IV. NONLINEAR ION LANDAU DAMPING OF KINETIC ALFVEN WAVES
As noted in Sec. I, the results obtained in Sec. II are rather general and, therefore, are also useful for describing nonlinear processes associated with low-frequency wave heating. To illustrate this potential application, we consider here nonlinear ion Landau damping of the mode-converted kinetic (shear) Alfven waves. [1] The major difference between our work and Ref. 16 is that we include the diamagnetic drift effects. Another minor difference is that no small ion-larmor-radius expansion is taken here.
For the sake of simplicity, we adopt the WKB description, Eqs. (49) and (50), and ignore the geometrical effects. Furthermore, Fo is taken to be Maxwellian and for the present purpose we can assume a weak-turbulence subsidiary ordering as well as ignore the compressional Alfven perturbations δA1=0 [i.e., β≡(plasma pressure/magnetic pressure)<<1 ]. Thus, the relevant equations are
where δL~(∼k)=δϕ~(∼k)−v∥δA~∥(∼k)/c,ω∗k=kycT/eBLn,Ln−1={dlnN0/dx} , x is the nonuniformity direction, and temperature gradients are ignored. Expanding δH~0=δH0(1)+δH0(2)+δH0(3)+… , we have, in the linear order,
To calculate δHo(2),(3) , we let (ω∼k,k) and (ω∼k′,k′) be the normal modes, while (ωq=ω∼k−ω∼k′,q=k−k′) is the virtual or quasi mode. Also, we have in mind
and δδ~≡eδϕ~/Te . In deriving Eq. (A5), we have observed that ∣δϕ~∣∼∣ωδAˉ∣∣/ck∣∣ for the k and k' modes. As to electron nonlinearity, it is 0(∣ω/ω∗e∣q) smaller and, hence, negligible. Furthermore, for β≪1 , we have ∣ωq∣≪∣q∥vA∣ , and the q-mode response is predominantly electrostatic. Here VA is the Aifven speed. The quasi-neutrality condition then yields
From Eq. (90) we can readily calculate δn1(3)(∼k) . Meanwhile, the electron nonlinearity, again is found to be O(∣ω/ω∗e∣∼g) smaller. Substituting δn1(3)(∼k) and the linear response into the quasi-neutrality condition and parallel Ampere's law, it is straightforward to derive the following nonlinear dispersion relation
zig≡z(ω/g/∣qij∣vi) with Z being the standard plasma dispersion function.
In the low- β limit of interest here, ∣ω∗e/kIVA∣k<1 , and assuming ωk=ωrk+iωik with ∣ωl/ωr∣<<1 , Eq. (92) gives the known result that there exist two branches of waves; the drift waves with
We note that nonlinear ion Landau damping of the drift waves, as described by Eq. (91), has previously been investigated. 17 Concentrating on the kinetic shear-Alfven waves, we find
Comparing the ωi given by Eq. (96) to that of Ref. 16 (which, we recall, does not include the ω∗ effects), we find the present parametric-decay growth rate is larger by a factor of 0(∣ω∗iq∣/∣q∥vi∣)>1 . Furthermore, noting that ImZiq=∣qi∣viπδ(ωk′−ωk) , G(k,k′)≥0 , and ωi∝(ω∗k′−ω∗k)∝(ky′−ky) ; hence, the daughter wave (ωk,k) has a smaller ky but the same ω as the pump wave (ωk′,k′) . This is qualitatively different from the results of Ref. 16, where ωi∝(ωk′−ωk) . In fact, as might be expected, one can show that the results of Ref. 16 are valid for ∣ω∗iq∣<∣q∥vi∣ , i.e., in the opposite limit.
Finally, we briefly discuss the implication of the present results on the shear Alfven wave heating scheme. Since the pump wave (i.e., the mode-converted kinetic shear Alfven wave) was kx′ρ1∼0(1) and ky′ρ1∼0(ρ1/a)≪1 with a being the tokamak minor radius, we have ωi=∣k×k′⋅en∣2=(kx′ky)2 , and hence nonlinear coupling becomes appreciable for kyρ1∼0(1)>>ky′ρ1 ; that is, ωi<0 . Thus, our results suggest that parametric decay through nonlinear ion Landau damping will be, in general, unlikely to occur.
5.2. The kinetic Alfvén wave ion-ion streaming instability
For our second test, we consider the excitation of kinetic Alfvén waves (KAW) by ion-ion streaming in 2D-3V [54]. This is a moderately magnetized case, with an imposed external magnetic field. The instability is caused by interactions between the wave and the streaming ions. The simulation parameters are similar to those used in Ref. [54]. The mass ratio is mi/me=25 , which is nominal (we will vary this parameter later). We use ions as the reference species. The simulated domain is 10(di)×10(di) (the unit length being the ion skin depth), with 64 64 uniformly distributed cells (with each cell size in each direction about 10 times larger than the Debye length) and periodic boundary conditions, and the average number of particles per cell of one species is 500. The external magnetic field is set to be B0=0.0667 along the x-axis. The plasma consists of Maxwellian electrons with veT=0.0745(βe=0.1) , and two singly charged ion components, i.e., an ambient ion component a and an ion beam component b, with number densities na=0.6ne and nb=0.4ne (where ne is the electron density). The two ion components have vaT=1.925×10−2 and vbT=7.45×10−3 , and a relative streaming speed with respect to each other of vad=1.5vA and vbd=vA , with vA=me/mi/3 the Alfvén speed along the external magnetic field direction. The timestep is again set to Δt=0.1ωpi−1 (about 10 times larger than the explicit CFL for the mass ratio considered). The simulation is started without perturbing a specific wavelength. Consequently, waves with all wavelengths and with all angles to B0 supported by the simulation domain are excited. Figure 6 shows the simulation result of the magnetic energy density, which is again in excellent agreement with linear theory (the growth rate for this configuration is found to be γ=0.225 , using the same linear Vlasov code as in Ref. [55]).
Table 3: Solver performance for the KAW case with and without preconditioning for Lx×Ly=22(di)×22(di)Nx×Ny=32×32 , and Npc=200 . For all the test cases, Δt=0.1ωpi−1 . Convergence data is averaged over 10 time steps.
Table 4: Solver performance for the KAW case with and without preconditioning, with fixed ωpe/ωce=3. Other parameters are the same as in Table 3. (NC denotes “no convergence”.)