返回词条

弗拉索夫方程 · 历史版本

版本编辑者时间修改说明
第 4 版站方导入2026/11/9 06:54:10百科字段核查:fields.content
第 3 版站方导入2026/10/15 23:22:19按统一模板整理并保留字段来源
第 2 版站方导入2026/10/10 19:10:46根据原始来源更新词条
第 1 版站方导入2026/10/10 19:00:12根据原始来源更新词条

第 1 版

弗拉索夫方程

2.2. The gyrokinetic Vlasov equation

A key element of gyrokinetic theory is the gyrokinetic Vlasov equation which describes the time evolution of the gyrocenter distribution function Fσ(X,ν∥,μ,t)F _ { \sigma } ( \mathbf { X } , \nu _ { \parallel } , \mu , t ) for each particle species σ in the absence of collisions,

∂Fσ∂t+dXdt⋅∇Fσ+dv∥dt∂Fσ∂v∥=0.(2)\frac {\partial F _ {\sigma}}{\partial t} + \frac {d \mathbf {X}}{d t} \cdot \nabla F _ {\sigma} + \frac {d v _ {\parallel}}{d t} \frac {\partial F _ {\sigma}}{\partial v _ {\parallel}} = 0.\tag{2}

The corresponding equations of motion are [3]

dXdt=v∥b0+vE0+vE1+v∇B0+vc,(3)\frac {d \mathbf {X}}{d t} = v _ {\parallel} \mathbf {b} _ {0} + \mathbf {v} _ {E _ {0}} + \mathbf {v} _ {E _ {1}} + \mathbf {v} _ {\nabla B _ {0}} + \mathbf {v} _ {c},\tag{3} dv∥dt=−dX/dtmσv∥⋅(qσ∇(ϕ0+ϕˉ1)+μ∇B0),(4)\frac {d v _ {\parallel}}{d t} = - \frac {d \mathbf {X} / d t}{m _ {\sigma} v _ {\parallel}} \cdot \left(q _ {\sigma} \nabla \left(\phi_ {0} + \bar {\phi} _ {1}\right) + \mu \nabla B _ {0}\right),\tag{4} dμdt=0.(5)\frac {d \mu}{d t} = 0.\tag{5}

Here, X is the gyrocenter position, μ=mν | 2/(2B0)\mu = m \nu _ { \textrm { | } } ^ { 2 } / ( 2 B _ { 0 } ) is the magnetic moment, and ν∥\nu _ { \parallel } and ν⊥\nu _ { \perp } are the velocity components parallel and perpendicular to the background magnetic field. Electromagnetic terms have been neglected for clarity, as they are currently not yet implemented in GENE-3D. The E×B,\mathbf { E } \times \mathbf { B } , , grad-B, and curvature drift velocities appearing in Eq. (3) are defined as

vE0=cB02B0×∇ϕ0,(6)\mathbf {v} _ {E _ {0}} = \frac {c}{B _ {0} ^ {2}} \mathbf {B} _ {0} \times \nabla \phi_ {0},\tag{6} vE1=cB02B0×∇ϕˉ1,(7)\mathbf {v} _ {E _ {1}} = \frac {c}{B _ {0} ^ {2}} \mathbf {B} _ {0} \times \nabla \bar {\phi} _ {1},\tag{7} v∇B0=μcqσB02B0×∇B0,(8)\mathbf {v} _ {\nabla B _ {0}} = \frac {\mu c}{q _ {\sigma} B _ {0} ^ {2}} \mathbf {B} _ {0} \times \nabla B _ {0},\tag{8} vc=v∥2Ωσ(∇×b0)⊥.(9)\mathbf {v} _ {c} = \frac {v _ {\parallel} ^ {2}}{\Omega_ {\sigma}} (\nabla \times \mathbf {b} _ {0}) _ {\perp}.\tag{9}

The equilibrium electrostatic potential ϕ0\phi _ { 0 } appearing in Eq. (6) can be employed to consider externally imposed (longwavelength) radial electric field effects, which in stellarators is generally determined by neoclassical processes [35]. A gyroaverage operation is indicated with an overbar, defined as

ϕˉ1(X)=12π∮ϕ1(X+r(α))dα,(10)\bar {\phi} _ {1} (\mathbf {X}) = \frac {1}{2 \pi} \oint \phi_ {1} (\mathbf {X} + \mathbf {r} (\alpha)) d \alpha ,\tag{10}

for the electrostatic potential, where the gyroradius vector, r(α), is orthogonal to the local magnetic field.

In line with the gyrokinetic ordering, the gyrocenter distribution function is split into an equilibrium part F0σF _ { 0 \sigma } and a fluctuation F1σF _ { 1 \sigma }

Fσ=F0σ+F1σ,(11)F _ {\sigma} = F _ {0 \sigma} + F _ {1 \sigma},\tag{11}

where

F1σ/F0σ∼ϵδ≪1.(12)F _ {1 \sigma} / F _ {0 \sigma} \sim \epsilon_ {\delta} \ll 1.\tag{12}

This splitting procedure can be used to separate the “macroscopic” evolution of the plasma from the microturbulence, reducing the computational cost of the simulations employing this “delta-F” approach significantly in many situations [5]. A local Maxwellian is considered as the background distribution function in GENE-3D. Keeping only first-order terms in the perturbed distribution function, the resulting gyrokinetic Vlasov equation reads

∂F1σ∂t+v∥b^0⋅Γσ−b^0⋅μmσ∇B0∂F1σ∂v∥+(v∇B0+vc)⋅Γσ+(vE0+vE1)⋅∇F1σ+vE1⋅[∇F0σ+μ∇B0F0σT0σ]+(v∇B+vc)⋅∇F0σ=0,(13)\begin{array}{r l} & {\frac {\partial F _ {1 \sigma}}{\partial t} + v _ {\parallel} \hat {\mathbf {b}} _ {0} \cdot \Gamma_ {\sigma} - \hat {\mathbf {b}} _ {0} \cdot \frac {\mu}{m _ {\sigma}} \nabla B _ {0} \frac {\partial F _ {1 \sigma}}{\partial v _ {\parallel}} + (\mathbf {v} _ {\nabla B _ {0}} + \mathbf {v} _ {c}) \cdot \Gamma_ {\sigma} + (\mathbf {v} _ {E _ {0}} + \mathbf {v} _ {E _ {1}}) \cdot \nabla F _ {1 \sigma}} \\ & {\qquad + \mathbf {v} _ {E _ {1}} \cdot [ \nabla F _ {0 \sigma} + \mu \nabla B _ {0} \frac {F _ {0 \sigma}}{T _ {0 \sigma}} ] + (\mathbf {v} _ {\nabla B} + \mathbf {v} _ {c}) \cdot \nabla F _ {0 \sigma} = 0,} \end{array}\tag{13}

where, for simplicity, the abbreviation

Γσ=∇F1σ+qσT0σF0σ∇ϕˉ1(14)\Gamma_ {\sigma} = \nabla F _ {1 \sigma} + \frac {q _ {\sigma}}{T _ {0 \sigma}} F _ {0 \sigma} \nabla \bar {\phi} _ {1}\tag{14}

has been introduced. The last term in Eq. (13) couples neoclassical and turbulent transport; it may affect the long-term evolution of the system in the presence of collisions [36]. In the present paper, neoclassical contributions to transport are neglected for simplicity, and the E×BE \times B nonlinearity is the only nonlinear term kept in the current implementation of GENE-3D.

In the following, we will describe how collisional effects are incorporated in GENE-3D, and how the self-consistent electromagnetic fields are calculated.

2. Electromagnetic Vlasov-Darwin model

The Vlasov-Darwin equations for a collisionless electromagnetic plasma can be written as [7, 19, 22, 37–39]:

∂tfα+v⋅∇fα+qαmα(E+v×B)⋅∇vfα=0,(1)\partial_ {t} f _ {\alpha} + \mathbf {v} \cdot \nabla f _ {\alpha} + \frac {q _ {\alpha}}{m _ {\alpha}} (\mathbf {E} + \mathbf {v} \times \mathbf {B}) \cdot \nabla_ {v} f _ {\alpha} = 0,\tag{1} 1μ0∇2A+j−ϵ0∂t∇ϕ=0,(2)\frac {1}{\mu_ {0}} \nabla^ {2} \mathbf {A} + \mathbf {j} - \epsilon_ {0} \partial_ {t} \nabla \phi = 0,\tag{2} ϵ0∇2ϕ+ρ=0,(3)\epsilon_ {0} \nabla^ {2} \phi + \rho = 0,\tag{3}

where fα(r,v)f _ { \alpha } ( \mathbf { r } , \mathbf { v } ) is the particle distribution function of species α in phase space, qαq _ { \alpha } and mαm _ { \alpha } are the species charge and mass respectively, ϵ0\epsilon _ { \mathrm { 0 } } and μ0\mu _ { 0 } are the vacuum permittivity and permeability respectively, ϕ\phi and A are the t scalar and vector potentials, respectively. The electric and magnetic fields are defined uniquely from ϕ,\phi , A as:

E=−∇ϕ−∂tA;B=∇×A.(4)\mathbf {E} = - \nabla \phi - \partial_ {t} \mathbf {A}; \mathbf {B} = \nabla \times \mathbf {A}.\tag{4}

The set of Darwin equations is driven by the plasma current density

j=∑αqα∫fαvdv.(5)\mathbf {j} = \sum_ {\alpha} q _ {\alpha} \int f _ {\alpha} \mathbf {v} d \mathbf {v}.\tag{5}

Unlike Maxwell’s equations, the Darwin model does not feature Gauge invariance, and only the Coulomb gauge

∇⋅A=0(6)\nabla \cdot \mathbf {A} = 0\tag{6}

is physically acceptable (to enforce charge conservation).

Note that the Vlasov-Darwin model (Eqs. 1-6) is overdetermined, as it has more equations than unknowns. It features two involutions [40, 41]: Poisson’s equation and the solenoidal constraint of the vector potential. (An involution is a constraint satisfied by the solution of the system at all times, if satisfied initially.) These two involutions are not redundant, and must be enforced numerically [1, p. 359] to prevent spurious modes from being excited [42].

To solve the above Vlasov-Darwin equations, we begin by realizing that the two involutions do not need to be enforced explicitly. In stead of Poisson’s equation, we consider the equation:

ϵ0∂t∇2ϕ−∇⋅j=0,(7)\epsilon_ {0} \partial_ {t} \nabla^ {2} \phi - \nabla \cdot \mathbf {j} = 0,\tag{7}

found by taking the divergence of Eq. 2 and using Eq. 6. The two involutions are then implied by Eqs. 1, 2 and 77 when the local charge conservation equation,

∂tρ+∇⋅j=0,(8)\partial_ {t} \rho + \nabla \cdot \mathbf {j} = 0,\tag{8}

(which is implied independently by the Vlasov equation, Eq. 1) is satisfied. In particular, Poisson’s equation is implied by Eq. 3 and Eq. 8. The solenoidal constraint is implied as well, as can be seen by taking the divergence of Eq. 2 and using Eq. 3, to find:

∇⋅∇2A=0,\nabla \cdot \nabla^ {2} \mathbf {A} = 0,

from which, with appropriate conditions, Eq. 6 follows [42]. The derivation requires that ∇⋅(∇2A)=∇2(∇⋅A)=0\nabla \cdot ( \nabla ^ { 2 } \mathbf { A } ) = \nabla ^ { 2 } ( \nabla \cdot \mathbf { A } ) = 0 , and the boundary conditions be consistent with ∇⋅A=0\nabla \cdot \mathbf { A } = 0 at the boundary [43] (i.e., they must enforce continuity of the normal component of the vector potential at the boundary).

Equations 1, 2, 4, 5, and 7 constitute the minimal Vlasov-Darwin equation set of choice in this study. We emphasize that the main advantage of this set is that the two involutions (Poisson’s equation and the solenoidal constraint of A) are built-in, and thus do not need to be enforced or solved explicitly when local charge is strictly conserved. This property, when implemented discretely, will be most advantageous for both accuracy (it avoids spurious modes) and eficiency (it avoids the extra divergence-cleaning step via conventional projection methods [44] or hyperbolic cleaning [45]). Most importantly, this formulation avoids explicitly enforcing Eq. 6, which has been a critical implementation roadblock for the Darwin approximation in multiple dimensions in previous studies [7, 10]. Carrying the involution enforcement to the discrete will require a very careful discrete treatment, however, and in particular one that strictly conserves local charge.

2.1. The gyrokinetic Vlasov-Poisson system

The gyrokinetic Vlasov equation is derived from the Liouville theorem stating that the distribution function is constant along particle trajectories

dfσdt=∂fσ∂t+R˙⋅∇fσ+v˙∥∂fσ∂v∥=0.(7)\frac {\mathrm{d} f _ {\sigma}}{\mathrm{d} t} = \frac {\partial f _ {\sigma}}{\partial t} + \dot {\boldsymbol {R}} \cdot \nabla f _ {\sigma} + \dot {v} _ {\parallel} \frac {\partial f _ {\sigma}}{\partial v _ {\parallel}} = 0.\tag{7}

The trajectories of the gyrocenters are derived by solving the Euler-Lagrange equations [34]

R˙=v∥B∗B∥∗+cqσB∥∗b×(μ∇B+qσ∇ϕ),(8)\dot {\boldsymbol {R}} = v _ {\parallel} \frac {\boldsymbol {B} ^ {*}}{B _ {\parallel} ^ {*}} + \frac {c}{q _ {\sigma} B _ {\parallel} ^ {*}} \boldsymbol {b} \times (\mu \nabla B + q _ {\sigma} \nabla \phi),\tag{8} v˙∥=−B∗mσB∥∗⋅(μ∇B+qσ∇ϕ).(9)\dot {v} _ {\parallel} = - \frac {\pmb {B} ^ {*}}{m _ {\sigma} B _ {\parallel} ^ {*}} \cdot (\mu \nabla B + q _ {\sigma} \nabla \phi) .\tag{9}

The quasi-neutrality equation is derived by requiring that the variation of the action functional, generated by the Lagrangian, with respect to electrostatic potential vanishes. Explicit calculation yields [34]

−∇⊥⋅(∑σmσc2n0σB2∇⊥ϕ)=∑σqσ∫fσdW,(10)- \nabla_ {\perp} \cdot \left(\sum_ {\sigma} \frac {m _ {\sigma} c ^ {2} n _ {0 \sigma}}{B ^ {2}} \nabla_ {\perp} \phi\right) = \sum_ {\sigma} q _ {\sigma} \int f _ {\sigma} d W,\tag{10}

where n0σn _ { 0 \sigma } is the initial density given by the Maxwellian distribution function f0σf _ { 0 \sigma } . Eqs. (7) - (10) describe the system solved in the code. Due to the derivation from a Lagrangian the system fulfills an energy conservation law. The conserved energy is [34]

E=∑σ∫(12mσv∥2+μB+12qσϕ)fσdVdW.(11)E = \sum_ {\sigma} \int \left(\frac {1}{2} m _ {\sigma} v _ {\parallel} ^ {2} + \mu B + \frac {1}{2} q _ {\sigma} \phi\right) f _ {\sigma} d V d W.\tag{11}

After having described the gyrokinetic model and the energy theorem we continue with deriving the coordinate system for our simulations in the next section.

参考资料