† Corresponding author. E-mail:
Project supported by the National Key R&D Program of China (Grant No. 2017YFE0301900), the National Natural Science Foundation of China (Grant No. 11675083), and the Fundamental Research Funds for the Central Universities of China (Grant No. DUT18ZD101).
Nonlinear evolution of multiple toroidal Alfvén eigenmodes (TAEs) driven by fast ions is self-consistently investigated by kinetic simulations in toroidal plasmas. To clearly identify the effect of nonlinear coupling on the beam ion loss, simulations over single-n modes are also carried out and compared with those over multiple-n modes, and the wave-particle resonance and particle trajectory of lost ions in phase space are analyzed in detail. It is found that in the multiple-n case, the resonance overlap occurs so that the fast ion loss level is rather higher than the sum loss level that represents the summation of loss over all single-n modes in the single-n case. Moreover, increasing fast ion beta βh can not only significantly increase the loss level in the multiple-n case but also significantly increase the loss level increment between the single-n and multiple-n cases. For example, the loss level in the multiple-n case for βh = 6.0% can even reach 13% of the beam ions and is 44% higher than the sum loss level calculated from all individual single-n modes in the single-n case. On the other hand, when the closely spaced resonance overlap occurs in the multiple-n case, the release of mode energy is increased so that the widely spaced resonances can also take place. In addition, phase space characterization is obtained in both single-n and multiple-n cases.
Alfvén instability, as a type of magnetohydrodynamic (MHD) instabilities, is very common in both space and laboratory plasmas.[1–3] In present-day fusion and future burning plasmas, Alfvén instability is easily driven to be unstable by the wave-particle interaction, including alpha particles from fusion and fast ions produced by neutral beam injection and other auxiliary heating methods. It can not only affect the plasma confinement and transport but also induce excessive fast ion loss and redistribution. On the other hand, the successful realization of the magnetically confined fusion depends on the satisfactory confinement of the fast ions. In fact, if the fast ion losses are sufficiently localized and intense, they could bring about an unplanned heat load and damage to the first wall, leading to the termination of discharges in tokamaks. Ideally, fast ions and fusion products could be well confined until their energy is transferred to the bulk plasmas. Fully understanding the wave–particle interaction between Alfvén waves and fast ions, the mechanism of fast ion loss, and then developing some methods to control fast ion loss, therefore, are of critical importance for design of plasma-facing materials and reliable predication[4] for the International Thermonuclear Experimental Reactor (ITER) and the China Fusion Engineering Test Reactor (CFETR).[5]
A well-known discrete Alfvén mode is toroidicity-induced Alfvén eigenmode (TAE). The resonant interaction between fast ions and TAE, especially multiple toroidal harmonic mode (multiple-n) TAE, is believed to represent one of the main mechanisms for fast ion loss in ITER.[6] Also, there have been extensive experimental and theoretical investigations on the fast ion losses induced by TAE, including single toroidal harmonic mode (single-n) and multiple-n TAEs, in present-day tokamaks and helical devices. It was observed that a single TAE with fluctuation amplitude above a certain threshold can induce diffusive fast ion loss, and an overlapping of TAE and Alfvén cascades of spatial structures leads to a large fast ion diffusion and loss.[7] The experimental results on the NSTX device showed that up to 40% of the fast-ion population can be expelled from confined plasmas after a TAE avalanche, thus the cumulative effect of a repetitive cycle of avalanches on the plasmas performance is dramatic.[8] In JET advanced tokamak scenario, the core-localized TAE was observed to have a significant impact both on internal redistribution and on the loss of fast ions.[9] Strong TAE activity can induce up to 70% of the fast ion loss.[10] On the DIII-D tokamak, multiple simultaneous small-amplitude Alfvén eigenmodes (AEs) can result in overlapping wave-particle resonances, leading to a critical gradient transport phenomenon.[11,12] The critical energetic particle density gradient in an AE stiff transport model has been verified against nonlinear GYRO simulations for a representative DIII-D discharge, and a simple and computationally inexpensive way to estimate the time-averaged steady-state energetic-particle (EP) density in a system with AE-induced EP transport was put forward by Bass and Waltz.[13] It should be noted that the simulation by Bass and Waltz was a local nonlinear simulation. We perform the global and self-consistent nonlinear simulations and assess the TAE-induced EP transport level self-consistently. In addition, HAGIS-LIGKA simulations by Schneller et al. showed that the global nonlinear effects are crucial for evolution of the multi-mode scenario, which not only grow amplitudes of multiple modes to higher values compared to the single mode case but also trigger strong redistribution.[14] Furthermore, Chen et al. theoretically showed that in future the presence of negative neutral beam injection induced energetic particle tail will make the system more unstable and prone to losses.[15] These results present a fairly complex pattern, only part of which is understood at present. Fast ions can be lost through various processes, such as prompt loss,[16] non-resonant loss,[16] mode-particle pumping,[17] and ripple loss.[18] Prompt losses are losses of particles born on perturbed orbits that collide with the first wall before they finish one or several poloidal transits, which may result from local wave particle resonant interaction.[16,19] The non-resonant ions can exchange energy with the wave, but the net contributed energy may be nearly zero, as the energy gained during one phase of wave is lost during the next opposite phase for these confined non-resonant ions. However, when a loss boundary exists, the case can vary. If an ion interacts with the wave for less than a full circle, the net contributed energy will not be negligible. Mode-particle pumping is a resonance phenomenon, which is an interaction of large-scale (global) MHD modes with the particle drift motion over many mode periods.[17] Accordingly, understanding the underlying physics of loss induced by single-n and multiple-n TAEs, especially in the loss mechanism and phase-space characterization, is still rare and badly needed.
In this work, taking into account the beam ion distribution function with the pitch angle scattering, we study the nonlinear evolution of multiple TAEs by means of kinetic simulations with the M3D-K code. Two kinds of cases are performed: one is the single-n TAE case, the other is the multiple-n case. The focus is on the understanding of the principle physics behind the loss induced by single-n and multiple-n TAEs and the phase-space characterization of losses. The wave-particle resonance and particle trajectory of lost ions in phase space in both the cases are obtained and compared. It is found that the fast ion loss level induced by the relatively strong resonance-overlapping multiple-n TAE is greatly enhanced. Finally, we also compare the percentages of different lost particles in different cases.
The remainder of this paper is organized as follows: Section
This work is performed by using the global nonlinear kinetic-MHD hybrid initial value code M3D-K for toroidal plasmas.[20,21] In the code, the plasmas consist of two components: the thermal component and the energetic component. The thermal component including thermal ions and electrons is described as a single fluid by resistive MHD equations solved via the finite element method. The energetic component is treated by drift-kinetic equations calculated via the δf particle-in-cell method. The energetic particle effect enters our model through coupling the energetic particle pressure tensor Ph to the momentum equation. The M3D-K code has been widely used to investigate the non-resonant kink mode,[22] nonlinear dynamics of fishbone mode,[23–25] TAE,[26,27] EPM,[28–30] and effects of energetic particles on tearing mode.[31]
In the simulation, the chosen safety factor q is q = 1.1 + ψ, where ψ is the normalized poloidal magnetic flux in code units varying from 0 at the magnetic axis to 1 at the edge. The main parameters are as follows: circular cross section R0/a = 3.11, Alfvén speed υA, Alfvén time τA = R/υA, the central beta of both thermal plasma and energetic particle βtotal, 0 = 3.43%, and the fraction of energetic particles βbeam = 1.81%.
The injection speed of beam ions is υh = 2.55υA in the simulation. The fast ion distribution is slowing down in energy, peaked in pitch angle parameter (Λ = μB0/E), where μ is the magnetic moment. The form is given by
We first study the nonlinear evolution of the TAE mode with single-n (nsingle = 1 or 2) as a reference case. Figure
Figure
In the late saturation phase of the evolution, however, the mode structure changes significantly with the dominant m/n = 2/1 component, as shown in Figs.
It is known that the fast ions can destabilize Alfvén waves by passing and/or trapped particle resonance via tapping the free energy associated with fast ion pressure nonuniformity. The TAE modes can be excited by the free energy associated with fast ion pressure gradient through wave particle resonant interaction due to the fact that they are discrete in nature and have weak continuum damping. The wave particle resonant interaction mathematically can be described as the resonant condition nωt + pωp – ω = 0, where p is an integer, ω is the mode frequency, ωt = Δϕ/Δt, ωp = 2π/Δt are both bounce-average frequency in the calculation and Δt is the time for each particle to complete one round poloidally. For passing particles, ωt is toroidal transit frequency and ωp is poloidal transit frequency. For trapped particles, ωt ≡ ωd is toroidal precession drift frequency and ωp is poloidal bounce frequency. If the right wave-particle phase is given, the linear momentum exchange may cause convective loss of fast ions. However, when the trajectories of resonant fast ions overlap with mode spatial structure, the fast ion stochastic diffusion is induced, thus a global redistribution of fast ions may occur. In Fig.
For the single-n case, the moment of a particle in phase space can be described mathematically as
To analyze the loss mechanism in the nsingle = 1 case, a typical lost particle is taken to represent the lost particle resonant with the nsingle = 1 mode. Figures
To study the effect of the beam ion injection speed on the beam ion loss, the time evolution of the beam ion loss induced by the nsingle = 1 TAE mode is given in Fig.
In this section, the results of simulations simultaneously including n (or nmultiple) = 1 – 4 modes (multiple-n case) are presented. The parameters used in the multiple-n simulation are the same as those in the nsingle = 1 simulation (see Section
In general, multiple modes can interact with each other indirectly via mutual scattering effect on their individual resonant particles. Such indirect interaction may occur in two ways. On the one hand, the modes have to share the energy delivered by the particles. On the other hand, the energy can also be transferred when a particle is scattered from one mode into the other mode instead of escaping from the wave. Such a particle recycling, as well as multiple resonances covering a large region of phase space, can also increase the total energy transferred to the fluctuating fields in the multiple-n simulation. This can be easily seen in Fig.
Moreover, the wave-particle resonance and particle trajectory in the phase space are analyzed. The orbits of beam ions are usually defined by three invariants of the motions: the energy E, the magnetic moment μ, and the toroidal canonical momentum Pϕ. When a resonant interaction between a mode and a beam ion takes place, the magnetic moment μ of the ion is conserved, whereas both the energy E and the toroidal canonical momentum Pϕ change during the process. In Fig.
Note that although a particle can be lost in the nsingle = 1 or 2 mode, or other TAE mode, it is only recorded as one lost particle in the sum loss of all individual single-n modes. That is, this sum loss is not the simple addition of all individual single-n losses. Therefore, it is shown in Fig.
Figure
The evolution of Pϕ is given in Fig.
Figure
Table
Another characteristic in common, as shown in Table
In summary, nonlinear evolution of multiple TAEs driven by fast ions is self-consistently investigated using kinetic simulations with the M3D-K code. A beam ion distribution function is considered, in which the pitch angle scattering is included. Some typical characteristics of single-n (n = 1, 2) TAE, for example, mode frequency chirping and the dynamics of resonant excitation, are obtained. Meanwhile, the increasing tendency for overall loss level with increasing beam ion injection speed is observed. This is mainly due to the fact that the large beam ion injection speed leads to a long slowing-down time and a high energy gained from the resonant beam ions. On the other hand, the larger bean ion injection speed results in a larger orbit width, leading to the higher level anomalous transport and higher level beam ion losses. In the self-consistent simulation, the mode’s growth and chirping induce island broadening and drift, which is the main reason of single-n TAE mode induced ion losses.
Even more importantly, it is interestingly found that in the multiple-n case, the fast ion loss level is rather higher than the sum loss level from all individual single-n modes in the single-n case. Moreover, increasing fast ion beta βh can not only significantly increase the loss level in the multiple-n case but also significantly increase the loss level increment between the single-n and multiple-n cases. For example, the loss level in the multiple-n case for βh = 6.0% can even reach 13% of the beam ions and is 44% higher than the sum loss level calculated from all individual single-n modes in the single-n case. The high loss level is indeed of critical importance for the fast ion confinement in high parameter plasmas like in ITER and CFETR. This significant increment may result from the strong resonance overlap. Once the closely spaced resonance overlap occurs in the multiple-n case, the release of mode energy is increased so that the widely spaced resonances can also take place. As a result, such a process can effectively enlarge the phase space region and then particles are redistributed or lost. It is also demonstrated that the trajectory of the same lost ions in the presence of multiple-n TAE is significantly different from that in the presence of single-n TAE.
On the other hand, a slight shift of the beam ion losses is towards the higher pitch angle Λ value in the single-n and multiple-n simulations. The main loss mechanism is through contributing energy to the wave, which is due to the fact that when a beam ion drives the mode, its energy decreases (ΔE < 0) and then it moves radially outward in the minor radius (ΔPϕ > 0) and eventually may be easily lost. The lost fraction of trapped particles is 2–4 times larger than that of passing particles. Particles traveling on banana orbits are more easily lost since the trapped orbits are wider. Especially for so-called ‘potato orbits’, they are more sensitive to the perturbation near their turning points.
In fact, fast ion loss is a very complicated issue, because it can be influenced by many elements such as the mode structure, the selected configuration, and the employed parameters. Usually, only one mechanism cannot account for the complicated physical process. Likewise, several kinds of fast ion loss mechanisms, such as prompt loss, mode-particle pumping and non-resonant loss mentioned in the introduction, coexist in our simulation. Prompt losses are the ones of particles born on perturbed orbits that collide with the first wall. In our simulation, a certain fraction of fast ions, passing through the plasma domain after several bounce periods, is prompt losses. The non-resonant losses are the ones of particles without resonance with the mode. Usually, some fraction of fast ions must become lost particles in the NBI-born fast ion simulation, regardless of the presence of the perturbation or not. The mode-particle pumping is actually a resonant phenomenon, where the loss results from the wave-particle resonance. This loss mechanism is also present because the TAE mode is mainly excited by the wave-particle resonance in our model. Therefore, fast ion loss is actually a very complicated important problem, and these several kinds of loss mechanisms mentioned above coexist in our simulation.
A major effort spent in the present work is aimed at understanding the general characteristic of beam ion loss in the presence of the single-n and multiple-n TAEs. It is also interesting to investigate and assess the beam ion loss in the coexistence of multiple kinds of instabilities, such as fishbone modes and TAEs under the realistic profiles observed experimentally or the predicted profiles on the CFETR. A hybrid gyro-kinetic linear simulation by Chen et al.[35] indicated that the most unstable modes in reactor, such as the ITER, lie in the range of 10 < n < 20. The physics of 10 < n < 20 Alfvén eigenmodes is significantly crucial for future reactor. Global nonlinear simulations of beam ion loss considering beam ion source and sink are also important to be studied. We leave these subjects to the future works.
[1] | |
[2] | |
[3] | |
[4] | |
[5] | |
[6] | |
[7] | |
[8] | |
[9] | |
[10] | |
[11] | |
[12] | |
[13] | |
[14] | |
[15] | |
[16] | |
[17] | |
[18] | |
[19] | |
[20] | |
[21] | |
[22] | |
[23] | |
[24] | |
[25] | |
[26] | |
[27] | |
[28] | |
[29] | |
[30] | |
[31] | |
[32] | |
[33] | |
[34] | |
[35] |