Dissipative Schrödinger Equation

Updated 4 July 2026

Dissipative Schrödinger equations are Schrödinger-type models that incorporate non-Hamiltonian damping terms to capture irreversible effects such as relaxation, viscosity, and decoherence.
They encompass a variety of formulations including strongly dissipative linear models, complex nonlinear equations, stochastic open-system approaches, and hydrodynamic analogs.
These models offer practical insights into enhanced smoothing, stability, and decay properties, aiding in the control of quantum systems and the simulation of fluid dynamics.

Searching arXiv for relevant papers on dissipative Schrödinger equations and closely related formulations. A dissipative Schrödinger equation is a Schrödinger-type evolution law in which non-Hamiltonian terms model damping, relaxation, viscosity, decoherence, reservoir coupling, or other irreversible effects while preserving, modifying, or explicitly breaking structures familiar from the conservative Schrödinger equation. Across the literature, the term encompasses several mathematically distinct classes: strongly dissipative linear equations with derivative damping in bounded domains; nonlinear Schrödinger equations with dissipative complex nonlinearities; damped and driven nonlinear Schrödinger equations in optics and fluid models; stochastic Schrödinger equations for open quantum systems; and hydrodynamic reformulations in which Schrödinger-type equations encode viscous Navier–Stokes or Navier–Stokes–Korteweg dynamics. A unifying theme is that dissipation changes both the spectral and asymptotic structure of the flow: smoothing can replace mere regularity propagation, trapped or uncontrolled rays need not obstruct stabilization, coherent structures can become attractors or metastable slow manifolds, and the long-time dynamics is often governed by decay or relaxation rather than unitary scattering (Aloui et al., 2012).

1. Definitions and principal model classes

One important linear model is the regularized, strongly dissipative Schrödinger equation on a bounded obstacle domain $\Omega_0$ ,

$\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$

where $\Delta_D$ is the Dirichlet Laplacian and $a$ is supported near the outer boundary $\partial B$ of a bounded domain containing strictly convex obstacles (Aloui et al., 2012). The dissipative operator is

$B_a = a(x)(-\Delta_D)^{1/2}a(x),$

and the generator is

$A_a = -\Delta_D + i B_a.$

Here dissipation is of order $1$, nonlocal, and frequency dependent; this is the sense in which the model is “strongly dissipative” (Aloui et al., 2012).

A second major class consists of dissipative nonlinear Schrödinger equations in $\mathbb{R}^d$ with complex nonlinear coefficient,

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$

with $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$0, and in particular the attractive-dissipative case $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$1 (Kita et al., 15 May 2026). In this setting the $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$2-norm is nonincreasing: $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$3 Related dissipative nonlinearities arise in the equation

for a class of arbitrarily large oscillatory initial data (Cazenave et al., 2020).

A third class includes damped NLS with linear damping. On the one-dimensional torus, the damped cubic equation

provides a non-Hamiltonian perturbation of the periodic cubic NLS and supports a stability theory for cnoidal-wave manifolds with exponentially decaying mass (Antonelli et al., 2022). On star-shaped networks, another model is

with damping on only one branch and at infinity (Ammari et al., 2019).

Driven-damped variants are equally central. The AC-driven damped nonlinear Schrödinger equation

and, after passage to the rotating frame,

$\Delta_D$ 0

form the Lugiato–Lefever-type model used for localized dissipative structures in Kerr resonators (Jang et al., 2015).

Another hydrodynamic family is the Schrödinger–Navier–Stokes equation

$\Delta_D$ 1

which is formally equivalent to the Navier–Stokes–Korteweg equations for capillary fluids and decomposes into a Korteweg conservative part and a Rayleigh dissipative part (Salasnich et al., 13 Apr 2026). Closely related wave formulations for dissipative fluids use

$\Delta_D$ 2

to generate viscous Navier–Stokes dynamics after a generalized Madelung transformation (Salasnich et al., 2023).

Finally, open-system formulations often take stochastic or effective non-Hermitian forms. One example is the probability-conserving dissipative Schrödinger equation

$\Delta_D$ 3

where $\Delta_D$ 4 is built from microscopic rate equations for basis-state probabilities and total probability is conserved although energy is not (Veenendaal et al., 2010). Another is the non-Markovian quantum state diffusion equation

$\Delta_D$ 5

which yields a stochastic dissipative Schrödinger equation with memory for a qubit–qutrit open system (Jing et al., 2011).

2. Strong dissipation, smoothing, and stabilization

The most striking linear result in this area is that geometric control is not necessary for strong smoothing and uniform stabilization when the damping has sufficient analytic strength. In the obstacle geometry of strictly convex obstacles inside a bounded domain $\Delta_D$ 6, there are trapped rays bouncing between obstacles that never enter the support of $\Delta_D$ 7, so the geometric control condition fails (Aloui et al., 2012). These are the non-controlled orbits.

Despite this, the strongly dissipative equation with $\Delta_D$ 8 satisfies a resolvent estimate near the real axis,

$\Delta_D$ 9

and, more strongly, a Sobolev smoothing estimate

$a$ 0

for every $a$ 1 and $a$ 2 (Aloui et al., 2012). By Fourier transform in time and Plancherel, this yields an inhomogeneous smoothing effect: $a$ 3 The gain is almost one derivative globally in space and on arbitrary finite time intervals (Aloui et al., 2012).

For the homogeneous problem, the same mechanism implies instantaneous regularization: $a$ 4 for every $a$ 5 (Aloui et al., 2012). This is much stronger than classical Kato $a$ 6-smoothing for conservative Schrödinger dynamics, which is only local in space and generally fails in trapping geometries.

The same resolvent control implies uniform exponential stabilization: $a$ 7 so the energy $a$ 8 decays exponentially (Aloui et al., 2012). The decisive point is that $a$ 9 is nonnegative and nonlocal; although the multiplier $\partial B$ 0 is supported near $\partial B$ 1, the operator acts through $\partial B$ 2 and high-frequency trapped components still feel the damping analytically. This distinguishes strong derivative damping from weak zeroth-order damping $\partial B$ 3, for which geometric control is typically essential.

A plausible implication is that, for certain dissipative Schrödinger models, analytic ellipticity can replace geometric optics control. That theme reappears in other settings where dissipation regularizes or stabilizes modes that would otherwise remain trapped or weakly damped.

3. Nonlinear dissipative equations: decay, asymptotics, and modulated coherent structures

For nonlinear dissipative Schrödinger equations with complex nonlinearity, the fundamental structural condition is $\partial B$ 4, which makes the $\partial B$ 5-norm monotone decreasing (Kawamoto et al., 2022, Kita et al., 15 May 2026). In the equation

$\partial B$ 6

large-data $\partial B$ 7-decay is proved in the sharp decay range

$\partial B$ 8

for arbitrary data $\partial B$ 9, even in the attractive-dissipative case $B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 0 and without the strong dissipative condition previously required in that setting (Kita et al., 15 May 2026). The key innovation is the augmented energy

$B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 1

where

$B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 2

Adding a suitable multiple of the decreasing mass term produces an extra dissipative contribution and yields a direct uniform-in-time $B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 3 bound (Kita et al., 15 May 2026). The resulting $B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 4-decay is

$B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 5

This establishes polynomial or logarithmic decay rather than exponential decay, reflecting that the dissipative term weakens with the amplitude (Kita et al., 15 May 2026).

A related asymptotic regime is analyzed for

$B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 6

with $B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 7 and

$B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 8

for a class of arbitrarily large oscillatory initial data (Cazenave et al., 2020). After a pseudo-conformal transformation, the long-time behavior is reduced to the study of a nonautonomous equation

$B_a = a(x)(-\Delta_D)^{1/2}a(x),$ 9

The main theorem gives a profile $A_a = -\Delta_D + i B_a.$ 0 such that

$A_a = -\Delta_D + i B_a.$ 1

and, crucially,

$A_a = -\Delta_D + i B_a.$ 2

The $A_a = -\Delta_D + i B_a.$ 3 decay rate is therefore universal within the considered class, while the $A_a = -\Delta_D + i B_a.$ 4-decay rate depends on the weighted-regularity parameter $A_a = -\Delta_D + i B_a.$ 5 in the initial-data space (Cazenave et al., 2020).

Time-dependent external potentials alter this decay/non-decay dichotomy in a precise way. For

$A_a = -\Delta_D + i B_a.$ 6

with $A_a = -\Delta_D + i B_a.$ 7, the critical quantity controlling whether the $A_a = -\Delta_D + i B_a.$ 8-mass decays is

$A_a = -\Delta_D + i B_a.$ 9

where $1$0 is a fundamental solution of

$1$1

(Kawamoto et al., 2022). If $1$2 diverges, the mass decays to zero; if it converges, small solutions can retain positive mass asymptotically. For $1$3, the critical exponent becomes

$1$4

where $1$5 (Kawamoto et al., 2022). This makes the potential-mediated modification of effective dispersion explicit.

Damped equations with linear dissipation alter coherent structures in a different way. On the torus, the linearly damped cubic NLS

$1$6

destroys exact cnoidal standing waves, since the mass decays exponentially,

$1$7

Nevertheless, the cnoidal family $1$8 remains a slow manifold, and the solution stays close to a mass-modulated profile $1$9 with

$\mathbb{R}^d$ 0

provided the initial perturbation is small and $\mathbb{R}^d$ 1 is sufficiently small (Antonelli et al., 2022). The main estimate is

$\mathbb{R}^d$ 2

Because $\mathbb{R}^d$ 3 is not an exact solution of the damped equation, the proof requires a first-order approximate profile $\mathbb{R}^d$ 4 and an exponentially decreasing Lyapunov functional around that corrected manifold (Antonelli et al., 2022). This suggests that, in dissipative NLS, orbital stability is often replaced by stability of a moving or decaying family.

4. Geometry, networks, and finite-frequency damping

Localized damping can stabilize nonlinear Schrödinger dynamics on graphs and other non-Euclidean geometries. On a star-shaped network with $\mathbb{R}^d$ 5 semi-infinite branches, damping on only one branch and only at infinity,

$\mathbb{R}^d$ 6

is enough to produce exponential decay of the global $\mathbb{R}^d$ 7-mass for the cubic case $\mathbb{R}^d$ 8, and for the quintic case $\mathbb{R}^d$ 9 under a smallness assumption (Ammari et al., 2019). The global energy identity is

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 0

where $i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 1 is the damped branch (Ammari et al., 2019). The key step is an observability-type estimate showing that the total damping integral controls the energy on undamped branches and on the undamped portion of the damped branch. The proof uses compactness–uniqueness, smoothing, and unique continuation rather than frequency-domain semigroup criteria (Ammari et al., 2019).

Another physically concrete dissipative NLS arises in ocean-wave modeling in the marginal ice zone: $i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 2 The damping coefficient

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 3

is frequency dependent, strongly attenuating high-frequency modes (Alberello et al., 2022). Because the dissipation is diagonal in the frequency domain, short-period waves decay fastest, which causes spectral peak downshift and less-than-exponential decay of integrated energy. Nonlinearity counteracts the linear attenuation by feeding lower, less dissipative frequencies, so the nonlinear dissipative model predicts weaker attenuation than the corresponding linear model with the same $i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 4 (Alberello et al., 2022). The paper also reports a tendency toward Gaussian wave statistics deeper into the ice zone as dissipation and downshift reduce steepness and suppress modulational instability (Alberello et al., 2022).

These examples underscore that “dissipative Schrödinger equation” is not a single operator-theoretic notion. It includes geometrically localized linear damping, branchwise damping on graphs, and strongly frequency-selective attenuation. What unifies them is the competition between dispersion and a loss mechanism that is localized either in physical space, Fourier space, or network topology.

5. Hydrodynamic and fluid-mechanical Schrödinger formulations

A major modern development is the use of Schrödinger-type equations as compact wave representations of viscous or capillary fluids. In the Schrödinger–Navier–Stokes equation,

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 5

the Madelung representation

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 6

yields the continuity equation

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 7

and the Navier–Stokes–Korteweg momentum equation

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 8

with

$i \partial_t u + \dfrac{1}{2}\Delta u = \lambda |u|^{p-1}u,$ 9

(Salasnich et al., 13 Apr 2026). The conservative and dissipative structures arise variationally from

and the Rayleigh functional

(Salasnich et al., 13 Apr 2026). In this framework, dissipation is not an ad hoc non-Hermitian perturbation but an Onsager–Rayleigh contribution encoding viscous loss.

A complementary construction starts from

and shifts the nonlinear potential by

Under the Madelung map

this yields

and

(Salasnich et al., 2023). In the incompressible case $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$10, one recovers a classical Navier–Stokes viscosity $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$11 (Salasnich et al., 2023).

These hydrodynamic models broaden the meaning of dissipative Schrödinger equation well beyond open-quantum-system damping. They use complex wave evolution as a representation of classical dissipative continuum mechanics. A plausible implication is that Schrödinger-type PDEs may serve as an interface language between quantum simulation, capillary-fluid theory, and computational fluid dynamics.

6. Open quantum systems, stochastic formulations, and non-Hermitian effective dynamics

In open-system theory, dissipative Schrödinger equations often arise as trajectory equations rather than closed deterministic PDEs. One direct construction begins from basis-state probabilities $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$12 satisfying microscopic or phenomenological rate equations

and builds an effective dissipative operator

The modified wave equation

then preserves total probability but not system energy (Veenendaal et al., 2010). The construction is illustrated for direct electronic decay and for phonon damping, where microscopic system–bath couplings yield rate equations such as

for local phonon occupations (Veenendaal et al., 2010). This formulation is computationally cheaper than density-matrix propagation because it evolves $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$17 amplitudes rather than $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$18 density-matrix entries (Veenendaal et al., 2010).

A more systematic stochastic formulation is quantum state diffusion. For the qubit–qutrit system coupled to a bosonic bath at zero temperature, the exact non-Markovian QSD equation is

where $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$20, $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$21 is complex Gaussian noise, and $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$22 is the bath correlation (Jing et al., 2011). Introducing an $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$23-operator

renders the dynamics time local: $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$25 (Jing et al., 2011). Averaging over trajectories recovers the density matrix. In the Markov limit, the corresponding master equation becomes Lindblad: $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$26 The dissipative Schrödinger equation here is therefore stochastic, non-Markovian, and trajectory based, rather than a deterministic nonlinear PDE (Jing et al., 2011).

At a more formal thermodynamic level, a stochastic, dissipative Schrödinger equation can be derived from a non-equilibrium entropy operator $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$27, with probability operator

and stochastic propagator satisfying average unitarity (Attard, 2014). The finite-step evolution has the form

anchoring the non-Hermitian subsystem dynamics in entropy production and average unitarity (Attard, 2014).

These open-system equations show that “dissipative Schrödinger equation” can mean: a deterministic effective non-Hermitian wave equation; a stochastic unraveling of a master equation; or a thermodynamic trajectory law derived from entropy and fluctuation–dissipation structure.

7. Driven, damped, and rotationally dissipative structures

Driven-damped nonlinear Schrödinger equations support localized dissipative structures that differ fundamentally from conservative solitons. In the AC-driven damped NLS

localized dissipative structures are phase-locked to the drive and exist as attractors on a homogeneous background (Jang et al., 2015). Their interactions are intrinsically inelastic: depending on the detuning $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$34 and drive strength $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$35, collisions induced by a phase-modulated driver can lead to merging into a single structure or annihilation into the homogeneous state (Jang et al., 2015). The driver phase profile $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$36 imposes a drift law

so collisions can be triggered on demand (Jang et al., 2015). This is qualitatively unlike conservative NLS solitons, whose collisions are elastic.

A different rotationally dissipative context is the dissipative Gross–Pitaevskii equation under rotation,

Here dissipation converts energetic instabilities of the vortex-free ground state into dynamical instabilities (Carretero-Gonzalez et al., 2014). Linear analysis yields a critical mode number

and critical rotation

so the most unstable mode scales like $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$41 (Carretero-Gonzalez et al., 2014). Nonlinearly, this mode first nucleates many vortices at the cloud periphery, but through symmetry breaking and dissipation only a much smaller number spiral inward and survive as stable vortex configurations (Carretero-Gonzalez et al., 2014). This is another instance where dissipation creates attractor selection rather than conservative persistence.

Even in quantum thermodynamic modeling, an effective driven-dissipative Schrödinger equation appears: $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$42 with $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$43 diagonal in the energy basis and built from bath-induced population rates (Fang et al., 2020). In the quantum Otto engine setting, this equation reproduces relaxation and thermalization of a harmonic oscillator working medium. Apparent transient efficiencies above the Otto or Carnot limits arise from energy stored in the initial state or supplied by pumping, not from a violation of thermodynamic bounds (Fang et al., 2020). This suggests that dissipative Schrödinger equations can also function as reduced thermodynamic simulators when coherence is secondary and population dynamics dominates.

Across these examples, dissipation does not merely damp amplitudes. It reshapes the attractor landscape, selects patterns, and defines whether coherent structures merge, decay, spiral, or relax.

8. Conceptual synthesis and recurrent themes

Several themes recur across the theory.

First, the form of dissipation matters more than the mere presence of loss. Strong derivative damping $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$44 yields smoothing and exponential stabilization without geometric control (Aloui et al., 2012), whereas weak zeroth-order damping often cannot overcome trapping. Nonlinear dissipative terms $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$45 with $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$46 produce algebraic or logarithmic decay (Kita et al., 15 May 2026, Cazenave et al., 2020), and linearly damped torus or graph models can stabilize only relative to moving or decaying manifolds (Antonelli et al., 2022, Ammari et al., 2019).

Second, dissipative Schrödinger equations often interpolate between ODE-like amplitude relaxation and PDE-like dispersive transport. The universal limit

for nonlinear dissipation in $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$48 (Cazenave et al., 2020) is ODE-like, while the spatial profile and $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$49-decay retain dispersive dependence on the initial state. The critical decay/non-decay criterion involving $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$50 under time-dependent harmonic potentials (Kawamoto et al., 2022) is another expression of this interplay.

Third, many dissipative Schrödinger equations admit hydrodynamic interpretations. Through Madelung variables, dissipation can appear as viscosity $\left\{ \begin{array}{ll} i\partial_t u(t,x) - \Delta_D u(t,x) + i a(x)(-\Delta_D)^{1/2} a(x) u(t,x) = 0, & (t>0,\ x\in\Omega_0),\[4pt] u(0,x) = u_0(x), & x\in\Omega_0,\[4pt] u(t,x)=0, & (t>0, x\in\partial\Omega_0), \end{array} \right.$51 (Salasnich et al., 13 Apr 2026, Salasnich et al., 2023), stochastic forcing as a Langevin term in the packet center (Mousavi et al., 2019), and derivative damping as a global smoothing mechanism (Aloui et al., 2012). This makes the topic a meeting point between PDE theory, open quantum systems, fluid dynamics, and nonequilibrium statistical mechanics.

Fourth, probability conservation is not universal. Some effective dissipative wave equations preserve norm by construction (Veenendaal et al., 2010) or maintain the standard continuity equation [(Mousavi et al., 2019), second generalized equation], while others yield source/sink terms in the continuity law and even violate standard Ehrenfest relations [(Mousavi et al., 2019), first generalized equation]. Stochastic trajectory formulations preserve physical normalization only after suitable normalization or ensemble averaging (Jing et al., 2011, Attard, 2014).

Finally, dissipation changes what “stability” means. In Hamiltonian Schrödinger equations, stability is usually orbital or scattering based. In dissipative equations, one instead encounters exponential stabilization (Aloui et al., 2012), attraction to the zero solution with sharp decay rates (Kita et al., 15 May 2026, Cazenave et al., 2020), convergence toward mass-decaying coherent structures (Antonelli et al., 2022), selection of stable vortex patterns (Carretero-Gonzalez et al., 2014), and relaxation toward entangled dark-state mixtures in open systems (Jing et al., 2011).

Taken together, these works show that the dissipative Schrödinger equation is best understood not as a single canonical PDE, but as a family of non-Hamiltonian Schrödinger-type evolutions whose analytic behavior is governed by the precise mechanism of loss: derivative damping, complex nonlinearity, localized dissipation, stochastic reservoir coupling, coherent drive, or hydrodynamic viscosity. The common consequence is that irreversibility becomes a structural part of the wave dynamics rather than an external perturbation.