Introduction
A reservoir with finite correlation time can return information to an open quantum system and alter its relaxation, coherence, and correlations [1 , 2 ] . These effects depend on the reservoir structure as well as on the system–bath coupling. Examples include entanglement revivals for qubits coupled to independent reservoirs [3 ] and noise-assisted coherence protection in hierarchical environments [4 ] . In a piezoelectric optomechanical model, Ding et al. reported singular coefficient functions in non-Markovian quantum-state diffusion (NMQSD) and distinct entanglement dynamics as the bath central frequency was varied [5 ] . Such coefficient behavior motivates a spectral question: can a degeneracy in the equations that encode environmental memory be identified through the quantum state that they generate?
Exceptional points (EPs), at which eigenvalues and eigenvectors of a non-Hermitian operator coalesce, provide a natural setting for this question [6 , 7 ] . Their Jordan structure produces polynomial factors multiplying exponential evolution. In an open quantum system, the operator whose spectrum is studied must be specified: effective Hamiltonians and Liouvillians can have different EPs because quantum jumps affect the density-operator evolution [8 ] . With reservoir memory, pseudomode and hierarchical descriptions explicitly project extended Jordan modes into reduced-state signals [9 ] , while reaction-coordinate treatments connect poles to observable-dependent polynomial factors and late-time relaxation [10 ] . These physical connections build on enlarged-space representations of structured reservoirs [11 , 12 , 13 ] . Zhang et al. describe an experimental superposition preparation and joint qubit–pseudomode tomography that showed non-Markovian EPs with quantum jumps retained [14 ] . Different coherence functions can display different orders [15 ] , so the preparation and readout matter as well as the spectrum.
The NMQSD formulation follows these connections through the stochastic system state [16 , 17 , 18 , 19 ] . In this formulation, the reservoir enters a stochastic pure-state equation through colored Gaussian noise and a functional derivative. An O O O operator represents that derivative and, for suitable system algebra, obeys a finite exact closure. Exact time-local equations have been constructed for interacting qubits and multilevel systems [20 , 21 ] . Averaging the same stochastic equation yields the reduced density operator; the corresponding master equation has been derived explicitly for quantum Brownian motion and multiple qubits [22 , 23 ] . This established route carries the coefficient spectrum through state reconstruction and noise averaging within the same reservoir model.
Here we ask which features of a coefficient EP survive the state map and complete noise average in two exactly closed models. A linear homogeneous lift makes the nonlinear coefficient dynamics accessible to spectral analysis, but its physical solution has a prescribed initial vector and a denominator projection. These select the reached Jordan chains. State reconstruction can retain, combine, or cancel their contributions. The noise average must also include correlations between the O O O operator and the stochastic projector; a finite coefficient closure does not make this step automatic.
For common-bath qubits in the zero- and single-excitation sector, the O O O operator is noise independent. We establish an affine-transpose relation between the coefficient and state-memory matrices and prove that every defective decaying mode retains its full pole order in a suitable unconditional excitation–reference coherence. The proof covers arbitrary Hermitian single-excitation Hamiltonians and nonzero collective rank-one coupling with a single-exponential bath correlation. A known third-order EP (EP3) configuration [9 ] illustrates the general result and cancellations in individual quadratures. A determinant identity then relates the coefficient chart to the amplitude map. On intervals where this map is invertible, its inversion identifies the memory coefficients from separately prepared ensembles and known model parameters.
The position-coupled oscillator tests the extra averaging step that the qubit closure avoids. In driven, linearized red-sideband optomechanics, reservoir memory shifts a hybrid second-order EP (EP2) that enters the probe-reflection response [24 ] . Here the photon number is conserved, with no drive or optical loss, and radiation pressure supplies an affine force. The exact O O O operator contains a scalar noise-history term [18 ] . On the regular normalizable branch studied here, the two coefficient roots produce a double-pole conditional displacement response to the NMQSD noise history at coalescence, in the long-preparation-age limit at fixed lag. The purely algebraic coefficient tail cancels from the Gaussian width. The full operator average then determines the effective master equation. Comparing the conditional and reduced-state covariances identifies how the spread of state centers changes the fluctuations after averaging. Together, these results give a readout-dependent interpretation of coefficient EPs within the exact NMQSD-to-master-equation route.
Common coefficient-to-state framework
Using ℏ = 1 \hbar= 1 ℏ = 1 , the linear NMQSD equation for an initially factorized system and Gaussian vacuum reservoir reads [18 , 22 ]
∂ t ∣ ψ t ⟩ = [ − i H S + L z t ∗ − L † O ˉ t ] ∣ ψ t ⟩ , \partial_t|\psi_t\rangle= \left[-iH_S + Lz_t^* - L^\dagger\bar{O}_t\right]|\psi_t\rangle, ∂ t ∣ ψ t ⟩ = [ − i H S + L z t ∗ − L † O ˉ t ] ∣ ψ t ⟩ ,
O ˉ t = ∫ 0 t α ( t , s ) O ( t , s , z ∗ ) d s , \bar{O}_t = \int_0^t \alpha(t,s)O(t,s,z^*)\,\mathrm{d}s, O ˉ t = ∫ 0 t α ( t , s ) O ( t , s , z ∗ ) d s ,
δ ∣ ψ t ⟩ δ z s ∗ = O ( t , s , z ∗ ) ∣ ψ t ⟩ . (1) \frac{\delta|\psi_t\rangle}{\delta z_s^*} = O(t,s,z^*)|\psi_t\rangle. \tag*{(1)} δ z s ∗ δ ∣ ψ t ⟩ = O ( t , s , z ∗ ) ∣ ψ t ⟩ . (1)
Here H S H_S H S is the system Hamiltonian and L L L is the system operator that couples to the reservoir. The reference noise has zero mean, E 0 [ z t z s ∗ ] = α ( t , s ) \mathbb{E}_0[z_tz_s^*] = \alpha(t,s) E 0 [ z t z s ∗ ] = α ( t , s ) , and E 0 [ z t z s ] = 0 \mathbb{E}_0[z_tz_s] = 0 E 0 [ z t z s ] = 0 . Both applications use the full prescribed correlation
α ( t , s ) = A 0 e − γ ∣ t − s ∣ − i Ω ( t − s ) , d = γ + i Ω , (2) \alpha(t,s) = A_0e^{-\gamma|t-s|-i\Omega(t-s)}, \qquad d = \gamma+ i\Omega, \tag*{(2)} α ( t , s ) = A 0 e − γ ∣ t − s ∣ − i Ω ( t − s ) , d = γ + i Ω , (2)
with A 0 , γ > 0 A_0,\gamma> 0 A 0 , γ > 0 and real Ω \Omega Ω . The parameters set the correlation amplitude, inverse memory time, and central frequency. We use this full kernel as the definition of the effective reservoir. Truncating its full-line Lorentzian transform to positive frequencies would change the prescribed correlation. The system preparations are specified separately.
In the vacuum Bargmann representation, fixed coherent-state labels generate the colored process
z t ∗ = − i ∑ k g k z k ∗ e i ω k t , E 0 [ z t ∗ ] = 0 , z_t^* = -i\sum_k g_k z_k^* e^{i\omega_k t}, \qquad\mathbb{E}_0[z_t^*] = 0, z t ∗ = − i k ∑ g k z k ∗ e i ω k t , E 0 [ z t ∗ ] = 0 ,
E 0 [ z t z s ∗ ] = α ( t , s ) , E 0 [ z t z s ] = 0. (3) \mathbb{E}_0[z_tz_s^*] = \alpha(t,s), \qquad\mathbb{E}_0[z_tz_s] = 0. \tag*{(3)} E 0 [ z t z s ∗ ] = α ( t , s ) , E 0 [ z t z s ] = 0. (3)
Here g k g_k g k and ω k \omega_k ω k denote the mode weights and frequency labels of a Bargmann representation chosen to reproduce the prescribed correlation in Eq. (2) ; its continuum realization is understood at the level of this full kernel. The reference Gaussian measure satisfies E 0 [ z k z l ∗ ] = δ k l \mathbb{E}_0[z_kz_l^*] = \delta_{kl} E 0 [ z k z l ∗ ] = δ k l . Bath annihilation operators become derivatives with respect to z k ∗ z_k^* z k ∗ , and the chain rule gives the history derivative in Eq. (1) . Thus the ket depends on the entire label configuration, not only on the instantaneous value z t ∗ z_t^* z t ∗ .
Commuting the history derivative with time evolution gives the consistency equation
∂ t O = [ − i H S + L z t ∗ − L † O ˉ t , O ] − L † δ O ˉ t δ z s ∗ , O ( t , t , z ∗ ) = L . (4) \partial_t O = [-iH_S + Lz_t^* - L^\dagger\bar{O}_t,O] - L^\dagger\frac{\delta\bar{O}_t}{\delta z_s^*}, \qquad O(t,t,z^*) = L. \tag*{(4)} ∂ t O = [ − i H S + L z t ∗ − L † O ˉ t , O ] − L † δ z s ∗ δ O ˉ t , O ( t , t , z ∗ ) = L . (4)
An exponential correlation simplifies the integrated coefficient equations, but a finite exact closure also requires a suitable operator algebra. For the qubits, the retained excitation sector closes on lowering operators [23 ] . For the oscillator, closure requires a scalar noise-history term as well as position and momentum [18 , 22 ] .
The exact density operator and normalized-state ensemble obey
ρ t = E 0 [ ∣ ψ t ⟩ ⟨ ψ t ∣ ] , d P t = ⟨ ψ t ∣ ψ t ⟩ d P 0 , (5) \rho_t = \mathbb{E}_0[|\psi_t\rangle\langle\psi_t|], \qquad\mathrm{d}\mathbb{P}_t = \langle\psi_t|\psi_t\rangle\,\mathrm{d}\mathbb{P}_0, \tag*{(5)} ρ t = E 0 [ ∣ ψ t ⟩ ⟨ ψ t ∣ ] , d P t = ⟨ ψ t ∣ ψ t ⟩ d P 0 , (5)
where E 0 ⟨ ψ t ∣ ψ t ⟩ = 1 \mathbb{E}_0\langle\psi_t|\psi_t\rangle= 1 E 0 ⟨ ψ t ∣ ψ t ⟩ = 1 for a normalized preparation. Thus averaging normalized kets with the reference Gaussian weight is generally insufficient. Equation (5) defines a fixed-time ensemble. For general non-Markovian dynamics, these pure-state ensembles cannot be joined into pure-state trajectories generated by continuous monitoring of the environment [25 ] .
Writing Π t = ∣ ψ t ⟩ ⟨ ψ t ∣ \Pi_t = \lvert\psi_t\rangle\langle\psi_t\rvert Π t = ∣ ψ t ⟩ ⟨ ψ t ∣ and retaining the operator–state correlations gives the formally exact non-Markovian master equation (NMME) [22 , 23 ] ,
ρ ˙ t = − i [ H S , ρ t ] + [ L , A t † ] + [ A t , L † ] , (6) \dot{\rho}_t=-i[H_S,\rho_t]+[L,A_t^\dagger]+[A_t,L^\dagger], \tag*{(6)} ρ ˙ t = − i [ H S , ρ t ] + [ L , A t † ] + [ A t , L † ] , (6)
A t = E 0 [ O ‾ t Π t ] . A_t=\mathbb{E}_0[\overline{O}_t\Pi_t]. A t = E 0 [ O t Π t ] .
This is the NMQSD-to-master-equation relation used in both applications. If O ‾ t \overline{O}_t O t is independent of the noise labels, A t = O ‾ t ρ t A_t=\overline{O}_t\rho_t A t = O t ρ t and the equation closes directly. For a noise-dependent O O O operator, the joint average must instead be evaluated completely. Zero-mean noise does not imply zero noise–state correlations. In the oscillator calculation these correlations carry the scalar history sector into the effective NMME coefficients. This distinction also appears in general Gaussian stochastic and master-equation constructions [26 , 27 ] .
Projective coefficients and the causal lift
The deterministic coefficient cores derived below have the form
y ˙ = b + K y − y ( r T y ) , y ( 0 ) = 0 , (7) \dot{y}=b+Ky-y(r^T y),\qquad y(0)=0, \tag*{(7)} y ˙ = b + K y − y ( r T y ) , y ( 0 ) = 0 , (7)
with constant, generally complex entries. The transpose denotes a bilinear coordinate pairing, not a Hermitian inner product. Introducing homogeneous coordinates gives
y = x q , v = ( x q ) , v ˙ = M v , v ( 0 ) = e q , y=\frac{x}{q},\qquad v=\begin{pmatrix}x\\q\end{pmatrix},\qquad\dot{v}=Mv,\qquad v(0)=e_q, y = q x , v = ( x q ) , v ˙ = M v , v ( 0 ) = e q ,
M = ( K b r T 0 ) , q ˙ q = r T y . (8) M=\begin{pmatrix}K&b\\r^T&0\end{pmatrix},\qquad\frac{\dot{q}}{q}=r^T y. \tag*{(8)} M = ( K r T b 0 ) , q q ˙ = r T y . (8)
Here e q e_q e q is the last coordinate vector. Differentiating x / q x/q x / q proves Eq. (7) wherever q ≠ 0 q\ne0 q = 0 . An eigenvector of M M M with nonzero last coordinate gives a finite stationary coefficient vector y ∗ y_* y ∗ ; its eigenvalue is r T y ∗ r^T y_* r T y ∗ . Eigenvectors with vanishing last coordinate lie outside this chart.
This representation also separates spectral branching from finite-time singular behavior. If the entries of M M M depend analytically on a control parameter, the finite-time matrix exponential is analytic even at an exceptional point. Individual labeled eigenvalues can have fractional-power splittings while their combined contribution to v ( t ) v(t) v ( t ) remains regular. The ratios inherit that local regularity wherever q ≠ 0 q\ne0 q = 0 . Consequently, identifying an exceptional point from a defective matrix does not require a discontinuity in the finite-time coefficient trajectory. Conversely, a pole caused by q = 0 q=0 q = 0 does not establish a spectral degeneracy.
A zero of q q q therefore marks failure of the finite coefficient coordinates. The homogeneous solution remains defined at every finite time. Regularity of the reconstructed physical state requires the model-specific state map and is not a consequence of this coordinate observation alone. In the qubit model that map yields an exact determinant identity; in the oscillator model the Gaussian state and its averaging chart must also be controlled.
The prescribed zero-coefficient preparation fixes both the initial vector and the denominator readout. The Laplace transform q ~ ( s ) \widetilde{q}(s) q ( s ) of q ( t ) q(t) q ( t ) , with complex frequency variable s s s , is
q ~ ( s ) = e q T ( s I − M ) − 1 e q . (9) \widetilde{q}(s)=e_q^T(sI-M)^{-1}e_q. \tag*{(9)} q ( s ) = e q T ( s I − M ) − 1 e q . (9)
An eigenvalue degeneracy is insufficient to fix the late-time coefficient tail. Suppose that q ~ \widetilde{q} q has a pole of order r r r at μ ∗ \mu_* μ ∗ , whose real part is strictly larger than that of every other pole of this scalar function. Partial fractions then give
q ( t ) = e μ ∗ t [ a r − 1 t r − 1 + O ( t r − 2 ) ] + R s u b ( t ) , q(t)=e^{\mu_*t}\left[a_{r-1}t^{r-1}+O(t^{r-2})\right]+R_{\mathrm{sub}}(t), q ( t ) = e μ ∗ t [ a r − 1 t r − 1 + O ( t r − 2 ) ] + R sub ( t ) ,
r T y ( t ) = q ˙ ( t ) q ( t ) = μ ∗ + r − 1 t + O ( t − 2 ) . (10) r^T y(t)=\frac{\dot{q}(t)}{q(t)}=\mu_*+\frac{r-1}{t}+O(t^{-2}). \tag*{(10)} r T y ( t ) = q ( t ) q ˙ ( t ) = μ ∗ + t r − 1 + O ( t − 2 ) . (10)
Here a r − 1 ≠ 0 a_{r-1}\ne0 a r − 1 = 0 and R s u b R_{\mathrm{sub}} R sub is exponentially subleading relative to the first term. For r = 1 r=1 r = 1 , the leading polynomial is constant. The second line follows by logarithmic differentiation. The pole order must survive both excitation from e q e_q e q and projection back onto e q e_q e q ; it need not equal the largest Jordan block of M M M . Equal-real-part poles or a subleading exceptional point do not satisfy the stated dominance assumption. The two applications explicitly establish the required noncancellation for their causal preparations.
A reached dominant pole of order r r r therefore contributes ( r − 1 ) / t (r-1)/t ( r − 1 ) / t to the contraction r T y r^T y r T y . The sign and scale of a renamed model coefficient follow from its relation to this contraction, as derived in the corresponding model section. State reconstruction and the specified readout then determine an observable’s polynomial degree, decay rate, and late-time dominance; their propagation matrix and projection can differ from those of the coefficient lift.
A specified readout selects the Jordan contribution
State reconstruction supplies the appropriate propagation operator for a proposed signal. When that particular problem reduces to a constant linear system, write
u ˙ = A u + B f , o = C u , T ( s ) = C ( s I − A ) − 1 B . (11) \dot{u}=Au+Bf,\qquad o=Cu,\qquad\mathcal{T}(s)=C(sI-A)^{-1}B. \tag*{(11)} u ˙ = A u + B f , o = C u , T ( s ) = C ( s I − A ) − 1 B . (11)
For zero initial response, C e A t B Ce^{At}B C e A t B is the impulse-response kernel. The same expression propagates an initial preparation of the form B u 0 Bu_0 B u 0 . The matrices A , B , C A,B,C A , B , C must be derived for the stated input, preparation, and output; A A A is not assumed to be the coefficient matrix M M M .
Let P λ P_\lambda P λ be the spectral projector onto the generalized eigenspace of A A A at λ \lambda λ , and put N λ = ( A − λ I ) P λ N_\lambda=(A-\lambda I)P_\lambda N λ = ( A − λ I ) P λ . If its nilpotency index is r λ r_\lambda r λ , the contribution of this eigenspace is
T λ ( s ) = ∑ j = 0 r λ − 1 C N λ j P λ B ( s − λ ) j + 1 . (12) \mathcal{T}_{\lambda}(s)=\sum_{j=0}^{r_{\lambda}-1}\frac{CN_{\lambda}^{j}P_{\lambda}B}{(s-\lambda)^{j+1}}. \tag*{(12)} T λ ( s ) = j = 0 ∑ r λ − 1 ( s − λ ) j + 1 C N λ j P λ B . (12)
This follows by expanding the inverse of ( s − λ ) I − N λ (s-\lambda)I-N_{\lambda} ( s − λ ) I − N λ as a finite geometric series. Equivalently,
C e A t P λ B = e λ t ∑ j = 0 r λ − 1 t j j ! C N λ j P λ B . (13) Ce^{At}P_{\lambda}B=e^{\lambda t}\sum_{j=0}^{r_{\lambda}-1}\frac{t^j}{j!}CN_{\lambda}^{j}P_{\lambda}B. \tag*{(13)} C e A t P λ B = e λ t j = 0 ∑ r λ − 1 j ! t j C N λ j P λ B . (13)
The highest-order pole survives precisely when C N λ r λ − 1 P λ B ≠ 0 CN_{\lambda}^{r_{\lambda}-1}P_{\lambda}B\ne0 C N λ r λ − 1 P λ B = 0 . For matrix-valued inputs or outputs, this guarantees a suitable pair of channels, not every preparation and readout. A particular quadrature can cancel the highest term while retaining lower-order terms. These are standard linear-algebra identities underlying observable-dependent EP signatures [9 , 15 ] ; the physical work is to identify and justify the relevant map.
For the qubits, the relevant input is a coherent system preparation and the output is an unconditional excitation–reference coherence. For the oscillator, the response variable is the stochastic displacement, and its constant propagation matrix is obtained only after a long-preparation-age limit at fixed lag. The corresponding input varies the NMQSD history coordinate. These maps will be derived from the respective state equations. Neither Eq. (12) nor coefficient spectral similarity replaces nonlinear state reconstruction or the average in Eq. (5) . A Gaussian width can cancel a coefficient tail, and a reduced covariance combines conditional widths with the weighted spread of state centers. The two applications below show explicitly where these operations preserve the Jordan contribution and where they change its signature. The analysis concerns exact closures for an initially factorized vacuum reservoir with a prescribed exponential correlation; the spectral statements are not a general criterion for information backflow or for an EP of the instantaneous reduced-state generator.
Common-reservoir multiqubit systems
A common reservoir can couple several system transitions through one lowering operator. Restricting the system to at most one excitation makes its exact O O O operator noise independent. This provides a first test of the coefficient-to-state connection: the same deterministic row enters both the coefficient EP and the unconditional system coherence.
Exact NMQSD coefficients and their linear lift
Consider N q N_q N q qubits restricted to their zero- and single-excitation sector. The state ∣ G S ⟩ = ∣ 0 ⋯ 0 ⟩ |G_S\rangle=|0\cdots0\rangle ∣ G S ⟩ = ∣0 ⋯ 0 ⟩ is the system’s zero-excitation reference, distinct from the reservoir vacuum, and { ∣ j ⟩ } j = 1 N q \{|j\rangle\}_{j=1}^{N_q} { ∣ j ⟩ } j = 1 N q spans its single-excitation subspace. Subtracting the zero-excitation energy from the entire system Hamiltonian gives H S = 0 ⊕ H 1 H_S=0\oplus H_1 H S = 0 ⊕ H 1 , where the arbitrary time-independent Hermitian matrix H 1 H_1 H 1 contains transition energies relative to that reference. The common lowering operator is
L = ∑ j = 1 N q κ j ∣ G S ⟩ ⟨ j ∣ , κ = ( κ 1 , … , κ N q ) . (14) L=\sum_{j=1}^{N_q}\kappa_j|G_S\rangle\langle j|,\qquad\boldsymbol{\kappa}=(\kappa_1,\ldots,\kappa_{N_q}). \tag*{(14)} L = j = 1 ∑ N q κ j ∣ G S ⟩ ⟨ j ∣ , κ = ( κ 1 , … , κ N q ) . (14)
We choose L L L and κ \boldsymbol{\kappa} κ dimensionless. The coefficients κ j \kappa_j κ j weight system transitions in L L L , whereas the microscopic bath weights g k g_k g k specify the reservoir correlation in Eq. (2) . The interaction conserves total excitation number, and the initially independent reservoir is in its Gaussian vacuum with that exponential correlation. Thus A 0 A_0 A 0 has units of frequency squared and d = γ + i Ω d=\gamma+i\Omega d = γ + i Ω .
Within this invariant sector, the O O O operator closes exactly as
O ( t , s ) = ∑ j f q , j ( t , s ) ∣ G S ⟩ ⟨ j ∣ , O(t,s)=\sum_j f_{q,j}(t,s)|G_S\rangle\langle j|, O ( t , s ) = j ∑ f q , j ( t , s ) ∣ G S ⟩ ⟨ j ∣ ,
O ˉ t = ∑ j F q , j ( t ) ∣ G S ⟩ ⟨ j ∣ , \bar{O}_t=\sum_j F_{q,j}(t)|G_S\rangle\langle j|, O ˉ t = j ∑ F q , j ( t ) ∣ G S ⟩ ⟨ j ∣ ,
F q ( t ) = ∫ 0 t α ( t , s ) f q ( t , s ) d s . (15) F_{q}(t)=\int_0^t\alpha(t,s)f_q(t,s)\,\mathrm{d}s. \tag*{(15)} F q ( t ) = ∫ 0 t α ( t , s ) f q ( t , s ) d s . (15)
The rows f q f_q f q and F q F_q F q are noise independent, and both operators annihilate ∣ G S ⟩ |G_S\rangle ∣ G S ⟩ . This property uses the vacuum preparation and excitation restriction. General multi-excitation qubit closures contain additional noise terms [20 , 23 ] . The consistency equation and f q ( t , t ) = κ f_q(t,t)=\kappa f q ( t , t ) = κ give
∂ t f q = i f q H 1 + ( f q κ † ) F q , (16) \partial_t f_q=if_qH_1+(f_q\kappa^\dagger)F_q, \tag*{(16)} ∂ t f q = i f q H 1 + ( f q κ † ) F q , (16)
F ˙ q = A 0 κ − d F q + i F q H 1 + ( F q κ † ) F q , F q ( 0 ) = 0. (17) \dot{F}_q=A_0\kappa-dF_q+iF_qH_1+(F_q\kappa^\dagger)F_q,
\qquad F_q(0)=0. \tag*{(17)} F ˙ q = A 0 κ − d F q + i F q H 1 + ( F q κ † ) F q , F q ( 0 ) = 0. (17)
Appendix A derives this closure from the consistency equation. The coefficient spectrum follows directly from this exact NMQSD initial-value problem. Set F q = x T / q F_q=x^T/q F q = x T / q on q ≠ 0 q\ne0 q = 0 . Its linear lift is
v ˙ = M q v , v ( 0 ) = e a = ( 0 N q 1 ) , \dot{v}=M_qv,\qquad v(0)=e_a=\begin{pmatrix}0_{N_q}\\1\end{pmatrix}, v ˙ = M q v , v ( 0 ) = e a = ( 0 N q 1 ) ,
M q = ( − d I N q + i H 1 A 0 κ T − κ ∗ 0 ) . (18) M_q=\begin{pmatrix}
-dI_{N_q}+iH_1 & A_0\kappa^T\\
-\kappa^* & 0
\end{pmatrix}. \tag*{(18)} M q = ( − d I N q + i H 1 − κ ∗ A 0 κ T 0 ) . (18)
Here e a e_a e a is the last-coordinate unit column, and c q = F q κ † = − q ˙ / q c_q=F_q\kappa^\dagger=-\dot{q}/q c q = F q κ † = − q ˙ / q . A finite stationary direction with eigenvalue μ \mu μ gives
F q , ∗ ( μ ) = A 0 κ [ ( μ + d ) I − i H 1 ] − 1 , c q , ∗ = − μ , (19) F_{q,*}(\mu)=A_0\kappa[(\mu+d)I-iH_1]^{-1},
\qquad c_{q,*}=-\mu, \tag*{(19)} F q , ∗ ( μ ) = A 0 κ [( μ + d ) I − i H 1 ] − 1 , c q , ∗ = − μ , (19)
whenever the displayed inverse exists. Coefficient EPs are identified from the defective eigenvalues of M q M_q M q .
A third-order EP of the coefficient equations
For N q = 2 N_q=2 N q = 2 , take ∣ G S ⟩ = ∣ 00 ⟩ |G_S\rangle=|00\rangle ∣ G S ⟩ = ∣00 ⟩ , ∣ 1 ⟩ = ∣ 10 ⟩ |1\rangle=|10\rangle ∣1 ⟩ = ∣10 ⟩ , ∣ 2 ⟩ = ∣ 01 ⟩ |2\rangle=|01\rangle ∣2 ⟩ = ∣01 ⟩ , A 0 = γ 2 A_0=\gamma^2 A 0 = γ 2 , Ω = 0 \Omega=0 Ω = 0 , and
H 1 = γ ( 1 / 6 1 / ( 6 3 ) 1 / ( 6 3 ) − 1 / 6 ) , κ = ( 6 / 9 2 / 3 ) . (20) H_1=\gamma\begin{pmatrix}1/6 & 1/(6\sqrt{3}) \\ 1/(6\sqrt{3}) & -1/6\end{pmatrix},\qquad
\kappa=\begin{pmatrix}\sqrt{6}/9 & \sqrt{2}/3\end{pmatrix}. \tag*{(20)} H 1 = γ ( 1/6 1/ ( 6 3 ) 1/ ( 6 3 ) − 1/6 ) , κ = ( 6 /9 2 /3 ) . (20)
The characteristic polynomial calculated from Eq. (18) is det ( μ I 3 − M q ) = ( μ + 2 γ / 3 ) 3 \det(\mu I_3-M_q)=(\mu+2\gamma/3)^3 det ( μ I 3 − M q ) = ( μ + 2 γ /3 ) 3 . For N M = M q + 2 γ I 3 / 3 N_M=M_q+2\gamma I_3/3 N M = M q + 2 γ I 3 /3 , rank N M = 2 \operatorname{rank}N_M=2 rank N M = 2 , rank N M 2 = 1 \operatorname{rank}N_M^2=1 rank N M 2 = 1 , and N M 3 = 0 N_M^3=0 N M 3 = 0 . Thus the coefficient dynamics has one size-three Jordan block at μ E P = − 2 γ / 3 \mu_{\mathrm{EP}}=-2\gamma/3 μ EP = − 2 γ /3 . The parameter set is related by a basis rotation to the known third-order configuration of Ref. [9 ] ; here the degeneracy is diagnosed in the NMQSD coefficient matrix itself. Appendix C gives the explicit basis rotation and exact benchmark derivation.
With dimensionless time τ = γ t \tau=\gamma t τ = γ t , the prescribed coefficient solution is
q ( t ) = e − 2 τ / 3 ( 1 + 2 τ 3 + 2 τ 2 27 ) , (21) q(t)=e^{-2\tau/3}\left(1+\frac{2\tau}{3}+\frac{2\tau^2}{27}\right), \tag*{(21)} q ( t ) = e − 2 τ /3 ( 1 + 3 2 τ + 27 2 τ 2 ) , (21)
c q ( t ) = γ [ 2 3 − 2 / 3 + 4 τ / 27 1 + 2 τ / 3 + 2 τ 2 / 27 ] . (22) c_q(t)=\gamma\left[\frac{2}{3}-\frac{2/3+4\tau/27}{1+2\tau/3+2\tau^2/27}\right]. \tag*{(22)} c q ( t ) = γ [ 3 2 − 1 + 2 τ /3 + 2 τ 2 /27 2/3 + 4 τ /27 ] . (22)
The chart remains finite because q ( t ) > 0 q(t)>0 q ( t ) > 0 for t ≥ 0 t\ge0 t ≥ 0 . Its asymptotic contraction is c q = 2 γ / 3 − 2 / t + O ( ( γ t 2 ) − 1 ) c_q=2\gamma/3-2/t+O((\gamma t^2)^{-1}) c q = 2 γ /3 − 2/ t + O (( γ t 2 ) − 1 ) . Figure 1 displays the stationary branches and this coefficient tail.
Figure 1. EP3 obtained from the NMQSD coefficient equations. (a) Real parts of the stationary contraction c q , ∗ / γ = − μ / γ c_{q,*}/\gamma=-\mu/\gamma c q , ∗ / γ = − μ / γ under H 1 , 11 ↦ H 1 , 11 + ϵ H_{1,11}\mapsto H_{1,11}+\epsilon H 1 , 11 ↦ H 1 , 11 + ϵ . The branches are obtained from the coefficient lift in Eq. (18) ; the coalescence is fixed by the exact size-three Jordan block. (b) The ordinate t [ 2 γ / 3 − Re c q ( t ) ] t[2\gamma/3-\operatorname{Re}c_q(t)] t [ 2 γ /3 − Re c q ( t )] approaches 2. The analytic expression in Eq. (22) is compared with direct integration of the coefficient ordinary differential equation (ODE), Eq. (17) , using DOP853. This long-time comparison concerns coefficient dynamics.
For the local frequency shift H 1 , 11 ↦ H 1 , 11 + ϵ H_{1,11}\mapsto H_{1,11}+\epsilon H 1 , 11 ↦ H 1 , 11 + ϵ , define ε = ϵ / γ \varepsilon=\epsilon/\gamma ε = ϵ / γ and ζ = ( μ + 2 γ / 3 ) / γ \zeta=(\mu+2\gamma/3)/\gamma ζ = ( μ + 2 γ /3 ) / γ . The exact dimensionless coefficient polynomial is
p M ( ζ ) = ζ 3 − i ε ζ 2 + ε ( 1 / 6 + i / 3 ) ζ − ε / 9. (23) p_M(\zeta)=\zeta^3-i\varepsilon\zeta^2+\varepsilon(1/6+i/3)\zeta-\varepsilon/9. \tag*{(23)} p M ( ζ ) = ζ 3 − i ε ζ 2 + ε ( 1/6 + i /3 ) ζ − ε /9. (23)
It gives ζ 3 ∼ ε / 9 \zeta^3\sim\varepsilon/9 ζ 3 ∼ ε /9 and hence a cube-root splitting of the coefficient branches. The characteristic polynomial and the Jordan-rank test specify the EP without assuming its visibility in a particular state variable.
Stochastic-state propagation and the exact NMME
Let G ( t ) G(t) G ( t ) propagate the excitation amplitude of the linear NMQSD state. Its system-excitation block is deterministic in this sector:
G ˙ = ( − i H 1 − κ † F q ) G , G ( 0 ) = I N q . (24) \dot{G}=(-iH_1-\kappa^\dagger F_q)G,\qquad G(0)=I_{N_q}. \tag*{(24)} G ˙ = ( − i H 1 − κ † F q ) G , G ( 0 ) = I N q . (24)
The Gaussian history enters the zero-excitation amplitude. Consequently, noise-independent coefficients do not imply noise-independent stochastic kets.
The deterministic operator row factors out of the projector average in Eq. (6) : E 0 [ O ˉ t Π t ] = O ˉ t ρ t \mathbb{E}_0[\bar{O}_t\Pi_t]=\bar{O}_t\rho_t E 0 [ O ˉ t Π t ] = O ˉ t ρ t . This is the noise-independent case of the quantum-trajectory construction in Ref. [23 ] . Thus, on the original retained system space,
ρ ˙ t = − i [ H S , ρ t ] + [ L , ρ t O ˉ t † ] + [ O ˉ t ρ t , L † ] . (25) \dot{\rho}_t=-i[H_S,\rho_t]+[L,\rho_t\bar{O}_t^\dagger]+[\bar{O}_t\rho_t,L^\dagger]. \tag*{(25)} ρ ˙ t = − i [ H S , ρ t ] + [ L , ρ t O ˉ t † ] + [ O ˉ t ρ t , L † ] . (25)
The time dependence of O ˉ t \bar{O}_t O ˉ t retains the full exponential memory. Equation (25) is exact in the stated sector, without a Markov or weak-coupling approximation. Appendix A gives its block reduction and integrated state map. Noise-dependent sectors in the full multiqubit problem require the additional averages treated in Ref. [23 ] .
Write ρ e e \rho_{ee} ρ ee for the system excitation block, ρ e G = ( ρ 1 G , … , ρ N q G ) T \rho_{eG}=(\rho_{1G},\ldots,\rho_{N_qG})^T ρ e G = ( ρ 1 G , … , ρ N q G ) T for its coherence with ∣ G S ⟩ |G_S\rangle ∣ G S ⟩ , and ρ G G \rho_{GG} ρ GG for the zero-excitation population. Inserting the explicit L L L and O ˉ t \bar{O}_t O ˉ t into Eq. (25) gives
ρ ˙ e G = ( − i H 1 − κ † F q ) ρ e G , ρ ˙ e e = ( − i H 1 − κ † F q ) ρ e e + ρ e e ( i H 1 − F q † κ ) , ρ ˙ G G = F q ρ e e κ † + κ ρ e e F q † . (26) \begin{aligned}
\dot{\rho}_{eG}&=(-iH_1-\kappa^\dagger F_q)\rho_{eG},\\
\dot{\rho}_{ee}&=(-iH_1-\kappa^\dagger F_q)\rho_{ee}
+\rho_{ee}(iH_1-F_q^\dagger\kappa),\\
\dot{\rho}_{GG}&=F_q\rho_{ee}\kappa^\dagger+\kappa\rho_{ee}F_q^\dagger. \tag*{(26)}
\end{aligned} ρ ˙ e G ρ ˙ ee ρ ˙ GG = ( − i H 1 − κ † F q ) ρ e G , = ( − i H 1 − κ † F q ) ρ ee + ρ ee ( i H 1 − F q † κ ) , = F q ρ ee κ † + κ ρ ee F q † . (26)
For a normalized initial excitation u 0 u_0 u 0 , these equations imply
ρ t = ∣ G ( t ) u 0 ⟩ ⟨ G ( t ) u 0 ∣ + [ 1 − ∥ G ( t ) u 0 ∥ 2 ] ∣ G S ⟩ ⟨ G S ∣ . (27) \rho_t=|G(t)u_0\rangle\langle G(t)u_0|
+[1-\|G(t)u_0\|^2]|G_S\rangle\langle G_S|. \tag*{(27)} ρ t = ∣ G ( t ) u 0 ⟩ ⟨ G ( t ) u 0 ∣ + [ 1 − ∥ G ( t ) u 0 ∥ 2 ] ∣ G S ⟩ ⟨ G S ∣. (27)
For the coherent preparation ( ∣ G S ⟩ + ∣ u 0 ⟩ ) / 2 (|G_S\rangle+|u_0\rangle)/\sqrt{2} ( ∣ G S ⟩ + ∣ u 0 ⟩) / 2 , they instead give
ρ e G ( t ) = 1 2 G ( t ) u 0 . (28) \rho_{eG}(t)=\frac{1}{2}G(t)u_0. \tag*{(28)} ρ e G ( t ) = 2 1 G ( t ) u 0 . (28)
This is an unconditional system coherence obtained from the NMME. The reference superposition makes the readout linear in G G G , whereas populations involve amplitude products. These distinct maps explain why one coefficient degeneracy need not produce the same time dependence in every observable.
Internal memory coordinates and pole transfer
Define the row W = F q G W=F_qG W = F q G . Its derivative, using Eqs. (17) and (24), is
W ˙ = A 0 κ G − d W , W ( 0 ) = 0. \dot{W}=A_0\kappa G-dW,\qquad W(0)=0. W ˙ = A 0 κ G − d W , W ( 0 ) = 0.
Equivalently, W ( t ) = A 0 ∫ 0 t e − d ( t − s ) κ G ( s ) d s W(t)=A_0\int_0^t e^{-d(t-s)}\kappa G(s)\,\mathrm{d}s W ( t ) = A 0 ∫ 0 t e − d ( t − s ) κ G ( s ) d s . It is a mathematical coordinate for the amplitude history. The coupled propagation is therefore
d d t ( G W ) = A q ( G W ) , ( G ( 0 ) W ( 0 ) ) = ( I N q 0 ) , (29) \frac{\mathrm{d}}{\mathrm{d}t}\begin{pmatrix}G\\W\end{pmatrix}
=A_q\begin{pmatrix}G\\W\end{pmatrix},\qquad
\begin{pmatrix}G(0)\\W(0)\end{pmatrix}
=\begin{pmatrix}I_{N_q}\\0\end{pmatrix}, \tag*{(29)} d t d ( G W ) = A q ( G W ) , ( G ( 0 ) W ( 0 ) ) = ( I N q 0 ) , (29)
A q = ( − i H 1 − κ † A 0 κ − d ) . A_q=\begin{pmatrix}-iH_1&-\kappa^\dagger\\A_0\kappa&-d\end{pmatrix}. A q = ( − i H 1 A 0 κ − κ † − d ) .
This linear equation is derived from the NMQSD coefficients and state amplitudes. It also supplies a global continuation when a finite coefficient chart fails.
The coefficient and state-memory matrices satisfy
M q = J ( − A q T − d I N q + 1 ) J , J = diag ( I N q , − 1 ) , μ = − λ − d , (30) M_q=J(-A_q^{T}-dI_{N_q+1})J,\qquad
J=\operatorname{diag}(I_{N_q},-1),\qquad
\mu=-\lambda-d, \tag*{(30)} M q = J ( − A q T − d I N q + 1 ) J , J = diag ( I N q , − 1 ) , μ = − λ − d , (30)
where λ \lambda λ is an eigenvalue of A q A_q A q . Their Jordan orders therefore coincide, although their eigenvalues and late-time ordering differ. Both matrices have been obtained from the same NMQSD coefficients and state map. For Re λ < 0 \operatorname{Re}\lambda<0 Re λ < 0 , the matrix λ I + i H 1 \lambda I+iH_1 λ I + i H 1 is invertible, and the corresponding eigenspace of A q A_q A q is one dimensional. The prescribed coefficient initial condition retains the full pole order because the Laplace transform of q ( t ) q(t) q ( t ) , with complex frequency s s s , obeys
q ~ ( s ) = e a T ( s I − M q ) − 1 e a = det [ ( s + d ) I − i H 1 T ] det ( s I − M q ) (31) \widetilde{q}(s)=e_a^{T}(sI-M_q)^{-1}e_a
=\frac{\det[(s+d)I-iH_1^{T}]}{\det(sI-M_q)} \tag*{(31)} q ( s ) = e a T ( s I − M q ) − 1 e a = det ( s I − M q ) det [( s + d ) I − i H 1 T ] (31)
has a nonzero numerator at s = − λ − d s=-\lambda-d s = − λ − d . A positive weighted-norm identity in Appendix B shows that Re λ ≤ 0 \operatorname{Re}\lambda\leq0 Re λ ≤ 0 and that modes on the imaginary axis are dark and semisimple. Thus a defective pole in this class belongs to the decaying sector just considered.
If a reached order-r r r pole μ ∗ \mu_* μ ∗ is strictly dominant in real part, the common projective criterion gives
c q ( t ) = − μ ∗ − r − 1 t + O ( t − 2 ) . (32) c_q(t)=-\mu_*-\frac{r-1}{t}+O(t^{-2}). \tag*{(32)} c q ( t ) = − μ ∗ − t r − 1 + O ( t − 2 ) . (32)
The affine map implies
Re μ = − Re λ − γ . (33) \operatorname{Re}\mu=-\operatorname{Re}\lambda-\gamma. \tag*{(33)} Re μ = − Re λ − γ . (33)
The dominant coefficient pole therefore corresponds to faster amplitude decay. Equal-real-part poles or a reached but subleading EP require separate asymptotic analysis.
Let U q ( t ) = e A q t U_q(t)=e^{A_qt} U q ( t ) = e A q t , whose system block is G ( t ) G(t) G ( t ) . The inverse-transpose relation following from Eq. (30) gives
q ( t ) = e i tr ( H 1 ) t det G ( t ) . (34) q(t)=e^{i\operatorname{tr}(H_1)t}\det G(t). \tag*{(34)} q ( t ) = e i tr ( H 1 ) t det G ( t ) . (34)
On an invertible chart, F q = W G − 1 F_q=WG^{-1} F q = W G − 1 and f q ( t , s ) = κ G ( s ) G ( t ) − 1 f_q(t,s)=\kappa G(s)G(t)^{-1} f q ( t , s ) = κ G ( s ) G ( t ) − 1 . An isolated order-m m m zero of q q q at t 0 t_0 t 0 produces c q ∼ − m / ( t − t 0 ) c_q\sim-m/(t-t_0) c q ∼ − m / ( t − t 0 ) , while the global state propagation remains regular. Such coefficient-coordinate singularities are compatible with regular reduced dynamics [28 ] . They are distinct from spectral EPs; Appendix D separates chart poles, dark degeneracies, and subleading EPs.
The system readout must still be checked. For
D q ( z ) = z I + i H 1 , h q ( z ) = z + d + A 0 κ D q ( z ) − 1 κ † , (35) D_q(z)=zI+iH_1,\qquad
h_q(z)=z+d+A_0\kappa D_q(z)^{-1}\kappa^\dagger, \tag*{(35)} D q ( z ) = z I + i H 1 , h q ( z ) = z + d + A 0 κ D q ( z ) − 1 κ † , (35)
the Laplace transform of G ( t ) G(t) G ( t ) is the system block of the state-memory resolvent at complex frequency z z z :
G ~ ( z ) = D q − 1 − A 0 D q − 1 κ † h q − 1 κ D q − 1 . (36) \widetilde{G}(z)=D_q^{-1}-A_0D_q^{-1}\kappa^\dagger h_q^{-1}\kappa D_q^{-1}. \tag*{(36)} G ( z ) = D q − 1 − A 0 D q − 1 κ † h q − 1 κ D q − 1 . (36)
At a decaying order-r r r EP, h q h_q h q has an order-r r r zero at λ \lambda λ . The highest Laurent coefficient in this system block is
( R r ) e e = − A 0 r ! h q ( r ) ( λ ) D q ( λ ) − 1 κ † K D q ( λ ) − 1 ≠ 0. (37) (R_r)_{\mathrm{ee}}=-\frac{A_0r!}{h_q^{(r)}(\lambda)}D_q(\lambda)^{-1}\kappa^\dagger K D_q(\lambda)^{-1}\ne0. \tag*{(37)} ( R r ) ee = − h q ( r ) ( λ ) A 0 r ! D q ( λ ) − 1 κ † K D q ( λ ) − 1 = 0. (37)
Both factors are nonzero for the nonzero coupling required by the EP. Consequently, every defective decaying mode contributes t r − 1 e λ t t^{r-1}e^{\lambda t} t r − 1 e λ t to at least one system coherence for a suitable preparation. This establishes transfer of the full pole order to an unconditional readout. It does not require every quadrature to retain the term, as the EP3 quadratures in Figure 2 show.
Figure 2. Unconditional coherence from NMQSD and the exact NMME. The initial system state is ( ∣ G S ⟩ + ∣ 1 ⟩ ) / 2 (|G_S\rangle+ |1\rangle)/\sqrt{2} ( ∣ G S ⟩ + ∣1 ⟩) / 2 . Parameters are those of the coefficient EP3, Eq. (20) , with controls H 1 , 11 ↦ H 1 , 11 + ϵ H_{1,11} \mapsto H_{1,11} + \epsilon H 1 , 11 ↦ H 1 , 11 + ϵ at ϵ = ε / γ = ± 0.1 \epsilon= \varepsilon/\gamma= \pm0.1 ϵ = ε / γ = ± 0.1 . Here X 1 = 2 Re ρ 1 G X_1 = 2\operatorname{Re}\rho_{1G} X 1 = 2 Re ρ 1 G and Y 1 = 2 Im ρ 1 G Y_1 = 2\operatorname{Im}\rho_{1G} Y 1 = 2 Im ρ 1 G are the phase-referenced system-coherence quadratures. (a) Their ( X 1 , Y 1 ) (X_1,Y_1) ( X 1 , Y 1 ) trajectories. Arrows indicate increasing time; circles on the EP curve mark τ = γ t = 0 , 3 , 6 , 12 , 18 \tau= \gamma t = 0, 3, 6, 12, 18 τ = γ t = 0 , 3 , 6 , 12 , 18 . (b) Raw Y 1 ( t ) Y_1(t) Y 1 ( t ) . The state-response curves are compared on the EP trace with NMQSD coefficient propagation (circles) and direct NMME density propagation (squares). (c) Compensation gives e τ / 3 Y 1 ( t ) = − τ / 6 − τ 2 / 18 e^{\tau/3}Y_1(t) = -\tau/6 - \tau^2/18 e τ /3 Y 1 ( t ) = − τ /6 − τ 2 /18 . The dashed linear term and shaded quadratic contribution expose the nonzero third-order coherence pole; the raw signal decays. Here A 0 / γ 2 = 1 A_0/\gamma^2 = 1 A 0 / γ 2 = 1 , Ω / γ = 0 \Omega/\gamma= 0 Ω/ γ = 0 , and the laboratory carrier is ω 0 / γ = 2 \omega_0/\gamma= 2 ω 0 / γ = 2 . Numerical settings and discrepancies are given in Appendix § 1 .
EP3 in system quadratures
For the coefficient EP3 in Eq. (20) , the state-memory eigenvalue is λ E P = − γ / 3 \lambda_{\mathrm{EP}}=-\gamma/3 λ EP = − γ /3 . Prepare ( ∣ G S ⟩ + ∣ 1 ⟩ ) / 2 (|G_S\rangle+|1\rangle)/\sqrt{2} ( ∣ G S ⟩ + ∣1 ⟩) / 2 . The Hermitian observables
X ^ 1 = ∣ G S ⟩ ⟨ 1 ∣ + ∣ 1 ⟩ ⟨ G S ∣ , \hat{X}_1=|G_S\rangle\langle1|+|1\rangle\langle G_S|, X ^ 1 = ∣ G S ⟩ ⟨ 1∣ + ∣1 ⟩ ⟨ G S ∣ ,
Y ^ 1 = − i ∣ G S ⟩ ⟨ 1 ∣ + i ∣ 1 ⟩ ⟨ G S ∣ \hat{Y}_1=-i|G_S\rangle\langle1|+i|1\rangle\langle G_S| Y ^ 1 = − i ∣ G S ⟩ ⟨ 1∣ + i ∣1 ⟩ ⟨ G S ∣
have means X 1 = 2 Re ρ 1 G X_1=2\operatorname{Re}\rho_{1G} X 1 = 2 Re ρ 1 G and Y 1 = 2 Im ρ 1 G Y_1=2\operatorname{Im}\rho_{1G} Y 1 = 2 Im ρ 1 G . The NMQSD propagation and Eq. (28) give (see Appendix C for the amplitude derivation)
X 1 ( t ) = e − τ / 3 ( 1 + τ / 3 ) , X_1(t)=e^{-\tau/3}(1+\tau/3), X 1 ( t ) = e − τ /3 ( 1 + τ /3 ) ,
Y 1 ( t ) = − e − τ / 3 ( τ / 6 + τ 2 / 18 ) . (38) Y_1(t)=-e^{-\tau/3}(\tau/6+\tau^2/18). \tag*{(38)} Y 1 ( t ) = − e − τ /3 ( τ /6 + τ 2 /18 ) . (38)
Equivalently, the Laplace transform of C 1 = ρ 1 G C_1=\rho_{1G} C 1 = ρ 1 G is
C ~ 1 ( s ) = 1 2 ( s + γ / 3 ) + ( 2 − i ) γ 12 ( s + γ / 3 ) 2 − i γ 2 18 ( s + γ / 3 ) 3 . (39) \widetilde{C}_1(s)=\frac{1}{2(s+\gamma/3)}+\frac{(2-i)\gamma}{12(s+\gamma/3)^2}
-\frac{i\gamma^2}{18(s+\gamma/3)^3}. \tag*{(39)} C 1 ( s ) = 2 ( s + γ /3 ) 1 + 12 ( s + γ /3 ) 2 ( 2 − i ) γ − 18 ( s + γ /3 ) 3 i γ 2 . (39)
The same preparation thus gives a quadratic Jordan term in Y 1 Y_1 Y 1 and only a linear term in X 1 X_1 X 1 . The difference is a readout projection, not a change in the underlying EP. A common laboratory carrier ω 0 = 2 γ \omega_0=2\gamma ω 0 = 2 γ gives H 1 , l a b = H 1 + ω 0 I 2 H_{1,\mathrm{lab}}=H_1+\omega_0 I_2 H 1 , lab = H 1 + ω 0 I 2 and Ω l a b = ω 0 \Omega_{\mathrm{lab}}=\omega_0 Ω lab = ω 0 . The coefficient matrix is unchanged, while C 1 , l a b = e − i ω 0 t C 1 C_{1,\mathrm{lab}}=e^{-i\omega_0t}C_1 C 1 , lab = e − i ω 0 t C 1 ; a phase reference selects the quadratures plotted in Figure 2 .
The same preparation also allows the raw local selection
X 2 ( t ) = 2 Re ρ 2 G = − 3 27 τ 2 e − τ / 3 . (40) X_2(t)=2\operatorname{Re}\rho_{2G}=-\frac{\sqrt{3}}{27}\tau^2e^{-\tau/3}. \tag*{(40)} X 2 ( t ) = 2 Re ρ 2 G = − 27 3 τ 2 e − τ /3 . (40)
On the retained sector this is the mean of I ⊗ σ x I\otimes\sigma_x I ⊗ σ x . It isolates the quadratic term without compensation and reaches a maximum magnitude of approximately 0.31 at τ = 6 \tau=6 τ = 6 . Appendix I discusses phase selection and finite-signal diagnostics. The exact polynomial and the nonzero residue establish this signature; a short-time quadratic fit alone would not certify an EP.
Recovering coefficients from system readouts
For separate preparations ( ∣ G S ⟩ + ∣ j ⟩ ) / 2 (|G_S\rangle+|j\rangle)/\sqrt{2} ( ∣ G S ⟩ + ∣ j ⟩) / 2 , the phase-calibrated unconditional coherences determine each column:
G i j ( t ) = 2 ρ i G ( j ) ( t ) = X i ( j ) ( t ) + i Y i ( j ) ( t ) . (41) G_{ij}(t)=2\rho_{iG}^{(j)}(t)=X_i^{(j)}(t)+iY_i^{(j)}(t). \tag*{(41)} G ij ( t ) = 2 ρ i G ( j ) ( t ) = X i ( j ) ( t ) + i Y i ( j ) ( t ) . (41)
On an invertible chart, Eq. (24) gives
F q = − κ ( G G ˙ − 1 + i H 1 ) κ κ ∗ , (42) F_q=-\frac{\kappa(G\dot{G}^{-1}+iH_1)}{\kappa\kappa^*}, \tag*{(42)} F q = − κ κ ∗ κ ( G G ˙ − 1 + i H 1 ) , (42)
c q = − tr ( G G ˙ − 1 ) − i tr H 1 . (43) c_q=-\operatorname{tr}(G\dot{G}^{-1})-i\operatorname{tr}H_1. \tag*{(43)} c q = − tr ( G G ˙ − 1 ) − i tr H 1 . (43)
The row reconstruction requires known H 1 H_1 H 1 and known nonzero κ \kappa κ in a fixed normalization and phase convention; the trace formula needs only known tr H 1 \operatorname{tr}H_1 tr H 1 . Under a common laboratory rotation, G l a b = e − i ω 0 t G r o t G_{\mathrm{lab}}=e^{-i\omega_0t}G_{\mathrm{rot}} G lab = e − i ω 0 t G rot , H 1 , l a b = H 1 , r o t + ω 0 I N q H_{1,\mathrm{lab}}=H_{1,\mathrm{rot}}+\omega_0I_{N_q} H 1 , lab = H 1 , rot + ω 0 I N q , and Ω l a b = Ω + ω 0 \Omega_{\mathrm{lab}}=\Omega+\omega_0 Ω lab = Ω + ω 0 . The determinant-phase derivative and the trace shift cancel in Eq. (43) . This application of evolution-map inversion [28 ] identifies the integrated coefficients of the specified NMQSD representation. Different readouts require separately prepared ensembles. Differentiation and small singular values of G G G can amplify errors even while the physical state stays regular. For the two-qubit example, both basis preparations are needed to recover the complete map; the selected X 2 X_2 X 2 record alone does not determine the full coefficient row. The interval reconstruction and conditioning checks are given in Appendix I3.
Position-coupled oscillator in an optomechanical sector
Fixed-sector closure and Gaussian propagation
Position coupling requires a noise-dependent Q Q Q operator even for a harmonic oscillator. This makes the model complementary to the qubits: a coefficient branch that is subleading in the causal solution can still enter the state through its width, and a narrow conditional state can become broad after averaging over its centers. We retain the full noise-history term to follow both effects.
We use dimensionless mechanical quadratures with [ Q , P ] = i [Q,P]=i [ Q , P ] = i and vacuum covariance I 2 / 2 I_2/2 I 2 /2 . The system Hamiltonian and bath coupling operator are
H S = ν 2 ( Q 2 + P 2 ) + η N Q + H N , L = ℓ Q . (44) H_S=\frac{\nu}{2}(Q^2+P^2)+\eta NQ+H_N,\qquad L=\ell Q. \tag*{(44)} H S = 2 ν ( Q 2 + P 2 ) + η N Q + H N , L = ℓ Q . (44)
For the optomechanical realization, let a c a_c a c be a cavity annihilation operator, N = a c † a c N=a_c^\dagger a_c N = a c † a c , and H N = ω c N H_N=\omega_cN H N = ω c N . The mechanical position is Q Q Q , and η N Q \eta NQ η N Q is the radiation-pressure interaction in the chosen sign convention [29 ] . The photon-number dependence of the mechanical force is the familiar nonlinear optomechanical coupling used in state-preparation models [30 ] . Here [ N , Q ] = [ N , P ] = 0 [N,Q]=[N,P]=0 [ N , Q ] = [ N , P ] = 0 and the photon number is conserved. We restrict the system preparation to one photon-number sector, N ∣ n ⟩ = n ∣ n ⟩ N|n\rangle=n|n\rangle N ∣ n ⟩ = n ∣ n ⟩ . No cavity drive or optical loss is included in this fixed-sector model.
We fix n > 0 n>0 n > 0 and write its constant energy as E n E_n E n . The oscillator frequency ν \nu ν , coupling scale ℓ \ell ℓ , and force coefficient η \eta η are positive. Within this sector, radiation pressure is an affine force. The coefficient core and its exceptional-point locus below are independent of η \eta η and $n; the result does not characterize driven cavity–mechanical hybrid-mode exceptional points.
The initial oscillator state is a normalized pure Gaussian, independent of the vacuum Gaussian bath. We use the common correlation in Eq. (2) , now with A 0 , γ , Ω > 0 A_0,\gamma,\Omega> 0 A 0 , γ , Ω > 0 , and define B = A 0 ℓ 2 B = A_0\ell^2 B = A 0 ℓ 2 . Position coupling retains both rotating and counter-rotating terms, placing the oscillator in the quantum Brownian-motion setting [22 , 31 ] . The exact NMQSD and consistency equations are Eqs. (1) and (4) .
For Eq. (44) , the operator algebra closes on Q Q Q , P P P , N N N , I \mathbb{I} I . The noise-complete oscillator O O O operator of Ref. [18 ] extends to this conserved affine sector. In particular, [ ℓ Q z t ∗ , g O P ] = i ℓ g O z t ∗ I [\ell Qz_t^*,g_O P]=i\ell g_O z_t^*\mathbb{I} [ ℓ Q z t ∗ , g O P ] = i ℓ g O z t ∗ I requires the same scalar noise term:
O ( t , s , z ∗ ) = f ( t , s ) Q + g O ( t , s ) P + r O ( t , s ) N + i ∫ 0 t j ( t , s , u ) z u ∗ d u I . (45) O(t,s,z^*)=f(t,s)Q+g_O(t,s)P+r_O(t,s)N+i\int_0^t j(t,s,u)z_u^*\,du\,\mathbb{I}. \tag*{(45)} O ( t , s , z ∗ ) = f ( t , s ) Q + g O ( t , s ) P + r O ( t , s ) N + i ∫ 0 t j ( t , s , u ) z u ∗ d u I . (45)
Writing O ˉ t = F Q + G P + R N + i Ξ t I \bar O_t=FQ+GP+RN+i\Xi_t\mathbb{I} O ˉ t = F Q + GP + R N + i Ξ t I , define F ( t ) = ∫ 0 t α ( t , s ) f ( t , s ) d s F(t)=\int_0^t\alpha(t,s)f(t,s)\,ds F ( t ) = ∫ 0 t α ( t , s ) f ( t , s ) d s , with G ( t ) G(t) G ( t ) and R ( t ) R(t) R ( t ) given by the same integral of g O ( t , s ) g_O(t,s) g O ( t , s ) and r O ( t , s ) r_O(t,s) r O ( t , s ) . The integrated scalar noise history is Ξ t = ∫ 0 t J ( t , u ) z u ∗ d u \Xi_t=\int_0^t J(t,u)z_u^*\,du Ξ t = ∫ 0 t J ( t , u ) z u ∗ d u , where J ( t , u ) = ∫ 0 t α ( t , s ) j ( t , s , u ) d s J(t,u)=\int_0^t\alpha(t,s)j(t,s,u)\,ds J ( t , u ) = ∫ 0 t α ( t , s ) j ( t , s , u ) d s . Its deterministic bath integral J ( t ) = ∫ 0 t α ( t , u ) J ( t , u ) d u J(t)=\int_0^t\alpha(t,u)J(t,u)\,du J ( t ) = ∫ 0 t α ( t , u ) J ( t , u ) d u enters F ˙ \dot F F ˙ through the functional-derivative contribution − i ℓ J -i\ell J − i ℓ J . The ( F , G , J ) (F,G,J) ( F , G , J ) equations are the exponential-kernel oscillator equations of Ref. [18 ] , here with the complex decay d d d and the affine coefficient R R R :
F ˙ = A 0 ℓ − d F + ν G − i ℓ F G − i ℓ J , G ˙ = − d G − ν F − i ℓ G 2 , J ˙ = A 0 ℓ G − 2 d J − i ℓ G J , R ˙ = − d R + η G − i ℓ G R , (46) \begin{aligned}
\dot F&=A_0\ell-dF+\nu G-i\ell FG-i\ell J,\\
\dot G&=-dG-\nu F-i\ell G^2,\\
\dot J&=A_0\ell G-2dJ-i\ell GJ,\\
\dot R&=-dR+\eta G-i\ell GR, \tag*{(46)}
\end{aligned} F ˙ G ˙ J ˙ R ˙ = A 0 ℓ − d F + ν G − i ℓ F G − i ℓ J , = − d G − ν F − i ℓ G 2 , = A 0 ℓ G − 2 dJ − i ℓ G J , = − d R + η G − i ℓ GR , (46)
with F ( 0 ) = G ( 0 ) = J ( 0 ) = R ( 0 ) = 0 F(0)=G(0)=J(0)=R(0)=0 F ( 0 ) = G ( 0 ) = J ( 0 ) = R ( 0 ) = 0 . Appendix § 1.2.5 gives the two-time equations and both noise boundaries. This closure uses no weak-coupling, rotating-wave, or noise-order truncation. A Gaussian system preparation is needed for the state representation below, but not for the operator closure. The scalar Ξ ˉ t \bar\Xi_t Ξ ˉ t still multiplies Q Q Q in Eq. (1) , so it changes the displacement rather than merely the overall phase.
The deterministic core ( F , G , J ) (F,G,J) ( F , G , J ) fixes the coefficient spectrum, while R R R carries the affine force. The scalar history remains necessary for the displacement and for the operator average.
Write the unnormalized stochastic wavefunction as ψ t ( x ; z ∗ ) = exp [ − a t x 2 / 2 + b t x + c ψ , t ] \psi_t(x;z^*)=\exp[-a_t x^2/2+b_t x+c_{\psi,t}] ψ t ( x ; z ∗ ) = exp [ − a t x 2 /2 + b t x + c ψ , t ] , with a R = Re a > 0 a_R=\operatorname{Re}a>0 a R = Re a > 0 and a I = Im a a_I=\operatorname{Im}a a I = Im a . Substitution into Eq. (1) yields
a ˙ = i ν ( 1 − a 2 ) + 2 ℓ F + 2 i ℓ G a , b ˙ = ( i ℓ G − i ν a ) b + ℓ z t ∗ − i ℓ Ξ t − n ( i η + ℓ R ) , c ˙ ψ = i ν 2 ( b 2 − a ) − i Ξ t n , Ξ ˙ = ℓ G z t ∗ − ( d + i ℓ G ) Ξ , Ξ ( 0 ) = 0. (47) \begin{aligned}
\dot a&=i\nu(1-a^2)+2\ell F+2i\ell Ga,\\
\dot b&=(i\ell G-i\nu a)b+\ell z_t^*-i\ell\Xi_t-n(i\eta+\ell R),\\
\dot c_\psi&=\frac{i\nu}{2}(b^2-a)-i\Xi_t n,\\
\dot\Xi&=\ell Gz_t^*-(d+i\ell G)\Xi,\qquad\Xi(0)=0. \tag*{(47)}
\end{aligned} a ˙ b ˙ c ˙ ψ Ξ ˙ = i ν ( 1 − a 2 ) + 2 ℓ F + 2 i ℓ G a , = ( i ℓ G − i ν a ) b + ℓ z t ∗ − i ℓ Ξ t − n ( i η + ℓ R ) , = 2 i ν ( b 2 − a ) − i Ξ t n , = ℓ G z t ∗ − ( d + i ℓ G ) Ξ , Ξ ( 0 ) = 0. (47)
We order the quadratures as r = ( Q , P ) T r=(Q,P)^{T} r = ( Q , P ) T and use the mean m = ⟨ r ⟩ m=\langle r\rangle m = ⟨ r ⟩ and symmetrized covariance V i j = ⟨ { r i − m i , r j − m j } ⟩ / 2 V_{ij}=\langle\{r_i-m_i,r_j-m_j\}\rangle/2 V ij = ⟨{ r i − m i , r j − m j }⟩ /2 for a normalized state [32 ] . For the normalized stochastic Gaussian, these are
m ψ = T a ( Re b Im b ) , T a = ( 1 / a R 0 − a I / a R 1 ) , (48) m_{\psi}=T_a\begin{pmatrix}\operatorname{Re}b\\ \operatorname{Im}b\end{pmatrix},\qquad T_a=\begin{pmatrix}1/a_R&0\\-a_I/a_R&1\end{pmatrix}, \tag*{(48)} m ψ = T a ( Re b Im b ) , T a = ( 1/ a R − a I / a R 0 1 ) , (48)
V ψ = 1 2 a R ( 1 − a I − a I ∣ a ∣ 2 ) . (49) V_{\psi}=\frac{1}{2a_R}\begin{pmatrix}1&-a_I\\-a_I&|a|^2\end{pmatrix}. \tag*{(49)} V ψ = 2 a R 1 ( 1 − a I − a I ∣ a ∣ 2 ) . (49)
The width is deterministic, while the displacement responds linearly to the noise history. The scalar wavefunction coefficient c ψ c_{\psi} c ψ determines the squared norm used in the average, even though it cancels from the normalized moments in Eqs. (48) and (49) . It is distinct from the scalar-noise sector Ξ \Xi Ξ of the O O O operator. The sector energy E n E_n E n contributes only a phase.
Coefficient EPs and the two-root state response
The causal coefficient solution
The nonlinear subsystem ( F , G , J ) (F,G,J) ( F , G , J ) has a four-dimensional linear representation:
v ˙ = M o v , v ( 0 ) = e 4 , M o = ( − d ν − i ℓ A 0 ℓ − ν − d 0 0 0 A 0 ℓ − 2 d 0 0 i ℓ 0 0 ) . (50) \dot{v}=M_o v,\qquad v(0)=e_4,\qquad M_o=\begin{pmatrix}-d&\nu&-i\ell&A_0\ell\\-\nu&-d&0&0\\0&A_0\ell&-2d&0\\0&i\ell&0&0\end{pmatrix}. \tag*{(50)} v ˙ = M o v , v ( 0 ) = e 4 , M o = − d − ν 0 0 ν − d A 0 ℓ i ℓ − i ℓ 0 − 2 d 0 A 0 ℓ 0 0 0 . (50)
Here ( F , G , J ) = ( x F , x G , x J ) / q (F,G,J)=(x_F,x_G,x_J)/q ( F , G , J ) = ( x F , x G , x J ) / q , v = ( x F , x G , x J , q ) T v=(x_F,x_G,x_J,q)^T v = ( x F , x G , x J , q ) T , and e 4 = ( 0 , 0 , 0 , 1 ) T e_4=(0,0,0,1)^T e 4 = ( 0 , 0 , 0 , 1 ) T is the fourth unit vector. The characteristic polynomial and its shifted discriminant are
P a ( μ ) = μ ( μ + 2 d ) [ ( μ + d ) 2 + ν 2 ] + 2 i B ν d , (51) P_a(\mu)=\mu(\mu+2d)\left[(\mu+d)^2+\nu^2\right]+2iB\nu d, \tag*{(51)} P a ( μ ) = μ ( μ + 2 d ) [ ( μ + d ) 2 + ν 2 ] + 2 i B ν d , (51)
Δ a = ( ν 2 + d 2 ) 2 − 8 i B ν d . (52) \Delta_a=(\nu^2+d^2)^2-8iB\nu d. \tag*{(52)} Δ a = ( ν 2 + d 2 ) 2 − 8 i B ν d . (52)
Their positive-parameter EP branch is
Ω a 2 = ν 2 − γ 2 + 2 ν 4 + ν 2 γ 2 + γ 4 3 , \Omega_a^2=\frac{\nu^2-\gamma^2+2\sqrt{\nu^4+\nu^2\gamma^2+\gamma^4}}{3}, Ω a 2 = 3 ν 2 − γ 2 + 2 ν 4 + ν 2 γ 2 + γ 4 ,
B a = Ω a ( ν 2 + γ 2 − Ω a 2 ) 2 ν . (53) B_a=\frac{\Omega_a(\nu^2+\gamma^2-\Omega_a^2)}{2\nu}. \tag*{(53)} B a = 2 ν Ω a ( ν 2 + γ 2 − Ω a 2 ) . (53)
In this oscillator section, λ = B / B a \lambda=B/B_a λ = B / B a denotes the dimensionless coupling ratio. Away from coalescence, define the two shifted roots with positive real parts,
w 1 , 2 = d 2 − ν 2 ± Δ a 2 , Re w 1 , 2 > 0. (54) w_{1,2}=\sqrt{\frac{d^2-\nu^2\pm\sqrt{\Delta_a}}{2}},\qquad\operatorname{Re}w_{1,2}>0. \tag*{(54)} w 1 , 2 = 2 d 2 − ν 2 ± Δ a , Re w 1 , 2 > 0. (54)
The four eigenvalues μ = − d ± w 1 , − d ± w 2 \mu=-d\pm w_1,-d\pm w_2 μ = − d ± w 1 , − d ± w 2 give four finite algebraic fixed points away from coalescence. Figure 3 follows the two positive-shift branches, G ∗ ( j ) = ( − d + w j ) / ( i ℓ ) G_*^{(j)}=(-d+w_j)/(i\ell) G ∗ ( j ) = ( − d + w j ) / ( i ℓ ) . The causal trajectory G ( t ) = q ˙ ( t ) / ( i ℓ q ( t ) ) G(t)=\dot{q}(t)/(i\ell q(t)) G ( t ) = q ˙ ( t ) / ( i ℓ q ( t )) approaches the branch with larger Re w j \operatorname{Re}w_j Re w j for simple roots of unequal real parts, provided the prescribed initial vector reaches that component and the ratio chart is regular. Other reached components remain at finite age. State reconstruction also requires the Gaussian width.
Figure 3. Coefficient branches and the two channels of the stochastic-state response. (a),(b) The two positive-shift fixed-point branches G ∗ ( j ) = ( − d + w j ) / ( i ℓ ) G_*^{(j)}=(-d+w_j)/(i\ell) G ∗ ( j ) = ( − d + w j ) / ( i ℓ ) along λ = B / B a \lambda=B/B_a λ = B / B a at Ω = Ω a \Omega=\Omega_a Ω = Ω a coalesce at λ = 1 \lambda=1 λ = 1 . The subscript * denotes a fixed point. The solid branch is selected by the causal zero initial condition; the dashed companion still enters the state response through the second root. Right axes show the corresponding response poles p = − w 1 , 2 p=-w_{1,2} p = − w 1 , 2 on the same curves. (c) Real (solid) and imaginary (dashed) parts of the full kernel h ( θ ) h(\theta) h ( θ ) at λ = 0.7 , 1 , 1.3 \lambda=0.7,1,1.3 λ = 0.7 , 1 , 1.3 , identified respectively by sparse circles, squares, and triangles. At coalescence, the response contains both e − w θ e^{-w\theta} e − w θ and θ e − w θ \theta e^{-w\theta} θ e − w θ . The horizontal axis is the dimensionless lag ν θ \nu\theta ν θ . This is the fixed-lag response after a long preparation age, with 0 ≤ ν θ ≤ 20 0\leq\nu\theta\leq20 0 ≤ ν θ ≤ 20 , rather than the finite-age evolution shown in Sec. IV C 2 and Appendix J. Parameters are ν = ℓ = 1 \nu=\ell=1 ν = ℓ = 1 , γ = 0.5 \gamma=0.5 γ = 0.5 , η = 0.25 \eta=0.25 η = 0.25 , and n = 1 n=1 n = 1 .
At the EP, put w 2 = ( d 2 − ν 2 ) / 2 w^2=(d^2-\nu^2)/2 w 2 = ( d 2 − ν 2 ) /2 with Re w > 0 \operatorname{Re}w>0 Re w > 0 . Each of μ ± = − d ± w \mu_{\pm}=-d\pm w μ ± = − d ± w has algebraic multiplicity two and geometric multiplicity one. The initial vector e 4 e_4 e 4 reaches the leading Jordan chain, giving
G ( t ) = μ + i ℓ + 1 i ℓ t + O ( t − 2 ) + O [ poly ( t ) e − 2 Re w t ] . (55) G(t)=\frac{\mu_+}{i\ell}+\frac{1}{i\ell t}+O(t^{-2})+O\left[\operatorname{poly}(t)e^{-2\operatorname{Re}w\,t}\right]. \tag*{(55)} G ( t ) = i ℓ μ + + i ℓ t 1 + O ( t − 2 ) + O [ poly ( t ) e − 2 Re w t ] . (55)
Appendix F proves the positive branch and its causal reachability. A zero of q q q is a pole of this coefficient chart; the homogeneous evolution in Eq. (50) remains regular.
Width selection and displacement response
In the EP neighborhood, label the simple roots locally so that Re w 1 > Re w 2 > 0 \operatorname{Re}w_1>\operatorname{Re}w_2>0 Re w 1 > Re w 2 > 0 , and set μ 1 = w 1 − d \mu_1=w_1-d μ 1 = w 1 − d . The selected coefficient solution approaches
G ∞ = μ 1 i ℓ , F ∞ = i μ 1 ( μ 1 + d ) ν ℓ . (56) G_{\infty}=\frac{\mu_1}{i\ell},\qquad F_{\infty}=\frac{i\mu_1(\mu_1+d)}{\nu\ell}. \tag*{(56)} G ∞ = i ℓ μ 1 , F ∞ = ν ℓ i μ 1 ( μ 1 + d ) . (56)
Since w 1 2 + w 2 2 = d 2 − ν 2 w_1^2+w_2^2=d^2-\nu^2 w 1 2 + w 2 2 = d 2 − ν 2 , the stable normalizable width continued from the EP is
a ∞ = i ( d − w 1 − w 2 ) ν . (57) a_{\infty}=\frac{i(d-w_1-w_2)}{\nu}. \tag*{(57)} a ∞ = ν i ( d − w 1 − w 2 ) . (57)
Its linearization rate is − 2 w 2 -2w_2 − 2 w 2 ; the other width root has rate + 2 w 2 +2w_2 + 2 w 2 . Thus the companion coefficient root appears through the width, while the selected root remains in the noise-history channel. Comparing histories at the same width preparation and the same nonzero affine force removes the common force term. For a history input at time s s s and endpoint T T T , let θ = T − s ≥ 0 \theta=T-s\geq0 θ = T − s ≥ 0 be the lag. In the long-preparation limit T → ∞ T\to\infty T → ∞ at fixed θ \theta θ , the variations satisfy
d d θ ( δ b δ Ξ ) = A ∞ ( δ b δ Ξ ) + B ∞ δ z ∗ , \frac{d}{d\theta}\begin{pmatrix}\delta b\\\delta\Xi\end{pmatrix}=A_{\infty}\begin{pmatrix}\delta b\\\delta\Xi\end{pmatrix}+B_{\infty}\delta z^*, d θ d ( δ b δ Ξ ) = A ∞ ( δ b δ Ξ ) + B ∞ δ z ∗ ,
A ∞ = ( − w 2 − i ℓ 0 − w 1 ) , B ∞ = ( ℓ − i μ 1 ) . (58) A_{\infty}=\begin{pmatrix}-w_2&-i\ell\\0&-w_1\end{pmatrix},\qquad B_{\infty}=\begin{pmatrix}\ell\\-i\mu_1\end{pmatrix}. \tag*{(58)} A ∞ = ( − w 2 0 − i ℓ − w 1 ) , B ∞ = ( ℓ − i μ 1 ) . (58)
For zero initial response, the Laplace transform in θ \theta θ , with complex variable p p p , gives the transfer function
H b z ( p ) = ℓ ( p + d ) ( p + w 1 ) ( p + w 2 ) . (59) H_{bz}(p)=\frac{\ell(p+d)}{(p+w_1)(p+w_2)}. \tag*{(59)} H b z ( p ) = ( p + w 1 ) ( p + w 2 ) ℓ ( p + d ) . (59)
The direct noise input and the history stored in Ξ \Xi Ξ both reach b b b . For B ν > 0 B\nu>0 B ν > 0 and d ≠ 0 d\ne0 d = 0 , neither root equals d d d , so both residues are nonzero away from coalescence. Combining the two poles before taking their coalescence gives a regular expression for the complete kernel,
h ( θ ) = ℓ e − w ˉ θ [ cosh ( ϵ w θ ) + ( d − w ˉ ) θ sinh ( ϵ w θ ) ] , h(\theta)=\ell e^{-\bar{w}\theta}\left[\cosh(\epsilon_w\theta)+(d-\bar{w})\theta\sinh(\epsilon_w\theta)\right], h ( θ ) = ℓ e − w ˉ θ [ cosh ( ϵ w θ ) + ( d − w ˉ ) θ sinh ( ϵ w θ ) ] ,
w ˉ = ( w 1 + w 2 ) / 2 , ϵ w = ( w 1 − w 2 ) / 2. (60) \bar{w}=(w_1+w_2)/2,\qquad\epsilon_w=(w_1-w_2)/2. \tag*{(60)} w ˉ = ( w 1 + w 2 ) /2 , ϵ w = ( w 1 − w 2 ) /2. (60)
where sinhc x = sinh ( x ) / x \operatorname{sinhc} x = \sinh(x)/x sinhc x = sinh ( x ) / x and sinhc 0 = 1 \operatorname{sinhc} 0 = 1 sinhc 0 = 1 . At the EP,
h E P ( θ ) = ℓ [ 1 + ( d − w ) θ ] e − w θ . (61) h_{\mathrm{EP}}(\theta)=\ell[1+(d-w)\theta]e^{-w\theta}. \tag*{(61)} h EP ( θ ) = ℓ [ 1 + ( d − w ) θ ] e − w θ . (61)
The nonzero coefficient ℓ ( d − w ) \ell(d-w) ℓ ( d − w ) gives a double pole in the stochastic displacement response. The polynomial–exponential form is the familiar time-domain consequence of a higher-order pole [33 ] ; here its residue is fixed by the causal preparation and the specified history-to-displacement map. Through T a T_a T a in Eq. (48) , this kernel quantifies how a variation δ z ∗ \delta z^* δ z ∗ of the NMQSD Bargmann history changes the normalized conditional centroid. A force susceptibility requires an applied-force input, while bath-averaged two-time correlations require their own averaging construction. Such correlations can be obtained by propagating paired trajectories [34 ] ; the optomechanical treatment in Ref. [35] uses a common phononic history with optical no-jump evolution.
Along Ω = Ω a \Omega= \Omega_a Ω = Ω a , the splitting in the coupling ratio is
w 1 − w 2 = − 8 i B a ν d ( λ − 1 ) 2 w + O ( ∣ λ − 1 ∣ 3 / 2 ) . (62) w_1-w_2=\frac{\sqrt{-8\mathrm{i}B_a\nu d(\lambda-1)}}{2w}+O\left(\lvert\lambda-1\rvert^{3/2}\right). \tag*{(62)} w 1 − w 2 = 2 w − 8 i B a ν d ( λ − 1 ) + O ( ∣ λ − 1 ∣ 3/2 ) . (62)
The roots split as a square root, whereas Eq. (60) is analytic in λ \lambda λ near λ = 1 \lambda= 1 λ = 1 at every fixed finite lag. Figure 3 shows both the branch geometry and the damped complex response.
The width has a different late-time consequence. On the regular normalizable branch at the EP, for 0 < γ < ν 0 < \gamma< \nu 0 < γ < ν ,
a ∗ = i ( d − 2 w ) ν , a ( t ) − a ∗ = O ( t 2 e − 2 Re w t ) . (63) a_*=\frac{\mathrm{i}(d-2w)}{\nu},\qquad a(t)-a_*=O\left(t^2e^{-2\operatorname{Re}w t}\right). \tag*{(63)} a ∗ = ν i ( d − 2 w ) , a ( t ) − a ∗ = O ( t 2 e − 2 Re w t ) . (63)
The purely algebraic coefficient tail cancels from the width equation; polynomially modified exponential corrections remain. The interval 0 < γ < ν 0 < \gamma< \nu 0 < γ < ν places the bath correlation decay rate below the oscillator frequency and is the working interval used here together with the regular normalizable branch condition. Appendix § 1.2.6 proves that the stationary width has Re a ∗ > 0 \operatorname{Re}a_* > 0 Re a ∗ > 0 for all positive γ , Ω , ν \gamma,\Omega,\nu γ , Ω , ν ; this algebraic positivity alone does not establish finite-time normalizability. On the regular branch, preservation of the mean squared norm in linear NMQSD excludes the nonnormalizable asymptotic solution and gives Eq. (63) . The finite-time reduced state requires the weighted average developed next.
Complete noise average and finite-time state dynamics
The common average in Eq. (5) can be evaluated directly for the Gaussian state. The reference measure P 0 \mathbb{P}_0 P 0 averages unnormalized projectors, whereas the norm-weighted measure P t \mathbb{P}_t P t averages normalized-state observables. Because c ψ c_\psi c ψ determines that weight, it must be kept even when the desired observable depends only on a a a and b b b .
The Gaussian ensemble and its norm weight
Applying the same mean and symmetrized-covariance convention to the reduced state, the norm-weighted ensemble of the conditional moments in Eqs. (48) and (49) gives
m = E P t [ m ψ ] , m=\mathbb{E}_{\mathbb{P}_t}[m_\psi], m = E P t [ m ψ ] ,
V = V ψ + V c e n t r o i d , (64) V=V_\psi+V_{\mathrm{centroid}}, \tag*{(64)} V = V ψ + V centroid , (64)
V c e n t r o i d = Cov P t ( m ψ ) . V_{\mathrm{centroid}}=\operatorname{Cov}_{\mathbb{P}_t}(m_\psi). V centroid = Cov P t ( m ψ ) .
Because the Gaussian width is independent of the noise labels, the first covariance contribution is the same for every normalized stochastic state. The second measures the spread of their centers under the norm-weighted measure. A reduction of a conditional quadrature variance therefore does not determine the corresponding reduced-state variance, as Figure 4 illustrates for the preparations in Sec. IV C 2.
Figure 4. Conditional width and centroid spread resolve the reduced-state variance. (a) conditional covariance V ψ V_\psi V ψ ; (b) centroid covariance V c e n t r o i d = Cov P t ( m ψ ) V_{\mathrm{centroid}} = \operatorname{Cov}_{P_t}(m_\psi) V centroid = Cov P t ( m ψ ) ; (c) their relation to the total variance at ν t = 8 \nu t = 8 ν t = 8 . In (a),(b), j = Q , P j = Q, P j = Q , P : solid and dashed lines denote Q Q Q and P P P , respectively; all time-series markers are hollow. Color and marker shape identify the three fixed couplings. Each row in (c) identifies one quadrature and coupling; a gray filled symbol marks the conditional variance and a hollow symbol marks the total. Their horizontal separation equals the centroid contribution. The rows are discrete cases, with no interpolation between them. Faint reference lines mark the initial vacuum variance 1 / 2 1/2 1/2 . Across the three momentum rows, increasing λ \lambda λ moves the conditional endpoint left and the total endpoint right: centroid spread outweighs conditional narrowing. Parameters and the common Gaussian initial state are specified in Sec. IV C 2.
For the numerical average, the finite Gaussian source in Appendix K makes the displacement affine and the scalar c ψ c_\psi c ψ quadratic in the source coordinates. Appendix H evaluates the resulting Gaussian norm integral and moments, retaining c ψ c_\psi c ψ and checking the absolute total weight Z Z Z against its exact value one. Source refinement and agreement with the effective master equation are reported in Appendix K.
Finite-time reduced-state dynamics
We compare the coefficient EP and its two sides at fixed ν = ℓ = 1 \nu=\ell=1 ν = ℓ = 1 , γ = 0.5 \gamma=0.5 γ = 0.5 , η = 0.25 \eta=0.25 η = 0.25 , n = 1 n=1 n = 1 , and Ω = Ω a \Omega=\Omega_a Ω = Ω a . The three couplings B = λ B a B=\lambda B_a B = λ B a , with λ = 0.7 , 1 , 1.3 \lambda=0.7,1,1.3 λ = 0.7 , 1 , 1.3 , share the normalized initial wavefunction a ( 0 ) = 1 a(0)=1 a ( 0 ) = 1 , b ( 0 ) = 0 b(0)=0 b ( 0 ) = 0 , c ψ ( 0 ) = − log π / 4 c_\psi(0)=-\log\pi/4 c ψ ( 0 ) = − log π /4 , so that m ( 0 ) = 0 m(0)=0 m ( 0 ) = 0 and V ( 0 ) = I 2 / 2 V(0)=\mathbb{I}_2/2 V ( 0 ) = I 2 /2 . We follow the full nonautonomous equations over 0 ≤ ν t ≤ 8 0\leq\nu t\leq8 0 ≤ ν t ≤ 8 . The spectral coalescence is identified analytically by Eq. (53) ; these three trajectories describe the state evolution at the specified couplings.
The variance decomposition in Figure 4 separates the width of a normalized stochastic pure state from the spread of its center in the norm-weighted ensemble. Panels (a) and (b) follow these two contributions on the same time grid as the total variances reported in Appendix J. Their sum is the reduced-state variance in Eq. (64) ; a narrower conditional state therefore need not imply a narrower reduced state.
At the common endpoint ν t = 8 \nu t=8 ν t = 8 , increasing the coupling across the three examples narrows the conditional momentum distribution, while the spread of its centers grows enough to raise the total momentum variance. At λ = 1 \lambda=1 λ = 1 , for example, the conditional contribution is about 0.48, below the initial vacuum value 1/2, but a centroid contribution of about 0.08 raises the total to 0.56.
At finite times, the combined state can remain smooth in the parameters on a regular coefficient and averaging chart, even though the labeled roots split as a square root. The three fixed-coupling trajectories describe this finite-time averaged evolution; the analytic EP result concerns the coalescence of the two stochastic history-response channels and their nonzero Jordan term at long preparation age and fixed lag.
The effective master equation
Averaging Eq. (1) with A t = E 0 [ O ˉ t Π t ] A_t=\mathbb{E}_0[\bar{O}_t\Pi_t] A t = E 0 [ O ˉ t Π t ] gives the formal master equation [22 , 23 ]
ρ ˙ t = − i [ H S , ρ t ] + [ L , A t † ] − [ L † , A t ] . (65) \dot{\rho}_t=-i[H_S,\rho_t]+[L,A_t^\dagger]-[L^\dagger,A_t]. \tag*{(65)} ρ ˙ t = − i [ H S , ρ t ] + [ L , A t † ] − [ L † , A t ] . (65)
Following the paired operator-average construction of Strunz and Yu [22 ] , resolve A t A_t A t into left and right actions,
A t = ∑ k ( U k X k ρ t + V k ρ t X k ) , A_t=\sum_k(U_kX_k\rho_t+V_k\rho_tX_k), A t = k ∑ ( U k X k ρ t + V k ρ t X k ) ,
with X k ∈ { Q , P , N } X_k\in\{Q,P,N\} X k ∈ { Q , P , N } . On the regular averaging chart specified in Appendix § 2.9 , the effective coefficients are
f = U Q − V Q ∗ , g = U P − V P ∗ , r = U N − V N ∗ . (66) f=U_Q-V_Q^*,\qquad g=U_P-V_P^*,\qquad r=U_N-V_N^*. \tag*{(66)} f = U Q − V Q ∗ , g = U P − V P ∗ , r = U N − V N ∗ . (66)
The coefficients U k , V k U_k,V_k U k , V k are the bath-weighted integrals of the left/right components in Eq. (105) . They retain scalar-noise correlations, so the effective coefficients generally differ from F , G , R F,G,R F , G , R . Appendix § 2.9 derives the paired integral inverse for the present exponential kernel and affine sector, including its pointwise-conjugation convention and invertibility condition.
Thus Im f \operatorname{Im} f Im f changes the restoring force, Im g \operatorname{Im} g Im g the damping, and Im r \operatorname{Im} r Im r the nonzero affine force. The real parts of f f f and g g g determine momentum and cross diffusion.
In the sector N = n N = n N = n , the closed master equation is
ρ ˙ = − i [ H e f f , ρ ] − ℓ Re f [ Q , [ Q , ρ ] ] − ℓ Re g [ Q , [ P , ρ ] ] − i ℓ Im g [ Q , { P , ρ } ] , (67) \dot{\rho} = -i[H_{\mathrm{eff}}, \rho] - \ell\operatorname{Re} f[Q, [Q, \rho]]
-\ell\operatorname{Re} g[Q, [P, \rho]] - i\ell\operatorname{Im} g[Q, \{P, \rho\}], \tag*{(67)} ρ ˙ = − i [ H eff , ρ ] − ℓ Re f [ Q , [ Q , ρ ]] − ℓ Re g [ Q , [ P , ρ ]] − i ℓ Im g [ Q , { P , ρ }] , (67)
H e f f = ν 2 ( Q 2 + P 2 ) + ℓ Im f Q 2 + ( η n + 2 ℓ n Im r ) Q + E n . (68) H_{\mathrm{eff}} = \frac{\nu}{2}(Q^2 + P^2) + \ell\operatorname{Im} f Q^2
+(\eta n + 2\ell n \operatorname{Im} r)Q + E_n. \tag*{(68)} H eff = 2 ν ( Q 2 + P 2 ) + ℓ Im f Q 2 + ( η n + 2 ℓ n Im r ) Q + E n . (68)
The reduced-state moments defined above obey
m ˙ = A m + b , V ˙ = A V + V A T + D , (69) \dot{m} = A m + b,\qquad\dot{V} = A V + V A^T + D, \tag*{(69)} m ˙ = A m + b , V ˙ = A V + V A T + D , (69)
where
A = ( 0 ν − ν − 2 ℓ Im f − 2 ℓ Im g ) , b = ( 0 − η n − 2 ℓ n Im r ) , D = ( 0 − ℓ Re g − ℓ Re g 2 ℓ Re f ) . (70) A = \begin{pmatrix}
0 & \nu\\
-\nu- 2\ell\operatorname{Im} f & -2\ell\operatorname{Im} g
\end{pmatrix},\qquad
b = \begin{pmatrix}
0 \\
-\eta n - 2\ell n \operatorname{Im} r
\end{pmatrix},\qquad
D = \begin{pmatrix}
0 & -\ell\operatorname{Re} g \\
-\ell\operatorname{Re} g & 2\ell\operatorname{Re} f
\end{pmatrix}. \tag*{(70)} A = ( 0 − ν − 2 ℓ Im f ν − 2 ℓ Im g ) , b = ( 0 − η n − 2 ℓ n Im r ) , D = ( 0 − ℓ Re g − ℓ Re g 2 ℓ Re f ) . (70)
Physicality concerns the propagated state; the instantaneous diffusion matrix need not be positive semidefinite in this non-Markovian equation. For a Gaussian preparation, m m m and V V V specify the reduced Gaussian state. The accompanying effective-coefficient and full-state comparisons are given in Appendix J.
Physical meaning and scope of the signatures
The two models answer the same question at different stages of the reconstruction. In the qubits, the deterministic lowering row passes directly through the noise average, and a reference coherence is linear in the system amplitude. A nonzero highest-order resolvent residue therefore gives an unconditional Jordan contribution. In the oscillator, the scalar noise history survives in the state equation and in the operator average. The width combines both coefficient roots before the displacement is formed, and now weighting adds the spread of centers to the conditional covariance.
Three selections remain distinct: the causal initial vector must reach a Jordan chain; dominance in the denominator projection determines whether that chain sets a coefficient’s algebraic tail; the chosen state or observable must retain its residue. The qubit eigenvalue relation μ = − λ − d \mu= -\lambda- d μ = − λ − d reverses coefficient dominance and amplitude-decay ordering, while the oscillator’s causal coefficient approaches one branch whose companion remains in the normalizable width and displacement response.
The observable determines what a measured time trace can establish. In the qubit example, a calibrated phase reference selects an unconditional coherence with a quadratic-exponential term. Both basis preparations, known model parameters, and an invertible amplitude map then allow the full coefficient row to be reconstructed. The oscillator’s double pole describes the conditional displacement response of the chosen NMQSD representation to a history variation at long preparation age and fixed lag. Applied-force susceptibilities and bath-averaged two-time correlation problems involve separate propagation problems [34 , 35 ] . The finite-time variance comparison establishes the effect of complete averaging for the displayed parameters; it does not isolate an EP-specific enhancement.
Finite correlation time specifies a reservoir property, whereas operational non-Markovianity is assessed from the reduced dynamical map through information-backflow or divisibility criteria [36 , 37 , 38 , 39 ] . Their relation to a coefficient EP is model dependent, as is the occurrence of revivals in reaction-coordinate dynamics [10 ] . Likewise, an instantaneous time-local NMME generator has its own spectral structure, which must be determined separately from that of the coefficient lift.
The exact results apply to the stated closures and preparations. Multiple excitations introduce additional operator sectors [23 ] , while finite temperature extends the stochastic trajectories and their averages [40 ] . Stochastic hierarchies accommodate more general reservoir correlations [41 ] ; their spectral and averaging analysis must be established for the enlarged hierarchy, with master-equation consistency and positivity checked for approximate closures [42 ] . In the oscillator calculation, conserved photon number and a pure Gaussian mechanical preparation specify the state family. Its long-age results are established on a regular normalizable branch in the chosen interval 0 < γ < ν 0 < \gamma< \nu 0 < γ < ν , and the explicit master equation uses a regular averaging chart. These conditions connect the coefficient EP to the particular state signature under study.
Conclusions
We have connected exceptional points of exact NMQSD coefficient equations to state dynamics by carrying the same operator closure through reconstruction and noise averaging. The coefficient lift identifies reached Jordan chains and their asymptotic contributions, while the state map and the readout determine which terms survive. This separation also distinguishes a spectral EP from a singularity of finite coefficient coordinates.
The two models show why each step matters. For common-bath qubits in the zero- and single-excitation sector, every defective decaying mode contributes its full pole order to a suitable unconditional excitation–reference coherence. The EP3 example displays a quadratic Jordan term in one quadrature and its cancellation in another. For the fixed-photon-number optomechanical sector, the noise-history term and the Gaussian width combine two coefficient roots into a double-pole conditional displacement response at long preparation age and fixed lag on the regular normalizable branch. The purely algebraic coefficient tail cancels from the width, and complete averaging can turn conditional momentum narrowing into reduced-state broadening.
A coefficient EP therefore has a physical interpretation through a specified causal preparation, state reconstruction, and average. The exact NMQSD-to-NMME route makes those connections explicit without assigning a universal EP order to all observables or to the instantaneous reduced-state generator.
Acknowledgments
This material is based upon work supported by the National Science Foundation under Grant Nos. 2328948 and 2519355.
Data availability
The underlying data and code can be obtained from the corresponding author upon reasonable request.
Direct NMQSD closure and the non-Markovian master equation
In the retained zero- and single-excitation sector, take H S = 0 ⊕ H 1 H_S = 0 \oplus H_1 H S = 0 ⊕ H 1 with time-independent Hermitian H 1 H_1 H 1 , the lowering operator in Eq. (14) , and the independent vacuum reservoir specified in Sec. III A. The operators in Eq. (15) act on the whole retained system space, including their zero action on ∣ G S ⟩ \lvert G_S\rangle ∣ G S ⟩ . Products of two lowering operators vanish in this sector, so the noise commutator in the consistency equation vanishes. The remaining commutators give Eq. (16) , with boundary condition f q ( t , t ) = κ f_q(t,t) = \kappa f q ( t , t ) = κ . Differentiating the integral in Eq. (15) , using α ( t , t ) = A 0 \alpha(t,t) = A_0 α ( t , t ) = A 0 and ∂ t α ( t , s ) = − ∂ α ( t , s ) \partial_t\alpha(t,s) = -\partial\alpha(t,s) ∂ t α ( t , s ) = − ∂ α ( t , s ) , then yields the Riccati equation (17) and its initial condition F q ( 0 ) = 0 F_q(0) = 0 F q ( 0 ) = 0 .
Because O ˉ t \bar{O}_t O ˉ t is noise independent, the exact projector average in Eq. (6) gives Eq. (25) . Direct multiplication of the ground-excitation blocks yields Eq. (26) . Integrating those equations with the amplitude propagator in Eq. (24) gives
ρ e e ( t ) = G ( t ) ρ e e ( 0 ) G ( t ) † , ρ e G ( t ) = G ( t ) ρ e G ( 0 ) , ρ G G ( t ) = ρ G G ( 0 ) + tr [ ( I − G † G ) ρ e e ( 0 ) ] . (71) \rho_{\mathrm{ee}}(t)=G(t)\rho_{\mathrm{ee}}(0)G(t)^\dagger,
\qquad
\rho_{\mathrm{eG}}(t)=G(t)\rho_{\mathrm{eG}}(0),
\qquad
\rho_{\mathrm{GG}}(t)=\rho_{\mathrm{GG}}(0)+\operatorname{tr}\left[(I-G^\dagger G)\rho_{\mathrm{ee}}(0)\right]. \tag*{(71)} ρ ee ( t ) = G ( t ) ρ ee ( 0 ) G ( t ) † , ρ eG ( t ) = G ( t ) ρ eG ( 0 ) , ρ GG ( t ) = ρ GG ( 0 ) + tr [ ( I − G † G ) ρ ee ( 0 ) ] . (71)
In particular, the preparation ( ∣ G S ⟩ + ∣ u 0 ⟩ ) / 2 (\lvert G_S\rangle+\lvert u_0\rangle)/\sqrt{2} (∣ G S ⟩ + ∣ u 0 ⟩) / 2 gives ρ e G ( t ) = G ( t ) u 0 / 2 \rho_{\mathrm{eG}}(t)=G(t)u_0/2 ρ eG ( t ) = G ( t ) u 0 /2 . This is an unconditional subsystem coherence obtained directly from the NMME.
The same result follows at the NMQSD-state level. The excitation amplitude is the deterministic vector G ( t ) u 0 G(t)u_0 G ( t ) u 0 , while noise enters the ground amplitude with zero reference mean. On a regular chart the functional derivative then gives f q ( t , s ) = κ G ( s ) G ( t ) − 1 f_q(t,s)=\kappa G(s)G(t)^{-1} f q ( t , s ) = κ G ( s ) G ( t ) − 1 . Thus the coefficient closure, state propagation, and NMME refer to the same preparation and reservoir, without an additional physical evolution model.
State-memory lift, global chart, and pole visibility
Define the row W = F q G \mathbf{W}=F_qG W = F q G on a regular chart. Using Eqs. (17) and (24) cancels both the i F q H 1 G iF_qH_1G i F q H 1 G terms and the quadratic terms, yielding the coupled equations in Eq. (29) . Since W ( 0 ) = 0 W(0)=0 W ( 0 ) = 0 , this auxiliary variable is exactly the original memory convolution,
W ( t ) = A 0 ∫ 0 t e − d ( t − s ) κ G ( s ) d s . (72) W(t)=A_0\int_0^t e^{-d(t-s)}\kappa G(s)\,\mathrm{d}s. \tag*{(72)} W ( t ) = A 0 ∫ 0 t e − d ( t − s ) κ G ( s ) d s . (72)
It introduces no additional system state or physical preparation. Eliminating it gives the causal amplitude equation
G ˙ ( t ) = − i H 1 G ( t ) − ∫ 0 t α ( t , s ) κ † κ G ( s ) d s . (73) \dot{G}(t)=-iH_1G(t)-\int_0^t\alpha(t,s)\kappa^\dagger\kappa G(s)\,\mathrm{d}s. \tag*{(73)} G ˙ ( t ) = − i H 1 G ( t ) − ∫ 0 t α ( t , s ) κ † κ G ( s ) d s . (73)
The constant matrix A q A_q A q and initial block column ( I N q , 0 ) T (I_{N_q},0)^T ( I N q , 0 ) T are given in Eq. (29) . Thus G G G and W W W are blocks of the globally defined matrix U q ( t ) = e A q t U_q(t)=e^{A_qt} U q ( t ) = e A q t .
Dual lifts and the finite coefficient chart
Direct block multiplication establishes Eq. (30) , where J − 1 = J J^{-1}=J J − 1 = J . Its spectral map is μ = − λ − d \mu=-\lambda-d μ = − λ − d , with μ \mu μ and λ \lambda λ eigenvalues of M q M_q M q and A q A_q A q , respectively. The transpose and invertible similarity preserve Jordan orders. The propagators obey e M q t = e − d t J U q ( t ) − T J e^{M_qt}=e^{-dt}JU_q(t)^{-T}J e M q t = e − d t J U q ( t ) − T J . Because det U q = e ( − i tr H 1 − d ) t \det U_q=e^{(-i\operatorname{tr}H_1-d)t} det U q = e ( − i tr H 1 − d ) t , the complementary minor formula gives
q ( t ) = e − d t ( U q − 1 ) a a = e i tr ( H 1 ) t det G ( t ) . (74) q(t)=e^{-dt}(U_q^{-1})_{aa}=e^{i\operatorname{tr}(H_1)t}\det G(t). \tag*{(74)} q ( t ) = e − d t ( U q − 1 ) aa = e i tr ( H 1 ) t det G ( t ) . (74)
This identity is global: U q U_q U q is invertible at every finite time, even when its system block is singular. On det G ≠ 0 \det G\ne0 det G = 0 , F q = W G − 1 F_q=WG^{-1} F q = W G − 1 . An isolated order-m m m zero of det G \det G det G gives c q = − q ˙ / q ∼ − m / ( t − t 0 ) c_q=-\dot q/q\sim-m/(t-t_0) c q = − q ˙ / q ∼ − m / ( t − t 0 ) . It is a boundary of the finite coefficient chart.
A positive metric and global reduced-state regularity
The Hermiticity of H 1 H_1 H 1 and A 0 , γ > 0 A_0,\gamma>0 A 0 , γ > 0 give the positive metric
P q = diag ( I N q , A 0 − 1 ) , P_q=\operatorname{diag}(I_{N_q},A_0^{-1}), P q = diag ( I N q , A 0 − 1 ) ,
A q † P q + P q A q = − 2 γ A 0 e a e a † . (75) A_q^\dagger P_q+P_qA_q=-\frac{2\gamma}{A_0}e_a e_a^\dagger. \tag*{(75)} A q † P q + P q A q = − A 0 2 γ e a e a † . (75)
Applying this identity to the block column in Eq. (29) and integrating yields
G † G + W † W A 0 + 2 γ A 0 ∫ 0 t W ( s ) † W ( s ) d s = I N q . (76) G^\dagger G+\frac{W^\dagger W}{A_0}+\frac{2\gamma}{A_0}\int_0^t W(s)^\dagger W(s)\,\mathrm{d}s=I_{N_q}. \tag*{(76)} G † G + A 0 W † W + A 0 2 γ ∫ 0 t W ( s ) † W ( s ) d s = I N q . (76)
Hence G † G ⪯ I G^\dagger G\preceq I G † G ⪯ I . The reduced map (Eq. (71) ) stays finite and completely positive across a coefficient-chart zero. For example, a Kraus representation consists of K 0 = ∣ G S ⟩ ⟨ G S ∣ + ∑ i j G i j ∣ i ⟩ ⟨ j ∣ K_0=|G_S\rangle\langle G_S|+\sum_{ij}G_{ij}|i\rangle\langle j| K 0 = ∣ G S ⟩ ⟨ G S ∣ + ∑ ij G ij ∣ i ⟩ ⟨ j ∣ and ground-transfer rows obtained from a factorization of I − G † G I-G^\dagger G I − G † G . They satisfy ∑ α K α † K α = I \sum_\alpha K_\alpha^\dagger K_\alpha=I ∑ α K α † K α = I . The time-local NMME uses finite coefficients only on its regular chart; Eq. (71) supplies the continuous global map.
For an eigenvector v = ( u , w ) T v=(u,w)^T v = ( u , w ) T of A q A_q A q , Eq. (75) gives
Re λ = − γ ∣ w ∣ 2 / A 0 ∥ u ∥ 2 + ∣ w ∣ 2 / A 0 ≤ 0. (77) \operatorname{Re}\lambda=-\gamma\frac{|w|^2/A_0}{\lVert u\rVert^2+|w|^2/A_0}\le0. \tag*{(77)} Re λ = − γ ∥ u ∥ 2 + ∣ w ∣ 2 / A 0 ∣ w ∣ 2 / A 0 ≤ 0. (77)
If Re λ = 0 \operatorname{Re}\lambda=0 Re λ = 0 , then w = 0 w=0 w = 0 , κ u = 0 \kappa u=0 κ u = 0 , and − i H 1 u = λ u -iH_1u=\lambda u − i H 1 u = λ u . Moreover v † P q v^\dagger P_q v † P q is a left eigenvector at λ \lambda λ . A generalized vector satisfying ( A q − λ I ) v 1 = v (A_q-\lambda I)v_1=v ( A q − λ I ) v 1 = v would imply 0 = v † P q v > 0 0=v^\dagger P_qv>0 0 = v † P q v > 0 , a contradiction. These boundary modes are therefore dark and semisimple. Every defective eigenvalue has Re λ < 0 \operatorname{Re}\lambda<0 Re λ < 0 .
Causal reachability and a surviving system pole
For Re λ < 0 \operatorname{Re}\lambda<0 Re λ < 0 , the matrix λ I + i H 1 \lambda I+iH_1 λ I + i H 1 is invertible. The eigenvector equations fix u = − ( λ I + i H 1 ) − 1 κ † w u=-(\lambda I+iH_1)^{-1}\kappa^\dagger w u = − ( λ I + i H 1 ) − 1 κ † w uniquely, with w ≠ 0 w\ne0 w = 0 . Each such eigenspace is one dimensional; an algebraic multiplicity r r r therefore corresponds to one size-r r r block. The Laplace transform of the prescribed coefficient denominator is given by Eq. (31) . At μ = − λ − d \mu=-\lambda-d μ = − λ − d its numerator equals ( − 1 ) N q det ( λ I + i H 1 ) ≠ 0 (-1)^{N_q}\det(\lambda I+iH_1)\ne0 ( − 1 ) N q det ( λ I + i H 1 ) = 0 . Thus the zero-coefficient initial condition and denominator readout retain the full order of every defective pole. The coefficient tail follows from Eq. (10) only when that reached pole is strictly dominant. Writing Γ λ = − Re λ \Gamma_\lambda=-\operatorname{Re}\lambda Γ λ = − Re λ gives Re μ = Γ λ − γ \operatorname{Re}\mu=\Gamma_\lambda-\gamma Re μ = Γ λ − γ : coefficient dominance selects faster amplitude decay.
For D q ( z ) D_q(z) D q ( z ) and h q ( z ) h_q(z) h q ( z ) defined in Eq. (35) , the Schur complement of z I − A q zI-A_q z I − A q gives the system resolvent block in Eq. (36) . At a defective eigenvalue λ \lambda λ of order r r r , det D q ( λ ) ≠ 0 \det D_q(\lambda)\ne0 det D q ( λ ) = 0 and h q h_q h q has an order-r r r zero. Expanding h q − 1 h_q^{-1} h q − 1 about that zero gives the highest system Laurent coefficient in Eq. (37) . Both outer-product factors are nonzero: otherwise invertibility would imply κ = 0 \kappa=0 κ = 0 , which cannot produce a defective eigenvalue in this class. Consequently,
G ( t ) ⊃ t r − 1 e λ t ( r − 1 ) ! ( R r ) e e . (78) G(t)\supset\frac{t^{r-1}e^{\lambda t}}{(r-1)!}(R_r)_{ee}. \tag*{(78)} G ( t ) ⊃ ( r − 1 )! t r − 1 e λ t ( R r ) ee . (78)
Some normalized system input u 0 u_0 u 0 and system row v † v^\dagger v † have v † ( R r ) e e u 0 ≠ 0 v^\dagger(R_r)_{ee}u_0\ne0 v † ( R r ) ee u 0 = 0 . The NMME coherence G ( t ) u 0 / 2 G(t)u_0/2 G ( t ) u 0 /2 then exposes the term in a suitable Hermitian quadrature. This argument also covers an EP embedded in a larger spectrum. It does not assign the same order to every quadrature or to the complete density-operator generator.
Exact EP3 solution and coefficient spectral splitting
Use the two-qubit parameters of Eq. (20) and τ = γ t \tau=\gamma t τ = γ t . Their relation to the configuration of Ref. [9 ] is explicit under the real orthogonal basis rotation
U 30 = 1 2 ( 3 1 − 1 3 ) , κ U 30 = ( 0 , 8 / 27 ) , U_{30}=\frac{1}{2}\begin{pmatrix}\sqrt{3} & 1 \\ -1 & \sqrt{3}\end{pmatrix}, \qquad\kappa U_{30}=(0,\sqrt{8/27}), U 30 = 2 1 ( 3 − 1 1 3 ) , κ U 30 = ( 0 , 8/27 ) ,
U 30 T H 1 U 30 = γ 3 3 ( 0 1 1 0 ) . U_{30}^{T}H_{1}U_{30}=\frac{\gamma}{3\sqrt{3}}\begin{pmatrix}0 & 1 \\ 1 & 0\end{pmatrix}. U 30 T H 1 U 30 = 3 3 γ ( 0 1 1 0 ) .
Thus the exchange magnitude is γ / ( 3 3 ) \gamma/(3\sqrt{3}) γ / ( 3 3 ) and the total bath coupling weight is A 0 ∥ κ ∥ 2 = 8 γ 2 / 27 A_{0}\lVert\kappa\rVert^{2}=8\gamma^{2}/27 A 0 ∥ κ ∥ 2 = 8 γ 2 /27 , corresponding to Λ = γ \Lambda=\gamma Λ = γ and Γ = 16 γ / 27 \Gamma=16\gamma/27 Γ = 16 γ /27 in that reference. The following calculation uses the original NMQSD coefficient lift. Direct evaluation gives
det ( μ I − M q ) = ( μ + 2 γ / 3 ) 3 . (79) \det(\mu I-M_{q})=(\mu+2\gamma/3)^{3}. \tag*{(79)} det ( μ I − M q ) = ( μ + 2 γ /3 ) 3 . (79)
For N M = M q + 2 γ I 3 / 3 N_{M}=M_{q}+2\gamma I_{3}/3 N M = M q + 2 γ I 3 /3 , N M 3 = 0 N_{M}^{3}=0 N M 3 = 0 and ( N M 2 ) a a = 4 γ 2 / 27 ≠ 0 (N_{M}^{2})_{aa}=4\gamma^{2}/27\ne0 ( N M 2 ) aa = 4 γ 2 /27 = 0 . Hence its nilpotency index is three, establishing the EP3 directly in the coefficient equations. Exponentiating this block gives Eqs. (21) and (22) . The positive coefficients of the polynomial in Eq. (21) ensure a regular chart for t ≥ 0 t\ge0 t ≥ 0 , and its quadratic degree yields the stated − 2 / t -2/t − 2/ t correction to c q c_{q} c q .
The derived state-memory matrix has N A = A q + γ I 3 / 3 N_{A}=A_{q}+\gamma I_{3}/3 N A = A q + γ I 3 /3 with N A 3 = 0 N_{A}^{3}=0 N A 3 = 0 , so its system block is
G ( t ) = e − τ / 3 [ I 3 + τ γ N A + τ 2 2 γ 2 N A 2 ] e e . (80) G(t)=e^{-\tau/3}\left[I_{3}+\frac{\tau}{\gamma}N_{A}+\frac{\tau^{2}}{2\gamma^{2}}N_{A}^{2}\right]_{\mathrm{ee}}. \tag*{(80)} G ( t ) = e − τ /3 [ I 3 + γ τ N A + 2 γ 2 τ 2 N A 2 ] ee . (80)
The diagonal similarity R = diag ( 1 , 3 , 6 ) R=\operatorname{diag}(1,\sqrt{3},\sqrt{6}) R = diag ( 1 , 3 , 6 ) makes R − 1 A q R R^{-1}A_{q}R R − 1 A q R rational-complex at γ = 1 \gamma=1 γ = 1 . Exact arithmetic gives ranks 2, 1, 0 for the first three powers of its shifted nilpotent matrix; the affine-transpose identity gives the coefficient result.
For u 0 = ( 1 , 0 ) T u_{0}=(1,0)^{T} u 0 = ( 1 , 0 ) T , its two components are
a 1 ( t ) = e − τ / 3 [ 1 + ( 2 − i ) τ 6 − i τ 2 18 ] , a 2 ( t ) = − 3 54 e − τ / 3 τ [ ( 2 + i ) τ + 3 i ] . (81) \begin{aligned}
a_{1}(t)&=e^{-\tau/3}\left[1+\frac{(2-i)\tau}{6}-\frac{i\tau^{2}}{18}\right],\\
a_{2}(t)&=-\frac{\sqrt{3}}{54}e^{-\tau/3}\tau\left[(2+i)\tau+3i\right]. \tag*{(81)}
\end{aligned} a 1 ( t ) a 2 ( t ) = e − τ /3 [ 1 + 6 ( 2 − i ) τ − 18 i τ 2 ] , = − 54 3 e − τ /3 τ [ ( 2 + i ) τ + 3 i ] . (81)
The coherent preparation gives ρ j G = a j / 2 \rho_{jG}=a_{j}/2 ρ j G = a j /2 . Thus X 1 X_{1} X 1 has a linear prefactor and Y 1 Y_{1} Y 1 retains the quadratic term, as in Eq. (38) . The selected signal in Eq. (40) has Laplace transform
X ~ 2 ( s ) = − 2 3 γ 2 27 ( s + γ / 3 ) 3 . (82) \widetilde{X}_{2}(s)=-\frac{2\sqrt{3}\gamma^{2}}{27(s+\gamma/3)^{3}}. \tag*{(82)} X 2 ( s ) = − 27 ( s + γ /3 ) 3 2 3 γ 2 . (82)
The nonzero cubic pole is therefore a consequence of the direct NMQSD-to-NMME map. For the pure initial excitation ∣ 1 ⟩ \lvert1\rangle ∣ 1 ⟩ , the total excitation population P e x c = tr ρ e e P_{\mathrm{exc}}=\operatorname{tr}\rho_{\mathrm{ee}} P exc = tr ρ ee is
P e x c ( t ) = e − 2 τ / 3 2 τ 4 + 6 τ 3 + 36 τ 2 + 162 τ + 243 243 . (83) P_{\mathrm{exc}}(t)=e^{-2\tau/3}\frac{2\tau^{4}+6\tau^{3}+36\tau^{2}+162\tau+243}{243}. \tag*{(83)} P exc ( t ) = e − 2 τ /3 243 2 τ 4 + 6 τ 3 + 36 τ 2 + 162 τ + 243 . (83)
Its fourth-degree prefactor comes from products of amplitudes and does not by itself identify a fifth-order EP.
Under the frequency shift defined above Eq. (23) , the coefficient characteristic polynomial in that equation gives ζ 3 ∼ ε / 9 \zeta^{3}\sim\varepsilon/9 ζ 3 ∼ ε /9 . The state-memory eigenvalues obey ( λ + γ / 3 ) / γ = − ζ (\lambda+\gamma/3)/\gamma=-\zeta ( λ + γ /3 ) / γ = − ζ and have the same pairwise spectral gaps. This splitting is derived from M q M_{q} M q ; the state map fixes its readout consequences. Numerical route comparisons and their finite-grid accuracy are reported separately in Appendix § 1 .
Three coefficient and state boundary cases
All three controls use A 0 = γ 2 A_{0}=\gamma^{2} A 0 = γ 2 , Ω = 0 \Omega=0 Ω = 0 , and τ = γ t \tau=\gamma t τ = γ t . They follow directly from the same coefficient and NMME equations.
Dark degeneracy. Take N q = 3 N_{q}=3 N q = 3 , H 1 ∼ diag ( 0 , 0 , 1 ) H_{1}\sim\operatorname{diag}(0,0,1) H 1 ∼ diag ( 0 , 0 , 1 ) , and κ = ( 0 , 0 , 1 / 2 ) \kappa=(0,0,1/2) κ = ( 0 , 0 , 1/2 ) . The first two excitation states have constant amplitudes. Their repeated eigenvalue λ = 0 \lambda=0 λ = 0 of A q A_{q} A q is semisimple, as is the corresponding μ = − γ \mu=-\gamma μ = − γ of M q M_{q} M q . The coefficient eigenvectors have last coordinate zero and therefore lie outside the finite stationary-row chart. This does not mean that the prescribed trajectory q ( t ) q(t) q ( t ) vanishes identically.
Off-EP chart pole. Take N q = 1 N_{q}=1 N q = 1 , H 1 = 0 H_{1}=0 H 1 = 0 , and κ = 1 \kappa=1 κ = 1 . The two eigenvalues of M q M_{q} M q are the distinct values μ ± = − γ / 2 ± i 3 γ / 2 \mu_{\pm}=-\gamma/2\pm i\sqrt{3}\gamma/2 μ ± = − γ /2 ± i 3 γ /2 . With β 0 = 3 / 2 \beta_{0}=\sqrt{3}/2 β 0 = 3 /2 ,
q ( t ) = G ( t ) = e − τ / 2 ( cos β 0 τ + sin β 0 τ 3 ) , q(t)=G(t)=e^{-\tau/2}\left(\cos\beta_{0}\tau+\frac{\sin\beta_{0}\tau}{\sqrt{3}}\right), q ( t ) = G ( t ) = e − τ /2 ( cos β 0 τ + 3 sin β 0 τ ) ,
q ˙ ( t ) = − γ e − τ / 2 sin β 0 τ β 0 . (84) \dot{q}(t)=-\gamma e^{-\tau/2}\frac{\sin\beta_{0}\tau}{\beta_{0}}. \tag*{(84)} q ˙ ( t ) = − γ e − τ /2 β 0 sin β 0 τ . (84)
The first zero is t 0 = 4 π / ( 3 3 γ ) t_{0}=4\pi/(3\sqrt{3}\gamma) t 0 = 4 π / ( 3 3 γ ) , with q ˙ ( t 0 ) = − γ e − γ t 0 / 2 ≠ 0 \dot{q}(t_{0})=-\gamma e^{-\gamma t_{0}/2}\ne0 q ˙ ( t 0 ) = − γ e − γ t 0 /2 = 0 . Thus c q ∼ − 1 / ( t − t 0 ) c_{q}\sim-1/(t-t_{0}) c q ∼ − 1/ ( t − t 0 ) without a coefficient EP. Equation (71) gives the regular ground-state density at t 0 t_{0} t 0 for both an initial excitation and its equal superposition with ∣ G S ⟩ \lvert G_{S}\rangle ∣ G S ⟩ . The global map continues through this point; no finite coefficient value is assigned there.
Reached but subleading EP. Take H 1 = H_{1}= H 1 = γ diag ( 1 / ( 4 2 ) , − 1 / ( 4 2 ) ) \gamma\operatorname{diag}(1/(4\sqrt{2}),-1/(4\sqrt{2})) γ diag ( 1/ ( 4 2 ) , − 1/ ( 4 2 )) and κ = ( 3 / 8 , 3 / 8 ) \kappa=(3/8,3/8) κ = ( 3/8 , 3/8 ) .
q ( t ) = 9 2 e − τ / 2 − ( 7 2 + 3 τ 8 ) e − 3 τ / 4 = e − τ / 2 Q ( τ ) , Q ( τ ) = 9 2 − ( 7 2 + 3 τ 8 ) e − τ / 4 , Q ′ ( τ ) = ( 1 2 + 3 τ 32 ) e − τ / 4 > 0. (85) \begin{aligned}
q(t)&=\frac{9}{2}e^{-\tau/2}-\left(\frac{7}{2}+\frac{3\tau}{8}\right)e^{-3\tau/4}\\
&=e^{-\tau/2}Q(\tau),\\
Q(\tau)&=\frac{9}{2}-\left(\frac{7}{2}+\frac{3\tau}{8}\right)e^{-\tau/4},\\
Q'(\tau)&=\left(\frac{1}{2}+\frac{3\tau}{32}\right)e^{-\tau/4}>0. \tag*{(85)}
\end{aligned} q ( t ) Q ( τ ) Q ′ ( τ ) = 2 9 e − τ /2 − ( 2 7 + 8 3 τ ) e − 3 τ /4 = e − τ /2 Q ( τ ) , = 2 9 − ( 2 7 + 8 3 τ ) e − τ /4 , = ( 2 1 + 32 3 τ ) e − τ /4 > 0. (85)
Here q ( t ) q(t) q ( t ) is the prescribed denominator. The coefficient spectrum has a size-two block at μ = − 3 γ / 4 \mu=-3\gamma/4 μ = − 3 γ /4 and a simple pole at μ = − γ / 2 \mu=-\gamma/2 μ = − γ /2 . Since Q ( 0 ) = 1 Q(0)=1 Q ( 0 ) = 1 , the chart is finite for all t ≥ 0 t\geq0 t ≥ 0 . The EP contribution is reached, but c q = γ [ 1 / 2 − Q ′ ( τ ) / Q ( τ ) ] → γ / 2 c_q=\gamma[1/2-Q'(\tau)/Q(\tau)]\to\gamma/2 c q = γ [ 1/2 − Q ′ ( τ ) / Q ( τ )] → γ /2 has no leading 1 / t 1/t 1/ t term. The state-memory eigenvalues are the defective λ = − γ / 4 \lambda=-\gamma/4 λ = − γ /4 and simple λ = − γ / 2 \lambda=-\gamma/2 λ = − γ /2 . Thus the faster-decaying amplitude mode selects the coefficient tail, while the EP polynomial remains explicitly present in the global state propagation.
Noise-complete two-time coefficients
The oscillator terms below follow the noise-complete O O O -operator equations of Ref. [18 ] , with the conserved affine term included. The integrated coefficients F , G , R F,G,R F , G , R are defined in Sec. IV A; the scalar-history definitions are
J ( t , u ) = ∫ 0 t α ( t , s ) j ( t , s , u ) d s , Ξ t = ∫ 0 t J ( t , u ) z u ∗ d u , J ( t ) = ∫ 0 t α ( t , u ) J ( t , u ) d u . (86) \begin{aligned}
\mathcal{J}(t,u)&=\int_0^t\alpha(t,s)j(t,s,u)\,ds,\\
\Xi_t&=\int_0^t\mathcal{J}(t,u)z_u^*\,du,\\
J(t)&=\int_0^t\alpha(t,u)\mathcal{J}(t,u)\,du. \tag*{(86)}
\end{aligned} J ( t , u ) Ξ t J ( t ) = ∫ 0 t α ( t , s ) j ( t , s , u ) d s , = ∫ 0 t J ( t , u ) z u ∗ d u , = ∫ 0 t α ( t , u ) J ( t , u ) d u . (86)
Matching coefficients in Eq. (4) gives
∂ t f = ν g O − 2 i ℓ F g O + i ℓ G f − i ℓ J ( t , s ) , ∂ t g O = − ν f − i ℓ G g O , ∂ t r O = η g O − i ℓ R g O , ∂ t j ( t , s , v ) = − i ℓ g O ( t , s ) J ( t , v ) . (87) \begin{aligned}
\partial_t f&=\nu g_O-2i\ell Fg_O+i\ell Gf-i\ell\mathcal{J}(t,s),\\
\partial_tg_O&=-\nu f-i\ell Gg_O,\\
\partial_tr_O&=\eta g_O-i\ell Rg_O,\\
\partial_tj(t,s,v)&=-i\ell g_O(t,s)\mathcal{J}(t,v). \tag*{(87)}
\end{aligned} ∂ t f ∂ t g O ∂ t r O ∂ t j ( t , s , v ) = ν g O − 2 i ℓ F g O + i ℓ G f − i ℓ J ( t , s ) , = − ν f − i ℓ G g O , = η g O − i ℓ R g O , = − i ℓ g O ( t , s ) J ( t , v ) . (87)
The endpoint conditions are
f ( t , t ) = ℓ , g O ( t , t ) = r O ( t , t ) = 0 , f(t,t)=\ell,\qquad g_O(t,t)=r_O(t,t)=0, f ( t , t ) = ℓ , g O ( t , t ) = r O ( t , t ) = 0 ,
j ( t , t , v ) = 0 , j ( t , s , t ) = ℓ g O ( t , s ) . (88) j(t,t,v)=0,\qquad j(t,s,t)=\ell g_O(t,s). \tag*{(88)} j ( t , t , v ) = 0 , j ( t , s , t ) = ℓ g O ( t , s ) . (88)
The last condition comes from the upper limit of the noise integral and [ L z s ∗ , g O P ] [Lz_s^*,g_OP] [ L z s ∗ , g O P ] . All integrated coefficients initially vanish. Since ∂ t α ( t , s ) = − d α ( t , s ) \partial_t\alpha(t,s)=-d\alpha(t,s) ∂ t α ( t , s ) = − d α ( t , s ) for t > s t>s t > s ,
∂ t J ( t , v ) = − ( d + i ℓ G ) J ( t , v ) , J ( t , t ) = ℓ G ( t ) . (89) \partial_t\mathcal{J}(t,v)=-(d+i\ell G)\mathcal{J}(t,v),\qquad\mathcal{J}(t,t)=\ell G(t). \tag*{(89)} ∂ t J ( t , v ) = − ( d + i ℓ G ) J ( t , v ) , J ( t , t ) = ℓ G ( t ) . (89)
Differentiating the integrals in Eq. (86) gives J ˙ = A 0 ℓ G − ( 2 d + i ℓ G ) J \dot J=A_0\ell G-(2d+i\ell G)J J ˙ = A 0 ℓ G − ( 2 d + i ℓ G ) J and the Ξ \Xi Ξ equation in Eq. (47) . Integration of the first line of Eq. (87) combines − 2 i ℓ F G + i ℓ G F -2i\ell FG+i\ell GF − 2 i ℓ F G + i ℓ GF into − i ℓ F G -i\ell FG − i ℓ F G , yielding Eq. (46) . These identities hold on regular coefficient charts; a ratio-coordinate singularity is continued through the homogeneous variables.
Coefficient exceptional point and Gaussian response
Shifting the spectral variable in Eq. (51) gives
P a ( w − d ) = ( w 2 − d 2 ) ( w 2 + ν 2 ) + 2 i B ν d . (90) P_a(w-d)=(w^2-d^2)(w^2+\nu^2)+2iB\nu d. \tag*{(90)} P a ( w − d ) = ( w 2 − d 2 ) ( w 2 + ν 2 ) + 2 i B ν d . (90)
A multiple root requires Δ a = 0 \Delta_a=0 Δ a = 0 or w = 0 w=0 w = 0 . The latter would require d ν = 2 i B d\nu=2iB d ν = 2 i B , impossible for γ > 0 \gamma>0 γ > 0 . Write y = Ω 2 y=\Omega^2 y = Ω 2 and s a = ν 2 + γ 2 − y s_a=\nu^2+\gamma^2-y s a = ν 2 + γ 2 − y . The imaginary and real parts of Δ a = 0 \Delta_a=0 Δ a = 0 give B = Ω s a / ( 2 ν ) B=\Omega s_a/(2\nu) B = Ω s a / ( 2 ν ) and s a 2 + 4 y ( ν 2 − y ) = 0 s_a^2+4y(\nu^2-y)=0 s a 2 + 4 y ( ν 2 − y ) = 0 . Their only positive solution is Eq. (53) , for which ν 2 < y < ν 2 + γ 2 \nu^2<y<\nu^2+\gamma^2 ν 2 < y < ν 2 + γ 2 .
The geometric multiplicities can be read from a finite-matrix similarity. Set s 0 = 2 A 0 s_0=\sqrt{2A_0} s 0 = 2 A 0 , g a = 2 B g_a=\sqrt{2B} g a = 2 B , and
S = ( 0 1 0 0 − 1 0 0 0 0 0 − i / s 0 A 0 / s 0 0 0 − 1 / s 0 i A 0 / s 0 ) . (91) S=\begin{pmatrix}
0&1&0&0\\
-1&0&0&0\\
0&0&-i/s_0&A_0/s_0\\
0&0&-1/s_0&iA_0/s_0
\end{pmatrix}. \tag*{(91)} S = 0 − 1 0 0 1 0 0 0 0 0 − i / s 0 − 1/ s 0 0 0 A 0 / s 0 i A 0 / s 0 . (91)
Direct multiplication gives det S = 1 \det S=1 det S = 1 and
S ( M o + d I ) S − 1 = ( 0 ν 0 0 − ν 0 − g a 0 0 0 0 − i d − g a 0 i d 0 ) ≡ K a l g . (92) S(M_o+dI)S^{-1}=
\begin{pmatrix}
0&\nu&0&0\\
-\nu&0&-g_a&0\\
0&0&0&-id\\
-g_a&0&id&0
\end{pmatrix}
\equiv K_{\mathrm{alg}}. \tag*{(92)} S ( M o + d I ) S − 1 = 0 − ν 0 − g a ν 0 0 0 0 − g a 0 i d 0 0 − i d 0 ≡ K alg . (92)
Each eigenspace of K a l g K_{\mathrm{alg}} K alg is one-dimensional: its first coordinate fixes the others when ν , g a , d ≠ 0 \nu,g_a,d\ne0 ν , g a , d = 0 . Hence the two repeated roots μ ± = − d ± w \mu_\pm=-d\pm w μ ± = − d ± w are defective. This similarity is used only to establish the Jordan structure of the coefficient matrix.
Causal reachability follows from the denominator generated by v ( 0 ) = e 4 v(0)=e_4 v ( 0 ) = e 4 . Its Laplace transform is
q ~ ( s ) = N o ( s ) P a ( s ) , N o ( s ) = ( s + 2 d ) [ ( s + d ) 2 + ν 2 ] − i B ν . (93) \widetilde q(s)=\frac{\mathcal{N}_o(s)}{P_a(s)},\qquad\mathcal{N}_o(s)=(s+2d)\left[(s+d)^2+\nu^2\right]-iB\nu. \tag*{(93)} q ( s ) = P a ( s ) N o ( s ) , N o ( s ) = ( s + 2 d ) [ ( s + d ) 2 + ν 2 ] − i B ν . (93)
At any root μ \mu μ of P a P_a P a , N o ( μ ) = − i B ν ( μ + 2 d ) / μ ≠ 0 \mathcal{N}_o(\mu)=-iB\nu(\mu+2d)/\mu\ne0 N o ( μ ) = − i B ν ( μ + 2 d ) / μ = 0 . Thus
q ( t ) = ( C + t + C 0 ) e μ + t + ( C − t + C 1 ) e μ − t , C + ≠ 0. (94) q(t)=(C_+t+C_0)e^{\mu_+t}+(C_-t+C_1)e^{\mu_-t},\qquad C_+\ne0. \tag*{(94)} q ( t ) = ( C + t + C 0 ) e μ + t + ( C − t + C 1 ) e μ − t , C + = 0. (94)
Using G = q ˙ / ( i ℓ q ) G=\dot q/(i\ell q) G = q ˙ / ( i ℓ q ) gives Eq. (55) . Equivalently, the initial vector is cyclic because det [ e 4 , M o e 4 , M o 2 e 4 , M o 3 e 4 ] = − ( A 0 ℓ ) 4 ν 2 ≠ 0 \det[e_4,M_oe_4,M_o^2e_4,M_o^3e_4]=-(A_0\ell)^4\nu^2\ne0 det [ e 4 , M o e 4 , M o 2 e 4 , M o 3 e 4 ] = − ( A 0 ℓ ) 4 ν 2 = 0 . The constants in Eq. (94) are coefficients of the homogeneous solution and are unrelated to the wavefunction scalar c ψ , t c_{\psi,t} c ψ , t .
Width cancellation and the response kernel. For the root w = x w + i y w w=x_w+iy_w w = x w + i y w with x w > 0 x_w>0 x w > 0 , the identities 2 x w y w = γ Ω 2x_wy_w=\gamma\Omega 2 x w y w = γ Ω and x w 2 − y w 2 = ( γ 2 − Ω 2 − ν 2 ) / 2 x_w^2-y_w^2=(\gamma^2-\Omega^2-\nu^2)/2 x w 2 − y w 2 = ( γ 2 − Ω 2 − ν 2 ) /2 give y w > 0 y_w>0 y w > 0 and
( y w 2 − Ω 2 2 ) ( 1 + γ 2 2 y w 2 ) = ν 2 2 > 0. \left(y_w^2-\frac{\Omega^2}{2}\right)\left(1+\frac{\gamma^2}{2y_w^2}\right)=\frac{\nu^2}{2}>0. ( y w 2 − 2 Ω 2 ) ( 1 + 2 y w 2 γ 2 ) = 2 ν 2 > 0.
Consequently, 2 γ w > Ω 2\gamma_w > \Omega 2 γ w > Ω and Re a ∗ = ( 2 γ w − Ω ) / ν > 0 \operatorname{Re} a_* = (2\gamma_w-\Omega)/\nu> 0 Re a ∗ = ( 2 γ w − Ω ) / ν > 0 for all positive γ , Ω , ν \gamma,\Omega,\nu γ , Ω , ν . The working interval 0 < γ < ν 0 < \gamma< \nu 0 < γ < ν in Eq. (63) is therefore separate from stationary-width positivity and is used together with the regular normalizable branch hypothesis.
The coefficient equations give F = i ( q ¨ + d q ˙ ) / ( ν ℓ q ) F = i(\ddot{q}+d\dot{q})/(\nu\ell q) F = i ( q ¨ + d q ˙ ) / ( ν ℓ q ) . At the leading EP root μ = w − d \mu= w-d μ = w − d , substitution of a ∗ = i ( d − 2 w ) / ν a_* = i(d-2w)/\nu a ∗ = i ( d − 2 w ) / ν into the width equation yields
a ˙ ∣ a = a ∗ = 2 i ( ∂ t − μ ) 2 q ν q . (95) \dot{a}\big|_{a=a_*} = \frac{2i(\partial_t-\mu)^2q}{\nu q}. \tag*{(95)} a ˙ a = a ∗ = ν q 2 i ( ∂ t − μ ) 2 q . (95)
The differential operator annihilates the entire leading block ( C + t + C 0 ) e μ t (C_+t+C_0)e^{\mu t} ( C + t + C 0 ) e μ t . Its forcing comes only from the exponentially subleading block. Of the two stationary widths, a ∗ a_* a ∗ and a u = i d / ν a_u = id/\nu a u = i d / ν , the latter has Re a u = − Ω / ν < 0 \operatorname{Re}a_u=-\Omega/\nu<0 Re a u = − Ω/ ν < 0 .
For the relaxation estimate, put a = u ˙ / ( i ν u ) a=\dot{u}/(i\nu u) a = u ˙ / ( i ν u ) and u = q v u=qv u = q v . With ϑ = t + C 0 / C + \vartheta=t+C_0/C_+ ϑ = t + C 0 / C + ,
v ¨ + [ − w 2 + 2 w ϑ − 2 ϑ 2 + O ( e − 2 Re w t ) ] v = 0. (96) \ddot{v}+\left[-w^2+\frac{2w}{\vartheta}-\frac{2}{\vartheta^2}+O(e^{-2\operatorname{Re}wt})\right]v=0. \tag*{(96)} v ¨ + [ − w 2 + ϑ 2 w − ϑ 2 2 + O ( e − 2 Re w t ) ] v = 0. (96)
The leading solutions are
v 1 = e w t ϑ , v_1=\frac{e^{wt}}{\vartheta}, v 1 = ϑ e w t ,
v 2 = e − w t ( ϑ 2 w + 1 2 w 2 + 1 4 w 3 ϑ ) . (97) v_2=e^{-wt}\left(\frac{\vartheta}{2w}+\frac{1}{2w^2}+\frac{1}{4w^3\vartheta}\right). \tag*{(97)} v 2 = e − w t ( 2 w ϑ + 2 w 2 1 + 4 w 3 ϑ 1 ) . (97)
The two leading solutions have a nonzero constant Wronskian. Variation of constants multiplies the remainder in Eq. (96) by v 1 2 v_1^2 v 1 2 , v 1 v 2 v_1v_2 v 1 v 2 , or v 2 2 v_2^2 v 2 2 , with respective integrable late-time bounds O ( t − 2 ) O(t^{-2}) O ( t − 2 ) , O ( e − 2 Re w t ) O(e^{-2\operatorname{Re}wt}) O ( e − 2 Re w t ) , and O ( t 2 e − 4 Re w t ) O(t^2e^{-4\operatorname{Re}wt}) O ( t 2 e − 4 Re w t ) . A vanishing v 1 v_1 v 1 amplitude would select a u a_u a u . For the normalized initial state, however, the exact NMQSD identity E 0 ∥ ψ t ∥ 2 = 1 \mathbb{E}_0\lVert\psi_t\rVert^2=1 E 0 ∥ ψ t ∥ 2 = 1 requires the nonzero Gaussian kets to remain normalizable on the regular branch. Since a a a is independent of the noise labels, this enforces Re a > 0 \operatorname{Re}a>0 Re a > 0 and excludes convergence to Re a u < 0 \operatorname{Re}a_u<0 Re a u < 0 . The homogeneous coefficient variables continue across any earlier zero of q q q ; Gaussian normalizability remains a separate condition. The surviving v 1 v_1 v 1 contribution gives Eq. (63) , including the polynomial–exponential correction.
At finite preparation age, the response of ( b , Ξ ) T (b,\Xi)^T ( b , Ξ ) T is governed by
A b ( t ) = ( i ℓ G − i ν a − i ℓ 0 − d − i ℓ G ) , B b ( t ) = ( ℓ ℓ G ( t ) ) . (98) A_b(t)=\begin{pmatrix}i\ell G-i\nu a & -i\ell\\ 0 & -d-i\ell G\end{pmatrix},
\qquad
B_b(t)=\begin{pmatrix}\ell\\ \ell G(t)\end{pmatrix}. \tag*{(98)} A b ( t ) = ( i ℓ G − i ν a 0 − i ℓ − d − i ℓ G ) , B b ( t ) = ( ℓ ℓ G ( t ) ) . (98)
If Φ ( t , s ) \Phi(t,s) Φ ( t , s ) is its fundamental propagator, the finite-age kernel is h T ( s ) = ( 1 , 0 ) Φ ( T , s ) B b ( s ) h_T(s)=(1,0)\Phi(T,s)B_b(s) h T ( s ) = ( 1 , 0 ) Φ ( T , s ) B b ( s ) . Equivalently, the row v T ( s ) v^T(s) v T ( s ) satisfies v ˙ T ( s ) = − v T ( s ) A b ( s ) \dot{v}^T(s)=-v^T(s)A_b(s) v ˙ T ( s ) = − v T ( s ) A b ( s ) with v T ( T ) = ( 1 , 0 ) v^T(T)=(1,0) v T ( T ) = ( 1 , 0 ) . The stationary-lag kernel in Eq. (60) is obtained only after the long-age limit, rather than by replacing T T T with the lag.
Paired Novikov average
For distinct w 1 , w 2 w_1,w_2 w 1 , w 2 , the inverse transform of Eq. (59) is
h ( θ ) ℓ = ( d − w 1 ) e − w 1 θ − ( d − w 2 ) e − w 2 θ w 2 − w 1 . (99) \frac{h(\theta)}{\ell}=\frac{(d-w_1)e^{-w_1\theta}-(d-w_2)e^{-w_2\theta}}{w_2-w_1}. \tag*{(99)} ℓ h ( θ ) = w 2 − w 1 ( d − w 1 ) e − w 1 θ − ( d − w 2 ) e − w 2 θ . (99)
Combining the two terms gives Eq. (60) . Locally, w ˉ \bar{w} w ˉ and ϵ w 2 \epsilon_w^2 ϵ w 2 are analytic in Δ a \Delta_a Δ a , while cosh ( ϵ w θ ) \cosh(\epsilon_w\theta) cosh ( ϵ w θ ) and sinhc ( ϵ w θ ) \operatorname{sinhc}(\epsilon_w\theta) sinhc ( ϵ w θ ) are even in ϵ w \epsilon_w ϵ w . Hence the kernel is regular at every fixed finite lag. At the EP, A ∞ = − w I + N J A_\infty=-wI+N_J A ∞ = − w I + N J , with ( N J ) 12 = − i ℓ (N_J)_{12}=-i\ell ( N J ) 12 = − i ℓ and N J 2 = 0 N_J^2=0 N J 2 = 0 , which yields Eq. (61) . The coefficient d − w d-w d − w cannot vanish: w = d w=d w = d in w 2 = ( d 2 − ν 2 ) / 2 w^2=(d^2-\nu^2)/2 w 2 = ( d 2 − ν 2 ) /2 would require d 2 = − ν 2 d^2=-\nu^2 d 2 = − ν 2 , inconsistent with γ , Ω > 0 \gamma,\Omega>0 γ , Ω > 0 . The same substitution in Eq. (90) excludes a canceled pole away from the EP when B v d ≠ 0 Bvd\ne0 B v d = 0 .
The coupled noise average and its adjoint follow the exact quantum-Brownian-motion construction of Ref. [22 ] . The correlation convention in Eq. (3) fixes both contractions:
E 0 [ z u Π t ] = ∫ 0 t α ( u , v ) E 0 [ O ( t , v ) Π t ] d v , \mathbb{E}_0[z_u\Pi_t]=\int_0^t\alpha(u,v)\mathbb{E}_0[O(t,v)\Pi_t]\,\mathrm{d}v, E 0 [ z u Π t ] = ∫ 0 t α ( u , v ) E 0 [ O ( t , v ) Π t ] d v ,
E 0 [ z u ∗ Π t ] = ∫ 0 t α ( v , u ) E 0 [ Π t O † ( t , v ) ] d v . (100) \mathbb{E}_0[z_u^*\Pi_t]=\int_0^t\alpha(v,u)\mathbb{E}_0[\Pi_tO^\dagger(t,v)]\,\mathrm{d}v. \tag*{(100)} E 0 [ z u ∗ Π t ] = ∫ 0 t α ( v , u ) E 0 [ Π t O † ( t , v )] d v . (100)
At fixed endpoint t t t , set A t , s = f ( t , s ) Q + g 0 ( t , s ) P + r 0 ( t , s ) N A_{t,s}=f(t,s)Q+g_0(t,s)P+r_0(t,s)N A t , s = f ( t , s ) Q + g 0 ( t , s ) P + r 0 ( t , s ) N and Y s = E 0 [ O ( t , s , z ∗ ) Π t ] Y_s=\mathbb{E}_0[O(t,s,z^*)\Pi_t] Y s = E 0 [ O ( t , s , z ∗ ) Π t ] . Novikov averaging couples Y s Y_s Y s to its adjoint:
Y s = A t , s ρ t + ∫ 0 t K t ( s , v ) Y v † d v , Y_s=A_{t,s}\rho_t+\int_0^t\mathcal{K}_t(s,v)Y_v^\dagger\,\mathrm{d}v, Y s = A t , s ρ t + ∫ 0 t K t ( s , v ) Y v † d v ,
K t ( s , v ) = i ∫ 0 t j ( t , s , u ) α ( v , u ) d u . (101) \mathcal{K}_t(s,v)=i\int_0^t j(t,s,u)\alpha(v,u)\,\mathrm{d}u. \tag*{(101)} K t ( s , v ) = i ∫ 0 t j ( t , s , u ) α ( v , u ) d u . (101)
The adjoint equation is
Y s † = ρ t A t , s † + ∫ 0 t K t ( s , v ) ‾ Y v d v . (102) Y_s^\dagger=\rho_t A_{t,s}^\dagger+\int_0^t\overline{\mathcal{K}_t(s,v)}Y_v\,\mathrm{d}v. \tag*{(102)} Y s † = ρ t A t , s † + ∫ 0 t K t ( s , v ) Y v d v . (102)
For X k ∈ { Q , P , N } X_k\in\{Q,P,N\} X k ∈ { Q , P , N } , expand Y s = ∑ k [ u k ( s ) X k ρ t + v k ( s ) ρ t X k ] Y_s=\sum_k[u_k(s)X_k\rho_t+v_k(s)\rho_tX_k] Y s = ∑ k [ u k ( s ) X k ρ t + v k ( s ) ρ t X k ] and denote the corresponding coefficients A t , s A_{t,s} A t , s by a k ( s ) a_k(s) a k ( s ) . Since Q , P , N Q,P,N Q , P , N are Hermitian, matching their left and right actions on ρ t \rho_t ρ t gives
u k = a k + K t v k ∗ , v k = K t u k ∗ . (103) u_k=a_k+\mathcal{K}_t v_k^*,
\qquad
v_k=\mathcal{K}_t u_k^*. \tag*{(103)} u k = a k + K t v k ∗ , v k = K t u k ∗ . (103)
Here a Q = f a_Q=f a Q = f , a P = g 0 a_P=g_0 a P = g 0 , and a N = r 0 a_N=r_0 a N = r 0 . The product K t K t ‾ \mathcal{K}_t\overline{\mathcal{K}_t} K t K t means integral-operator composition with pointwise complex conjugation, without exchanging the kernel arguments; it is not K t K t † \mathcal{K}_t\mathcal{K}_t^\dagger K t K t † . Eliminating v k v_k v k gives
u k = ( I − K t K t ‾ ) − 1 a k , v k = K t u k ∗ . (104) u_k=(I-\mathcal{K}_t\overline{\mathcal{K}_t})^{-1}a_k,
\qquad
v_k=\mathcal{K}_t u_k^*. \tag*{(104)} u k = ( I − K t K t ) − 1 a k , v k = K t u k ∗ . (104)
Invertibility of I − K t K t ‾ I-\mathcal{K}_t\overline{\mathcal{K}_t} I − K t K t specifies the time-local averaging chart; ∥ K t ∥ < 1 \lVert\mathcal{K}_t\rVert<1 ∥ K t ∥ < 1 is sufficient, but is not assumed as a universal bound. This inverse supplies the present kernel representation of that paired closure.
Define
U k = ∫ 0 t α ( t , s ) u k ( s ) d s , V k = ∫ 0 t α ( t , s ) v k ( s ) d s . (105) U_k=\int_0^t \alpha(t,s)u_k(s)\,ds,\qquad V_k=\int_0^t \alpha(t,s)v_k(s)\,ds. \tag*{(105)} U k = ∫ 0 t α ( t , s ) u k ( s ) d s , V k = ∫ 0 t α ( t , s ) v k ( s ) d s . (105)
Then A t = ∑ k ( U k X k ρ t + V k ρ t X k ) A_t=\sum_k(U_kX_k\rho_t+V_k\rho_tX_k) A t = ∑ k ( U k X k ρ t + V k ρ t X k ) . Substitution into Eq. (65) , using L = L † = ℓ Q L=L^\dagger=\ell Q L = L † = ℓ Q , combines the coefficients into U k − V k ∗ U_k-V_k^* U k − V k ∗ and gives Eqs. (67) –(70) . The scalar-history correlations remain in these effective coefficients.
Complete Gaussian norm integral
For the finite Gaussian source in Appendix K, the linear displacement and quadratic scalar equations in Eq. (47) give
b = b 0 + β T x , b=b_0+\beta^T x, b = b 0 + β T x ,
c ψ = c ψ , 0 + ζ T x + 1 2 x T Γ x , Γ T = Γ . (106) c_\psi=c_{\psi,0}+\zeta^T x+\frac{1}{2}x^T\Gamma x,\qquad\Gamma^T=\Gamma. \tag*{(106)} c ψ = c ψ , 0 + ζ T x + 2 1 x T Γ x , Γ T = Γ. (106)
Here x ∼ N ( 0 , I ) x\sim N(0,I) x ∼ N ( 0 , I ) is a finite real Gaussian source vector, distinct from the position coordinate in the wavefunction. The source-independent terms b 0 b_0 b 0 , c ψ , 0 c_{\psi,0} c ψ , 0 and response coefficients β \beta β , ζ \zeta ζ , Γ \Gamma Γ depend on time. The Gaussian integrals retain the full scalar c ψ c_\psi c ψ . The squared norm of the stochastic ket, which supplies its statistical weight, is
w ( x ) = ∥ ψ t ∥ 2 = π a R exp [ 2 Re c ψ + ( Re b ) 2 a R ] . (107) w(x)=\|\psi_t\|^2
=\sqrt{\frac{\pi}{a_R}}\exp\left[2\operatorname{Re}c_\psi+\frac{(\operatorname{Re}b)^2}{a_R}\right]. \tag*{(107)} w ( x ) = ∥ ψ t ∥ 2 = a R π exp [ 2 Re c ψ + a R ( Re b ) 2 ] . (107)
Combining this weight with the unit Gaussian source measure gives the real precision matrix and linear term
P = I − 2 Re Γ − 2 a R β R β R T , P=I-2\operatorname{Re}\Gamma-\frac{2}{a_R}\beta_R\beta_R^T, P = I − 2 Re Γ − a R 2 β R β R T ,
q x = 2 Re ζ + 2 Re b 0 a R β R . (108) q_x=2\operatorname{Re}\zeta+\frac{2\operatorname{Re}b_0}{a_R}\beta_R. \tag*{(108)} q x = 2 Re ζ + a R 2 Re b 0 β R . (108)
Here β R = Re β \beta_R=\operatorname{Re}\beta β R = Re β and β I = Im β \beta_I=\operatorname{Im}\beta β I = Im β are real column vectors. When P ≻ 0 P\succ0 P ≻ 0 , completion of the square gives
x ˉ = P − 1 q x , Σ = P − 1 , (109) \bar{x}=P^{-1}q_x,\qquad\Sigma=P^{-1}, \tag*{(109)} x ˉ = P − 1 q x , Σ = P − 1 , (109)
with absolute total weight
Z = π a R det P exp ( χ ) , Z=\sqrt{\frac{\pi}{a_R\det P}}\exp(\chi), Z = a R det P π exp ( χ ) ,
χ = 2 Re c ψ , 0 + ( Re b 0 ) 2 a R + 1 2 q x T P − 1 q x . (110) \chi=2\operatorname{Re}c_{\psi,0}+\frac{(\operatorname{Re}b_0)^2}{a_R}+\frac{1}{2}q_x^TP^{-1}q_x. \tag*{(110)} χ = 2 Re c ψ , 0 + a R ( Re b 0 ) 2 + 2 1 q x T P − 1 q x . (110)
The matrix mapping source fluctuations to the state centroid is
B = T a ( β R T β I T ) , m 0 = T a ( Re b 0 Im b 0 ) . (111) B=T_a\begin{pmatrix}\beta_R^T\\ \beta_I^T\end{pmatrix},\qquad
m_0=T_a\begin{pmatrix}\operatorname{Re}b_0\\ \operatorname{Im}b_0\end{pmatrix}. \tag*{(111)} B = T a ( β R T β I T ) , m 0 = T a ( Re b 0 Im b 0 ) . (111)
Qubit readout selection and direct numerical comparisons
Direct NMQSD and NMME calculations
We propagate system amplitudes and the full system density operator. We set γ = A 0 = 1 \gamma=A_0=1 γ = A 0 = 1 , Ω = 0 \Omega=0 Ω = 0 , and use the three shifts ϵ / γ = 0 , − 0.1 , + 0.1 \epsilon/\gamma=0,-0.1,+0.1 ϵ / γ = 0 , − 0.1 , + 0.1 of Eq. (20) . Two initial states, ( ∣ G S ⟩ + ∣ j ⟩ ) / 2 (|GS\rangle+|j\rangle)/\sqrt{2} ( ∣ GS ⟩ + ∣ j ⟩) / 2 with j = 1 , 2 j=1,2 j = 1 , 2 , determine both columns of the system propagator. The nonlinear coefficient row in Sec. III is integrated first; its dense interpolant drives separate integrations of G ˙ = ( − i H 1 − κ † F q ) G \dot{G}=(-iH_1-\kappa^\dagger F_q)G G ˙ = ( − i H 1 − κ † F q ) G and of the full 3 × 3 3\times3 3 × 3 system density operator in Eq. (25) . The two routes share the exact coefficient closure and provide different objectives. Their agreement tests the internal NMQSD-to-NMME calculation. Figure 5 shows selected coherence channels and their finite-time ratio.
Figure 5. Local quadrature selection within the exact NMME. The preparation is ( ∣ G S ⟩ + ∣ 1 ⟩ ) / 2 (|G_S\rangle+|1\rangle)/\sqrt{2} ( ∣ G S ⟩ + ∣1 ⟩) / 2 , with the benchmark parameters of Eq. (20) and a fixed rotating-phase reference. (a) At the EP, X 1 X_1 X 1 has a linear polynomial prefactor, Y 1 Y_1 Y 1 has linear and quadratic terms, and X 2 X_2 X 2 selects the quadratic term. (b) Direct NMME evolution of X 2 X_2 X 2 for the EP and the two shifts ε = ϵ / γ = ± 0.1 \varepsilon=\epsilon/\gamma=\pm0.1 ε = ϵ / γ = ± 0.1 . The circle marks the EP maximum magnitude at τ = 6 \tau=6 τ = 6 . (c) The signed three-time ratio at the same 161 centers for each evolution. The EP value is 9 / 16 9/16 9/16 ; the detuned controls depart at finite times. For R ≤ 0 R\le0 R ≤ 0 , the real logarithmic degree ln R / ln ( 3 / 4 ) \ln R/\ln(3/4) ln R / ln ( 3/4 ) is undefined. This ratio is a model-dependent diagnostic, not a general EP criterion.
The integrations use DOP853 [43 ] with relative tolerance 2 × 10 − 12 2\times10^{-12} 2 × 1 0 − 12 and absolute tolerance 2 × 10 − 14 2\times10^{-14} 2 × 1 0 − 14 . We save 7201 uniformly spaced times over 0 ≤ τ = γ t ≤ 18 0\le\tau=\gamma t\le18 0 ≤ τ = γ t ≤ 18 , with spacing 0.0025 0.0025 0.0025 . Across both preparations, all system-coherence components, and the three shifts, the maximum sampled absolute difference between NMQSD and NMME is 8.50 × 10 − 12 8.50\times10^{-12} 8.50 × 1 0 − 12 . At the EP, the maximum absolute discrepancy of the doubled quadratures X 1 , Y 1 , X 2 X_1,Y_1,X_2 X 1 , Y 1 , X 2 from their analytic expressions is 1.28 × 10 − 12 1.28\times10^{-12} 1.28 × 1 0 − 12 . All reported numerical discrepancies and conditioning diagnostics are discrete samples on the stated grids; they provide neither continuous-time error bounds nor a guarantee of invertibility between samples.
The system block of e A q t e^{A_qt} e A q t , derived from the internal coordinate W = F q G W=F_qG W = F q G , agrees with the NMQSD amplitude matrix within 1.69 × 10 − 11 1.69\times10^{-11} 1.69 × 1 0 − 11 and with the NMME coherence matrix within 6.44 × 10 − 13 6.44\times10^{-13} 6.44 × 1 0 − 13 on the same grid. Density-matrix trace and Hermiticity discrepancies remain below 7 × 10 − 16 7\times10^{-16} 7 × 1 0 − 16 , at roundoff scale. The minimum sampled density eigenvalue is − 7.93 × 10 − 13 -7.93\times10^{-13} − 7.93 × 1 0 − 13 , at the numerical error scale; positivity of the exact map follows from the analytic contraction argument in Appendix B.
A separate coefficient-only integration reaches τ = 1000 \tau=1000 τ = 1000 without dividing underflowing homogeneous coordinates. Its maximum sampled absolute discrepancy in c q / γ c_q/\gamma c q / γ is 2.07 × 10 − 12 2.07\times10^{-12} 2.07 × 1 0 − 12 , and the final value of t [ 2 γ / 3 − Re c q ( t ) ] t[2\gamma/3-\operatorname{Re}c_q(t)] t [ 2 γ /3 − Re c q ( t )] is approximately 1.99.
Selected quadrature and finite-time ratio
The selected X 2 X_2 X 2 signal in Eq. (40) gives a one-channel diagnostic. For a single component S ( t ) = A t m e − β t S(t)=At^m e^{-\beta t} S ( t ) = A t m e − β t ,
R ( t , h ) = S ( t − h ) S ( t + h ) S ( t ) 2 = [ 1 − ( h t ) 2 ] m , t > h > 0. (112) R(t,h)=\frac{S(t-h)S(t+h)}{S(t)^2}
=\left[1-\left(\frac{h}{t}\right)^2\right]^m,\qquad t>h>0. \tag*{(112)} R ( t , h ) = S ( t ) 2 S ( t − h ) S ( t + h ) = [ 1 − ( t h ) 2 ] m , t > h > 0. (112)
The denominator must be nonzero. A constant gain and the exponential factor cancel. For the EP signal X 2 ( t ) = − ( 3 / 27 ) τ 2 e − τ / 3 X_2(t)=-(\sqrt{3}/27)\tau^2e^{-\tau/3} X 2 ( t ) = − ( 3 /27 ) τ 2 e − τ /3 , this gives
R ( t , t / 2 ) = X 2 ( t / 2 ) X 2 ( 3 t / 2 ) X 2 ( t ) 2 = 9 16 . (113) R(t,t/2)=\frac{X_2(t/2)X_2(3t/2)}{X_2(t)^2}=\frac{9}{16}. \tag*{(113)} R ( t , t /2 ) = X 2 ( t ) 2 X 2 ( t /2 ) X 2 ( 3 t /2 ) = 16 9 . (113)
Figure 5 compares this signed ratio from the unconditional NMME quadratures at 161 centers 2 ≤ τ ≤ 10 2\le\tau\le10 2 ≤ τ ≤ 10 for the EP and both detunings. For the negative and positive shifts, respectively, its values at τ = 6 \tau=6 τ = 6 are approximately 0.51 and 0.48; at τ = 10 \tau=10 τ = 10 they are approximately − 0.33 -0.33 − 0.33 and 0.08.
For these noiseless, model-dependent readouts, additive offsets, a shifted time origin, phase mixing, and a small denominator can alter the ratio. Both detuned controls also begin quadratically, which precludes using the short-time ratio alone as an EP test. Reconstructing the full coefficient row requires the separately prepared ensembles in Eq. (41) .
Coefficient reconstruction and conditioning
Equations (42) and (43) reconstruct the specified coefficients from the separate preparations and phase-calibrated coherences in Eq. (41) , on an invertible chart. A derivative-free interval form follows on a path where the determinant never vanishes:
1 Δ t ∫ t t + Δ t c q ( u ) d u = − i tr H 1 − 1 Δ t Log c o n t det G ( t + Δ t ) det G ( t ) . (114) \frac{1}{\Delta t}\int_t^{t+\Delta t}c_q(u)\,\mathrm{d}u
=-\mathrm{i}\operatorname{tr}H_1
-\frac{1}{\Delta t}\operatorname{Log}_{\mathrm{cont}}\frac{\det G(t+\Delta t)}{\det G(t)}. \tag*{(114)} Δ t 1 ∫ t t + Δ t c q ( u ) d u = − i tr H 1 − Δ t 1 Log cont det G ( t ) det G ( t + Δ t ) . (114)
The logarithm follows the determinant phase continuously; nonzero endpoints alone do not establish a regular path. Pointwise differentiation amplifies measurement and discretization errors, while small singular values of G G G amplify the inverse. A condition number alone is not an absolute guard: a scalar map has condition number one at every nonzero value. The reconstructed contraction is compared with its coefficient solution using the direct integrations in Appendix I1.
The numerical reconstruction uses G i j = 2 ρ i G ( j ) G_{ij}=2\rho^{(j)}_{iG} G ij = 2 ρ i G ( j ) from each NMME preparation. At the 7193 admissible centers 0.01 ≤ τ ≤ 17.99 0.01\le\tau\le17.99 0.01 ≤ τ ≤ 17.99 , a centered five-point derivative with step h = 0.005 / γ h=0.005/\gamma h = 0.005/ γ supplies G ˙ \dot{G} G ˙ in Eqs. (42) and (43) .
Figure 6 and Table I compare the resulting F q F_q F q and c q c_q c q with the separately integrated coefficient solution.
Figure 6. System readouts recover the coefficient contraction at the EP. The line gives the direct NMQSD coefficient solution Re c q / γ \operatorname{Re} c_q/\gamma Re c q / γ . Open circles reconstruct the contraction from the exact NMME coherences for the two preparations ( ∣ G S ⟩ + ∣ j ⟩ ) / 2 (|G_S\rangle+ |j\rangle)/\sqrt{2} ( ∣ G S ⟩ + ∣ j ⟩) / 2 , j = 1 , 2 j=1,2 j = 1 , 2 , using a five-point derivative. Both calculations use the same specified H 1 H_1 H 1 , coupling convention, and phase reference. The plot is noiseless; the inverse is model dependent and requires an invertible system propagator.
Table I. Sampled errors in the noiseless NMME-based coefficient reconstruction. In the first row, the maximum is over the 7193 admissible times and both components of Δ F q \Delta F_q Δ F q ; the second row uses the same times. The third row takes the maximum over all 7201 saved times on 0 ≤ τ ≤ 18 0 \leq\tau\leq18 0 ≤ τ ≤ 18 , including the endpoints.
The minimum sampled singular values of G G G over the full time grid are approximately 1.9 × 10 − 3 1.9 \times10^{-3} 1.9 × 1 0 − 3 , 4.6 × 10 − 4 4.6 \times10^{-4} 4.6 × 1 0 − 4 , and 1.1 × 10 − 2 1.1 \times10^{-2} 1.1 × 1 0 − 2 for the EP, negative shift, and positive shift. The exact EP denominator in Eq. (21) is nonzero for all t ≥ 0 t \geq0 t ≥ 0 .
Finite-time oscillator coefficient and state comparisons
Figure 7 and Figure 8 compare the coefficients and full Gaussian-state trajectories for the preparation in Sec. IV C 2; Table II gives their endpoint values.
Table II. Reduced Gaussian state at ν t = 8 \nu t=8 ν t = 8 . The three fixed couplings share the initial state and bath frequency specified in Sec. IV C 2. Values are rounded from the 256-mode norm-weighted NMQSD averages, with refinement checks in Appendix K.
Figure 7. The noise average changes the coefficients that govern the reduced state. The upper and lower rows show F , f F,f F , f and G , g G,g G , g , respectively, for λ = 0.7 , 1 , 1.3 \lambda=0.7,1,1.3 λ = 0.7 , 1 , 1.3 . (a),(b),(d),(e) real and imaginary parts of the raw O O O -operator coefficients (dashed) and effective master-equation coefficients (solid); (c),(f) their noise-induced corrections, with real parts solid and imaginary parts dotted. Parameter identities are distinguished by circles, squares, and triangles. The imaginary parts of f f f and g g g enter the restoring force and damping, while their real parts enter diffusion. Parameters are those of Figure 3 , with a ( 0 ) = 1 a(0)=1 a ( 0 ) = 1 , b ( 0 ) = 0 b(0)=0 b ( 0 ) = 0 , and the normalized vacuum wavefunction. The three fixed couplings compare finite-time dynamics across the coefficient exceptional point.
Figure 8. Reduced-state evolution from the same Gaussian initial state. Solid lines show independent integration of the master-equation moment equations; hollow symbols show the norm-weighted NMQSD average at every third saved time and the final endpoint. The panels give the two means, purity P = 1 / ( 2 det V ) P = 1/(2\sqrt{\det V}) P = 1/ ( 2 det V ) , and three independent covariance entries. Their agreement tests the complete noise average over 0 ≤ ν t ≤ 8 0 \le\nu t \le8 0 ≤ ν t ≤ 8 for the three fixed couplings. The shared initial state has m ( 0 ) = 0 m(0) = 0 m ( 0 ) = 0 and V ( 0 ) = I 2 / 2 V(0) = I_2/2 V ( 0 ) = I 2 /2 ; other parameters are those of Figure 3 .
The complete average changes the coefficients that act on the density operator. Figure 7 shows the original F F F , G G G , the effective f f f , g g g , and the corrections separately. For example, at ν t = 8 \nu t=8 ν t = 8 and λ = 1 \lambda=1 λ = 1 , Re G ≈ − 0.033 \operatorname{Re}G \approx-0.033 Re G ≈ − 0.033 whereas Re g ≈ − 0.044 \operatorname{Re}g \approx-0.044 Re g ≈ − 0.044 . This change enters the cross diffusion in Eq. (70) .
The corresponding imaginary parts modify the deterministic motion. The full-state trajectories in Figure 8 show oscillatory mean motion while the initially pure reduced state becomes mixed. The three couplings produce different displacement, covariance, and purity trajectories over the common interval. For a single-mode Gaussian state, the purity is P = tr ρ 2 = 1 / ( 2 det V ) P=\operatorname{tr}\rho^2=1/(2\sqrt{\det V}) P = tr ρ 2 = 1/ ( 2 det V ) [44 ] . Across these three couplings at ν t = 8 \nu t=8 ν t = 8 , stronger coupling lowers the purity and raises the position variance. These finite-time comparisons do not by themselves establish an exceptional point of the reduced-state generator.
Numerical representation and convergence
The diagnostics below concern the stated preparation and finite comparison grids. Both propagation routes share the exact coefficient core, so their agreement tests the subsequent averaging and moment propagation; it is not an absolute error bound for the continuous exact solution. The source-kernel discrepancy is likewise evaluated on the specified grid.
On the fixed interval 0 ≤ t ≤ T 0 \leq t \leq T 0 ≤ t ≤ T , with T = 8 / ν T=8/\nu T = 8/ ν , the source uses M s r c = 128 M_{\mathrm{src}}=128 M src = 128 , 256 filtered temporal modes and an independent initial mode. The corresponding dimensions are 129 and 257 complex coordinates, or 258 and 514 real coordinates, respectively. The temporal basis consists of the orthonormal shifted Legendre functions
e j ( t ) = 2 j + 1 T P j ( 2 t / T − 1 ) . (115) e_j(t)=\sqrt{\frac{2j+1}{T}}P_j(2t/T-1). \tag*{(115)} e j ( t ) = T 2 j + 1 P j ( 2 t / T − 1 ) . (115)
Here j = 0 , … , M s r c − 1 j=0,\ldots,M_{\mathrm{src}}-1 j = 0 , … , M src − 1 and P j P_j P j is the Legendre polynomial of degree j j j . Their filtered components satisfy
ϕ ˙ j ( t ) = e j ( t ) − d ∗ ϕ j ( t ) , ϕ j ( 0 ) = 0. (116) \dot{\phi}_j(t)=e_j(t)-d^*\phi_j(t),\qquad\phi_j(0)=0. \tag*{(116)} ϕ ˙ j ( t ) = e j ( t ) − d ∗ ϕ j ( t ) , ϕ j ( 0 ) = 0. (116)
The finite colored source is
z M s r c ∗ ( t ) = A 0 e − d ∗ t ξ 0 + 2 γ A 0 ∑ j = 0 M s r c − 1 ϕ j ( t ) ξ j + 1 . (117) z_{M_{\mathrm{src}}}^*(t)=\sqrt{A_0}e^{-d^*t}\xi_0
+\sqrt{2\gamma A_0}\sum_{j=0}^{M_{\mathrm{src}}-1}\phi_j(t)\xi_{j+1}. \tag*{(117)} z M src ∗ ( t ) = A 0 e − d ∗ t ξ 0 + 2 γ A 0 j = 0 ∑ M src − 1 ϕ j ( t ) ξ j + 1 . (117)
where the ξ k \xi_k ξ k are independent standard proper complex Gaussian variables, with E 0 [ ξ k ξ l ∗ ] = δ k l \mathbb{E}_0[\xi_k\xi_l^*]=\delta_{kl} E 0 [ ξ k ξ l ∗ ] = δ k l and E 0 [ ξ k ξ l ] = 0 \mathbb{E}_0[\xi_k\xi_l]=0 E 0 [ ξ k ξ l ] = 0 . The independent initial term retains the stationary Ornstein–Uhlenbeck initial distribution. The real source in Eq. (106) is obtained by writing ξ k = ( x 2 k + 1 + i x 2 k + 2 ) / 2 \xi_k=(x_{2k+1}+ix_{2k+2})/\sqrt{2} ξ k = ( x 2 k + 1 + i x 2 k + 2 ) / 2 , with independent x j ∼ N ( 0 , 1 ) x_j\sim N(0,1) x j ∼ N ( 0 , 1 ) . The norm integral retains c ψ ( t , z ∗ ) c_\psi(t,z^*) c ψ ( t , z ∗ ) and its derivative and is evaluated analytically, without random draws.
Writing z M s r c ∗ ( t ) = ∑ k s k ( t ) ξ k z^*_{M_{\mathrm{src}}}(t)=\sum_k s_k(t)\xi_k z M src ∗ ( t ) = ∑ k s k ( t ) ξ k , the reconstructed source covariance is
α M s r c ( t , s ) = ∑ k s k ∗ ( t ) s k ( s ) . (118) \alpha_{M_{\mathrm{src}}}(t,s)=\sum_k s_k^*(t)s_k(s). \tag*{(118)} α M src ( t , s ) = k ∑ s k ∗ ( t ) s k ( s ) . (118)
In the numerical units ν = 1 \nu=1 ν = 1 , the kernel comparison grid T \mathcal{T} T is the union of the 33 output times { 0 , 0.25 , … , 8 } \{0,0.25,\ldots,8\} { 0 , 0.25 , … , 8 } and the 128-point Gauss–Legendre nodes on [ 0 , 8 ] [0,8] [ 0 , 8 ] . Its maximum-entry discrepancy is
ϵ k e r ( M s r c ) = max t , s ∈ T ∣ α M s r c ( t , s ) − α ( t , s ) ∣ max { 1 , max t , s ∈ T ∣ α ( t , s ) ∣ } . (119) \epsilon_{\mathrm{ker}}^{(M_{\mathrm{src}})}=\frac{\max_{t,s\in\mathcal{T}}\left|\alpha_{M_{\mathrm{src}}}(t,s)-\alpha(t,s)\right|}{\max\left\{1,\max_{t,s\in\mathcal{T}}\left|\alpha(t,s)\right|\right\}}. \tag*{(119)} ϵ ker ( M src ) = max { 1 , max t , s ∈ T ∣ α ( t , s ) ∣ } max t , s ∈ T ∣ α M src ( t , s ) − α ( t , s ) ∣ . (119)
For M s r c = 256 M_{\mathrm{src}}=256 M src = 256 , this diagnostic ranges from approximately 4.13 × 10 − 4 4.13\times10^{-4} 4.13 × 1 0 − 4 to 7.68 × 10 − 4 7.68\times10^{-4} 7.68 × 1 0 − 4 across the three couplings.
The master-equation calculation starts from the same initial state and the same exact coefficient core as the NMQSD calculation, then independently computes the paired average and propagates the moments. The paired average uses order-64 quadrature at 129 coefficient times, and order-128 comparisons at ν t = 0.5 , 1 , 2 , 4 , 6 , 8 \nu t=0.5,1,2,4,6,8 ν t = 0.5 , 1 , 2 , 4 , 6 , 8 . The largest recorded condition number of the discretized paired-averaging matrix I − K t K ~ t I-K_t\widetilde{K}_t I − K t K t in Eq. (104) is below 1.6.
The master-equation propagation compares coefficient spacings ν Δ t = 0.125 \nu\Delta t=0.125 ν Δ t = 0.125 and 0.0625 0.0625 0.0625 , with relative and absolute integration tolerances 10 − 10 10^{-10} 1 0 − 10 and 10 − 12 10^{-12} 1 0 − 12 . Real and imaginary cubic interpolation supplies the coefficients between their saved times. The 33 common output times are separated by ν Δ t = 0.25 \nu\Delta t=0.25 ν Δ t = 0.25 . Interpolation is used for propagation; the figures connect the original saved coefficient or state nodes.
For a compared quantity X X X , define
ϵ ( X ) = ∥ X − X r e f ∥ max max ( 1 , ∥ X r e f ∥ max ) . (120) \epsilon(X) = \frac{\lVert X-X_{\mathrm{ref}}\rVert_{\max}}{\max(1,\lVert X_{\mathrm{ref}}\rVert_{\max})}. \tag*{(120)} ϵ ( X ) = max ( 1 , ∥ X ref ∥ m a x ) ∥ X − X ref ∥ m a x . (120)
The reference is the finer representation for source and mesh comparisons, and the 256-mode norm-weighted NMQSD result for the master-equation comparison. The denominators are one for the reported moment discrepancies; these are not percentages relative to each small component. Table III separates source refinement, time-mesh refinement, and agreement of the two propagation routes.
Table III. Maximum covariance discrepancies over the 33 output times. The columns compare source modes 128/256, master-equation coefficient spacings ν Δ t = 0.125 / 0.0625 \nu\Delta t = 0.125/0.0625 ν Δ t = 0.125/0.0625 , and the finer master-equation trajectory against the 256-mode NMQSD average, respectively. Discrepancies use Eq. (120) .
The absolute norm checks at the higher source resolution give ∣ log Z ∣ < 10 − 7 \lvert\log Z\rvert< 10^{-7} ∣ log Z ∣ < 1 0 − 7 and ∣ ∂ t log Z ∣ < 10 − 7 \lvert\partial_t \log Z\rvert< 10^{-7} ∣ ∂ t log Z ∣ < 1 0 − 7 . Over the 33 common output times for each coupling, the scaled discrepancies between the finer master-equation trajectory and the 256-mode NMQSD average are below 10 − 7 10^{-7} 1 0 − 7 in both m m m and V V V . The instantaneous covariance-derivative discrepancy is also below 10 − 7 10^{-7} 1 0 − 7 . The largest 128/256 source difference among m m m , V V V , m ˙ \dot{m} m ˙ , V ˙ \dot{V} V ˙ is below 10 − 6 10^{-6} 1 0 − 6 .
Physicality checks cover the reduced, conditional, and centroid covariances. In the convention [ Q , P ] = i [Q,P] = i [ Q , P ] = i , the Gaussian quantum condition [44 ] is V + i J 2 / 2 ⪰ 0 V+iJ_2/2 \succeq0 V + i J 2 /2 ⪰ 0 , where J 2 = ( 0 1 − 1 0 ) J_2 = \begin{pmatrix} 0 & 1 \\ -1 & 0 \end{pmatrix} J 2 = ( 0 − 1 1 0 ) , and V c e n t r o i d ⪰ 0 V_{\mathrm{centroid}} \succeq0 V centroid ⪰ 0 . The saved Gaussian states satisfy these conditions within numerical tolerance; the master-equation output and accepted adaptive states are checked as well. Purity uses valid covariance nodes, including every displayed node. These sampled-state checks do not prove continuous-time physicality or complete positivity for all preparations.
References