Caldeira-Leggett Master Equation

• Caldeira–Leggett master equation is a high-temperature, Markovian model for quantum Brownian motion that couples a system to an Ohmic bath of harmonic oscillators.
• It integrates unitary dynamics with friction and diffusion, and requires Lindblad-completion corrections to ensure complete positivity of the quantum evolution.
• Recent generalizations and non-Markovian extensions broaden its scope, enabling improved numerical methods and deeper insights into decoherence, thermalization, and dissipation.

Searching arXiv for the cited paper and closely related Caldeira–Leggett master-equation work. The Caldeira–Leggett master equation is the standard high-temperature, Markovian equation for quantum Brownian motion of a particle or harmonic oscillator linearly coupled to a bath of harmonic oscillators. In its commonly used form it combines unitary dynamics, friction, and diffusion; in its original form it is not in Lindblad form and therefore does not, in general, generate a completely positive trace-preserving semigroup. The modern literature treats it simultaneously as an effective equation for decoherence and dissipation, as the Markovian limit of more general reduced dynamics, and as a starting point for non-Markovian, nonlinear, multidimensional, and numerically exact generalizations (Bernád et al., 2017).

1. Microscopic model and regime of validity

The underlying Caldeira–Leggett construction couples a system coordinate to an environment made of many harmonic oscillators. A standard form is

$$H = H_0 + H_\mathcal{E} + H_\mathrm{int},\qquad H_0=\frac{p^2}{2M}+V(x),\qquad H_\mathcal{E}=\sum_{k=1}^{N_\mathcal{E}}\left(\frac{p_k^2}{2m}+\frac{1}{2}m\omega_k^2 q_k^2\right),\qquad H_\mathrm{int}=-x\sum_{k=1}^{N_\mathcal{E}} c_k q_k,$$

with factorized initial data $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$ and a thermal bath $$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$ (Nishimura et al., 26 Mar 2025). In field-theoretical language the same structure is written as $H=H_S+H_B+H_{SB}$, with $H_{SB}=\sum_\alpha S^\alpha B^\alpha$, a Gaussian oscillator bath, and bath influence fully encoded by two-point correlators or, equivalently, by a spectral density $g(\Omega)$ (Fogedby, 2022).

The Markovian Caldeira–Leggett regime is associated with an Ohmic bath and a large separation of scales. In one formulation the assumptions are

$$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$

which yield the familiar high-temperature coefficients

$$\gamma=\frac{\eta}{2m},\qquad D_{pp}=\eta k_BT,\qquad D_{px}=\frac{\gamma k_BT}{\Omega},\qquad D_{pp}=2m\gamma k_BT$$

for the harmonic-oscillator problem (Bernád et al., 2017). In the influence-functional treatment the Ohmic spectrum with cutoff $\omega_\mathrm{cut}$, together with the high-temperature approximation $\coth(\beta\omega/2)\approx 2/(\beta\omega)$ and a short-memory approximation, yields the local reduced dynamics usually identified as the Caldeira–Leggett master equation (Nishimura et al., 26 Mar 2025).

At the Hamiltonian level, the model also admits a continuum-spectrum interpretation in which the apparent damping of the discrete oscillator arises by phase mixing rather than by non-Hamiltonian friction. In that formulation the decay mechanism is mathematically analogous to Landau damping in the linearized Vlasov–Poisson system, and the oscillator-plus-bath dynamics can be transformed to a continuum of uncoupled harmonic oscillators (Hagstrom et al., 2010). This suggests a useful distinction between microscopic reversible dynamics and effective reduced dissipation.

2. Standard operator and coordinate-space forms

For a harmonic oscillator of mass $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$0 and frequency $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$1 with $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$2, the extended Markovian master equation studied in the modern Lindblad-completion literature is

$$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$3

The original Caldeira–Leggett master equation is the same equation without the last term $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$4 (Bernád et al., 2017).

In coordinate representation, for a one-dimensional particle of mass $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$5, the high-temperature Markovian equation is written as

$$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$6

The kinetic term gives the unitary contribution, the term proportional to $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$7 gives dissipation or relaxation, and the term proportional to $$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$8 suppresses off-diagonal density-matrix elements in the position basis (Mousavi et al., 2022).

In the free-particle and high-temperature weak-coupling setting, the same coordinate-space structure appears as

$$\hat\rho(0)=\hat\rho_\mathcal{S}(0)\otimes \hat\rho_\mathcal{E}(0)$$9

again with $$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$0, or $$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$1 in the dimensionless convention $$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$2 (Mousavi, 17 Sep 2025).

The high-temperature influence-functional derivation also makes explicit the renormalized Hamiltonian,

$$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$3

so the master equation carries both dissipative terms and bath-induced frequency renormalization (Nishimura et al., 26 Mar 2025).

3. Complete positivity, Lindblad completion, and the $$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$4 problem

A central structural issue is that the original Caldeira–Leggett generator is Markovian but not of Lindblad form. In the harmonic-oscillator notation above, complete positivity of the extended equation requires the Dekker inequality

$$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$5

Using the high-temperature coefficients, the minimal allowed position-diffusion coefficient is

$$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$6

The term $$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$7 is therefore the standard correction used to make the generator Lindblad-compatible (Bernád et al., 2017).

The same issue appears in the familiar “fixed” completely positive version

$$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$8

which is described as a standard Lindblad-completing correction, and is associated in the literature block with Diósi (Ferialdi, 2017). The position-diffusion term has no classical analog in ordinary Brownian motion, but in the quantum setting it arises naturally from complete positivity and the Lindblad structure; the same literature also notes a physical interpretation in terms of continuous momentum measurements in a collisional model (Bernád et al., 2017).

Positivity failure is not merely formal. An exact analysis of the non-Lindblad Caldeira–Leggett equation shows that positivity violations can occur for certain pure initial states at intermediate times, even within the usual approximation regime. The main diagnostics used are the purity $$\hat\rho_\mathcal{E}(0)\propto e^{-\beta H_\mathcal{E}}$$9, for which $H=H_S+H_B+H_{SB}$0 is a sufficient witness of non-positivity, and the Robertson–Schrödinger uncertainty relation. The same work reports that Cases I–III in its classification have physical steady states in the parameter regimes assumed in their derivations, whereas the truncated Dekker-type case is “more drastic” and “cannot be truncated at will” (Homa et al., 2018).

The entropy-production approach to $H=H_S+H_B+H_{SB}$1 addresses a narrower ambiguity. For Gaussian states evolving under the extended equation, one computes the relative entropy to the stationary state and defines entropy production as the negative initial decay rate,

$H=H_S+H_B+H_{SB}$2

The proposed selection rule is to view a global minimum or maximum of $H=H_S+H_B+H_{SB}$3, subject to the Dekker bound, as the most plausible value of $H=H_S+H_B+H_{SB}$4. The strongest conclusion reported is that for initial states near the steady state the extremal criterion most naturally selects the Dekker lower bound, whereas for more remote initial states the result may be interior or initial-state dependent (Bernád et al., 2017).

4. Gaussian solutions, stationary states, and thermalization

For the harmonic oscillator, the Fourier-transformed density matrix

$H=H_S+H_B+H_{SB}$5

admits the Gaussian ansatz

$H=H_S+H_B+H_{SB}$6

with coefficients satisfying a closed linear system. The stationary solution has $H=H_S+H_B+H_{SB}$7, while $H=H_S+H_B+H_{SB}$8 depend on $H=H_S+H_B+H_{SB}$9. The corresponding Gaussian state is identified as a displaced squeezed thermal state, and its spectrum matches that of a thermal oscillator with effective mean occupation number

$H_{SB}=\sum_\alpha S^\alpha B^\alpha$0

This representation is the basis for both stationary-state analysis and entropy-production calculations (Bernád et al., 2017).

Exact solution methods also exist for the non-Lindblad equation without the $H_{SB}=\sum_\alpha S^\alpha B^\alpha$1 term. In center-of-mass and relative coordinates,

$H_{SB}=\sum_\alpha S^\alpha B^\alpha$2

Fourier transformation in $H_{SB}=\sum_\alpha S^\alpha B^\alpha$3 turns the equation into a first-order PDE that can be solved by the method of characteristics, reducing the problem to coupled ODEs for $H_{SB}=\sum_\alpha S^\alpha B^\alpha$4, $H_{SB}=\sum_\alpha S^\alpha B^\alpha$5, and $H_{SB}=\sum_\alpha S^\alpha B^\alpha$6 (Homa et al., 2018). In the exact stochastic derivation of dissipative harmonic-oscillator dynamics, the resulting time-dependent master equation

$H_{SB}=\sum_\alpha S^\alpha B^\alpha$7

has coefficients determined by integral equations and is proved equivalent to the Hu–Paz–Zhang equation (Li et al., 2011).

Thermalization is visible at the level of reduced density matrices and correlation functions. For a harmonic system coupled to a thermal bath, the finite-time reduced density matrix remains Gaussian, and its coefficients stabilize at late times; in particular $H_{SB}=\sum_\alpha S^\alpha B^\alpha$8 in one explicit parametrization, while the asymptotic coefficients agree with those of a thermal oscillator (Ayyar et al., 2012). The same work shows

$H_{SB}=\sum_\alpha S^\alpha B^\alpha$9

and the unequal-time correlator approaches the thermal form

$g(\Omega)$0

including the late-time emergence of imaginary-time periodicity of KMS type (Ayyar et al., 2012). Nonperturbative real-time path-integral calculations likewise confirm decoherence and thermalization from first principles, while finding quantitative deviations from the naive master-equation coefficient outside the clean high-temperature, weak-coupling, large-cutoff regime (Nishimura et al., 26 Mar 2025).

5. Non-Markovian extensions and the debate over dissipation

The exact reduced dynamics of the oscillator bath can be written as a non-Markovian master equation before any Markov or rotating-wave limit is imposed. In the transmission-matrix formulation,

$g(\Omega)$1

and the reduced density matrix obeys

$g(\Omega)$2

Within that framework, the Born approximation gives the kernel $g(\Omega)$3 in terms of system propagators and bath correlators; a quasiparticle or pole approximation yields a local Markovian equation; and the rotating-wave approximation produces the GKSL generator (Fogedby, 2022).

A complementary exact route uses stochastic decoupling of the bath. For the dissipative harmonic oscillator, the bath-induced stochastic field is expressed through the bath correlation function

$g(\Omega)$4

and stochastic averaging yields a closed master equation with time-dependent coefficients. This equation is proved equivalent to the Hu–Paz–Zhang master equation, thereby identifying the exact non-Markovian benchmark underlying the Markovian Caldeira–Leggett limit (Li et al., 2011).

Several papers in the literature block retain finite bath memory explicitly. One approach expands the high-temperature Ohmic dynamics in inverse powers of the cutoff $g(\Omega)$5, obtaining a local but non-Markovian correction

$g(\Omega)$6

where $g(\Omega)$7 mixes $g(\Omega)$8 with $g(\Omega)$9. In the free-particle case this leads to a resurgence of coherence at intermediate times, termed “lateral coherences,” accompanied by a transient negative entropy production rate (Lally et al., 2019). Another route, based on dynamical quantization of a non-Markovian Fokker–Planck equation, produces a non-Markovian Caldeira–Leggett equation with memory kernel $$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$0, reducing to the original high-temperature form when $$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$1 and $$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$2 [(Bolivar, 2010); (Bolivar, 2010)].

The status of dissipation in the strict Markovian limit is contested. One line of analysis argues that if the bath correlation is taken to be truly delta-correlated, then for position coupling $$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$3 the exact Markovian limit is not the standard dissipative Caldeira–Leggett equation but the purely decohering Joos–Zeh form

$$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$4

with the friction term absent. On that reading, the usual friction term survives only when the bath retains finite memory, so dissipation is “a genuinely non-Markovian feature” of the microscopic model (Ferialdi, 2017). Other parts of the literature, including the standard influence-functional, stochastic, and kernel-based derivations, continue to use the familiar dissipative Caldeira–Leggett equation as the effective high-temperature Markovian description [(Nishimura et al., 26 Mar 2025); (Li et al., 2011); (Fogedby, 2022)]. A plausible implication is that the equation occupies an intermediate conceptual status: microscopically motivated, but not identical to every possible notion of a strict Markov limit.

6. Applications, variants, and contemporary generalizations

The equation’s most direct application is decoherence. In the high-temperature master equation, the off-diagonal damping factor

$$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$5

implies the decoherence time scale

$$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$6

so larger separation, stronger coupling, and higher temperature produce faster coherence loss (Nishimura et al., 26 Mar 2025). For Schrödinger-cat superpositions in the zero-dissipation approximation, the attenuation is summarized by $$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$7 with

$$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$8

and the same study reports that a constant force field shifts trajectories and fringes but does not change the decoherence function itself, while positive values of a Gaussian stretching parameter reduce the rate of decoherence (Mousavi et al., 2022).

Bipartite extensions lead to the “double CL” formalism. For two particles in a common bath, the master equation acquires cross terms absent for distinct baths, and the resulting Bohmian nonlocality measure starts at zero, grows to a maximum, and then decays in both cases; however, common baths generate revivals and oscillations, whereas local baths give smooth monotonic decay. The work explicitly emphasizes that these oscillations do not imply non-Markovianity of the master equation itself, but rather reflect bath structure within a formally Markovian regime (Mousavi, 17 Sep 2025).

Recent generalizations modify either the environment or the coupling. When the bath oscillators themselves are damped through a Caldirola–Kanai prescription, the reduced master equation acquires a new term

$$\Omega,\,k_BT \gg \gamma,\qquad \Omega,\,k_BT \gg \omega,$$9

equivalently an induced inverted-harmonic contribution to the effective potential,

$$\gamma=\frac{\eta}{2m},\qquad D_{pp}=\eta k_BT,\qquad D_{px}=\frac{\gamma k_BT}{\Omega},\qquad D_{pp}=2m\gamma k_BT$$0

In the double-well example this changes both short-time decoherence and long-time well-transfer probabilities, and with an appropriate rescaling condition stronger environmental damping slows decoherence (Buxton et al., 2023). In driven nonlinear oscillators, retaining the full nonlinear and time-dependent system motion inside the dissipator yields a generalized Caldeira–Leggett equation with dynamically dressed dissipation, new nonlinear and drive-dependent channels, nonlinear damping, suppression of bistability, and asymmetric resonance responses (Wagner et al., 8 May 2026). The phase-coupled Caldeira–Leggett model replaces linear bath coupling by $$\gamma=\frac{\eta}{2m},\qquad D_{pp}=\eta k_BT,\qquad D_{px}=\frac{\gamma k_BT}{\Omega},\qquad D_{pp}=2m\gamma k_BT$$1, producing an exact non-Markovian hierarchical equation with noncanonical steady states and strong Hamiltonian renormalization; in the weak-$$\gamma=\frac{\eta}{2m},\qquad D_{pp}=\eta k_BT,\qquad D_{px}=\frac{\gamma k_BT}{\Omega},\qquad D_{pp}=2m\gamma k_BT$$2 limit it reduces to CL-like dissipation (Chang et al., 29 Oct 2025).

The model has also become a benchmark for numerical methods that avoid or postpone the master-equation approximation. Real-time path integrals evaluated on a Lefschetz thimble reproduce the qualitative dependence of decoherence on $$\gamma=\frac{\eta}{2m},\qquad D_{pp}=\eta k_BT,\qquad D_{px}=\frac{\gamma k_BT}{\Omega},\qquad D_{pp}=2m\gamma k_BT$$3 and $$\gamma=\frac{\eta}{2m},\qquad D_{pp}=\eta k_BT,\qquad D_{px}=\frac{\gamma k_BT}{\Omega},\qquad D_{pp}=2m\gamma k_BT$$4 in the regime where the master equation is expected to hold, while exposing finite-cutoff and finite-temperature deviations outside that regime (Nishimura et al., 26 Mar 2025). A finite-dimensional “adapted Caldeira–Leggett” construction evolves the full system and environment unitarily in a truncated Hilbert space, thereby avoiding Born and Markov assumptions while reproducing key decoherence and einselection phenomena (Albrecht et al., 2021). For the multidimensional model, a Dyson-series reformulation combined with low-rank approximation and frozen Gaussian approximation reduces the spatial dimensionality by half and collapses the relevant time integrals to one- and two-dimensional integrals independent of the Dyson truncation level; the method is validated on several examples, including a two-dimensional double-slit simulation (Zhan et al., 29 Sep 2025).

Taken together, these developments preserve the core role of the Caldeira–Leggett master equation as the canonical high-temperature quantum Brownian-motion model, while making clear that its precise interpretation depends on the intended level of description. As a reduced equation it remains a standard tool for dissipation, decoherence, and thermalization; as a generator it raises enduring questions about positivity and Lindblad form; and as a limiting case it now sits within a broader hierarchy of exact, non-Markovian, nonlinear, and computationally explicit theories (Bernád et al., 2017).