One scaling framework for three Lindblad limits

This note shows how three microscopic derivations of a Markovian Lindblad (GKSL) generator are really just variations on the same basic framework, differing in the relative strengths of the system, bath, and interaction Hamiltonians:

• the weak-coupling/Davies limit (Born–Markov plus a secular average),
• the singular-coupling/white-noise limit (fast bath with appropriately rescaled coupling), and
• the fast-system limit (“super-secular”).

The first two are familiar. The third is partially degenerate, and so less often discussed, but appears, e.g., in the classic decoherence literature on chiral molecules. The three variations are often presented in ways that make them look completely distinct, which obscures their close relationship.

The scaling ansatz and the Markov limit

Fix a system–bath decomposition and consider a one-parameter family of Hamiltonians

(1)  

where acts on the system Hilbert space, on the bath Hilbert space, and couples them. Think of as encoding how quickly the system rotates (), how quickly the bath decorrelates (), and how strongly the coupling acts (), all while keeping the physical time variable fixed.

We assume the usual minimal structure used in essentially all microscopic derivations of the Lindblad equation:

• Factorized initial state: .
• Stationary bath state: .
• Centered coupling: , i.e., no first-order drift.

A convenient form is

(2)  

The question is: for which can the reduced dynamics

have a nontrivial, finite limit that is a time-homogeneous quantum dynamical semigroup, i.e., a Lindblad equation?

Why the scaling constraint 2c = b is essentially forced

Go to the interaction picture with respect to the free Hamiltonian

Then the interaction-picture coupling is

with

Bath correlations in the stationary state are

Because , one has

(3)  

Now recall what every Born–Markov style derivation is doing at second order: the reduced generator is built from a memory integral of the form

(4)  

where “” means

1. truncate at second order in the interaction,
2. replace by inside the integral, and
3. extend the upper limit to using the decay of bath correlations.

Equation (4) is not a proof, but it is the correct scaling for essentially all rigorous versions.

Now insert (3) into (4). The only explicit prefactor coming from the coupling strength is . The bath memory integral contributes a factor by the change of variables :

(5)  

assuming (componentwise). Importantly, the function itself has no dependence.

Thus, at the level of scaling,

This immediately implies:

• If , the dissipator vanishes as (trivial unitary limit).
• If , the dissipator blows up so a finite Lindblad limit does not exist (although one can get Zeno behavior; see the Appendix).
• A finite nonzero limit at fixed physical time forces
  

  (6)  

So, in this framework, is not a choice: it is the balance condition that keeps the second-order noise strength when the bath correlation time is being squeezed to as .

At this point there is only one remaining scaling question with real dynamical content: does the system have time to rotate during the bath memory window? That is controlled by comparing and .

The three regimes: b > a, b = a, and b < a

The bath memory window is . Over such a window, the system interaction-picture operators change by a phase scale

So the dimensionless parameter controlling whether the system rotates appreciably during the memory time is .

Regime 1: b > a : fast bath correlations

Then , so for in the memory window ,

to leading order in . In words: on the timescale at which bath correlations decay, the system barely evolves. Once has fixed the overall strength, the remaining structure of the limit is white noise acting through the undressed operators . (No Bohr-frequency decomposition is enforced by fast system oscillations, because the system is not fast here.)

This is what is usually called the singular-coupling or white-noise limit: the bath becomes -correlated in time, and the system looks quasi-static on that correlation time.

Regime 2: b = a : system and bath correlations on same scale

Then , so the system does rotate by phases during the memory window. This is the regime where Bohr frequencies of survive into the coarse-grained generator. Concretely, contains Fourier components at the Bohr frequencies of . If one only does Born–Markov, these components can remain coupled (off-diagonal in Bohr frequency), and complete positivity is not guaranteed. The standard additional step is precisely to average out the cross-terms oscillating at different Bohr frequencies: this is the secular (rotating-wave) approximation.

This is what is usually called the weak-coupling (Davies/van Hove) limit: it is still second order, still Markovian, but now frequency resolution matters, so the secular projection is the natural way to obtain a time-homogeneous Lindblad generator.

Regime 3: b < a : fast system

Then , so the system completes many free rotations during a single bath memory window . In the Bohr-frequency decomposition , the second-order kernel contains phases ; for every fixed these oscillate arbitrarily fast as and average to zero (assuming the bath correlations are integrable in time). Thus only the components survive, i.e. only the part of the coupling that commutes with . When this yields a restricted Lindblad limit that (at least) dephases between distinct energy eigenspaces (and may also act within degenerate energy blocks); when for all the dissipator is trivial. Unlike the case, there’s no need for an extra hypothesis about Bohr-frequency separation; the limit itself cleanly separates exactly-zero from nonzero Bohr frequencies.

Summary of the logic

Within (1), the demand for a finite nontrivial second-order Markovian limit at fixed physical time forces . Once this balance is imposed, the remaining structure is controlled by whether the system rotates appreciably during a bath memory window , i.e. by . There are three nondegenerate regimes: (system quasi-static on the memory time, giving the singular/white-noise limit), (system rotates by phases and one needs a secular average, giving the Davies limit), and (system rotates many times and the limit automatically projects onto , giving at most energy-basis dephasing). In this sense the usual “weak coupling” and “singular coupling” derivations, together with the fast-system-dephasing regime, are the three canonical limits inside the single scaling framework (1).

Traditional presentation and regime-specific issues

The same three regimes are often presented using three different “small-parameter stories”: a long-time (van Hove) limit for weak coupling, a fast-bath/strong-coupling scaling for white noise, and (less often) a fast-system scaling in which secularization becomes exact. The first paragraph of each subsection below states the usual presentation and shows how it maps onto the exponent relations together with , , or . The remaining paragraphs collect issues that are specific to that regime: what must be assumed about bath correlations for the limit to be well behaved, what “extra step” is typically invoked (e.g. secular averaging). For the last regime, I also briefly discuss the historically clarifying case of environmental monitoring in chiral molecules.

Traditional singular-coupling (fast-bath/white-noise): b > a

A common parameterization uses and

Setting gives

which is (1) with , , and hence . Choosing yields , which lies in Regime 1 (), i.e. the bath correlation time collapses faster than any system rotation time.

Regime-specific issues

Here is scaled by , so the bath correlation time shrinks as . The balance is precisely the requirement that the integrated noise strength remain as . For the limit to be well behaved one needs the rescaled correlations to converge to a delta distribution:

A simple sufficient condition is together with finiteness and continuity of the noise spectrum at , in which case . If vanishes or diverges at the limiting rates vanish or blow up, respectively, so one does not obtain a finite nontrivial Lindbladian without modifying additional bath parameters (e.g. temperature, cutoffs, density) as part of the scaling. In practice this regime is often best viewed as an idealization: a controlled way to approximate a bath by white noise on the system time scale.

Traditional weak-coupling (Davies/van Hove): b = a

Start from the physical Hamiltonian

and study the reduced dynamics on long times (van Hove scaling). To cast this long-time limit into the fixed- scaling form (1), keep fixed and globally rescale the Hamiltonian by :

This is (1) with , which satisfies and , i.e. it sits in Regime 2, where secular averaging is the standard extra step.

Regime-specific issues

In this regime the bath is not being made fast: its correlation (memory) time stays in physical time units. The long-time scaling (or, equivalently, the global rescaling ) is what separates the system relaxation time from the bath memory time . (The system relaxation time is not the characteristic unitary timescale of the system Hamiltonian, but rather the dissipative timescale on which the reduced state changes by due to the interaction. At second order one has rates of order : schematically , with coefficients in given by bath correlation integrals ; if these integrals are then .)

A minimal sufficient assumption for the Markov step is that the bath correlations are integrable (or at least decay fast enough) so that ; then the memory kernel has a finite limit and, combined with , one can replace by over the memory window. However, this is not itself sufficient to get the Lindblad form.

In order to ensure complete positivity of the dynamics, one commonly further invokes a secular approximation, which can be phrased as the existence of a coarse-graining window satisfying for distinct Bohr frequencies of . Exact degeneracies simply mean keeping the corresponding blocks (terms with ); no further gap assumption is needed unless Bohr frequencies form clusters on scales comparable to the relaxation rate.

Traditional fast-system limit (automatic super-secular projection): b < a

A representative way to parameterize the regime is to scale only the system Hamiltonian,

and take at fixed physical time . This is (1) with , so and . In the interaction picture with respect to , one has

where ranges over Bohr frequencies of and are the corresponding eigenoperators of . In the second-order Born–Markov kernel, each channel appears under an integral of the form

with . Assuming , the Riemann–Lebesgue lemma implies this integral vanishes as for every fixed . Hence only the components survive, i.e., , and the limiting Lindblad generator (if nontrivial) has Lindblad operators commuting with . Equivalently: the usual secular projection becomes exact because all nonzero Bohr frequencies are sent to infinity by the scaling .

Regime-specific issues

The hallmark of this regime is that secularization is automatic: scaling sends every nonzero Bohr frequency to infinity, so only the exactly block survives in the limit. For a finite-dimensional system this cleanly separates commuting from noncommuting parts of without any additional assumption about the spacing between distinct Bohr frequencies (beyond exact degeneracies).

Decoherence of chirality

The fast-system regime famously arose in the classical decoherence literature answering this question: Due to parity symmetry, energy eigenstates of molecules are superposition of chiral eigenstates (say, ), so why do we treat some molecules, like ammonia, as energy eigenstates, while other molecules, like a sugar, as either chiral eigenstate but not superpositions thereof?F. Hund, “On the Interpretation of Molecular Spectra. I (Zur Deutung der Molekelspektren. I),” Z. Phys. 40, 742 (1927).J. D. Macomber, “Loss of phase coherence in the inversion of ammonia as a model for the racemization of chiral molecules,” J. Chem. Phys. 82, 4551 (1985).J. Trost and K. Hornberger, “Hund’s Paradox and the Collisional Stabilization of Chiral Molecules,” Phys. Rev. Lett. 103, 023202 (2009). [arXiv:0811.2140]P. J. Coles, V. Gheorghiu, and R. B. Griffiths, “Consistent histories for tunneling molecules subject to collisional decoherence,” Phys. Rev. A 86, 042111 (2012). [arXiv:1205.6188]^a A standard toy model of this is a tunneling double well with chiral (enantiomer) states and tunnel splitting between the symmetric/antisymmetric energy eigenstates . If the environment couples primarily to the localizing coordinate (so ), then is off-diagonal in the energy basis and . In the strict fast-system limit the leading dissipator therefore vanishes, and one expects coherent inversion oscillations with only small subleading decoherence. This is the regime relevant to ammonia inversion at sufficiently low pressure, where the inversion splitting is large compared to collisional decoherence rates. For larger chiral molecules, like sugar, the tunnel splitting is typically exponentially small while environmental scattering remains comparatively fast, pushing the effective dynamics toward the opposite ordering (closer to the “measurement” regime), selecting localized chiral pointer states and stabilizing chirality.

Appendix: Zeno limit for diverging dissipation

In the scaling framework , the regime means the second-order dissipative contribution grows like with . After the usual Born–Markov reduction, this typically produces a reduced generator of the form

where is a Lindblad dissipator (completely positive, trace preserving in semigroup form) and collects the remaining terms (Hamiltonian part and any weaker dissipation). As at fixed physical time , the relaxation time associated with goes to zero like , so the dynamics is no longer a small perturbation of unitary motion; instead it is a strong-dissipation limit.

If has a nontrivial steady manifold

then rapidly projects states onto (on times ), and the limit can be a well-defined “quantum Zeno” evolution confined to . Let be the projection (in the Banach-space sense appropriate to the setup) onto , and . If is primitive (unique full-rank steady state), then is one-dimensional and the limit at any fixed is typically trivial: instantaneous convergence to that steady state. If has dimension (decoherence-free subspace/algebra, noiseless subsystems, or conserved quantities), then the limit retains nontrivial dynamics within .

In the nontrivial case, one can often derive an effective Lindblad generator on by adiabatic elimination. Formally, assuming is invertible on (or has a bounded pseudo-inverse there), the leading effective generator is

which includes both the projected “slow” dynamics and a second-order correction capturing virtual excursions out of that are immediately damped back. Physically this is the continuous-measurement/Zeno picture: the dominant dissipation enforces a constraint (the Zeno manifold), and the remaining terms generate an induced (often still Lindblad) evolution within that constrained subspace.