Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “PLASMA FLUID EQUATIONS”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,333 records · Page 74

Cross-beam energy transfer in direct-drive ICF. II. Theory and simulation of mitigation through increased laser bandwidth

The response of the cross-beam energy transfer instability (CBET) to laser bandwidth is investigated through a combination of theory and simulation. Existing linear theory is generalized to treat broadband lasers, demonstrating that CBET is most effectively suppressed when the bandwidth exceeds the ion-acoustic wave (IAW) frequency. It is shown that for such bandwidths, reverse (seed to pump) transfer becomes possible, which reduces the net energy transfer rapidly as bandwidth is increased. The CBET gain exponent in this regime scales with bandwidth (Δω) as Δω−3 for Gaussian or Lorentzian laser spectra with different scalings possible for other spectra. Comparison of our theory with linearized fluid and particle-in-cell simulations, performed with the laser-plasma simulation environment (LPSE) and vector particle in cell (VPIC) codes, respectively, finds that the model is accurate in the absence of nonlinear processes. However, linear analysis also finds that the IAW energy density scales as Δω−1, implying that nonlinear effects may be more difficult to control than the CBET scaling would suggest. Indeed, nonlinear effects are found to be present in VPIC simulations with high-intensity lasers, despite minimal apparent CBET. Nonlinear processes in the VPIC cases include particle trapping, the two-ion wave decay, and ion wave self-focusing. In some high intensity VPIC cases, these effects lead to net energy transfer from seed to pump and increases to backscatter stimulated Brillouin scattering reflectivities. Finally, for a given bandwidth, we show that improved control of nonlinear processes can be achieved via smoothing by spectral dispersion.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Dynamics of potential vorticity staircase evolution and step mergers in a reduced model of beta-plane turbulence

A two-field model of potential vorticity (PV) staircase structure and dynamics relevant to both beta-plane and drift-wave plasma turbulence is studied numerically and analytically. The model evolves averaged PV whose flux is both driven by and regulates a potential enstrophy field, ε. The model employs a closure using a mixing length model. Its link to bistability, vital to staircase generation, is analyzed and verified by integrating the equations numerically. Long-time staircase evolution consistently manifests a pattern of metastable quasiperiodic configurations, lasting for hundreds of time units, yet interspersed with abrupt ( Δt << 1 ) mergers of adjacent steps in the staircase. The mergers occur at the staircase lattice defects where the pattern has not completely relaxed to a strictly periodic solution that can be obtained analytically. Other types of stationary solutions are solitons and kinks in the PV gradient and ε-profiles. The waiting time between mergers increases strongly as the number of steps in the staircase decreases. This is because of an exponential decrease in interstep coupling strength with growing spacing. The long-time staircase dynamics is shown numerically be determined by local interaction with adjacent steps. In conclusion, mergers reveal themselves through the explosive growth of the turbulent PV flux, which, however, abruptly drops to its global constant value once the merger is completed.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

An h-adaptive mesh method for Boltzmann-BGK/hydrodynamics coupling

We introduce a coupled method for hydrodynamic and kinetic equations on 2-dimensional h-adaptive meshes. We adopt the Euler equations with a fast kinetic solver in the region near thermodynamical equilibrium, while use the Boltzmann-BGK equation in kinetic regions where fluids are far from equilibrium. A buffer zone is created around the kinetic regions, on which a gradually varying numerical flux is adopted. Based on the property of a continuously discretized cut-off function which describes how the flux varies, the coupling will be conservative. In order for the conservative 2-dimensional specularly reflective boundary condition to be implemented conveniently, the discrete Maxwellian is approximated by a high order continuous formula with improved accuracy on a disc instead of on a square domain. The h-adaptive method can work smoothly with a time-split numerical scheme. Through h-adaptation, the cell number is greatly reduced. This method is particularly suitable for problems with hydrodynamics breakdown on only a small part of the whole domain, so that the total efficiency of the algorithm can be greatly improved. Three numerical examples are presented to validate the proposed method and demonstrate its efficiency.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Geostrophic adjustment in a shallow-water numerical model as it relates to thermospheric dynamics

The theory of geostrophic adjustment and its application to the dynamics of the high latitude thermosphere have been discussed in previous papers based on a linearized treatment of the fluid dynamical equations. However, a linearized treatment is only valid for small Rossby numbers given by Ro = V/fL, where V is the wind speed, f is the local value of the Coriolis parameter, and L is a characteristic horizontal scale for the flow. For typical values in the auroral zone, the approximation is not reasonable for wind speeds greater than 25 m/s or so. A shallow-water (one layer) model was developed that includes the spherical geometry and full nonlinear dynamics in the momentum equations in order to isolate the effects of the nonlinearities on the adjustment process. A belt of accelerated winds between 60 deg and 70 deg latitude was used as the initial condition. The adjustment process was found to proceed as expected from the linear formulation, but that an asymmetry between the response for an eastward and westward flow results from the nonlineawr curvature (centrifugal) terms. In general, the amplitude of an eastward flowing wind will be less after adjustment than a westward wind. For instance, if the initial wind velocity is 300 m/s, the linearized theory predicts a final wind speed of 240 m/s, regardless of the flow direction. However, the nonlinear curvature terms modify the response and produce a final wind speed of only 200 m/s for an initial eastward wind and a final wind speed of almost 300 m/s for an initial westward flow direction. Also, less gravity wave energy is produced by the adjustment of the westward flow than by the adjustment of the eastward flow. The implications are that the response of the thermosphere should be significantly different on the dawn and dusk sides of the auroral oval. Larger flow velocities would be expected on the dusk side since the plasma will accelerate the flow in a westward direction in that sector.

Larsen, M. F.↗

Variational symplectic algorithm for guiding center dynamics in the inner magnetosphere

Charged particle dynamics in magnetosphere has temporal and spatial multiscale; therefore, numerical accuracy over a long integration time is required. A variational symplectic integrator (VSI) [H. Qin and X. Guan, Phys. Rev. Lett. 100, 035006 (2008) and H. Qin, X. Guan, and W. M. Tang, Phys. Plasmas 16, 042510 (2009)] for the guiding-center motion of charged particles in general magnetic field is applied to study the dynamics of charged particles in magnetosphere. Instead of discretizing the differential equations of the guiding-center motion, the action of the guiding-center motion is discretized and minimized to obtain the iteration rules for advancing the dynamics. The VSI conserves exactly a discrete Lagrangian symplectic structure and has better numerical properties over a long integration time, compared with standard integrators, such as the standard and adaptive fourth order Runge-Kutta (RK4) methods. Applying the VSI method to guiding-center dynamics in the inner magnetosphere, we can accurately calculate the particles'orbits for an arbitrary long simulating time with good conservation property. When a time-independent convection and corotation electric field is considered, the VSI method can give the accurate single particle orbit, while the RK4 method gives an incorrect orbit due to its intrinsic error accumulation over a long integrating time.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Non-conformal attractor in boost-invariant plasmas

We study the dissipative evolution of (0+1)-dimensionally expanding media with Bjorken symmetry using the Boltzmann equation for massive particles in relaxation-time approximation. Breaking conformal symmetry by a mass induces a non-zero bulk viscous pressure in the medium. It is shown that even a small mass (in units of the local temperature) drastically modifies the well-known attractor for the shear Reynolds number previously observed in massless systems. For generic nonzero particle mass, neither the shear nor the bulk viscous pressure relax quickly to a non-equilibrium attractor; they approach the hydrodynamic limit only late, at small values of the inverse Reynolds numbers. Only the longitudinal pressure, which is a combination of thermal, shear and bulk viscous pressures, continues to show early approach to a far-off-equilibrium attractor, driven by the rapid longitudinal expansion at early times. Second-order dissipative hydrodynamics based on a gradient expansion around locally isotropic thermal equilibrium fails to reproduce this attractor.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Solving the Orszag–Tang vortex magnetohydrodynamics problem with physics-constrained convolutional neural networks

We study the 2D Orszag–Tang vortex magnetohydrodynamics (MHD) problem through the use of physics-constrained convolutional neural networks (PCNNs) for forecasting the density, ρ, and the magnetic field, B, as well as the prediction of B given the velocity field v of the fluid. In addition to translation equivariance from the convolutional architecture, other physics constraints were embedded: absence of magnetic monopoles, non-negativity of ρ, use of only relevant variables, and the periodic boundary conditions of the problem. The use of only relevant variables and the hard constraint of non-negative ρ were found to facilitate learning greatly. The divergenceless condition ∇·B=0 was implemented as a hard constraint up to machine precision through the use of a magnetic potential to define B=∇×A. Residual networks and data augmentation were also used to improve performance. This allowed for some of the residual models to function as surrogate models and provide reasonably accurate simulations. For the prediction task, the PCNNs were evaluated against a physics-informed neural network, which had the ideal MHD induction equation as a soft constraint. Several models were able to generate highly accurate fields, which are visually almost indistinguishable and have low mean squared error. Only methods with built-in hard constraints produced physical fields with ∇·B=0. The use of PCNNs for MHD has the potential to produce physically consistent real-time simulations to serve as virtual diagnostics in cases where inferences must be made with limited observables.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Weakly nonlinear Rayleigh-Taylor instability of a finite-thickness fluid layer

A weakly nonlinear (WN) model has been developed for the Rayleigh-Taylor instability of a finite-thickness incompressible fluid layer (slab). We derive the coupling evolution equations for perturbations on the (upper) “linearly stable” and (lower) “linearly unstable” interfaces of the slab. Expressions of temporal evolutions of the amplitudes of the perturbation first three harmonics on the upper and lower interfaces are obtained. The classical feedthrough (interface coupling) solution obtained by Taylor [Proc. R. Soc. London A 201, 192 (1950)] is readily recovered by the first-order results. Our third-order model can depict the WN perturbation growth and the saturation of linear (exponential) growth of the perturbation fundamental mode on both interfaces. The dependence of the WN perturbation growth and the slab distortion on the normalized layer thickness (kd) is analytically investigated via the third-order solutions. Comparison is made with Jacobs-Catton's formula [J. W. Jacobs and I. Catton, J. Fluid Mech. 187, 329 (1988)] of the position of the “linearly unstable” interface. Using a reduced formula, the saturation amplitude of linear growth of the perturbation fundamental mode is studied. It is found that the finite-thickness effects play a dominant role in the WN evolution of the slab, especially when kd < 1. Thus, it should be included in applications where the interface coupling effects are important, such as inertial confinement fusion implosions and supernova explosions.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Calculating electron cyclotron current drive stabilization of resistive tearing modes in a nonlinear magnetohydrodynamic model

A model which incorporates the effects of electron cyclotron current drive (ECCD) into the magnetohydrodynamic equations is implemented in the NIMROD code [C. R. Sovinec et al., J. Comput. Phys. 195, 355 (2004)] and used to investigate the effect of ECCD injection on the stability, growth, and dynamical behavior of magnetic islands associated with resistive tearing modes. In addition to qualitatively and quantitatively agreeing with numerical results obtained from the inclusion of localized ECCD deposition in static equilibrium solvers [A. Pletzer and F. W. Perkins, Phys. Plasmas 6, 1589 (1999)], predictions from the model further elaborate the role which rational surface motion plays in these results. The complete suppression of the (2,1) resistive tearing mode by ECCD is demonstrated and the relevant stabilization mechanism is determined. Consequences of the shifting of the mode rational surface in response to the injected current are explored, and the characteristic short-time responses of resistive tearing modes to spatial ECCD alignments which are stabilizing are also noted. We discuss the relevance of this work to the development of more comprehensive predictive models for ECCD-based mitigation and control of neoclassical tearing modes.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Nonlinear dynamics of small-scale Alfvén waves

Herein we study the nonlinear evolution of very oblique small-scale Alfvén waves with k ⊥ d i ≳1. At these scales, the waves become significantly compressive, unlike in magnetohydrodynamics, due to the Hall term in the equations. We demonstrate that when frequencies are small compared to the ion gyrofrequency and amplitudes are small compared to unity, no new nonlinear interaction appears due to the Hall term alone at the lowest non-trivial order, even when k ⊥ d i ~1. However, at the second non-trivial order, we discover that the Hall physics leads to a slow but resonant nonlinear interaction between co-propagating Alfvén waves, an inherently three-dimensional effect. Including the effects of finite temperature, finite frequency, and electron inertia, the two-fluid Alfvén wave also becomes dispersive once one or more of k ⊥ ρ s , k ⊥ d e , or k ∥ d i becomes significant: for oblique waves at low β as studied here, this can be at a much smaller scale than d i . We show that the timescale for one-dimensional steepening of two-fluid Alfvén waves is only significant at these smaller dispersive scales, and also derive an expression for the amplitude of driven harmonics of a primary wave. Importantly, both new effects are absent in gyrokinetics and other commonly used reduced two-fluid models. Our calculations have relevance for the interpretation of laboratory Alfvén wave experiments, as well as shedding light on the physics of turbulence in the solar corona and inner solar wind, where the dominant nonlinear interaction between counter-propagating waves is suppressed, allowing these new effects to become important.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Towards QCD-assisted hydrodynamics for heavy-ion collision phenomenology

Heavy-ion collisions are well described by a dynamical evolution with a long hydrodynamical phase. In this phase the properties of the strongly coupled quark-gluon plasma are reflected in the equation of state (EoS) and the transport coefficients, most prominently by the shear and bulk viscosity over entropy density ratios η/s( T ) and ς/s( T ), respectively. While the EoS is by now known to a high accuracy, the transport coefficients and in particular their temperature and density dependence are not well known from first-principle computations yet, as well as the possible influence they can have once used in hydrodynamical simulations. In this work, the most recent QCD-based parameters are provided as input to the MUSIC framework. A ratio η/s( T ) computed with a QCD based approach is used for the first time [1, 2]. The IP-Glasma model is used to describe the initial energy density distribution, and UrQMD for the dilute hadronic phase. Simulations are performed for Pb–Pb collisions at √ s NN = 2.76 TeV, for different centrality intervals. The resulting kinematic distributions of the particles produced in the collisions are compared to data from the LHC, for several experimental observables. Furthermore, the high precision of the experimental results and the broad variety of observables considered allow to critically verify the quality of the description based on first-principle input to the hydrodynamic evolution.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

On fast solid-body rotation of the solar core and differential (liquid-like) rotation of the solar surface

On the basis of a two-component (two-fluid) hydrodynamic model, it is shown that the probable phenomenon of solar core rotation with a velocity higher than the average velocity of global rotation of the Sun, discovered by the SOHO mission, can be related to fast solid-body rotation of the light hydrogen component of the solar plasma, which is caused by thermonuclear fusion of hydrogen into helium inside the hot dense solar core. Thermonuclear fusion of four protons into a helium nucleus (α-particle) creates a large free specific volume per unit particle due to the large difference between the densities of the solar plasma and nuclear matter. As a result, an efficient volumetric sink of one of the components of the solar substance—hydrogen—forms inside the solar core. Therefore, a steady-state radial proton flux converging to the center should exist inside the Sun, which maintains a constant concentration of hydrogen as it burns out in the solar core. It is demonstrated that such a converging flux of hydrogen plasma with the radial velocity v{sub r}(r) = −βr creates a convective, v{sub r}∂v{sub φ}/∂r, and a local Coriolis, v{sub r}v{sub φ}/r,φ nonlinear hydrodynamic forces in the solar plasma, rotating with the azimuthal velocity v{sub φ}. In the absence of dissipation, these forces should cause an exponential growth of the solid-body rotation velocity of the hydrogen component inside the solar core. However, friction between the hydrogen and helium components of the solar plasma due to Coulomb collisions of protons with α-particles results in a steady-state regime of rotation of the hydrogen component in the solar core with an angular velocity substantially exceeding the global rotational velocity of the Sun. It is suggested that the observed differential (liquid-like) rotation of the visible surface of the Sun (photosphere) with the maximum angular velocity at the equator is caused by sold-body rotation of the solar plasma in the radiation zone and strong turbulence in the tachocline layer, where the turbulent viscosity reaches its maximum value at the equator. There, the tachocline layer exerts the most efficient drag on the less dense outer layers of the solar plasma, which are slowed down due to the interaction with the ambient space plasma (solar wind).

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Hall effect in a strong magnetic field: Direct comparisons of compressible magnetohydrodynamics and the reduced Hall magnetohydrodynamic equations

In this work we numerically test a model of Hall magnetohydrodynamics in the presence of a strong mean magnetic field: the reduced Hall magnetohydrodynamic model (RHMHD) derived by [Gomez et al., Phys. Plasmas 15, 102303 (2008)] with the addition of weak compressible effects. The main advantage of this model lies in the reduction of computational cost. Nevertheless, up until now the degree of agreement with the original Hall MHD system and the range of validity in a regime of turbulence were not established. In this work direct numerical simulations of three-dimensional Hall MHD turbulence in the presence of a strong mean magnetic field are compared with simulations of the weak compressible RHMHD model. The results show that the degree of agreement is very high (when the different assumptions of RHMHD, such as spectral anisotropy, are satisfied). Nevertheless, when the initial conditions are isotropic but the mean magnetic field is maintained strong, the results differ at the beginning but asymptotically reach a good agreement at relatively short times. We also found evidence that the compressibility still plays a role in the dynamics of these systems, and the weak compressible RHMHD model is able to capture these effects. In conclusion the weak compressible RHMHD model is a valid approximation of the Hall MHD turbulence in the relevant physical context.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Is Quadrupole Structure of Out-of-Plane Magnetic Field Evidence for Hall Reconnection?

The experimental observation of a quadrupole pattern in the ''out-of-plane'' field B{sub y} is usually interpreted as a tell-tale sign of Hall MHD mediated magnetic reconnection in the vicinity of a reconnecting X-point. In contrast to these conventional expectations, it is shown B{sub y} means of analytical equations describing the generation of B{sub y} that quadrupole structure can be formed even in the absence of Hall effects. In general, B{sub y} can be generated due to inhomogeneous ion flow, inhomogeneous current (Hall effect), or terms proportional to a pre-existing B{sub y}. This prediction is validated in a 3D global MHD simulation where it is shown that vortex flows at the dayside magnetopause lead to formation of a finite B{sub y} even in the absence of a pre-existing IMF B{sub y}. The generated field is shown to have a quadrupolar pattern in B{sub y} with peak amplitude {approx}0.5 of the total field. Similar vortex flows are also observed in 3D global hybrid simulations but the resulting B{sub y} structure in this case is a combination of the Hall effects and the inhomogeneous ion flows. Thus special care must be taken in using observations of quadrupole B{sub y} in satellite and experiments to infer information about its particular generation mechanism or attributing it to the Hall MHD reconnection regime.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Nonlinear instability of an Oldroyd elastico–viscous magnetic nanofluid saturated in a porous medium

Through viscoelastic potential theory, a Kelvin-Helmholtz instability of two semi-infinite fluid layers, of Oldroydian viscoelastic magnetic nanofluids (MNF), is investigated. The system is saturated by porous medium through two semi-infinite fluid layers. The Oldroyd B model is utilized to describe the rheological behavior of viscoelastic MNF. The system is influenced by uniform oblique magnetic field that acts at the surface of separation. The model is used for the MNF incorporated the effects of uniform basic streaming and viscoelasticity. Therefore, a mathematical simplification must be considered. A linear stability analysis, based upon the normal modes analysis, is utilized to find out the solutions of the equations of motion. The onset criterion of stability is derived; analytically and graphs have been plotted by giving numerical values to the various parameters. These graphs depict the stability characteristics. Regions of stability and instability are identified and discussed in some depth. Some previous studies are recovered upon appropriate data choices. The stability criterion in case of ignoring the relaxation stress times is also derived. To relax the mathematical manipulation of the nonlinear approach, the linearity of the equations of motion is taken into account in correspondence with the nonlinear boundary conditions. Taylor's theory is adopted to expand the governing nonlinear characteristic equation according to of the multiple time scales technique. This analysis leads to the well-known Ginzburg–Landau equation, which governs the stability criteria. The stability criteria are achieved theoretically. To simplify the mathematical manipulation, a special case is considered to achieve the numerical estimations. The influence of orientation of the magnetic fields on the stability configuration, in linear as well as nonlinear approaches, makes a dual role for the magnetic field strength in the stability graphs. Stability diagram is plotted for different sets of physical parameters. A new stability as well as instability region, in the parameter space, appears due to the nonlinear effects.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Fokker-Planck simulations for core heating in subignition cone-guiding fast ignition targets

On the basis of two-dimensional coupled relativistic Fokker-Planck and fluid simulations, the core heating by fast electron beam in subignition cone-guiding fast ignition targets is investigated. It was found that the magnetic field due to the thermoelectric current reduces the beam pinching due to the resistive magnetic field. Compared with the carbon-deuterium fuel used in the early phase of Fast Ignition Realization Experiment, phase-I, the core heating of deuterium-tritium (DT) fuel used in the later phase is less effective. One reason is the slower temperature relaxation between bulk electron and ion for the DT case, which is previously reported by Johzaki et al. [Phys. Plasmas 15, 062702 (2008)]. In addition, in the lower-Z DT case, the pinching effects by the resistive magnetic field are weaker due to smaller resistivity; the optical thickness for fast electron beam is smaller, which leads to further reduction in core heating efficiency. It was also found that the fast electron beam quality deteriorates during the transport in the only 10 {mu}m thickness high-Z (e.g., Au and Cu) cone tip because of the collisional scattering and drag. As the result, the core heating efficiency considerably goes down compared to the case neglecting the transport in the cone tip.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

IONIC COMPOSITION STRUCTURE OF CORONAL MASS EJECTIONS IN AXISYMMETRIC MAGNETOHYDRODYNAMIC MODELS

We present the ionic charge state composition structure derived from axisymmetric MHD simulations of coronal mass ejections (CMEs), initiated via the flux-cancellation and magnetic breakout mechanisms. The flux-cancellation CME simulation is run on the Magnetohydrodynamics-on-A-Sphere code developed at Predictive Sciences, Inc., and the magnetic breakout CME simulation is run on ARC7 developed at NASA GSFC. Both MHD codes include field-aligned thermal conduction, radiative losses, and coronal heating terms which make the energy equations sufficient to calculate reasonable temperatures associated with the steady-state solar wind and model the eruptive flare heating during CME formation and eruption. We systematically track a grid of Lagrangian plasma parcels through the simulation data and calculate the coronal density and temperature history of the plasma in and around the CME magnetic flux ropes. The simulation data are then used to integrate the continuity equations for the ionic charge states of several heavy ion species under the assumption that they act as passive tracers in the MHD flow. We construct two-dimensional spatial distributions of commonly measured ionic charge state ratios in carbon, oxygen, silicon, and iron that are typically elevated in interplanetary coronal mass ejection (ICME) plasma. We find that the slower CME eruption has relatively enhanced ionic charge states and the faster CME eruption shows basically no enhancement in charge states-which is the opposite trend to what is seen in the in situ ICME observations. The primary cause of the difference in the ionic charge states in the two simulations is not due to the different CME initiation mechanisms per se. Rather, the difference lies in their respective implementation of the coronal heating which governs the steady-state solar wind, density and temperature profiles, the duration of the connectivity of the CME to the eruptive flare current sheet, and the contribution of the flare-heated plasma associated with the reconnection jet outflow into the ejecta. Despite the limitations inherent in the first attempt at this novel procedure, the simulation results provide strong evidence in support of the conclusion that enhanced heavy ion charge states within CMEs are a direct consequence of flare heating in the low corona. We also discuss future improvements through combining numerical CME modeling with quantitative ionic charge state calculations.

79 ASTRONOMY AND ASTROPHYSICS↗

Sparsified time-dependent Fourier neural operators for fusion simulations

This paper presents a sparsified Fourier neural operator for coupled time-dependent partial differential equations (ST-FNO) as an efficient machine learning surrogate for fluid and particle-based fusion codes such as NIMROD (Non-Ideal Magnetohydrodynamics with Rotation - Open Discussion) and GTC (Gyrokinetic Toroidal Code). ST-FNO leverages the structures in the governing equations and utilizes neural operators to represent Green's function-like numerical operators in the corresponding numerical solvers. Once trained, ST-FNO can rapidly and accurately predict dynamics in fusion devices compared with first-principle numerical algorithms. In general, ST-FNO represents an efficient and accurate machine learning surrogate for numerical simulators for multi-variable nonlinear time-dependent partial differential equations, with the proposed architectures and loss functions. The efficacy of ST-FNO has been demonstrated using quiescent H-mode simulation data from NIMROD and kink-mode simulation data from GTC. The ST-FNO H-mode results show orders of magnitude reduction in memory and central processing unit usage in comparison with the numerical solvers in NIMROD when computing fields over a selected poloidal plane. The ST-FNO kink-mode results achieve a factor of 2 reduction in the number of parameters compared to baseline FNO models without accuracy loss.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗