Self-Consistent Saturation Models

• Self-consistent saturation models are theoretical frameworks in which the saturation level emerges from internal feedback loops rather than being externally imposed.
• They integrate variables such as order parameters, fluctuation amplitudes, and electromagnetic fields using techniques like mean-field decoupling, gyrokinetic ordering, and Bayesian active learning.
• These models apply across disciplines—from correlated electron systems and plasma instabilities to machine-learned dielectric responses—highlighting both their practical utility and inherent domain limitations.

“Self-consistent saturation model” denotes a class of formulations in which the saturated state is not imposed externally but is obtained from a closure loop internal to the theory. In the supplied literature, that loop takes different technical forms: mean-field order parameters determined by expectation values of the same Hamiltonian in which they appear, fluctuation amplitudes fixed by gyrokinetic ordering, Bayesian posteriors restricted to admissible monotone saturation curves, self-consistent electromagnetic evolution from spontaneous start-up to deep saturation, and nonlinear field or phase-space structures that regulate further growth. Taken together, these works suggest that the expression is best understood as a general principle of feedback-constrained saturation rather than as a single formalism (Matveenko et al., 2021, Yamagishi et al., 27 Apr 2026, Navabi et al., 25 Jun 2026, Dhattarwal et al., 2023).

1. Definitional framework

Across the cited works, self-consistent means that the relevant fields, amplitudes, or response curves are determined from the same state that they generate. In the two-dimensional $t\text{–}U\text{–}V$ Hubbard-type model, the PDW, SDW, and CDW order parameters are not imposed; they are determined by expectation values in the ground state of the Hamiltonian containing those very order parameters. In the quasilinear gyrokinetic transport model of Yamagishi and Watanabe, the saturation amplitude is fixed by multiscale gyrokinetic ordering rather than by calibration to nonlinear simulations. In the Bayesian I-spline framework, self-consistency is statistical: every posterior draw satisfies $f(0)=0$, monotonicity, and plateau behavior within the observed window (Matveenko et al., 2021, Yamagishi et al., 27 Apr 2026, Navabi et al., 25 Jun 2026).

Saturation likewise has domain-specific meanings. In correlated-electron models it denotes a fully developed modulated ground state with fixed amplitude and periodicity; in transport theory it denotes a fluctuation level consistent with the ordering and the flux closure; in active learning it denotes a monotone rise to a plateau; in plasma and FEL problems it denotes a finite-amplitude state reached through back-reaction, trapping, or mode coupling. A plausible implication is that the term retains its coherence only at the level of structure: a feedback loop selects a terminal or plateaued state under internal constraints.

2. Comparative realizations

The principal realizations represented in the supplied papers are summarized below.

Domain / representative paper	Self-consistent quantity	Saturated state or constraint
Correlated electrons (Matveenko et al., 2021)	Bond PDW, SDW, and CDW from expectation values of the same mean-field Hamiltonian	Stripe lattice; same CDW and PDW period; pseudogap
Quasilinear gyrokinetics (Yamagishi et al., 27 Apr 2026)	Fluctuation amplitude from gyrokinetic ordering	Weighted flux spectrum; $Q_i\sim Q_e$ as a closed conclusion
Bayesian active learning (Navabi et al., 25 Jun 2026)	Monotone I-spline posterior with $w_m \ge 0$	Plateau at $x_{\max}$; unique level crossing
FEL simulation (Litvinenko, 2015)	MP/C particle-field evolution under the same Maxwell–Lorentz dynamics	Start-up from spontaneous radiation through deep saturation
Bounded plasma (Xu et al., 2020)	Virtual-anode potential, beam ions, and trapped ions	Oscillating virtual anode and system-scale ion hole
Granular-fluid saturation (Melnikov et al., 2015)	Cluster pressure from volume conservation and interface geometry	Bridges, menisci, filled pores, and Haines jumps
Dielectric saturation in water (Dhattarwal et al., 2023)	Long-range field and induced dipoles in an SCFNN loop	Non-linear polarization and saturation of the dielectric constant
EP-driven modes with zonals (Barberis et al., 25 Jul 2025)	Pump and zonal amplitudes constrained by energy conservation	Reduced saturated mode amplitude when zonal effects modify sinks or sources
Bell instability (Zacharegkas et al., 2019)	CRs, background plasma, and fields in hybrid evolution	Pressure equilibration; Bell and WICE saturation regimes

This comparison shows that the same label is attached to microscopically different closures. The shared content is not the detailed mathematics but the refusal to prescribe the saturated state independently of the dynamics that produce it.

3. Ordered states and self-organized structures

In the pair-density-wave study on the square-lattice $t\text{–}U\text{–}V$ model, self-consistency is explicit at the level of the mean-field decoupling: $$\Delta_{sc}(i,j) = -g_{sc} \langle c_{j\downarrow} c_{i\uparrow} \rangle,\quad \Delta_s(i,j) = -g_s \sum_{\sigma}\sigma \langle c^{\dagger}_{i\sigma} c_{j\sigma} \rangle,\quad \Delta_c(i,j) = -g_c \sum_{\sigma} \langle c^{\dagger}_{i\sigma} c_{j\sigma} \rangle.$$ The model yields analytic ground-state solutions with coexisting bond-centered $d_{x^2-y^2}$-symmetric PDW and SDW or CDW. In the SDW+PDW sector, the linearized $4\times 4$ BdG-type problem reduces to an effective $2\times 2$ system, and the single-stripe solution is written in hyperbolic-tangent form; with increasing doping it saturates into a periodic stripe lattice described by a Jacobi elliptic $f(0)=0$0 function. In the CDW+PDW sector, the ordering wavevectors coincide,

$f(0)=0$1

so CDW and PDW share the same period. The same construction produces a pseudogap in the Fermi-excitation spectrum and encodes direct competition between SDW or CDW order and superconducting order. The paper ties this locking to cuprate observations discussed in connection with Lake et al., Aeppli et al., and STM work associated with Hamidian, Davis, and Edkins (Matveenko et al., 2021).

A formally different but structurally similar example appears in the bounded ion-beam–plasma study of Pierce-type ion-sound instability. There, 1D PIC simulations with Boltzmann electrons show that saturation is mediated by an oscillating virtual anode potential structure near the injection boundary. The potential barrier separates incoming ions into a supersonic beam accelerated to $f(0)=0$2 and a trapped population forming an ion hole with effective temperature

$f(0)=0$3

In the final stage the ion hole expands over the whole system length. This suggests a second archetype of self-consistent saturation: not amplitude clipping, but reorganization into a stable phase-space structure that both stores and redistributes the instability energy (Xu et al., 2020).

4. Saturation amplitudes as transport and wave closures

In the quasilinear gyrokinetic transport model, saturation is specified by the ordering relation rather than by empirical mixing-length calibration. The central prescription is

$f(0)=0$4

which is then inserted into the quasilinear flux. The resulting wavenumber-resolved flux is expressed in ion gyro-Bohm units with a weighting factor $f(0)=0$5, so that the area integral in a log-linear representation yields the total flux. For systems with comparable ion and electron temperature gradients, the QL ion energy flux reproduces nonlinear simulation results in both wavenumber dependence and absolute magnitude, while the QL electron flux remains predominantly electron-scale. The relation $f(0)=0$6 is presented as a closed conclusion of the model and may be predictive if the area-integrated flux is conserved in the nonlinear energy cascade process (Yamagishi et al., 27 Apr 2026).

The perturbative model for energetic-particle-driven modes limited by self-generated zonal modes uses a different closure variable: the pump and zonal amplitudes are coupled by energy conservation. If $f(0)=0$7 and $f(0)=0$8 denote pump and zonal amplitudes, the wave–wave transfer obeys

$f(0)=0$9

while the beat-driven zonal growth is assumed to be

$Q_i\sim Q_e$0

In the collisionless regime, zonal generation reduces the saturated pump amplitude because part of the free energy is transferred into the self-generated zonal mode. In the scattering-dominated regime, beat-driven zonal generation by itself does not change the final saturation level; saturation is altered only when a finite zonal mode reduces microturbulent particle scattering and therefore limits the energetic-particle source. This is an explicit example in which self-consistent saturation depends on sources and sinks as much as on nonlinear coupling (Barberis et al., 25 Jul 2025).

5. Statistical, electrodynamic, and machine-learned realizations

In the Bayesian active-learning framework for self-limiting saturation curves, the surrogate is

$Q_i\sim Q_e$1

with monotone I-splines $Q_i\sim Q_e$2 satisfying $Q_i\sim Q_e$3 and HalfNormal priors enforcing $Q_i\sim Q_e$4. Because

$Q_i\sim Q_e$5

every posterior draw is monotone non-decreasing, starts at zero, and approaches a plateau given by $Q_i\sim Q_e$6, anchored by a measurement at maximum exposure $Q_i\sim Q_e$7 rather than by a parametric asymptote. The platform reaches noise-floor accuracy within a 20-measurement budget in every benchmarked regime, in as few as seven measurements, and the predicted pulse times near saturation err only on the conservative side. Here, self-consistency is statistical and operational: the posterior family, the acquisition rule, and the inferred saturation levels all respect the same admissible shape constraints (Navabi et al., 25 Jun 2026).

In Litvinenko’s macro-particle / clone model for FELs, self-consistency refers to the joint treatment of spontaneous and induced radiation by the same Maxwell–Lorentz evolution. Each electron macro-particle is paired with a positron clone so that shot-noise start-up, gain, harmonic generation, and saturation are simulated without anomalously strong spontaneous radiation and without a seed. The model is stated to provide statistically exact simulation of multi-mode, multi-harmonic and multi-frequency short-wavelength 3-D FELs including the high power and saturation effects. The saturation process is therefore not seeded phenomenologically; it emerges from self-consistent evolution of particle motion and radiation field (Litvinenko, 2015).

In the SCFNN description of water, Remsing and Gao’s long-range machine-learning framework provides a self-consistent field loop in which the long-range electrostatic field and the induced molecular dipoles are iterated to convergence. The model predicts non-linear response at high electric fields, including saturation of the dielectric constant, without training on those high field strengths or on the resulting liquid configurations. The macroscopic dielectric response is read from

$Q_i\sim Q_e$8

and the simulations identify a crossover from nearly linear response at low field to dielectric saturation at larger field, with both orientational and electronic contributions resolved microscopically. This is a saturation model in which the plateau arises from learned long-range response rather than from a hard-coded constitutive law (Dhattarwal et al., 2023).

6. Granular, astrophysical, and cross-domain limitations

At grain scale in random packings, arbitrary fluid saturation is modeled by explicitly combining capillary bridges, menisci, and fully saturated pores into liquid clusters. The packing geometry is obtained by Delaunay triangulation, and each cluster pressure is updated so that the cluster volume $Q_i\sim Q_e$9 matches the actual liquid volume through a false-position root search. Redistribution occurs through thin films according to

$w_m \ge 0$0

and local instabilities, including Haines jumps, trigger topological changes such as trimer formation, pore imbibition, and drainage. The model is self-consistent because grain geometry, Laplace pressure, cluster volume, and connectivity are updated together; saturation is represented as the evolution from isolated bridges to connected clusters and eventually to a percolating liquid body (Melnikov et al., 2015).

In hybrid simulations of the Bell instability, saturation is likewise not prescribed. CRs, background ions, electrons as a neutralizing fluid, and fields evolve self-consistently, allowing CR back-reaction and gas acceleration to regulate growth. In the Bell regime, saturation occurs when the gas perpendicular pressure reaches approximately $w_m \ge 0$1, with magnetic pressure scaling close to a fraction of the initial CR pressure. In the WICE regime, magnetic pressure remains well below that level, indicating thermal suppression and additional damping. The saturated magnetic energy also correlates linearly with CR kinetic energy per particle. This establishes a pressure-equilibration picture of saturation rather than a purely kinematic threshold (Zacharegkas et al., 2019).

A recurrent misconception is to treat “self-consistent saturation model” as a unique formalism. The papers instead use the phrase for different closures that share only the feedback principle. Domain-specific limitations are therefore essential. In the active-learning setting, the plateau is defined relative to the measurement window at $w_m \ge 0$2, and a single fixed log-spaced knot grid reaches a capacity boundary at the sharpest sigmoidal onset (Navabi et al., 25 Jun 2026). In the quasilinear gyrokinetic setting, the shift of electron-scale transport toward ion scales observed in nonlinear simulations is not captured within the present linear framework (Yamagishi et al., 27 Apr 2026). In the MP/C FEL framework, coherent radiation at even harmonics of the fundamental is suppressed (Litvinenko, 2015). These caveats indicate that self-consistency constrains the internal logic of a model, but does not eliminate approximation, regime dependence, or the need to specify which saturation mechanism is being closed.