Introduction
Modeling the boundary region of magnetically confined fusion plasmas remains a challenge for the fusion community. Fluctuations at all scales need to be considered, as well as kinetic effects and collisional processes. Developing a model for simulating the plasma boundary requires overcoming the limitations of existing models. Full-f fluid codes focus on the long-wavelength limit [32 , 14 ] , and their validity is limited to the highly collisional regime. Gyrofluid codes apply Padé approximation to relax the long-wavelength approximation [17 , 18 ] . Additionally, they use kinetic closures to avoid considering the highly collisional limit, although the success of these closures is highly case dependent [3 , 40 , 41 ] . On the other hand, gyrokinetic (GK) δ \delta δ -f models, which offer a good description of the core of fusion plasmas [29 , 21 ] , present the challenge of including collisions at the level of relevance for the boundary. Furthermore, their application to the boundary is questionable due to large-scale fluctuations present in this region. Full-f GK codes often focus on the long-wavelength limit [5 , 7 , 16 , 26 , 39 ] and thus compromise physical fidelity at the cost of computational performance.
To overcome the limitations of these approaches and accurately describe the boundary, [36 , 15 , 9 ] , propose a model that couples a DK and GK ordering. Here, we leverage the work of Frei et al. [9 ] and present a first implementation of a model that couples DK and GK orderings. More precisely, this work relies on splitting the distribution functions into a drift-kinetic (DK) large-scale slowly-varying part and a GK small-scale rapidly-varying part. The DK fields are allowed to be large in amplitude but only vary on long scale, employing a long-wavelength limit, while the GK fields are small in amplitude but allowed to vary on all scales.
The model is based on single particle equations of motion derived from first principles starting from the particle one-form. From the single particle equations of motion, the gyro-averaged Boltzmann equation is derived, split into a GK part and a DK part, and projected onto a Hermite-Laguerre basis giving a DK and a GK hierarchy equation for the projection coefficients, denoted as gyromoments. The Hermite-Laguerre projection allows for an efficient implementation of the plasma dynamics, including multiple different collision operators. Finally, we split the electrostatic potential into a DK part and a GK part, similarly to the distribution functions, and we derive a DK and GK Poisson equation from first-principles by using a variational principle.
Previous δ \delta δ -f GK gyromoment simulations based on the Hermite-Laguerre projection are shown to efficiently reduce the computational complexity of solving the GK system of equations compared to continuum δ \delta δ -f f f GK simulations Hoffmann et al. [20 ] ; Mandell et al. [24 ] . Similarly, full-f f f gyromoment simulations were performed in a linear plasma device, within a DK ordering Frei et al. [10 ] ; Mencke & Ricci [25 ] . This work combines the advantages of the two previously developed models allowing for order-one DK fluctuations with small wavenumbers and, at the same time, small-amplitude GK fluctuations with arbitrary wavenumbers. It then enables the study of instabilities on all scales while still removing the single-particle gyromotion from the description.
Although our approach ultimately is intended for the boundary of magnetically confined fusion devices, we consider the simpler case of a linear plasma device in this work. A linear plasma device offers an ideal testbed for new global turbulent plasma simulations. The absence of drifts associated with the curvature of the magnetic field, together with the smaller size of the simulations compared to toroidal devices reduces the complexity and cost of the simulations and allows for a deeper understanding of the physics at play Friedman et al. [11 ] . We note that, despite their simple geometries, previous simulations of linear plasma devices show complex turbulent behavior Shi et al. [38 ] ; Pan et al. [29 , 30 ] ; Popovich et al. [34 ] ; Rogers & Ricci [37 ] ; Frei et al. [10 ] ; Mencke & Ricci [25 ] . Finally, the perpendicular incidence of the magnetic field with the walls at the end of the plasma column simplify the boundary conditions of the simulations.
This paper is structured as follows. In section § 2 we present the ordering used as well as the equations describing the evolution of the distribution functions. The DK model for the ions and electrons is derived in § 3 from a kinetic description, and similarly, the GK model is derived in § 4 . We close the system in § 5 , by deriving the DK and GK Poisson equations used to solve for the two electrostatic potentials. The numerical implementation of the physical model is described in § 6 . In § 7 we report the results of the comprehensive turbulent simulations comparing it with a full-f f f DK simulation where the GK perturbations are suppressed. § 8 provides an analysis of the findings in § 7 by presenting linear investigations of the turbulence-driving instabilities. Finally, the conclusions follow in § 9 .
Ordering assumptions and equations of motion
We focus on the case of a fully-ionized plasma with one single-ionized ion species such that the plasma is composed of species a = i , e a=i,e a = i , e with e e e denoting the electrons and i i i the only ion species with charge q i = e q_i=e q i = e . We split the electrostatic potential
ϕ ( x ) = ϕ D K ( x ) + ϕ G K ( x ) , (1) \phi(\mathbf{x})=\phi_{DK}(\mathbf{x})+\phi_{GK}(\mathbf{x}), \tag*{(1)} ϕ ( x ) = ϕ D K ( x ) + ϕ G K ( x ) , (1)
where x \mathbf{x} x is the particle position and we impose that the GK part of the electrostatic potential has a small amplitude compared to the DK part, ϕ G K / ϕ D K ∼ ϵ δ < 1 \phi_{GK}/\phi_{DK}\sim\epsilon_\delta<1 ϕ G K / ϕ D K ∼ ϵ δ < 1 . We order the DK and GK quantities such that they vary on different length scales. More precisely, we order the perpendicular gradient lengths, k ⊥ D K ∼ ∣ ∇ ⊥ ln F a D K ∣ ∼ ∣ ∇ ⊥ ln ϕ D K ∣ ∼ ϵ ⊥ / ρ s k_{\perp DK}\sim\lvert\nabla_\perp\ln F_{aDK}\rvert\sim\lvert\nabla_\perp\ln\phi_{DK}\rvert\sim\epsilon_\perp/\rho_s k ⊥ D K ∼ ∣ ∇ ⊥ ln F a D K ∣ ∼ ∣ ∇ ⊥ ln ϕ D K ∣ ∼ ϵ ⊥ / ρ s with ϵ ⊥ < 1 \epsilon_\perp<1 ϵ ⊥ < 1 and k ⊥ G K ∼ ∣ ∇ ⊥ ln ϕ G K ∣ ∼ 1 / ρ s k_{\perp GK}\sim\lvert\nabla_\perp\ln\phi_{GK}\rvert\sim1/\rho_s k ⊥ G K ∼ ∣ ∇ ⊥ ln ϕ G K ∣ ∼ 1/ ρ s , being ρ s = c s / Ω i \rho_s=c_s/\Omega_i ρ s = c s / Ω i the ion sound Larmor radius with c s = T e / m i c_s=\sqrt{T_e/m_i} c s = T e / m i the sound speed, T e T_e T e the electron temperature, and Ω i = q i B / m i \Omega_i=q_iB/m_i Ω i = q i B / m i the ion gyro frequency, with B B B the magnetic field strength. Furthermore, we order the parallel length scale as k ∥ ∼ ϵ k ⊥ D K k_\parallel\sim\epsilon k_{\perp DK} k ∥ ∼ ϵ k ⊥ D K and the frequency as ω ∼ Ω i ϵ ϵ ⊥ \omega\sim\Omega_i\epsilon\epsilon_\perp ω ∼ Ω i ϵ ϵ ⊥ . This ordering imposes that ϕ D K \phi_{DK} ϕ D K gives a contribution to the E × B E\times B E × B velocity on the order of ϵ ⊥ / ϵ δ \epsilon_\perp/\epsilon_\delta ϵ ⊥ / ϵ δ with respect to that of ϕ G K \phi_{GK} ϕ G K Frei et al. [9 ] . By further ordering ϵ ⊥ ∼ ϵ δ \epsilon_\perp\sim\epsilon_\delta ϵ ⊥ ∼ ϵ δ , we impose that the E × B E\times B E × B drift from the DK electrostatic potential contributes to the drifts, comparably to the GK counterpart.
Based on this ordering, we derive the single-particle dynamics in § 2.1 . The behavior of the full distribution function is described in § 2.2 using the Boltzmann equation. Finally, the evolution of the DK and GK parts of the distribution function, is derived from the Boltzmann equation in § 2.3 .
Single-particle dynamics
We assume that the collisionless particle dynamics be governed by the gyrocenter single-particle Lagrangian, Γ A \Gamma_A Γ A . By using a Lie approach and including all terms up to second order in the GK and DK parts of the E × B E\times B E × B drift (i.e., to order c s 2 ϵ ⊥ 2 c_s^2\epsilon_\perp^2 c s 2 ϵ ⊥ 2 and c s 2 ϵ δ 2 c_s^2\epsilon_\delta^2 c s 2 ϵ δ 2 ), while assuming a straight and constant magnetic field, we have Frei et al. [9 ]
Γ a ( R , μ , v ∥ , t ) = q a A a ∗ ⋅ R ˙ − μ B Ω i θ ˙ − m a v ∥ 2 / 2 − q a Φ a ∗ + O ( m a m i T a ϵ ⊥ 3 , m a m i T a ϵ δ 3 ) , (2) \Gamma_a(\mathbf{R},\mu,v_\parallel,t)=q_a\mathbf{A}_a^*\cdot\dot{\mathbf{R}}-\frac{\mu B}{\Omega_i}\dot{\theta}-m_av_\parallel^2/2-q_a\Phi_a^*+O\left(\frac{m_a}{m_i}T_a\epsilon_\perp^3,\frac{m_a}{m_i}T_a\epsilon_\delta^3\right), \tag*{(2)} Γ a ( R , μ , v ∥ , t ) = q a A a ∗ ⋅ R ˙ − Ω i μ B θ ˙ − m a v ∥ 2 /2 − q a Φ a ∗ + O ( m i m a T a ϵ ⊥ 3 , m i m a T a ϵ δ 3 ) , (2)
where v ∥ = v ⋅ b v_\parallel=\mathbf{v}\cdot\mathbf{b} v ∥ = v ⋅ b is the velocity component parallel to the magnetic field, b \mathbf{b} b being the unit vector in the direction of the magnetic field, μ = m a v ⊥ 2 / B \mu=m_av_\perp^2/B μ = m a v ⊥ 2 / B is the adiabatic invariant, m a m_a m a is the mass of species a a a , and v ⊥ = ∣ v − b v ∥ ∣ v_\perp=\lvert\mathbf{v}-\mathbf{b}v_\parallel\rvert v ⊥ = ∣ v − b v ∥ ∣ is the velocity component perpendicular to the magnetic field. We also define the fields q a Φ a ∗ = q a ϕ D K + μ B + q a Φ E + q a ⟨ ψ ⟩ R q_a\Phi_a^*=q_a\phi_{DK}+\mu B+q_a\Phi_E+q_a\langle\psi\rangle_{\mathbf{R}} q a Φ a ∗ = q a ϕ D K + μ B + q a Φ E + q a ⟨ ψ ⟩ R and q a A a ∗ = q a A + m a v ∥ b + m a u E q_a\mathbf{A}_a^*=q_a\mathbf{A}+m_av_\parallel\mathbf{b}+m_a\mathbf{u}_E q a A a ∗ = q a A + m a v ∥ b + m a u E , where q a Φ E = m a ∣ u E ∣ 2 / 2 + μ B ∇ ⊥ 2 ϕ D K / ( 2 B Ω a ) q_a\Phi_E=m_a\lvert\mathbf{u}_E\rvert^2/2+\mu B\nabla_\perp^2\phi_{DK}/(2B\Omega_a) q a Φ E = m a ∣ u E ∣ 2 /2 + μ B ∇ ⊥ 2 ϕ D K / ( 2 B Ω a ) and u E = − ∇ ϕ D K × B / B 2 \mathbf{u}_E=-\nabla\phi_{DK}\times\mathbf{B}/B^2 u E = − ∇ ϕ D K × B / B 2 is the E × B E\times B E × B velocity for the DK potential. The effective GK gyro-averaged potential is defined as
⟨ ψ ⟩ R = ⟨ ϕ G K ⟩ R + q a 2 2 m a Ω a ∂ μ ( ⟨ ϕ G K 2 ⟩ R − ⟨ ϕ G K ⟩ R 2 ) + q a 2 m a Ω a 2 ⟨ ( b × ∇ ϕ ~ G K ) ⋅ ∇ ϕ ~ G K ⟩ R (3) \langle\psi\rangle_R = \langle\phi_{GK} \rangle_R + \frac{q_a^2}{2m_a\Omega_a}\partial_\mu\left(\langle\phi_{GK}^2 \rangle_R-\langle\phi_{GK}\rangle_R^2\right)+\frac{q_a}{2m_a\Omega_a^2}\left\langle\left(\mathbf{b}\times\nabla\widetilde{\phi}_{GK}\right)\cdot\nabla\widetilde{\phi}_{GK}\right\rangle_R \tag*{(3)} ⟨ ψ ⟩ R = ⟨ ϕ G K ⟩ R + 2 m a Ω a q a 2 ∂ μ ( ⟨ ϕ G K 2 ⟩ R − ⟨ ϕ G K ⟩ R 2 ) + 2 m a Ω a 2 q a ⟨ ( b × ∇ ϕ G K ) ⋅ ∇ ϕ G K ⟩ R (3)
where the gyroaverage operator is introduced,
⟨ f ⟩ R = 1 2 π ∫ 0 2 π d θ ∫ d x δ ( R + ρ − x ) f ( x , μ , v ∥ , t ) (4) \langle f\rangle_R=\frac{1}{2\pi}\int_0^{2\pi}d\theta\int d\mathbf{x}\,\delta(\mathbf{R}+\boldsymbol{\rho}-\mathbf{x})f(\mathbf{x},\mu,v_\parallel,t) \tag*{(4)} ⟨ f ⟩ R = 2 π 1 ∫ 0 2 π d θ ∫ d x δ ( R + ρ − x ) f ( x , μ , v ∥ , t ) (4)
= 1 2 π ∫ 0 2 π d θ ∑ n = 0 ∞ ( ρ a ⋅ ∇ ) n n ! f ( R , μ , v ∥ , t ) (5) =\frac{1}{2\pi}\int_0^{2\pi}d\theta\sum_{n=0}^{\infty}\frac{(\boldsymbol{\rho}_a\cdot\nabla)^n}{n!}f(\mathbf{R},\mu,v_\parallel,t) \tag*{(5)} = 2 π 1 ∫ 0 2 π d θ n = 0 ∑ ∞ n ! ( ρ a ⋅ ∇ ) n f ( R , μ , v ∥ , t ) (5)
= ∑ n = 0 ∞ ( 2 μ q a Ω a ) n ∇ 2 n 2 n n ! n ! f ( R , μ , v ∥ , t ) , (6) =\sum_{n=0}^{\infty}\left(\frac{2\mu}{q_a\Omega_a}\right)^n\frac{\nabla^{2n}}{2^n n!n!}f(\mathbf{R},\mu,v_\parallel,t), \tag*{(6)} = n = 0 ∑ ∞ ( q a Ω a 2 μ ) n 2 n n ! n ! ∇ 2 n f ( R , μ , v ∥ , t ) , (6)
with δ ( ⋅ ) \delta(\cdot) δ ( ⋅ ) the Dirac delta function and ρ a = a v ⊥ / Ω a \boldsymbol{\rho}_a=a\mathbf{v}_\perp/\Omega_a ρ a = a v ⊥ / Ω a , the Larmor radius, being a a unit vector perpendicular to both the magnetic field and the gyromotion of the particle Frei et al. [9 ] . Finally, we define
ϕ ~ G K = ϕ G K − ⟨ ϕ G K ⟩ R , (7) \widetilde{\phi}_{GK}=\phi_{GK}-\langle\phi_{GK}\rangle_R, \tag*{(7)} ϕ G K = ϕ G K − ⟨ ϕ G K ⟩ R , (7)
and
ϕ ‾ G K = ∫ 0 θ d θ ′ ϕ ~ G K ( x ( R , θ ′ ) , μ , v ∥ , t ) , (8) \overline{\phi}_{GK}=\int_0^\theta d\theta'\,\widetilde{\phi}_{GK}\left(\mathbf{x}(\mathbf{R},\theta'),\mu,v_\parallel,t\right), \tag*{(8)} ϕ G K = ∫ 0 θ d θ ′ ϕ G K ( x ( R , θ ′ ) , μ , v ∥ , t ) , (8)
in Eq. (3) .
Collective behavior
The collective behavior of the distribution function is governed by the gyroaveraged Boltzmann equation Frei et al. [9 ] ,
∂ t F a + ∇ ⋅ ( R ˙ F a ) + v ˙ ∥ ∂ v ∥ F a = ⟨ S a ⟩ R + ∑ b = i , e ⟨ C a b ⟩ R , (9) \partial_t F_a+\nabla\cdot\left(\dot{\mathbf{R}}F_a\right)+\dot{v}_\parallel\partial_{v_\parallel}F_a=\langle S_a\rangle_R+\sum_{b=i,e}\langle C_{ab}\rangle_R, \tag*{(9)} ∂ t F a + ∇ ⋅ ( R ˙ F a ) + v ˙ ∥ ∂ v ∥ F a = ⟨ S a ⟩ R + b = i , e ∑ ⟨ C ab ⟩ R , (9)
where the sources for species a a a , S a S_a S a , are introduced along with the collision operator between species a a a and b b b , C a b C_{ab} C ab , the parallel acceleration, v ˙ ∥ \dot{v}_\parallel v ˙ ∥ , and gyrocenter drift, R ˙ \dot{\mathbf{R}} R ˙ .
The Euler-Lagrange equations of the ion Lagrangian, Eq. (2) , give
v ∥ = b ⋅ R ˙ , (10) v_\parallel=\mathbf{b}\cdot\dot{\mathbf{R}}, \tag*{(10)} v ∥ = b ⋅ R ˙ , (10)
μ ˙ = 0 , (11) \dot{\mu}=0, \tag*{(11)} μ ˙ = 0 , (11)
− q a B ∗ × R ˙ − m a v ˙ ∥ b = q a ∇ Φ a ∗ + q a ∂ t A ∗ , (12) -q_a\mathbf{B}^{*}\times\dot{\mathbf{R}}-m_a\dot{v}_\parallel\mathbf{b}=q_a\nabla\Phi_a^{*}+q_a\partial_t\mathbf{A}^{*}, \tag*{(12)} − q a B ∗ × R ˙ − m a v ˙ ∥ b = q a ∇ Φ a ∗ + q a ∂ t A ∗ , (12)
introducing the effective magnetic field, B ∗ = ∇ × A ∗ \mathbf{B}^{*}=\nabla\times\mathbf{A}^{*} B ∗ = ∇ × A ∗ . Now, projecting Eq. (2.7c) on the B ∗ \mathbf{B}^{*} B ∗ direction and crossing with b \mathbf{b} b , we obtain respectively
v ˙ ∥ B ∗ = − q a m a B ∗ ⋅ ( ∇ Φ a ∗ + ∂ t A ∗ ) , (13) \dot{v}_{\parallel\mathbf{B}^{*}}=-\frac{q_a}{m_a}\mathbf{B}^{*}\cdot\left(\nabla\Phi_a^{*}+\partial_t\mathbf{A}^{*}\right), \tag*{(13)} v ˙ ∥ B ∗ = − m a q a B ∗ ⋅ ( ∇ Φ a ∗ + ∂ t A ∗ ) , (13)
R ˙ B ∥ ∗ = B ∗ v ∥ + b × ( ∇ Φ a ∗ + ∂ t A ∗ ) , (14) \dot{\mathbf{R}}B_{\parallel}^{*}=\mathbf{B}^{*}v_{\parallel}+\mathbf{b}\times\left(\nabla\Phi_a^{*}+\partial_t\mathbf{A}^{*}\right), \tag*{(14)} R ˙ B ∥ ∗ = B ∗ v ∥ + b × ( ∇ Φ a ∗ + ∂ t A ∗ ) , (14)
where B ∥ ∗ = B ∗ ⋅ b B_{\parallel}^{*}=\mathbf{B}^{*}\cdot\mathbf{b} B ∥ ∗ = B ∗ ⋅ b . Equations (2.8a) and (2.8b) are used to evolve the distribution functions through the Boltzmann equation.
Considering the lowest order in m e / m i m_e/m_i m e / m i in (2) , the electron one-form becomes
Γ e ( R , μ , v ∥ , t ) = q e A ⋅ R ˙ − μ B Ω e θ ˙ − m e v ∥ 2 2 − q e ( ϕ D K + ϕ G K ) + O ( m e m i T a ) , (15) \Gamma_e(\mathbf{R},\mu,v_{\parallel},t)=q_e\mathbf{A}\cdot\dot{\mathbf{R}}-\frac{\mu B}{\Omega_e}\dot{\theta}-\frac{m_ev_{\parallel}^{2}}{2}-q_e\left(\phi_{DK}+\phi_{GK}\right)+O\left(\frac{m_e}{m_i}T_a\right), \tag*{(15)} Γ e ( R , μ , v ∥ , t ) = q e A ⋅ R ˙ − Ω e μ B θ ˙ − 2 m e v ∥ 2 − q e ( ϕ D K + ϕ G K ) + O ( m i m e T a ) , (15)
with the equations of motion identical to that of the ions in eq:2.8 , with B ∥ ∗ = B B_{\parallel}^{*}=B B ∥ ∗ = B , m a = m e m_a=m_e m a = m e , Φ e ∗ = ϕ D K + ϕ G K − μ B e \Phi_e^{*}=\phi_{DK}+\phi_{GK}-\frac{\mu B}{e} Φ e ∗ = ϕ D K + ϕ G K − e μ B , B ∗ = B \mathbf{B}^{*}=\mathbf{B} B ∗ = B , and ∂ t A ∗ = 0 \partial_t\mathbf{A}^{*}=0 ∂ t A ∗ = 0 .
DK and GK Boltzmann equations
Given the equations of motion in eq:2.8 , (9) can, in principle, be solved with appropriate numerical methods once the sources, S a S_a S a , and collision operators, C a b C_{ab} C ab , are defined. However, we rely, in the following, on the splitting of the distribution function in a similar way as ϕ \phi ϕ in (1) . That is
F a ( R , v ∥ , μ ) = F a D K ( R , v ∥ , μ ) + F a G K ( R , v ∥ , μ ) . (16) F_a(\mathbf{R},v_{\parallel},\mu)=F_{aDK}(\mathbf{R},v_{\parallel},\mu)+F_{aGK}(\mathbf{R},v_{\parallel},\mu). \tag*{(16)} F a ( R , v ∥ , μ ) = F a D K ( R , v ∥ , μ ) + F a G K ( R , v ∥ , μ ) . (16)
Analogously to the electrostatic potential, we impose, F a G K ∼ ϵ δ F a D K F_{aGK}\sim\epsilon_{\delta}F_{aDK} F a G K ∼ ϵ δ F a D K , ∣ ∇ ⊥ ln F a D K ∣ ∼ k ⊥ D K |\nabla_{\perp}\ln F_{aDK}|\sim k_{\perp DK} ∣ ∇ ⊥ ln F a D K ∣ ∼ k ⊥ D K , and ∣ ∇ ⊥ ln F a G K ∣ ∼ k ⊥ G K ∼ 1 / ρ s |\nabla_{\perp}\ln F_{aGK}|\sim k_{\perp GK}\sim1/\rho_s ∣ ∇ ⊥ ln F a G K ∣ ∼ k ⊥ G K ∼ 1/ ρ s .
To separate the DK and GK part of the distribution function, we average (9) over small scales, defining the averaging operator such that for any DK-ordered field, g a D K g_{aDK} g a D K ,
⟨ g a D K ⟩ D K = g a D K , (17) \left\langle g_{aDK}\right\rangle_{DK}=g_{aDK}, \tag*{(17)} ⟨ g a D K ⟩ D K = g a D K , (17)
and for the GK fields
⟨ F a G K ⟩ D K = 0 , (18) \left\langle F_{aGK}\right\rangle_{DK}=0, \tag*{(18)} ⟨ F a G K ⟩ D K = 0 , (18)
⟨ ∇ ⊥ ⟨ ϕ G K ⟩ R ⟩ D K = ∇ ⊥ ϕ G K , (19) \left\langle\nabla_{\perp}\left\langle\phi_{GK}\right\rangle_{\mathbf{R}}\right\rangle_{DK}=\nabla_{\perp}\phi_{GK}, \tag*{(19)} ⟨ ∇ ⊥ ⟨ ϕ G K ⟩ R ⟩ D K = ∇ ⊥ ϕ G K , (19)
⟨ ∇ ∥ ⟨ ϕ G K ⟩ R ⟩ D K = 0 , (20) \left\langle\nabla_{\parallel}\left\langle\phi_{GK}\right\rangle_{\mathbf{R}}\right\rangle_{DK}=0, \tag*{(20)} ⟨ ∇ ∥ ⟨ ϕ G K ⟩ R ⟩ D K = 0 , (20)
with ∇ ∥ = b ⋅ ∇ \nabla_{\parallel}=\mathbf{b}\cdot\nabla ∇ ∥ = b ⋅ ∇ . The averaging operation removes all fluctuations on the Larmor-radius scale and we assume that only the perpendicular electric field has a large-scale component due to the assumption ∇ ⊥ ϕ D K ∼ ∇ ⊥ ϕ G K \nabla_{\perp}\phi_{DK}\sim\nabla_{\perp}\phi_{GK} ∇ ⊥ ϕ D K ∼ ∇ ⊥ ϕ G K . The leading order (in ϵ ⊥ \epsilon_{\perp} ϵ ⊥ ) DK average of Eq. (2.6) is
∂ t F a D K + ∇ ⋅ ( R ˙ D K F a D K ) + v ˙ ∥ D K ∂ v ∥ F a D K = S a D K + ∑ b = i , e C a b D K + O ( F a Ω i ϵ ⊥ 2 ) , (21) \partial_t F_{aDK} + \nabla\cdot\left(\dot{\mathbf{R}}_{DK} F_{aDK}\right) + \dot{v}_{\parallel DK}\partial_{v_{\parallel}} F_{aDK} = S_{aDK} + \sum_{b=i,e} C_{abDK} + O\left(F_a\Omega_i\epsilon_\perp^2\right), \tag*{(21)} ∂ t F a D K + ∇ ⋅ ( R ˙ D K F a D K ) + v ˙ ∥ D K ∂ v ∥ F a D K = S a D K + b = i , e ∑ C ab D K + O ( F a Ω i ϵ ⊥ 2 ) , (21)
where
R ˙ D K = b v ∥ + 1 B b × ∇ ( ϕ D K + ϕ G K ) , v ˙ ∥ D K = q a m a ∇ ∥ ϕ D K , (22) \dot{\mathbf{R}}_{DK} = b v_{\parallel} + \frac{1}{B}\mathbf{b} \times\nabla\left(\phi_{DK}+\phi_{GK}\right), \qquad\dot{v}_{\parallel DK} = \frac{q_a}{m_a}\nabla_{\parallel}\phi_{DK}, \tag*{(22)} R ˙ D K = b v ∥ + B 1 b × ∇ ( ϕ D K + ϕ G K ) , v ˙ ∥ D K = m a q a ∇ ∥ ϕ D K , (22)
S a D K = ⟨ S a ⟩ D K , C a b D K = ⟨ C a b ⟩ D K = C a b ( F a D K , F b D K ) . (23) S_{aDK} = \langle S_a\rangle_{DK}, \qquad C_{abDK} = \langle C_{ab}\rangle_{DK} = C_{ab}\left(F_{aDK},F_{bDK}\right). \tag*{(23)} S a D K = ⟨ S a ⟩ D K , C ab D K = ⟨ C ab ⟩ D K = C ab ( F a D K , F b D K ) . (23)
Finally, the GK Boltzmann equation is retrieved by subtracting Eq. (2.12a) from Eq. (9) . That is,
∂ t F a G K + ∇ ⋅ ( ( R ˙ − R ˙ D K ) F a D K + R ˙ F a G K ) + v ˙ ∥ D K ∂ v ∥ F a G K + v ˙ ∥ G K ∂ v ∥ F a D K = ⟨ S a ⟩ R − S a D K + ∑ b = i , e ( ⟨ C a b ⟩ R − C a b D K ) + O ( F a Ω i ϵ ⊥ 2 ϵ δ ) , (24) \partial_t F_{aGK} + \nabla\cdot\left(\left(\dot{\mathbf{R}}-\dot{\mathbf{R}}_{DK}\right)F_{aDK}+\dot{\mathbf{R}}F_{aGK}\right) + \dot{v}_{\parallel DK}\partial_{v_{\parallel}}F_{aGK}+\dot{v}_{\parallel GK}\partial_{v_{\parallel}}F_{aDK}
= \langle S_a\rangle_R-S_{aDK}+\sum_{b=i,e}\left(\langle C_{ab}\rangle_R-C_{abDK}\right)+O\left(F_a\Omega_i\epsilon_\perp^2\epsilon_\delta\right), \tag*{(24)} ∂ t F a G K + ∇ ⋅ ( ( R ˙ − R ˙ D K ) F a D K + R ˙ F a G K ) + v ˙ ∥ D K ∂ v ∥ F a G K + v ˙ ∥ G K ∂ v ∥ F a D K = ⟨ S a ⟩ R − S a D K + b = i , e ∑ ( ⟨ C ab ⟩ R − C ab D K ) + O ( F a Ω i ϵ ⊥ 2 ϵ δ ) , (24)
with
v ˙ ∥ G K = q a m a ∇ ∥ ⟨ ϕ G K ⟩ R . (25) \dot{v}_{\parallel GK}=\frac{q_a}{m_a}\nabla_{\parallel}\langle\phi_{GK}\rangle_R. \tag*{(25)} v ˙ ∥ G K = m a q a ∇ ∥ ⟨ ϕ G K ⟩ R . (25)
Eqs. (2.12a) and (24) describe the evolution of the distribution functions given the electrostatic potentials, ϕ D K \phi_{DK} ϕ D K and ϕ G K \phi_{GK} ϕ G K .
Drift-kinetic governing equations
This section presents the derivation of the part of the model that governs the dynamics of the distribution functions on the DK scale. We simplify the electron evolution described by the DK Boltzmann Eq. (2.12a) using the ordering in § 3.1 . Furthermore, also starting from Eq. (2.12a), we derive a gyromoment hierarchy equation for the ions in § 3.2 by expanding the ion distribution function on a Hermite-Laguerre basis.
Electron drift-kinetic model
For the electrons, we notice that the leading order term in Eq. (2.12a) is order F a Ω i ϵ ⊥ m i / m e F_a\Omega_i\epsilon_\perp\sqrt{m_i/m_e} F a Ω i ϵ ⊥ m i / m e . Expanding the electron distribution function in terms of square root of the mass ratio and keeping terms only to first order, F e D K = F e D K 0 + F e D K 1 F_{eDK}=F_{eDK0}+F_{eDK1} F eD K = F eD K 0 + F eD K 1 , with F e D K 1 ∼ ϵ m F e D K 0 F_{eDK1}\sim\epsilon_mF_{eDK0} F eD K 1 ∼ ϵ m F eD K 0 and ϵ m = m e / m i \epsilon_m=\sqrt{m_e/m_i} ϵ m = m e / m i , the lowest order terms appearing in Eq. (2.12a) are
v ∥ ∥ ∇ ∥ F e D K 0 − e m e ∇ ∥ ϕ D K ∂ v ∥ F e D K 0 = C e e ( F e D K 0 , F e D K 0 ) + C e i ( F e D K 0 , F i D K ) , (26) v_{\parallel}\lVert\nabla_{\parallel} F_{eDK0} - \frac{e}{m_e}\nabla_{\parallel}\phi_{DK}\partial_{v_{\parallel}}F_{eDK0} = C_{ee}(F_{eDK0},F_{eDK0}) + C_{ei}(F_{eDK0},F_{iDK}), \tag*{(26)} v ∥ ∥ ∇ ∥ F eD K 0 − m e e ∇ ∥ ϕ D K ∂ v ∥ F eD K 0 = C ee ( F eD K 0 , F eD K 0 ) + C e i ( F eD K 0 , F i D K ) , (26)
where C e i C_{ei} C e i , at leading order in ϵ m \epsilon_m ϵ m , is given by [4 ]
C e i ( F e D K , F i D K ) = − L e i N e D K 8 π ∂ v γ ( V β γ ∂ v β F e D K ) = L e i ( F e D K ) , (27) C_{ei}(F_{eDK},F_{iDK}) = -\frac{L_{ei}N_{eDK}}{8\pi}\partial_{v_\gamma}\left(V_{\beta\gamma}\partial_{v_\beta}F_{eDK}\right) = \mathcal{L}_{ei}(F_{eDK}), \tag*{(27)} C e i ( F eD K , F i D K ) = − 8 π L e i N eD K ∂ v γ ( V β γ ∂ v β F eD K ) = L e i ( F eD K ) , (27)
with
L e i = 3 π 4 ν e i N e D K v T h e 3 , v T h e = 2 T e m e , (28) L_{ei} = \frac{3\sqrt{\pi}}{4}\frac{\nu_{ei}}{N_{eDK}}v_{The}^{3}, \qquad v_{The} = \sqrt{\frac{2T_e}{m_e}}, \tag*{(28)} L e i = 4 3 π N eD K ν e i v T h e 3 , v T h e = m e 2 T e , (28)
V β γ = 1 v 3 ( v 2 δ β γ − v β v γ ) . (29) V_{\beta\gamma} = \frac{1}{v^3}\left(v^2\delta_{\beta\gamma}-v_\beta v_\gamma\right). \tag*{(29)} V β γ = v 3 1 ( v 2 δ β γ − v β v γ ) . (29)
A solution of Eq. (3.1a) is given by
F e D K 0 = N e 0 ( R , t ) m e 3 / 2 ( 2 π T e 0 ( R ⊥ , t ) ) 3 / 2 exp ( − m e v 2 2 T e 0 ( R ⊥ , t ) ) , (30) F_{eDK0} = N_{e0}(\mathbf{R},t)\frac{m_e^{3/2}}{(2\pi T_{e0}(\mathbf{R}_{\perp},t))^{3/2}}\exp\left(-\frac{m_ev^2}{2T_{e0}(\mathbf{R}_{\perp},t)}\right), \tag*{(30)} F eD K 0 = N e 0 ( R , t ) ( 2 π T e 0 ( R ⊥ , t ) ) 3/2 m e 3/2 exp ( − 2 T e 0 ( R ⊥ , t ) m e v 2 ) , (30)
N e 0 ( R , t ) = N e ⊥ ( R ⊥ , t ) exp ( − e ϕ D K T e 0 ( R ⊥ , t ) ) . (31) N_{e0}(\mathbf{R},t) = N_{e\perp}(\mathbf{R}_{\perp},t)\exp\left(-\frac{e\phi_{DK}}{T_{e0}(\mathbf{R}_{\perp},t)}\right). \tag*{(31)} N e 0 ( R , t ) = N e ⊥ ( R ⊥ , t ) exp ( − T e 0 ( R ⊥ , t ) e ϕ D K ) . (31)
In order to obtain an evolution equation for N e 0 N_{e0} N e 0 and T e 0 T_{e0} T e 0 , we consider the next order in ϵ m \epsilon_m ϵ m of the Boltzmann equation, Eq. (2.12a), that is
∂ t F e D K 0 + 1 B [ ϕ D K + ϕ G K , F e D K 0 ] + v ∥ ∥ ∇ ∥ F e D K 1 − e m e ∇ ∥ ϕ D K ∂ v ∥ F e D K 1 = C e e ( 1 ) ( F e D K 1 ) + L e i ( F e D K 1 − m e v ∥ J ∥ N e 0 T e 0 F e 0 ) + S e D K , (32) \begin{aligned}
\partial_t F_{eDK0} + \frac{1}{B}[\phi_{DK}+\phi_{GK},F_{eDK0}] + v_{\parallel}\lVert\nabla_{\parallel}F_{eDK1} - \frac{e}{m_e}\nabla_{\parallel}\phi_{DK}\partial_{v_{\parallel}}F_{eDK1} \\
= C_{ee}^{(1)}(F_{eDK1}) + \mathcal{L}_{ei}\left(F_{eDK1}-\frac{m_ev_{\parallel}J_{\parallel}}{N_{e0}T_{e0}}F_{e0}\right)+S_{eDK}, \tag*{(32)}
\end{aligned} ∂ t F eD K 0 + B 1 [ ϕ D K + ϕ G K , F eD K 0 ] + v ∥ ∥ ∇ ∥ F eD K 1 − m e e ∇ ∥ ϕ D K ∂ v ∥ F eD K 1 = C ee ( 1 ) ( F eD K 1 ) + L e i ( F eD K 1 − N e 0 T e 0 m e v ∥ J ∥ F e 0 ) + S eD K , (32)
with [ f , g ] = b ⋅ ∇ f × ∇ g [f,g]=\mathbf{b}\cdot\nabla f\times\nabla g [ f , g ] = b ⋅ ∇ f × ∇ g . The F e D K 1 F_{eDK1} F eD K 1 , that solves Eq. (32) , is
F e D K 1 = [ N e 1 N e 0 + P e 1 N e 0 T e 0 ( m e v 2 2 T e 0 − 3 2 ) ] F e D K 0 + f s h , (33) F_{eDK1} = \left[\frac{N_{e1}}{N_{e0}}+\frac{P_{e1}}{N_{e0}T_{e0}}\left(\frac{m_ev^2}{2T_{e0}}-\frac{3}{2}\right)\right]F_{eDK0}+f_{sh}, \tag*{(33)} F eD K 1 = [ N e 0 N e 1 + N e 0 T e 0 P e 1 ( 2 T e 0 m e v 2 − 2 3 ) ] F eD K 0 + f s h , (33)
where
f s h = ∑ k = 0 ∞ v ∥ F e D K N e 0 A k ( R ) L k 3 / 2 ( m e v 2 2 T e 0 ) . (34) f_{sh}=\sum_{k=0}^{\infty}v_{\parallel}\frac{F_{eDK}}{N_{e0}}A_k(\mathbf{R})L_k^{3/2}\left(\frac{m_ev^2}{2T_{e0}}\right). \tag*{(34)} f s h = k = 0 ∑ ∞ v ∥ N e 0 F eD K A k ( R ) L k 3/2 ( 2 T e 0 m e v 2 ) . (34)
Truncating f s h f_{sh} f s h at k = 3 k=3 k = 3 , we determine A k A_k A k , as described in App. § 11 , obtaining
A 0 = 2 m e T e 0 N e 0 U ∥ e D K , A 1 = 1.266 C P 1 + 0.286 m e T e J ∥ , (35) A_0=2\frac{m_e}{T_{e0}}N_{e0}U_{\parallel eDK},\qquad A_1=1.266C_{P1}+0.286\frac{m_e}{T_e}J_{\parallel}, \tag*{(35)} A 0 = 2 T e 0 m e N e 0 U ∥ eD K , A 1 = 1.266 C P 1 + 0.286 T e m e J ∥ , (35)
A 2 = − 0.654 C P 1 + 0.055 m e T e J ∥ , A 3 = 0.0193 C P 1 + 0.0238 m e T e J ∥ , (36) A_2=-0.654C_{P1}+0.055\frac{m_e}{T_e}J_{\parallel},\qquad A_3=0.0193C_{P1}+0.0238\frac{m_e}{T_e}J_{\parallel}, \tag*{(36)} A 2 = − 0.654 C P 1 + 0.055 T e m e J ∥ , A 3 = 0.0193 C P 1 + 0.0238 T e m e J ∥ , (36)
being
C P 1 = 1 T e 0 ν e e ∇ ∥ T e 1 , and T e 1 = P e 1 N e 0 . (37) C_{P1}=\frac{1}{T_{e0}\nu_{ee}}\nabla_{\parallel}T_{e1},\qquad\text{and}\qquad T_{e1}=\frac{P_{e1}}{N_{e0}}. \tag*{(37)} C P 1 = T e 0 ν ee 1 ∇ ∥ T e 1 , and T e 1 = N e 0 P e 1 . (37)
By taking the moments of Eq. (32) with weight 1 and m e v 2 / 2 m_e v^2/2 m e v 2 /2 , we determine an evolution equation for N e D K = N e 0 + N e 1 N_{eDK}=N_{e0}+N_{e1} N eD K = N e 0 + N e 1 and T e D K = T e 0 + T e 1 T_{eDK}=T_{e0}+T_{e1} T eD K = T e 0 + T e 1 , respectively, that is
∂ t N e D K + 1 B [ ϕ D K + ϕ G K , N e D K ] + ∇ ∥ ( U ∥ e D K N e D K ) = S N + O ( N e Ω i ϵ e 3 ) , (38) \partial_t N_{eDK}+\frac{1}{B}\left[\phi_{DK}+\phi_{GK},N_{eDK}\right]+\nabla_{\parallel}\left(U_{\parallel eDK}N_{eDK}\right)=S_N+O\left(N_e\Omega_i\epsilon_e^3\right), \tag*{(38)} ∂ t N eD K + B 1 [ ϕ D K + ϕ G K , N eD K ] + ∇ ∥ ( U ∥ eD K N eD K ) = S N + O ( N e Ω i ϵ e 3 ) , (38)
and
3 2 ( ∂ t T e D K + 1 B [ ϕ D K + ϕ G K , T e D K ] + U ∥ e D K ∇ ∥ T e D K ) + T e D K ∇ ∥ U ∥ e D K − 0.71 T e D K N e D K ∇ ∥ J ∥ + 1 N e D K ∇ ∥ q ∥ e r = 3 2 S T e + O ( T e Ω i ϵ e 3 ) , (39) \frac{3}{2}\left(\partial_t T_{eDK}+\frac{1}{B}\left[\phi_{DK}+\phi_{GK},T_{eDK}\right]+U_{\parallel eDK}\nabla_{\parallel}T_{eDK}\right)+T_{eDK}\nabla_{\parallel}U_{\parallel eDK}-\frac{0.71T_{eDK}}{N_{eDK}}\nabla_{\parallel}J_{\parallel}+\frac{1}{N_{eDK}}\nabla_{\parallel}q_{\parallel er}=\frac{3}{2}S_{Te}+O\left(T_e\Omega_i\epsilon_e^3\right), \tag*{(39)} 2 3 ( ∂ t T eD K + B 1 [ ϕ D K + ϕ G K , T eD K ] + U ∥ eD K ∇ ∥ T eD K ) + T eD K ∇ ∥ U ∥ eD K − N eD K 0.71 T eD K ∇ ∥ J ∥ + N eD K 1 ∇ ∥ q ∥ er = 2 3 S T e + O ( T e Ω i ϵ e 3 ) , (39)
being
q ∥ e r = − 3.16 N e D K T e D K m e ν e e ∇ ∥ T e D K , (40) q_{\parallel er}=-\frac{3.16N_{eDK}T_{eDK}}{m_e\nu_{ee}}\nabla_{\parallel}T_{eDK}, \tag*{(40)} q ∥ er = − m e ν ee 3.16 N eD K T eD K ∇ ∥ T eD K , (40)
(see App. A for the details of the derivation). We note that Eqs. (3.5a) and (3.5b) correspond to the evolution of the density and electron temperature in the drift-reduced Braginskii model with the electrostatic potential expressed as the sum of the GK and DK electrostatic potentials, ϕ D K + ϕ G K \phi_{DK}+\phi_{GK} ϕ D K + ϕ G K [4 , 42 , 31 ] .
Ion moment hierarchy
For the ions, instead of ordering the ion-ion collision frequency as ν i i ≪ ω \nu_{ii}\ll\omega ν ii ≪ ω and recover the drift-reduced Braginskii through a similar calculation as in § 3.1 , we assume ν i i ∼ ω \nu_{ii}\sim\omega ν ii ∼ ω , similar to [9 , 10 , 25 ] , and expand the DK part of the distribution functions on a Hermite-Laguerre basis
F i D K ( R , v ∥ , μ ) = ∑ p , j = 0 ∞ N i D K p j ( R ) H p ( s ∥ i ) L j ( x i ) 2 p p ! F M i ( s ∥ i , x i ) , (41) F_{iDK}(\mathbf{R},v_{\parallel},\mu)=\sum_{p,j=0}^{\infty}N_{iDK}^{pj}(\mathbf{R})\frac{H_p(s_{\parallel i})L_j(x_i)}{\sqrt{2^p p!}}F_{Mi}(s_{\parallel i},x_i), \tag*{(41)} F i D K ( R , v ∥ , μ ) = p , j = 0 ∑ ∞ N i D K p j ( R ) 2 p p ! H p ( s ∥ i ) L j ( x i ) F M i ( s ∥ i , x i ) , (41)
where we introduce
s ∥ i = v ∥ − U ∥ i D K ( R ) v T h i , x i = μ B T i 0 , v T h i = 2 T i 0 m i , (42) s_{\parallel i}=\frac{v_{\parallel}-U_{\parallel iDK}(\mathbf{R})}{v_{Thi}},\qquad x_i=\frac{\mu B}{T_{i0}},\qquad v_{Thi}=\sqrt{\frac{2T_{i0}}{m_i}}, \tag*{(42)} s ∥ i = v T hi v ∥ − U ∥ i D K ( R ) , x i = T i 0 μ B , v T hi = m i 2 T i 0 , (42)
F M i = e − s ∥ i 2 − x i ( π v T h i ) 3 , (43) F_{Mi}=\frac{e^{-s_{\parallel i}^{2}-x_i}}{\left(\sqrt{\pi}v_{Thi}\right)^3}, \tag*{(43)} F M i = ( π v T hi ) 3 e − s ∥ i 2 − x i , (43)
and the background temperature of the ions, T i 0 T_{i0} T i 0 , along with the physicists’ Hermite polynomials, H p H_p H p , and the Laguerre polynomials, L j L_j L j . These polynomials satisfy the following orthogonality relations,
∫ − ∞ ∞ d x H p ( x ) H p ′ ( x ) e − x 2 = 2 p p ! δ p ′ p (44) \int_{-\infty}^{\infty} \mathrm{d}x H_p(x) H_{p'}(x) e^{-x^2}=2^p p!\delta_{p'}^p \tag*{(44)} ∫ − ∞ ∞ d x H p ( x ) H p ′ ( x ) e − x 2 = 2 p p ! δ p ′ p (44)
∫ 0 ∞ d x L j ( x ) L j ′ ( x ) e − x = δ j ′ j (45) \int_0^{\infty} \mathrm{d}x L_j(x)L_{j'}(x)e^{-x}=\delta_{j'}^j \tag*{(45)} ∫ 0 ∞ d x L j ( x ) L j ′ ( x ) e − x = δ j ′ j (45)
with δ n m \delta_n^m δ n m the Kronecker delta. For simplicity, T i 0 T_{i0} T i 0 is chosen as a constant in Eq. (3.6), in contrast to Frei et al. [9 ] where a dynamical bi-Maxwellian is used as the weight function for the Hermite and Laguerre polynomial basis. We note that the generalization of the projection in Eq. (3.6) onto a basis with anisotropic temperatures is straightforward.
We define the DK velocity-space projection of a general phase-space function, χ \chi χ , on the p p p ’th Hermite and j j j ’th Laguerre polynomial, that is
∥ χ ∥ i p j = ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ B m i ∫ 0 2 π d θ F i D K H p L j 2 p p ! χ ( R , v ∥ , μ , θ ) . (46) \|\chi\|_i^{pj}=\int_{-\infty}^{\infty}\mathrm{d}v_{\parallel}\int_0^{\infty}\mathrm{d}\mu\frac{B}{m_i}\int_0^{2\pi}\mathrm{d}\theta F_i^{DK}\frac{H_pL_j}{\sqrt{2^p p!}}\chi\left(\boldsymbol{R},v_{\parallel},\mu,\theta\right). \tag*{(46)} ∥ χ ∥ i p j = ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ m i B ∫ 0 2 π d θ F i D K 2 p p ! H p L j χ ( R , v ∥ , μ , θ ) . (46)
And we further introduce the shorthand notation,
∥ χ ∥ i = ∥ χ ∥ i 00 . (47) \|\chi\|_i=\|\chi\|_i^{00}. \tag*{(47)} ∥ χ ∥ i = ∥ χ ∥ i 00 . (47)
We notice that in terms of the projector in Eq. (3.8b) and the gyromoments in Eq. (3.6a), the DK ion density is given by
N i D K = ∥ 1 ∥ i = N i D K 00 , (48) N_{iDK}=\|1\|_i=N_{iDK}^{00}, \tag*{(48)} N i D K = ∥1 ∥ i = N i D K 00 , (48)
the DK ion parallel velocity by
U ∥ i D K = ∥ v ∥ ∥ i / N i D K , (49) U_{\parallel iDK}=\|v_{\parallel}\|_i/N_{iDK}, \tag*{(49)} U ∥ i D K = ∥ v ∥ ∥ i / N i D K , (49)
the DK ion parallel pressure by
P ∥ i D K = m i ∥ ( v ∥ − U ∥ i D K ) 2 ∥ i = T i 0 ( 2 N i D K 20 + N i D K ) , (50) P_{\parallel iDK}=m_i\left\|(v_{\parallel}-U_{\parallel iDK})^2\right\|_i=T_{i0}\left(\sqrt{2}N_{iDK}^{20}+N_{iDK}\right), \tag*{(50)} P ∥ i D K = m i ( v ∥ − U ∥ i D K ) 2 i = T i 0 ( 2 N i D K 20 + N i D K ) , (50)
and the DK ion perpendicular pressure by
P ⊥ i D K = ∥ μ B ∥ i = T i 0 ( N i D K − N i D K 01 ) . (51) P_{\perp iDK}=\|\mu B\|_i=T_{i0}\left(N_{iDK}-N_{iDK}^{01}\right). \tag*{(51)} P ⊥ i D K = ∥ μ B ∥ i = T i 0 ( N i D K − N i D K 01 ) . (51)
Finally, from Eq. (3.9b), we have that
N i D K 10 = 2 ∥ s ∥ i ∥ i = 0. (52) N_{iDK}^{10}=\sqrt{2}\|s_{\parallel i}\|_i=0. \tag*{(52)} N i D K 10 = 2 ∥ s ∥ i ∥ i = 0. (52)
Projecting Eq. (2.12a) for a = i a=i a = i on the Hermite Laguerre basis gives a gyromoment hierarchy
∂ ∂ t N i D K p j + 2 p v T h i N i D K p − 1 j ∂ ∂ t U ∥ i D K + 2 p v T h i ∥ R ⃗ ∥ i 0 p − 1 j ⋅ ∇ U ∥ i D K + ∇ ⋅ ∥ R ⃗ ∥ i 0 p j − 2 p v T h i ∥ v ⃗ ∥ i 0 p − 1 j = C i i 0 p j + C i e 0 p j + S i 0 p j , (53) \frac{\partial}{\partial t}N_{iDK}^{pj}+\frac{\sqrt{2p}}{v_{Thi}}N_{iDK}^{p-1j}\frac{\partial}{\partial t}U_{\|iDK}+\frac{\sqrt{2p}}{v_{Thi}}\left\lVert\vec{R}\right\rVert_{i0}^{p-1j}\cdot\nabla U_{\|iDK}
+\nabla\cdot\left\lVert\vec{R}\right\rVert_{i0}^{pj}-\frac{\sqrt{2p}}{v_{Thi}}\left\lVert\vec{v}\right\rVert_{i0}^{p-1j}=C_{ii0}^{pj}+C_{ie0}^{pj}+S_{i0}^{pj}, \tag*{(53)} ∂ t ∂ N i D K p j + v T hi 2 p N i D K p − 1 j ∂ t ∂ U ∥ i D K + v T hi 2 p R i 0 p − 1 j ⋅ ∇ U ∥ i D K + ∇ ⋅ R i 0 p j − v T hi 2 p ∥ v ∥ i 0 p − 1 j = C ii 0 p j + C i e 0 p j + S i 0 p j , (53)
with
∇ ⋅ ∥ R ⃗ ∥ i 0 p j = ∇ ∥ v ⃗ ∥ i p j + 1 B [ ϕ D K + ϕ G K , N i D K p j ] , (54) \nabla\cdot\left\lVert\vec{R}\right\rVert_{i0}^{pj}=\nabla\left\lVert\vec{v}\right\rVert_i^{pj}+\frac{1}{B}\left[\phi_{DK}+\phi_{GK},N_{iDK}^{pj}\right], \tag*{(54)} ∇ ⋅ R i 0 p j = ∇ ∥ v ∥ i p j + B 1 [ ϕ D K + ϕ G K , N i D K p j ] , (54)
∥ R ⃗ ∥ i 0 p j ⋅ ∇ U ∥ i D K = ∥ v ⃗ ∥ i p j ∇ ∥ U ∥ i D K + N i D K p j B [ ϕ D K + ϕ G K , U ∥ i D K ] , (55) \left\lVert\vec{R}\right\rVert_{i0}^{pj}\cdot\nabla U_{\|iDK}=\left\lVert\vec{v}\right\rVert_i^{pj}\nabla_{\|}U_{\|iDK}+\frac{N_{iDK}^{pj}}{B}\left[\phi_{DK}+\phi_{GK},U_{\|iDK}\right], \tag*{(55)} R i 0 p j ⋅ ∇ U ∥ i D K = ∥ v ∥ i p j ∇ ∥ U ∥ i D K + B N i D K p j [ ϕ D K + ϕ G K , U ∥ i D K ] , (55)
∥ v ⃗ ∥ i p j = − q i m i N i D K p j ∇ ∥ ϕ D K , (56) \left\lVert\vec{v}\right\rVert_i^{pj}=-\frac{q_i}{m_i}N_{iDK}^{pj}\nabla_{\|}\phi_{DK}, \tag*{(56)} ∥ v ∥ i p j = − m i q i N i D K p j ∇ ∥ ϕ D K , (56)
and
∥ v ⃗ ∥ i p j = v T h i 2 ( p + 1 N i D K p + 1 j + p N i D K p − 1 j ) + U ∥ i D K N i D K p j , (57) \left\lVert\vec{v}\right\rVert_i^{pj}=\frac{v_{Thi}}{\sqrt{2}}\left(\sqrt{p+1}N_{iDK}^{p+1j}+\sqrt{p}N_{iDK}^{p-1j}\right)+U_{\|iDK}N_{iDK}^{pj}, \tag*{(57)} ∥ v ∥ i p j = 2 v T hi ( p + 1 N i D K p + 1 j + p N i D K p − 1 j ) + U ∥ i D K N i D K p j , (57)
where we introduce the small-scale-averaged projection of the ion-ion collision operator, C i i 0 p j = ∥ C i i D K / F i D K ∥ p j C_{ii0}^{pj}=\left\lVert C_{iiDK}/F_{iDK}\right\rVert^{pj} C ii 0 p j = ∥ C ii D K / F i D K ∥ p j , the ion-electron collision operator, C i e 0 p j = ∥ C i e D K / F i D K ∥ p j C_{ie0}^{pj}=\left\lVert C_{ieDK}/F_{iDK}\right\rVert^{pj} C i e 0 p j = ∥ C i eD K / F i D K ∥ p j , and the ion source, S i 0 p j = ∥ S i D K / F i D K ∥ p j S_{i0}^{pj}=\left\lVert S_{iDK}/F_{iDK}\right\rVert^{pj} S i 0 p j = ∥ S i D K / F i D K ∥ p j .
The equation for U ∥ i D K U_{\|iDK} U ∥ i D K is retrieved by setting ( p , j ) = ( 1 , 0 ) (p,j)=(1,0) ( p , j ) = ( 1 , 0 ) in Eq. (3.10a), that is
m i N i D K ∂ t U ∥ i D K = − m i N i D K ( U ∥ i D K ∇ ∥ U ∥ i D K + 1 B [ ϕ D K + ϕ G K , U ∥ i D K ] ) − ∇ ∥ P ∥ i D K − e ∇ ∥ ϕ D K − F e i + m i 2 v T h i S i 0 10 + O ( m i N i D K c s ω ) , (58) m_iN_{iDK}\partial_tU_{\|iDK}=-m_iN_{iDK}\left(U_{\|iDK}\nabla_{\|}U_{\|iDK}+\frac{1}{B}\left[\phi_{DK}+\phi_{GK},U_{\|iDK}\right]\right)-\nabla_{\|}P_{\|iDK}-e\nabla_{\|}\phi_{DK}-F_{ei}+\frac{m_i}{\sqrt{2}}v_{Thi}S_{i0}^{10}+O\left(m_iN_{iDK}c_s\omega\right), \tag*{(58)} m i N i D K ∂ t U ∥ i D K = − m i N i D K ( U ∥ i D K ∇ ∥ U ∥ i D K + B 1 [ ϕ D K + ϕ G K , U ∥ i D K ] ) − ∇ ∥ P ∥ i D K − e ∇ ∥ ϕ D K − F e i + 2 m i v T hi S i 0 10 + O ( m i N i D K c s ω ) , (58)
with
F e i = m e ∫ d v v ∥ C e i = − m i ∫ d v v ∥ C i e , (59) F_{ei}=m_e\int\mathrm{d}v\,v_{\|}C_{ei}=-m_i\int\mathrm{d}v\,v_{\|}C_{ie}, \tag*{(59)} F e i = m e ∫ d v v ∥ C e i = − m i ∫ d v v ∥ C i e , (59)
due to conservation of momentum.
Explicitly inserting the relation for ∂ t U ∥ i D K \partial_tU_{\|iDK} ∂ t U ∥ i D K , Eq. (3.11a), in Eq. (3.10a) we get
∂ t N i D K p j + ∇ ∥ ∥ v ⃗ ∥ i p j + p ∥ S ∥ i p − 1 j ∇ ∥ U ∥ i D K + 1 B [ ϕ D K + ϕ G K , N i D K p j ] − 2 p N i D K p − 1 j N i D K 1 m i v T h i ∇ ∥ P ∥ i D K = ∑ s C i s 0 p j + S i 0 p j − p N i D K p − 1 j N i D K m i v T h i S U , (60) \partial_tN_{iDK}^{pj}+\nabla_{\|}\left\lVert\vec{v}\right\rVert_i^{pj}+\sqrt{p}\left\lVert S\right\rVert_i^{p-1j}\nabla_{\|}U_{\|iDK}+\frac{1}{B}\left[\phi_{DK}+\phi_{GK},N_{iDK}^{pj}\right]
-\frac{\sqrt{2p}N_{iDK}^{p-1j}}{N_{iDK}}\frac{1}{m_iv_{Thi}}\nabla_{\|}P_{\|iDK}
=\sum_s C_{is0}^{pj}+S_{i0}^{pj}-\frac{\sqrt{p}N_{iDK}^{p-1j}}{N_{iDK}m_iv_{Thi}}S_U, \tag*{(60)} ∂ t N i D K p j + ∇ ∥ ∥ v ∥ i p j + p ∥ S ∥ i p − 1 j ∇ ∥ U ∥ i D K + B 1 [ ϕ D K + ϕ G K , N i D K p j ] − N i D K 2 p N i D K p − 1 j m i v T hi 1 ∇ ∥ P ∥ i D K = s ∑ C i s 0 p j + S i 0 p j − N i D K m i v T hi p N i D K p − 1 j S U , (60)
∥ S ∥ i p j = p + 1 N i D K p + 1 j + p N i D K p − 1 j . (61) \left\lVert S\right\rVert_i^{pj}=\sqrt{p+1}N_{iDK}^{p+1j}+\sqrt{p}N_{iDK}^{p-1j}. \tag*{(61)} ∥ S ∥ i p j = p + 1 N i D K p + 1 j + p N i D K p − 1 j . (61)
Finally, we use a long-wavelength Dougherty collision operator Dougherty [8 ] in Eq. (60) whose projection is given by Frei et al. [10 ] ; Mencke & Ricci [25 ]
C i i 0 p j = ν i [ − ( p + 2 j ) N i D K p j + ( T i D K − 1 ) ( p ( p − 1 ) N i D K p − 2 j − 2 j N i D K p j − 1 ) ] , (62) \mathcal{C}_{ii0}^{pj}=\nu_i\left[-(p+2j)\mathcal{N}_{iDK}^{pj}+(T_{iDK}-1)\left(\sqrt{p(p-1)}\mathcal{N}_{iDK}^{p-2j}-2j\mathcal{N}_{iDK}^{pj-1}\right)\right], \tag*{(62)} C ii 0 p j = ν i [ − ( p + 2 j ) N i D K p j + ( T i D K − 1 ) ( p ( p − 1 ) N i D K p − 2 j − 2 j N i D K p j − 1 ) ] , (62)
with ν i = ν 0 1.38 m i / m e ( T i D K / T e 0 ) − 3 / 2 \nu_i=\nu_0 1.38\sqrt{m_i/m_e}(T_{iDK}/T_{e0})^{-3/2} ν i = ν 0 1.38 m i / m e ( T i D K / T e 0 ) − 3/2 and ν 0 = 4 2 π e 4 N 0 m e ln λ / [ 3 m i T e 0 3 / 2 1.96 ( 4 π ϵ 0 ) 2 ] \nu_0=4\sqrt{2}\pi e^4N_0\sqrt{m_e}\ln\lambda/[3m_iT_{e0}^{3/2}1.96(4\pi\epsilon_0)^2] ν 0 = 4 2 π e 4 N 0 m e ln λ / [ 3 m i T e 0 3/2 1.96 ( 4 π ϵ 0 ) 2 ] .
Gyrokinetic governing equations
For the governing equations of the GK fields, we first develop the GK part of the electron distribution function in a similar way as the DK counterpart assuming an adiabatic response to the GK electrostatic potential. We then derive a moment hierarchy for the GK ion gyromoments, similarly as for the DK ion gyromoments.
Focusing first on the electron species, we perturbatively solve the Boltzmann equation by expanding in orders of ϵ m \epsilon_m ϵ m . We first write down the order ϵ m − 1 ∂ t F e G K \epsilon_m^{-1}\partial_tF_{eGK} ϵ m − 1 ∂ t F e G K of Eq. (2.13a) with a = e a=e a = e , that is
v ∥ ∇ ∥ F e G K − e m e ∇ ∥ ϕ D K ∂ v ∥ F e G K − e m e ∇ ∥ ϕ G K ∂ v ∥ F e D K = C e e 1 ( F e D K , F e G K ) + L e i ( F e G K ) . (63) v_{\parallel}\nabla_{\parallel}F_{eGK}-\frac{e}{m_e}\nabla_{\parallel}\phi_{DK}\partial_{v_{\parallel}}F_{eGK}-\frac{e}{m_e}\nabla_{\parallel}\phi_{GK}\partial_{v_{\parallel}}F_{eDK}=C_{ee1}(F_{eDK},F_{eGK})+L_{ei}(F_{eGK}). \tag*{(63)} v ∥ ∇ ∥ F e G K − m e e ∇ ∥ ϕ D K ∂ v ∥ F e G K − m e e ∇ ∥ ϕ G K ∂ v ∥ F eD K = C ee 1 ( F eD K , F e G K ) + L e i ( F e G K ) . (63)
Since Eq. (63) is linear in F e G K F_{eGK} F e G K , its general solution is given by
F e G K = ( δ N e G K ( R ⊥ , t ) N e D K + e ϕ G K T e D K ) F e D K , (64) F_{eGK}=\left(\frac{\delta N_{eGK}(\mathbf{R}_{\perp},t)}{N_{eDK}}+\frac{e\phi_{GK}}{T_{eDK}}\right)F_{eDK}, \tag*{(64)} F e G K = ( N eD K δ N e G K ( R ⊥ , t ) + T eD K e ϕ G K ) F eD K , (64)
where δ N e G K \delta N_{eGK} δ N e G K is a free parameter. We set δ N e G K = 0 \delta N_{eGK}=0 δ N e G K = 0 , allowing us to close our system of equations. This corresponds to an adiabatic Boltzmann response for the GK part of the electron distribution function.
For the ions, similarly to the DK system, we expand the GK part of the ion distribution function on a Hermite-Laguerre basis
F i G K ( R , v ∥ , μ ) = ∑ p , j = 0 ∞ N i G K p j ( R ) H p ( s ∥ i ) L j ( x i ) 2 p p ! F M i ( s ∥ i , x i ) . (65) F_{iGK}(\mathbf{R},v_{\parallel},\mu)=\sum_{p,j=0}^{\infty}\mathcal{N}_{iGK}^{pj}(\mathbf{R})\frac{H_p(s_{\parallel i})L_j(x_i)}{\sqrt{2^p p!}}F_{Mi}(s_{\parallel i},x_i). \tag*{(65)} F i G K ( R , v ∥ , μ ) = p , j = 0 ∑ ∞ N i G K p j ( R ) 2 p p ! H p ( s ∥ i ) L j ( x i ) F M i ( s ∥ i , x i ) . (65)
The GK velocity-space projection of a general phase-space function, χ \chi χ , is defined
∥ χ ∥ i δ p j = ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ B m i ∫ 0 2 π d θ F i G K H p L j 2 p p ! χ ( R , v ∥ , μ , θ ) , (66) \|\chi\|_{i}^{\delta pj}=\int_{-\infty}^{\infty}dv_{\parallel}\int_{0}^{\infty}d\mu\frac{B}{m_i}\int_{0}^{2\pi}d\theta F_{iGK}\frac{H_pL_j}{\sqrt{2^p p!}}\chi(\mathbf{R},v_{\parallel},\mu,\theta), \tag*{(66)} ∥ χ ∥ i δ p j = ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ m i B ∫ 0 2 π d θ F i G K 2 p p ! H p L j χ ( R , v ∥ , μ , θ ) , (66)
where we additionally define,
∥ χ ∥ i δ = ∥ χ ∥ i δ 00 . (67) \|\chi\|_{i}^{\delta}=\|\chi\|_{i}^{\delta00}. \tag*{(67)} ∥ χ ∥ i δ = ∥ χ ∥ i δ 00 . (67)
The first few moments of the GK distribution function can be expanded in terms of the GK and DK fluid quantities and gyromoments as follows
∥ u ∥ i δ = N i G K 00 = N i G K , (68) \lVert\mathbf{u} \rVert_i^\delta= \mathcal{N}^{00}_{iGK} = N_{iGK}, \tag*{(68)} ∥ u ∥ i δ = N i G K 00 = N i G K , (68)
∥ v ∥ ∥ i δ = U ∥ i D K N i G K + v T h i 2 N i G K 10 = N i D K U ∥ i G K + U ∥ i D K N i G K , (69) \lVert v_\parallel\rVert_i^\delta= U_{\parallel iDK}N_{iGK} + \frac{v_{Thi}}{\sqrt{2}}\mathcal{N}^{10}_{iGK} = N_{iDK}U_{\parallel iGK} + U_{\parallel iDK}N_{iGK}, \tag*{(69)} ∥ v ∥ ∥ i δ = U ∥ i D K N i G K + 2 v T hi N i G K 10 = N i D K U ∥ i G K + U ∥ i D K N i G K , (69)
m i ∥ ( v ∥ − U ∥ i D K ) 2 ∥ i δ − 2 m i ∥ ( v ∥ − U ∥ i D K ) U ∥ i G K ∥ i = T i 0 ( 2 N i G K 20 + N i G K ) = P ∥ i G K , (70) m_i\lVert(v_\parallel-U_{\parallel iDK})^2\rVert_i^\delta- 2m_i\lVert(v_\parallel-U_{\parallel iDK})U_{\parallel iGK}\rVert_i
= T_{i0}\left(\sqrt{2}\mathcal{N}^{20}_{iGK}+N_{iGK}\right)=P_{\parallel iGK}, \tag*{(70)} m i ∥( v ∥ − U ∥ i D K ) 2 ∥ i δ − 2 m i ∥( v ∥ − U ∥ i D K ) U ∥ i G K ∥ i = T i 0 ( 2 N i G K 20 + N i G K ) = P ∥ i G K , (70)
∥ μ B ⟧ i δ = T i 0 ( N i G K − N i G K 01 ) = P ⊥ i G K , (71) \lVert\mu B\rrbracket_i^\delta= T_{i0}\left(N_{iGK}-N^{01}_{iGK}\right)=P_{\perp iGK}, \tag*{(71)} ∥ μ B ] ] i δ = T i 0 ( N i G K − N i G K 01 ) = P ⊥ i G K , (71)
and, from Eq. (4.3e), it follows that
U ∥ i G K = v T h i 2 N i G K 10 N i D K . (72) U_{\parallel iGK}=\frac{v_{Thi}}{\sqrt{2}}\frac{\mathcal{N}^{10}_{iGK}}{N_{iDK}}. \tag*{(72)} U ∥ i G K = 2 v T hi N i D K N i G K 10 . (72)
In order to derive the evolution equations for N i G K p j \mathcal{N}^{pj}_{iGK} N i G K p j , we project Eq. (2.13a) with a = i a=i a = i on the Hermite-Laguerre basis, thus obtaining
∂ ∂ t N i G K p j + 2 p v T h i ∥ R ˙ ∥ i 1 p − 1 j ⋅ ∇ U ∥ i D K + ∇ ⋅ ∥ R ˙ ∥ i 1 p j − 2 p v T h i ∥ v ˙ ∥ i 1 p − 1 j + 2 p v T h i N i G K p − 1 j ∂ U ∥ i D K ∂ t + 2 p v T h i ∥ R ˙ ∥ i 1 δ p − 1 j ⋅ ∇ U ∥ i D K + ∇ ⋅ ∥ R ˙ ∥ i 1 δ p j − 2 p v T h i ∥ v ˙ ∥ i 1 δ p − 1 j = C i i 1 p j + C i e 1 p j + δ S i p j , (73) \begin{aligned}
\frac{\partial}{\partial t}\mathcal{N}^{pj}_{iGK}
&+\frac{\sqrt{2p}}{v_{Thi}}\lVert\dot{\mathbf{R}}\rVert^{p-1j}_{i1}\cdot\nabla U_{\parallel iDK}
+\nabla\cdot\lVert\dot{\mathbf{R}}\rVert^{pj}_{i1}
-\frac{\sqrt{2p}}{v_{Thi}}\lVert\dot{\mathbf{v}}\rVert^{p-1j}_{i1}\\
&+\frac{\sqrt{2p}}{v_{Thi}}\mathcal{N}^{p-1j}_{iGK}\frac{\partial U_{\parallel iDK}}{\partial t}
+\frac{\sqrt{2p}}{v_{Thi}}\lVert\dot{\mathbf{R}}\rVert^{\delta p-1j}_{i1}\cdot\nabla U_{\parallel iDK}
+\nabla\cdot\lVert\dot{\mathbf{R}}\rVert^{\delta pj}_{i1}
-\frac{\sqrt{2p}}{v_{Thi}}\lVert\dot{\mathbf{v}}\rVert^{\delta p-1j}_{i1}\\
&=C^{pj}_{ii1}+C^{pj}_{ie1}+\delta S^{pj}_{i}, \tag*{(73)}
\end{aligned} ∂ t ∂ N i G K p j + v T hi 2 p ∥ R ˙ ∥ i 1 p − 1 j ⋅ ∇ U ∥ i D K + ∇ ⋅ ∥ R ˙ ∥ i 1 p j − v T hi 2 p ∥ v ˙ ∥ i 1 p − 1 j + v T hi 2 p N i G K p − 1 j ∂ t ∂ U ∥ i D K + v T hi 2 p ∥ R ˙ ∥ i 1 δ p − 1 j ⋅ ∇ U ∥ i D K + ∇ ⋅ ∥ R ˙ ∥ i 1 δ p j − v T hi 2 p ∥ v ˙ ∥ i 1 δ p − 1 j = C ii 1 p j + C i e 1 p j + δ S i p j , (73)
where
∥ R ˙ ∥ i 1 p j ⋅ ∇ U ∥ = 1 B ∥ b × ∇ ( ⟨ ψ ⟩ R − ⟨ ϕ G K ⟩ R ) ∥ i p j ⋅ ∇ U ∥ + 1 B ∥ b × ∇ ⟨ ϕ G K ⟩ R ∥ i p j ⋅ ∇ U ∥ , (74) \lVert\dot{\mathbf{R}}\rVert^{pj}_{i1}\cdot\nabla U_\parallel
=\frac{1}{B}\lVert\mathbf{b}\times\nabla\left(\langle\psi\rangle_R-\langle\phi_{GK}\rangle_R\right)\rVert^{pj}_{i}\cdot\nabla U_\parallel
+\frac{1}{B}\lVert\mathbf{b}\times\nabla\langle\phi_{GK}\rangle_R\rVert^{pj}_{i}\cdot\nabla U_\parallel, \tag*{(74)} ∥ R ˙ ∥ i 1 p j ⋅ ∇ U ∥ = B 1 ∥ b × ∇ ( ⟨ ψ ⟩ R − ⟨ ϕ G K ⟩ R ) ∥ i p j ⋅ ∇ U ∥ + B 1 ∥ b × ∇ ⟨ ϕ G K ⟩ R ∥ i p j ⋅ ∇ U ∥ , (74)
∇ ⋅ ∥ R ˙ ∥ i 1 p j = 1 B ∇ ⋅ ∥ b × ∇ ( ⟨ ψ ⟩ R − ⟨ ϕ G K ⟩ R ) ∥ i p j + 1 B ∇ ⋅ ∥ b × ∇ ⟨ ϕ G K ⟩ R ∥ i p j , (75) \nabla\cdot\lVert\dot{\mathbf{R}}\rVert^{pj}_{i1}
=\frac{1}{B}\nabla\cdot\lVert\mathbf{b}\times\nabla\left(\langle\psi\rangle_R-\langle\phi_{GK}\rangle_R\right)\rVert^{pj}_{i}
+\frac{1}{B}\nabla\cdot\lVert\mathbf{b}\times\nabla\langle\phi_{GK}\rangle_R\rVert^{pj}_{i}, \tag*{(75)} ∇ ⋅ ∥ R ˙ ∥ i 1 p j = B 1 ∇ ⋅ ∥ b × ∇ ( ⟨ ψ ⟩ R − ⟨ ϕ G K ⟩ R ) ∥ i p j + B 1 ∇ ⋅ ∥ b × ∇ ⟨ ϕ G K ⟩ R ∥ i p j , (75)
∥ v ˙ ∥ i 1 p j = − q i m i ∥ ∇ ∥ ⟨ ϕ G K ⟩ R ∥ i p j , (76) \lVert\dot{\mathbf{v}}\rVert^{pj}_{i1}
=-\frac{q_i}{m_i}\lVert\nabla_\parallel\langle\phi_{GK}\rangle_R\rVert^{pj}_{i}, \tag*{(76)} ∥ v ˙ ∥ i 1 p j = − m i q i ∥ ∇ ∥ ⟨ ϕ G K ⟩ R ∥ i p j , (76)
N i G K p − 1 j ∂ U ∥ ∂ t + ∥ R ˙ ∥ i 1 δ p − 1 j ⋅ ∇ U ∥ − ∥ v ˙ ∥ i 1 δ p − 1 j = v T h i 2 ∥ s ∥ i δ p − 1 j ∇ ∥ U ∥ − N i G K p − 1 j N i D K ( 1 m i ∇ ∥ P ∥ − S U m i ) , (77) \mathcal{N}^{p-1j}_{iGK}\frac{\partial U_\parallel}{\partial t}
+\lVert\dot{\mathbf{R}}\rVert^{\delta p-1j}_{i1}\cdot\nabla U_\parallel
-\lVert\dot{\mathbf{v}}\rVert^{\delta p-1j}_{i1}
=\frac{v_{Thi}}{\sqrt{2}}\lVert s\rVert^{\delta p-1j}_{i}\nabla_\parallel U_\parallel
-\frac{\mathcal{N}^{p-1j}_{iGK}}{N_{iDK}}\left(\frac{1}{m_i}\nabla_\parallel P_\parallel-\frac{S_U}{m_i}\right), \tag*{(77)} N i G K p − 1 j ∂ t ∂ U ∥ + ∥ R ˙ ∥ i 1 δ p − 1 j ⋅ ∇ U ∥ − ∥ v ˙ ∥ i 1 δ p − 1 j = 2 v T hi ∥ s ∥ i δ p − 1 j ∇ ∥ U ∥ − N i D K N i G K p − 1 j ( m i 1 ∇ ∥ P ∥ − m i S U ) , (77)
and
∇ ⋅ ∥ R ˙ ∥ i 1 δ p j = ∇ ∥ ∥ v ˙ ∥ i δ p j + 1 B [ ϕ D K , N i G K p j ] + 1 B ∇ ⋅ ∥ b × ∇ ⟨ ϕ G K ⟩ R ∥ i δ p j . (78) \nabla\cdot\lVert\dot{\mathbf{R}}\rVert^{\delta pj}_{i1}
=\nabla_\parallel\lVert\dot{\mathbf{v}}\rVert^{\delta pj}_{i}
+\frac{1}{B}\left[\phi_{DK},\mathcal{N}^{pj}_{iGK}\right]
+\frac{1}{B}\nabla\cdot\lVert\mathbf{b}\times\nabla\langle\phi_{GK}\rangle_R\rVert^{\delta pj}_{i}. \tag*{(78)} ∇ ⋅ ∥ R ˙ ∥ i 1 δ p j = ∇ ∥ ∥ v ˙ ∥ i δ p j + B 1 [ ϕ D K , N i G K p j ] + B 1 ∇ ⋅ ∥ b × ∇ ⟨ ϕ G K ⟩ R ∥ i δ p j . (78)
In Eq. (4.4a) we introduce the linearized (around F i D K F_{iDK} F i D K and F e D K F_{eDK} F eD K ) ion-ion and ion-electron collision operator, C i i 1 p j \mathcal{C}^{pj}_{ii1} C ii 1 p j and C i e 1 p j \mathcal{C}^{pj}_{ie1} C i e 1 p j , and the small-scale source term, $\delta S^{pj}_{i1} = ∫ d v H p j ( S i − ⟨ S i ⟩ D K ) \int\mathrm{d}v H^{pj}\left(S_i-\langle S_i\rangle_{DK}\right) ∫ d v H p j ( S i − ⟨ S i ⟩ D K ) . The expressions of the projection of the gyroaveraged quantities involving ⟨ ψ ⟩ R \langle\psi\rangle_R ⟨ ψ ⟩ R and ⟨ ϕ G K ⟩ R \langle\phi_{GK}\rangle_R ⟨ ϕ G K ⟩ R are reported in App. B. We use the GK Dougherty collision operator in Eq. (4.4a) Frei et al. [9 ]
C i i p j = ν i 0 [ − ( p + 2 j ) N i G K p j + τ i T i D K ∇ ⊥ 2 N i G K p j + ( T i D K − 1 ) × ( p ( p − 1 ) N i G K p − 2 j − 2 j N i G K p j − 1 ) ] (79) C_{ii}^{pj}=\nu_{i0}\left[-(p+2j)N_{iGK}^{pj}+\tau_i T_{iDK}\nabla_{\perp}^{2}N_{iGK}^{pj}+(T_{iDK}-1)\times\left(\sqrt{p(p-1)}N_{iGK}^{p-2j}-2jN_{iGK}^{pj-1}\right)\right] \tag*{(79)} C ii p j = ν i 0 [ − ( p + 2 j ) N i G K p j + τ i T i D K ∇ ⊥ 2 N i G K p j + ( T i D K − 1 ) × ( p ( p − 1 ) N i G K p − 2 j − 2 j N i G K p j − 1 ) ] (79)
where we approximate T i = T i D K T_i=T_{iDK} T i = T i D K .
Field equations
In order to close our system of equations, we now determine the governing equations for ϕ D K \phi_{DK} ϕ D K and ϕ G K \phi_{GK} ϕ G K . In the process, we will also determine a relation between N i D K N_{iDK} N i D K and N e D K N_{eDK} N eD K which allows us to avoid solving both Eq. (3.5a) and Eq. (3.12a) with ( p , j ) = ( 0 , 0 ) (p,j)=(0,0) ( p , j ) = ( 0 , 0 ) . For this purpose, we perform the variation of the action,
A = ∑ a ∫ d t ∫ d x ∫ d v F a Γ a , (80) \mathcal{A}=\sum_a\int\mathrm{d}t\int\mathrm{d}x\int\mathrm{d}v F_a\Gamma_a, \tag*{(80)} A = a ∑ ∫ d t ∫ d x ∫ d v F a Γ a , (80)
with respect to ϕ G K \phi_{GK} ϕ G K where quasineutrality is imposed by neglecting the field functional action term Frei et al. [9 ] , and Γ i \Gamma_i Γ i and Γ e \Gamma_e Γ e are defined in Eqs. (2) and (15) , respectively. We note that, the variation with respect to ϕ D K \phi_{DK} ϕ D K , yields an identical result in the long wavelength limit, being accurate only to order ( ρ s k ⊥ G K ) 2 (\rho_s k_{\perp GK})^2 ( ρ s k ⊥ G K ) 2 since we only allow ϕ D K \phi_{DK} ϕ D K to vary on long scales.
The variation of the action, Eq. (80) , with respect to ϕ G K \phi_{GK} ϕ G K gives the following Poisson equation,
B ∗ B N i D K − N e D K + 1 2 m i Ω i 2 ∇ ⊥ 2 P ⊥ i D K + ρ i ∗ − N e G K + ∇ G K ( ⟨ ϕ G K ⟩ R ) = O ( n e ϵ ⊥ 3 , n e ϵ ⊥ ϵ δ , n e ϵ δ 2 ) , (81) \frac{B^*}{B}N_{iDK}-N_{eDK}+\frac{1}{2m_i\Omega_i^2}\nabla_{\perp}^{2}P_{\perp iDK}+\rho_i^*-N_{eGK}+\nabla_{GK}\left(\langle\phi_{GK}\rangle_R\right)=\mathcal{O}\left(n_e\epsilon_\perp^3,n_e\epsilon_\perp\epsilon_\delta,n_e\epsilon_\delta^2\right), \tag*{(81)} B B ∗ N i D K − N eD K + 2 m i Ω i 2 1 ∇ ⊥ 2 P ⊥ i D K + ρ i ∗ − N e G K + ∇ G K ( ⟨ ϕ G K ⟩ R ) = O ( n e ϵ ⊥ 3 , n e ϵ ⊥ ϵ δ , n e ϵ δ 2 ) , (81)
where we introduce the GK Poisson operator
∇ G K ( ⟨ ϕ G K ⟩ R ) = 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ B m i ( F i D K ∂ ∂ μ ⟨ ⟨ ϕ G K ⟩ R ⟩ x † ) , (82) \nabla_{GK}\left(\langle\phi_{GK}\rangle_R\right)=2\pi\int_{-\infty}^{\infty}\mathrm{d}v_{\parallel}\int_0^{\infty}\mathrm{d}\mu\frac{B}{m_i}\left(F_{iDK}\frac{\partial}{\partial\mu}\langle\langle\phi_{GK}\rangle_R\rangle_{x}^{\dagger}\right), \tag*{(82)} ∇ G K ( ⟨ ϕ G K ⟩ R ) = 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ m i B ( F i D K ∂ μ ∂ ⟨⟨ ϕ G K ⟩ R ⟩ x † ) , (82)
and the ion GK particle density
ρ i ∗ = 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ B m i ⟨ F i G K ⟩ x † , (83) \rho_i^*=2\pi\int_{-\infty}^{\infty}\mathrm{d}v_{\parallel}\int_0^{\infty}\mathrm{d}\mu\frac{B}{m_i}\langle F_{iGK}\rangle_x^{\dagger}, \tag*{(83)} ρ i ∗ = 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ m i B ⟨ F i G K ⟩ x † , (83)
{#eq:5.2d}
and where the adjoint gyroaverage operator has been defined as
⟨ f ⟩ x † = 1 2 π ∫ 0 2 π d θ ∫ d R δ ( R + ρ − x ) f ( R , μ , v ∥ , t ) (84) \langle f\rangle_{\mathbf{x}}^\dagger= \frac{1}{2\pi}\int_0^{2\pi}d\theta\int d\mathbf{R}\,\delta(\mathbf{R}+\boldsymbol{\rho}-\mathbf{x})f(\mathbf{R},\mu,v_\parallel,t) \tag*{(84)} ⟨ f ⟩ x † = 2 π 1 ∫ 0 2 π d θ ∫ d R δ ( R + ρ − x ) f ( R , μ , v ∥ , t ) (84)
= ∑ n = 0 ∞ ( 2 μ q a Ω a ) n ∇ ⊥ 2 n 2 n ( n ! ) 2 f ( x , μ , v ∥ , t ) , (85) = \sum_{n=0}^{\infty}\left(\frac{2\mu}{q_a\Omega_a}\right)^n\frac{\nabla_\perp^{2n}}{2^n(n!)^2}f(\mathbf{x},\mu,v_\parallel,t), \tag*{(85)} = n = 0 ∑ ∞ ( q a Ω a 2 μ ) n 2 n ( n ! ) 2 ∇ ⊥ 2 n f ( x , μ , v ∥ , t ) , (85)
valid for constant B B B (in this case the gyroaverage operator is self-adjoint).
We impose that Eq. (5.2a) is satisfied independently order by order, up to leading order in ϵ ⊥ \epsilon_\perp ϵ ⊥ and ϵ δ \epsilon_\delta ϵ δ . For this purpose, we introduce the higher order DK densities such that
N a D K = N a D K 0 + N a D K 1 , (86) N_{aDK}=N_{aDK0}+N_{aDK1}, \tag*{(86)} N a D K = N a D K 0 + N a D K 1 , (86)
with N a D K 1 ∼ ϵ ⊥ 2 N a D K N_{aDK1}\sim\epsilon_\perp^2N_{aDK} N a D K 1 ∼ ϵ ⊥ 2 N a D K . This expansion removes the need of evaluating the Pfirsch-Kaufman polarization Pfirsch & Morrison [33 ] ; Kaufman [22 ] ; Frei et al. [9 ] . At order N e N_e N e , Eq. (5.2a) reads
N i D K 0 − N e D K 0 = 0 , (87) N_{iDK0}-N_{eDK0}=0, \tag*{(87)} N i D K 0 − N eD K 0 = 0 , (87)
thus we impose N i D K 0 = N e D K 0 = N D K N_{iDK0}=N_{eDK0}=N_{DK} N i D K 0 = N eD K 0 = N D K . By taking the time derivative of Eq. (87) , using Eq. (3.5a) and Eq. (3.10a) with ( p , j ) = ( 0 , 0 ) (p,j)=(0,0) ( p , j ) = ( 0 , 0 ) , we relate the electron and ion parallel velocities as follows
∇ ∥ ∫ d v ∥ v ∥ ( F i D K − F e D K ) ≡ ∇ ∥ J ∥ = O ( n e Ω i ϵ ⊥ 3 ) . (88) \nabla_\parallel\int dv_\parallel v_\parallel\left(F_{iDK}-F_{eDK}\right)\equiv\nabla_\parallel J_\parallel=\mathcal{O}\left(n_e\Omega_i\epsilon_\perp^3\right). \tag*{(88)} ∇ ∥ ∫ d v ∥ v ∥ ( F i D K − F eD K ) ≡ ∇ ∥ J ∥ = O ( n e Ω i ϵ ⊥ 3 ) . (88)
In order to satisfy Eq. (88) , we assume J ∥ = O ( c s n e ϵ ⊥ 2 ) J_\parallel=\mathcal{O}\left(c_sn_e\epsilon_\perp^2\right) J ∥ = O ( c s n e ϵ ⊥ 2 ) . This means that to the lowest order in ϵ ⊥ \epsilon_\perp ϵ ⊥ , the electrons and ions move with one fluid velocity, U ∥ U_\parallel U ∥ . Therefore, we define
m i N D K U ∥ ≡ 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ B m i v ∥ ( m e F e D K + m i F i D K ) + O ( c s ϵ ⊥ 2 ) . (89) m_iN_{DK}U_\parallel\equiv2\pi\int_{-\infty}^{\infty}dv_\parallel\int_0^\infty d\mu\frac{B}{m_i}v_\parallel\left(m_eF_{eDK}+m_iF_{iDK}\right)+\mathcal{O}\left(c_s\epsilon_\perp^2\right). \tag*{(89)} m i N D K U ∥ ≡ 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ m i B v ∥ ( m e F eD K + m i F i D K ) + O ( c s ϵ ⊥ 2 ) . (89)
We use that to the lowest order in ϵ m \epsilon_m ϵ m , U ∥ i D K = U ∥ e D K + O ( c s ϵ ⊥ 2 , c s ϵ m 2 ) U_{\parallel iDK}=U_{\parallel eDK}+\mathcal{O}\left(c_s\epsilon_\perp^2,c_s\epsilon_m^2\right) U ∥ i D K = U ∥ eD K + O ( c s ϵ ⊥ 2 , c s ϵ m 2 ) and we determine the governing equation for U ∥ U_\parallel U ∥ as a total momentum equation, adding m e N D K ∂ t U ∥ e D K + m i N D K ∂ t U ∥ i D K m_eN_{DK}\partial_tU_{\parallel eDK}+m_iN_{DK}\partial_tU_{\parallel iDK} m e N D K ∂ t U ∥ eD K + m i N D K ∂ t U ∥ i D K , as it is done in the drift-reduced Braginskii model Braginskii [4 ] ; Zeiler et al. [42 ] ; Parra [31 ] . The moment with m e v ∥ m_ev_\parallel m e v ∥ of Eq. (32) gives
m e N D K ∂ t U ∥ e D K = e N D K ∇ ∥ ϕ D K − ∇ ∥ P e D K + F e i + m e 2 v T h e S e 0 10 + O ( m i N D K c s Ω i ϵ ⊥ 3 , m e N D K c s Ω i ϵ ⊥ ) , (90) \begin{aligned}
m_eN_{DK}\partial_tU_{\parallel eDK}
&=eN_{DK}\nabla_\parallel\phi_{DK}-\nabla_\parallel P_{eDK}+F_{ei}+\frac{m_e}{\sqrt{2}}v_{Th e}S_{e0}^{10}\\
&\quad+\mathcal{O}\left(m_iN_{DK}c_s\Omega_i\epsilon_\perp^3,m_eN_{DK}c_s\Omega_i\epsilon_\perp\right), \tag*{(90)}
\end{aligned} m e N D K ∂ t U ∥ eD K = e N D K ∇ ∥ ϕ D K − ∇ ∥ P eD K + F e i + 2 m e v T h e S e 0 10 + O ( m i N D K c s Ω i ϵ ⊥ 3 , m e N D K c s Ω i ϵ ⊥ ) , (90)
and combining Eqs. (3.11a) and (5.8a) gives
m i N D K ( ∂ t U ∥ + U ∥ ∇ ∥ U ∥ + 1 B [ ϕ D K + ϕ G K , U ∥ ] ) + ∇ ∥ P ∥ i D K + ∇ ∥ P ∥ e D K = S U + O ( m i N D K c s Ω i ϵ ⊥ 3 , m e N D K c s Ω i ϵ ⊥ ϵ m 2 ) . (91) m_iN_{DK}\left(\partial_t U_{\parallel}+U_{\parallel}\nabla_{\parallel}U_{\parallel}+\frac{1}{B}\left[\phi_{DK}+\phi_{GK},U_{\parallel}\right]\right)+\nabla_{\parallel}P_{\parallel iDK}+\nabla_{\parallel}P_{\parallel eDK}=S_U+\mathcal{O}\left(m_iN_{DK}c_s\Omega_i\epsilon_{\perp}^{3},m_eN_{DK}c_s\Omega_i\epsilon_{\perp}\epsilon_m^2\right). \tag*{(91)} m i N D K ( ∂ t U ∥ + U ∥ ∇ ∥ U ∥ + B 1 [ ϕ D K + ϕ G K , U ∥ ] ) + ∇ ∥ P ∥ i D K + ∇ ∥ P ∥ eD K = S U + O ( m i N D K c s Ω i ϵ ⊥ 3 , m e N D K c s Ω i ϵ ⊥ ϵ m 2 ) . (91)
Thus ∂ t U ∥ \partial_t U_{\parallel} ∂ t U ∥ is independent of the collision frequencies and we avoid to numerically satisfy momentum conservation, m e ∫ d v v ∥ C e i 10 = − m i ∫ d v v ∥ C i e 10 m_e\int dvv_{\parallel}C_{ei}^{10}=-m_i\int dvv_{\parallel}C_{ie}^{10} m e ∫ d v v ∥ C e i 10 = − m i ∫ d v v ∥ C i e 10 , at high collisionality. In Eq. (5.8b), we define
S U = 1 2 ( m i v T h i S i 0 10 + m e v T h e S e 0 10 ) . (92) S_U=\frac{1}{\sqrt{2}}\left(m_iv_{T hi}S_{i0}^{10}+m_ev_{T he}S_{e0}^{10}\right). \tag*{(92)} S U = 2 1 ( m i v T hi S i 0 10 + m e v T h e S e 0 10 ) . (92)
We evolve U ∥ U_{\parallel} U ∥ through Eq. (5.8b) and impose U ∥ i = U ∥ e = U ∥ U_{\parallel i}=U_{\parallel e}=U_{\parallel} U ∥ i = U ∥ e = U ∥ in Eqs. (3.5a), (3.5b), and (3.12a) in our simulations. Furthermore, Eq. (3.5a) is identical to Eq. (3.12a) with ( p , j ) = ( 0 , 0 ) (p,j)=(0,0) ( p , j ) = ( 0 , 0 ) so we only evolve Eq. (3.5a) and impose N i D K 00 = N e D K = N D K N_{iDK}^{00}=N_{eDK}=N_{DK} N i D K 00 = N eD K = N D K . Finally, as an aside, we note that the implicit condition J ∥ = O ( n e c s ϵ ⊥ 2 , n e c s ϵ m 2 ) J_{\parallel}=\mathcal{O}\left(n_ec_s\epsilon_{\perp}^{2},n_ec_s\epsilon_m^2\right) J ∥ = O ( n e c s ϵ ⊥ 2 , n e c s ϵ m 2 ) implies parallel force balance given as
e m e ∇ ∥ ϕ D K − 1 m e ∇ ∥ P e D K + v T h e 2 S e 0 10 − 1 m e F e i − v T h i 2 S i 0 10 + 1 m i ∇ ∥ P ∥ i D K = O ( N D K c s Ω i ϵ ⊥ 3 , N D K c s Ω i ϵ ⊥ ϵ m 2 ) . (93) \frac{e}{m_e}\nabla_{\parallel}\phi_{DK}-\frac{1}{m_e}\nabla_{\parallel}P_{eDK}+\frac{v_{T he}}{\sqrt{2}}S_{e0}^{10}-\frac{1}{m_e}F_{ei}-\frac{v_{T hi}}{\sqrt{2}}S_{i0}^{10}+\frac{1}{m_i}\nabla_{\parallel}P_{\parallel iDK}=\mathcal{O}\left(N_{DK}c_s\Omega_i\epsilon_{\perp}^{3},N_{DK}c_s\Omega_i\epsilon_{\perp}\epsilon_m^2\right). \tag*{(93)} m e e ∇ ∥ ϕ D K − m e 1 ∇ ∥ P eD K + 2 v T h e S e 0 10 − m e 1 F e i − 2 v T hi S i 0 10 + m i 1 ∇ ∥ P ∥ i D K = O ( N D K c s Ω i ϵ ⊥ 3 , N D K c s Ω i ϵ ⊥ ϵ m 2 ) . (93)
In other words, in order for quasi-neutrality to be satisfied to a certain order in ϵ ⊥ \epsilon_{\perp} ϵ ⊥ , the acceleration of ions and electrons must balance to the same order.
To obtain an equation for ϕ D K \phi_{DK} ϕ D K , we have to consider the order ϵ ⊥ 2 \epsilon_{\perp}^{2} ϵ ⊥ 2 terms of Eq. (5.2a), that is
N e D K 1 − N i D K 1 = 1 B Ω i N D K ∇ ⊥ 2 ϕ D K + 1 2 m i Ω i 2 ∇ ⊥ 2 P ⊥ i D K . (94) N_{eDK1}-N_{iDK1}=\frac{1}{B\Omega_i}N_{DK}\nabla_{\perp}^{2}\phi_{DK}+\frac{1}{2m_i\Omega_i^{2}}\nabla_{\perp}^{2}P_{\perp iDK}. \tag*{(94)} N eD K 1 − N i D K 1 = B Ω i 1 N D K ∇ ⊥ 2 ϕ D K + 2 m i Ω i 2 1 ∇ ⊥ 2 P ⊥ i D K . (94)
In order to avoid a closure problem in ϵ ⊥ \epsilon_{\perp} ϵ ⊥ , which arises by evolving N e D K 1 N_{eDK1} N eD K 1 and N i D K 1 N_{iDK1} N i D K 1 independently, we define the vorticity,
Ω ≡ N e D K 1 − N i D K 1 + 1 B Ω i ∇ ⊥ N D K ⋅ ∇ ⊥ ϕ D K + 1 2 m i Ω i 2 ∇ ⊥ 2 P ⊥ i D K (95) \Omega\equiv N_{eDK1}-N_{iDK1}+\frac{1}{B\Omega_i}\nabla_{\perp}N_{DK}\cdot\nabla_{\perp}\phi_{DK}+\frac{1}{2m_i\Omega_i^{2}}\nabla_{\perp}^{2}P_{\perp iDK} \tag*{(95)} Ω ≡ N eD K 1 − N i D K 1 + B Ω i 1 ∇ ⊥ N D K ⋅ ∇ ⊥ ϕ D K + 2 m i Ω i 2 1 ∇ ⊥ 2 P ⊥ i D K (95)
as is custom in fluid approaches Zeiler et al. [42 ] . Using Eq. (94) , Ω \Omega Ω can also be written as
Ω = 1 B Ω i ∇ ⋅ ( N D K ∇ ⊥ ϕ D K + 1 q i Ω i ∇ ⊥ P ⊥ i D K ) . (96) \Omega=\frac{1}{B\Omega_i}\nabla\cdot\left(N_{DK}\nabla_{\perp}\phi_{DK}+\frac{1}{q_i\Omega_i}\nabla_{\perp}P_{\perp iDK}\right). \tag*{(96)} Ω = B Ω i 1 ∇ ⋅ ( N D K ∇ ⊥ ϕ D K + q i Ω i 1 ∇ ⊥ P ⊥ i D K ) . (96)
The addition of the pressure contribution in Eq. (95) makes Ω \Omega Ω independent of collisions. Indeed, the particle density is independent of collision such that ∂ t n i \partial_t n_i ∂ t n i , with
n i ≡ 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ B m i ⟨ F i D K ⟩ x † = N D K + N i D K 1 + 1 2 m i Ω i 2 ∇ ⊥ 2 P ⊥ i D K + O ( N D K ϵ ⊥ 4 ) . n_i\equiv2\pi\int_{-\infty}^{\infty}dv_{\parallel}\int_{0}^{\infty}d\mu\,\frac{B}{m_i}\left\langle F_{iDK}\right\rangle_x^{\dagger}=N_{DK}+N_{iDK1}+\frac{1}{2m_i\Omega_i^{2}}\nabla_{\perp}^{2}P_{\perp iDK}+\mathcal{O}\left(N_{DK}\epsilon_{\perp}^{4}\right). n i ≡ 2 π ∫ − ∞ ∞ d v ∥ ∫ 0 ∞ d μ m i B ⟨ F i D K ⟩ x † = N D K + N i D K 1 + 2 m i Ω i 2 1 ∇ ⊥ 2 P ⊥ i D K + O ( N D K ϵ ⊥ 4 ) .
is independent of collisions, but not ∂ t N i D K 1 \partial_t N_{iDK1} ∂ t N i D K 1 on its own.
We differentiate each term of Eq. (95) with respect to t t t . For this purpose, we note that, by using the Boltzmann equation, Eq. (9) , averaged over small-scale fluctuations but including ϵ ⊥ 3 \epsilon_{\perp}^{3} ϵ ⊥ 3 terms, we have
∂ t N i D K 1 = − ∇ ∥ ( U ∥ N i D K 1 + N D K U i ∥ D K 1 ) − 1 B [ ϕ D K + ϕ G K , N i D K 1 ] − 1 B [ ϕ D K 1 + ϕ G K 1 , N D K ] + S N 1 + 1 B Ω i S N ∇ ⊥ 2 ϕ D K − 1 B Ω i ∂ t ( N D K ∇ ⊥ 2 ϕ D K ) + 1 B 2 Ω i ∇ ⊥ ⋅ ( N D K ∇ ⊥ ∂ t ϕ D K ) − 1 B Ω i ∇ ⊥ 2 ϕ D K ∇ ∥ ( N D K U ∥ ) − m i 2 B 3 [ ∣ ∇ ⊥ ϕ D K ∣ 2 , N D K ] − 1 2 m i Ω i 2 B [ ∇ ⊥ 2 ϕ D K , P ⊥ i D K ] + 1 B Ω i ∇ ∥ ∇ ⊥ ϕ D K ⋅ ∇ ⊥ ( N D K U ∥ ) + 1 2 m i Ω i 2 ∇ ⊥ 2 ∥ μ B ( S i + C i i ) / F i D K ∥ , (97) \begin{aligned}
\partial_t N_{iDK1}={}&-\nabla_{\parallel}\left(U_{\parallel}N_{iDK1}+N_{DK}U_{i\parallel DK1}\right)-\frac{1}{B}\left[\phi_{DK}+\phi_{GK},\mathcal{N}_{iDK1}\right]\\
&-\frac{1}{B}\left[\phi_{DK1}+\phi_{GK1},N_{DK}\right]+S_{N1}+\frac{1}{B\Omega_i}S_N\nabla_{\perp}^{2}\phi_{DK}-\frac{1}{B\Omega_i}\partial_t\left(N_{DK}\nabla_{\perp}^{2}\phi_{DK}\right)\\
&+\frac{1}{B^{2}\Omega_i}\nabla_{\perp}\cdot\left(N_{DK}\nabla_{\perp}\partial_t\phi_{DK}\right)-\frac{1}{B\Omega_i}\nabla_{\perp}^{2}\phi_{DK}\nabla_{\parallel}\left(N_{DK}U_{\parallel}\right)\\
&-\frac{m_i}{2B^{3}}\left[\left|\nabla_{\perp}\phi_{DK}\right|^{2},N_{DK}\right]-\frac{1}{2m_i\Omega_i^{2}B}\left[\nabla_{\perp}^{2}\phi_{DK},P_{\perp iDK}\right]\\
&+\frac{1}{B\Omega_i}\nabla_{\parallel}\nabla_{\perp}\phi_{DK}\cdot\nabla_{\perp}\left(N_{DK}U_{\parallel}\right)+\frac{1}{2m_i\Omega_i^{2}}\nabla_{\perp}^{2}\left\|\mu B\left(S_i+C_{ii}\right)/F_{iDK}\right\|, \tag*{(97)}
\end{aligned} ∂ t N i D K 1 = − ∇ ∥ ( U ∥ N i D K 1 + N D K U i ∥ D K 1 ) − B 1 [ ϕ D K + ϕ G K , N i D K 1 ] − B 1 [ ϕ D K 1 + ϕ G K 1 , N D K ] + S N 1 + B Ω i 1 S N ∇ ⊥ 2 ϕ D K − B Ω i 1 ∂ t ( N D K ∇ ⊥ 2 ϕ D K ) + B 2 Ω i 1 ∇ ⊥ ⋅ ( N D K ∇ ⊥ ∂ t ϕ D K ) − B Ω i 1 ∇ ⊥ 2 ϕ D K ∇ ∥ ( N D K U ∥ ) − 2 B 3 m i [ ∣ ∇ ⊥ ϕ D K ∣ 2 , N D K ] − 2 m i Ω i 2 B 1 [ ∇ ⊥ 2 ϕ D K , P ⊥ i D K ] + B Ω i 1 ∇ ∥ ∇ ⊥ ϕ D K ⋅ ∇ ⊥ ( N D K U ∥ ) + 2 m i Ω i 2 1 ∇ ⊥ 2 ∥ μ B ( S i + C ii ) / F i D K ∥ , (97)
where the 1 subscript indicates order ϵ ⊥ 2 \epsilon_{\perp}^{2} ϵ ⊥ 2 corrections. The last term in Eq. (5.14a) comes from the sources and collisions acting on the particle position and not gyrocenter position. For the time derivative of the higher order electron density, we get
∂ t N e D K 1 = − ∇ ∥ ( U ∥ N e D K 1 + N D K U e ∥ D K 1 ) − 1 B [ ϕ D K + ϕ G K , N i D K e ] − 1 B [ ϕ D K 1 + ϕ G K 1 , N D K ] + S N 1 , (98) \begin{aligned}
\partial_t N_{eDK1}={}&-\nabla_{\parallel}\left(U_{\parallel}N_{eDK1}+N_{DK}U_{e\parallel DK1}\right)-\frac{1}{B}\left[\phi_{DK}+\phi_{GK},\mathcal{N}_{iDKe}\right]\\
&-\frac{1}{B}\left[\phi_{DK1}+\phi_{GK1},N_{DK}\right]+S_{N1}, \tag*{(98)}
\end{aligned} ∂ t N eD K 1 = − ∇ ∥ ( U ∥ N eD K 1 + N D K U e ∥ D K 1 ) − B 1 [ ϕ D K + ϕ G K , N i D K e ] − B 1 [ ϕ D K 1 + ϕ G K 1 , N D K ] + S N 1 , (98)
using Eq. (3.12a) and Eq. (3.9d), the time derivative of the pressure term in Eq. (95) is given by
∂ t ∇ ⊥ 2 P ⊥ i D K = − ∇ ∥ ∇ ⊥ 2 ( P ⊥ i D K U ∥ + Q ⊥ i D K ) − 1 B ∇ ⊥ 2 [ ϕ D K + ϕ G K , P ⊥ i D K ] + ∇ ⊥ 2 ∥ μ B ( S i + C i i ) / F i D K ∥ , (99) \partial_t\nabla_{\perp}^{2}P_{\perp iDK}=-\nabla_{\parallel}\nabla_{\perp}^{2}\left(P_{\perp iDK}U_{\parallel}+Q_{\perp iDK}\right)-\frac{1}{B}\nabla_{\perp}^{2}\left[\phi_{DK}+\phi_{GK},P_{\perp iDK}\right]+\nabla_{\perp}^{2}\left\|\mu B\left(S_i+C_{ii}\right)/F_{iDK}\right\|, \tag*{(99)} ∂ t ∇ ⊥ 2 P ⊥ i D K = − ∇ ∥ ∇ ⊥ 2 ( P ⊥ i D K U ∥ + Q ⊥ i D K ) − B 1 ∇ ⊥ 2 [ ϕ D K + ϕ G K , P ⊥ i D K ] + ∇ ⊥ 2 ∥ μ B ( S i + C ii ) / F i D K ∥ , (99)
with Q ⊥ i D K = ∥ v ∥ μ B ∥ i = − v T h i T i 0 N i D K 1 / 2 Q_{\perp iDK}=\lVert v\rVert\mu B\lVert_i=-v_T h_iT_{i0}N^1_{iDK}/\sqrt{2} Q ⊥ i D K = ∥ v ∥ μ B ∥ i = − v T h i T i 0 N i D K 1 / 2 . Adding Eqs. (5.14a), (5.14b), and (5.14c) while using Eq. (95) , we notice that all terms including a time derivative combine to ∇ ⊥ ⋅ ( ∂ t N D K ∇ ⊥ ϕ D K ) \nabla_\perp\cdot(\partial_tN_{DK}\nabla_\perp\phi_{DK}) ∇ ⊥ ⋅ ( ∂ t N D K ∇ ⊥ ϕ D K ) . Finally, using Eq. (3.5a), we get
∂ t Ω = ∇ ∥ ( J ∥ − ( N e D K 1 − N i D K 1 ) U ∥ − 1 2 m i Ω i 2 ∇ ⊥ 2 ( P ⊥ i D K U ∥ + Q ⊥ i D K ) ) − 1 B [ ϕ D K + ϕ G K , N e D K 1 − N i D K 1 ] − 1 2 m i Ω i 2 B 2 ∇ ⊥ 2 [ ϕ D K + ϕ G K , P ⊥ i D K ] − 1 B Ω i ∇ ⊥ ⋅ [ ∇ ⊥ ϕ D K ( ∇ ∥ ( U ∥ N D K ) + 1 B [ ϕ D K + ϕ G K , N D K ] − S N ) ] + 1 B Ω i ∇ ⊥ ∂ t ϕ D K ⋅ ∇ ⊥ N D K + m i 2 B 3 [ ∣ ∇ ⊥ ϕ D K ∣ 2 , N D K ] − 1 2 m i Ω i 2 B [ ∇ ⊥ 2 ϕ D K , P ⊥ i D K ] − 1 B Ω i ∇ ⊥ ⋅ ( N D K ∇ ⊥ ∂ t ϕ D K ) + 1 B Ω i ( N D K ∇ ⊥ 2 ∂ t ϕ D K ) + 1 B Ω i S N ∇ ⊥ 2 ϕ D K , (100) \begin{aligned}
\partial_t\Omega={}&\nabla_\parallel\left(J_\parallel-(N_{eDK1}-N_{iDK1})U_\parallel-\frac{1}{2m_i\Omega_i^2}\nabla_\perp^2(P_{\perp iDK}U_\parallel+Q_{\perp iDK})\right)\\
&-\frac{1}{B}[\phi_{DK}+\phi_{GK},N_{eDK1}-N_{iDK1}]-\frac{1}{2m_i\Omega_i^2B^2}\nabla_\perp^2[\phi_{DK}+\phi_{GK},P_{\perp iDK}]\\
&-\frac{1}{B\Omega_i}\nabla_\perp\cdot\left[\nabla_\perp\phi_{DK}\left(\nabla_\parallel(U_\parallel N_{DK})+\frac{1}{B}[\phi_{DK}+\phi_{GK},N_{DK}]-S_N\right)\right]\\
&+\frac{1}{B\Omega_i}\nabla_\perp\partial_t\phi_{DK}\cdot\nabla_\perp N_{DK}+\frac{m_i}{2B^3}[\lvert\nabla_\perp\phi_{DK}\rvert^2,N_{DK}]\\
&-\frac{1}{2m_i\Omega_i^2B}[\nabla_\perp^2\phi_{DK},P_{\perp iDK}]-\frac{1}{B\Omega_i}\nabla_\perp\cdot(N_{DK}\nabla_\perp\partial_t\phi_{DK})\\
&+\frac{1}{B\Omega_i}(N_{DK}\nabla_\perp^2\partial_t\phi_{DK})+\frac{1}{B\Omega_i}S_N\nabla_\perp^2\phi_{DK}, \tag*{(100)}
\end{aligned} ∂ t Ω = ∇ ∥ ( J ∥ − ( N eD K 1 − N i D K 1 ) U ∥ − 2 m i Ω i 2 1 ∇ ⊥ 2 ( P ⊥ i D K U ∥ + Q ⊥ i D K ) ) − B 1 [ ϕ D K + ϕ G K , N eD K 1 − N i D K 1 ] − 2 m i Ω i 2 B 2 1 ∇ ⊥ 2 [ ϕ D K + ϕ G K , P ⊥ i D K ] − B Ω i 1 ∇ ⊥ ⋅ [ ∇ ⊥ ϕ D K ( ∇ ∥ ( U ∥ N D K ) + B 1 [ ϕ D K + ϕ G K , N D K ] − S N ) ] + B Ω i 1 ∇ ⊥ ∂ t ϕ D K ⋅ ∇ ⊥ N D K + 2 B 3 m i [∣ ∇ ⊥ ϕ D K ∣ 2 , N D K ] − 2 m i Ω i 2 B 1 [ ∇ ⊥ 2 ϕ D K , P ⊥ i D K ] − B Ω i 1 ∇ ⊥ ⋅ ( N D K ∇ ⊥ ∂ t ϕ D K ) + B Ω i 1 ( N D K ∇ ⊥ 2 ∂ t ϕ D K ) + B Ω i 1 S N ∇ ⊥ 2 ϕ D K , (100)
with J ∥ = N D K ( U ∥ i D K 1 − U ∥ e D K 1 ) J_\parallel=N_{DK}(U_{\parallel iDK1}-U_{\parallel eDK1}) J ∥ = N D K ( U ∥ i D K 1 − U ∥ eD K 1 ) . Finally, through tedious elementary vector calculus and algebra, we get
∂ t Ω = ∇ ∥ J ∥ − ∇ ∥ ( Ω U ∥ i D K ) − 1 B [ ϕ D K + ϕ G K , Ω ] − 1 2 m i Ω i 2 ∇ ∥ ( P ⊥ i D K ∇ ⊥ 2 U ∥ i D K ) − 1 B [ ∇ ⊥ ( ϕ D K + ϕ G K ) ⋅ , ω ] − ∇ ∥ ( ω ⋅ ∇ ⊥ U ∥ i D K ) + 1 B Ω i ∇ ⊥ S N ⋅ ∇ ⊥ ϕ D K + 1 B 2 Ω i [ ∇ ⊥ ϕ D K ⋅ , N i D K ∇ ⊥ ϕ G K ] + 1 2 m i Ω i 2 [ ∇ ⊥ 2 ϕ G K , P ⊥ i D K ] − 1 2 m i Ω i 2 ∇ ∥ ( ∇ ⊥ 2 Q ⊥ i D K ) , (101) \begin{aligned}
\partial_t\Omega={}&\nabla_\parallel J_\parallel-\nabla_\parallel(\Omega U_{\parallel iDK})-\frac{1}{B}[\phi_{DK}+\phi_{GK},\Omega]-\frac{1}{2m_i\Omega_i^2}\nabla_\parallel(P_{\perp iDK}\nabla_\perp^2U_{\parallel iDK})\\
&-\frac{1}{B}[\nabla_\perp(\phi_{DK}+\phi_{GK})\cdot,\boldsymbol{\omega}]-\nabla_\parallel(\boldsymbol{\omega}\cdot\nabla_\perp U_{\parallel iDK})+\frac{1}{B\Omega_i}\nabla_\perp S_N\cdot\nabla_\perp\phi_{DK}\\
&+\frac{1}{B^2\Omega_i}[\nabla_\perp\phi_{DK}\cdot,N_{iDK}\nabla_\perp\phi_{GK}]+\frac{1}{2m_i\Omega_i^2}[\nabla_\perp^2\phi_{GK},P_{\perp iDK}]\\
&-\frac{1}{2m_i\Omega_i^2}\nabla_\parallel(\nabla_\perp^2Q_{\perp iDK}), \tag*{(101)}
\end{aligned} ∂ t Ω = ∇ ∥ J ∥ − ∇ ∥ ( Ω U ∥ i D K ) − B 1 [ ϕ D K + ϕ G K , Ω ] − 2 m i Ω i 2 1 ∇ ∥ ( P ⊥ i D K ∇ ⊥ 2 U ∥ i D K ) − B 1 [ ∇ ⊥ ( ϕ D K + ϕ G K ) ⋅ , ω ] − ∇ ∥ ( ω ⋅ ∇ ⊥ U ∥ i D K ) + B Ω i 1 ∇ ⊥ S N ⋅ ∇ ⊥ ϕ D K + B 2 Ω i 1 [ ∇ ⊥ ϕ D K ⋅ , N i D K ∇ ⊥ ϕ G K ] + 2 m i Ω i 2 1 [ ∇ ⊥ 2 ϕ G K , P ⊥ i D K ] − 2 m i Ω i 2 1 ∇ ∥ ( ∇ ⊥ 2 Q ⊥ i D K ) , (101)
with
ω = 1 B Ω i N D K ∇ ⊥ ϕ D K + 1 m i Ω i 2 ∇ ⊥ P ⊥ i D K , (102) \boldsymbol{\omega}=\frac{1}{B\Omega_i}N_{DK}\nabla_\perp\phi_{DK}+\frac{1}{m_i\Omega_i^2}\nabla_\perp P_{\perp iDK}, \tag*{(102)} ω = B Ω i 1 N D K ∇ ⊥ ϕ D K + m i Ω i 2 1 ∇ ⊥ P ⊥ i D K , (102)
and as a consequence
Ω = ∇ ⊥ ⋅ ω . (103) \Omega=\nabla_\perp\cdot\boldsymbol{\omega}. \tag*{(103)} Ω = ∇ ⊥ ⋅ ω . (103)
Since Eq. (5.14e) requires J ∥ J_\parallel J ∥ to first non-vanishing order, we derive an equation for J ∥ J_\parallel J ∥ , describing it as a dynamical field. Noticing that
J ∥ ≡ N D K ( U ∥ i D K − U ∥ e D K ) = N D K ( U ∥ i D K 1 − U ∥ e D K 1 ) ∼ O ( N D K c s ϵ ⊥ 2 , N D K c s ϵ m 2 ) , (104) \begin{aligned}
J_\parallel&\equiv N_{DK}(U_{\parallel iDK}-U_{\parallel eDK})\\
&=N_{DK}(U_{\parallel iDK1}-U_{\parallel eDK1})\sim O(N_{DK}c_s\epsilon_\perp^2,N_{DK}c_s\epsilon_m^2), \tag*{(104)}
\end{aligned} J ∥ ≡ N D K ( U ∥ i D K − U ∥ eD K ) = N D K ( U ∥ i D K 1 − U ∥ eD K 1 ) ∼ O ( N D K c s ϵ ⊥ 2 , N D K c s ϵ m 2 ) , (104)
we remark that that J ∥ J_{\parallel} J ∥ is the DK current density and does not include the GK distribution functions. Taking the derivative of Eq. (5.15a), we get
∂ t J ∥ + 1 B [ ϕ D K + ϕ G K , J ∥ ] + ∇ ∥ ( J ∥ U ∥ ) + J ∥ ∇ ∥ U ∥ + e m e N D K ∇ ∥ ϕ D K + 1 m i ∇ ∥ P ∥ i D K + 1 m e ∇ ∥ ( N D K T e D K ) = S N N D K J ∥ − ν ∥ N D K m i N 0 m e J ∥ − 0.71 N D K m e ∇ ∥ T e D K + S J , (105) \begin{aligned}
\partial_t J_{\parallel} &+ \frac{1}{B}\left[\phi_{DK}+\phi_{GK},J_{\parallel}\right]+\nabla_{\parallel}\left(J_{\parallel}U_{\parallel}\right)+J_{\parallel}\nabla_{\parallel}U_{\parallel}+\frac{e}{m_e}N_{DK}\nabla_{\parallel}\phi_{DK}+\frac{1}{m_i}\nabla_{\parallel}P_{\parallel iDK}\\
&+\frac{1}{m_e}\nabla_{\parallel}\left(N_{DK}T_{eDK}\right)=\frac{S_N}{N_{DK}}J_{\parallel}-\nu_{\parallel}\frac{N_{DK}m_i}{N_0m_e}J_{\parallel}-0.71\frac{N_{DK}}{m_e}\nabla_{\parallel}T_{eDK}+S_J, \tag*{(105)}
\end{aligned} ∂ t J ∥ + B 1 [ ϕ D K + ϕ G K , J ∥ ] + ∇ ∥ ( J ∥ U ∥ ) + J ∥ ∇ ∥ U ∥ + m e e N D K ∇ ∥ ϕ D K + m i 1 ∇ ∥ P ∥ i D K + m e 1 ∇ ∥ ( N D K T eD K ) = N D K S N J ∥ − ν ∥ N 0 m e N D K m i J ∥ − 0.71 m e N D K ∇ ∥ T eD K + S J , (105)
having introduced the source term, S J ≡ ∫ d v ( ( v ∥ − U ∥ i D K ) S i − ( v ∥ − U ∥ e D K ) S e ) S_J \equiv\int\mathrm{d}v\left((v_{\parallel}-U_{\parallel iDK})S_i-(v_{\parallel}-U_{\parallel eDK})S_e\right) S J ≡ ∫ d v ( ( v ∥ − U ∥ i D K ) S i − ( v ∥ − U ∥ eD K ) S e ) and the Spitzer resistivity ν ∥ = ν 0 ( T e / T e 0 ) − 3 / 2 \nu_{\parallel}=\nu_0\left(T_e/T_{e0}\right)^{-3/2} ν ∥ = ν 0 ( T e / T e 0 ) − 3/2 .
Eqs. (5.14e) and (5.15b) are equivalent to the vorticity and current density equation for the drift-reduced Braginskii model Zeiler et al. [42 ] except for the addition of the source term, ∼ ∇ ⊥ S N \sim\nabla_{\perp}S_N ∼ ∇ ⊥ S N , arising from considering the presence of sources, and the terms ∼ ∇ ∥ ( P ⊥ i D K ∇ ⊥ 2 U ∥ i D K + ∇ ⊥ 2 Q ⊥ i D K ) \sim\nabla_{\parallel}\left(P_{\perp iDK}\nabla_{\perp}^{2}U_{\parallel iDK}+\nabla_{\perp}^{2}Q_{\perp iDK}\right) ∼ ∇ ∥ ( P ⊥ i D K ∇ ⊥ 2 U ∥ i D K + ∇ ⊥ 2 Q ⊥ i D K ) in Eq. (5.14e) due to the distinction between particle and gyrocenter parallel velocity. In this work, we use the gyrocenter parallel velocity while in drift-reduced Braginskii equations, the particle parallel velocity is used. We remark that the term T e D K ∇ ∥ J ∥ / N e D K ∼ ϵ ⊥ 2 ∂ t T e T_{eDK}\nabla_{\parallel}J_{\parallel}/N_{eDK}\sim\epsilon_{\perp}^{2}\partial_tT_e T eD K ∇ ∥ J ∥ / N eD K ∼ ϵ ⊥ 2 ∂ t T e in Eq. (3.5b) is of higher order, therefore we choose to neglect it in our simulations.
Finally, in order to obtain an equation for ϕ G K \phi_{GK} ϕ G K , we evaluate the order ϵ δ \epsilon_{\delta} ϵ δ terms in Eq. (5.2a) and insert N e G K N_{eGK} N e G K from Eq. (64)
ρ i ∗ = e ϕ G K T e D K N D K − q i B ∥ ∂ μ ⟨ ⟨ ϕ G K ⟩ R ⟩ x † ∥ i ∗ . (106) \rho_i^{*}=\frac{e\phi_{GK}}{T_{eDK}}N_{DK}-\frac{q_i}{B}\left\|\partial_{\mu}\left\langle\left\langle\phi_{GK}\right\rangle_R\right\rangle_x^{\dagger}\right\|_i^{*}. \tag*{(106)} ρ i ∗ = T eD K e ϕ G K N D K − B q i ∂ μ ⟨ ⟨ ϕ G K ⟩ R ⟩ x † i ∗ . (106)
We note that the left hand side of Eq. (106) only depends on N i G K p j \mathcal{N}_{iGK}^{pj} N i G K p j while the right hand side is an operator, depending on the DK fields, acting on ϕ G K \phi_{GK} ϕ G K . In the limit where the DK distribution functions are constant Maxwellians, Eq. (106) reduces to the standard δ \delta δ -f GK Poisson equation
ρ i ∗ = e N D K [ 1 T e D K + 1 T i D K − 1 T i D K exp ( − T i D K k ⊥ 2 m i Ω i 2 ) I 0 ( T i D K k ⊥ 2 m i Ω i 2 ) ] ϕ G K , (107) \rho_i^{*}=eN_{DK}\left[\frac{1}{T_{eDK}}+\frac{1}{T_{iDK}}-\frac{1}{T_{iDK}}\exp\left(-\frac{T_{iDK}k_{\perp}^{2}}{m_i\Omega_i^{2}}\right)I_0\left(\frac{T_{iDK}k_{\perp}^{2}}{m_i\Omega_i^{2}}\right)\right]\phi_{GK}, \tag*{(107)} ρ i ∗ = e N D K [ T eD K 1 + T i D K 1 − T i D K 1 exp ( − m i Ω i 2 T i D K k ⊥ 2 ) I 0 ( m i Ω i 2 T i D K k ⊥ 2 ) ] ϕ G K , (107)
with I 0 I_0 I 0 the modified Bessel function of order zero.
Simulation setup and numerical implementation
In order to implement the system of Eqs. (3.5a), (3.5b), (3.10a) for ( p , j ) ≠ ( 0 , 0 ) (p,j)\ne(0,0) ( p , j ) = ( 0 , 0 ) and ( p , j ) ≠ ( 1 , 0 ) (p,j)\ne(1,0) ( p , j ) = ( 1 , 0 ) , (4.4a), (5.8b), (5.14e), (5.14g), (5.15b), and (5.16), closed by enforcing N i D K 00 = N e D K = N D K \mathcal{N}_{iDK}^{00}=N_{eDK}=N_{DK} N i D K 00 = N eD K = N D K , N i D K 10 = 0 \mathcal{N}_{iDK}^{10}=0 N i D K 10 = 0 , and U ∥ i D K = U ∥ e D K = U ∥ U_{\parallel iDK}=U_{\parallel eDK}=U_{\parallel} U ∥ i D K = U ∥ eD K = U ∥ , we normalize time, t t t , to R / c s 0 R/c_{s0} R / c s 0 (c s 0 = T e 0 / m i c_{s0}=\sqrt{T_{e0}/m_i} c s 0 = T e 0 / m i is the ion sound speed at the constant reference electron temperature, T e 0 T_{e0} T e 0 , and R R R is the characteristic length that indicates the size of the plasma chamber in the direction perpendicular to B \mathbf{B} B . The electrostatic potentials, ϕ D K \phi_{DK} ϕ D K and ϕ G K \phi_{GK} ϕ G K , are normalized to T e 0 / e T_{e0}/e T e 0 / e , the parallel (to B \mathbf{B} B ) spatial scale to R R R , and the perpendicular ones to ρ s 0 = c s 0 / Ω i \rho_{s0}=c_{s0}/\Omega_i ρ s 0 = c s 0 / Ω i . The densities, N D K N_{DK} N D K and N G K N_{GK} N G K , are normalized to the constant reference density N 0 N_0 N 0 , the parallel fluid velocities, U ∥ U_{\parallel} U ∥ , to c s 0 c_{s0} c s 0 , the electron and ion temperatures, T e T_e T e , T ∥ i T_{\parallel i} T ∥ i , and T ⊥ i T_{\perp i} T ⊥ i , to T e 0 T_{e0} T e 0 , T i 0 T_{i0} T i 0 , and T i 0 T_{i0} T i 0 , respectively. Finally, we normalize the parallel current density, J ∥ J_{\parallel} J ∥ , to N 0 c s 0 N_0c_{s0} N 0 c s 0 .
The domain we consider is a cuboid with sides L x L_x L x , L y L_y L y , and L z L_z L z . We introduce the spatial coordinates ( x , y , z ) (x,y,z) ( x , y , z ) , where z z z is parallel to the magnetic field of the device, and x x x and y y y are perpendicular to z z z , forming a right-handed coordinate system.
To evolve the DK quantities, N i D K p j \mathcal{N}_{iDK}^{pj} N i D K p j , U ∥ U_{\parallel} U ∥ , N D K N_{DK} N D K , T e D K T_{eDK} T eD K , ϕ D K \phi_{DK} ϕ D K , Ω \Omega Ω and J ∥ J_{\parallel} J ∥ , we use a finite-difference scheme similar to the one described in Ref. Frei et al. [10 ] ; Mencke & Ricci [25 ] . The x x x direction is discretized in N x N_x N x equally sized grid points such that the grid spacing is Δ x = L x / N x \Delta x=L_x/N_x Δ x = L x / N x . An equivalent discretization is used for y y y and z z z . Similarly to the GBS code Giacomin et al. [14 ] , derivatives are calculated with a fourth order central difference scheme, except for the Poisson bracket operator, [ f , g ] [f,g] [ f , g ] , which is evaluated using the fourth-order Arakawa algorithm Arakawa [1 ] . A staggered grid approach Paruta et al. [32 ] is used in the z z z direction with U ∥ U_{\parallel} U ∥ , J ∥ J_{\parallel} J ∥ , and the N i D K p j \mathcal{N}_{iDK}^{pj} N i D K p j with odd p p p being evolved on grid points shifted by a distance Δ z / 2 \Delta z/2 Δ z /2 with respect to the grid points for N D K N_{DK} N D K , T e D K T_{eDK} T eD K , ϕ D K \phi_{DK} ϕ D K , Ω \Omega Ω , and N i D K p j \mathcal{N}_{iDK}^{pj} N i D K p j with even p p p .
To evolve the GK fields, N i G K p j \mathcal{N}_{iGK}^{pj} N i G K p j and ϕ G K \phi_{GK} ϕ G K , we use a spectral method in the ( x , y ) (x,y) ( x , y ) plane, and finite differences along z z z . More precisely, we evolve N k x × ⌊ N k y / 2 + 1 ⌋ N_{kx}\times\lfloor N_{ky}/2+1\rfloor N k x × ⌊ N k y /2 + 1 ⌋ Fourier modes with wavevectors k x k_x k x and k y k_y k y with spacing Δ k x = 2 π / L x \Delta k_x=2\pi/L_x Δ k x = 2 π / L x and Δ k y = 2 π / L y \Delta k_y=2\pi/L_y Δ k y = 2 π / L y , respectively. Non-linear terms are evaluated in real space by inverse Fourier transforming the fields, multiplying them, and Fourier transforming back. All Fourier transforms are performed with the FFTW library Frigo & Johnson [12 ] . We note that ϕ G K \phi_{GK} ϕ G K and N i G K p j \mathcal{N}_{iGK}^{pj} N i G K p j with even p p p are evolved on the same z z z -coordinates as ϕ D K \phi_{DK} ϕ D K , while the N i G K p j \mathcal{N}_{iGK}^{pj} N i G K p j with odd p p p are evolved on the same z z z -coordinates as U ∥ U_{\parallel} U ∥ . Finally, all fields are antialiased using the 2/3 rule Hoffmann et al. [20 ] ; Orszag [28 ] .
For the DK and GK gyromoments, we close the hierarchy using a simple truncation approach. Thus, we evolve N i D K p j \mathcal{N}_{iDK}^{pj} N i D K p j with p p p from 0 to P D K P_{DK} P D K and with j j j from 0 to J D K J_{DK} J D K for the DK moments. Similarly, for the GK moments, we evolve N i G K p j \mathcal{N}_{iGK}^{pj} N i G K p j with p p p from 0 to P G K P_{GK} P G K and with j j j from 0 to J G K J_{GK} J G K .
Similarly to Mencke & Ricci [25 ] , a fifth-order adaptive time-step Runge–Kutta scheme is used to advance all fields Press et al. [35 ] . Equation (5.14g) is solved for ϕ D K \phi_{DK} ϕ D K using an approach similar to the one used in the two-fluid GBS code Giacomin et al. [14 ] . The discrete operator of the ∇ ⋅ ( N D K ∇ ϕ D K ) \nabla\cdot(N_{DK}\nabla\phi_{DK}) ∇ ⋅ ( N D K ∇ ϕ D K ) term is constructed at each time step and the linear system is solved for ϕ D K \phi_{DK} ϕ D K using the PETSCs library Balay et al. [2 ] . Equation (106) is solved for ϕ G K \phi_{GK} ϕ G K using an iterative approach. Since the electron adiabatic response generally dominates the right hand side of Eq. (106) , for each z z z plane we solve
ϕ G K n + 1 = T e D K N e D K ( ρ i ∗ + q i B ∥ ∂ μ ⟨ ⟨ ϕ G K ⟩ R x ⃗ † ⟩ i ∥ ) \phi_{GK}^{n+1} = \frac{T_{eDK}}{N_{eDK}}\left(\rho_i^* + \frac{q_i}{B}\left\Vert\partial_\mu\left\langle\left\langle\phi_{GK}\right\rangle_R \vec{x}^{\dagger}\right\rangle_i\right\Vert\right) ϕ G K n + 1 = N eD K T eD K ( ρ i ∗ + B q i ∂ μ ⟨ ⟨ ϕ G K ⟩ R x † ⟩ i )
while:
max k x , k y ∣ ϕ G K n + 1 − ϕ G K n ∣ > η G K , (108) \max_{k_x,k_y}\left|\phi_{GK}^{n+1}-\phi_{GK}^{n}\right| > \eta_{GK}, \tag*{(108)} k x , k y max ϕ G K n + 1 − ϕ G K n > η G K , (108)
and initialize ϕ G K 0 \phi_{GK}^{0} ϕ G K 0 using ϕ G K \phi_{GK} ϕ G K from the previous timestep. The parameter η G K > 0 \eta_{GK} > 0 η G K > 0 is the error threshold chosen small enough not to impact the results (typically η G K ∼ 10 − 7 \eta_{GK} \sim10^{-7} η G K ∼ 1 0 − 7 ).
For stability reasons, numerical diffusion is added to the right-hand side of all equations for all evolved DK fields, f f f , that is
D ( f ) = η ⊥ ( ∂ x 2 + ∂ y 2 ) f + η z ∂ z 2 f . (109) D(f) = \eta_{\perp}\left(\partial_x^2+\partial_y^2\right)f+\eta_z\partial_z^2f. \tag*{(109)} D ( f ) = η ⊥ ( ∂ x 2 + ∂ y 2 ) f + η z ∂ z 2 f . (109)
The numerical parameter η ⊥ > 0 \eta_{\perp} > 0 η ⊥ > 0 and η z > 0 \eta_z > 0 η z > 0 are chosen as small as possible to guarantee the numerical stability of the simulations, without significantly affecting the results. For the GK field, we do not add numerical diffusion in the perpendicular plane, i.e. η ⊥ = 0 \eta_{\perp} = 0 η ⊥ = 0 . However, we do include numerical diffusion in the z z z -direction.
Following the approach of Frei et al. [10 ] and Mencke & Ricci [25 ] , we use Bohm fluid boundary conditions for U ∥ U_{\parallel} U ∥ and J ∥ J_{\parallel} J ∥ Loizu et al. [23 ] ; ); Mosetto et al. [27 ] at the sheath entrances
U ∥ ( z = ± L z / 2 ) = ± c s = ± T e D K + τ i T i D K , (110) \begin{aligned}
U_{\parallel}(z=\pm L_z/2) &= \pm c_s \\
&= \pm\sqrt{T_{eDK}+\tau_iT_{iDK}}, \tag*{(110)}
\end{aligned} U ∥ ( z = ± L z /2 ) = ± c s = ± T eD K + τ i T i D K , (110)
J ∥ ( z = ± L z / 2 ) = ± N D K ( c s − T e D K e Λ − ϕ D K / T e D K ) , (111) J_{\parallel}(z=\pm L_z/2)=\pm N_{DK}\left(c_s-\sqrt{T_{eDK}}e^{\Lambda-\phi_{DK}/T_{eDK}}\right), \tag*{(111)} J ∥ ( z = ± L z /2 ) = ± N D K ( c s − T eD K e Λ − ϕ D K / T eD K ) , (111)
while homogeneous Neumann boundary conditions are used for the remaining boundaries for J ∥ J_{\parallel} J ∥ and U ∥ U_{\parallel} U ∥ . For all other DK fields, homogeneous Neumann boundary conditions are applied at all boundaries while we apply homogeneous Neumann boundary conditions in the parallel direction for the GK fields and periodic boundary conditions in the perpendicular plane.
The plasma sources are chosen as
S i = A N ( x ) F M i + A E ( x ) ( s ∥ i 2 + x i − 3 2 ) N D K F M i , (112) S_i = A_N(\boldsymbol{x})F_{Mi} + A_E(\boldsymbol{x})\left(s_{\parallel i}^{2} + x_i - \frac{3}{2}\right)N_{DK}F_{Mi}, \tag*{(112)} S i = A N ( x ) F M i + A E ( x ) ( s ∥ i 2 + x i − 2 3 ) N D K F M i , (112)
where ⟨ ⟨ A N ( x ) ⟩ R ⟩ D K = A N \langle\langle A_N(\boldsymbol{x})\rangle_R\rangle_{DK}=A_N ⟨⟨ A N ( x ) ⟩ R ⟩ D K = A N is assumed and analogously for A E A_E A E . Therefore, we order S i S_i S i on the DK timescale (S i ∼ ∂ t F i D K S_i \sim\partial_t F_{iDK} S i ∼ ∂ t F i D K ). The projection in the DK hierarchy, Eq. (3.12a), is
S i 0 p j = A N ( R ) δ 00 p j + δ 02 p j 2 N D K A E ( R ) − δ 01 p j N D K A E ( R ) , (113) S_{i0}^{pj} = A_N(\boldsymbol{R})\delta_{00}^{pj} + \frac{\delta_{02}^{pj}}{\sqrt{2}}N_{DK}A_E(\boldsymbol{R}) - \delta_{01}^{pj}N_{DK}A_E(\boldsymbol{R}), \tag*{(113)} S i 0 p j = A N ( R ) δ 00 p j + 2 δ 02 p j N D K A E ( R ) − δ 01 p j N D K A E ( R ) , (113)
and the projections in the GK hierarchy, Eq. (4.4a), is given by
δ S i 1 p j = δ 0 p ( K j ( b i ) − δ 0 j ) A N + δ 2 p 2 N D K ( K j ( b i ) − δ 0 j ) A E \delta S_{i1}^{pj} = \delta_0^p\left(K_j(b_i)-\delta_0^j\right)A_N + \frac{\delta_2^p}{\sqrt{2}}N_{DK}\left(K_j(b_i)-\delta_0^j\right)A_E δ S i 1 p j = δ 0 p ( K j ( b i ) − δ 0 j ) A N + 2 δ 2 p N D K ( K j ( b i ) − δ 0 j ) A E
+ δ 0 p N D K ( 2 j K j ( b i ) − ( j + 1 ) K j + 1 ( b i ) − j K j − 1 ( b i ) + δ 1 j ) A E . (114) {}+\delta_0^pN_{DK}\left(2jK_j(b_i)-(j+1)K_{j+1}(b_i)-jK_{j-1}(b_i)+\delta_1^j\right)A_E. \tag*{(114)} + δ 0 p N D K ( 2 j K j ( b i ) − ( j + 1 ) K j + 1 ( b i ) − j K j − 1 ( b i ) + δ 1 j ) A E . (114)
For the electrons in Eqs. (3.5a) and (3.5b), the sources are
S N = A N ( R ) , S T e = A T e ( R ) , (115) S_N=A_N(\boldsymbol{R}),\qquad S_{T_e}=A_{T_e}(\boldsymbol{R}), \tag*{(115)} S N = A N ( R ) , S T e = A T e ( R ) , (115)
with
K n ( x ) = 1 n ! ( x 2 ) 2 n e − ( x / 2 ) 2 , b i = v T h i k ⊥ Ω i . (116) K_n(x)=\frac{1}{n!}\left(\frac{x}{2}\right)^{2n}e^{-(x/2)^2},\qquad b_i=\frac{v_{T h i}k_\perp}{\Omega_i}. \tag*{(116)} K n ( x ) = n ! 1 ( 2 x ) 2 n e − ( x /2 ) 2 , b i = Ω i v T hi k ⊥ . (116)
Finally, we do not include a momentum source such that S U = 0 S_U=0 S U = 0 in Eq. (5.8b).
Simulations that solve the full system, including DK and GK fluctuations, are referred to as full simulations. For comparison, simulations are run where N i G K p j = 0 N_{iGK}^{pj}=0 N i G K p j = 0 and ϕ G K = 0 \phi_{GK}=0 ϕ G K = 0 . These are referred to as DK simulations.
We start the simulations from smooth analytical profiles satisfying the boundary conditions in Eq. (6.3) and evolve the DK simulations at first. Quasi-steady state is achieved after approximately 165 c s 0 / R 165\,c_{s0}/R 165 c s 0 / R time units using a small number of gyromoments (P D K = 2 P_{DK}=2 P D K = 2 and J D K = 1 J_{DK}=1 J D K = 1 ). Afterwards, the number of moments, P D K P_{DK} P D K and J D K J_{DK} J D K , is increased and the evolution of ϕ G K \phi_{GK} ϕ G K and N i G K p j N_{iGK}^{pj} N i G K p j is included. We choose P D K = 2 J D K P_{DK}=2J_{DK} P D K = 2 J D K for the DK moments and P G K ≲ J G K P_{GK}\lesssim J_{GK} P G K ≲ J G K for the GK moments to study the influence of FLR effects on the simulations. When a quasi-steady state is reached, the particles and energy injected by the sources are transported slowly across the field lines by turbulent transport and rapidly along the field lines to the sheaths where the plasma outflow at the sheath entrances balances the sources. All analyses are carried out during this quasi-steady state.
For our simulations, we consider a helium LAPD plasma Gekelman et al. [13 ] with parameters as in [37 , 10 , 25 ] . The simulation parameters are: n e 0 = 2 × 10 12 c m − 3 n_{e0}=2\times10^{12}\,\mathrm{cm}^{-3} n e 0 = 2 × 1 0 12 cm − 3 , T e 0 = 6 e V T_{e0}=6\,\mathrm{eV} T e 0 = 6 eV , T i 0 = 3 e V T_{i0}=3\,\mathrm{eV} T i 0 = 3 eV , Ω i ∼ 960 k H z \Omega_i\sim960\,\mathrm{kHz} Ω i ∼ 960 kHz , ρ s 0 = 1.4 c m \rho_{s0}=1.4\,\mathrm{cm} ρ s 0 = 1.4 cm , c s 0 = 1.3 × 10 6 c m / s c_{s0}=1.3\times10^6\,\mathrm{cm/s} c s 0 = 1.3 × 1 0 6 cm/s , m i / m e = 400 m_i/m_e=400 m i / m e = 400 , ν 0 = 0.03 \nu_0=0.03 ν 0 = 0.03 , and τ i = 0.5 \tau_i=0.5 τ i = 0.5 . The LAPD vacuum chamber has a radius R ≃ 0.56 m R \simeq0.56\,\mathrm{m} R ≃ 0.56 m , therefore we assume R = 40 ρ s 0 R = 40\rho_{s0} R = 40 ρ s 0 and, with a parallel extension L z ≃ 18 m L_z \simeq18\,\mathrm{m} L z ≃ 18 m , we impose L z = 36 R L_z = 36R L z = 36 R . With these parameters, the reference time is R / c s 0 ∼ 43 μ s R/c_{s0} \sim43\,\mu\mathrm{s} R / c s 0 ∼ 43 μ s . We set the perpendicular extension of the domain L x = L y = 100 ρ s 0 L_x = L_y = 100\rho_{s0} L x = L y = 100 ρ s 0 , sufficiently large for the turbulent structures not to be affected by the perpendicular boundary conditions. We choose our sources in Eq. (6.4) to have radial extension from the axis r s r_s r s and radial gradient length L s L_s L s , that is
A N ( x , y ) = A N 0 2 [ 1 − tanh ( r − r s L s ) ] + A N ∞ , (117) A_N(x,y) = \frac{A_{N0}}{2}\left[1-\tanh\left(\frac{r-r_s}{L_s}\right)\right]+A_{N\infty}, \tag*{(117)} A N ( x , y ) = 2 A N 0 [ 1 − tanh ( L s r − r s ) ] + A N ∞ , (117)
where r = x 2 + y 2 r = \sqrt{x^2+y^2} r = x 2 + y 2 and equivalent expressions are used for A E A_E A E , A T e A_{T_e} A T e , and A N A_N A N with A N 0 A_{N0} A N 0 , A E 0 A_{E0} A E 0 , and A T e 0 A_{T_e0} A T e 0 representing the bulk plasma sources. The constant terms, A N ∞ , A E ∞ , A T e ∞ ≪ A N 0 , A T i 0 , A T e 0 A_{N\infty}, A_{E\infty}, A_{T_e\infty} \ll A_{N0}, A_{Ti0}, A_{T_e0} A N ∞ , A E ∞ , A T e ∞ ≪ A N 0 , A T i 0 , A T e 0 , avoid negative values of temperature and density and are chosen to have small amplitude. The source parameters are chosen as A N ∞ = A T e ∞ = 0.004 A_{N\infty} = A_{T_e\infty} = 0.004 A N ∞ = A T e ∞ = 0.004 , A E ∞ = 0.001 A_{E\infty} = 0.001 A E ∞ = 0.001 , A N 0 = A T e 0 = A E 0 = 0.04 A_{N0} = A_{T_e0} = A_{E0} = 0.04 A N 0 = A T e 0 = A E 0 = 0.04 , L s = 1 ρ s 0 L_s = 1\rho_{s0} L s = 1 ρ s 0 , r s = 20 ρ s 0 r_s = 20\rho_{s0} r s = 20 ρ s 0 .
The perpendicular grid resolution is N x = N y = 192 N_x = N_y = 192 N x = N y = 192 , which corresponds to a grid spacing of ∼ 0.52 ρ s 0 \sim0.52\rho_{s0} ∼ 0.52 ρ s 0 . The number of parallel grid points is N z = 64 N_z = 64 N z = 64 , which corresponds to a grid spacing of ∼ 0.56 R \sim0.56R ∼ 0.56 R . For the DK fields (n e n_e n e , T e T_e T e , U ∥ U_{\parallel} U ∥ , Ω \Omega Ω , J ∥ J_{\parallel} J ∥ , and N i D K p j \mathcal{N}^{pj}_{iDK} N i D K p j ) diffusion parameters are chosen as η z D K ≃ 4 \eta_{zDK} \simeq4 η z D K ≃ 4 and η ⊥ ≃ 0.5 \eta_{\perp} \simeq0.5 η ⊥ ≃ 0.5 . For the Fourier grid, we impose N k x = N k y = 96 N_{kx} = N_{ky} = 96 N k x = N k y = 96 such that max ( k x ) = max ( k y ) = 3.08 / ρ s 0 \max(k_x) = \max(k_y) = 3.08/\rho_{s0} max ( k x ) = max ( k y ) = 3.08/ ρ s 0 . For N i G K p j \mathcal{N}^{pj}_{iGK} N i G K p j , η z G K ≃ 4 \eta_{zGK} \simeq4 η z G K ≃ 4 is the same value as for the DK fields. We note that no numerical diffusion for the GK fields is required since the ∇ ⊥ 2 N i G K p j \nabla_{\perp}^2\mathcal{N}^{pj}_{iGK} ∇ ⊥ 2 N i G K p j term in Eq. (79) provides sufficiently large damping for high k ⊥ k_{\perp} k ⊥ modes and also ensures strong damping for modes with k ⊥ ≳ 1 / ρ s 0 k_{\perp} \gtrsim1/\rho_{s0} k ⊥ ≳ 1/ ρ s 0 . Finally, we use η G K = 10 − 7 \eta_{GK} = 10^{-7} η G K = 1 0 − 7 .
Simulation results
In this section, we present simulations based on the model described in § 3 , § 4 , and § 5 , using the parameters of the LAPD experiment described in § 6 . We show the equilibrium DK and GK gyromoments in § 7.1 . In Sec. 7.2 we demonstrate spectral convergence of our simulations by comparing the ion distribution function with a local bi-Maxwellian and by investigating the effect of the number of gyromoments on the statistical properties of ϕ D K \phi_{DK} ϕ D K . We investigate the turbulence properties of the DK fields in Sec. 7.3. Finally, in Sec. 7.4 we analyze the properties of the GK fields.
Equilibrium gyromoments
The spatial and azimuthal average, defined as
⟨ ⋅ ⟩ t , ϑ = 1 2 π ∫ 0 2 π d ϑ 1 Δ t ∫ t 0 t 0 + Δ t d t ( ⋅ ) , (118) \langle\cdot\rangle_{t,\vartheta}=\frac{1}{2\pi}\int_{0}^{2\pi}d\vartheta\frac{1}{\Delta t}\int_{t_0}^{t_0+\Delta t}dt\,(\cdot), \tag*{(118)} ⟨ ⋅ ⟩ t , ϑ = 2 π 1 ∫ 0 2 π d ϑ Δ t 1 ∫ t 0 t 0 + Δ t d t ( ⋅ ) , (118)
of the DK gyromoments, ⟨ N i D K p j ⟩ t , ϑ \langle N^{pj}_{iDK}\rangle_{t,\vartheta} ⟨ N i D K p j ⟩ t , ϑ , with ϑ = arctan y / x \vartheta=\arctan y/x ϑ = arctan y / x are shown in Figure 1 . The gyromoments with odd (even) p p p are antisymmetric (symmetric) around z = L z / 2 z=L_z/2 z = L z /2 . Furthermore, the amplitude of the moments decreases with p p p and j j j . We note that the plasma equilibrium is very similar to previous gyromoment simulations of an LAPD plasma Mencke & Ricci [25 ] .
Figure 1. The spatial and azimuthal average of the DK moments, ⟨ N i D K p j ⟩ t , ϑ \langle N^{pj}_{iDK}\rangle_{t,\vartheta} ⟨ N i D K p j ⟩ t , ϑ , for a full simulation with ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{DK},J_{DK},P_{GK},J_{GK})=(6,3,6,5) ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) for p = 1 p=1 p = 1 –4 4 4 (rows) and j = 1 , 2 j=1,2 j = 1 , 2 (columns). Moments with odd p p p are antisymmetric around z = L z / 2 z=L_z/2 z = L z /2 and moments with even p p p are symmetric around z = L z / 2 z=L_z/2 z = L z /2 .
Figure 2 shows the spatial and azimuthal average of the GK moments, ⟨ N i G K p j ⟩ t , ϑ \langle N^{pj}_{iGK}\rangle_{t,\vartheta} ⟨ N i G K p j ⟩ t , ϑ . Similar symmetry properties as for the DK moments are observed, although the GK moments have significantly smaller amplitude. Furthermore, the amplitude of ⟨ N i G K p j ⟩ t , ϑ \langle N^{pj}_{iGK}\rangle_{t,\vartheta} ⟨ N i G K p j ⟩ t , ϑ is localized around the source gradient r ∼ 20 ρ s 0 r\sim20\rho_{s0} r ∼ 20 ρ s 0 since δ S i 1 p j \delta S^{pj}_{i1} δ S i 1 p j is the largest in that region.
Figure 2. The spatial and azimuthal average of N i G K p j N^{pj}_{iGK} N i G K p j , ⟨ N i G K / t , ϑ p j ⟩ \langle N^{pj}_{iGK/t,\vartheta}\rangle ⟨ N i G K / t , ϑ p j ⟩ , for a full simulation with ( P D K , ϑ D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{DK}, \vartheta_{DK}, P_{GK}, J_{GK}) = (6,3,6,5) ( P D K , ϑ D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) for p = 1 − 4 p = 1 - 4 p = 1 − 4 (rows) and j = 1 , 2 j = 1,2 j = 1 , 2 (columns). Moments with odd p p p are antisymmetric around z = L z / 2 z = L_z/2 z = L z /2 and moments with even p p p are symmetric around z = L z / 2 z = L_z/2 z = L z /2 .
In order to quantify the kinetic part of the distribution function, we compare the moments from the full simulations with those of a local bi-Maxwellian Mencke & Ricci
(2025)
F b M = N i m i 3 / 2 ( 2 π ) 3 / 2 T ∥ i T ⊥ i exp ( − m i ( v ∥ − U ∥ i ) 2 T ∥ i − μ B T ⊥ i ) , (119) F_{bM} = \frac{N_i m_i^{3/2}}{(2\pi)^{3/2}\sqrt{T_{\parallel i}T_{\perp i}}}\exp\left(-\frac{m_i\left(v_{\parallel}-U_{\parallel i}\right)^2}{T_{\parallel i}}-\frac{\mu B}{T_{\perp i}}\right), \tag*{(119)} F b M = ( 2 π ) 3/2 T ∥ i T ⊥ i N i m i 3/2 exp ( − T ∥ i m i ( v ∥ − U ∥ i ) 2 − T ⊥ i μ B ) , (119)
with the projection on the Hermite-Laguerre basis in Eq. (3.6a) given by
N b M p j = { N i ( − 1 ) j p ! 2 p ( p / 2 ) ! ( T ∥ i T ∥ 0 − 1 ) p / 2 ( T ⊥ i T ⊥ 0 − 1 ) j for even p , 0 for odd p . (120) \mathcal{N}^{pj}_{bM} =
\begin{cases}
N_i\frac{(-1)^j\sqrt{p!}}{\sqrt{2^p(p/2)!}}\left(\frac{T_{\parallel i}}{T_{\parallel0}}-1\right)^{p/2}\left(\frac{T_{\perp i}}{T_{\perp0}}-1\right)^j & \text{for even }p,\\
0 & \text{for odd }p.
\end{cases} \tag*{(120)} N b M p j = ⎩ ⎨ ⎧ N i 2 p ( p /2 )! ( − 1 ) j p ! ( T ∥ 0 T ∥ i − 1 ) p /2 ( T ⊥ 0 T ⊥ i − 1 ) j 0 for even p , for odd p . (120)
We use the local T ⊥ i T_{\perp i} T ⊥ i , T ∥ i T_{\parallel i} T ∥ i , and N i = N D K N_i = N_{DK} N i = N D K in Eq. (7.2b) and show the spatially maximal deviation for the gyromoments in the full-f f f simulation for each p p p and j j j in Figure 3 . Small deviations are observed due to the high collisionality in the simulations, similarly to DK-ordered gyromoment simulations of the same system Mencke & Ricci [25 ] .
Figure 3. The spatial maximum of the difference between the DK moments and the moments of a bi-Maxwellian calculated using Eq. (7.2b) with the local parallel and perpendicular temperatures normalized to the local density for ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{DK},J_{DK},P_{GK},J_{GK}) = (6,3,6,5) ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) . Small deviations from a bi-Maxwellian are observed.
Convergence study
Since deviations from a bi-Maxwellian are small, a small number of moments is needed to reach convergence. We investigate the statistical properties of the DK electrostatic potential, ϕ D K \phi_{DK} ϕ D K , and how these are affected by the number of evolved moments in Figure 4 . We compare the temporal and azimuthal average of ϕ D K \phi_{DK} ϕ D K , defined in Eq. (118) , the root mean square,
Figure 4. The azimuthal and temporal average, root mean square, and skewness of ϕ D K \phi_{DK} ϕ D K at z = L z / 2 z = L_z/2 z = L z /2 for different choices of ( P D K , J D K , P G K , J G K ) (P_{DK}, J_{DK}, P_{GK}, J_{GK}) ( P D K , J D K , P G K , J G K ) . For the GK simulations, P D K = P G K P_{DK} = P_{GK} P D K = P G K and J G K = 5 J_{GK} = 5 J G K = 5 . The full (solid) and DK (dashed) models are compared. The simulations are considered converged at ( P D K , J D K ) = ( 2 , 1 ) (P_{DK}, J_{DK}) = (2, 1) ( P D K , J D K ) = ( 2 , 1 ) . The presence of GK fields does not impact ϕ D K \phi_{DK} ϕ D K noticeably.
R M S ( ϕ D K ) = ⟨ ( ( ϕ D K ) − ⟨ ϕ D K ⟩ t , ϑ ) 2 ⟩ t , ϑ , (121) RMS(\phi_{DK}) = \sqrt{\left\langle\left((\phi_{DK})-\langle\phi_{DK}\rangle_{t,\vartheta}\right)^2\right\rangle_{t,\vartheta}}, \tag*{(121)} R M S ( ϕ D K ) = ⟨ ( ( ϕ D K ) − ⟨ ϕ D K ⟩ t , ϑ ) 2 ⟩ t , ϑ , (121)
and the skewness
S K W ( ϕ D K ) = ⟨ ( ( ϕ D K ) − ⟨ ϕ D K ⟩ t , ϑ ) 3 ⟩ t , ϑ / R M S 3 ( ϕ D K ) , (122) SKW(\phi_{DK}) = \left\langle\left((\phi_{DK})-\langle\phi_{DK}\rangle_{t,\vartheta}\right)^3\right\rangle_{t,\vartheta}/RMS^3(\phi_{DK}), \tag*{(122)} S K W ( ϕ D K ) = ⟨ ( ( ϕ D K ) − ⟨ ϕ D K ⟩ t , ϑ ) 3 ⟩ t , ϑ / R M S 3 ( ϕ D K ) , (122)
for different choices of P D K , J D K , P G K , J G K P_{DK},J_{DK},P_{GK},J_{GK} P D K , J D K , P G K , J G K and for the full and DK simulations. It is observed that spectral convergence is obtained at ( P D K , J D K ) = ( 2 , 1 ) (P_{DK},J_{DK}) = (2,1) ( P D K , J D K ) = ( 2 , 1 ) for both the full and DK simulations and the average, RMS, and skewness of ϕ D K \phi_{DK} ϕ D K are not significantly affected by the presence of GK fields. A similar trend is observed for all other DK fields. Both models and all choices of ( P D K , J D K ) (P_{DK},J_{DK}) ( P D K , J D K ) display a negative skewness at r ≲ 20 ρ s 0 r \lesssim20\rho_{s0} r ≲ 20 ρ s 0 , associated with the propagation of holes, and a positive skewness for r ≳ 20 ρ s 0 r \gtrsim20\rho_{s0} r ≳ 20 ρ s 0 , associated with blobs, similar to simulations of the tokamak boundary. Similar statistical behavior of the fields was reported in previous simulations of LAPD plasmas Frei et al. [10 ] ; Mencke & Ricci [25 ] . In the following, we analyse the results of the simulations with ( P D K , J D K ) = ( 6 , 3 ) (P_{DK},J_{DK}) = (6,3) ( P D K , J D K ) = ( 6 , 3 ) .
Properties of the DK fields
We consider the turbulent behavior of the DK fields. Figure 5 shows a perpendicular snapshot of the DK fields for the full simulation, while Figure 6 shows the same snapshot for the DK simulation. We note that for both simulations, all fields display large turbulent structures in the perpendicular plane. The DK and full simulations are very similar, indicating that the inclusion of the GK fields do not significantly affect the system. A snapshot of the DK fields in the parallel plane is shown in Figure 7 for the full simulations. We observe elongated structures in the z z z -direction. Similar properties are observed for the DK simulations (not shown here).
Figure 5. Perpendicular, ( x , y ) (x,y) ( x , y ) , snapshot of N D K N_{\mathrm{DK}} N DK (top left), T e D K T_{\mathrm{eDK}} T eDK (top center), J ∥ J_{\parallel} J ∥ (top right), ϕ D K \phi_{\mathrm{DK}} ϕ DK (bottom left), Ω \Omega Ω (bottom center), and U ∥ U_{\parallel} U ∥ (bottom right) for a full simulation with ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{\mathrm{DK}},J_{\mathrm{DK}},P_{\mathrm{GK}},J_{\mathrm{GK}})=(6,3,6,5) ( P DK , J DK , P GK , J GK ) = ( 6 , 3 , 6 , 5 ) at z = L z / 2 z=L_z/2 z = L z /2 . Turbulent structures are observed at the source-gradient region ( r ∼ r s ∼ 20 ρ s 0 ) (r\sim r_s\sim20\rho_{s0}) ( r ∼ r s ∼ 20 ρ s 0 ) .
Figure 6. Same as Figure 5 for a DK simulation with ( P D K , J D K ) = ( 6 , 3 ) (P_{\mathrm{DK}},J_{\mathrm{DK}})=(6,3) ( P DK , J DK ) = ( 6 , 3 ) . Similar structures as for the full model in Figure 5 are observed.
Figure 7. Parallel, ( z , y ) (z,y) ( z , y ) , snapshot of N D K N_{\mathrm{DK}} N DK (top left), T e D K T_{e\mathrm{DK}} T e DK (top center), J ∥ J_{\parallel} J ∥ (top right), ϕ D K \phi_{\mathrm{DK}} ϕ DK (bottom left), Ω \Omega Ω (bottom center), and U ∥ U_{\parallel} U ∥ (bottom right) for a full simulation with ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{\mathrm{DK}},J_{\mathrm{DK}},P_{\mathrm{GK}},J_{\mathrm{GK}})=(6,3,6,5) ( P DK , J DK , P GK , J GK ) = ( 6 , 3 , 6 , 5 ) at x = 0 x=0 x = 0 . The turbulent structures are very elongated along z z z .
Figure 8 compares the power spectrum of Ω \Omega Ω as a function of the parallel wave number, k z k_z k z , and azimuthal mode number, m m m . We observe low-m m m modes elongated in z z z . The full and DK simulations have similar power spectrum. To test the interaction between DK and GK fields, we artificially lower the collisionality in Eq. (79) by a factor of 1000, setting ν i 0 ∼ 0.00234 c s 0 / R \nu_{i0} \sim0.00234c_{s0}/R ν i 0 ∼ 0.00234 c s 0 / R , and we increase A E A_E A E and A N A_N A N by a factor 10 in Eq. (6.4c). By reducing the GK collisional term and increasing the GK source term, we observe an increase in high m m m modes indicating that in a system with lower collisionality and larger sources, the GK fields can introduce high k ⊥ k_\perp k ⊥ -modes in the DK fields.
Figure 8. Power spectrum of Ω = ∑ m , ω Ω ^ e i ( m ϑ + k z z ) \Omega=\sum_{m,\omega}\widehat{\Omega}e^{i(m\vartheta+k_z z)} Ω = ∑ m , ω Ω e i ( m ϑ + k z z ) at the radial distance, r = x 2 + y 2 = 20 ρ s 0 r=\sqrt{x^2+y^2}=20\rho_{s0} r = x 2 + y 2 = 20 ρ s 0 for a full simulation with ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{\mathrm{DK}},J_{\mathrm{DK}},P_{\mathrm{GK}},J_{\mathrm{GK}})=(6,3,6,5) ( P DK , J DK , P GK , J GK ) = ( 6 , 3 , 6 , 5 ) (left), the DK simulation with ( P D K , J D K ) = ( 6 , 3 ) (P_{\mathrm{DK}},J_{\mathrm{DK}})=(6,3) ( P DK , J DK ) = ( 6 , 3 ) (center), and a full simulation with ν i 0 \nu_{i0} ν i 0 in Eq. (79) reduced by a factor of 1000 and with A E A_E A E and A N A_N A N increased by a factor 10 in Eq. (6.4c) (right).
Properties of gyrokinetic fields
In Figure 9 , ϕ G K \phi_{GK} ϕ G K is shown in real space ( x , y , z ) (x,y,z) ( x , y , z ) while Figure 10 shows ϕ G K \phi_{GK} ϕ G K in ( k x , k y , z ) (k_x,k_y,z) ( k x , k y , z ) space. The GK potential, ϕ G K \phi_{GK} ϕ G K , depends weakly on the z z z -direction and is dominated by
Figure 9. Snapshots of ϕ G K \phi_{\mathrm{GK}} ϕ GK in real space at z = L z / 2 z=L_z/2 z = L z /2 (left), x = 0 x=0 x = 0 (middle), and y = 0 y=0 y = 0 (right) for ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{\mathrm{DK}},J_{\mathrm{DK}},P_{\mathrm{GK}},J_{\mathrm{GK}})=(6,3,6,5) ( P DK , J DK , P GK , J GK ) = ( 6 , 3 , 6 , 5 ) . We note that ϕ G K \phi_{\mathrm{GK}} ϕ GK has small amplitude with small scale structures in the perpendicular plane and is almost constant along z z z .
Figure 10. Snapshots of ϕ G K \phi_{\mathrm{GK}} ϕ GK in Fourier ( k x , k y , z ) (k_x,k_y,z) ( k x , k y , z ) space at z = 0 z=0 z = 0 (left), k x = 0 k_x=0 k x = 0 (middle), and k y = 0 k_y=0 k y = 0 (right) real part (top) and imaginary part (bottom) for ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{\mathrm{DK}},J_{\mathrm{DK}},P_{\mathrm{GK}},J_{\mathrm{GK}})=(6,3,6,5) ( P DK , J DK , P GK , J GK ) = ( 6 , 3 , 6 , 5 ) .
k ⊥ ρ s ≲ 1 k_{\perp}\rho_s\lesssim1 k ⊥ ρ s ≲ 1 modes. Comparing with Figure 5 , it is seen that ϕ G K \phi_{\mathrm{GK}} ϕ GK is one order of magnitude smaller than ϕ D K \phi_{\mathrm{DK}} ϕ DK and from Figure 10 , k ⊥ ρ s 0 ∼ 1 k_{\perp}\rho_{s0}\sim1 k ⊥ ρ s 0 ∼ 1 modes are observed to exist even though they are not dominant. This is consistent with the ordering imposed in § 2 . Due to the high collisionality, the spatial diffusion induced by the ∇ ⊥ 2 N G K p j \nabla_{\perp}^{2}N_{\mathrm{GK}}^{pj} ∇ ⊥ 2 N GK p j term in Eq. (79) suppresses high k ⊥ k_\perp k ⊥ modes for N G K p j \mathcal{N}_{\mathrm{GK}}^{pj} N GK p j . Since in Eq. (106) , ρ i ∗ ∝ exp ( − 2 T i 0 k ⊥ 2 / m i Ω i 2 ) \rho_i^* \propto\exp(-2T_{i0}k_\perp^2/m_i\Omega_i^2) ρ i ∗ ∝ exp ( − 2 T i 0 k ⊥ 2 / m i Ω i 2 ) [see Eq. (B1b)], we further dampen high k ⊥ k_\perp k ⊥ modes in ϕ G K \phi_{\mathrm{GK}} ϕ GK .
Finally, the amplitude of the GK moments are shown in Figure 11 . The amplitude of the moments, N G K p j \mathcal{N}_{\mathrm{GK}}^{pj} N GK p j , rapidly decreases with p p p and j j j . This is due to the large collisionality which suppresses high p p p and j j j moments, as seen from Eq. (79) .
Figure 11. Spatial maximum of the amplitude of the GK moments, max k x , k y , z ( ∣ N i G K p j ∣ ) \max_{k_x,k_y,z}(\lvert N^{pj}_{iGK}\rvert) max k x , k y , z (∣ N i G K p j ∣) , for ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) (P_{DK},J_{DK},P_{GK},J_{GK})=(6,3,6,5) ( P D K , J D K , P G K , J G K ) = ( 6 , 3 , 6 , 5 ) . The magnitude quickly decreases with p p p and j j j .
Linear analysis of the turbulent modes
For studying the turbulent behavior of the simulations observed in Figure 8 , the system of Eqs. (3.5a), (3.5b), (3.10a), (4.4a), (5.8b), (5.14e), (5.14g), (5.15b), and (106) is linearized. Specifically, each DK field, f D K f_{DK} f D K , is split as
f D K = f ˉ D K ( r ) + f ~ D K ( r ) e i ( m ϑ + k ∥ z − ω t ) (123) f_{DK}=\bar{f}_{DK}(r)+\tilde{f}_{DK}(r)e^{i(m\vartheta+k_{\parallel}z-\omega t)} \tag*{(123)} f D K = f ˉ D K ( r ) + f ~ D K ( r ) e i ( m ϑ + k ∥ z − ω t ) (123)
where the equilibrium, f ˉ D K \bar{f}_{DK} f ˉ D K , is calculated from the full 3D turbulent simulations by performing a temporal and azimuthal averaging at z = L z / 2 z=L_z/2 z = L z /2 . For the GK field, f G K f_{GK} f G K , we apply a local approximation,
f G K = f ˉ G K ( r ) + f ~ G K δ ( r − r g ) e i ( m ϑ + k ∥ z − ω t ) (124) f_{GK}=\bar{f}_{GK}(r)+\tilde{f}_{GK}\delta(r-r_g)e^{i(m\vartheta+k_{\parallel}z-\omega t)} \tag*{(124)} f G K = f ˉ G K ( r ) + f ~ G K δ ( r − r g ) e i ( m ϑ + k ∥ z − ω t ) (124)
where r g r_g r g is the radial position at which the GK fluctuations are localized. To calculate f ˉ G K \bar{f}_{GK} f ˉ G K we first perform an inverse Fourier transform in the perpendicular plane of the fields in the global full 3D simulations. Then, the temporal and azimuthal averaging at z = L z / 2 z=L_z/2 z = L z /2 are evaluated.
In Figure 12 we show the dominant linear growth-rate, γ ≡ Im ω \gamma\equiv\operatorname{Im}\omega γ ≡ Im ω , as function of k ∥ k_{\parallel} k ∥ and m m m for the full model, the DK model (where we impose N i G K p j = 0 \mathcal{N}^{pj}_{iGK}=0 N i G K p j = 0 and ϕ G K = 0 \phi_{GK}=0 ϕ G K = 0 ), and the GK model (where we impose f ~ D K = 0 \tilde{f}_{DK}=0 f ~ D K = 0 ). We only consider m ≤ 40 m \le40 m ≤ 40 since the numerical diffusion for the DK field significantly dampens modes with higher m m m . The equilibrium fields, f ˉ \bar{f} f ˉ , for the DK model are calculated using the simulations with physical collisionality and sources. With the physical parameters, the full model yields linear growth rates identical to that of the DK model while the GK model is stable for all choices of m m m and k ∥ k_{\parallel} k ∥ . For this reason, for the full and GK model, we calculate the equilibrium fields using the simulations where ν i 0 \nu_{i0} ν i 0 in Eq. (79) is reduced by a factor of 1000 and where A E A_E A E and A N A_N A N are increased by a factor 10 in Eq. (6.4c). Furthermore, we neglect C i i p j C_{ii}^{pj} C ii p j in Eq. (4.4a).
Figure 12. The linear growth rates of the system of Eqs. (3.5a), (3.5b), (5.14e), (5.14g), (5.15b), (5.8b), and (106) with ( P D K , J D K , P G K , J G K ) = ( 4 , 2 , 4 , 3 ) (P_{DK}, J_{DK}, P_{GK}, J_{GK}) = (4, 2, 4, 3) ( P D K , J D K , P G K , J G K ) = ( 4 , 2 , 4 , 3 ) , for the full system (left), for the DK system with f G K = 0 f_{GK} = 0 f G K = 0 (center), and for the GK system with f ~ D K = 0 \tilde{f}_{DK} = 0 f ~ D K = 0 (right), r g = 22 ρ s 0 r_g = 22\rho_{s0} r g = 22 ρ s 0 . The simulations with physical GK collisionality and sources are used to calculate the equilibrium fields for the DK system while the simulations with ν i 0 \nu_{i0} ν i 0 in Eq. (79) reduced by a factor of 1000 and with A E A_E A E and A N A_N A N increased by a factor 10 in Eq. (6.4c) are used for the two other systems. The maximum growth rates for the full model is max m , k ∥ ( γ max ) = 1.62 c s 0 / R \max_{m,k_\parallel}(\gamma_{\max}) = 1.62c_{s0}/R max m , k ∥ ( γ m a x ) = 1.62 c s 0 / R , for the DK model it is max m , k ∥ ( γ max ) = 0.91 c s 0 / R \max_{m,k_\parallel}(\gamma_{\max}) = 0.91c_{s0}/R max m , k ∥ ( γ m a x ) = 0.91 c s 0 / R and for the GK model it is max m , k ∥ ( γ max ) = 0.014 c s 0 / R \max_{m,k_\parallel}(\gamma_{\max}) = 0.014c_{s0}/R max m , k ∥ ( γ m a x ) = 0.014 c s 0 / R . The highest growth rate for the full system is observed at m = 33 m = 33 m = 33 and k ∥ = 0 k_\parallel= 0 k ∥ = 0 .
By artificially suppressing different terms, we study the linear drive of the dominant instabilities. For the DK model, the peak growth rate is a Kelvin-Helmholtz instability D’Angelo [6 ] . More specifically, the modes are driven by the [ ϕ D K , Ω ] [\phi_{DK},\Omega] [ ϕ D K , Ω ] and the [ ∇ ⊥ ϕ D K ⋅ , ω ] [\nabla_\perp\phi_{DK}\cdot,\omega] [ ∇ ⊥ ϕ D K ⋅ , ω ] terms in Eq. (5.14e). This finding is consistent with previous long-wavelength simulations of linear plasma devices Frei et al. [10 ] ; Mencke & Ricci [25 ] ; Rogers & Ricci [37 ] .
For the GK model, the dominant instability is at k ∥ = 0.087 / R k_\parallel=0.087/R k ∥ = 0.087/ R and m = 5 m=5 m = 5 and it is driven by the [ ϕ D K , N i G K p j ] [\phi_{DK},\mathcal{N}_{iGK}^{pj}] [ ϕ D K , N i G K p j ] term in Eq. (4.4a) for p = j = 0 p=j=0 p = j = 0 . It is driven by shear flow and a density gradient since both the GK density gradient and the gradient of the DK electrostatic potential are necessary for the instability to appear. We note that for m = 5 m=5 m = 5 , using Eq. (79) , C i i 00 ∼ − 0.051 ν i 0 τ i N i G K 00 C_{ii}^{00}\sim-0.051\nu_{i0}\tau_i\mathcal{N}_{iGK}^{00} C ii 00 ∼ − 0.051 ν i 0 τ i N i G K 00 . Therefore, collisions damp this instability with γ ∼ 0.014 c s 0 / R \gamma\sim0.014c_{s0}/R γ ∼ 0.014 c s 0 / R when using the physical collisionality.
The dominant instability for the full model is a Kelvin-Helmholtz mode with the growth rate enhanced by the [ ϕ D K , N i G K p j ] [\phi_{DK},\mathcal{N}_{iGK}^{pj}] [ ϕ D K , N i G K p j ] term in Eq. (4.4a) for p = j = 0 p=j=0 p = j = 0 and the [ ϕ G K , N i D K p j ] [\phi_{GK},\mathcal{N}_{iDK}^{pj}] [ ϕ G K , N i D K p j ] term in Eq. (3.12a) for p = j = 0 p=j=0 p = j = 0 . Through a DK density gradient, the Kelvin-Helmholtz mode is coupled to the GK fields. We note that ϕ ˉ G K \bar{\phi}_{GK} ϕ ˉ G K is an order of magnitude smaller than the amplitude of ϕ ˉ G K \bar{\phi}_{GK} ϕ ˉ G K . Therefore, we do not observe a significant influence of ϕ ˉ G K \bar{\phi}_{GK} ϕ ˉ G K in the 1D linear simulations.
In order to study the development of small-scale structures observed for the low-collisionality and larger-GK-sources simulations observed in Figure 8 , we assume that k ∥ = 0 k_\parallel=0 k ∥ = 0 , F i D K F_{iDK} F i D K is a Maxwellian constant in time and space, and F i G K F_{iGK} F i G K is an isotropic Maxwellian. Focusing on a small gradient region with a gradient length of x 0 < ρ s 0 x_0<\rho_{s0} x 0 < ρ s 0 , and assuming that ∣ ∇ ⊥ ϕ G K ∣ ∼ T e e N e ∣ ∇ ⊥ ϱ i ∗ ∣ ≫ ∣ ∇ ⊥ ϕ D K ∣ |\nabla_\perp\phi_{GK}|\sim\frac{T_e}{eN_e}|\nabla_\perp\varrho_i^*|\gg|\nabla_\perp\phi_{DK}| ∣ ∇ ⊥ ϕ G K ∣ ∼ e N e T e ∣ ∇ ⊥ ϱ i ∗ ∣ ≫ ∣ ∇ ⊥ ϕ D K ∣ , we obtain the following condition for an instability,
1 T e D K + 1 T i [ 1 − exp ( − T i D K k ⊥ 2 Ω i 2 ) I 0 ( T i D K k ⊥ 2 Ω i 2 ) ] > ∣ ∇ ⊥ ϱ i ∗ ∣ e N D K ∣ ∇ ⊥ ϕ G K ∣ . (125) \frac{1}{T_{eDK}}+\frac{1}{T_i}\left[1-\exp\left(-\frac{T_{iDK}k_\perp^2}{\Omega_i^2}\right)I_0\left(\frac{T_{iDK}k_\perp^2}{\Omega_i^2}\right)\right]>\frac{|\nabla_\perp\varrho_i^*|}{eN_{DK}|\nabla_\perp\phi_{GK}|}. \tag*{(125)} T eD K 1 + T i 1 [ 1 − exp ( − Ω i 2 T i D K k ⊥ 2 ) I 0 ( Ω i 2 T i D K k ⊥ 2 ) ] > e N D K ∣ ∇ ⊥ ϕ G K ∣ ∣ ∇ ⊥ ϱ i ∗ ∣ . (125)
The details of the calculation leading to Eq. (125) are specified in App. § 13 . For small structures where v T h i k ⊥ ≫ Ω i v_{Thi}k_\perp\gg\Omega_i v T hi k ⊥ ≫ Ω i , we expect ϕ G K ∼ T e D K ϱ i ∗ / N D K \phi_{GK}\sim T_{eDK}\varrho_i^*/N_{DK} ϕ G K ∼ T eD K ϱ i ∗ / N D K . Indeed, in Eq. (106) , using Eqs. (B 1a) and (B 1b), ∣ ∂ μ ⟨ ⟨ ϕ G K ⟩ R ⟩ x † ∣ ∝ exp ( − T i 0 k ⊥ 2 / m i Ω i 2 ) → 0 \left|\partial_\mu\langle\langle\phi_{GK}\rangle_{R}\rangle_{x}^{\dagger}\right|\propto\exp\left(-T_{i0}k_\perp^2/m_i\Omega_i^2\right)\to0 ∂ μ ⟨⟨ ϕ G K ⟩ R ⟩ x † ∝ exp ( − T i 0 k ⊥ 2 / m i Ω i 2 ) → 0 when v T h i k ⊥ ≫ Ω i v_{Thi}k_\perp\gg\Omega_i v T hi k ⊥ ≫ Ω i . Therefore, Eq. (125) is in general satisfied for large enough k ⊥ k_\perp k ⊥ . This instability would allow for a non-linear growth of small-scale structures (due to the gradient size, x 0 x_0 x 0 , of ϕ G K \phi_{GK} ϕ G K further decreasing). On the other hand, the ∼ ν i 0 ∇ ⊥ 2 \sim\nu_{i0}\nabla_\perp^2 ∼ ν i 0 ∇ ⊥ 2 term in Eq. (79) and the exp ( − T i 0 k 2 Ω i 2 ) \exp\left(-\frac{T_{i0}k^2}{\Omega_i^2}\right) exp ( − Ω i 2 T i 0 k 2 ) factor in ρ i ∗ \rho_i^* ρ i ∗ in Eq. (B1b) inhibits the buildup of large gradient regions for ϕ G K \phi_{GK} ϕ G K . As a direct consequence, in the global simulations, this instability is only observed in the simulation with low GK collisionality.
Conclusion
First of a kind simulations that self-consistently couple full-f f f drift-kinetic (DK) and δ \delta δ -f f f gyrokinetic (GK) fields are performed in a linear plasma device, relying on a gyromoment approach. We consistently derive the ion and electron system of equations starting from the gyro-averaged Boltzmann equations. Adopting a critical balance ordering, we find that the electrons are described by a drift-reduced Braginskii model. The DK and GK parts of the ion distribution function are projected on a Hermite-Laguerre basis yielding a hierarchy equation for the resulting gyromoments of the distribution function.
The physical model is numerically implemented and simulations are carried out using LAPD parameters. Spectral convergence is observed with a small amount of moments, due to the high ion-ion collisionality that leads to an ion distribution function close to a bi-Maxwellian. The inclusion of the small-scale GK part of the electrostatic potential and ion gyromoments does not affect the DK fields at the physical collisionality. When artificially reducing the GK collision term and increasing the GK sources, a small increase in small-scale turbulent structures in the plane perpendicular to the magnetic field is observed for the DK fields.
The GK electrostatic potential is dominated by long-wavelength modes that have significantly smaller perpendicular sizes than its DK counterpart as well as smaller amplitude satisfying the assumptions behind the physical model. The amplitude of the GK gyromoments quickly decreases with increasing order of the Hermite and Laguerre polynomials due to the high ion-ion collisionality.
Linearly, the growth-rates of the system of equations are driven by the DK fields supporting the findings from the global full-f simulations that the GK fields play a subdominant role in driving turbulence. The driving instability has a Kelvin-Helmholz nature. If the amplitude of the GK sources are amplified and collisions for the GK fields are suppressed, increased growth rates are observed caused by the GK density gradients. In this scenario, the growth rate of unstable modes at large perpendicular wavenumbers is amplified. We show analytically that small gradient length scales for the GK potential and density can drive a GK Kelvin-Helmholz-like instability. Additionally, this instability can non-linearly cause small-scale structures to develop in the full 3D simulations.
Acknowledgements
The authors thank, Louis N. Stenger, Brenno Jason Sanzio Peter De Lucca, Samuel Ernst, and Zeno Tecchiolli for fruitful discussions. Simulations were carried out due to the grant EHPC-REG-2023R03-144 on the Discoverer supercomputer. This work has been carried out within the framework of the EUROfusion Consortium, partially funded by the European Union via the Euratom Research and Training Programme (Grant Agreement No 101052200 — EUROfusion). The Swiss contribution to this work has been funded in part by the Swiss State Secretariat for Education, Research and Innovation (SERI). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union, the European Commission or SERI. Neither the European Union nor the European Commission nor SERI can be held responsible for them.
Calculation of higher order electron distribution function
In order to solve for F e D K 1 F_{eDK1} F eD K 1 , we use the ansatz Eq. (3.4a) with the additional ansatz for f s h f_{sh} f s h in Eq. (3.4b). We project Eq. (32) on the v ∥ L j ( 3 / 2 ) v_{\parallel}L_j^{(3/2)} v ∥ L j ( 3/2 ) basis, using parity in v ∥ v_{\parallel} v ∥ (notice that the collision operators conserve parity in v)
4 3 π ( 2 Γ ( 5 2 ) δ 0 j ( ∇ ∥ P e 1 + e N e 1 ∇ ∥ ϕ D K ) − 2 Γ ( 7 2 ) δ 1 j ( ∇ ∥ P e 1 + e P e 1 T e 0 ∇ ∥ ϕ D K ) ) = m e ( C e e 1 j + C e i 1 j ) (126) \frac{4}{3\sqrt{\pi}}\left(2\Gamma\left(\frac{5}{2}\right)\delta_{0}^{j}\left(\nabla_{\parallel}P_{e1}+eN_{e1}\nabla_{\parallel}\phi_{DK}\right)-2\Gamma\left(\frac{7}{2}\right)\delta_{1}^{j}\left(\nabla_{\parallel}P_{e1}+e\frac{P_{e1}}{T_{e0}}\nabla_{\parallel}\phi_{DK}\right)\right)=m_{e}\left(C_{ee1}^{j}+C_{ei1}^{j}\right) \tag*{(126)} 3 π 4 ( 2Γ ( 2 5 ) δ 0 j ( ∇ ∥ P e 1 + e N e 1 ∇ ∥ ϕ D K ) − 2Γ ( 2 7 ) δ 1 j ( ∇ ∥ P e 1 + e T e 0 P e 1 ∇ ∥ ϕ D K ) ) = m e ( C ee 1 j + C e i 1 j ) (126)
C e e 1 j = v T h e 2 v e e ∑ p K e e j p A p , (127) C_{ee1}^{j}=v_{The}^{2}v_{ee}\sum_{p}K_{ee}^{jp}A_{p}, \tag*{(127)} C ee 1 j = v T h e 2 v ee p ∑ K ee j p A p , (127)
C e i 1 j = v T h e 2 v e i ( ∑ p K e i j p A p − m e T e 0 K e i j 0 J ∥ ) , (128) C_{ei1}^{j}=v_{The}^{2}v_{ei}\left(\sum_{p}K_{ei}^{jp}A_{p}-\frac{m_{e}}{T_{e0}}K_{ei}^{j0}J_{\parallel}\right), \tag*{(128)} C e i 1 j = v T h e 2 v e i ( p ∑ K e i j p A p − T e 0 m e K e i j 0 J ∥ ) , (128)
and finally,
δ 0 j ( ∇ ∥ P e 1 + e N e 1 ∇ ∥ ϕ D K ) + m e v e i K e i j 0 J ∥ + 5 2 δ 1 j ( ∇ ∥ P e 1 + e P e 1 T e 0 ∇ ∥ ϕ D K ) = T e 0 ∑ k ( v e e K e e j k + v e i K e i j k ) A k , (129) \begin{aligned}
\delta_{0}^{j}\left(\nabla_{\parallel}P_{e1}+eN_{e1}\nabla_{\parallel}\phi_{DK}\right)+m_{e}v_{ei}K_{ei}^{j0}J_{\parallel} \\
&+\frac{5}{2}\delta_{1}^{j}\left(\nabla_{\parallel}P_{e1}+e\frac{P_{e1}}{T_{e0}}\nabla_{\parallel}\phi_{DK}\right)=T_{e0}\sum_{k}\left(v_{ee}K_{ee}^{jk}+v_{ei}K_{ei}^{jk}\right)A_{k}, \tag*{(129)}
\end{aligned} δ 0 j ( ∇ ∥ P e 1 + e N e 1 ∇ ∥ ϕ D K ) + m e v e i K e i j 0 J ∥ + 2 5 δ 1 j ( ∇ ∥ P e 1 + e T e 0 P e 1 ∇ ∥ ϕ D K ) = T e 0 k ∑ ( v ee K ee j k + v e i K e i j k ) A k , (129)
where we have defined Braginskii (1965); Parra (2019)
\begin{pmatrix}
K_{ee}^{00} & K_{ee}^{01} & K_{ee}^{02} & K_{ee}^{03} & \cdots\\
K_{ee}^{10} & K_{ee}^{11} & K_{ee}^{12} & K_{ee}^{13} & \cdots\\
K_{ee}^{20} & K_{ee}^{21} & K_{ee}^{22} & K_{ee}^{23} & \cdots\\
K_{ee}^{30} & K_{ee}^{31} & K_{ee}^{32} & K_{ee}^{33} & \cdots\\
\vdots& \vdots& \vdots& \vdots& \ddots
\end{pmatrix}
=\sqrt{2}
\begin{pmatrix}
0 & 0 & 0 & 0 & \cdots\\
0 & 1 & 3/4 & 15/32 & \cdots\\
0 & 3/4 & 45/16 & 309/128 & \cdots\\
0 & 15/32 & 309/128 & 5657/1024 & \cdots\\
\vdots& \vdots& \vdots& \vdots& \ddots
\end{pmatrix}, \tag*{(130)}\end{pmatrix}
\begin{pmatrix}
K_{ei}^{00} & K_{ei}^{01} & K_{ei}^{02} & K_{ei}^{03} & \cdots\\
K_{ei}^{10} & K_{ei}^{11} & K_{ei}^{12} & K_{ei}^{13} & \cdots\\
K_{ei}^{20} & K_{ei}^{21} & K_{ei}^{22} & K_{ei}^{23} & \cdots\\
K_{ei}^{30} & K_{ei}^{31} & K_{ei}^{32} & K_{ei}^{33} & \cdots\\
\vdots& \vdots& \vdots& \vdots& \ddots
\end{pmatrix}
=
\begin{pmatrix}
1 & 3/2 & 15/8 & 35/16 & \cdots\\
3/2 & 13/4 & 69/16 & 165/32 & \cdots\\
15/8 & 69/16 & 433/64 & 1077/128 & \cdots\\
35/16 & 165/32 & 1077/128 & 2957/256 & \cdots\\
\vdots& \vdots& \vdots& \vdots& \ddots
\end{pmatrix}. \tag*{(131)}\end{pmatrix}
We note that the j = 0 j=0 j = 0 equation of Eq. (A1d) is solved for using J ∥ = O ( c s n e ϵ ⊥ ) J_{\parallel}=\mathcal{O}(c_{s}n_{e}\epsilon_{\perp}) J ∥ = O ( c s n e ϵ ⊥ ) which gives the additional condition by combining Eq. (3.10a) with ( p , j ) = ( 1 , 0 ) (p,j)=(1,0) ( p , j ) = ( 1 , 0 ) and Eq. (A1d) with j = 0 j=0 j = 0 ,
1 m i ∇ ∥ P ∥ i D K − 1 m e ∇ ∥ P e 1 − e N e 1 m e ∇ ∥ ϕ D K = v e i ( J ∥ − T e 0 m e ∑ k K e i 0 k A k ) + O ( Ω i c s ϵ ⊥ 2 ) . (132) \frac{1}{m_{i}}\nabla_{\parallel}P_{\parallel iDK}-\frac{1}{m_{e}}\nabla_{\parallel}P_{e1}-\frac{eN_{e1}}{m_{e}}\nabla_{\parallel}\phi_{DK}=v_{ei}\left(J_{\parallel}-\frac{T_{e0}}{m_{e}}\sum_{k}K_{ei}^{0k}A_{k}\right)+\mathcal{O}\left(\Omega_{i}c_{s}\epsilon_{\perp}^{2}\right). \tag*{(132)} m i 1 ∇ ∥ P ∥ i D K − m e 1 ∇ ∥ P e 1 − m e e N e 1 ∇ ∥ ϕ D K = v e i ( J ∥ − m e T e 0 k ∑ K e i 0 k A k ) + O ( Ω i c s ϵ ⊥ 2 ) . (132)
We obtain from Eq. (A 1d) the following system,
∑ k ( K e e i k + K e i i k ) A k = [ C P 1 0 0 ] + m e n e T e U ∥ i D K K e i i 0 , (133) \sum_k \left(K_{ee}^{ik}+K_{ei}^{ik}\right) A_k =
\begin{bmatrix}
C_{P1}\\
0\\
0
\end{bmatrix}
+\frac{m_e n_e}{T_e}U_{\parallel iDK}K_{ei}^{i0}, \tag*{(133)} k ∑ ( K ee ik + K e i ik ) A k = C P 1 0 0 + T e m e n e U ∥ i D K K e i i 0 , (133)
with
C P 1 = 1 T e 0 ν e e ( ∇ ∥ P e 1 + e P e 1 T e 0 ∇ ∥ ϕ D K ) = N e 0 T e 0 ν e e ∇ ∥ T e 1 , (134) C_{P1}=\frac{1}{T_{e0}\nu_{ee}}\left(\nabla_{\parallel}P_{e1}+e\frac{P_{e1}}{T_{e0}}\nabla_{\parallel}\phi_{DK}\right)=\frac{N_{e0}}{T_{e0}\nu_{ee}}\nabla_{\parallel}T_{e1}, \tag*{(134)} C P 1 = T e 0 ν ee 1 ( ∇ ∥ P e 1 + e T e 0 P e 1 ∇ ∥ ϕ D K ) = T e 0 ν ee N e 0 ∇ ∥ T e 1 , (134)
where we have defined T e 1 = P e 1 / N e 0 T_{e1}=P_{e1}/N_{e0} T e 1 = P e 1 / N e 0 and used Eq. (3.2b) to relate ∇ ∥ ϕ D K \nabla_{\parallel}\phi_{DK} ∇ ∥ ϕ D K to ∇ ∥ N e 0 \nabla_{\parallel}N_{e0} ∇ ∥ N e 0 .
The solution to Eq. (A 4) truncated at j = 3 j=3 j = 3 gives Braginskii [4 ] ; Parra [31 ]
A 1 = 1.266 C P 1 + 0.286 m e T e 0 J ∥ , (135) A_1=1.266C_{P1}+0.286\frac{m_e}{T_{e0}}J_{\parallel}, \tag*{(135)} A 1 = 1.266 C P 1 + 0.286 T e 0 m e J ∥ , (135)
A 2 = − 0.654 C P 1 + 0.055 m e T e 0 J ∥ , (136) A_2=-0.654C_{P1}+0.055\frac{m_e}{T_{e0}}J_{\parallel}, \tag*{(136)} A 2 = − 0.654 C P 1 + 0.055 T e 0 m e J ∥ , (136)
A 3 = 0.0193 C P 1 + 0.0238 m e T e 0 J ∥ . (137) A_3=0.0193C_{P1}+0.0238\frac{m_e}{T_{e0}}J_{\parallel}. \tag*{(137)} A 3 = 0.0193 C P 1 + 0.0238 T e 0 m e J ∥ . (137)
Taking the moments of Eq. (32) with 1 1 1 , and m e v 2 m_e v^2 m e v 2 gives
∂ t N e 0 + 1 B [ ϕ D K + ϕ G K , N e 0 ] + ∇ ∥ ( N e 0 U ∥ i D K − J ∥ ) = S N , (138) \partial_tN_{e0}+\frac{1}{B}[\phi_{DK}+\phi_{GK},N_{e0}]+\nabla_{\parallel}\left(N_{e0}U_{\parallel iDK}-J_{\parallel}\right)=S_N, \tag*{(138)} ∂ t N e 0 + B 1 [ ϕ D K + ϕ G K , N e 0 ] + ∇ ∥ ( N e 0 U ∥ i D K − J ∥ ) = S N , (138)
3 ∂ t T e 1 + 3 B [ ϕ D K + ϕ G K , T e 0 ] + 2 N e 0 ∇ ∥ q ∥ e − 2 U ∥ e D K T e 0 N e 0 ∇ ∥ n e − 3 T e 0 N e 0 ∇ ∥ ( N e 0 U ∥ e D K ) = C e i 1 m v 2 + 2 S P e N e 0 − 3 T e 0 N e 0 S n , (139) \begin{aligned}
3\partial_tT_{e1}+\frac{3}{B}[\phi_{DK}+\phi_{GK},T_{e0}]+\frac{2}{N_{e0}}\nabla_{\parallel}q_{\parallel e}\\
-2U_{\parallel eDK}\frac{T_{e0}}{N_{e0}}\nabla_{\parallel}n_e-3\frac{T_{e0}}{N_{e0}}\nabla_{\parallel}\left(N_{e0}U_{\parallel eDK}\right)
&=C_{ei1}^{mv^2}+2\frac{S_{Pe}}{N_{e0}}-3\frac{T_{e0}}{N_{e0}}S_n, \tag*{(139)}
\end{aligned} 3 ∂ t T e 1 + B 3 [ ϕ D K + ϕ G K , T e 0 ] + N e 0 2 ∇ ∥ q ∥ e − 2 U ∥ eD K N e 0 T e 0 ∇ ∥ n e − 3 N e 0 T e 0 ∇ ∥ ( N e 0 U ∥ eD K ) = C e i 1 m v 2 + 2 N e 0 S P e − 3 N e 0 T e 0 S n , (139)
3 ∂ t T e 0 + 3 B [ ϕ D K + ϕ G K , T e 0 ] + 1 N e 0 ∇ ∥ ( 2 q ∥ e + 3 T e 0 J ∥ ) − 2 U ∥ i D K T e 0 N e 0 ∇ ∥ N e 0 + 2 J ∥ T e 0 N e 0 2 ∇ ∥ N e 0 − 3 T e 0 N e 0 ∇ ∥ ( N e 0 U ∥ i D K ) = C e i 1 m v 2 + 2 S P e N e 0 − 3 T e 0 N e 0 S n , (140) \begin{aligned}
3\partial_tT_{e0}+\frac{3}{B}[\phi_{DK}+\phi_{GK},T_{e0}]+\frac{1}{N_{e0}}\nabla_{\parallel}\left(2q_{\parallel e}+3T_{e0}J_{\parallel}\right)\\
-2U_{\parallel iDK}\frac{T_{e0}}{N_{e0}}\nabla_{\parallel}N_{e0}+2J_{\parallel}\frac{T_{e0}}{N_{e0}^{2}}\nabla_{\parallel}N_{e0}-3\frac{T_{e0}}{N_{e0}}\nabla_{\parallel}\left(N_{e0}U_{\parallel iDK}\right)
&=C_{ei1}^{mv^2}+2\frac{S_{Pe}}{N_{e0}}-3\frac{T_{e0}}{N_{e0}}S_n, \tag*{(140)}
\end{aligned} 3 ∂ t T e 0 + B 3 [ ϕ D K + ϕ G K , T e 0 ] + N e 0 1 ∇ ∥ ( 2 q ∥ e + 3 T e 0 J ∥ ) − 2 U ∥ i D K N e 0 T e 0 ∇ ∥ N e 0 + 2 J ∥ N e 0 2 T e 0 ∇ ∥ N e 0 − 3 N e 0 T e 0 ∇ ∥ ( N e 0 U ∥ i D K ) = C e i 1 m v 2 + 2 N e 0 S P e − 3 N e 0 T e 0 S n , (140)
with
q ∥ e = 1 2 ∫ d v m e v ∥ v 2 F e 1 = T e 0 ( 5 2 N e 0 U ∥ e D K − 5 2 A 1 m e ) , (141) q_{\parallel e}=\frac{1}{2}\int dv\,m_e v_{\parallel}v^2F_{e1}=T_{e0}\left(\frac{5}{2}N_{e0}U_{\parallel eDK}-\frac{5}{2}\frac{A_1}{m_e}\right), \tag*{(141)} q ∥ e = 2 1 ∫ d v m e v ∥ v 2 F e 1 = T e 0 ( 2 5 N e 0 U ∥ eD K − 2 5 m e A 1 ) , (141)
∥ S e ∥ e = S N \lVert S_e\rVert_e=S_N ∥ S e ∥ e = S N , ∥ m e v 2 S e / 2 ∥ e = S P e \lVert m_e v^2S_e/2\rVert_e=S_{Pe} ∥ m e v 2 S e /2 ∥ e = S P e , and ∥ m e v 2 C e i / 2 ∥ e = C e i 1 m v 2 \lVert m_e v^2C_{ei}/2\rVert_e=C_{ei1}^{mv^2} ∥ m e v 2 C e i /2 ∥ e = C e i 1 m v 2 .
Finally, we define N e D K = N e 0 + N e 1 N_{eDK}=N_{e0}+N_{e1} N eD K = N e 0 + N e 1 and T e D K = T e 0 + T e 1 T_{eDK}=T_{e0}+T_{e1} T eD K = T e 0 + T e 1 , neglecting C e i 1 m v 2 C_{ei1}^{mv^2} C e i 1 m v 2 and adding a few terms to Eqs. (A 7a) and (A 7c) noticing that these does not change the order of validity regime of the equations, we arrive at Eqs. (3.5a) and (39) having defined S T e = S P e / 3 N e D K − T e D K S N / N e D K S_{T_e}=S_{Pe}/3N_{eDK}-T_{eDK}S_N/N_{eDK} S T e = S P e /3 N eD K − T eD K S N / N eD K . It is noted that Eqs. (A 7a) and (A 7c) have some higher order terms, however, these are added such that they exactly correspond to the drift-reduced-Braginskii model Parra [31 ] ; Zeiler et al. [42 ] .
Equation (A 3) represents nothing more than force balance for electrons and ions. In order to maintain quasi neutrality, we need J ∥ ∼ 0 J_{\parallel} \sim0 J ∥ ∼ 0 which for large ν e i \nu_{ei} ν e i is obviously satisfied. For low ν e i \nu_{ei} ν e i , if there is an imbalance in the accelerations from the electron and ion pressure, a large electric field builds up to accelerate the electrons such that U ∥ e ∼ U ∥ i U_{\parallel e} \sim U_{\parallel i} U ∥ e ∼ U ∥ i .
Hermite-Laguerre projection of gyroaveraged quantities
For completeness, the complicated expression for the gyroaveraged quantities in Eqs. (4.4a) and (106) are reported. The linear expressions (in ϕ G K \phi_{GK} ϕ G K ) are given by Frei et al. [9 ]
∥ ∂ μ ⟨ ⟨ ϕ G K ⟩ R / x † ∥ i = − ∑ k B T i 0 ∑ n ≥ 0 , m ≥ 1 K n ( b i ) K m ( b i ) ∑ l = 0 m − 1 ∑ f = 0 n + l d n l 0 f N i D K 0 f ϕ G K ( k ) e i k ⋅ x , (142) \left\lVert\partial_{\mu}\left\langle\left\langle\phi_{GK}\right\rangle_{R}/x^{\dagger}\right\rVert_{i}\right. = -\sum_{k}\frac{B}{T_{i0}}\sum_{\substack{n\geq0,m\geq1}}K_{n}(b_{i})K_{m}(b_{i})\sum_{l=0}^{m-1}\sum_{f=0}^{n+l}d_{nl}^{0f}\mathcal{N}_{iDK}^{0f}\phi_{GK}(k)e^{i\boldsymbol{k}\cdot\boldsymbol{x}}, \tag*{(142)} ∂ μ ⟨ ⟨ ϕ G K ⟩ R / x † i = − k ∑ T i 0 B n ≥ 0 , m ≥ 1 ∑ K n ( b i ) K m ( b i ) l = 0 ∑ m − 1 f = 0 ∑ n + l d n l 0 f N i D K 0 f ϕ G K ( k ) e i k ⋅ x , (142)
Q i ∗ = ∑ k ∑ n ≥ 0 K n ( b i ) N i G K 0 n ( k ) e i k ⋅ x , (143) Q_{i}^{*}=\sum_{k}\sum_{n\geq0}K_{n}(b_{i})\mathcal{N}_{iGK}^{0n}(k)e^{i\boldsymbol{k}\cdot\boldsymbol{x}}, \tag*{(143)} Q i ∗ = k ∑ n ≥ 0 ∑ K n ( b i ) N i G K 0 n ( k ) e i k ⋅ x , (143)
∥ ⟨ ϕ G K ⟩ R ∥ i p j = ∑ k ∑ n ≥ 0 D n p j n ϕ G K ( k ) e i k ⋅ R , (144) \left\lVert\left\langle\phi_{GK}\right\rangle_{R}\right\rVert_{i}^{pj}=\sum_{k}\sum_{n\geq0}D_{n}^{pjn}\phi_{GK}(k)e^{i\boldsymbol{k}\cdot\boldsymbol{R}}, \tag*{(144)} ∥ ⟨ ϕ G K ⟩ R ∥ i p j = k ∑ n ≥ 0 ∑ D n p j n ϕ G K ( k ) e i k ⋅ R , (144)
D n p k l = ∑ r = 0 l + k d l k r 0 N i D K p r K n ( b i ) , (145) D_{n}^{pkl}=\sum_{r=0}^{l+k}d_{lkr}^{0}\mathcal{N}_{iDK}^{pr}K_{n}(b_{i}), \tag*{(145)} D n p k l = r = 0 ∑ l + k d l k r 0 N i D K p r K n ( b i ) , (145)
d ˉ r j f = ∑ r 1 = 0 r ∑ j 1 = 0 j ∑ f 1 = 0 f L j 1 j − 1 / 2 L r 1 r n − 1 / 2 L f 1 f − 1 / 2 ( r 1 + j 1 + f 1 ) ! , (146) \bar{d}_{rjf}=\sum_{r_{1}=0}^{r}\sum_{j_{1}=0}^{j}\sum_{f_{1}=0}^{f}L_{j_{1}j}^{-1/2}L_{r_{1}r}^{n-1/2}L_{f_{1}f}^{-1/2}(r_{1}+j_{1}+f_{1})!, \tag*{(146)} d ˉ r j f = r 1 = 0 ∑ r j 1 = 0 ∑ j f 1 = 0 ∑ f L j 1 j − 1/2 L r 1 r n − 1/2 L f 1 f − 1/2 ( r 1 + j 1 + f 1 )! , (146)
L n l m = ( − 1 ) l ( m + n + 1 / 2 ) ! ( n − l ) ! ( m + l + 1 / 2 ) ! l ! , (147) L_{nl}^{m}=\frac{(-1)^{l}(m+n+1/2)!}{(n-l)!(m+l+1/2)!l!}, \tag*{(147)} L n l m = ( n − l )! ( m + l + 1/2 )! l ! ( − 1 ) l ( m + n + 1/2 )! , (147)
K n ( b i ) = 1 n ! ( b i 2 ) 2 n e − ( b i / 2 ) 2 , b i = k ⊥ 2 T i 0 Ω i m i , (148) K_{n}(b_{i})=\frac{1}{n!}\left(\frac{b_{i}}{2}\right)^{2n}e^{-(b_{i}/2)^{2}},\qquad b_{i}=\frac{k_{\perp}\sqrt{2T_{i0}}}{\Omega_{i}\sqrt{m_{i}}}, \tag*{(148)} K n ( b i ) = n ! 1 ( 2 b i ) 2 n e − ( b i /2 ) 2 , b i = Ω i m i k ⊥ 2 T i 0 , (148)
while the non-linear projections are
∥ ⟨ ψ ⟩ R − ⟨ ϕ G K ⟩ R ∥ i p j = q i 2 2 m i Ω i 2 ∥ ∂ μ ( ⟨ ϕ G K ⟩ R 2 − ⟨ ϕ G K 2 ⟩ R ) ∥ i p j + q i 2 m i Ω i 2 ∥ ⟨ ( b × ∇ ϕ ~ G K ) ⋅ ∇ ϕ ~ G K ⟩ R ∥ i p j (149) \left\lVert\langle\psi\rangle_R-\langle\phi_{GK}\rangle_R\right\rVert_i^{pj}
=
\frac{q_i^2}{2m_i\Omega_i^2}
\left\lVert\partial_\mu\left(\langle\phi_{GK}\rangle_R^2-\langle\phi_{GK}^2\rangle_R\right)\right\rVert_i^{pj}
+
\frac{q_i}{2m_i\Omega_i^2}
\left\lVert\left\langle\left(\mathbf{b}\times\nabla\widetilde{\phi}_{GK}\right)\cdot\nabla\widetilde{\phi}_{GK}\right\rangle_R\right\rVert_i^{pj} \tag*{(149)} ∥ ⟨ ψ ⟩ R − ⟨ ϕ G K ⟩ R ∥ i p j = 2 m i Ω i 2 q i 2 ∂ μ ( ⟨ ϕ G K ⟩ R 2 − ⟨ ϕ G K 2 ⟩ R ) i p j + 2 m i Ω i 2 q i ⟨ ( b × ∇ ϕ G K ) ⋅ ∇ ϕ G K ⟩ R i p j (149)
∥ ∂ μ ( ⟨ ϕ G K ⟩ R 2 − ⟨ ϕ G K 2 ⟩ R ) ∥ i p j = ∑ k , k ′ , n = 0 ∞ 1 n + 1 [ K n ( ∣ b i + b i ′ ∣ ) ∣ k ⊥ + k ⊥ ′ ∣ 2 ∑ f = 0 j + n d n j f 1 N i D K p f − ∑ m = 0 ∞ ( K n ( b i ) K m ( b i ′ ) k ⊥ ′ 2 + K n ( b i ′ ) K m ( b i ) ( k ⊥ ′ ) 2 ) ∑ f = 0 m + n d n m f 1 ∑ s = 0 f + j d f j s 0 N i D K p s ] ϕ G K ( k ) ϕ G K ( k ′ ) e i ( k + k ′ ) ⋅ R (150) \left\lVert\partial_\mu\left(\langle\phi_{GK}\rangle_R^2-\langle\phi_{GK}^2\rangle_R\right)\right\rVert_i^{pj}
=
\sum_{k,k',n=0}^{\infty}\frac{1}{n+1}
\left[
K_n\left(\lvert\mathbf{b}_i+\mathbf{b}'_i\rvert\right)\lvert\mathbf{k}_{\perp}+\mathbf{k}'_{\perp}\rvert^2
\sum_{f=0}^{j+n}d_{njf}^{1}\mathcal{N}_{iDK}^{pf}
-
\sum_{m=0}^{\infty}
\left(K_n(b_i)K_m(b'_i)k_{\perp}^{\prime2}+K_n(b'_i)K_m(b_i)(k'_{\perp})^2\right)
\sum_{f=0}^{m+n}d_{nmf}^{1}\sum_{s=0}^{f+j}d_{fjs}^{0}\mathcal{N}_{iDK}^{ps}
\right]\phi_{GK}(\mathbf{k})\phi_{GK}(\mathbf{k}')e^{i(\mathbf{k}+\mathbf{k}')\cdot\mathbf{R}} \tag*{(150)} ∂ μ ( ⟨ ϕ G K ⟩ R 2 − ⟨ ϕ G K 2 ⟩ R ) i p j = k , k ′ , n = 0 ∑ ∞ n + 1 1 K n ( ∣ b i + b i ′ ∣ ) ∣ k ⊥ + k ⊥ ′ ∣ 2 f = 0 ∑ j + n d nj f 1 N i D K p f − m = 0 ∑ ∞ ( K n ( b i ) K m ( b i ′ ) k ⊥ ′2 + K n ( b i ′ ) K m ( b i ) ( k ⊥ ′ ) 2 ) f = 0 ∑ m + n d nm f 1 s = 0 ∑ f + j d f j s 0 N i D K p s ϕ G K ( k ) ϕ G K ( k ′ ) e i ( k + k ′ ) ⋅ R (150)
while the final term is a bit more tricky, without projecting, it can be expressed as
⟨ ( b × ∇ ϕ ~ G K ) ⋅ ∇ ϕ ~ G K ⟩ R = 2 ∑ k , k ′ , n > 0 b × k ⋅ k ′ ( − 1 ) n n J n ( b ′ ) J n ( b ) sin n α ϕ G K ( k ) ϕ G K ( k ′ ) e i ( k + k ′ ) ⋅ R , (151) \left\langle\left(\mathbf{b}\times\nabla\widetilde{\phi}_{GK}\right)\cdot\nabla\widetilde{\phi}_{GK}\right\rangle_R
=
2\sum_{\mathbf{k},\mathbf{k}',n>0}\mathbf{b}\times\mathbf{k}\cdot\mathbf{k}'\frac{(-1)^n}{n}J_n(b')J_n(b)\sin n\alpha\,
\phi_{GK}(\mathbf{k})\phi_{GK}(\mathbf{k}')e^{i(\mathbf{k}+\mathbf{k}')\cdot\mathbf{R}}, \tag*{(151)} ⟨ ( b × ∇ ϕ G K ) ⋅ ∇ ϕ G K ⟩ R = 2 k , k ′ , n > 0 ∑ b × k ⋅ k ′ n ( − 1 ) n J n ( b ′ ) J n ( b ) sin n α ϕ G K ( k ) ϕ G K ( k ′ ) e i ( k + k ′ ) ⋅ R , (151)
where we have introduced α \alpha α as the angle between k \mathbf{k} k and k ′ \mathbf{k}' k ′ such that k ⋅ k ′ = k k ′ cos α \mathbf{k}\cdot\mathbf{k}'=kk'\cos\alpha k ⋅ k ′ = k k ′ cos α and the Bessel function of the first kind of order n n n , J n J_n J n . Finally we truncate the sum in Eq. (B 2) at n = 1 n=1 n = 1 for simplicity and get
∥ ⟨ ( b × ∇ ϕ ~ G K ) ⋅ ∇ ϕ ~ G K ⟩ R ∥ i p j = T i 0 m i Ω i 2 ( b × k ⋅ k ′ ) 2 ∑ n = 0 ∞ ∑ m = 0 ∞ 1 n + 1 K n ( b i ) K m ( b i ′ ) ( ∑ f = 0 m + n d n m f 1 − ∑ f = 0 m + 1 + n d n ( m + 1 ) f 1 ) ∑ s = 0 f + j d f j s 0 N i D K p s ϕ G K ( k ) ϕ G K ( k ′ ) e i ( k + k ′ ) ⋅ R . (152) \left\lVert\left\langle\left(\mathbf{b}\times\nabla\widetilde{\phi}_{GK}\right)\cdot\nabla\widetilde{\phi}_{GK}\right\rangle_R\right\rVert_i^{pj}
=
\frac{T_{i0}}{m_i\Omega_i^2}(\mathbf{b}\times\mathbf{k}\cdot\mathbf{k}')^2
\sum_{n=0}^{\infty}\sum_{m=0}^{\infty}\frac{1}{n+1}K_n(b_i)K_m(b'_i)
\left(
\sum_{f=0}^{m+n}d_{nmf}^{1}
-
\sum_{f=0}^{m+1+n}d_{n(m+1)f}^{1}
\right)
\sum_{s=0}^{f+j}d_{fjs}^{0}\mathcal{N}_{iDK}^{ps}\phi_{GK}(\mathbf{k})\phi_{GK}(\mathbf{k}')e^{i(\mathbf{k}+\mathbf{k}')\cdot\mathbf{R}}. \tag*{(152)} ⟨ ( b × ∇ ϕ G K ) ⋅ ∇ ϕ G K ⟩ R i p j = m i Ω i 2 T i 0 ( b × k ⋅ k ′ ) 2 n = 0 ∑ ∞ m = 0 ∑ ∞ n + 1 1 K n ( b i ) K m ( b i ′ ) f = 0 ∑ m + n d nm f 1 − f = 0 ∑ m + 1 + n d n ( m + 1 ) f 1 s = 0 ∑ f + j d f j s 0 N i D K p s ϕ G K ( k ) ϕ G K ( k ′ ) e i ( k + k ′ ) ⋅ R . (152)
Finally for the projection on the GK distribution function, we simply replace N i D K p j \mathcal{N}_{iDK}^{pj} N i D K p j with N i G K p j \mathcal{N}_{iGK}^{pj} N i G K p j in Eq. (B 1c). For all the sums, we truncate maximally at J D K J_{DK} J D K and J G K J_{GK} J G K for the DK and GK projections, respectively [20 a] .
Linear dispersion relation of the GK Kelvin-Helmholtz mode
In this appendix, we derive a dispersion relation for the k ∥ = 0 k_{\parallel}=0 k ∥ = 0 , m ≫ 1 m\gg1 m ≫ 1 mode observed for the full model in Figure 12 . We go to a slab geometry with x x x the radial direction and y y y the poloidal direction. We assume that the ion DK distribution function is a fixed Maxwellian with density N D K N_{DK} N D K and temperature T i T_i T i , furthermore, we approximate the ion GK distribution function as an isotropic distribution function, also, with temperature T i T_i T i . We linearize all fields, f f f , with the equilibrium f ˉ = f ˉ ( x ) \bar{f}=\bar{f}(x) f ˉ = f ˉ ( x ) and the fluctuating part f ~ ( x ) exp ( − i ω t + i k y y ) \tilde{f}(x)\exp(-i\omega t+ik_y y) f ~ ( x ) exp ( − iω t + i k y y ) . We use Eq. (5.14e), perform the gyroaverage operation on Eq. (4.4a) with ( p , j ) = ( 0 , 0 ) (p,j)=(0,0) ( p , j ) = ( 0 , 0 ) while only including terms linear in ε 0 \varepsilon_0 ε 0 . Finally, we close the system with Eq. (106) and to get
∇ ⊥ ⋅ ( − i ω ∗ N ˉ D K ∇ ⊥ ϕ ~ D K − i k y ∗ N ˉ D K ∂ x ∇ ⊥ ϕ ˉ D K ϕ ~ t o t ) = 0 , (153) \nabla_\perp\cdot\left(-i\omega^*\bar{N}_{DK}\nabla_\perp\tilde{\phi}_{DK}-ik_y^*\bar{N}_{DK}\partial_x\nabla_\perp\bar{\phi}_{DK}\tilde{\phi}_{tot}\right)=0, \tag*{(153)} ∇ ⊥ ⋅ ( − i ω ∗ N ˉ D K ∇ ⊥ ϕ ~ D K − i k y ∗ N ˉ D K ∂ x ∇ ⊥ ϕ ˉ D K ϕ ~ t o t ) = 0 , (153)
− i ω ρ ~ i ∗ + i k y ∗ ( ∂ x ϕ ~ D K ρ ˉ i ∗ − ∂ x ρ ˉ i ∗ ϕ ~ D K ) = 0 , (154) -i\omega\tilde{\rho}_i^*+ik_y^*\left(\partial_x\tilde{\phi}_{DK}\bar{\rho}_i^*-\partial_x\bar{\rho}_i^*\tilde{\phi}_{DK}\right)=0, \tag*{(154)} − iω ρ ~ i ∗ + i k y ∗ ( ∂ x ϕ ~ D K ρ ˉ i ∗ − ∂ x ρ ˉ i ∗ ϕ ~ D K ) = 0 , (154)
ρ ~ i ∗ = N ˉ D K k y ϕ ~ G K , (155) \tilde{\rho}_i^*= \bar{N}_{DK}k_y\tilde{\phi}_{GK}, \tag*{(155)} ρ ~ i ∗ = N ˉ D K k y ϕ ~ G K , (155)
with
I y = e ( 1 T e + 1 T i − 1 T i exp ( − b i y 2 2 ) I 0 ( b i y 2 2 ) ) , (156) I_y=e\left(\frac{1}{T_e}+\frac{1}{T_i}-\frac{1}{T_i}\exp\left(-\frac{b_{iy}^2}{2}\right)I_0\left(\frac{b_{iy}^2}{2}\right)\right), \tag*{(156)} I y = e ( T e 1 + T i 1 − T i 1 exp ( − 2 b i y 2 ) I 0 ( 2 b i y 2 ) ) , (156)
k y ∗ = k y B ρ ∗ , ϕ t o t = ϕ D K + ϕ G K , and ω ∗ = ω − k y ∗ ∂ x ϕ ˉ t o t . (157) k_y^*=\frac{k_y}{B\rho_*},\qquad\phi_{tot}=\phi_{DK}+\phi_{GK},\qquad\text{and}\qquad\omega^*=\omega-k_y^*\partial_x\bar{\phi}_{tot}. \tag*{(157)} k y ∗ = B ρ ∗ k y , ϕ t o t = ϕ D K + ϕ G K , and ω ∗ = ω − k y ∗ ∂ x ϕ ˉ t o t . (157)
Neglecting the long-wavelength contribution of Eq. (C1a), i.e. only keeping terms linear in ϕ G K \phi_{GK} ϕ G K and combining the equations lead to
− ∇ ⊥ ⋅ ( k y ∗ ∂ x ϕ G K ∇ ⊥ ϕ ~ D K + ( k y ∗ ) 2 ( ω − k y ∗ ∂ x ϕ ˉ D K ) I y ∂ x ∇ ⊥ ϕ D K ∂ x ρ ~ i ∗ ϕ ~ D K ) = 0. (158) -\nabla_\perp\cdot\left(k_y^*\partial_x\phi_{GK}\nabla_\perp\tilde{\phi}_{DK}+\frac{(k_y^*)^2}{(\omega-k_y^*\partial_x\bar{\phi}_{DK})I_y}\partial_x\nabla_\perp\phi_{DK}\partial_x\tilde{\rho}_i^*\tilde{\phi}_{DK}\right)=0. \tag*{(158)} − ∇ ⊥ ⋅ ( k y ∗ ∂ x ϕ G K ∇ ⊥ ϕ ~ D K + ( ω − k y ∗ ∂ x ϕ ˉ D K ) I y ( k y ∗ ) 2 ∂ x ∇ ⊥ ϕ D K ∂ x ρ ~ i ∗ ϕ ~ D K ) = 0. (158)
Assuming that ∂ x ϕ D K \partial_x\phi_{DK} ∂ x ϕ D K and ∂ x ϕ G K \partial_x\phi_{GK} ∂ x ϕ G K are piecewise linear around a region of x 0 x_0 x 0
∂ x ϕ ˉ D K = { − V 0 , x < − x 0 , V 0 x x 0 , − x 0 < x < x 0 , V 0 , x > x 0 , (159) \partial_x\bar{\phi}_{DK}=
\begin{cases}
-V_0, & x<-x_0,\\
V_0\frac{x}{x_0}, & -x_0<x<x_0,\\
V_0, & x>x_0,
\end{cases} \tag*{(159)} ∂ x ϕ ˉ D K = ⎩ ⎨ ⎧ − V 0 , V 0 x 0 x , V 0 , x < − x 0 , − x 0 < x < x 0 , x > x 0 , (159)
∂ x ϕ ˉ G K = { − V 1 , x < − x 0 , V 1 x x 0 , − x 0 < x < x 0 , V 1 , x > x 0 , (160) \partial_x\bar{\phi}_{GK}=
\begin{cases}
-V_1, & x<-x_0,\\
V_1\frac{x}{x_0}, & -x_0<x<x_0,\\
V_1, & x>x_0,
\end{cases} \tag*{(160)} ∂ x ϕ ˉ G K = ⎩ ⎨ ⎧ − V 1 , V 1 x 0 x , V 1 , x < − x 0 , − x 0 < x < x 0 , x > x 0 , (160)
while ϕ ~ D K = V n \tilde{\phi}_{DK}=V_n ϕ ~ D K = V n is constant. We assume that V 0 ≪ V 1 V_0\ll V_1 V 0 ≪ V 1 such that a general solution for ϕ ~ D K \tilde{\phi}_{DK} ϕ ~ D K is
ϕ ~ D K = { ( A 1 + A 2 exp ( 2 k y x 0 ) ) exp ( k y x ) , x < − x 0 , A 1 exp ( k y x ) + A 2 exp ( − k y x ) , − x 0 < x < x 0 , ( A 2 + A 1 exp ( 2 k y x 0 ) ) exp ( − k y x ) , x > x 0 , (161) \tilde{\phi}_{DK}=
\begin{cases}
(A_1+A_2\exp(2k_yx_0))\exp(k_yx), & x<-x_0,\\
A_1\exp(k_yx)+A_2\exp(-k_yx), & -x_0<x<x_0,\\
(A_2+A_1\exp(2k_yx_0))\exp(-k_yx), & x>x_0,
\end{cases} \tag*{(161)} ϕ ~ D K = ⎩ ⎨ ⎧ ( A 1 + A 2 exp ( 2 k y x 0 )) exp ( k y x ) , A 1 exp ( k y x ) + A 2 exp ( − k y x ) , ( A 2 + A 1 exp ( 2 k y x 0 )) exp ( − k y x ) , x < − x 0 , − x 0 < x < x 0 , x > x 0 , (161)
where we have used that ϕ ~ D K \tilde{\phi}_{DK} ϕ ~ D K vanishes at − ∞ -\infty − ∞ and ∞ \infty ∞ and is continuous. Finally, we integrate Eq. (158) over each discontinuity (x = ± x 0 x=\pm x_0 x = ± x 0 ) and obtain a matching condition
\begin{bmatrix}
\frac{(k_y^*)^2 V_0 V_n}{(\omega+k_y^* V_0)x_0 I_y}e^{-2k_yx_0}-2k_yk_y^*V_1+\frac{(k_y^*)^2 V_0 V_n}{(\omega+k_y^* V_0)x_0 I_y} & -2k_yk_y^*V_1+\frac{(k_y^*)^2 V_0 V_n}{(\omega+k_y^* V_0)x_0 I_y}\\
2k_yk_y^*V_1+\frac{(k_y^*)^2 V_0 V_n}{(\omega-k_y^* V_0)x_0 I_y} & \frac{(k_y^*)^2 V_0 V_n}{(\omega-k_y^* V_0)x_0 I_y}e^{-2k_yx_0}
\end{bmatrix}
\begin{bmatrix}
A_1\\
A_2
\end{bmatrix}=0. \tag*{(162)}\end{bmatrix}
For ϕ ~ D K \tilde{\phi}_{DK} ϕ ~ D K to be able to grow, this system of equations needs to have infinitely many solutions, implying that the determinant of the matrix in Eq. (162) needs to be 0. This requirement restricts ω \omega ω
ω 2 = ( k y ∗ V 0 V n k y V 1 x 0 I y − 2 k y ∗ V 0 ) 2 − ( k y ∗ V 0 V n k y V 1 x 0 I y ) 2 e − 4 k y x 0 . (163) \omega^2=\left(\frac{k_y^*V_0V_n}{k_yV_1x_0I_y}-2k_y^*V_0\right)^2-\left(\frac{k_y^*V_0V_n}{k_yV_1x_0I_y}\right)^2e^{-4k_yx_0}. \tag*{(163)} ω 2 = ( k y V 1 x 0 I y k y ∗ V 0 V n − 2 k y ∗ V 0 ) 2 − ( k y V 1 x 0 I y k y ∗ V 0 V n ) 2 e − 4 k y x 0 . (163)
For the instability to grow, we need ω 2 < 0 \omega^2<0 ω 2 < 0 which implies
( k y ∗ V 0 V n k y V 1 x 0 I y − 2 k y ∗ V 0 ) 2 < ( k y ∗ V 0 V n k y V 1 x 0 I y ) 2 e − 4 k y x 0 . (164) \left(\frac{k_y^*V_0V_n}{k_yV_1x_0I_y}-2k_y^*V_0\right)^2<\left(\frac{k_y^*V_0V_n}{k_yV_1x_0I_y}\right)^2e^{-4k_yx_0}. \tag*{(164)} ( k y V 1 x 0 I y k y ∗ V 0 V n − 2 k y ∗ V 0 ) 2 < ( k y V 1 x 0 I y k y ∗ V 0 V n ) 2 e − 4 k y x 0 . (164)
As can be seen, if V n V_n V n and V 1 V_1 V 1 have the same sign, this relation can only be satisfied for small k y x 0 < 1 / 4 k_yx_0<1/4 k y x 0 < 1/4 (necessary but not sufficient condition). In the limit x 0 → 0 x_0\to0 x 0 → 0 , we get 1 > V n / ( V 1 I y ) 1>V_n/(V_1I_y) 1 > V n / ( V 1 I y ) and the instability condition in Eq. (125) is recovered. It is noted that this instability is limited by diffusion (numeric and physical through the collision operator) which restricts the development of small x 0 x_0 x 0 . Finally, the instability gives rise to more small-scale gradients in ϕ ~ D K \tilde{\phi}_{DK} ϕ ~ D K which can drive a cascade of small-scale instabilities.