Source-linked AI summary
Explicit charge-conserving and modified energy-conserving PIC methods using the SAV approach
Chengyu Fan, Nicolas Crouseilles, Jingwei Hu
TL;DR
In long-time PIC simulations, violated energy conservation can allow numerical errors and particle noise to accumulate into grid instability, while violated charge conservation can cause field oscillations and filamentary structures [5]. The paper develops explicit second-order SAV-based PIC schemes for both systems, including predictor-based and symmetric-splitting schemes using a scaled electric field. The schemes preserve charge and a total modified energy in the Vlasov–Ampère and Vlasov–Maxwell systems.
Problem
Method
The paper develops explicit second-order SAV-based PIC schemes for both systems, including predictor-based and symmetric-splitting schemes using a scaled electric field.
Source: §1 · p. 2Results
The schemes preserve charge and a total modified energy in the Vlasov–Ampère and Vlasov–Maxwell systems.
Source: §5 · p. 23Takeaways & Limitations
Using an auxiliary scalar to enforce modified-energy balance avoids the imaginary correction factors and problematic particles associated with the velocity-correction strategy [5].
Source: §1 · p. 2For the Vlasov–Ampère system, the manuscript presents VA-I and VA-II, while noting that other time discretizations may be derived from the SAV semi-discretization.
Source: §2.3.4 · p. 10
Abstract
from arXiv · showhide
We develop explicit second-order particle-in-cell (PIC) methods for the Vlasov--Amp\`ere and Vlasov--Maxwell systems using the scalar auxiliary variable (SAV) approach. Two schemes are proposed for each system: a predictor-based scheme and a symmetric splitting scheme using a scaled electric field. The methods exactly preserve the discrete Gauss law and a modified total energy. Numerical experiments for electrostatic and electromagnetic problems demonstrate the expected dynamics as well as the conservation properties of both invariants.
1 Introduction
Energy and charge conservation are important for reliable PIC simulations, but explicit methods face challenges in preserving both. This paper uses the SAV approach to develop explicit second-order PIC schemes for the Vlasov–Ampère and Vlasov–Maxwell systems that conserve charge and modified energy.
1 Introduction
- Energy or charge conservation violations can cause unphysical energy growth, field oscillations, or filamentary structures in long-time PIC simulations [5].For electromagnetic particle solvers, failure to preserve Gauss’s law can also produce spurious virtual charges,.Source: §1 · p. 1
- Explicit PIC methods have historically often conserved momentum but not energy, while later energy-conserving methods were fully or semi-implicit [14] [15] [16].The cited explicit velocity-correction method [5] may produce problematic particles because of imaginary correction factors.Source: §1 · pp. 1–2
- Charge-conserving approaches include electric-field corrections and the Villasenor-Buneman method, which modifies the shape function in the Ampère source term to match the charge update.Later extensions accommodate more shape functions and improve efficiency.Source: §1 · p. 2
- Prior SAV work reformulated free energy into a quadratic form with an auxiliary scalar, and its Vlasov-system adaptation enabled explicit schemes conserving charge and modified total energy.The scalar adds a degree of freedom to the system.Source: §1 · p. 2
- This paper develops explicit second-order SAV-based PIC schemes for both systems, using a predictor-based scheme and a symmetric splitting scheme with a scaled electric field.Unlike the local velocity-correction strategy [5], the modified-energy balance is enforced through an auxiliary scalar, avoiding imaginary correction factors and problematic particles.Source: §1 · p. 2
2 Electrostatic Vlasov-Amp`ere system
This section presents the 1D1V Vlasov–Ampère system and SAV-based particle schemes designed to preserve charge and modified energy. It describes a predictor scheme and a scaled-electric-field splitting scheme.
2.1 SAV formulation and the particle method
- The SAV particle formulation introduces an auxiliary scalar and represents the distribution function with particles coupled to spatial-grid fields.The formulation chooses a constant C0 so the modified-energy expression is positive and defines particle and grid quantities through the particle-mesh coupling.Source: §2.1 · pp. 3–4
2.2 Scheme VA-I
- VA-I is a predictor-based, second-order time scheme for the SAV-reformulated system.The Villasenor–Buneman-style current used in its electric-field update is justified as a second-order approximation to the exact current.Source: §2.2 · pp. 4–5
- VA-I preserves the total modified energy and maintains Gauss law from one time step to the next.Its electric-field update uses a Villasenor–Buneman-style current for charge conservation, while the scalar-variable update uses a standard particle-current approximation.Source: §2.2 · pp. 5–6
2.3 Scheme VA-II
- VA-II reformulates the system using a scaled electric field and advances it with a second-order time-splitting scheme.The full scheme preserves charge through the same electric-field update as VA-I and preserves total modified energy.Source: §2.3 · p. 7, §2.3.4 · p. 9
2.3.2 Second subsystem: φ[2]
- In the second subsystem, all components of the scaled field share a scalar growth rate, allowing that subsystem to be solved exactly..Source: §2.3.2 · p. 8
2.3.4 Full scheme
- The full VA-II scheme uses a Strang composition for second-order accuracy and preserves charge and modified total energy.The scheme uses the discrete current in φ for charge preservation and a midpoint method in φ for modified-energy preservation.Source: §2.3.4 · pp. 9–10
3 Electromagnetic Vlasov-Maxwell system
The section formulates the electromagnetic Vlasov–Maxwell system in 1D2V and specifies initial conditions under which its equations propagate Gauss’s law. It motivates an SAV reformulation to address the challenge of preserving charge and total energy in the discretized system.
3 Electromagnetic Vlasov-Maxwell system
- The 1D2V VM system evolves a distribution over position and two velocity components under electric and magnetic forces.The kinetic equation includes the longitudinal and transverse electric fields and the magnetic field.Source: §3 · p. 10
- With the initial longitudinal electric field constrained by Gauss’s law, the system equations propagate Gauss’s law at later times.The stated setting assumes periodic spatial boundaries and vanishing velocity boundaries.Source: §3 · p. 11
- The section identifies simultaneous preservation of total energy and charge as a discretization challenge and introduces SAV reformulation toward that goal.Source: §3 · p. 11
3.1 SAV formulation and the particle method
The section introduces two SAVs for the VM reformulation and describes its particle approximation on a spatial mesh. The mesh places field and charge/current quantities on specified Yee-lattice locations and uses B-spline particle-mesh coupling.
3.1 SAV formulation and the particle method
- The VM reformulation introduces two scalar auxiliary variables.Source: §3.1 · p. 11
- The reformulation includes a constant C0 chosen to keep the right-hand side positive.Source: §3.1 · p. 11
- The VM kinetic equation is approximated with particles and a uniform spatial mesh using B-spline particle-mesh coupling.On the Yee lattice, Ex, Jx, and B are at cell centers, while Ey, Jy, and charge density are at cell interfaces.Source: §3.1 · pp. 11–12
3.2 Scheme VM-I
VM-I uses predictors in a second-order scheme and reformulates its partly implicit electromagnetic field equations as a linear system. The scheme preserves total modified energy and Gauss’s law.
3.2 Scheme VM-I
- Predictors are used to obtain a second-order VM-I scheme.Source: §3.2 · p. 12
- Although VM-I’s electromagnetic field equations are not completely explicit, they can be reformulated as a linear system.The passage states that this system can be solved at the cost of an explicit method and that its implicitness ensures unconditional stability of the field solver.Source: §3.2 · p. 13
- VM-I preserves the total modified energy and Gauss’s law.Source: §3.2 · pp. 13–14
3.3 Scheme VM-II
VM-II uses scaled electric fields and a symmetric second-order splitting of the VM system into subflows. The full scheme preserves modified energy and Gauss’s law, with the stated extension to 3D3V and compatible current deposition.
3.3 Scheme VM-II
- VM-II scales each electric-field component by its SAV and splits the reformulated system into four subflows.A symmetric second-order composition combines the subsystem updates.Source: §3.3 · p. 15, §3.3.5 · p. 18
3.3.2 Second subsystem: φ[2]
- The second subsystem updates the scaled fields gx and gy.Source: §3.3.2 · p. 16
3.3.3 Third subsystem: φ[3]
- The third subsystem calculates its corresponding split update.Source: §3.3.3 · p. 17
3.3.4 Fourth subsystem: φ[4]
- The fourth subsystem uses midpoint quantities to update the auxiliary variable and particle velocities.Its field update is expressed through G and the scaled field, with a linear-system step also described.Source: §3.3.4 · pp. 17–18
3.3.5 Full scheme
- The full VM-II scheme preserves total modified energy and Gauss’s law.Source: §3.3.5 · p. 19
- The VM-II formulation extends to 3D3V with one SAV per electric-field component and the same symmetric splitting structure.Source: §3.3.5 · pp. 19–20
- A suitable current deposition, including Esirkepov’s method, is used to satisfy the discrete Gauss law on a Yee lattice.The stated discrete-operator identity makes the divergence of the magnetic curl vanish, propagating Gauss’s law through the Ampère update.Source: §3.3.5 · pp. 19–20
4 Numerical results
The VA schemes reproduce expected Landau damping and growth while preserving modified energy and Gauss law to machine precision; both VM schemes also match the expected growth rate and preserve these invariants. In the VA two-stream test, removing charge conservation destabilizes VA-I during the nonlinear phase.
4.1 The Vlasov-Amp`ere system
- Both VA schemes match the theoretical Landau damping and growth rates, with modified total energy and Gauss-law errors remaining at machine precision.VA-II runs roughly twice as long as VA-I.Source: §4.1 · p. 21
- Both VA schemes show second-order temporal convergence in maximum total-energy error, but spatial or particle-approximation errors can obscure the rate at small time steps.The original total energy is not preserved, so its error reflects temporal discretization error.Source: §4.1 · p. 21, Figure 2 · p. 23
- VA-I without charge conservation develops nonlinear instability, larger Gauss-law error, and larger maximum momentum error; VA-II without it retains reasonable electric-energy behavior.All four schemes initially follow the theoretical electric-energy growth rate γ = 0.5.Source: §4.1 · p. 21
4.2 The Vlasov-Maxwell system
- Both VM schemes match the theoretical growth rate and preserve modified total energy and Gauss law to machine precision.VM-II runs roughly twice as long as VM-I.Source: §4.2 · p. 23
5 Conclusions and future work
The paper develops two explicit second-order PIC schemes each for the Vlasov–Amp`ere and Vlasov–Maxwell systems, preserving discrete charge and modified total energy. The schemes are free from the Maxwell CFL condition, and further comparisons among SAV schemes are left for future work.
5 Conclusions and future work
- The proposed VA and VM schemes are second order in time, linearly implicit in fields, explicit in particles, and free from the Maxwell CFL condition.They preserve discrete charge and modified total energy exactly.Source: §5 · p. 23
- Each system has two schemes: one uses an additional scalar-variable update, while the other includes the scalar variable in the pusher and solves the electric field afterward.The first scheme is inspired by [5].Source: §5 · p. 23
- Future work will further investigate differences among these SAV schemes and compare them with other SAV schemes.Source: §5 · p. 23