Wave-Particle Turbulence Simulation (WPTS) Overview

• Wave-Particle Turbulence Simulation (WPTS) is a multiscale framework that represents turbulent flow as a coupled wave field and discrete particle transport model.
• It employs a BGK-type kinetic model alongside Eulerian finite-volume and stochastic Lagrangian methods to implicitly capture unresolved turbulent kinetic energy.
• The method adapts between laminar and turbulent regimes and has been validated in canonical shear flows and round-jet cases with promising coarse-grid performance.

Searching arXiv for the WPTS papers and closely related work to ground the article in the cited literature. arXiv Search Query: "Wave-Particle Turbulence Simulation" Wave-Particle Turbulence Simulation (WPTS) denotes a multiscale, non-equilibrium turbulence framework in which the flow is represented as a coupled mixture of a wave component and a particle component. In the engineering-flow formulation, the wave component captures resolved, cell-scale structures in an Eulerian finite-volume setting, while the particle component represents unresolved turbulent motion through stochastic Lagrangian transport derived from a relaxation-type kinetic equation. The same expression also appears, in a broader and less method-specific sense, across simulations that couple wave dynamics, particle transport, and turbulent energy transfer; this suggests that WPTS is both a named turbulence solver and a wider research style for studying multiscale cascade and energization processes (Yang et al., 10 Mar 2025, Yang et al., 19 Jul 2025).

1. Conceptual basis

The core hypothesis of WPTS is that turbulence should not be treated only as an enhanced diffusion of the mean flow. In the formulation developed for non-equilibrium turbulent flows, turbulence emergence is attributed to the breakdown of continuously connected fluid elements under the cell resolution. The unresolved part of the flow is then represented by discrete fluid elements carrying turbulent kinetic energy, while the organized part continues to evolve as a wave-like background field (Yang et al., 10 Mar 2025).

This construction differs explicitly from both major reduced descriptions. It does not rely on a mean/fluctuation decomposition of the RANS type, and it does not reduce unresolved turbulence to a purely local eddy-viscosity correction of the LES type. Instead, unresolved transport is carried by particles that can move non-equilibrium, penetrate across cells, collide or relax into the background, and exchange mass, momentum, and energy with the wave field. In the round-jet implementations, the method automatically collapses to a pure wave formulation in well-resolved or laminar regions and behaves like a high-order gas-kinetic scheme, whereas particles appear adaptively in under-resolved turbulent regions (Yang et al., 19 Jul 2025).

A common misconception is therefore that WPTS is merely “particles added to CFD.” The published formulations reject that reading. The particle component is not an auxiliary tracer population but the explicit representation of unresolved turbulent transport, and the laminar–turbulent distinction is controlled by the local particle density in the wave-particle decomposition inside each cell (Yang et al., 10 Mar 2025, Yang et al., 10 Feb 2026).

2. Kinetic formulation and wave-particle decomposition

The mathematical backbone of WPTS is a BGK-type kinetic model,

$$\frac{\partial f}{\partial t} + \nabla_x \cdot (u f) = \frac{g-f}{\tau},$$

where $f$ is the distribution function, $u$ is particle velocity, $g$ is the equilibrium distribution, and $\tau$ is the relaxation time. Macroscopic variables are obtained from velocity moments with

$$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$

subject to the compatibility condition

$\int \psi\,(g-f)\,d\Xi = 0.$

In turbulent regions, the equilibrium state is extended so that the energy contains kinetic, thermal, and turbulent contributions; the turbulent part is written as

$\rho E_t = \frac{3}{2}\rho \Theta_t,$

so that the effective temperature entering the equilibrium distribution includes a turbulence temperature $\Theta_t$ (Yang et al., 19 Jul 2025, Yang et al., 10 Mar 2025).

The decisive step is the integral solution of the kinetic equation,

$$f(x,t,u)=\frac{1}{\tau}\int_0^t g(x',t',u)e^{-(t-t')/\tau}\,dt' + e^{-t/\tau}f_0(x-ut,u).$$

WPTS interprets the first term as the accumulated equilibrium contribution and the second term as free streaming from the initial distribution. In the published formulation, the equilibrium part becomes the wave flux $f$0, and the free-streaming part becomes the particle component (Yang et al., 19 Jul 2025).

This decomposition has an important consequence for turbulence closure. The turbulent kinetic energy is not advanced by a separate closed transport equation. Instead, it is evolved implicitly through particle sampling, advection, deletion, and merging. In the broader non-equilibrium WPTS formulation, the particle population itself defines the unresolved turbulent content; in the compressible mixing-layer formulation, turbulent kinetic energy can be written directly from particle fluctuations as

$f$1

so the unresolved energy is measured from the simulated microstate rather than from a separate turbulence-energy PDE (Yang et al., 19 Jul 2025, Yang et al., 10 Mar 2025).

3. Numerical realization

In the round-jet implementation, the wave part is advanced by a gas-kinetic scheme using fifth-order WENO-AO reconstruction for the resolved conserved variables, second-order reconstruction with a van Leer limiter for the subgrid wave flux, and a two-step fourth-order method for time integration. The wave flux at an interface is written as

$f$2

with

$f$3

and

$f$4

If a cell contains no particles, the method reverts automatically to the high-order GKS flux (Yang et al., 19 Jul 2025).

The particle part is modeled as stochastic, non-equilibrium transport. In the 2025 round-jet paper, particle dynamics follow

$f$5

with effective collision time

$f$6

Particles are sampled from the wave component through

$f$7

then transported by operator splitting,

$f$8

$f$9

and deleted when their stochastic lifetime ends; the surviving or returning content is merged back into the cell averages (Yang et al., 19 Jul 2025).

The full update of the conservative state is

$u$0

Each stochastic particle carries mass, momentum, and energy,

$u$1

This is the operative form of the wave-particle coupling: equilibrium transport is computed analytically through the gas-kinetic flux, while non-equilibrium transport is counted through particle crossings (Yang et al., 19 Jul 2025).

4. Characteristic-time closures and canonical validations

Although WPTS avoids a conventional eddy-viscosity closure, it still depends critically on a modeled turbulence characteristic time $u$2. In the 2025 round-jet validation, $u$3 is described as central to both accuracy and efficiency, and it is modeled by a previously proposed expression involving both resolved and particle contributions, with $u$4, $u$5, and $u$6 when $u$7 and zero otherwise (Yang et al., 19 Jul 2025). In the 2026 jet study, this closure is replaced by a Prandtl-inspired mixing-length form,

$u$8

so that the particle fraction is governed primarily by the turbulence characteristic time rather than by a separate transport equation (Yang et al., 10 Feb 2026).

The validation history is centered on canonical shear flows. In the earlier non-equilibrium WPTS study of a temporal compressible mixing layer at $u$9 and $g$0, the method captured vorticity growth, mixing-layer thickening, and self-similar development. The momentum-thickness growth became linear after transition with slope $g$1, close to the cited DNS value $g$2, and Reynolds stresses $g$3 agreed well with DNS and experiment after averaging over time. On the same coarse mesh, plain GKS and LES with Smagorinsky were reported to perform worse (Yang et al., 10 Mar 2025).

For the spatially developing round jet at

$g$4

the 2025 paper used a domain of $g$5 and a nonuniform $g$6 Cartesian mesh, described as only about 2% of the number of cells used in DNS. The method reproduced the characteristic linear centerline decay, radial mean-velocity profiles, and anisotropic Reynolds stresses. Using the similarity law

$g$7

the reported fit was $g$8 and $g$9, compared with $\tau$0 from Hussein, $\tau$1 from Panchapakesan, and $\tau$2 from the DNS of Sharan et al. Statistics accumulated over the final $\tau$3 also showed that changing the inlet perturbation frequency ratio from $\tau$4 to $\tau$5 shifted the transition region only slightly while leaving the self-similar far-field jet statistics essentially unchanged (Yang et al., 19 Jul 2025).

The 2026 jet paper retained the same general WPTS framework but used the mixing-length characteristic time. For $\tau$6, it reported $\tau$7 and $\tau$8, with a fitted centerline decay constant $\tau$9. For $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$0, the tuned coefficient became $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$1, and the centerline decay slope was reported as $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$2, differing by about $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$3 from the cited reference value $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$4. The same paper notes that the outer radial region is less well captured, which it attributes to the simple $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$5-dependent mixing-length form and limited radial freedom in the $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$6 closure (Yang et al., 10 Feb 2026).

5. Broader research landscape

The surrounding literature suggests that WPTS also names, or at least aligns with, a broader class of simulations in which turbulent transfer is studied through explicit wave dynamics, particle kinetics, or both. In that wider sense, the engineering-flow WPTS method sits alongside several neighboring traditions rather than exhausting the term’s meaning.

One branch concerns wave-turbulence DNS. A three-dimensional direct numerical simulation of free-surface wave turbulence on a ferrofluid in a horizontal magnetic field showed a transition from isotropic capillary-wave turbulence to strongly anisotropic magneto-capillary turbulence. In the strong-field regime, energy transfer occurred mainly perpendicular to the magnetic field, with

$$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$7

and scaling $$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$8, closely analogous to anisotropic Alfvén-wave turbulence (Kochurin et al., 2022). A distinct quasi-one-dimensional capillary-wave study found that, even though exact three-wave and nontrivial exact four-wave resonances are absent in the ideal collinear geometry, nonlinear broadening still supports a cascade dominated by four-wave quasi-resonant interactions, with measured spectrum

$$\psi = \left(1, u, \frac{1}{2}(u^2+\xi^2)\right)^T, \qquad W=(\rho,\rho U,\rho E)^T,$$9

consistent with $\int \psi\,(g-f)\,d\Xi = 0.$0 (Kochurin et al., 2020). In the Gross–Pitaevskii system, direct numerical simulations of the GPE were compared with the wave-kinetic equation and found to agree accurately for both the wave-action spectrum and the probability density functions of Fourier-mode intensities for about two nonlinear kinetic times, with Gaussianization occurring on the same kinetic time scale (Zhu et al., 2021).

A second branch concerns wave-particle energization in plasmas. In a hybrid Vlasov-Maxwell simulation of Alfvén-ion cyclotron turbulence, field-particle correlations separated Landau and cyclotron channels from single-point data. Across 64 diagnosed spatial points, the parallel electric field accounted for $\int \psi\,(g-f)\,d\Xi = 0.$1 of the total energization, or about $\int \psi\,(g-f)\,d\Xi = 0.$2 depending on the averaging convention, while the perpendicular field remained significant and localized in the expected cyclotron-resonant velocity range (Klein et al., 2020). A theoretical and numerical study of relativistic turbulence tracked test-particle acceleration in synthetic fast, slow, and Alfvén wave fields, showing that at high rigidity the contributions of the three mode families become comparable to within an order of magnitude once resonance broadening is included (Demidem et al., 2019). In lower-hybrid/whistler turbulence driven by a cold ion ring beam, a reanalysis of three-dimensional PIC data argued that nonlinear induced scattering by thermal plasma particles dominates over three-wave decay, with the ratio of particle-scattering to wave-decay rates reported as $\int \psi\,(g-f)\,d\Xi = 0.$3 in the low-$\int \psi\,(g-f)\,d\Xi = 0.$4 regime (Rudakov et al., 2012).

A third branch emphasizes structure-preserving particle-field algorithms. A geometric PIC formulation for low-frequency electrostatic perturbations with fully kinetic ions and adiabatic electrons preserved the non-canonical symplectic structure and gauge symmetry, remained charge-conserving, and was applied to drift-wave and ion-temperature-gradient turbulence; at early times the observed diffusion lay between Bohm and gyro-Bohm scaling, while at later times it moved closer to gyro-Bohm scaling, with density blobs prominent in fully developed ITG turbulence (Xiao et al., 2019). In periodic slow-wave structures such as traveling-wave tubes, a multi-particle self-consistent Hamiltonian and a one-dimensional symplectic multi-particle algorithm reproduced linear growth, bunching, trapping oscillations, and power evolution in a measured 3-meter helix TWT, providing a reduced yet self-consistent wave-particle model for long devices (Minenna et al., 2021).

6. Limitations, points of debate, and likely directions

The main limitation of current engineering-flow WPTS implementations is not the wave-particle decomposition itself but the closure of the turbulence characteristic time. The published papers state explicitly that $\int \psi\,(g-f)\,d\Xi = 0.$5 is central to both accuracy and efficiency, that the current implementation uses pressure-gradient forcing as the main particle driving mechanism, and that the quality of the closure remains decisive for predictive performance (Yang et al., 19 Jul 2025). The broader non-equilibrium WPTS paper describes the turbulent collision-time model as preliminary and somewhat heuristic, while the jet-specific paper notes that broader validation is still needed (Yang et al., 10 Mar 2025, Yang et al., 10 Feb 2026).

Another point of discussion is conceptual rather than purely numerical. WPTS does not advance a separate turbulence-energy equation; unresolved transport is encoded directly in the stochastic particle population. This is a major distinction from conventional two-equation closures, but it also means that transition, dissipation, and non-local transport are all mediated through sampling, survival, deletion, and reinjection rules. That design is central to the method’s identity, yet it concentrates modeling responsibility in the particle dynamics and characteristic-time law (Yang et al., 19 Jul 2025).

The current evidence base is strongest for free shear flows and canonical validation cases. The published demonstrations emphasize compressible mixing layers and round jets, with promising coarse-grid performance and direct recovery of Reynolds stress from the simulated field. At the same time, the papers identify open areas: near-wall turbulence is not yet developed in the broader WPTS formulation; the simple jet mixing-length closure lacks explicit radial dependence; and more sophisticated forcing models and improved $\int \psi\,(g-f)\,d\Xi = 0.$6 closures are suggested for stronger separation, curvature, compressibility, and multiphysics coupling (Yang et al., 10 Mar 2025, Yang et al., 10 Feb 2026).

Taken together, these results place WPTS in a distinctive position within turbulence research. In its narrow sense, it is a kinetic, multiscale solver that replaces part of conventional subgrid modeling with non-equilibrium particle transport. In its broader sense, the surrounding literature suggests a family of simulations organized around the same question: how turbulent cascades, waves, and particles exchange energy across scales when neither a purely continuum nor a purely particle description is sufficient.