Stochastic Floquet Engineering

• Stochastic Floquet Engineering is a framework that studies periodically driven systems where intrinsic noise, dissipation, or bath coupling shape the effective dynamics.
• It employs techniques such as Floquet–Magnus expansion and master-equation linearization to derive effective static drifts and renormalized diffusion from nonlinear stochastic systems.
• The approach enables stabilization of non-equilibrium steady states and synthesis of non-Hermitian Hamiltonians, revealing prethermal behavior and convergence in many-body settings.

Stochastic Floquet engineering denotes the control and analysis of periodically driven systems in settings where stochasticity, dissipation, or bath coupling is intrinsic to the effective dynamics rather than a negligible perturbation. In the literature considered here, the term spans several closely related constructions: Floquet–Magnus reduction of nonlinear classical stochastic equations through their master equations, dissipative stabilization of Floquet-engineered many-body phases by thermal baths, synthesis of non-Hermitian target Hamiltonians by noisy periodic drives, and Floquet-modified Langevin descriptions of nonadiabatic motion near metal surfaces (Higashikawa et al., 2018, Wanckel et al., 1 Apr 2026, Guo et al., 14 Jun 2026, Wang et al., 2023).

1. Linearization through generators, densities, and master equations

A central obstacle in stochastic Floquet engineering is that the ordinary Floquet theorem applies to linear ordinary differential equations, whereas many driven stochastic systems are governed by nonlinear stochastic differential equations or by open-system quantum dynamics. Higashikawa et al. address this difficulty by moving from the stochastic trajectory equation to the master-equation level, where linearity is restored in the probability density $P$ (Higashikawa et al., 2018).

For a general classical stochastic system under periodic driving, the Stratonovich equation is written as

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$

with Gaussian white noise

$$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$

The associated Fokker–Planck equation is

$$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$

with

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$

Because $L_t$ is linear in $P$ and $T$-periodic, one may apply the Floquet theorem to

$$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$

so that

$$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$

A common misconception is that the same procedure can be applied directly to the stochastic equation of motion. The explicit statement in the classical formulation is that this fails because the stochastic differential equation is nonlinear in $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$0 and contains noise, so the usual Floquet theorem does not apply (Higashikawa et al., 2018).

An analogous lifting to a linear equation appears in the non-Hermitian construction with noisy periodic driving. There the averaged density matrix obeys

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$1

which is already in Lindblad form after averaging over Gaussian white noise $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$2 (Guo et al., 14 Jun 2026). In both classical and quantum settings, stochastic Floquet engineering proceeds by replacing a nonlinear or noise-driven trajectory description with a linear periodic generator acting on a density.

2. Floquet–Magnus structure and effective stochastic equations

Once a periodic linear generator is available, the high-frequency expansion becomes systematic. Writing the Fourier decomposition $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$3, Higashikawa et al. expand

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$4

with $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$5 (Higashikawa et al., 2018). The first terms are

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$6

and

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$7

Truncating the series to order $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$8 yields an effective Fokker–Planck equation

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$9

If the resulting operator contains at most second-order derivatives in $$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$0, one may read off an effective Langevin equation. At first order, assuming time-independent diffusion so that only the drift renormalizes, and writing

$$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$1

the effective drift is

$$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$2

The effective Langevin equation then becomes

$$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$3

At second order, both drift and diffusion are renormalized. The kick operator $$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$4 produces small rapid displacements, identified as micromotion in the classical formulation. This generator-based reduction is the basic mechanism by which a periodically driven stochastic system acquires an effective static drift, diffusion matrix, or potential landscape (Higashikawa et al., 2018).

3. Convergence, prethermal behavior, and dissipative stabilization

The high-frequency expansion is not purely formal in the regimes examined. For few-body nonchaotic systems, if the instantaneous Fokker–Planck generator $$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$5 is norm-bounded, then for $$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$6 the Floquet–Magnus series converges, and the truncated series approximates the dynamics for all times, including relaxation to a steady state (Higashikawa et al., 2018).

For many-body systems, the same work states that the $$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$7th Floquet–Magnus term scales as

$$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$8

for a $$\langle h_i(t)\,h_j(t')\rangle=2D\,\delta_{ij}\,\delta(t-t').$$9-local Fokker–Planck operator of scale $$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$0. Asymptotically, the series is well behaved up to

$$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$1

and the effective generator is nearly conserved over exponentially long times

$$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$2

In isolated systems this produces a Floquet-prethermal plateau, while in open systems the corresponding description is a non-equilibrium steady state accurately given by the truncated $$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$3 (Higashikawa et al., 2018).

Dissipative Floquet engineering with thermal baths provides a complementary stabilization mechanism in interacting quantum systems. The stated objective is to suppress Floquet heating and guide the system into a non-equilibrium steady state with a large occupation of the effective ground-state, but generally non-thermal occupations of excited states of the effective Hamiltonian (Wanckel et al., 1 Apr 2026). The key hierarchy of scales is

$$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$4

together with an intermediate system-bath coupling satisfying

$$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$5

Under these conditions, the bath is chosen so that it both suppresses Floquet heating and preserves cooling transitions into the gapped effective ground state.

Another misconception is that dissipation necessarily destroys Floquet engineering. The bath-based protocol explicitly uses dissipation as a design parameter: it is required to be strong enough to diabatically suppress Floquet-induced avoided crossings, but weak enough not to smear out the many-body gap (Wanckel et al., 1 Apr 2026).

4. Classical and micromagnetic realizations

The classical benchmark in Higashikawa et al. is the Kapitza pendulum with friction. For a rigid pendulum of length $$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$6 with vertical fast drive $$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$7 and viscous damping $$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$8,

$$\partial_t P = -\frac{\partial}{\partial\phi_i}\bigl[\mathcal F_i(\bm\phi,t)\,P\bigr] +\frac{\partial^2}{\partial\phi_i\partial\phi_j}\bigl[\mathcal D_{ij}(\bm\phi,t)\,P\bigr] \equiv L_t\,P,$$9

With $$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$0, the Fourier components of the drift are

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$1

The second-order effective drift is

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$2

equivalent to the effective potential

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$3

The second term stabilizes the inverted point $$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$4 once

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$5

and numerics show that the Floquet–Magnus prediction for attraction basins and relaxation dynamics agrees excellently with direct integration (Higashikawa et al., 2018).

The same formalism applies to laser-driven magnets described by the stochastic Landau–Lifshitz–Gilbert equation,

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$6

with

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$7

For a circularly polarized field

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$8

the first Floquet–Magnus correction is

$$\mathcal F_i= -f_i -D\,g_{kl}\,\partial_k g_{il},\qquad \mathcal D_{ij}=D\,g_{ik}g_{jk}.$$9

Since

$L_t$0

the drive induces the synthetic static field

$L_t$1

The effective stochastic LLG reproduces a net magnetization $L_t$2 for $L_t$3, in quantitative agreement with full stochastic LLG numerics over a broad range of $L_t$4, $L_t$5, and $L_t$6 (Higashikawa et al., 2018).

A multiferroic extension adds the inverse Dzyaloshinskii–Moriya coupling

$L_t$7

so that a circularly polarized electric field generates an effective Dzyaloshinskii–Moriya interaction

$L_t$8

In one dimension, with a small counter-field to cancel $L_t$9, the non-equilibrium steady state is a chiral spiral with vector chirality per site

$P$0

again in excellent agreement with direct stochastic LLG simulation up to $P$1 and moderate $P$2 (Higashikawa et al., 2018).

5. Dissipative many-body phases and non-Hermitian target synthesis

In the bath-stabilized quantum setting, the driven many-body Hamiltonian is

$P$3

and for the driven Bose–Hubbard chain at unit filling,

$P$4

The time-averaged part is

$P$5

The effective Hamiltonian satisfies

$P$6

and coupling to a thermal bath produces a Floquet–Born–Markov master equation in Lindblad or Redfield form for the reduced density matrix (Wanckel et al., 1 Apr 2026).

The bath is modeled by

$P$7

with

$P$8

For the Bose–Hubbard example, $P$9. In the Mott regime, $T$0 is a Mott insulator for

$T$1

with Mott gap

$T$2

to leading order. The reported numerical parameters are

$T$3

and chains up to $T$4 show

$T$5

deep in the Mott regime at moderate $T$6, with clear suppression of heating-induced dips when $T$7 is ramped into the optimal window (Wanckel et al., 1 Apr 2026).

A distinct quantum branch of stochastic Floquet engineering uses classical white-noise modulation of a periodic Hermitian drive to synthesize an arbitrary non-Hermitian target Hamiltonian. The starting point is

$T$8

with $T$9 and $$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$0 Hermitian and $$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$1-periodic, and

$$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$2

After noise averaging and stroboscopic Floquet averaging, the effective Hamiltonian is

$$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$3

To match a desired target

$$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$4

the construction chooses $$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$5, imposes

$$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$6

and sets

$$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$7

with $$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$8. One then obtains

$$\partial_t P=L_tP,\qquad U(t_2,t_1)=\mathcal T\exp\!\Bigl[\int_{t_1}^{t_2}L_t\,dt\Bigr],$$9

Up to an overall decay factor, the stroboscopic map is the non-unitary map generated by $$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$0. With post-selection of the no-jump outcome, the deterministic non-unitary evolution is

$$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$1

Applications explicitly discussed include a cavity Hamiltonian with dissipative coupling between Fock states, arbitrary pure-state preparation from essentially any $$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$2 with nonzero target overlap, and the realization of non-unitary quantum gates without ancillae or state-dependent updating (Guo et al., 14 Jun 2026).

These two quantum lines differ in objective but share a common structural feature: stochasticity and dissipation are engineered to shape the effective stroboscopic generator rather than treated solely as errors.

6. Floquet electronic friction, fluctuation–dissipation violation, and regime boundaries

Wang and Dou develop a Floquet classical master equation for nonadiabatic dynamics near metal surfaces under Floquet engineering and then reduce it, in a fast-driving and fast-electron limit, to a Floquet Fokker–Planck equation and equivalent Langevin dynamics (Wang et al., 2023). The model is an Anderson–Holstein impurity with periodically driven on-site energy,

$$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$3

where

$$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$4

The Floquet-modified hopping rates are

$$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$5

with

$$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$6

In the limit $$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$7, cycle averaging gives

$$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$8

With fast electronic relaxation $$U(t_2,t_1)=e^{G_F(t_2)}e^{(t_2-t_1)L_F}e^{-G_F(t_1)}.$$9, elimination of the nonadiabatic correction yields the Floquet Fokker–Planck equation

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$00

where

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$01

and the cycle-averaged potential of mean force is

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$02

The Floquet-modified friction and noise coefficients are

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$03

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$04

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$05

The equivalent Langevin equations are

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$06

with Gaussian noise

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$07

The significant nonequilibrium feature is that, in general,

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$08

so the second fluctuation–dissipation theorem is violated. The effective temperature is therefore

$$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$09

For strong Floquet driving $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$10, the stated result is $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$11 over much of configuration space, hence $$\dot\phi_i(t)=f_i\bigl(\bm\phi,t\bigr)+\sum_j g_{ij}\bigl(\bm\phi,t\bigr)\,h_j(t),$$12, which is identified as Floquet-induced heating (Wang et al., 2023).

The benchmark structure is regime-dependent. In the slow-driving or weak-coupling regime, Floquet surface hopping agrees with the exact Floquet quantum master equation, whereas the friction/Langevin approach begins to break down. In the fast-driving, strong-coupling regime, Floquet electronic friction with cycle-averaged potential of mean force reproduces both the steady-state electronic population and nuclear kinetic energy of the Floquet quantum master equation and Floquet surface hopping. If one retains the full time-dependent potential of mean force, residual oscillations in the electronic population can also be captured (Wang et al., 2023).

Taken together, these results show that stochastic Floquet engineering is not restricted to a single formalism. In the examples surveyed here it includes master-equation Floquet–Magnus expansions for classical noisy dynamics, bath-assisted stabilization of gapped effective ground states, noisy-drive synthesis of non-Hermitian stroboscopic evolution, and Floquet-modified frictional Langevin theories. The shared technical core is the replacement of explicitly driven stochastic dynamics by an effective generator whose validity is controlled by scale separation, convergence or asymptotic estimates, and the structure of the noise or bath.