\omega\kappa^4}{2c_U}(p_c-q_c)^2 + O(\kappa^6). \tag*{(194)} Δ G corr eq = 2 c U ω κ 4 ( p c − q c ) 2 + O ( κ 6 ) . (194)
This result shows that the additional global work is determined entirely by the difference between the mode-dependent momentum and position corrections. Using the previously defined weighted spatial correlation
η E = I c , E I 0 , (195) \eta_E = \frac{I_{c,E}}{I_0}, \tag*{(195)} η E = I 0 I c , E , (195)
and introducing the leading-order local-work coefficient for λ = 0 \lambda=0 λ = 0 ,
N C = ω I 0 2 8 π 4 c U , (196) \mathcal{N}_C = \frac{\omega I_0^2}{8\pi^4 c_U}, \tag*{(196)} N C = 8 π 4 c U ω I 0 2 , (196)
the local and global Gaussian ergotropies can be expressed as
G l o c e q = κ 4 N C + O ( κ 6 ) , \mathcal{G}_{\mathrm{loc}}^{\mathrm{eq}} = \kappa^4\mathcal{N}_C + O(\kappa^6), G loc eq = κ 4 N C + O ( κ 6 ) ,
G g l o b e q = κ 4 N C ( 1 + η E 2 ) + O ( κ 6 ) . (197) \mathcal{G}_{\mathrm{glob}}^{\mathrm{eq}} = \kappa^4\mathcal{N}_C(1+\eta_E^2) + O(\kappa^6). \tag*{(197)} G glob eq = κ 4 N C ( 1 + η E 2 ) + O ( κ 6 ) . (197)
The difference between the global and local Gaussian ergotropies is therefore given by
Δ G c o r r e q = κ 4 N C η E 2 + O ( κ 6 ) . (198) \Delta\mathcal{G}_{\mathrm{corr}}^{\mathrm{eq}} = \kappa^4\mathcal{N}_C\eta_E^2 + O(\kappa^6). \tag*{(198)} Δ G corr eq = κ 4 N C η E 2 + O ( κ 6 ) . (198)
This relation establishes a direct connection between the spatial correlation induced by the common environment and the additional work accessible through joint Gaussian operations. The common response of the two collective modes determines the leading local contribution, whereas their different responses give rise to the additional global contribution.
We finally examine whether the additional global work requires entanglement between the two detectors. For a two-mode Gaussian state, this can be assessed using the positive-partial-transpose (PPT) criterion. Partial transposition with respect to detector B B B reverses the sign of p B p_B p B . In the collective-mode representation, this operation interchanges p + p_+ p + and p − p_- p − while leaving q + q_+ q + and q − q_- q − unchanged.
The partially transposed equilibrium covariance matrix therefore takes the form
σ A B e q , P T ∼ diag ( Q + e q , P − e q , Q − e q , P + e q ) . (199) \sigma_{AB}^{\mathrm{eq,PT}} \sim\operatorname{diag}\left(Q_{+}^{\mathrm{eq}},P_{-}^{\mathrm{eq}},Q_{-}^{\mathrm{eq}},P_{+}^{\mathrm{eq}}\right). \tag*{(199)} σ A B eq , PT ∼ diag ( Q + eq , P − eq , Q − eq , P + eq ) . (199)
Its symplectic eigenvalues are
ν ~ 1 e q = Q + e q P − e q , ν ~ 2 e q = Q − e q P + e q . (200) \widetilde{\nu}_1^{\mathrm{eq}} = \sqrt{Q_{+}^{\mathrm{eq}}P_{-}^{\mathrm{eq}}}, \qquad\widetilde{\nu}_2^{\mathrm{eq}} = \sqrt{Q_{-}^{\mathrm{eq}}P_{+}^{\mathrm{eq}}}. \tag*{(200)} ν 1 eq = Q + eq P − eq , ν 2 eq = Q − eq P + eq . (200)
In our covariance convention, the two-mode Gaussian state is separable if and only if
min ( ν ~ 1 e q , ν ~ 2 e q ) ≥ 1 2 . (201) \min\left(\widetilde{\nu}_1^{\mathrm{eq}},\widetilde{\nu}_2^{\mathrm{eq}}\right) \ge\frac{1}{2}. \tag*{(201)} min ( ν 1 eq , ν 2 eq ) ≥ 2 1 . (201)
At zero detector–field coupling, the two collective modes reduce to identical thermal states, with
Q ± e q ( 0 ) = P ± e q ( 0 ) = c U 2 . (202) Q_{\pm}^{\mathrm{eq}}(0)=P_{\pm}^{\mathrm{eq}}(0)=\frac{c_U}{2}. \tag*{(202)} Q ± eq ( 0 ) = P ± eq ( 0 ) = 2 c U . (202)
The corresponding PPT symplectic eigenvalues therefore satisfy
min ( ν ~ 1 e q , ν ~ 2 e q ) − 1 2 = 1 2 coth ( ω 2 T U ) − 1 2 = 1 e ω / T U − 1 > 0. (203) \begin{aligned}
\min\left(\widetilde{\nu}_1^{\mathrm{eq}},\widetilde{\nu}_2^{\mathrm{eq}}\right)-\frac{1}{2}
= \frac{1}{2}\coth\left(\frac{\omega}{2T_U}\right)-\frac{1}{2} \\
&= \frac{1}{e^{\omega/T_U}-1}>0. \tag*{(203)}
\end{aligned} min ( ν 1 eq , ν 2 eq ) − 2 1 = 2 1 coth ( 2 T U ω ) − 2 1 = e ω / T U − 1 1 > 0. (203)
At zero detector–field coupling, the smallest PPT symplectic eigenvalue is strictly larger than the separability threshold 1 / 2 1/2 1/2 for T U > 0 T_U>0 T U > 0 . Since the regulated equilibrium covariance matrix and its partially transposed symplectic eigenvalues depend continuously on κ 2 \kappa^2 κ 2 , the smallest PPT symplectic eigenvalue remains above 1 / 2 1/2 1/2 for sufficiently weak but finite coupling. Therefore, there exists a κ ∗ 2 > 0 \kappa_*^2>0 κ ∗ 2 > 0 such that the two-detector equilibrium state remains separable for 0 ≤ κ 2 < κ ∗ 2 0\le\kappa^2<\kappa_*^2 0 ≤ κ 2 < κ ∗ 2 .
Since I c , E > 0 I_{c,E}>0 I c , E > 0 , Eq. (198) implies that the additional global work is positive for sufficiently weak but nonzero coupling. At the same time, the PPT criterion shows that the two-detector equilibrium state remains separable over a finite interval of weak coupling for T U > 0 T_U>0 T U > 0 . Hence, within this interval, the detectors can remain separable, while joint Gaussian operations still extract more work than independent local operations. This demonstrates that entanglement between the detectors is not necessary for the additional global work. At T U = 0 T_U=0 T U = 0 , the PPT symplectic eigenvalues equal 1 / 2 1/2 1/2 at zero coupling, exactly at the separability boundary.
Finite-Time Dynamics of Gaussian Ergotropy
Having established the equilibrium relation between spatial correlations and the additional global work, we now investigate the instantaneous dynamics of the Gaussian ergotropy in the accelerated-vacuum environment. Using the covariance-matrix propagation developed in Sec. III, we calculate the time-dependent Gaussian ergotropy G U ( τ ) \mathcal{G}_{U}(\tau) G U ( τ ) of the detector pair with respect to the detector Hamiltonian H S H_{S} H S , as defined in Eq. (86) .
For the numerical propagation, we specify the bath discretization introduced in Sec. II by employing an N N N -point Gauss–Legendre quadrature on the finite frequency interval [ Ω min , Ω max ] [\Omega_{\min},\Omega_{\max}] [ Ω m i n , Ω m a x ] . The quadrature nodes z j z_j z j are the zeros of the Legendre polynomial P N ( z ) P_N(z) P N ( z ) on [ − 1 , 1 ] [-1,1] [ − 1 , 1 ] :
P N ( z j ) = 0 , − 1 < z 1 < z 2 < ⋯ < z N < 1. (204) P_N(z_j)=0,\qquad-1<z_1<z_2<\cdots<z_N<1. \tag*{(204)} P N ( z j ) = 0 , − 1 < z 1 < z 2 < ⋯ < z N < 1. (204)
The corresponding reference weights are
w ~ j = 2 ( 1 − z j 2 ) [ P N ′ ( z j ) ] 2 . (205) \widetilde{w}_j=\frac{2}{(1-z_j^2)[P_N'(z_j)]^2}. \tag*{(205)} w j = ( 1 − z j 2 ) [ P N ′ ( z j ) ] 2 2 . (205)
Mapping the reference interval onto the physical frequency interval gives
Ω j = Ω max − Ω min 2 z j + Ω max + Ω min 2 , \Omega_j=\frac{\Omega_{\max}-\Omega_{\min}}{2}z_j+\frac{\Omega_{\max}+\Omega_{\min}}{2}, Ω j = 2 Ω m a x − Ω m i n z j + 2 Ω m a x + Ω m i n ,
w j = Ω max − Ω min 2 w ~ j . (206) w_j=\frac{\Omega_{\max}-\Omega_{\min}}{2}\widetilde{w}_j. \tag*{(206)} w j = 2 Ω m a x − Ω m i n w j . (206)
The frequency integrals are then evaluated according to
∫ Ω min Ω max f ( Ω ) d Ω ≃ ∑ j = 1 N w j f ( Ω j ) . (207) \int_{\Omega_{\min}}^{\Omega_{\max}} f(\Omega)\,\mathrm{d}\Omega\simeq\sum_{j=1}^{N}w_jf(\Omega_j). \tag*{(207)} ∫ Ω m i n Ω m a x f ( Ω ) d Ω ≃ j = 1 ∑ N w j f ( Ω j ) . (207)
We take Ω min = 10 − 5 \Omega_{\min}=10^{-5} Ω m i n = 1 0 − 5 and Ω max = 50 \Omega_{\max}=50 Ω m a x = 50 . The numerical results use the exponential frequency regulator, with Ω c = 5 \Omega_c=5 Ω c = 5 unless otherwise specified.
We first examine the dependence of the transient Gaussian ergotropy on the detector–field coupling strength κ 2 \kappa^2 κ 2 . Figure 4 shows that G U ( τ ) \mathcal{G}_{U}(\tau) G U ( τ ) exhibits an oscillatory decay as the proper time increases. The successive local maxima become larger as κ 2 \kappa^2 κ 2 increases. For the three displayed coupling strengths, the early peak amplitudes approximately follow a ratio of 1 : 4 : 9 1:4:9 1 : 4 : 9 , indicating a quadratic dependence on the plotted parameter κ 2 \kappa^2 κ 2 . This is consistent with the leading weak-coupling scaling G U ( τ ) ∝ κ 4 \mathcal{G}_{U}(\tau)\propto\kappa^4 G U ( τ ) ∝ κ 4 . The oscillation period is approximately 3.6 3.6 3.6 , while the oscillation envelope decreases progressively. At stronger coupling, the transient response decays more rapidly, and the differences between the curves become less pronounced at later times.
Figure 4. Time evolution of the Gaussian ergotropy for different values of the detector–field coupling strength κ 2 \kappa^2 κ 2 , with fixed parameters λ = 0.5 \lambda=0.5 λ = 0.5 , ω = 1 \omega=1 ω = 1 , d = 1 d=1 d = 1 , a = 2 π a=2\pi a = 2 π , and N = 20000 N=20000 N = 20000 .
We next investigate the effect of the detector separation d d d . As shown in Figure 5 , the Gaussian ergotropy exhibits an oscillatory decay for all displayed separations, and its local maxima increase with increasing d d d . The oscillation period remains approximately 3.6 3.6 3.6 , while the envelopes decay progressively. At fixed acceleration, changing d d d leaves the local spectral contribution and the Unruh temperature unchanged, but modifies the spatial cross-correlation factor γ U ( Ω , a , d ) \gamma_U(\Omega,a,d) γ U ( Ω , a , d ) . As the separation increases, the cross-correlation envelope becomes weaker, and the two collective spectral densities progressively approach the same local spectrum. The distance dependence of the transient response consequently becomes less pronounced at sufficiently large separation.
Figure 5. Time evolution of the Gaussian ergotropy for different values of the detector separation d d d , with fixed parameters λ = 0.5 \lambda=0.5 λ = 0.5 , ω = 1 \omega=1 ω = 1 , κ 2 = 0.05 \kappa^2=0.05 κ 2 = 0.05 , a = 2 π a=2\pi a = 2 π , and N = 20000 N=20000 N = 20000 .
We further examine the influence of acceleration. Figure 6 shows that the transient Gaussian ergotropy increases in magnitude with increasing acceleration, while its oscillation envelope decays more rapidly. The oscillation periods remain approximately unchanged over the displayed range. Notably, the increase in ergotropy indicates that stronger Unruh-induced thermal fluctuations do not necessarily suppress work extraction; instead, they are accompanied by an enhancement of the transient Gaussian ergotropy in the parameter range considered. This behavior reflects the combined influence of the increasing Unruh temperature T U = a / ( 2 π ) T_U=a/(2\pi) T U = a / ( 2 π ) and the acceleration-dependent spatial correlation factor γ U ( Ω , a , d ) \gamma_U(\Omega,a,d) γ U ( Ω , a , d ) . Thus, stronger thermal fluctuations can enhance the transient work signal, even though the corresponding oscillations decay more rapidly.
Figure 6. Time evolution of the Gaussian ergotropy for different values of the acceleration a a a , with fixed parameters λ = 0.5 \lambda= 0.5 λ = 0.5 , ω = 1 \omega= 1 ω = 1 , κ 2 = 0.05 \kappa^2 = 0.05 κ 2 = 0.05 , d = 1 d = 1 d = 1 , and N = 20000 N = 20000 N = 20000 .
We now turn to the effect of the direct interaction between the detectors. Figure 7 shows the time evolution for different values of λ \lambda λ . The Gaussian ergotropy exhibits an oscillatory decay for all displayed values. Increasing λ \lambda λ produces higher peaks and more widely separated oscillations, whereas at small λ \lambda λ the ergotropy remains relatively small and the curves nearly coincide. This behavior is related to the collective detector frequency
Figure 7. Time evolution of the Gaussian ergotropy for different values of the direct detector coupling strength λ \lambda λ , with fixed parameters ω = 1 \omega= 1 ω = 1 , κ 2 = 0.05 \kappa^2 = 0.05 κ 2 = 0.05 , d = 1 d = 1 d = 1 , a = 2 π a = 2\pi a = 2 π , and N = 20000 N = 20000 N = 20000 .
ν S = ω 2 − λ 2 , (208) \nu_{S} = \sqrt{\omega^2 - \lambda^2}, \tag*{(208)} ν S = ω 2 − λ 2 , (208)
which decreases as λ \lambda λ increases. The resulting longer oscillation period is accompanied by enhanced two-mode squeezing of the detector ground state, thereby modifying the response of the detector pair to the thermal field.
Finally, we investigate the dependence on the detector frequency ω \omega ω . Figure 8 shows that increasing ω \omega ω suppresses the transient peak amplitudes and produces faster oscillations. At fixed λ \lambda λ , increasing ω \omega ω raises the collective frequency ν S \nu_S ν S . This reduces the thermal occupation around the detector frequency and also changes the response kernel B 1 B_1 B 1 . Together, these changes affect how much Gaussian ergotropy can be extracted from the field fluctuations.
Figure 8. Time evolution of the Gaussian ergotropy for different values of the detector frequency ω \omega ω , with fixed parameters λ = 0.5 \lambda= 0.5 λ = 0.5 , κ 2 = 0.05 \kappa^2 = 0.05 κ 2 = 0.05 , d = 1 d = 1 d = 1 , a = 2 π a = 2\pi a = 2 π , and N = 20000 N = 20000 N = 20000 .
We finally examine the long-time evolution of the Gaussian ergotropy. Figure 9 shows its time dependence over an extended proper-time interval. The Gaussian ergotropy exhibits persistent oscillations with a gradually decreasing envelope. The oscillation amplitude decreases rapidly at early times and then decays more slowly at later times, while the oscillatory behavior remains visible throughout the displayed interval.
Figure 9. Time evolution of the Gaussian ergotropy in the accelerated vacuum, with fixed parameters λ = 0.5 \lambda= 0.5 λ = 0.5 , ω = 1 \omega= 1 ω = 1 , κ 2 = 0.05 \kappa^2 = 0.05 κ 2 = 0.05 , d = 1 d = 1 d = 1 , a = 2 π a = 2\pi a = 2 π , Ω c = 5 \Omega_c = 5 Ω c = 5 , and N = 30000 N = 30000 N = 30000 .
Conclusions
In this work, we have investigated how the spatial correlations of a bosonic environment are encoded in the Gaussian ergotropy of two harmonic-oscillator detectors.
By comparing the accelerated vacuum with an inertial thermal field at the same Unruh temperature, we have shown that identical local spectra do not imply identical collective responses. The leading equilibrium Gaussian ergotropy is governed by the response-weighted spatial correlation η E \eta_E η E , establishing a direct connection between environmental spatial correlations and extractable work.
We have further shown that a single interaction setting exhibits a unique finite-distance blind point, at which the two environments yield identical leading-order ergotropies despite having different spatial correlations. This ambiguity can be removed by considering two opposite direct-coupling settings, λ = ± g \lambda= \pm g λ = ± g . The resulting normalized work asymmetry A E A_E A E is a monotonic function of η E \eta_E η E and can be inverted to reconstruct the weighted spatial correlation. Thus, the two-setting scheme eliminates the blind point and distinguishes environments that cannot be distinguished through their local spectra alone. Calculations using the unexpanded interacting Green function further confirm the validity of this reconstruction at finite coupling. Moreover, when the direct detector coupling vanishes, the additional work accessible through joint Gaussian operations is proportional to η E 2 \eta_E^2 η E 2 . This advantage can persist even when the detector pair remains separable, showing that entanglement is not a necessary resource for additional global work extraction.
Finally, the finite-time dynamics reveal an oscillatory decay of Gaussian ergotropy, whose magnitude and decay behavior depend on the detector–field coupling, separation, acceleration, and detector parameters. Stronger acceleration enhances the transient work while accelerating its decay, whereas increasing the detector frequency suppresses work extraction and speeds up the oscillations. These findings establish Gaussian ergotropy as a useful probe of both the spatial structure of quantum environments and their dynamical influence on detector systems.
Acknowledgments
This work is supported by the Natural Science Foundation of Hainan Province under Grant No. 125RC744; the China Scholarship Council (CSC).
Gaussian conventions, field spectra, and switching preparation
Gaussian-state conventions and quadratic Hamiltonians
Throughout this paper, we work in natural units:
ℏ = c = k B = 1. (209) \hbar= c = k_{\mathrm{B}} = 1. \tag*{(209)} ℏ = c = k B = 1. (209)
Consider an n n n -mode bosonic system described by n n n pairs of canonical conjugate coordinates and momenta ( q i , p i ) (q_i,p_i) ( q i , p i ) . Since the full Hamiltonian is quadratic in these variables, we assemble them into the column vector R = ( q 1 , q 2 , … , q n , p 1 , p 2 , … , p n ) T \mathbf{R} = (q_1,q_2,\ldots,q_n,p_1,p_2,\ldots,p_n)^{\mathrm{T}} R = ( q 1 , q 2 , … , q n , p 1 , p 2 , … , p n ) T and introduce the symmetric matrix H H H via
H = 1 2 R T H R . (210) \mathcal{H} = \frac{1}{2}\mathbf{R}^{\mathrm{T}}H\mathbf{R}. \tag*{(210)} H = 2 1 R T H R . (210)
Note that H H H must be positive semidefinite; otherwise the Hamiltonian H \mathcal{H} H would have a spectrum unbounded from below, which is unphysical.
For each oscillator in the system with canonical coordinate q q q and momentum p p p , we define the annihilation operator a a a and its Hermitian conjugate a † a^\dagger a † via [99 , 100 ]
q = a + a † 2 , p = − i a − a † 2 , [ q , p ] = i . (211) q = \frac{a+a^\dagger}{\sqrt{2}}, \qquad p = -i\frac{a-a^\dagger}{\sqrt{2}}, \qquad[q,p] = i. \tag*{(211)} q = 2 a + a † , p = − i 2 a − a † , [ q , p ] = i . (211)
Inverting these relations gives a = ( q + i p ) / 2 a = (q+ip)/\sqrt{2} a = ( q + i p ) / 2 , which leads to the identity
a † a + 1 2 = 1 2 ( q 2 + p 2 ) , (212) a^\dagger a + \frac{1}{2} = \frac{1}{2}\left(q^2+p^2\right), \tag*{(212)} a † a + 2 1 = 2 1 ( q 2 + p 2 ) , (212)
Thus, the free Hamiltonian of a harmonic oscillator with frequency ω \omega ω is
H = ω ( a † a + 1 2 ) = ω 2 ( q 2 + p 2 ) . (213) \mathcal{H} = \omega\left(a^\dagger a+\frac{1}{2}\right) = \frac{\omega}{2}\left(q^2+p^2\right). \tag*{(213)} H = ω ( a † a + 2 1 ) = 2 ω ( q 2 + p 2 ) . (213)
For a Gaussian state with vanishing first moments, let us furthermore define the covariance matrix σ \sigma σ as [96 , 100 ]
σ i j = 1 2 ⟨ R i R j + R j R i ⟩ . (214) \sigma_{ij} = \frac{1}{2}\langle R_iR_j+R_jR_i\rangle. \tag*{(214)} σ ij = 2 1 ⟨ R i R j + R j R i ⟩ . (214)
It should be noted that a covariance matrix σ \sigma σ is symmetric and must satisfy the uncertainty principle [96 , 101 , 102 ]
σ + i 2 Ω c a n ≥ 0 , \sigma+\frac{\mathrm{i}}{2}\Omega_{\mathrm{can}}\ge0, σ + 2 i Ω can ≥ 0 ,
where, in the ordering of R \mathbf{R} R above,
Ω c a n = ( 0 I n − I n 0 ) , \Omega_{\mathrm{can}}=\begin{pmatrix}0&I_n\\-I_n&0\end{pmatrix}, Ω can = ( 0 − I n I n 0 ) ,
with I n I_n I n the n × n n\times n n × n identity. in particular σ > 0 \sigma>0 σ > 0 as σ + i 2 Ω c a n ≥ 0 \sigma+\frac{\mathrm{i}}{2}\Omega_{\mathrm{can}}\ge0 σ + 2 i Ω can ≥ 0 .
Self-consistency between the switching function and the initial state
To avoid exciting high-frequency transients induced by abrupt switching on, we adopt a C ∞ C^\infty C ∞ S-shaped ramp function. Define
f ( x ) = e − 1 / x 1 x > 0 , f(x)=e^{-1/x}\mathbb{1}_{x>0}, f ( x ) = e − 1/ x 1 x > 0 ,
S ( x ) = f ( x ) f ( x ) + f ( 1 − x ) , (215) S(x)=\frac{f(x)}{f(x)+f(1-x)}, \tag*{(215)} S ( x ) = f ( x ) + f ( 1 − x ) f ( x ) , (215)
o = τ − τ o n τ s , o=\frac{\tau-\tau_{\mathrm{on}}}{\tau_s}, o = τ s τ − τ on ,
so
χ ( τ ) = { 0 , τ ≤ τ o n , S ( o ) , τ o n < τ < τ o n + τ o , 1 , τ ≥ τ o n + τ o . (216) \chi(\tau)=\begin{cases}
0,&\tau\le\tau_{\mathrm{on}},\\
S(o),&\tau_{\mathrm{on}}<\tau<\tau_{\mathrm{on}}+\tau_o,\\
1,&\tau\ge\tau_{\mathrm{on}}+\tau_o.
\end{cases} \tag*{(216)} χ ( τ ) = ⎩ ⎨ ⎧ 0 , S ( o ) , 1 , τ ≤ τ on , τ on < τ < τ on + τ o , τ ≥ τ on + τ o . (216)
The function S ( x ) S(x) S ( x ) has vanishing derivatives of all orders at x = 0 x=0 x = 0 and x = 1 x=1 x = 1 (all derivatives of e − 1 / x e^{-1/x} e − 1/ x vanish at 0 + 0^+ 0 + ), hence χ ∈ C ∞ \chi\in C^\infty χ ∈ C ∞ [103 ] . In this work, we set τ o n = 0 \tau_{\mathrm{on}}=0 τ on = 0 and τ o = 1 \tau_o=1 τ o = 1 .
Since χ ( 0 ) = 0 ⟹ H S B ( 0 ) = H C T ( 0 ) = 0 ⟹ H t o t ( 0 ) = H S + H B \chi(0)=0\Longrightarrow H_{SB}(0)=H_{CT}(0)=0\Longrightarrow H_{\mathrm{tot}}(0)=H_S+H_B χ ( 0 ) = 0 ⟹ H S B ( 0 ) = H C T ( 0 ) = 0 ⟹ H tot ( 0 ) = H S + H B , the initial state can be legitimately taken as the product state [104 ]
ρ ( 0 ) = ρ S ( 0 ) ⊗ ρ B ( 0 ) , (217) \rho(0)=\rho_S(0)\otimes\rho_B(0), \tag*{(217)} ρ ( 0 ) = ρ S ( 0 ) ⊗ ρ B ( 0 ) , (217)
where ρ S ( 0 ) \rho_S(0) ρ S ( 0 ) is the ground state of H S H_S H S , and ρ B ( 0 ) \rho_B(0) ρ B ( 0 ) is the thermal state at temperature T U T_U T U .
Before proceeding with the calculation, we briefly recall the residue theorem.
If f ( z ) f(z) f ( z ) is analytic inside a simple positively oriented closed contour Γ \Gamma Γ except for isolated singularities z k z_k z k , then
∮ Γ f ( z ) d z = 2 π i ∑ k Res ( f , z k ) . (218) \oint_{\Gamma}f(z)\,\mathrm{d}z=2\pi\mathrm{i}\sum_k\operatorname{Res}(f,z_k). \tag*{(218)} ∮ Γ f ( z ) d z = 2 π i k ∑ Res ( f , z k ) . (218)
If f ( z ) f(z) f ( z ) has a second-order pole at z n z_n z n and can be expressed as f ( z ) = h ( z ) / ( z − z n ) 2 f(z)=h(z)/(z-z_n)^2 f ( z ) = h ( z ) / ( z − z n ) 2 , where h ( z ) h(z) h ( z ) is analytic at z n z_n z n , then the residue of f ( z ) f(z) f ( z ) at z n z_n z n reads
Res ( f , z n ) = h ′ ( z n ) . (219) \operatorname{Res}(f,z_n)=h'(z_n). \tag*{(219)} Res ( f , z n ) = h ′ ( z n ) . (219)
From Eq. (16) , it follows that
W ~ A A ( Ω ) = W ~ B B ( Ω ) = ∫ − ∞ ∞ d Δ τ e i Ω Δ τ W A A ( Δ τ ) = − a 2 16 π 2 ∫ e i Ω τ sinh 2 [ a 2 ( τ − i ϵ ) ] d τ . (220) \widetilde{W}_{AA}(\Omega)=\widetilde{W}_{BB}(\Omega)=\int_{-\infty}^{\infty}\mathrm{d}\Delta\tau\,e^{\mathrm{i}\Omega\Delta\tau}W_{AA}(\Delta\tau)
=-\frac{a^2}{16\pi^2}\int\frac{e^{\mathrm{i}\Omega\tau}}{\sinh^2\left[\frac{a}{2}(\tau-\mathrm{i}\epsilon)\right]}\,\mathrm{d}\tau. \tag*{(220)} W AA ( Ω ) = W B B ( Ω ) = ∫ − ∞ ∞ d Δ τ e i ΩΔ τ W AA ( Δ τ ) = − 16 π 2 a 2 ∫ sinh 2 [ 2 a ( τ − i ϵ ) ] e i Ω τ d τ . (220)
Here we restrict our attention to the core integral:
∫ e i Ω τ sinh 2 [ a 2 ( τ − i ϵ ) ] d τ . (221) \int\frac{e^{\mathrm{i}\Omega\tau}}{\sinh^2\left[\frac{a}{2}(\tau-\mathrm{i}\epsilon)\right]}\,\mathrm{d}\tau. \tag*{(221)} ∫ sinh 2 [ 2 a ( τ − i ϵ ) ] e i Ω τ d τ . (221)
Performing the substitution z = a 2 ( τ − i ϵ ) z=\frac{a}{2}(\tau-\mathrm{i}\epsilon) z = 2 a ( τ − i ϵ ) . The integration contour L ϵ L_\epsilon L ϵ is given by
L ϵ : z = x − i a ϵ 2 , x ∈ R , x : − ∞ → + ∞ , (222) L_\epsilon:\quad z=x-\frac{\mathrm{i}a\epsilon}{2},\qquad x\in\mathbb{R},\qquad x:-\infty\to+\infty, \tag*{(222)} L ϵ : z = x − 2 i a ϵ , x ∈ R , x : − ∞ → + ∞ , (222)
we obtain
∫ e i Ω τ sinh 2 [ a 2 ( τ − i ϵ ) ] d τ = 2 a ∫ e 2 i Ω z / a sinh 2 z d z . (223) \int\frac{e^{\mathrm{i}\Omega\tau}}{\sinh^2\left[\frac{a}{2}(\tau-\mathrm{i}\epsilon)\right]}\,\mathrm{d}\tau
=\frac{2}{a}\int\frac{e^{2\mathrm{i}\Omega z/a}}{\sinh^2z}\,\mathrm{d}z. \tag*{(223)} ∫ sinh 2 [ 2 a ( τ − i ϵ ) ] e i Ω τ d τ = a 2 ∫ sinh 2 z e 2 i Ω z / a d z . (223)
Since the zeros of sinh z \sinh z sinh z , given by z n = i n π z_n=\mathrm{i}n\pi z n = i nπ with n ∈ Z n\in\mathbb{Z} n ∈ Z , are all second-order poles of 1 / sinh 2 z 1/\sinh^2z 1/ sinh 2 z , and hence they are second-order poles of the function
f ( z ) = e 2 i Ω z / a sinh 2 z (224) f(z)=\frac{e^{2\mathrm{i}\Omega z/a}}{\sinh^2z} \tag*{(224)} f ( z ) = sinh 2 z e 2 i Ω z / a (224)
Applying the residue theorem, we obtain
∫ L ϵ f ( z ) d z = 2 π i ∑ n = 0 ∞ Res ( f , z n ) . (225) \int_{L_\epsilon}f(z)\,\mathrm{d}z=2\pi\mathrm{i}\sum_{n=0}^{\infty}\operatorname{Res}(f,z_n). \tag*{(225)} ∫ L ϵ f ( z ) d z = 2 π i n = 0 ∑ ∞ Res ( f , z n ) . (225)
Using the residue formula for second-order poles, we obtain
Res ( f , z n ) = 2 i Ω a e − 2 π Ω n / a ( n = 0 , 1 , 2 , … ) . (226) \operatorname{Res}(f,z_n)=\frac{2\mathrm{i}\Omega}{a}e^{-2\pi\Omega n/a}\qquad(n=0,1,2,\ldots). \tag*{(226)} Res ( f , z n ) = a 2 i Ω e − 2 π Ω n / a ( n = 0 , 1 , 2 , … ) . (226)
It is straightforward to see that this is a convergent geometric series:
∑ n = 0 ∞ Res ( f , z n ) = 2 i Ω a ∑ n = 0 ∞ e − 2 π Ω n / a = 2 i Ω a 1 1 − e − 2 π Ω / a . (227) \sum_{n=0}^{\infty}\operatorname{Res}(f,z_n)=\frac{2\mathrm{i}\Omega}{a}\sum_{n=0}^{\infty}e^{-2\pi\Omega n/a}=\frac{2\mathrm{i}\Omega}{a}\frac{1}{1-e^{-2\pi\Omega/a}}. \tag*{(227)} n = 0 ∑ ∞ Res ( f , z n ) = a 2 i Ω n = 0 ∑ ∞ e − 2 π Ω n / a = a 2 i Ω 1 − e − 2 π Ω/ a 1 . (227)
Therefore,
∫ L ϵ f ( z ) d z = 2 π i 2 i Ω a 1 1 − e − 2 π Ω / a = − 4 π Ω a 1 1 − e − 2 π Ω / a . (228) \int_{L_\epsilon}f(z)\,\mathrm{d}z=2\pi\mathrm{i}\frac{2\mathrm{i}\Omega}{a}\frac{1}{1-e^{-2\pi\Omega/a}}=-\frac{4\pi\Omega}{a}\frac{1}{1-e^{-2\pi\Omega/a}}. \tag*{(228)} ∫ L ϵ f ( z ) d z = 2 π i a 2 i Ω 1 − e − 2 π Ω/ a 1 = − a 4 π Ω 1 − e − 2 π Ω/ a 1 . (228)
Hence we obtain
W ~ A A ( Ω ) = W ~ B B ( Ω ) = Ω 2 π 1 1 − e − 2 π Ω / a , Ω > 0. (229) \widetilde{W}_{AA}(\Omega)=\widetilde{W}_{BB}(\Omega)=\frac{\Omega}{2\pi}\frac{1}{1-e^{-2\pi\Omega/a}},\qquad\Omega>0. \tag*{(229)} W AA ( Ω ) = W B B ( Ω ) = 2 π Ω 1 − e − 2 π Ω/ a 1 , Ω > 0. (229)
Following the same procedure, we can compute the Fourier transform of the Wightman function for two distinct detectors:
W ~ A B ( Ω ) = W ~ B A ( Ω ) = 1 2 π d 1 + ( a d 2 ) 2 sin [ 2 Ω a arsinh a d 2 ] 1 − e − 2 π Ω / a . (230) \widetilde{W}_{AB}(\Omega)=\widetilde{W}_{BA}(\Omega)=\frac{1}{2\pi d\sqrt{1+\left(\frac{ad}{2}\right)^2}}\frac{\sin\left[\frac{2\Omega}{a}\operatorname{arsinh}\frac{ad}{2}\right]}{1-e^{-2\pi\Omega/a}}. \tag*{(230)} W A B ( Ω ) = W B A ( Ω ) = 2 π d 1 + ( 2 a d ) 2 1 1 − e − 2 π Ω/ a sin [ a 2Ω arsinh 2 a d ] . (230)
Geometric correlation factor $\gamma_E`
We define the geometric correlation factor γ E \gamma_E γ E as:
γ E ( Ω , a , d ) ≡ W ~ A B ( Ω ) W ~ A A ( Ω ) = sin [ 2 Ω a arsinh a d 2 ] Ω d 1 + ( a d 2 ) 2 . (231) \gamma_E(\Omega,a,d)\equiv\frac{\widetilde{W}_{AB}(\Omega)}{\widetilde{W}_{AA}(\Omega)}=\frac{\sin\left[\frac{2\Omega}{a}\operatorname{arsinh}\frac{ad}{2}\right]}{\Omega d\sqrt{1+\left(\frac{ad}{2}\right)^2}}. \tag*{(231)} γ E ( Ω , a , d ) ≡ W AA ( Ω ) W A B ( Ω ) = Ω d 1 + ( 2 a d ) 2 sin [ a 2Ω arsinh 2 a d ] . (231)
We set
u = arsinh ( a d 2 ) ≥ 0 , X = 2 Ω a u ≥ 0 , (232) u=\operatorname{arsinh}\left(\frac{ad}{2}\right)\geq0,\qquad X=\frac{2\Omega}{a}u\geq0, \tag*{(232)} u = arsinh ( 2 a d ) ≥ 0 , X = a 2Ω u ≥ 0 , (232)
then
γ E = 2 u sinh 2 u sin X X . (233) \gamma_E=\frac{2u}{\sinh2u}\frac{\sin X}{X}. \tag*{(233)} γ E = sinh 2 u 2 u X sin X . (233)
For d → 0 + d\to0^{+} d → 0 + , we have
u = arsinh ( a d 2 ) = a d 2 + O ( d 3 ) , (234) u=\operatorname{arsinh}\left(\frac{ad}{2}\right)=\frac{ad}{2}+O(d^3), \tag*{(234)} u = arsinh ( 2 a d ) = 2 a d + O ( d 3 ) , (234)
and therefore
X = 2 Ω a u = Ω d + O ( d 3 ) . (235) X=\frac{2\Omega}{a}u=\Omega d+O(d^3). \tag*{(235)} X = a 2Ω u = Ω d + O ( d 3 ) . (235)
It follows that
2 u sinh 2 u → 1 , sin X X → 1 , (236) \frac{2u}{\sinh2u}\to1,\qquad\frac{\sin X}{X}\to1, \tag*{(236)} sinh 2 u 2 u → 1 , X sin X → 1 , (236)
which yields
lim d → 0 + γ E ( Ω , a , d ) = 1. (237) \lim_{d\to0^{+}}\gamma_E(\Omega,a,d)=1. \tag*{(237)} d → 0 + lim γ E ( Ω , a , d ) = 1. (237)
For a → 0 + a\to0^{+} a → 0 + , we have
u = arsinh ( a d 2 ) = a d 2 + O ( a 3 ) . (238) u=\operatorname{arsinh}\left(\frac{ad}{2}\right)=\frac{ad}{2}+O(a^3). \tag*{(238)} u = arsinh ( 2 a d ) = 2 a d + O ( a 3 ) . (238)
Substituting into the expression for X X X yields
X = Ω d + O ( a 2 ) . (239) X=\Omega d+O(a^2). \tag*{(239)} X = Ω d + O ( a 2 ) . (239)
Meanwhile,
2 u sinh 2 u → 1 , (240) \frac{2u}{\sinh2u}\to1, \tag*{(240)} sinh 2 u 2 u → 1 , (240)
and therefore
γ E = 2 u sinh 2 u sin X X → sin ( Ω d ) Ω d . (241) \gamma_E=\frac{2u}{\sinh2u}\frac{\sin X}{X}\to\frac{\sin(\Omega d)}{\Omega d}. \tag*{(241)} γ E = sinh 2 u 2 u X sin X → Ω d sin ( Ω d ) . (241)
That is,
lim a → 0 + γ E ( Ω , a , d ) = sin ( Ω d ) Ω d . (242) \lim_{a\to0^{+}}\gamma_E(\Omega,a,d)=\frac{\sin(\Omega d)}{\Omega d}. \tag*{(242)} a → 0 + lim γ E ( Ω , a , d ) = Ω d sin ( Ω d ) . (242)
From the Taylor expansion of the hyperbolic sine function
sinh x = x + x 3 6 + x 5 120 + ⋯ (243) \sinh x=x+\frac{x^3}{6}+\frac{x^5}{120}+\cdots \tag*{(243)} sinh x = x + 6 x 3 + 120 x 5 + ⋯ (243)
it follows that sinh x > x \sinh x>x sinh x > x for x > 0 x>0 x > 0 . Thus,
0 < 2 u sinh 2 u < 1 ( u > 0 ) , (244) 0<\frac{2u}{\sinh2u}<1\qquad(u>0), \tag*{(244)} 0 < sinh 2 u 2 u < 1 ( u > 0 ) , (244)
and this expression approaches 1 only when u → 0 u\to0 u → 0 , i.e., when d → 0 d\to0 d → 0 .
Furthermore, from the Taylor expansion of the sine function
sin x = x − x 3 6 + ⋯ (245) \sin x=x-\frac{x^3}{6}+\cdots \tag*{(245)} sin x = x − 6 x 3 + ⋯ (245)
it follows that ∣ sin x ∣ ≤ ∣ x ∣ |\sin x|\leq|x| ∣ sin x ∣ ≤ ∣ x ∣ for all real numbers x x x . Hence, for X > 0 X>0 X > 0 ,
∣ sin X X ∣ ≤ 1 , (246) \left|\frac{\sin X}{X}\right|\leq1, \tag*{(246)} X sin X ≤ 1 , (246)
and it tends to 1 only as X → 0 X\to0 X → 0 .
Multiplying the two inequalities yields
∣ γ E ∣ = 2 u sinh 2 u ∣ sin X X ∣ ≤ 2 u sinh 2 u < 1 ( u > 0 ) . (247) |\gamma_E|=\frac{2u}{\sinh2u}\left|\frac{\sin X}{X}\right|\leq\frac{2u}{\sinh2u}<1\qquad(u>0). \tag*{(247)} ∣ γ E ∣ = sinh 2 u 2 u X sin X ≤ sinh 2 u 2 u < 1 ( u > 0 ) . (247)
Therefore, for all Ω > 0 \Omega>0 Ω > 0 , a > 0 a>0 a > 0 , and d > 0 d>0 d > 0 , we have
∣ γ E ∣ < 1. (248) |\gamma_E|<1. \tag*{(248)} ∣ γ E ∣ < 1. (248)
When d = 0 d=0 d = 0 , the value is defined as the limit 1 by continuity, so
∣ γ E ( Ω , a , d ) ∣ ≤ 1. (249) |\gamma_E(\Omega,a,d)|\leq1. \tag*{(249)} ∣ γ E ( Ω , a , d ) ∣ ≤ 1. (249)
Equilibrium covariance and continuum-bath relaxation
In Sec. IV, the detector equilibrium covariance is obtained from the Euclidean Green function of the coupled detector–bath system. Here we provide the intermediate steps connecting the total Hamiltonian of each collective sector to the effective Green function and the leading-order Gaussian ergotropy. We then examine when the resulting Gibbs covariance can also describe the long-time limit of the reduced detector dynamics.
Bath elimination and equilibrium covariance
The equilibrium quantities of interest are the detector variances Q ζ e q Q_\zeta^{\rm eq} Q ζ eq and P ζ e q P_\zeta^{\rm eq} P ζ eq . Although the total Hamiltonian contains all bath oscillators, these variances can be calculated without retaining the bath coordinates explicitly. The reason is that the Euclidean action in Eq. (104) is quadratic, so the bath variables can be integrated out exactly.
We first show how integrating out the bath coordinates leads to the effective Green function used in the main text. For each collective sector ζ \zeta ζ , we perform a Matsubara Fourier expansion of the detector and bath coordinates:
q ζ ( u ) = T U ∑ n q ζ , n e i ν n u , Q j , ζ ( u ) = T U ∑ n Q j ζ , n e i ν n u , (250) q_\zeta(u)=\sqrt{T_U}\sum_n q_{\zeta,n}e^{i\nu_n u},
\qquad
Q_{j,\zeta}(u)=\sqrt{T_U}\sum_n Q_{j\zeta,n}e^{i\nu_n u}, \tag*{(250)} q ζ ( u ) = T U n ∑ q ζ , n e i ν n u , Q j , ζ ( u ) = T U n ∑ Q j ζ , n e i ν n u , (250)
where ν n = 2 π n T U \nu_n=2\pi nT_U ν n = 2 π n T U are the bosonic Matsubara frequencies. Since the coordinates are real, their Fourier coefficients satisfy q ζ , − n = q ζ , n ∗ q_{\zeta,-n}=q_{\zeta,n}^* q ζ , − n = q ζ , n ∗ and Q j ζ , − n = Q j ζ , n ∗ Q_{j\zeta,-n}=Q_{j\zeta,n}^* Q j ζ , − n = Q j ζ , n ∗ .
Substituting these expansions into
S E ( ζ ) = ∫ 0 1 / T U d u [ q ˙ ζ 2 2 β ζ + α ζ 2 q ζ 2 + δ ω ζ 2 q ζ 2 + ∑ j ( Q ˙ j , ζ 2 2 Ω j + Ω j 2 Q j , ζ 2 + g j , ζ q ζ Q j , ζ ) ] , (251) S_E^{(\zeta)}=\int_0^{1/T_U}du\left[
\frac{\dot q_\zeta^2}{2\beta_\zeta}
+\frac{\alpha_\zeta}{2}q_\zeta^2
+\frac{\delta\omega_\zeta}{2}q_\zeta^2
+\sum_j\left(
\frac{\dot Q_{j,\zeta}^2}{2\Omega_j}
+\frac{\Omega_j}{2}Q_{j,\zeta}^2
+g_{j,\zeta}q_\zeta Q_{j,\zeta}
\right)\right], \tag*{(251)} S E ( ζ ) = ∫ 0 1/ T U d u [ 2 β ζ q ˙ ζ 2 + 2 α ζ q ζ 2 + 2 δ ω ζ q ζ 2 + j ∑ ( 2 Ω j Q ˙ j , ζ 2 + 2 Ω j Q j , ζ 2 + g j , ζ q ζ Q j , ζ ) ] , (251)
the contribution of the j j j th bath oscillator at frequency ν n \nu_n ν n is
ν n 2 + Ω j 2 2 Ω j ∣ Q j ζ , n ∣ 2 + g j , ζ Re ( q ζ , n ∗ Q j ζ , n ) . (252) \frac{\nu_n^2+\Omega_j^2}{2\Omega_j}|Q_{j\zeta,n}|^2
+g_{j,\zeta}\operatorname{Re}(q_{\zeta,n}^*Q_{j\zeta,n}). \tag*{(252)} 2 Ω j ν n 2 + Ω j 2 ∣ Q j ζ , n ∣ 2 + g j , ζ Re ( q ζ , n ∗ Q j ζ , n ) . (252)
The first term is the free bath contribution, while the second describes its coupling to the detector coordinate.
To remove Q j ζ , n Q_{j\zeta,n} Q j ζ , n , we complete the square:
ν n 2 + Ω j 2 2 Ω j ∣ Q j ζ , n ∣ 2 + g j , ζ Re ( q ζ , n ∗ Q j ζ , n ) = ν n 2 + Ω j 2 2 Ω j ∣ Q j ζ , n + g j , ζ Ω j ν n 2 + Ω j 2 q ζ , n ∣ 2 − g j , ζ 2 Ω j 2 ( ν n 2 + Ω j 2 ) ∣ q ζ , n ∣ 2 . (253) \begin{aligned}
\frac{\nu_n^2+\Omega_j^2}{2\Omega_j}|Q_{j\zeta,n}|^2
+g_{j,\zeta}\operatorname{Re}(q_{\zeta,n}^*Q_{j\zeta,n})
&=\frac{\nu_n^2+\Omega_j^2}{2\Omega_j}
\left|Q_{j\zeta,n}
+\frac{g_{j,\zeta}\Omega_j}{\nu_n^2+\Omega_j^2}q_{\zeta,n}\right|^2\\
&\quad-\frac{g_{j,\zeta}^2\Omega_j}{2(\nu_n^2+\Omega_j^2)}|q_{\zeta,n}|^2. \tag*{(253)}
\end{aligned} 2 Ω j ν n 2 + Ω j 2 ∣ Q j ζ , n ∣ 2 + g j , ζ Re ( q ζ , n ∗ Q j ζ , n ) = 2 Ω j ν n 2 + Ω j 2 Q j ζ , n + ν n 2 + Ω j 2 g j , ζ Ω j q ζ , n 2 − 2 ( ν n 2 + Ω j 2 ) g j , ζ 2 Ω j ∣ q ζ , n ∣ 2 . (253)
The shifted bath coordinate appears only quadratically. Its Gaussian integral contributes a factor independent of the detector coordinate, which is absorbed into the normalization. The remaining term depends only on q ζ , n q_{\zeta,n} q ζ , n and therefore modifies the detector’s effective action.
After integrating out all bath oscillators, we obtain
S e f f ( ζ ) = 1 2 ∑ n [ α ζ + ν n 2 β ζ + δ ω ζ − ∑ j g j , ζ 2 Ω j ν n 2 + Ω j 2 ] ∣ q ζ , n ∣ 2 , S_{\mathrm{eff}}^{(\zeta)}
=\frac{1}{2}\sum_n\left[
\alpha_\zeta+\frac{\nu_n^2}{\beta_\zeta}
+\delta\omega_\zeta
-\sum_j\frac{g_{j,\zeta}^2\Omega_j}{\nu_n^2+\Omega_j^2}
\right]|q_{\zeta,n}|^2, S eff ( ζ ) = 2 1 n ∑ [ α ζ + β ζ ν n 2 + δ ω ζ − j ∑ ν n 2 + Ω j 2 g j , ζ 2 Ω j ] ∣ q ζ , n ∣ 2 ,
up to a detector-independent constant. The last two contributions are the static counterterm and the bath-induced correction. Using
δ ω ζ = ∑ j g j , ζ 2 Ω j , \delta\omega_\zeta=\sum_j\frac{g_{j,\zeta}^2}{\Omega_j}, δ ω ζ = j ∑ Ω j g j , ζ 2 ,
their sum becomes
Σ ζ ( ν n ) = ∑ j g j , ζ 2 [ 1 Ω j − Ω j ν n 2 + Ω j 2 ] = ν n 2 ∑ j g j , ζ 2 Ω j ( ν n 2 + Ω j 2 ) = ν n 2 Λ ζ ( ν n ) . (254) \begin{aligned}
\Sigma_\zeta(\nu_n)
&=\sum_jg_{j,\zeta}^2\left[
\frac{1}{\Omega_j}
-\frac{\Omega_j}{\nu_n^2+\Omega_j^2}
\right]\\
&=\nu_n^2\sum_j\frac{g_{j,\zeta}^2}{\Omega_j(\nu_n^2+\Omega_j^2)}\\
&=\nu_n^2\Lambda_\zeta(\nu_n). \tag*{(254)}
\end{aligned} Σ ζ ( ν n ) = j ∑ g j , ζ 2 [ Ω j 1 − ν n 2 + Ω j 2 Ω j ] = ν n 2 j ∑ Ω j ( ν n 2 + Ω j 2 ) g j , ζ 2 = ν n 2 Λ ζ ( ν n ) . (254)
Hence the effective action takes the Gaussian form
S e f f ( ζ ) = 1 2 ∑ n G ζ − 1 ( i ν n ) ∣ q ζ , n ∣ 2 , (255) S_{\mathrm{eff}}^{(\zeta)}
=\frac{1}{2}\sum_nG_\zeta^{-1}(i\nu_n)|q_{\zeta,n}|^2, \tag*{(255)} S eff ( ζ ) = 2 1 n ∑ G ζ − 1 ( i ν n ) ∣ q ζ , n ∣ 2 , (255)
where
G ζ − 1 ( i ν n ) = α ζ + ν n 2 β ζ + Σ ζ ( ν n ) . (256) G_\zeta^{-1}(i\nu_n)
=\alpha_\zeta+\frac{\nu_n^2}{\beta_\zeta}+\Sigma_\zeta(\nu_n). \tag*{(256)} G ζ − 1 ( i ν n ) = α ζ + β ζ ν n 2 + Σ ζ ( ν n ) . (256)
The Green function determines the position variance through Eq. (111) . To calculate the momentum variance using the stationary virial identity, we also need the detector–bath cross-correlation. This quantity follows from the same Gaussian integration and does not require a separate calculation of the bath dynamics.
In Eq. (253) , the shifted Gaussian has the conditional mean
⟨ Q j ζ , n ⟩ q ζ = − g j , ζ Ω j ν n 2 + Ω j 2 q ζ , n . (257) \langle Q_{j\zeta,n}\rangle_{q_\zeta}
=-\frac{g_{j,\zeta}\Omega_j}{\nu_n^2+\Omega_j^2}q_{\zeta,n}. \tag*{(257)} ⟨ Q j ζ , n ⟩ q ζ = − ν n 2 + Ω j 2 g j , ζ Ω j q ζ , n . (257)
Averaging over the detector fluctuations then gives
⟨ q ζ Q j , ζ ⟩ e q = − T U ∑ n g j , ζ Ω j ν n 2 + Ω j 2 G ζ ( i ν n ) . (258) \langle q_\zeta Q_{j,\zeta}\rangle_{\rm eq}
=-T_U\sum_n\frac{g_{j,\zeta}\Omega_j}{\nu_n^2+\Omega_j^2}G_\zeta(i\nu_n). \tag*{(258)} ⟨ q ζ Q j , ζ ⟩ eq = − T U n ∑ ν n 2 + Ω j 2 g j , ζ Ω j G ζ ( i ν n ) . (258)
Summing over the bath modes yields
∑ j g j , ζ ⟨ q ζ Q j , ζ ⟩ e q = − δ ω ζ Q ζ e q + T U ∑ n Σ ζ ( ν n ) G ζ ( i ν n ) . (259) \sum_jg_{j,\zeta}\langle q_\zeta Q_{j,\zeta}\rangle_{\rm eq}
=-\delta\omega_\zeta Q_\zeta^{\rm eq}
+T_U\sum_n\Sigma_\zeta(\nu_n)G_\zeta(i\nu_n). \tag*{(259)} j ∑ g j , ζ ⟨ q ζ Q j , ζ ⟩ eq = − δ ω ζ Q ζ eq + T U n ∑ Σ ζ ( ν n ) G ζ ( i ν n ) . (259)
The first contribution cancels the counterterm in the stationary virial identity, Eq. (114) . We therefore recover
P ζ e q = T U β ζ ∑ n [ α ζ + Σ ζ ( ν n ) ] G ζ ( i ν n ) , (260) P_\zeta^{\rm eq}
=\frac{T_U}{\beta_\zeta}\sum_n
\left[\alpha_\zeta+\Sigma_\zeta(\nu_n)\right]G_\zeta(i\nu_n), \tag*{(260)} P ζ eq = β ζ T U n ∑ [ α ζ + Σ ζ ( ν n ) ] G ζ ( i ν n ) , (260)
which is Eq. (117) .
We next connect these covariance formulas to the weak-coupling work coefficient. The leading corrections are obtained by expanding the Green function in the bath self-energy, which is of order κ 2 \kappa^2 κ 2 . The resulting Matsubara sums can be written in terms of two kernels,
B k ( Ω ) = T U ∑ n ν n 2 ( ν n 2 + Ω 2 ) ( ν n 2 + ν S 2 ) k , k = 1 , 2. (261) B_k(\Omega)=T_U\sum_n\frac{\nu_n^2}{(\nu_n^2+\Omega^2)(\nu_n^2+\nu_S^2)^k},\qquad k=1,2. \tag*{(261)} B k ( Ω ) = T U n ∑ ( ν n 2 + Ω 2 ) ( ν n 2 + ν S 2 ) k ν n 2 , k = 1 , 2. (261)
They arise from the same self-energy but with different powers of the free detector Green function. Indeed, with D 0 , ζ = ( ν n 2 + ν S 2 ) / β ζ D_{0,\zeta}=(\nu_n^2+\nu_S^2)/\beta_\zeta D 0 , ζ = ( ν n 2 + ν S 2 ) / β ζ ,
T U ∑ n ν n 2 Λ ζ ( ν n ) D 0 , ζ ( ν n ) k = β ζ k ∑ j g j , ζ 2 Ω j B k ( Ω j ) . (262) T_U\sum_n\frac{\nu_n^2\Lambda_\zeta(\nu_n)}{D_{0,\zeta}(\nu_n)^k}=\beta_\zeta^k\sum_j\frac{g_{j,\zeta}^2}{\Omega_j}B_k(\Omega_j). \tag*{(262)} T U n ∑ D 0 , ζ ( ν n ) k ν n 2 Λ ζ ( ν n ) = β ζ k j ∑ Ω j g j , ζ 2 B k ( Ω j ) . (262)
The elementary Matsubara identity
T U ∑ n 1 ν n 2 + x 2 = 1 2 x coth x 2 T U (263) T_U\sum_n\frac{1}{\nu_n^2+x^2}=\frac{1}{2x}\coth\frac{x}{2T_U} \tag*{(263)} T U n ∑ ν n 2 + x 2 1 = 2 x 1 coth 2 T U x (263)
gives
B 1 ( Ω ) = Ω coth [ Ω / ( 2 T U ) ] − ν S c U 2 ( Ω 2 − ν S 2 ) , c U = coth ν S 2 T U . (264) B_1(\Omega)=\frac{\Omega\coth[\Omega/(2T_U)]-\nu_S c_U}{2(\Omega^2-\nu_S^2)},\qquad c_U=\coth\frac{\nu_S}{2T_U}. \tag*{(264)} B 1 ( Ω ) = 2 ( Ω 2 − ν S 2 ) Ω coth [ Ω/ ( 2 T U )] − ν S c U , c U = coth 2 T U ν S . (264)
The second kernel follows by differentiation,
B 2 ( Ω ) = − 1 2 ν S ∂ B 1 ( Ω ) ∂ ν S , (265) B_2(\Omega)=-\frac{1}{2\nu_S}\frac{\partial B_1(\Omega)}{\partial\nu_S}, \tag*{(265)} B 2 ( Ω ) = − 2 ν S 1 ∂ ν S ∂ B 1 ( Ω ) , (265)
at fixed temperature and bath frequency. Both kernels have finite limits at Ω = ν S \Omega=\nu_S Ω = ν S , reproducing Eq. (125) .
The role of these kernels becomes clear from the first-order covariance corrections:
δ Q ζ = − β ζ 2 ∑ j g j , ζ 2 Ω j B 2 ( Ω j ) + O ( κ 4 ) , \delta Q_\zeta=-\beta_\zeta^2\sum_j\frac{g_{j,\zeta}^2}{\Omega_j}B_2(\Omega_j)+O(\kappa^4), δ Q ζ = − β ζ 2 j ∑ Ω j g j , ζ 2 B 2 ( Ω j ) + O ( κ 4 ) ,
δ P ζ = α ζ β ζ δ Q ζ + ∑ j g j , ζ 2 Ω j B 1 ( Ω j ) + O ( κ 4 ) . (266) \delta P_\zeta=\frac{\alpha_\zeta}{\beta_\zeta}\delta Q_\zeta+\sum_j\frac{g_{j,\zeta}^2}{\Omega_j}B_1(\Omega_j)+O(\kappa^4). \tag*{(266)} δ P ζ = β ζ α ζ δ Q ζ + j ∑ Ω j g j , ζ 2 B 1 ( Ω j ) + O ( κ 4 ) . (266)
Thus B 2 B_2 B 2 enters the individual covariance corrections, whereas the difference relevant to Gaussian work is
δ P ζ − α ζ β ζ δ Q ζ = ∑ j g j , ζ 2 Ω j B 1 ( Ω j ) + O ( κ 4 ) . (267) \delta P_\zeta-\frac{\alpha_\zeta}{\beta_\zeta}\delta Q_\zeta=\sum_j\frac{g_{j,\zeta}^2}{\Omega_j}B_1(\Omega_j)+O(\kappa^4). \tag*{(267)} δ P ζ − β ζ α ζ δ Q ζ = j ∑ Ω j g j , ζ 2 B 1 ( Ω j ) + O ( κ 4 ) . (267)
The B 2 B_2 B 2 contribution cancels from this difference.
To see why this cancellation matters, introduce X ζ = α ζ G ζ e q X_\zeta=\alpha_\zeta G_\zeta^{\mathrm{eq}} X ζ = α ζ G ζ eq and Y ζ = β ζ P ζ e q Y_\zeta=\beta_\zeta P_\zeta^{\mathrm{eq}} Y ζ = β ζ P ζ eq . The Gaussian work of one collective sector can be expressed as
G ζ e q = ( X ζ − Y ζ ) 2 2 ( X ζ + Y ζ + 2 X ζ Y ζ ) . (268) \mathcal{G}_\zeta^{\mathrm{eq}}=\frac{(X_\zeta-Y_\zeta)^2}{2(X_\zeta+Y_\zeta+2\sqrt{X_\zeta Y_\zeta})}. \tag*{(268)} G ζ eq = 2 ( X ζ + Y ζ + 2 X ζ Y ζ ) ( X ζ − Y ζ ) 2 . (268)
At zero detector–bath coupling, X ζ = Y ζ = ν S c U / 2 X_\zeta=Y_\zeta=\nu_Sc_U/2 X ζ = Y ζ = ν S c U /2 . Equation (B18) shows that Y ζ − X ζ Y_\zeta-X_\zeta Y ζ − X ζ is of order κ 2 \kappa^2 κ 2 . Consequently, the leading Gaussian work is of order κ 4 \kappa^4 κ 4 .
Moreover, its leading coefficient depends only on the B 1 B_1 B 1 integral. In the continuum limit,
∑ j g j , ζ 2 Ω j B 1 ( Ω j ) ⟶ κ 2 2 π 2 ∫ 0 ∞ d Ω r ( Ω ) [ 1 ± γ E ( Ω ) ] B 1 ( Ω ) . (269) \sum_j\frac{g_{j,\zeta}^2}{\Omega_j}B_1(\Omega_j)
\longrightarrow\frac{\kappa^2}{2\pi^2}\int_0^\infty d\Omega\,r(\Omega)[1\pm\gamma_E(\Omega)]B_1(\Omega). \tag*{(269)} j ∑ Ω j g j , ζ 2 B 1 ( Ω j ) ⟶ 2 π 2 κ 2 ∫ 0 ∞ d Ω r ( Ω ) [ 1 ± γ E ( Ω )] B 1 ( Ω ) . (269)
This is the response-weighted spectral integral that leads to the work coefficient ℵ E \aleph_E ℵ E in Eq. (143) . The bath integration therefore provides the connection between the microscopic environmental spectra and the equilibrium Gaussian work used in the main text.
Continuum relaxation and the dark mode
The preceding calculation determines the reduced Gibbs covariance of the interacting detector–bath Hamiltonian. A separate question is whether the detectors actually approach this covariance after the interaction is switched on. This requires examining the real-time response of the coupled system.
For a finite number of bath oscillators, the total system is closed and its dynamics is generally quasiperiodic. The continuum reservoir is therefore essential for irreversible local relaxation.
To examine this limit, we eliminate the bath coordinates from the real-time equations of motion. The retarded susceptibility of one collective sector is
χ ζ − 1 ( z ) = α ζ − z 2 β ζ + δ ω ζ − 2 π ∫ 0 ∞ d Ω Ω J E ζ ( Ω ) Ω 2 − z 2 , (270) \chi_\zeta^{-1}(z)=\alpha_\zeta-\frac{z^2}{\beta_\zeta}+\delta\omega_\zeta-\frac{2}{\pi}\int_0^\infty d\Omega\,\frac{\Omega J_{E_\zeta}(\Omega)}{\Omega^2-z^2}, \tag*{(270)} χ ζ − 1 ( z ) = α ζ − β ζ z 2 + δ ω ζ − π 2 ∫ 0 ∞ d Ω Ω 2 − z 2 Ω J E ζ ( Ω ) , (270)
where z z z lies in the upper half-plane.
Under full static compensation, the zero-frequency contribution of the bath cancels the counterterm, so that
χ ζ − 1 ( 0 ) = α ζ > 0. (271) \chi_\zeta^{-1}(0)=\alpha_\zeta>0. \tag*{(271)} χ ζ − 1 ( 0 ) = α ζ > 0. (271)
For positive real frequencies, the boundary value satisfies
Im χ ζ − 1 ( x + i 0 + ) = − J E ζ ( x ) , (272) \operatorname{Im}\chi_\zeta^{-1}(x+\mathrm{i}0^+)=-J_{E_\zeta}(x), \tag*{(272)} Im χ ζ − 1 ( x + i 0 + ) = − J E ζ ( x ) , (272)
and hence
Im χ ζ ( x + i 0 + ) = J E ζ ( x ) [ Re χ ζ − 1 ( x ) ] 2 + J E ζ ( x ) 2 . (273) \operatorname{Im}\chi_\zeta(x+\mathrm{i}0^+)=\frac{J_{E_\zeta}(x)}{[\operatorname{Re}\chi_\zeta^{-1}(x)]^2+J_{E_\zeta}(x)^2}. \tag*{(273)} Im χ ζ ( x + i 0 + ) = [ Re χ ζ − 1 ( x ) ] 2 + J E ζ ( x ) 2 J E ζ ( x ) . (273)
For nonzero separation d > 0 d>0 d > 0 and nonzero detector–bath coupling, both collective spectral densities are positive at every x > 0 x>0 x > 0 in the environments considered here. The response therefore has no isolated positive-frequency pole. The zero-frequency condition above, together with stability of the completed quadratic Hamiltonian, also excludes zero-frequency and unstable modes.
Under these conditions, the response contains no undamped discrete-mode contribution. Moreover, the terms in the detector covariance that depend on the initial detector state and the finite switching history are Fourier integrals over the regulated spectral functions. Since the relevant spectral weights are integrable, these terms vanish as τ → ∞ \tau\to\infty τ → ∞ by the Riemann–Lebesgue lemma. The loss of dependence on the initial preparation is the standard relaxation behavior of a damped oscillator coupled to a continuum bath [80 , 105 ] . The remaining stationary covariance is therefore determined by the bath fluctuations:
Q ζ e q = 1 π ∫ 0 ∞ d x coth x 2 T U Im χ ζ ( x + i 0 + ) , Q_\zeta^{\mathrm{eq}}=\frac{1}{\pi}\int_0^\infty dx\,\coth\frac{x}{2T_U}\operatorname{Im}\chi_\zeta(x+i0^+), Q ζ eq = π 1 ∫ 0 ∞ d x coth 2 T U x Im χ ζ ( x + i 0 + ) ,
P ζ e q = 1 π β ζ 2 ∫ 0 ∞ d x x 2 coth x 2 T U Im χ ζ ( x + i 0 + ) . (274) P_\zeta^{\mathrm{eq}}=\frac{1}{\pi\beta_\zeta^2}\int_0^\infty dx\,x^2\coth\frac{x}{2T_U}\operatorname{Im}\chi_\zeta(x+i0^+). \tag*{(274)} P ζ eq = π β ζ 2 1 ∫ 0 ∞ d x x 2 coth 2 T U x Im χ ζ ( x + i 0 + ) . (274)
Analytic continuation to imaginary frequency recovers the equilibrium Matsubara expressions in Eqs. (111) and (117) . This establishes the connection between the reduced Gibbs covariance and the long-time detector covariance in the regular continuum-reservoir limit.
There is an important exception at zero detector separation. When d = 0 d=0 d = 0 , the antisymmetric spectral density vanishes for both the accelerated-vacuum and matched thermal environments:
J E − ( Ω ) = 0 , E = U , T . (275) J_{E-}(\Omega)=0,\qquad E=U,T. \tag*{(275)} J E − ( Ω ) = 0 , E = U , T . (275)
The antisymmetric collective mode is then decoupled from the bath. Its retarded response retains isolated poles at ± ν S \pm\nu_S ± ν S , so it does not undergo bath-induced relaxation.
Correlation ordering and uniqueness of the blind separation
The main text shows that the weighted spatial correlations in the accelerated-vacuum and matched thermal environments are different and that their work contrast can vanish at a finite detector separation. Here we establish these results analytically. We first prove the ordering of the two correlations. We then show that both decrease strictly with separation, which allows us to prove the uniqueness of the finite blind separation and determine the ordering of the two work minima.
Throughout this section, the temperature T = T U > 0 T=T_U>0 T = T U > 0 , the detector parameters, and the regulator are fixed, while the separation d d d varies.
Weighted spatial correlations and their ordering
We begin by establishing the positivity and ordering of the weighted spatial correlations. The key step is to show that the function
f ( Ω ) = r ( Ω ) B 1 ( Ω ) f(\Omega)=r(\Omega)B_1(\Omega) f ( Ω ) = r ( Ω ) B 1 ( Ω )
has a strictly positive cosine transform. We therefore introduce
f ( Ω ) = r ( Ω ) B 1 ( Ω ) , f(\Omega)=r(\Omega)B_1(\Omega), f ( Ω ) = r ( Ω ) B 1 ( Ω ) ,
h ( x ) = ∫ 0 ∞ d Ω f ( Ω ) cos ( Ω x ) , h(x)=\int_0^\infty d\Omega\,f(\Omega)\cos(\Omega x), h ( x ) = ∫ 0 ∞ d Ω f ( Ω ) cos ( Ω x ) ,
H ( x ) = ∫ 0 x h ( y ) d y . (276) H(x)=\int_0^x h(y)\,dy. \tag*{(276)} H ( x ) = ∫ 0 x h ( y ) d y . (276)
Here B 1 ( Ω ) B_1(\Omega) B 1 ( Ω ) is the positive kernel defined in Eq. (125) , and r ( Ω ) r(\Omega) r ( Ω ) is the regulator. We also write
I 0 = ∫ 0 ∞ f ( Ω ) d Ω . I_0=\int_0^\infty f(\Omega)\,d\Omega. I 0 = ∫ 0 ∞ f ( Ω ) d Ω.
The regulated integrals satisfy ∫ 0 ∞ f ( Ω ) d Ω < ∞ \int_0^\infty f(\Omega)\,d\Omega<\infty ∫ 0 ∞ f ( Ω ) d Ω < ∞ and ∫ 0 ∞ Ω f ( Ω ) d Ω < ∞ \int_0^\infty\Omega f(\Omega)\,d\Omega<\infty ∫ 0 ∞ Ω f ( Ω ) d Ω < ∞ .
To determine the sign of h ( x ) h(x) h ( x ) , we use the Matsubara representation of B 1 B_1 B 1 :
B 1 ( Ω ) = 2 T ∑ n = 1 ∞ ν n 2 ν n 2 + ν S 2 1 Ω 2 + ν n 2 . (277) B_1(\Omega)=2T\sum_{n=1}^{\infty}\frac{\nu_n^2}{\nu_n^2+\nu_S^2}\frac{1}{\Omega^2+\nu_n^2}. \tag*{(277)} B 1 ( Ω ) = 2 T n = 1 ∑ ∞ ν n 2 + ν S 2 ν n 2 Ω 2 + ν n 2 1 . (277)
Every term in this representation is positive. Moreover, the regulated sum is integrable, since
2 T ∑ n = 1 ∞ ν n 2 ν n 2 + ν S 2 ∫ 0 ∞ r ( Ω ) Ω 2 + ν n 2 d Ω = I 0 < ∞ . (278) 2T\sum_{n=1}^{\infty}\frac{\nu_n^2}{\nu_n^2+\nu_S^2}\int_0^\infty\frac{r(\Omega)}{\Omega^2+\nu_n^2}\,d\Omega=I_0<\infty. \tag*{(278)} 2 T n = 1 ∑ ∞ ν n 2 + ν S 2 ν n 2 ∫ 0 ∞ Ω 2 + ν n 2 r ( Ω ) d Ω = I 0 < ∞. (278)
This justifies interchanging the frequency integral and the Matsubara sum.
We now use the properties of the two regulators. Extend r ( Ω ) r(\Omega) r ( Ω ) evenly to the whole frequency axis and denote its cosine transform by
R ( x ) = ∫ 0 ∞ r ( Ω ) cos ( Ω x ) d Ω . R(x)=\int_0^\infty r(\Omega)\cos(\Omega x)\,d\Omega. R ( x ) = ∫ 0 ∞ r ( Ω ) cos ( Ω x ) d Ω.
For the exponential and Gaussian regulators, respectively, direct integration gives
R exp ( x ) = p p 2 + x 2 , p = Ω c − 1 , R_{\exp}(x)=\frac{p}{p^2+x^2},\qquad p=\Omega_c^{-1}, R e x p ( x ) = p 2 + x 2 p , p = Ω c − 1 ,
R G ( x ) = Ω c exp ( − Ω c 2 x 2 π ) . (279) R_G(x)=\Omega_c\exp\left(-\frac{\Omega_c^2x^2}{\pi}\right). \tag*{(279)} R G ( x ) = Ω c exp ( − π Ω c 2 x 2 ) . (279)
Both transforms are positive, even, integrable, and strictly decreasing for x > 0 x>0 x > 0 .
The remaining frequency integral can be evaluated using
∫ 0 ∞ cos ( Ω x ) Ω 2 + ν 2 d Ω = π 2 ν e − ν ∣ x ∣ , ν > 0. (280) \int_0^\infty\frac{\cos(\Omega x)}{\Omega^2+\nu^2}\,d\Omega=\frac{\pi}{2\nu}e^{-\nu|x|},\qquad\nu>0. \tag*{(280)} ∫ 0 ∞ Ω 2 + ν 2 cos ( Ω x ) d Ω = 2 ν π e − ν ∣ x ∣ , ν > 0. (280)
Defining K ν ( x ) = e − ν ∣ x ∣ K_\nu(x)=e^{-\nu|x|} K ν ( x ) = e − ν ∣ x ∣ , the convolution theorem for the even extensions gives
h ( x ) = T ∑ n = 1 ∞ ν n ν n 2 + ν S 2 ( R ∗ K ν n ) ( x ) . (281) h(x)=T\sum_{n=1}^{\infty}\frac{\nu_n}{\nu_n^2+\nu_S^2}(R*K_{\nu_n})(x). \tag*{(281)} h ( x ) = T n = 1 ∑ ∞ ν n 2 + ν S 2 ν n ( R ∗ K ν n ) ( x ) . (281)
where
( R ∗ K ) ( x ) = ∫ − ∞ ∞ R ( x − y ) K ( y ) d y . (R \ast K)(x) = \int_{-\infty}^{\infty} R(x-y)K(y)\,\mathrm{d}y. ( R ∗ K ) ( x ) = ∫ − ∞ ∞ R ( x − y ) K ( y ) d y .
Since both R R R and K ν n K_{\nu n} K ν n are strictly positive, every convolution in Eq. (281) is positive. All coefficients in the Matsubara sum are also positive. Consequently,
h ( x ) > 0 , H ( x ) > 0 ( x > 0 ) . (282) h(x)>0,\qquad H(x)>0\qquad(x>0). \tag*{(282)} h ( x ) > 0 , H ( x ) > 0 ( x > 0 ) . (282)
We can now express the spatial correlation integrals in terms of H H H . Integrating the cosine transform over the finite interval 0 ≤ y ≤ x 0\le y\le x 0 ≤ y ≤ x gives
H ( x ) = ∫ 0 ∞ d Ω f ( Ω ) sin ( Ω x ) Ω . (283) H(x)=\int_{0}^{\infty}\mathrm{d}\Omega\,f(\Omega)\frac{\sin(\Omega x)}{\Omega}. \tag*{(283)} H ( x ) = ∫ 0 ∞ d Ω f ( Ω ) Ω sin ( Ω x ) . (283)
The interchange of the integrals is justified by the regulated convergence established above. Using the geometric relations between the two environments, we obtain
I c , T = H ( d ) d , I c , U = H ( L ) D , (284) I_{c,T}=\frac{H(d)}{d},\qquad I_{c,U}=\frac{H(L)}{D}, \tag*{(284)} I c , T = d H ( d ) , I c , U = D H ( L ) , (284)
where
L = 2 a arsinh a d 2 , D = d 1 + ( a d 2 ) 2 . L=\frac{2}{a}\operatorname{arsinh}\frac{ad}{2},\qquad D=d\sqrt{1+\left(\frac{ad}{2}\right)^2}. L = a 2 arsinh 2 a d , D = d 1 + ( 2 a d ) 2 .
For a , d > 0 a,d>0 a , d > 0 , these geometric quantities satisfy
Since H ( x ) H(x) H ( x ) is strictly increasing, we have
H ( L ) < H ( d ) . H(L)<H(d). H ( L ) < H ( d ) .
Together with D > d D>d D > d , this yields
0 < I c , U < I c , T . (285) 0<I_{c,U}<I_{c,T}. \tag*{(285)} 0 < I c , U < I c , T . (285)
In addition, the thermal cross integral can be written as
I c , T = ∫ 0 ∞ d Ω f ( Ω ) sinc ( Ω d ) , I_{c,T}=\int_{0}^{\infty}\mathrm{d}\Omega\,f(\Omega)\operatorname{sinc}(\Omega d), I c , T = ∫ 0 ∞ d Ω f ( Ω ) sinc ( Ω d ) ,
where sinc ( x ) = sin x / x \operatorname{sinc}(x)=\sin x/x sinc ( x ) = sin x / x . Since sinc ( Ω d ) < 1 \operatorname{sinc}(\Omega d)<1 sinc ( Ω d ) < 1 for Ω , d > 0 \Omega,d>0 Ω , d > 0 and f ( Ω ) > 0 f(\Omega)>0 f ( Ω ) > 0 , it follows that I c , T < I 0 I_{c,T}<I_0 I c , T < I 0 .
Using the definition η E = I c , E / I 0 \eta_E=I_{c,E}/I_0 η E = I c , E / I 0 , we therefore obtain
0 < η U < η T < 1 , d > 0. (286) 0<\eta_U<\eta_T<1,\qquad d>0. \tag*{(286)} 0 < η U < η T < 1 , d > 0. (286)
This establishes the correlation ordering in Eq. (159) .
Monotonicity and uniqueness of the blind separation
The ordering established above shows that the accelerated-vacuum correlation is always smaller than the thermal one. To prove that the blind separation is unique, we need a further result: both correlations must decrease strictly with d d d .
We first show that h ( x ) h(x) h ( x ) is strictly decreasing. The two regulator transforms in Eq. (279) are even and strictly decreasing for positive arguments. For K ν ( x ) = e − ν ∣ x ∣ K_\nu(x)=e^{-\nu|x|} K ν ( x ) = e − ν ∣ x ∣ , differentiation of the convolution gives, for x > 0 x>0 x > 0 ,
( R ∗ K ν ) ′ ( x ) = ∫ − ∞ ∞ R ( x − y ) K ν ′ ( y ) d y = ∫ 0 ∞ K ν ′ ( y ) [ R ( x − y ) − R ( x + y ) ] d y . (287) (R\ast K_\nu)'(x)=\int_{-\infty}^{\infty}R(x-y)K_\nu'(y)\,\mathrm{d}y
=\int_{0}^{\infty}K_\nu'(y)[R(x-y)-R(x+y)]\,\mathrm{d}y. \tag*{(287)} ( R ∗ K ν ) ′ ( x ) = ∫ − ∞ ∞ R ( x − y ) K ν ′ ( y ) d y = ∫ 0 ∞ K ν ′ ( y ) [ R ( x − y ) − R ( x + y )] d y . (287)
The second equality follows by splitting the integral into positive and negative y y y and using the evenness of K ν K_\nu K ν and R R R .
For y > 0 y>0 y > 0 , we have
K ν ′ ( y ) = − ν e − ν y < 0. K_\nu'(y)=-\nu e^{-\nu y}<0. K ν ′ ( y ) = − ν e − ν y < 0.
Furthermore, since ∣ x − y ∣ < x + y |x-y|<x+y ∣ x − y ∣ < x + y and R R R is even and strictly decreasing with the absolute value of its argument,
R ( x − y ) > R ( x + y ) . R(x-y)>R(x+y). R ( x − y ) > R ( x + y ) .
Thus the integrand in the last line of Eq. (287) is strictly negative, giving
( R ∗ K ν ) ′ ( x ) < 0 , x > 0. (288) (R\ast K_\nu)'(x)<0,\qquad x>0. \tag*{(288)} ( R ∗ K ν ) ′ ( x ) < 0 , x > 0. (288)
The regulator also ensures that ∫ 0 ∞ Ω f ( Ω ) d Ω < ∞ \int_{0}^{\infty}\Omega f(\Omega)\,\mathrm{d}\Omega<\infty ∫ 0 ∞ Ω f ( Ω ) d Ω < ∞ . Consequently, the differentiated Matsubara series in Eq. (281) converges uniformly: the sum of the absolute bounds on its differentiated terms is bounded by ∫ 0 ∞ Ω f ( Ω ) d Ω \int_{0}^{\infty}\Omega f(\Omega)\,\mathrm{d}\Omega ∫ 0 ∞ Ω f ( Ω ) d Ω . We may therefore differentiate the series term by term. Since all its coefficients are positive, we conclude that
h ′ ( x ) < 0 , x > 0. (289) h'(x)<0,\qquad x>0. \tag*{(289)} h ′ ( x ) < 0 , x > 0. (289)
We next show that the normalized function
F ( x ) = H ( x ) I 0 x (290) F(x)=\frac{H(x)}{I_0x} \tag*{(290)} F ( x ) = I 0 x H ( x ) (290)
is also strictly decreasing. Because h ( x ) h(x) h ( x ) is strictly decreasing,
H ( x ) = ∫ 0 x h ( y ) d y > x h ( x ) . H(x)=\int_{0}^{x}h(y)\,\mathrm{d}y>xh(x). H ( x ) = ∫ 0 x h ( y ) d y > x h ( x ) .
It follows that
F ′ ( x ) = x h ( x ) − H ( x ) I 0 x 2 < 0. (291) F'(x)=\frac{xh(x)-H(x)}{I_0x^2}<0. \tag*{(291)} F ′ ( x ) = I 0 x 2 x h ( x ) − H ( x ) < 0. (291)
The two weighted correlations can now be written in the compact form
η T ( d ) = F ( d ) , η U ( d ) = b ( d ) F ( L ( d ) ) , b ( d ) = L ( d ) D ( d ) . (292) \eta_T(d)=F(d),\qquad\eta_U(d)=b(d)F(L(d)),\qquad b(d)=\frac{L(d)}{D(d)}. \tag*{(292)} η T ( d ) = F ( d ) , η U ( d ) = b ( d ) F ( L ( d )) , b ( d ) = D ( d ) L ( d ) . (292)
Since L ′ ( d ) > 0 L'(d)>0 L ′ ( d ) > 0 , the monotonicity of η U \eta_U η U also depends on the geometric factor b ( d ) b(d) b ( d ) . To determine its sign, introduce
u = arsinh a d 2 > 0. u=\operatorname{arsinh}\frac{ad}{2}>0. u = arsinh 2 a d > 0.
Then
b = 2 u sinh ( 2 u ) . b=\frac{2u}{\sinh(2u)}. b = sinh ( 2 u ) 2 u .
and
d b d u = 2 sinh ( 2 u ) − 4 u cosh ( 2 u ) sinh 2 ( 2 u ) < 0. (293) \frac{db}{du}=\frac{2\sinh(2u)-4u\cosh(2u)}{\sinh^2(2u)}<0. \tag*{(293)} d u d b = sinh 2 ( 2 u ) 2 sinh ( 2 u ) − 4 u cosh ( 2 u ) < 0. (293)
Indeed, the numerator has the opposite sign to 2 u cosh ( 2 u ) − sinh ( 2 u ) 2u\cosh(2u)-\sinh(2u) 2 u cosh ( 2 u ) − sinh ( 2 u ) . The latter vanishes at u = 0 u=0 u = 0 and has derivative 4 u sinh ( 2 u ) > 0 4u\sinh(2u)>0 4 u sinh ( 2 u ) > 0 for u > 0 u>0 u > 0 . Therefore b ′ ( d ) < 0 b'(d)<0 b ′ ( d ) < 0 . Together with F > 0 F>0 F > 0 , F ′ < 0 F'<0 F ′ < 0 , and L ′ ( d ) > 0 L'(d)>0 L ′ ( d ) > 0 , we obtain
η T ′ ( d ) = F ′ ( d ) < 0 , \eta'_T(d)=F'(d)<0, η T ′ ( d ) = F ′ ( d ) < 0 ,
η U ′ ( d ) = b ′ ( d ) F ( L ) + b ( d ) F ′ ( L ) L ′ ( d ) < 0. (294) \eta'_U(d)=b'(d)F(L)+b(d)F'(L)L'(d)<0. \tag*{(294)} η U ′ ( d ) = b ′ ( d ) F ( L ) + b ( d ) F ′ ( L ) L ′ ( d ) < 0. (294)
Thus both weighted spatial correlations decrease strictly with the detector separation.
Their endpoint limits follow directly from the integral representations. At coincidence, continuity of h h h gives
lim x → 0 + H ( x ) x = h ( 0 ) = I 0 , \lim_{x\to0^+}\frac{H(x)}{x}=h(0)=I_0, x → 0 + lim x H ( x ) = h ( 0 ) = I 0 ,
while b ( d ) → 1 b(d)\to1 b ( d ) → 1 as d → 0 + d\to0^+ d → 0 + . Hence
lim d → 0 + η U ( d ) = lim d → 0 + η T ( d ) = 1. \lim_{d\to0^+}\eta_U(d)=\lim_{d\to0^+}\eta_T(d)=1. d → 0 + lim η U ( d ) = d → 0 + lim η T ( d ) = 1.
At large separation, the geometric factors sinc ( Ω d ) \operatorname{sinc}(\Omega d) sinc ( Ω d ) and b ( d ) sinc ( Ω L ( d ) ) b(d)\operatorname{sinc}(\Omega L(d)) b ( d ) sinc ( Ω L ( d )) tend to zero for every fixed Ω > 0 \Omega>0 Ω > 0 . Their absolute values are bounded by one. Since f f f is integrable, dominated convergence then yields
lim d → ∞ η U ( d ) = lim d → ∞ η T ( d ) = 0. \lim_{d\to\infty}\eta_U(d)=\lim_{d\to\infty}\eta_T(d)=0. d → ∞ lim η U ( d ) = d → ∞ lim η T ( d ) = 0.
We can now establish the uniqueness of the finite blind separation. According to Eq. (158) , the work contrast factorizes as
N U − N T = N C ( η U − η T ) ( η U + η T − s ) . (295) \mathcal{N}_U-\mathcal{N}_T=\mathcal{N}_C(\eta_U-\eta_T)(\eta_U+\eta_T-s). \tag*{(295)} N U − N T = N C ( η U − η T ) ( η U + η T − s ) . (295)
For every finite d > 0 d>0 d > 0 , the ordering proved above ensures η U − η T < 0 \eta_U-\eta_T<0 η U − η T < 0 . Thus the finite blind separation is determined by
η U ( d 0 ) + η T ( d 0 ) = s . (296) \eta_U(d_0)+\eta_T(d_0)=s. \tag*{(296)} η U ( d 0 ) + η T ( d 0 ) = s . (296)
For 0 < λ < ω 0<\lambda<\omega 0 < λ < ω , we have 0 < s < 2 0<s<2 0 < s < 2 . The sum η U + η T \eta_U+\eta_T η U + η T is continuous and strictly decreasing, with limiting values 2 2 2 at d = 0 d=0 d = 0 and 0 0 0 as d → ∞ d\to\infty d → ∞ . It therefore crosses the value s s s exactly once. This proves the existence and uniqueness of the finite blind separation d 0 d_0 d 0 .
The crossing is also simple. Differentiating Eq. (295) and using Eq. (296) , we obtain
∂ d ( N U − N T ) ∣ d = d 0 = N C ( η U − η T ) ( η U ′ + η T ′ ) ∣ d = d 0 > 0. (297) \left.\partial_d(\mathcal{N}_U-\mathcal{N}_T)\right|_{d=d_0}
=\left.\mathcal{N}_C(\eta_U-\eta_T)(\eta'_U+\eta'_T)\right|_{d=d_0}>0. \tag*{(297)} ∂ d ( N U − N T ) ∣ d = d 0 = N C ( η U − η T ) ( η U ′ + η T ′ ) ∣ d = d 0 > 0. (297)
Both factors in parentheses are strictly negative, so the derivative is positive. The two work curves therefore cross with a nonzero slope difference at d 0 d_0 d 0 . The coincident-detector limit d = 0 d=0 d = 0 is a separate boundary case, where η U = η T = 1 \eta_U=\eta_T=1 η U = η T = 1 and the contrast also vanishes. It is not a finite blind separation.
Finally, we determine the ordering of the two work minima. From the dependence of R E R_E R E on η E \eta_E η E , its derivative is
R E ′ = ( 2 η E − s ) η E ′ . (298) R'_E=(2\eta_E-s)\eta'_E. \tag*{(298)} R E ′ = ( 2 η E − s ) η E ′ . (298)
Since each η E \eta_E η E decreases continuously from 1 1 1 to 0 0 0 , the corresponding R E R_E R E decreases until η E = s / 2 \eta_E=s/2 η E = s /2 and increases thereafter. Both environments therefore attain their minimum at η E = s / 2 \eta_E=s/2 η E = s /2 . The minimum values coincide, as given in Eq. (157) , although the corresponding separations are different.
At the accelerated-vacuum minimum, η U = s / 2 \eta_U=s/2 η U = s /2 . Since η T > η U \eta_T>\eta_U η T > η U , we have η U + η T > s \eta_U+\eta_T>s η U + η T > s at that separation. The strictly decreasing sum η U + η T \eta_U+\eta_T η U + η T reaches s s s only at a larger separation. Conversely, at the thermal minimum, η T = s / 2 \eta_T=s/2 η T = s /2 and η U < η T \eta_U<\eta_T η U < η T , so that η U + η T < s \eta_U+\eta_T<s η U + η T < s . The blind separation must therefore lie between the two minima:
d min , U < d 0 < d min , T . (299) d_{\min,U}<d_0<d_{\min,T}. \tag*{(299)} d m i n , U < d 0 < d m i n , T . (299)
For λ ≤ 0 \lambda\leq0 λ ≤ 0 , we have s ≤ 0 s\leq0 s ≤ 0 . Since 0 < η U < η T 0<\eta_U<\eta_T 0 < η U < η T , their difference is negative, whereas their sum minus s s s is strictly positive. Together with N C > 0 \mathcal{N}_C>0 N C > 0 , Eq. (295) therefore gives
N U − N T < 0 ( d > 0 ) . \mathcal{N}_U-\mathcal{N}_T<0\qquad(d>0). N U − N T < 0 ( d > 0 ) .
Thus, the two work coefficients never become equal at any finite detector separation, and no finite blind separation exists for λ ≤ 0 \lambda\leq0 λ ≤ 0 .
Finite-coupling corrections to the equilibrium readout
The main text establishes a relation between the equilibrium Gaussian ergotropy and the weighted spatial correlation in the weak-coupling regime. This relation is obtained from the leading O ( κ 4 ) O(\kappa^4) O ( κ 4 ) contribution to the ergotropy. At finite coupling, however, higher-order terms also contribute to the work readout. In particular, these corrections depend on additional frequency integrals and are not determined solely by the weighted spatial correlation η E \eta_E η E . Consequently, finite coupling can modify both the separation at which the two environments give the same work and the correlation inferred from the two-setting reconstruction. To quantify these effects, we expand the exact equilibrium covariance to the next order. in κ 2 \kappa^2 κ 2 . This gives the O ( κ 6 ) O(\kappa^6) O ( κ 6 ) correction to the ergotropy, which we then use to calculate the shift of the blind separation and the leading finite-coupling bias in the reconstructed spatial correlation.
Sixth-order correction to the equilibrium Gaussian ergotropy
We first derive the next-order correction to the equilibrium Gaussian ergotropy. The starting point is the exact covariance obtained from the Green function in Sec. IV. To expand these covariances systematically, we introduce
t = κ 2 , z n = ν n 2 , D n = z n + ν S 2 . t=\kappa^2,\qquad z_n=\nu_n^2,\qquad D_n=z_n+\nu_S^2. t = κ 2 , z n = ν n 2 , D n = z n + ν S 2 .
For a fixed environment E E E and collective sector ζ \zeta ζ , define, for n ≠ 0 n\ne0 n = 0 ,
ℓ E ζ n = 1 2 π 2 ∫ 0 ∞ d Ω r ( Ω ) [ 1 ± γ E ( Ω ) ] z n + Ω 2 . (300) \ell_{E\zeta n}=\frac{1}{2\pi^2}\int_0^\infty d\Omega\,\frac{r(\Omega)[1\pm\gamma_E(\Omega)]}{z_n+\Omega^2}. \tag*{(300)} ℓ E ζ n = 2 π 2 1 ∫ 0 ∞ d Ω z n + Ω 2 r ( Ω ) [ 1 ± γ E ( Ω )] . (300)
The fully compensated self-energy is then
Σ E ζ ( ν n ) = t z n ℓ E ζ n . \Sigma_{E\zeta}(\nu_n)=t z_n\ell_{E\zeta n}. Σ E ζ ( ν n ) = t z n ℓ E ζ n .
In particular, its zero-frequency contribution vanishes. We therefore omit the n = 0 n=0 n = 0 term in the following sums.
For later use, introduce three Matsubara sums:
A ζ = T ∑ n ≠ 0 z n ℓ E ζ n D n = I E , ζ 2 π 2 , A_\zeta=T\sum_{n\ne0}\frac{z_n\ell_{E\zeta n}}{D_n}=\frac{I_{E,\zeta}}{2\pi^2}, A ζ = T n = 0 ∑ D n z n ℓ E ζ n = 2 π 2 I E , ζ ,
B ζ = T ∑ n ≠ 0 z n 2 ℓ E ζ n D n 2 , B_\zeta=T\sum_{n\ne0}\frac{z_n^2\ell_{E\zeta n}}{D_n^2}, B ζ = T n = 0 ∑ D n 2 z n 2 ℓ E ζ n ,
C ζ = T ∑ n ≠ 0 z n ℓ E ζ n ( z n − ν S 2 ) D n 2 . (301) C_\zeta=T\sum_{n\ne0}\frac{z_n\ell_{E\zeta n}(z_n-\nu_S^2)}{D_n^2}. \tag*{(301)} C ζ = T n = 0 ∑ D n 2 z n ℓ E ζ n ( z n − ν S 2 ) . (301)
Here the environment label E E E is suppressed in A ζ A_\zeta A ζ , B ζ B_\zeta B ζ , and C ζ C_\zeta C ζ . The first sum reproduces the response-weighted spectral integral introduced in the main text, whereas the other two retain additional information about the frequency dependence of the self-energy.
For a fixed regulator, z n ℓ E ζ n z_n\ell_{E\zeta n} z n ℓ E ζ n approaches a finite constant as ∣ n ∣ → ∞ |n|\to\infty ∣ n ∣ → ∞ . The summands in Eq. (301) therefore decay as n − 2 n^{-2} n − 2 , n − 4 n^{-4} n − 4 , and n − 2 n^{-2} n − 2 , respectively. All three sums are consequently convergent.
We now use these sums to expand the exact equilibrium covariances. Define
X ζ = α ζ Q ζ e q , Y ζ = β ζ P ζ e q , E 0 = ν S c U 2 , X_\zeta=\alpha_\zeta Q_\zeta^{\rm eq},\qquad Y_\zeta=\beta_\zeta P_\zeta^{\rm eq},\qquad E_0=\frac{\nu_S c_U}{2}, X ζ = α ζ Q ζ eq , Y ζ = β ζ P ζ eq , E 0 = 2 ν S c U ,
where E 0 E_0 E 0 is the common uncoupled value of X ζ X_\zeta X ζ and Y ζ Y_\zeta Y ζ , with c U = coth [ ν S / ( 2 T U ) ] c_U=\coth[\nu_S/(2T_U)] c U = coth [ ν S / ( 2 T U )] .
The exact covariance formulas and the stationary virial identity derived in Sec. IV give
Y ζ − X ζ = t β ζ T ∑ n ≠ 0 z n ℓ E ζ n D n + t β ζ z n ℓ E ζ n , Y_\zeta-X_\zeta=t\beta_\zeta T\sum_{n\ne0}\frac{z_n\ell_{E\zeta n}}{D_n+t\beta_\zeta z_n\ell_{E\zeta n}}, Y ζ − X ζ = t β ζ T n = 0 ∑ D n + t β ζ z n ℓ E ζ n z n ℓ E ζ n ,
X ζ − E 0 = − t ν S 2 β ζ T ∑ n ≠ 0 z n ℓ E ζ n D n ( D n + t β ζ z n ℓ E ζ n ) . (302) X_\zeta-E_0=-t\nu_S^2\beta_\zeta T\sum_{n\ne0}\frac{z_n\ell_{E\zeta n}}{D_n(D_n+t\beta_\zeta z_n\ell_{E\zeta n})}. \tag*{(302)} X ζ − E 0 = − t ν S 2 β ζ T n = 0 ∑ D n ( D n + t β ζ z n ℓ E ζ n ) z n ℓ E ζ n . (302)
These expressions show explicitly how the bath-induced self-energy modifies the two detector variances. In particular, the difference between the momentum and position contributions is directly determined by the first expression.
Expanding the denominators for sufficiently small t t t at fixed temperature, cutoff, and stable detector parameters, we obtain
δ ζ ≡ Y ζ − X ζ = t β ζ A ζ − t 2 β ζ 2 B ζ + O ( t 3 ) , \delta_\zeta\equiv Y_\zeta-X_\zeta=t\beta_\zeta A_\zeta-t^2\beta_\zeta^2B_\zeta+O(t^3), δ ζ ≡ Y ζ − X ζ = t β ζ A ζ − t 2 β ζ 2 B ζ + O ( t 3 ) ,
S ζ ≡ X ζ + Y ζ = 2 E 0 + t β ζ C ζ + O ( t 2 ) . (303) S_\zeta\equiv X_\zeta+Y_\zeta=2E_0+t\beta_\zeta C_\zeta+O(t^2). \tag*{(303)} S ζ ≡ X ζ + Y ζ = 2 E 0 + t β ζ C ζ + O ( t 2 ) . (303)
Thus, A ζ A_\zeta A ζ determines the leading difference between the two covariance contributions, while B ζ B_\zeta B ζ and C ζ C_\zeta C ζ enter the next-order correction.
We can now obtain the finite-coupling correction to the Gaussian ergotropy. Using the expression for one collective sector,
G ζ e q = X ζ + Y ζ 2 − X ζ Y ζ , (304) \mathcal{G}^{\rm eq}_\zeta=\frac{X_\zeta+Y_\zeta}{2}-\sqrt{X_\zeta Y_\zeta}, \tag*{(304)} G ζ eq = 2 X ζ + Y ζ − X ζ Y ζ , (304)
and the identity
X ζ Y ζ = S ζ 2 − δ ζ 2 4 , X_\zeta Y_\zeta=\frac{S_\zeta^2-\delta_\zeta^2}{4}, X ζ Y ζ = 4 S ζ 2 − δ ζ 2 ,
we rewrite it as
G ζ e q = S ζ 2 [ 1 − 1 − δ ζ 2 S ζ 2 ] = δ ζ 2 4 S ζ + O ( t 4 ) . (305) \begin{aligned}
\mathcal{G}^{\rm eq}_\zeta=\frac{S_\zeta}{2}\left[1-\sqrt{1-\frac{\delta_\zeta^2}{S_\zeta^2}}\right] \\
&=\frac{\delta_\zeta^2}{4S_\zeta}+O(t^4). \tag*{(305)}
\end{aligned} G ζ eq = 2 S ζ [ 1 − 1 − S ζ 2 δ ζ 2 ] = 4 S ζ δ ζ 2 + O ( t 4 ) . (305)
Since δ ζ = O ( t ) \delta_\zeta=O(t) δ ζ = O ( t ) , the leading ergotropy is of order t 2 = κ 4 t^2=\kappa^4 t 2 = κ 4 . Retaining the next order requires the O ( t 2 ) O(t^2) O ( t 2 ) correction to δ ζ \delta_\zeta δ ζ and the O ( t ) O(t) O ( t ) correction to S ζ S_\zeta S ζ . Substituting Eq. (303) , we find
G ζ e q = t 2 β ζ 2 A ζ 2 4 ν S c U − t 3 β ζ 3 [ A ζ B ζ 2 ν S c U + A ζ 2 C ζ 4 ν S 2 c U 2 ] + O ( t 4 ) . (306) \mathcal{G}^{\rm eq}_\zeta=\frac{t^2\beta_\zeta^2A_\zeta^2}{4\nu_Sc_U}-t^3\beta_\zeta^3\left[\frac{A_\zeta B_\zeta}{2\nu_Sc_U}+\frac{A_\zeta^2C_\zeta}{4\nu_S^2c_U^2}\right]+O(t^4). \tag*{(306)} G ζ eq = 4 ν S c U t 2 β ζ 2 A ζ 2 − t 3 β ζ 3 [ 2 ν S c U A ζ B ζ + 4 ν S 2 c U 2 A ζ 2 C ζ ] + O ( t 4 ) . (306)
Summing the two collective sectors gives the equilibrium ergotropy of the detector pair:
G E e q = κ 4 N E + κ 6 M E + O ( κ 8 ) , \mathcal{G}^{\rm eq}_E=\kappa^4\mathcal{N}_E+\kappa^6\mathcal{M}_E+O(\kappa^8), G E eq = κ 4 N E + κ 6 M E + O ( κ 8 ) ,
M E = − ∑ ζ β ζ 3 [ A ζ B ζ 2 ν S c U + A ζ 2 C ζ 4 ν S 2 c U 2 ] . (307) \mathcal{M}_E=-\sum_\zeta\beta_\zeta^3\left[\frac{A_\zeta B_\zeta}{2\nu_Sc_U}+\frac{A_\zeta^2C_\zeta}{4\nu_S^2c_U^2}\right]. \tag*{(307)} M E = − ζ ∑ β ζ 3 [ 2 ν S c U A ζ B ζ + 4 ν S 2 c U 2 A ζ 2 C ζ ] . (307)
The leading coefficient is
N E = ∑ ζ β ζ 2 A ζ 2 4 ν S c U , (308) \mathcal{N}_E=\sum_\zeta\frac{\beta_\zeta^2A_\zeta^2}{4\nu_Sc_U}, \tag*{(308)} N E = ζ ∑ 4 ν S c U β ζ 2 A ζ 2 , (308)
which reproduces Eq. (143) upon using A ζ = I E , ζ / ( 2 π 2 ) A_\zeta=I_{E,\zeta}/(2\pi^2) A ζ = I E , ζ / ( 2 π 2 ) and β ζ = ν S 2 / α ζ \beta_\zeta=\nu_S^2/\alpha_\zeta β ζ = ν S 2 / α ζ .
The coefficient M E \mathcal{M}_E M E is the first finite-coupling correction beyond the leading weak-coupling result. Unlike N E \mathcal{N}_E N E , it depends on the additional sums B ζ B_\zeta B ζ and C ζ C_\zeta C ζ .
Shift of the blind separation at finite coupling
We next use the sixth-order correction to determine how the blind separation changes when the detector–field coupling is finite. At leading order, the blind separation d 0 d_0 d 0 is determined by the equality of the two work coefficients,
Δ N ( d 0 ) = 0 , Δ N = N U − N T . \Delta\mathcal{N}(d_0)=0,\qquad\Delta\mathcal{N}=\mathcal{N}_{U}-\mathcal{N}_{T}. Δ N ( d 0 ) = 0 , Δ N = N U − N T .
At finite coupling, the corresponding work difference also contains the sixth-order contribution. Defining
Δ M ( d ) = M U ( d ) − M T ( d ) , \Delta\mathcal{M}(d)=\mathcal{M}_{U}(d)-\mathcal{M}_{T}(d), Δ M ( d ) = M U ( d ) − M T ( d ) ,
the equality of the two equilibrium ergotropies becomes
0 = Δ N ( d ) + t Δ M ( d ) + O ( t 2 ) , (309) 0=\Delta\mathcal{N}(d)+t\Delta\mathcal{M}(d)+O(t^2), \tag*{(309)} 0 = Δ N ( d ) + t Δ M ( d ) + O ( t 2 ) , (309)
after dividing the work difference by t 2 t^2 t 2 . The leading blind separation is a simple root, as established in Eq. (297) . We can therefore expand the finite-coupling solution around d 0 d_0 d 0 :
d 0 ( t ) = d 0 + t δ d + O ( t 2 ) . (310) d_0(t)=d_0+t\delta d+O(t^2). \tag*{(310)} d 0 ( t ) = d 0 + t δ d + O ( t 2 ) . (310)
Substituting this expression into Eq. (309) and retaining terms through first order in t t t gives
Δ N ′ ( d 0 ) δ d + Δ M ( d 0 ) = 0. \Delta\mathcal{N}'(d_0)\delta d+\Delta\mathcal{M}(d_0)=0. Δ N ′ ( d 0 ) δ d + Δ M ( d 0 ) = 0.
Hence the leading shift is
δ d = − Δ M ( d 0 ) Δ N ′ ( d 0 ) . (311) \delta d=-\frac{\Delta\mathcal{M}(d_0)}{\Delta\mathcal{N}'(d_0)}. \tag*{(311)} δ d = − Δ N ′ ( d 0 ) Δ M ( d 0 ) . (311)
Thus, the sixth-order correction determines the leading O ( κ 2 ) O(\kappa^2) O ( κ 2 ) displacement of the blind separation.
Finite-coupling correction to the two-setting reconstruction
Finally, we examine how finite coupling affects the reconstruction of the weighted spatial correlation from two separately prepared detector settings, λ = ± g \lambda=\pm g λ = ± g . For a fixed environment E E E , write the corresponding ergotropies as
G ± ( t ) = t 2 N ± + t 3 M ± + O ( t 4 ) , (312) \mathcal{G}_{\pm}(t)=t^2\mathcal{N}_{\pm}+t^3\mathcal{M}_{\pm}+O(t^4), \tag*{(312)} G ± ( t ) = t 2 N ± + t 3 M ± + O ( t 4 ) , (312)
where N ± \mathcal{N}_{\pm} N ± and M ± \mathcal{M}_{\pm} M ± are evaluated at λ = ± g \lambda=\pm g λ = ± g . The two-setting readout is defined by the normalized difference
A W ( t ) = G − ( t ) − G + ( t ) G − ( t ) + G + ( t ) . (313) \mathcal{A}_{W}(t)=\frac{\mathcal{G}_{-}(t)-\mathcal{G}_{+}(t)}{\mathcal{G}_{-}(t)+\mathcal{G}_{+}(t)}. \tag*{(313)} A W ( t ) = G − ( t ) + G + ( t ) G − ( t ) − G + ( t ) . (313)
At leading order, this reduces to
A E = N − − N + N − + N + . (314) \mathcal{A}_{E}=\frac{\mathcal{N}_{-}-\mathcal{N}_{+}}{\mathcal{N}_{-}+\mathcal{N}_{+}}. \tag*{(314)} A E = N − + N + N − − N + . (314)
Expanding the ratio to first order in t t t then gives
A W ( t ) = A E + b E t + O ( t 2 ) , b E = M − − M + − A E ( M − + M + ) N − + N + . (315) \mathcal{A}_{W}(t)=\mathcal{A}_{E}+b_{E}t+O(t^2),
\qquad
b_{E}=\frac{\mathcal{M}_{-}-\mathcal{M}_{+}-\mathcal{A}_{E}(\mathcal{M}_{-}+\mathcal{M}_{+})}{\mathcal{N}_{-}+\mathcal{N}_{+}}. \tag*{(315)} A W ( t ) = A E + b E t + O ( t 2 ) , b E = N − + N + M − − M + − A E ( M − + M + ) . (315)
The coefficient b E b_E b E therefore quantifies the leading finite-coupling correction to the two-setting readout. To make its effect on the reconstructed spatial correlation explicit, recall that the fully compensated leading-order relation in the main text is
A E = s g η E 1 + η E 2 , s g = 4 ω g ω 2 + g 2 . (316) \mathcal{A}_{E}=\frac{s_g\eta_E}{1+\eta_E^2},
\qquad
s_g=\frac{4\omega g}{\omega^2+g^2}. \tag*{(316)} A E = 1 + η E 2 s g η E , s g = ω 2 + g 2 4 ω g . (316)
The reconstruction based on this leading-order relation gives
η ^ E ( t ) = 2 A W ( t ) s g + s g 2 − 4 A W ( t ) 2 . (317) \widehat{\eta}_{E}(t)=\frac{2\mathcal{A}_{W}(t)}{s_g+\sqrt{s_g^2-4\mathcal{A}_{W}(t)^2}}. \tag*{(317)} η E ( t ) = s g + s g 2 − 4 A W ( t ) 2 2 A W ( t ) . (317)
Expanding around the true weighted correlation η E \eta_E η E yields
η ^ E ( t ) = η E + ( 1 + η E 2 ) 2 s g ( 1 − η E 2 ) b E t + O ( t 2 ) , 0 < η E < 1. (318) \widehat{\eta}_{E}(t)=\eta_E+\frac{(1+\eta_E^2)^2}{s_g(1-\eta_E^2)}b_Et+O(t^2),
\qquad0<\eta_E<1. \tag*{(318)} η E ( t ) = η E + s g ( 1 − η E 2 ) ( 1 + η E 2 ) 2 b E t + O ( t 2 ) , 0 < η E < 1. (318)
Thus, the leading bias in the reconstructed correlation is proportional to κ 2 \kappa^2 κ 2 , with its coefficient determined by the sixth-order corrections to the two work measurements. The inversion becomes increasingly sensitive to finite-coupling corrections as η E \eta_E η E approaches unity, where the derivative of the leading-order readout relation vanishes.
Partial static compensation
The main text establishes the relation between the equilibrium Gaussian ergotropy and the weighted spatial correlation η E \eta_E η E under full static compensation. This compensation removes the static shift induced by the detector–field coupling, allowing us to isolate the contribution of the frequency-dependent bath response. Our aim is to determine how this residual interaction modifies the relation between the extracted work and the underlying spatial correlation.
We first derive the effective Green function and the corresponding equilibrium covariance. We then obtain the leading work coefficient and identify how the residual static interaction modifies the relation between the end_mid? weighted spatial correlation and the extracted work. Finally, we separate the additional global work and examine the classical equilibrium limit.
The coupling to the field induces both local and cross static contributions to the detector potential. We retain the full local compensation and introduce ρ c t ∈ [ 0 , 1 ] \rho_{\mathrm{ct}} \in[0,1] ρ ct ∈ [ 0 , 1 ] to specify the fraction of the cross contribution included in the counterterm.
To make this distinction explicit, define
μ 0 = 1 2 π 2 ∫ 0 ∞ r ( Ω ) d Ω , \mu_0 = \frac{1}{2\pi^2}\int_0^\infty r(\Omega)\,\mathrm{d}\Omega, μ 0 = 2 π 2 1 ∫ 0 ∞ r ( Ω ) d Ω ,
μ c , E = 1 2 π 2 ∫ 0 ∞ r ( Ω ) γ E ( Ω ) d Ω . (319) \mu_{\mathrm{c},E} = \frac{1}{2\pi^2}\int_0^\infty r(\Omega)\gamma_E(\Omega)\,\mathrm{d}\Omega. \tag*{(319)} μ c , E = 2 π 2 1 ∫ 0 ∞ r ( Ω ) γ E ( Ω ) d Ω. (319)
Here, μ 0 \mu_0 μ 0 describes the local static contribution, whereas μ c , E \mu_{\mathrm{c},E} μ c , E describes the cross contribution and depends on the environment through γ E ( Ω ) \gamma_E(\Omega) γ E ( Ω ) .
Under full static compensation, the counterterm coefficient in collective sector ζ \zeta ζ is κ 2 ( μ 0 ± μ c , E ) \kappa^2(\mu_0 \pm\mu_{\mathrm{c},E}) κ 2 ( μ 0 ± μ c , E ) . With partial compensation, the corresponding coordinate coefficient in the total Hamiltonian becomes
α ζ + κ 2 ( μ 0 ± ρ c t μ c , E ) . (320) \alpha_\zeta+\kappa^2(\mu_0\pm\rho_{\mathrm{ct}}\mu_{\mathrm{c},E}). \tag*{(320)} α ζ + κ 2 ( μ 0 ± ρ ct μ c , E ) . (320)
Thus, ρ c t = 1 \rho_{\mathrm{ct}}=1 ρ ct = 1 corresponds to full compensation, while ρ c t = 0 \rho_{\mathrm{ct}}=0 ρ ct = 0 means that only the local static contribution is compensated.
After integrating out the bath coordinates, the bath-induced static shift κ 2 ( μ 0 + ζ μ c , E ) \kappa^2(\mu_0+\zeta\mu_{\mathrm{c},E}) κ 2 ( μ 0 + ζ μ c , E ) is subtracted. The remaining coordinate coefficient is therefore
α ζ + κ 2 h E ζ , (321) \alpha_\zeta+\kappa^2h_{E\zeta}, \tag*{(321)} α ζ + κ 2 h E ζ , (321)
where
h E , + = − ( 1 − ρ c t ) μ c , E , h E , − = + ( 1 − ρ c t ) μ c , E . (322) h_{E,+}=-(1-\rho_{\mathrm{ct}})\mu_{\mathrm{c},E},\qquad h_{E,-}=+(1-\rho_{\mathrm{ct}})\mu_{\mathrm{c},E}. \tag*{(322)} h E , + = − ( 1 − ρ ct ) μ c , E , h E , − = + ( 1 − ρ ct ) μ c , E . (322)
The residual interaction vanishes when ρ c t = 1 \rho_{\mathrm{ct}}=1 ρ ct = 1 . Throughout the following calculation, we require
α ζ + κ 2 h E ζ > 0 , β ζ > 0 , \alpha_\zeta+\kappa^2h_{E\zeta}>0,\qquad\beta_\zeta>0, α ζ + κ 2 h E ζ > 0 , β ζ > 0 ,
so that the quadratic Hamiltonian remains stable.
We now calculate the equilibrium covariance of this modified detector–field system. Put
t = κ 2 , z n = ν n 2 , D n = z n + ν S 2 , t=\kappa^2,\qquad z_n=\nu_n^2,\qquad D_n=z_n+\nu_S^2, t = κ 2 , z n = ν n 2 , D n = z n + ν S 2 ,
and introduce
ℓ E ζ n = 1 2 π 2 ∫ 0 ∞ r ( Ω ) [ 1 ± γ E ( Ω ) ] z n + Ω 2 d Ω . (323) \ell_{E\zeta n}=\frac{1}{2\pi^2}\int_0^\infty\frac{r(\Omega)[1\pm\gamma_E(\Omega)]}{z_n+\Omega^2}\,\mathrm{d}\Omega. \tag*{(323)} ℓ E ζ n = 2 π 2 1 ∫ 0 ∞ z n + Ω 2 r ( Ω ) [ 1 ± γ E ( Ω )] d Ω. (323)
At zero Matsubara frequency, the product z n ℓ E ζ n z_n\ell_{E\zeta n} z n ℓ E ζ n is understood by continuity and vanishes.
Integrating out the bath gives the inverse coordinate Green function
G E ζ − 1 ( i ν n ) = D n β ζ + t R ζ n , R ζ n = z n ℓ E ζ n + h E ζ . (324) G^{-1}_{E\zeta}(i\nu_n)=\frac{D_n}{\beta_\zeta}+tR_{\zeta n},\qquad R_{\zeta n}=z_n\ell_{E\zeta n}+h_{E\zeta}. \tag*{(324)} G E ζ − 1 ( i ν n ) = β ζ D n + t R ζ n , R ζ n = z n ℓ E ζ n + h E ζ . (324)
The first term in R ζ n R_{\zeta n} R ζ n is the frequency-dependent bath contribution that remains under full static compensation. The second is the frequency-independent residual static interaction. In particular,
R ζ 0 = h E ζ . R_{\zeta0}=h_{E\zeta}. R ζ 0 = h E ζ .
This zero-frequency contribution is absent under full compensation but must be retained when the cross contribution is only partially compensated.
Let
X ζ = α ζ Q ζ , Y ζ = β ζ P ζ , X_\zeta=\alpha_\zeta Q_\zeta,\qquad Y_\zeta=\beta_\zeta P_\zeta, X ζ = α ζ Q ζ , Y ζ = β ζ P ζ ,
where Q ζ Q_\zeta Q ζ and P ζ P_\zeta P ζ are the equilibrium position and momentum variances, respectively. The coordinate covariance obtained from the Green function, together with the equilibrium virial identity, gives
X ζ = ν S 2 T U ∑ n ∈ Z 1 D n + t β ζ R ζ n , X_\zeta=\nu_S^2T_U\sum_{n\in\mathbb{Z}}\frac{1}{D_n+t\beta_\zeta R_{\zeta n}}, X ζ = ν S 2 T U n ∈ Z ∑ D n + t β ζ R ζ n 1 ,
Y ζ = T U ∑ n ∈ Z ν S 2 + t β ζ R ζ n D n + t β ζ R ζ n . (325) Y_\zeta=T_U\sum_{n\in\mathbb{Z}}\frac{\nu_S^2+t\beta_\zeta R_{\zeta n}}{D_n+t\beta_\zeta R_{\zeta n}}. \tag*{(325)} Y ζ = T U n ∈ Z ∑ D n + t β ζ R ζ n ν S 2 + t β ζ R ζ n . (325)
Both sums converge for the regulated spectrum. Subtracting the two expressions isolates the difference between the scaled momentum and position variances:
Y ζ − X ζ = t β ζ T U ∑ n ∈ Z R ζ n D n + t β ζ R ζ n . (326) Y_\zeta-X_\zeta=t\beta_\zeta T_U\sum_{n\in\mathbb{Z}}\frac{R_{\zeta n}}{D_n+t\beta_\zeta R_{\zeta n}}. \tag*{(326)} Y ζ − X ζ = t β ζ T U n ∈ Z ∑ D n + t β ζ R ζ n R ζ n . (326)
This difference determines the Gaussian ergotropy of each collective sector. In particular, the zero-frequency term contributes
t β ζ T U h E ζ ν S 2 + t β ζ h E ζ . (327) \frac{t\beta_\zeta T_U h_{E\zeta}}{\nu_S^2+t\beta_\zeta h_{E\zeta}}. \tag*{(327)} ν S 2 + t β ζ h E ζ t β ζ T U h E ζ . (327)
It is therefore important not to omit the n = 0 n=0 n = 0 term when a residual static interaction is present.
The coordinate variance can also be expressed relative to its uncoupled equilibrium value. Using
T U ∑ n 1 D n = c U 2 ν S , c U = coth ν S 2 T U , T_U\sum_n\frac{1}{D_n}=\frac{c_U}{2\nu_S},\qquad c_U=\coth\frac{\nu_S}{2T_U}, T U n ∑ D n 1 = 2 ν S c U , c U = coth 2 T U ν S ,
we obtain
X ζ − ν S c U 2 = − t ν S 2 β ζ T U ∑ n R ζ n D n ( D n + t β ζ R ζ n ) . (328) X_\zeta-\frac{\nu_S c_U}{2}=-t\nu_S^2\beta_\zeta T_U\sum_n\frac{R_{\zeta n}}{D_n(D_n+t\beta_\zeta R_{\zeta n})}. \tag*{(328)} X ζ − 2 ν S c U = − t ν S 2 β ζ T U n ∑ D n ( D n + t β ζ R ζ n ) R ζ n . (328)
We now expand the covariance difference to leading order in t = κ 2 t=\kappa^2 t = κ 2 . This gives
Y ζ − X ζ = t β ζ A ~ ζ + O ( t 2 ) , Y_\zeta-X_\zeta=t\beta_\zeta\widetilde{A}_\zeta+O(t^2), Y ζ − X ζ = t β ζ A ζ + O ( t 2 ) ,
A ~ ζ = I E , ζ 2 π 2 + h E ζ c U 2 ν S . (329) \widetilde{A}_\zeta=\frac{I_{E,\zeta}}{2\pi^2}+\frac{h_{E\zeta}c_U}{2\nu_S}. \tag*{(329)} A ζ = 2 π 2 I E , ζ + 2 ν S h E ζ c U . (329)
Here, I E , ζ I_{E,\zeta} I E , ζ is the frequency-weighted spectral integral defined in the main text:
I E , ζ = I 0 ± I c , E . I_{E,\zeta}=I_0\pm I_{\mathrm{c},E}. I E , ζ = I 0 ± I c , E .
The first term in A ~ ζ \widetilde{A}_{\zeta} A ζ is the contribution obtained under full static compensation. The second term is the correction from the residual static interaction. To see the origin of the first term, we use
T U ∑ n ∈ Z z n D n ( z n + Ω 2 ) = B 1 ( Ω ) , (330) T_U \sum_{n\in\mathbb{Z}} \frac{z_n}{D_n(z_n+\Omega^2)}=B_1(\Omega), \tag*{(330)} T U n ∈ Z ∑ D n ( z n + Ω 2 ) z n = B 1 ( Ω ) , (330)
and then perform the frequency integral in ℓ E ζ n \ell_{E\zeta n} ℓ E ζ n .
We can now calculate the Gaussian ergotropy. For one collective sector, it is
G ζ = X ζ + Y ζ 2 − X ζ Y ζ . (331) \mathcal{G}_{\zeta}=\frac{X_{\zeta}+Y_{\zeta}}{2}-\sqrt{X_{\zeta}Y_{\zeta}}. \tag*{(331)} G ζ = 2 X ζ + Y ζ − X ζ Y ζ . (331)
Equivalently, it can be written as
G ζ = ( Y ζ − X ζ ) 2 2 ( X ζ + Y ζ + 2 X ζ Y ζ ) . (332) \mathcal{G}_{\zeta}=\frac{(Y_{\zeta}-X_{\zeta})^2}{2\left(X_{\zeta}+Y_{\zeta}+2\sqrt{X_{\zeta}Y_{\zeta}}\right)}. \tag*{(332)} G ζ = 2 ( X ζ + Y ζ + 2 X ζ Y ζ ) ( Y ζ − X ζ ) 2 . (332)
At zero coupling, the two scaled variances coincide:
X ζ = Y ζ = ν S C U 2 . X_{\zeta}=Y_{\zeta}=\frac{\nu_{S}C_U}{2}. X ζ = Y ζ = 2 ν S C U .
The uncoupled detector therefore has zero Gaussian ergotropy. The leading nonzero contribution comes from the square of the coupling-induced variance difference. Substituting Eq. (329) , we obtain
G ζ = t 2 β ζ 2 4 ν S C U A ~ ζ 2 + O ( t 3 ) . (333) \mathcal{G}_{\zeta}=\frac{t^2\beta_{\zeta}^{2}}{4\nu_{S}C_U}\widetilde{A}_{\zeta}^{2}+O(t^3). \tag*{(333)} G ζ = 4 ν S C U t 2 β ζ 2 A ζ 2 + O ( t 3 ) . (333)
To express this result in terms of the spatial correlation defined in the main text, write
J c , E = 2 π 2 μ c , E , η E = I c , E I 0 . J_{c,E}=2\pi^2\mu_{c,E},\qquad\eta_E=\frac{I_{c,E}}{I_0}. J c , E = 2 π 2 μ c , E , η E = I 0 I c , E .
Using Eq. (322) , the coefficient in Eq. (329) becomes
A ~ ζ = I 0 2 π 2 ( 1 + ζ ξ E ) , (334) \widetilde{A}_{\zeta}=\frac{I_0}{2\pi^2}(1+\zeta\xi_E), \tag*{(334)} A ζ = 2 π 2 I 0 ( 1 + ζ ξ E ) , (334)
where
ξ E = η E − ( 1 − ρ c t ) C U J c , E 2 ν S I 0 . (335) \xi_E=\eta_E-(1-\rho_{ct})\frac{C_UJ_{c,E}}{2\nu_S I_0}. \tag*{(335)} ξ E = η E − ( 1 − ρ c t ) 2 ν S I 0 C U J c , E . (335)
Thus, the residual static interaction changes the parameter entering the work readout from η E \eta_E η E to ξ E \xi_E ξ E . When ρ c t = 1 \rho_{ct}=1 ρ c t = 1 , the residual term vanishes and ξ E = η E \xi_E=\eta_E ξ E = η E , recovering the fully compensated result.
Summing Eq. (333) over the two collective sectors gives
N E ( ρ c t ) = I 0 2 16 π 4 ν S C U [ ( ω − λ ) 2 ( 1 + ξ E ) 2 + ( ω + λ ) 2 ( 1 − ξ E ) 2 ] . (336) \mathcal{N}_{E}^{(\rho_{ct})}=\frac{I_0^2}{16\pi^4\nu_SC_U}\left[(\omega-\lambda)^2(1+\xi_E)^2+(\omega+\lambda)^2(1-\xi_E)^2\right]. \tag*{(336)} N E ( ρ c t ) = 16 π 4 ν S C U I 0 2 [ ( ω − λ ) 2 ( 1 + ξ E ) 2 + ( ω + λ ) 2 ( 1 − ξ E ) 2 ] . (336)
The part independent of the spatial correlation is the uncorrelated work coefficient
N C = I 0 2 ( ω 2 + λ 2 ) 8 π 4 ν S C U . (337) \mathcal{N}_C=\frac{I_0^2(\omega^2+\lambda^2)}{8\pi^4\nu_SC_U}. \tag*{(337)} N C = 8 π 4 ν S C U I 0 2 ( ω 2 + λ 2 ) . (337)
Dividing by N C \mathcal{N}_C N C and defining
s = 4 ω λ ω 2 + λ 2 , s=\frac{4\omega\lambda}{\omega^2+\lambda^2}, s = ω 2 + λ 2 4 ω λ ,
we obtain
N E ( ρ c t ) N C = 1 + ξ E 2 − s ξ E . (338) \frac{\mathcal{N}_{E}^{(\rho_{ct})}}{\mathcal{N}_C}=1+\xi_E^2-s\xi_E. \tag*{(338)} N C N E ( ρ c t ) = 1 + ξ E 2 − s ξ E . (338)
This is the relation used in the main text. It shows explicitly that incomplete static compensation changes the work readout through the replacement of η E \eta_E η E by ξ E \xi_E ξ E .
We can also identify how the residual static interaction changes the additional global work. At λ = 0 \lambda=0 λ = 0 , the leading work coefficient takes the form
N E ( ρ c t ) ∣ λ = 0 = N C ∣ λ = 0 ( 1 + ξ E 2 ) . (339) \left.\mathcal{N}_{E}^{(\rho_{ct})}\right|_{\lambda=0}=\left.\mathcal{N}_C\right|_{\lambda=0}(1+\xi_E^2). \tag*{(339)} N E ( ρ c t ) λ = 0 = N C ∣ λ = 0 ( 1 + ξ E 2 ) . (339)
The first term is the local-work contribution, whereas the second is the additional contribution obtained by operating jointly on the detector pair. Thus, the additional global-work coefficient is
( N E ( ρ c t ) − N C ) ∣ λ = 0 = N C ∣ λ = 0 ξ E 2 . (340) \left.\left(\mathcal{N}_{E}^{(\rho_{ct})}-\mathcal{N}_C\right)\right|_{\lambda=0}=\left.\mathcal{N}_C\right|_{\lambda=0}\xi_E^2. \tag*{(340)} ( N E ( ρ c t ) − N C ) λ = 0 = N C ∣ λ = 0 ξ E 2 . (340)
The residual static interaction therefore changes both the correlation parameter entering the work readout and the additional work available from the joint operation.
We now ask whether the work induced by incomplete static compensation requires quantum bath fluctuations or can also arise from the residual static interaction in the classical limit. To answer this question, we take the classical equilibrium limit of the detector–bath system and integrate out the bath coordinates. The resulting reduced equilibrium distribution of the detector pair is
ρ S , c l ∝ exp [ − 1 2 T U ∑ ζ ( ( α ζ + t h E ζ ) q ζ 2 + β ζ p ζ 2 ) ] . (341) \rho_{S,\mathrm{cl}}\propto\exp\left[-\frac{1}{2T_U}\sum_{\zeta}\left((\alpha_{\zeta}+th_E\zeta)q_{\zeta}^{2}+\beta_{\zeta}p_{\zeta}^{2}\right)\right]. \tag*{(341)} ρ S , cl ∝ exp − 2 T U 1 ζ ∑ ( ( α ζ + t h E ζ ) q ζ 2 + β ζ p ζ 2 ) . (341)
The corresponding classical variances are
Q ζ , c l = T U α ζ + t h E ζ , P ζ , c l = T U β ζ . (342) Q_{\zeta,\mathrm{cl}}=\frac{T_U}{\alpha_{\zeta}+th_E\zeta},\qquad P_{\zeta,\mathrm{cl}}=\frac{T_U}{\beta_{\zeta}}. \tag*{(342)} Q ζ , cl = α ζ + t h E ζ T U , P ζ , cl = β ζ T U . (342)
Evaluating the same covariance-based work expression relative to the original detector Hamiltonian H S \mathcal{H}_S H S , we find
G ζ , c l = T U 2 [ 1 + α ζ α ζ + t h E ζ − 2 α ζ α ζ + t h E ζ ] = T U t 2 h E 2 8 α ζ 2 + O ( t 3 ) . (343) \mathcal{G}_{\zeta,\mathrm{cl}}=\frac{T_U}{2}\left[1+\frac{\alpha_{\zeta}}{\alpha_{\zeta}+th_E\zeta}-2\sqrt{\frac{\alpha_{\zeta}}{\alpha_{\zeta}+th_E\zeta}}\right]
=\frac{T_Ut^2h_E^2}{8\alpha_{\zeta}^{2}}+O(t^3). \tag*{(343)} G ζ , cl = 2 T U [ 1 + α ζ + t h E ζ α ζ − 2 α ζ + t h E ζ α ζ ] = 8 α ζ 2 T U t 2 h E 2 + O ( t 3 ) . (343)
When the static contribution is fully compensated, h E ζ = 0 h_{E\zeta}=0 h E ζ = 0 , the classical equilibrium variances reduce to Q ζ , c l = T U / α ζ Q_{\zeta,\mathrm{cl}}=T_U/\alpha_\zeta Q ζ , cl = T U / α ζ and P ζ , c l = T U / β ζ P_{\zeta,\mathrm{cl}}=T_U/\beta_\zeta P ζ , cl = T U / β ζ . These are precisely the variances of the classical Gibbs state of the original detector Hamiltonian H S \mathcal{H}_S H S . The state is therefore passive with respect to H S \mathcal{H}_S H S , and its extractable work vanishes. In contrast, when h E ζ ≠ 0 h_{E\zeta}\ne0 h E ζ = 0 , the residual static interaction changes the effective stiffness from α ζ \alpha_\zeta α ζ to α ζ + t h E ζ \alpha_\zeta+t h_{E\zeta} α ζ + t h E ζ , while the momentum variance remains unchanged. The resulting classical equilibrium state is no longer the Gibbs state of H S \mathcal{H}_S H S . Consequently, it has nonzero ergotropy when the work is evaluated relative to the original detector Hamiltonian, as shown in Eq. (343) . This contribution arises from the residual static interaction itself and does not require quantum bath fluctuations. Thus, under incomplete static compensation, the extracted work contains a contribution from the change in the static detector potential, in addition to any contribution associated with quantum bath fluctuations.
Numerical methods and convergence checks
This appendix describes the numerical methods used in the transient and equilibrium calculations and provides the corresponding convergence checks.
Finite-coupling equilibrium covariance and numerical verification
The finite-coupling equilibrium work is calculated from the unexpanded Matsubara Green function, without performing a weak-coupling expansion in κ \kappa κ . The frequency integrals entering the self-energy are evaluated numerically. The zero-frequency contribution is treated separately, while the large-frequency tail of the Matsubara sum is evaluated using its asymptotic expansion and Hurwitz zeta functions. The resulting equilibrium covariances are then used to calculate the Gaussian ergotropy.
To independently verify the Matsubara results, we also calculate the equilibrium covariance by diagonalizing the full Hamiltonian of a discretized detector–bath system. In contrast to the Matsubara approach, which integrates out the bath degrees of freedom, this method retains the bath oscillators explicitly and evaluates the detector covariance directly from the interacting Gibbs state.
For each collective sector ζ = ± \zeta=\pm ζ = ± , the total quadratic Hamiltonian, including the static counterterm, can be written as
H t o t , ζ = 1 2 p ζ T B ζ p ζ + 1 2 q ζ T K ζ q ζ , (344) \mathcal{H}_{\mathrm{tot},\zeta}=\frac{1}{2}\mathbf{p}_{\zeta}^{T}B_{\zeta}\mathbf{p}_{\zeta}+\frac{1}{2}\mathbf{q}_{\zeta}^{T}K_{\zeta}\mathbf{q}_{\zeta}, \tag*{(344)} H tot , ζ = 2 1 p ζ T B ζ p ζ + 2 1 q ζ T K ζ q ζ , (344)
where
B ζ = diag ( β ζ , Ω 1 , … , Ω N ) , (345) B_{\zeta}=\operatorname{diag}(\beta_{\zeta},\Omega_1,\ldots,\Omega_N), \tag*{(345)} B ζ = diag ( β ζ , Ω 1 , … , Ω N ) , (345)
and the nonzero entries of K ζ K_{\zeta} K ζ are
( K ζ ) 00 = α ζ + δ ω ζ , (K_{\zeta})_{00}=\alpha_{\zeta}+\delta\omega_{\zeta}, ( K ζ ) 00 = α ζ + δ ω ζ ,
( K ζ ) j j = Ω j , (K_{\zeta})_{jj}=\Omega_j, ( K ζ ) j j = Ω j ,
( K ζ ) 0 j = ( K ζ ) j 0 = g j , ζ . (346) (K_{\zeta})_{0j}=(K_{\zeta})_{j0}=g_{j,\zeta}. \tag*{(346)} ( K ζ ) 0 j = ( K ζ ) j 0 = g j , ζ . (346)
Here, q ζ = ( q ζ , Q 1 , ζ , … , Q N , ζ ) T \mathbf{q}_{\zeta}=(q_{\zeta},Q_{1,\zeta},\ldots,Q_{N,\zeta})^{T} q ζ = ( q ζ , Q 1 , ζ , … , Q N , ζ ) T and p ζ = ( p ζ , P 1 , ζ , … , P N , ζ ) T \mathbf{p}_{\zeta}=(p_{\zeta},P_{1,\zeta},\ldots,P_{N,\zeta})^{T} p ζ = ( p ζ , P 1 , ζ , … , P N , ζ ) T . The index 0 denotes the detector, while j = 1 , … , N j=1,\ldots,N j = 1 , … , N labels the bath oscillators.
We introduce the canonical transformation
q ζ = B ζ 1 / 2 χ ζ , p ζ = B ζ − 1 / 2 π ζ , (347) \mathbf{q}_{\zeta}=B_{\zeta}^{1/2}\boldsymbol{\chi}_{\zeta},\qquad\mathbf{p}_{\zeta}=B_{\zeta}^{-1/2}\boldsymbol{\pi}_{\zeta}, \tag*{(347)} q ζ = B ζ 1/2 χ ζ , p ζ = B ζ − 1/2 π ζ , (347)
which brings the Hamiltonian into the form
H t o t , ζ = 1 2 π ζ T π ζ + 1 2 χ ζ T B ζ 1 / 2 K ζ B ζ 1 / 2 χ ζ . (348) \mathcal{H}_{\mathrm{tot},\zeta}=\frac{1}{2}\boldsymbol{\pi}_{\zeta}^{T}\boldsymbol{\pi}_{\zeta}+\frac{1}{2}\boldsymbol{\chi}_{\zeta}^{T}B_{\zeta}^{1/2}K_{\zeta}B_{\zeta}^{1/2}\boldsymbol{\chi}_{\zeta}. \tag*{(348)} H tot , ζ = 2 1 π ζ T π ζ + 2 1 χ ζ T B ζ 1/2 K ζ B ζ 1/2 χ ζ . (348)
The normal-mode frequencies are obtained from the orthogonal diagonalization
B ζ 1 / 2 K ζ B ζ 1 / 2 = O ζ diag ( f k 2 ) O ζ T , (349) B_{\zeta}^{1/2}K_{\zeta}B_{\zeta}^{1/2}=O_{\zeta}\operatorname{diag}(f_k^2)O_{\zeta}^{T}, \tag*{(349)} B ζ 1/2 K ζ B ζ 1/2 = O ζ diag ( f k 2 ) O ζ T , (349)
where f k > 0 f_k>0 f k > 0 are the normal-mode frequencies.
The transformed Hamiltonian describes independent harmonic oscillators. Using their thermal variances at temperature T U T_U T U , we obtain the equilibrium detector covariances
Q ζ = β ζ 2 ∑ k ( O ζ ) 0 k 2 coth ( f k / 2 T U ) f k , Q_{\zeta}=\frac{\beta_{\zeta}}{2}\sum_k(O_{\zeta})_{0k}^{2}\frac{\coth(f_k/2T_U)}{f_k}, Q ζ = 2 β ζ k ∑ ( O ζ ) 0 k 2 f k coth ( f k /2 T U ) ,
P ζ = 1 2 β ζ ∑ k ( O ζ ) 0 k 2 f k coth ( f k / 2 T U ) . (350) P_{\zeta}=\frac{1}{2\beta_{\zeta}}\sum_k(O_{\zeta})_{0k}^{2}f_k\coth(f_k/2T_U). \tag*{(350)} P ζ = 2 β ζ 1 k ∑ ( O ζ ) 0 k 2 f k coth ( f k /2 T U ) . (350)
These expressions give the detector covariance of the reduced interacting Gibbs state, including the effects of the detector–bath coupling and the static counterterm. No expansion in the coupling strength is required.
The Gaussian ergotropy is evaluated with respect to the original detector Hamiltonian
H S , ζ = 1 2 ( α ζ q ζ 2 + β ζ p ζ 2 ) , (351) \mathcal{H}_{S,\zeta}=\frac{1}{2}\left(\alpha_{\zeta}q_{\zeta}^{2}+\beta_{\zeta}p_{\zeta}^{2}\right), \tag*{(351)} H S , ζ = 2 1 ( α ζ q ζ 2 + β ζ p ζ 2 ) , (351)
consistent with the work definition used in the main text. For numerical stability, we use
G ζ = ( X ζ − Y ζ ) 2 2 ( X ζ + Y ζ + 2 X ζ Y ζ ) , \mathcal{G}_{\zeta}=\frac{(X_{\zeta}-Y_{\zeta})^{2}}{2\left(X_{\zeta}+Y_{\zeta}+2\sqrt{X_{\zeta}Y_{\zeta}}\right)}, G ζ = 2 ( X ζ + Y ζ + 2 X ζ Y ζ ) ( X ζ − Y ζ ) 2 ,
X ζ = α ζ Q ζ , Y ζ = β ζ P ζ . (352) X_{\zeta}=\alpha_{\zeta}Q_{\zeta},\qquad Y_{\zeta}=\beta_{\zeta}P_{\zeta}. \tag*{(352)} X ζ = α ζ Q ζ , Y ζ = β ζ P ζ . (352)
This expression is algebraically equivalent to the Gaussian ergotropy formula used in the main text, but avoids subtracting nearly equal energies when the extractable work is small.
To verify the finite-coupling equilibrium calculation, we compare two independent methods: the continuum Matsubara Green-function approach, in which the bath degrees of freedom are integrated out, and the finite-bath normal-mode approach, in which the full detector–bath Hamiltonian is diagonalized directly. Neither method involves a weak-coupling expansion.
For both environments, E = U , T E=U,T E = U , T , at d = 1 d=1 d = 1 and d = d 0 d=d_0 d = d 0 , with κ 2 = 0.05 \kappa^2=0.05 κ 2 = 0.05 , the relative difference in the equilibrium Gaussian ergotropy obtained from the two methods is below 1.9 × 10 − 8 1.9 \times10^{-8} 1.9 × 1 0 − 8 . This agreement provides an independent numerical check of the finite-coupling equilibrium results.
Numerical convergence of the dynamical calculations
We first check the convergence of the original finite-time calculations with respect to the discretization of the bath spectrum. The continuum frequency interval is fixed at
Ω ∈ [ 10 − 5 , 50 ] , \Omega\in[10^{-5},50], Ω ∈ [ 1 0 − 5 , 50 ] ,
and the spectral integral is approximated by an N N N -point Gauss–Legendre quadrature. To isolate the effect of the quadrature resolution, all other parameters and the propagation protocol are kept fixed.
The parameters used in this comparison are
Table 1.
We calculate the transient ergotropy for N = 5000 N=5000 N = 5000 , 10000 10000 10000 , and 20000 20000 20000 , and use the N = 40000 N=40000 N = 40000 result as the numerical reference. The differences are summarized in the following table.
Table 2.
Here, Δ \Delta Δ is the maximum absolute difference between the entire ergotropy curve and the N = 40000 N=40000 N = 40000 reference curve:
Δ = max 0 ≤ τ ≤ 50 ∣ G N ( τ ) − G 40000 ( τ ) ∣ . \Delta=\max_{0\leq\tau\leq50}\left|G_N(\tau)-G_{40000}(\tau)\right|. Δ = 0 ≤ τ ≤ 50 max ∣ G N ( τ ) − G 40000 ( τ ) ∣ .
The relative deviation is defined by
r e l . = Δ G p k , r e f × 100 % , \mathrm{rel.}=\frac{\Delta}{G_{\mathrm{pk,ref}}}\times100\%, rel. = G pk , ref Δ × 100% ,
where G p k , r e f G_{\mathrm{pk,ref}} G pk , ref is the main peak height of the reference curve. The quantities δ G p k \delta G_{\mathrm{pk}} δ G pk , δ T \delta T δ T , and δ G ( 50 ) \delta G(50) δ G ( 50 ) denote, respectively, the relative deviations of the main peak height, the separation between successive peaks, and the ergotropy at the endpoint τ = 50 \tau=50 τ = 50 , each evaluated relative to its corresponding reference value.
The maximum absolute difference is approximately 10 − 15 10^{-15} 1 0 − 15 for all three resolutions, indicating that the grid-refinement differences have reached the floating-point precision level of the calculation. The relative deviations of the peak height and endpoint value are also extremely small.
of the peak height and endpoint value are also extremely small. The entries displayed as 0.00 for δ T \delta T δ T and δ G ( 50 ) \delta G(50) δ G ( 50 ) are zero at the precision reported in the table. These results show that the transient ergotropy curve is insensitive to further refinement of the quadrature grid for the parameters considered here.
We next examine the dependence of the numerical results on the frequency interval. Taking
Ω ∈ [ 10 − 6 , 80 ] \Omega\in[10^{-6},80] Ω ∈ [ 1 0 − 6 , 80 ]
as the reference interval, we vary one endpoint at a time while keeping the other fixed. The maximum relative deviations, normalized by the reference peak amplitude, are listed below.
Table 3.
The upper-cutoff test is performed with Ω min = 10 − 6 \Omega_{\min}=10^{-6} Ω m i n = 1 0 − 6 fixed, whereas the lower-cutoff test is performed with Ω max = 80 \Omega_{\max}=80 Ω m a x = 80 fixed. For the frequency interval Ω ∈ [ 10 − 5 , 50 ] \Omega\in[10^{-5},50] Ω ∈ [ 1 0 − 5 , 50 ] adopted in this work, the corresponding deviations are 1.314 × 10 − 4 % 1.314 \times10^{-4}\% 1.314 × 1 0 − 4 % for the upper cutoff and 2.118 × 10 − 3 % 2.118 \times10^{-3}\% 2.118 × 1 0 − 3 % for the lower cutoff. Both are sufficiently small compared with the reference peak amplitude. These results support the use of Ω ∈ [ 10 − 5 , 50 ] \Omega\in[10^{-5},50] Ω ∈ [ 1 0 − 5 , 50 ] in the transient calculations.
We further examine how the transient work depends on the spectral cutoff Ω c \Omega_c Ω c . For each value of Ω c \Omega_c Ω c , we set Ω max = 10 Ω c \Omega_{\max}=10\Omega_c Ω m a x = 10 Ω c to ensure sufficient coverage of the regulated spectral tail. The first peak amplitude, the decay rate of the oscillation envelope, and the ergotropy at τ = 50 \tau=50 τ = 50 are compared for Ω c = 3 , 5 , 8 \Omega_c=3,5,8 Ω c = 3 , 5 , 8 .
Table 4.
Here, G p k G_{\rm pk} G pk denotes the height of the first ergotropy peak. The envelope rate is obtained by a least-squares fit of
log A m = r τ m + b \log A_m=r\tau_m+b log A m = r τ m + b
to the sequence of main peak amplitudes A m A_m A m , where r < 0 r<0 r < 0 is the fitted decay rate. The quantity G ( 50 ) G(50) G ( 50 ) denotes the ergotropy at τ = 50 \tau=50 τ = 50 .
As Ω c \Omega_c Ω c increases from 3 to 8, the first peak amplitude increases from 2.691 × 10 − 5 2.691 \times10^{-5} 2.691 × 1 0 − 5 to 3.152 × 10 − 5 3.152 \times10^{-5} 3.152 × 1 0 − 5 , while the envelope decay rate varies from − 1.104 × 10 − 2 -1.104 \times10^{-2} − 1.104 × 1 0 − 2 to − 1.215 × 10 − 2 -1.215 \times10^{-2} − 1.215 × 1 0 − 2 . The ergotropy at τ = 50 \tau=50 τ = 50 decreases over the same range. Thus, changing the spectral cutoff affects the quantitative work amplitude and its late-time value, while the characteristic decay rate remains within a relatively narrow range. The damped oscillatory behavior is preserved throughout the parameter range considered here.
References