---
title: Energy-Conserving Redfield Equation Analysis
url: https://www.emergentmind.com/topics/energy-conserving-redfield-equation
type: topic
---

# Energy-Conserving Redfield Equation Analysis

The expression **energy-conserving Redfield equation** does not denote a single universally standardized master equation across the literature. In the cited work, it refers most closely to two related constructions: first, the **secular** or **Bohr-frequency-resolved** reduction of Redfield dynamics, in which only frequency-matched terms are retained and the generator becomes of GKSL form; second, a more recent formulation in which **energy conservation is imposed already at the Born level**, yielding a kernel formally equivalent to the Lindblad equation without an additional rotating-wave approximation [1905.05068] [2602.13429]. Standard Redfield theory itself is already energy-weighted through bath correlation functions or frequency-domain response functions, but it is generally **non-secular**, can couple populations and coherences, and is not guaranteed to be completely positive [2108.03128] [2309.07105].

## 1. Standard Redfield equation and the origin of energy dependence

A standard starting point is a weak-coupling decomposition of the total Hamiltonian,
$$
H = H_S + H_R + \lambda H_I,\qquad H_I=\sum_\alpha T_\alpha\otimes B_\alpha,
$$
with reduced dynamics generated perturbatively from the full von Neumann equation [2108.03128]. In the Bogoliubov derivation, the second-order Schrödinger-picture master equation is
$$
\dot \rho_S = -i[H_S,\rho_S] +\lambda^2 \mathcal G_2 \rho_S,
$$
with
$$
\mathcal G_2 \rho_S = -\int_0^\infty ds\, \operatorname{Tr}_R \left[ \mathcal L_I \mathcal L_I(-s)\bigl(\rho_S\otimes \rho_R^{\rm ref}\bigr) \right].
$$
This is the standard Redfield generator in that framework [2108.03128].

An equivalent global formulation writes the reduced density matrix as
$$
\partial_t \varrho(t) = -i [H_S,\varrho(t)] + \left([u\,\varrho(t),v]+\mathrm{h.c.}\right),
$$
with Redfield jump operator
$$
u = \int_0^\infty d\tau\, \tilde v(-\tau)\, C(\tau),
$$
where \(C(\tau)\) is the bath correlation function [2309.07105]. In the system eigenbasis,
$$
u=\sum_{kq} \hat v_{kq}\, W(E_k-E_q),
$$
so each transition operator \(\hat v_{kq}\) is weighted by the bath transform \(W(E_k-E_q)\) at the corresponding transition energy [2309.07105].

This frequency weighting is the basic sense in which standard Redfield dynamics is already **energy aware**. It recognizes system transition energies through \(W(E_k-E_q)\), or equivalently through the Fourier-Laplace transforms of bath correlations evaluated at system transition frequencies, but it does **not** by itself impose strict frequency selection in the secular sense [1507.02755] [2309.07105].

## 2. Bohr frequencies, secularization, and the usual meaning of “energy-conserving”

Energy selection enters most transparently after decomposing system operators into Bohr-frequency components,
$$
T_\alpha(t)=\sum_\omega e^{-i\omega t} T_\alpha(\omega).
$$
The interaction-picture Redfield kernel then contains oscillatory factors \(e^{i(\omega-\omega')t}\), and the **secular approximation** drops terms with \(\omega\neq\omega'\), keeping only \(\omega=\omega'\) terms [2108.03128]. In this standard sense, the secular approximation is the usual mechanism by which Redfield dynamics becomes “energy-conserving” in the reduced-system description.

The same point appears in the HEOM-based derivation of higher-order Redfield corrections. There the Markovian Redfield equation contains sums over \(\omega,\omega'\in\Omega\) with factors \(e^{i(\omega-\omega')t}\); if the evolution time scale is much larger than \(\max_{\omega\neq\omega'}|\omega-\omega'|^{-1}\), the \(\omega\neq\omega'\) terms can be neglected as rapidly oscillating, yielding the secular approximation and a GKSL generator [1905.05068]. The resulting equation is
$$
\dot{\rho}(t) = -i[H_{\mathrm{LS}},\rho(t)] + v^2 \sum_{j=1}^N\sum_{\omega\in\Omega} 2\,\mathrm{Re}\,\Gamma_j(\omega) \left( V_j(\omega)\rho(t)V_j^\dagger(\omega) - \frac12\{V_j^\dagger(\omega)V_j(\omega),\rho(t)\} \right),
$$
which is the paper’s explicit positivity-preserving secular Redfield form [1905.05068].

For particle-exchange Redfield equations, the same resonance structure appears at the rate level. In the electronically open-molecule derivation, secularization of the Redfield tensor gives population rates
$$
\begin{aligned}
k_{b\leftarrow a} = \sum_{pq}&\langle a|\hat a_p^\dagger|b\rangle\langle b|\hat a_q|a\rangle J_{pq}(\omega_{ab})[1-\bar n(\omega_{ab})] \\
+ \sum_{pq}&\langle a|\hat a_q|b\rangle\langle b|\hat a_p^\dagger|a\rangle J_{pq}(\omega_{ba})\bar n(\omega_{ba}),
\end{aligned}
$$
with the resonance condition originating from
$$
\int_{-\infty}^{\infty}e^{i(\omega_{ab}-\omega)\tau}d\tau =2\pi\delta(\omega_{ab}-\omega).
$$
In that unbroadened limit, transitions are energy matched in the ordinary golden-rule sense [2406.16443].

The literature also emphasizes that this strict resonance condition can be deliberately relaxed. In the same molecular context, the replacement
$$
\delta(\omega_{ab}\pm\omega)\to \frac{1}{\sqrt{2\pi\gamma^2}} \exp\!\left[-\frac{(\omega_{ab}\pm\omega)^2}{2\gamma^2}\right]
$$
“relaxes the resonance requirement for transitions between molecule states” [2406.16443]. This makes clear that “energy conserving” in Redfield theory usually means **transition energy selectivity**, not exact conservation of a reduced-system energy functional.

## 3. Main variants associated with the term

The literature supports three distinct usages.

| Variant | Defining feature | Energy-selection status |
|---|---|---|
| Standard Redfield | Non-secular second-order master equation | Energy weighted, not strictly frequency diagonal |
| Secular Redfield / GKLS form | Keep only \(\omega=\omega'\) terms | Bohr-frequency matched transitions only |
| Born-level energy-conserving Redfield | Impose energy-conservation delta functions in the kernel | Formally equivalent to Lindblad without extra RWA |

In the first case, standard Redfield includes nonsecular couplings between populations and coherences, is not guaranteed to be completely positive, and does not explicitly enforce strict separation by Bohr frequency [2108.03128]. In the second case, secularization yields the familiar frequency-diagonal GKSL structure [1905.05068]. In the third case, a field-theoretical analysis identifies an inconsistency in the standard Markovian Redfield kernel and resolves it by imposing energy conservation on the Born level [2602.13429].

The corresponding modified kernel is
$$
\begin{aligned}
\tilde K_{pp',qq'}= &-\frac{1}{2}\delta_{p'q'}\sum_{\alpha\beta l}S_{pl}^\alpha S_{lq}^\beta \tilde D^{\alpha\beta}(E_{ql})\delta(E_{pl}+E_{lq}) \\
&-\frac{1}{2}\delta_{pq}\sum_{\alpha\beta l }S_{q'l}^\alpha S_{lp'}^\beta \tilde D^{\alpha\beta}(E_{q'l})\delta(E_{q'l}+E_{lp'}) \\
&+\sum_{\alpha\beta }S_{pq}^\beta S_{q'p'}^\alpha \tilde D^{\alpha\beta}(E_{q'p'})\delta(E_{pq}+E_{q'p'}).
\end{aligned}
$$
The authors show that this “energy-conserving Redfield equation” is formally equivalent to the energy-basis form of the Lindblad equation, without invoking an additional rotating-wave approximation [2602.13429].

This suggests a sharp distinction. In the older and broader usage, “energy-conserving Redfield” usually means **secular Redfield**. In the more specialized recent usage, it can mean a **Born-kernel corrected Redfield equation** whose delta-function selection rules enforce equal energy transfer already before the secular step [2602.13429].

## 4. Local-energy-resolved and Lindbladized Redfield constructions

A separate research direction starts from the global non-secular Redfield equation and asks how its energy structure can be approximated without full diagonalization of a many-body Hamiltonian. In that setting, the exact jump operator
$$
u=\sum_{kq}\hat v_{kq}W(E_k-E_q)
$$
is expanded around a chosen energy \(\varepsilon_0\) as
$$
u \approx \sum_{n=0}^{N} \frac{W^{(n)}(\varepsilon_0)}{n!} ([H_S,\cdot]-\varepsilon_0\cdot)^n[v] \equiv \mathcal T_N^{\varepsilon_0}[v].
$$
For local bath coupling, this becomes a local-energy expansion around local transition energies \(\varepsilon_{kq}^{i_0}\), producing an approximate local Redfield operator \(u_{\mathrm{loc}}\) [2309.07105].

This construction is explicitly **not** a secular or Davies-type energy-conserving master equation. It is an **approximate local-energy-resolved non-secular Redfield framework**. Its control parameters are a short bath-correlation time, \(\tau_B/\tau_S\ll 1\), and small deviation of relevant transition energies from the chosen expansion energies [2309.07105]. The resulting local Lindblad approximation inherits this local-energy structure but is no longer dynamically equivalent to exact Redfield, because the negative dissipative channel is neglected [2309.07105].

Another line of work “tames” the Bloch-Redfield equation by reconstructing a positivity-preserving non-secular Lindblad equation directly in the transition basis \(\sigma_j=|n_j\rangle\langle m_j|\), with transition frequencies
$$
\omega_j=\frac{E_{m_j}-E_{n_j}}{\hbar}.
$$
The key frequency criterion is
$$
|\omega_i-\omega_j| < \max\{\Gamma_{ij},\Lambda_{ij}\},
$$
in which case the non-secular couplings should be retained, whereas for sufficiently separated frequencies they can be discarded through secularization [2402.06354]. The proposed reconstruction uses an arithmetic mean for the energy shift,
$$
\tilde{\Lambda}_{ij} = \frac{\Lambda_{ij}(\omega_i)+\Lambda_{ij}(\omega_j)}{2},
$$
and a geometric mean for the dissipator,
$$
\tilde{\Gamma}_{ij} = \sqrt{\Gamma_{ij}(\omega_i)}\sqrt{\Gamma_{ij}(\omega_j)},
$$
followed by projection of the Kossakowski matrix onto the positive-semidefinite cone [2402.06354]. The authors explicitly state that they do **not** address “thermodynamic properties or local conservation laws” in this reconstruction [2402.06354].

A related benchmark of the Nathan–Rudner Lindbladization finds that the resulting equation is a GKSL/Lindblad-form modification of Redfield rather than a simple secular approximation. In the damped harmonic oscillator benchmark, the short-time dynamics is generally much better captured by the time-dependent Redfield equation, whereas the Nathan–Rudner equation delivers results comparable to those of the rotating-wave approximation; in the low-temperature steady-state regime the Lindbladized equation performs better, while in the high-temperature steady-state regime Redfield performs better [2403.08320].

## 5. Correlated initial states, higher-order corrections, and equilibrium structure

The Bogoliubov derivation of Redfield theory places the reduced dynamics on a **kinetic manifold**,
$$
\rho(t)=\mathcal R\, \rho_S(t),
$$
with
$$
\mathcal R = \mathcal R_0 + \lambda \mathcal R_1 + \lambda^2 \mathcal R_2 + \cdots.
$$
Its zeroth-order term is
$$
\mathcal R_0\rho_S=\rho_S\otimes\rho_R^{\rm ref},
$$
but the higher orders encode correlated kinetic states [2108.03128]. A distinctive conclusion is that initially correlated states generated by prior system-reservoir interaction are naturally incorporated, and **the Redfield equation does not require modification in this case** [2108.03128].

The same framework gives compact autonomous higher-order corrections. Under the stated assumptions on reservoir correlations,
$$
\mathcal G_1=0,\qquad \mathcal G_3=0,
$$
and the fourth-order generator \(\mathcal G_4\) is obtained in an explicit three-time integral form [2108.03128]. These corrections are presented as improved perturbative accuracy rather than as positivity-restoring or conservation-law-enforcing modifications [2108.03128].

The equilibrium question is more delicate. In the Bogoliubov framework, for a thermal bath at inverse temperature \(\beta\),
$$
\rho_{SR,\beta}=Z^{-1}e^{-\beta H},\qquad
\rho_{S,\beta}=\operatorname{Tr}_R(Z^{-1}e^{-\beta H}),
$$
and the recovery map satisfies
$$
\mathcal R \rho_{S,\beta} = \rho_{SR,\beta}.
$$
If the reduced semigroup has a unique stationary state, that stationary reduced state is the **mean-force Gibbs state** rather than the bare Gibbs state \(e^{-\beta H_S}/Z_S\) in general [2108.03128].

A complementary result is that the usual second-order Redfield steady state is generally **not correct to second order** in the system-bath coupling. A modified stationary construction based on analytic continuation of the off-diagonal Redfield solution recovers the reduced equilibrium state exactly up to \(O(\lambda^2)\), without requiring fourth-order relaxation tensors [1203.6207]. The target state is the coupling-dependent generalized quantum Gibbs state
$$
\rho^{\mathrm{eq}}\propto \operatorname{Tr}_B e^{-\beta H_{\mathrm{tot}}},
$$
not merely the bare canonical Gibbs state of \(H_S\) [1203.6207]. This suggests that any “energy-conserving” Redfield construction aimed at equilibrium consistency must address finite-coupling stationary structure, not only Bohr-frequency selection.

## 6. Positivity, Markovianity, and regime of validity

The best-known limitation of standard Redfield dynamics is positivity. The second-order Markovian Redfield equation generally does not preserve positivity, whereas the secular approximation yields a GKSL generator and therefore preserves positivity [1905.05068]. A more intermediate route is the **partial-secular approximation** based on coarse graining. There the Bohr-frequency couplings are weighted by
$$
S_{\omega-\omega'}^{(\Delta t)} := \operatorname{sinc}\!\left[\frac{(\omega-\omega')\Delta t}{2}\right],
$$
so off-diagonal frequency sectors are suppressed rather than removed completely [1903.07324]. For sufficiently large coarse-graining time \(\Delta t\), the coefficient matrix becomes positive semidefinite and the generator is GKLS-compatible [1903.07324].

The HEOM-based analysis gives a direct sufficient condition for the secular approximation:
$$
\max_{j,\omega} \frac{\lambda_j}{\sqrt{\gamma_j^2+\omega^2} \left(\frac{2}{\beta\gamma_j}+1\right)}
< \min_{\omega\neq\omega'}|\omega-\omega'|.
$$
This is the paper’s explicit criterion for when the energy-conserving secular reduction is justified [1905.05068]. A separate sufficient condition is given for the validity of second-order Redfield itself:
$$
\max_k \frac{\lambda_k}{\gamma_k} \left(\frac{2}{\beta\gamma_k}+1\right) < 2\min_j \gamma_j.
$$
So weak coupling is needed already before the further secular step is assessed [1905.05068].

Benchmark studies reinforce that these distinctions are operationally significant. In highly non-Markovian regimes, slow bath modes can make ordinary Redfield unreliable, and frozen-mode or hybrid Redfield constructions improve performance by converting part of the bath from dynamical dissipative modes into static or classical disorder [1507.02755]. In the damped harmonic oscillator, time-dependent Redfield captures short-time dynamics better than Lindbladized Redfield, while Lindbladized Redfield can outperform Redfield in the low-temperature steady state where Redfield may become unphysical [2403.08320].

Taken together, these results support a precise but limited conclusion. **Energy-conserving Redfield equation** is best understood either as the **secular, Bohr-frequency-diagonal Redfield/GKLS equation**, or as a more recent **Born-level energy-conserving reformulation** of the Redfield kernel [1905.05068] [2602.13429]. Standard Redfield remains the broader non-secular framework: it already contains the relevant transition-energy structure, but it does so without strict frequency selection, without guaranteed complete positivity, and without a universal finite-coupling equilibrium correction [2108.03128] [1203.6207].

Source: https://www.emergentmind.com/topics/energy-conserving-redfield-equation