WorksPhysics and Astronomy
Drift wave turbulence and transport analysis of hot ion L-mode plasma with fishbone instability on HL-3 tokamak
Abstract
An integrated analysis of the HL-3 hot-ion L-mode (shot #11972, , ,keV) establishes the cross-scale coupling by which a single fishbone mode governs both toroidal rotation and ion-scale drift-wave turbulence. Multi-diagnostic measurements, OMFIT integrated modeling, and TGLF quasilinear simulations show that a steep fast-ion pressure gradient under 2.5,MW co-injected NBI excites the fishbone. Its outward redistribution of resonant fast ions decouples the NBI torque, and the sawtooth crash it triggers collapses the rotation from above to ,krad/s. The same fast ions that drive the fishbone also stabilize ion-scale ITG/TEM turbulence, so their redistribution removes the ITG stabilization carried by the high temperature ratio, fast-ion dilution, and shear, and erodes the internal transport barrier. Two-dimensional flux and one-dimensional eigenvalue scans confirm that the high temperature ratio and the fast-ion population together suppress particle, energy, and momentum transport. These findings close one cross-scale coupling loop spanning MHD, fast-ion transport and microturbulence, in which one mode drives both momentum balance and turbulence stability. The results offer a predictive framework for fast-ion confinement and rotation sustainment in the upcoming deuterium tritium phase, and for burning plasmas where alpha particles likewise drive MHD activity and damp turbulence.
Introduction
The hot ion mode (HIM), a regime in which the ions run hotter than the electrons (), is a leading candidate operating scenario for next-generation reactors [1, 2]. Its elevated ion temperature strengthens fusion reactivity, and the pairing of a peaked density profile with a quiescent core raises the triple product and the fusion gain [3, 4]. JET and JT-60U reached central ion temperatures of 40–45 keV together with internal transport barriers (ITBs) [3, 4, 5], while DIII-D and ASDEX Upgrade established H-modes with ITBs at lower ion temperatures [6, 7]. KSTAR and ST40 extended the regime to long-pulse and compact devices [2, 8]. A recurrent theme across these devices is what sustains the ITB: reversed magnetic shear on the large tokamaks [5, 9], versus rotation-driven shear on the compact, strongly rotating ones, with fast ions damping the turbulence in both cases [10, 2]. Integrated OMFIT modeling [11] of HL-2A high- double transport barrier discharges attributed the core turbulence suppression to fast ions and shear [10]. The same workflow later addressed core-compatible H-mode divertor detachment [12]. HL-3 recently exceeded 10 keV in ion temperature () [13] and reached a triple product of [14], a national benchmark for the planned deuterium-tritium campaign [15]. These studies still center on producing and characterizing the hot ion state. None has traced how one macroscopic fast-ion-driven MHD mode, the fishbone, controls both momentum balance and microturbulence stability. The present work fills that gap.
Ion-scale turbulence, dominated by the ion temperature gradient (ITG) mode, sets the anomalous transport in the tokamak core [16]. Fast ions damp this turbulence and improve confinement [17]. How fast ions modify turbulent transport is an active question at the interface of these two fields. Three stabilizing mechanisms act together in the HIM. The elevated temperature ratio raises the ITG excitation threshold. Fast-ion dilution of the thermal-ion drive strengthens the threshold [18]. The shear generated by the co-injected NBI torque adds a third stabilizing channel. Their combined action gives the HIM its intrinsic transport advantage and frames the integrated analysis that follows. In high-power neutral beam injection (NBI) heated plasmas, energetic particle-driven instabilities decide the confinement [19]. Among these, the fishbone mode is particularly important [20]. It is an , internal kink that trapped fast ions resonantly excite [21, 22]. Its frequency chirps downward, convex during the initial burst and concave as saturation approaches [23]. This chirp encodes the phase-space dynamics of fast ions and the onset of velocity-shear damping [24]. It sets the nonlinear saturation amplitude and its impact on fast ion confinement [19]. The fishbone mode also helps ITB formation. The , resonance requires , and the mode holds near unity. This sustains a weak (near-zero) magnetic shear configuration favorable for ion internal transport barrier (iITB) establishment [25].
Toroidal rotation couples NBI momentum input to turbulence stabilization [26, 27]. Co-injected NBI beams drive core toroidal rotation above 100 krad/s in HL-3 discharges. The rotation gradient generates a radial electric field . Its flow shearing rate tears apart turbulent eddies and suppresses ion temperature gradient (ITG) and trapped electron mode (TEM) transport [28, 29], a mechanism central to transport-barrier formation [30, 31]. On HL-3, locked tearing modes can likewise induce a strong rotation shear that suppresses turbulence and forms an ITB with keV [32]. The fast ions injected by NBI act through several channels. They deposit the toroidal momentum that drives this rotation and its shear. Through dilution and wave-particle interaction, they directly raise the drift-wave threshold and stabilize the modes [2]. The fast-ion pressure gradient can further sustain a fast-ion-induced internal transport barrier (F-ATB) [33]. The spatial distribution of the fast ions that carry the NBI torque is critical [34]. When fast ion redistribution decouples the torque source from the plasma response, rotation collapse can cause ITB loss [5]. Fast-ion transport and its interplay with magnetohydrodynamic (MHD) activity govern HIM dynamics [20] [19]. During fishbone bursts, fast ions spread radially in an avalanche-like way. This releases the steep core fast ion pressure gradient [21, 24]. The transport is nonlinear. The mode excites phase-space structures (holes and clumps) that resonantly scatter fast ions, and the resulting gradient relaxation feeds back on the mode drive [23, 24].
The cross-scale coupling between macroscopic MHD and microscopic turbulence is the central theme of this work. At the macroscopic scale, the , fishbone mode governs global fast ion transport and cyclic sawtooth crash dynamics. During ITB rebuilding, the safety factor minimum oscillates around unity [21, 35]. At the microscopic scale, drift-wave instabilities (ITG, TEM, and ETG modes) set turbulent transport levels across the plasma cross section [36, 37]. The fishbone mode bridges these two scales. It redistributes resonant fast ions radially and decouples the NBI torque source from the core plasma, depleting the stabilizing fast ion population. Macroscopic MHD dynamics couple to microturbulence stability. The pre-existing tearing mode and coexisting kink mode accompany the fishbone mode in the Mirnov spectrum and play a secondary role.
Integrated modeling platforms such as OMFiT [11] bring together equilibrium reconstruction (EFIT) [38], heating and current-drive solvers (NUBEAM [34], ONETWO [39]), and turbulent transport models (TGYRO/TGLF [36, 37, 40] with NEO [41]) to unravel these multi-scale processes. A companion study [13] runs the same machinery to characterize turbulent transport under strong shear in the HL-3 HIM, including the fishbone mode and toroidal rotation. The present work shares that machinery but follows a distinct physics path, opened up by the high fast-ion beta of shot #11972. It extends the analysis to the momentum channel and isolates the fast-ion contribution by repeating each instability and flux scan with the fast-ion population removed, to establish the cross-scale coupling through which a single fishbone mode governs both momentum balance and microturbulence stability.
In this paper, we present an integrated experimental and simulation study of the cross-scale evolution of fishbone modes and their coupling to toroidal rotation and NBI torque deposition in the HL-3 HIM. The paper proceeds as follows. Section 2 describes the experimental observations of the fishbone mode. Section 3 presents the kinetic profile reconstruction and integrated OMFiT simulation results. Section 4 presents the deposition, beta, and transport simulation results. Section 5 provides a microturbulence instability analysis using TGLF-scan simulations. Section 6 summarizes the principal findings and discusses their broader implications.
Hot ion fishbone mode experimental phenomena and results
In the companion discharge #13243, recent HL-3 campaigns have already exceeded a central ion temperature of 10 keV [13], alongside a fusion triple product of 0.65 m · keV · s, the current HL-3 benchmark. The present study turns to shot #11972, a HIM discharge with strong fishbone mode activity. The plasma current was 0.62 MA and the toroidal magnetic field was 1.67 T (Figure 1a). The stored energy reached a maximum of 0.191 MJ, and the normalized beta was about 1.3 (Figure 1b). The auxiliary heating used two co-injected neutral beams [42] with a combined power of 2.5 MW. The first beam reached 1.2 MW in three steps, and the second reached 1.3 MW in two steps (Figure 1c). The line-averaged electron density reached a maximum of (Figure 1d). At about 600 ms, when the NBI power reached 2 MW, fishbone modes began to appear [21], and a core iITB formed. After the NBI power reached its peak, small sawtooth oscillations developed, and the plasma density kept rising. Around 1080 ms, a large sawtooth crash occurred (Figure 1d), together with fluctuations of the core ITB (Figure 1e). Throughout the discharge, the electron temperature showed no significant variation, and the ratio exceeded 3.5 at its maximum.
Figure 1. Key parameter evolution for HL-3 shot #11972. (a) plasma current and toroidal magnetic field; (b) stored energy and normalized beta ; (c) NBI heating power of the two co-injected beams and emission; (d) line-averaged electron density, soft X-ray signal, and core electron density; (e) core ion temperature, core electron temperature, and ratio. The four dashed lines mark the selected simulation time slices ms, ms, ms, and ms.
At 600 ms, the rising NBI power enables the core thermal ion profile to develop a steep internal transport barrier (ITB). The co-injected NBI fast ions transfer tangential momentum to the background thermal ions. This supplies the torque that drives the plasma to toroidal rotation above 100 krad/s. The resulting rotation shear generates an intense shearing rate. This suppresses micro-scale drift-wave turbulence, mainly the ion temperature gradient (ITG) mode [28, 29]. This rotation-driven shear sustains the extreme HIM (, Figure 1e) and the high-confinement ITB at 816 ms. The same shear also disrupts the localized linear eigenmode structure of the , fishbone mode. It suppresses the growth rate through nonlinear turbulent damping and prevents unbounded growth. Between 700 ms and 1000 ms, the fishbone mode reaches a dynamically stable state. It keeps a finite saturated amplitude at 816 ms even with the large fast ion energy.
A fishbone mode also appears in the L-mode phase of the companion discharge #13243 [13], indicating that the fishbone is not an artifact of a single shot; the cross-scale coupling documented here is, however, established in detail for shot #11972 alone. The Mirnov probe array spectrogram in Figure 2 shows a fishbone mode in the 25–45 kHz frequency gap. Mirnov probe phase measurements determine its poloidal and toroidal mode numbers and . The fundamental mode at about 40 kHz further excites sideband modes at 60 kHz, 80 kHz, and 100 kHz, each with a distinct mode number. Apart from the fishbone mode and its sidebands, the spectrogram is otherwise quiet. No toroidal or elliptical Alfvén eigenmodes (TAE/EAE) or other energetic-particle-driven instabilities are present. The steep fast ion pressure gradient from the large NBI-driven fast ion population drives this mode. The 2.5 MW of NBI power deposited in the core produces a very high fast ion density and a sharply peaked fast ion pressure gradient. Starting at about 1080 ms, the fishbone mode loses stability and evolves into a large sawtooth crash. Each sawtooth crash and fishbone re-excitation goes with a radial redistribution of core fast ions and outward transport. The core thermal ion temperature and toroidal rotation crash and rebuild together.

Figure 2. SXR, Mirnov, and CXRS measurements for HL-3 shot #11972. (a) Mirnov raw signal; (b) Mirnov spectrogram showing the , fishbone mode in the 0–150 kHz frequency range, together with an kink mode near 25 kHz and an tearing mode near 6 kHz, overlaid with ion temperature (blue solid line) and toroidal rotation velocity (blue dashed line) from the CXRS system at m; (c) SXR spectrogram in the 715–725 ms window showing the convex downward chirping; (d) SXR spectrogram in the 792–802 ms window showing the concave downward chirping. The four dashed lines mark –.
To further resolve the spatial structure of the fishbone-driven fast ion transport, we performed soft X-ray (SXR) tomographic inversion at three time windows. The first is 707.97 ms (Figure 3a), the convex downward chirping phase during the early nonlinear burst. The second is 810.74 ms (Figure 3b), the concave chirping phase approaching nonlinear saturation. The third is 1074 ms (Figure 3c), just before the sawtooth crash at 1080 ms. The SXR emissivity is dominated by bremsstrahlung and recombination radiation from the thermal plasma, and depends on the electron temperature, electron density, and impurity density [35]. The reconstruction therefore tracks the core plasma response to the fishbone-driven fast-ion redistribution.
Figure 3. Soft X-ray tomographic inversion of the core plasma redistribution accompanying the fishbone-driven fast-ion transport at three characteristic time windows: (a) 707.97 ms, (b) 810.74 ms, and (c) 1074 ms.
The inversion results show a clear outward radial movement of the core emissivity as the fishbone mode evolves. At 707.97 ms, the emissivity concentrates strongly in the core region, consistent with the sharply peaked core fast-ion pressure gradient that drives the initial fishbone burst. By 810.74 ms, the emissivity peak broadens and shifts outward, reaching its maximum intensity, consistent with the phase-space hole and clump structure resonantly ejecting fast ions as the fishbone mode approaches its nonlinear saturation amplitude. At 1074 ms, just before the sawtooth crash, the core emissivity drops to its minimum among the three windows, indicating that a large part of the resonant fast ions has moved radially outward and the core fast-ion pressure gradient has largely relaxed, in agreement with the transport modeling of § 4. This progression, from core peaking through outward expansion to core depletion, gives tomographic evidence, complementary to the modeling, that the fishbone mode drives a continuous radial redistribution of fast ions through its nonlinear saturation cycle, ending in the sawtooth crash.
Figure 2 compares the fishbone mode chirping during two different phases. Panel (c) shows the SXR spectrogram in the 715–725 ms window, the early nonlinear saturation stage, where the fishbone mode displays a convex downward chirp. In the phase-space hole–clump picture [23, 24], this chirp reflects a hole that experiences an intensifying drag as the system enters the strongly nonlinear initial-burst regime, and the high fast-ion energy speeds its outward radial drift toward the resonance boundary. The accompanying avalanche-like ejection of fast ions from the core relaxes the steep fast-ion pressure gradient .
Panel (d) shows the SXR spectrogram in the 792–802 ms window, the post-saturation phase, where the chirp lengthens and turns to a concave downward (decelerating) pattern. Within the same hole–clump picture [23, 24], the hole now meets strong damping as it moves through the localized velocity shear, slowing the outward escape of fast ions. The concave chirp thus indicates that the fishbone mode and the fast-ion population have reached a local kinetic equilibrium, with the ambient shear barrier holding back the ejection rate and briefly accumulating fast ions at a localized radial-electric-field barrier. This state precedes the macroscopic sawtooth crash at 1080 ms.
The Mirnov coil spectrogram (Figure 2b) shows two modes that accompany the fishbone mode (Figure 2a). One is an kink mode near 25 kHz that grows together with the fishbone mode from 620 ms. The other is an tearing mode near 6 kHz in the low-frequency band. It appears as early as 400 ms and lasts through the discharge. The kink mode frequency band overlaps with the lower-frequency edge of the fishbone mode. This points to nonlinear wave-wave coupling between the two modes. The tearing mode, with its much lower frequency, is a long-lived electromagnetic perturbation that exists before the fishbone mode. The charge exchange recombination spectroscopy (CXRS) diagnostic gives the time evolution of ion temperature and toroidal rotation velocity from the plasma core ( to 2.4 m) over 300–1500 ms, as shown in Figure 4. During the early NBI phase (340–600 ms), the co-injected fast ion fraction is too small to give much tangential momentum to the background plasma. The toroidal rotation stays low. The core electron density and ion temperature then rise quickly as the beam power reaches its full 2.5 MW. The ion temperature profile (Figure 4a) develops an iITB structure, and the toroidal rotation (Figure 4b) forms a steep shear layer in the core. The NBI power mainly lands in the core, and the edge region ( m) does not develop an ion temperature edge transport barrier. A velocity shear layer still forms there.
Figure 4. CXRS measurements of (a) ion temperature and (b) toroidal rotation velocity as functions of major radius and time for HL-3 shot #11972. The four dashed lines mark the selected simulation time slices –.
In the post-sawtooth phase after 1080 ms, the ion temperature profile keeps a large gradient in the radial range m. The toroidal rotation shear layer at this location becomes flat. Both profiles keep their gradients in . The gradients of both quantities drop clearly near the ITB foot (). This shows that the ITB contracts inward and the parameter values at its foot rise. After 1300 ms, as the NBI power declines, the plasma disrupts and the ITB collapses.
Kinetic profile reconstruction and integrated simulation results
The detailed physical workflow and setup of each modeling module follow the description in Ref. [13]. For each of the four time slices, we jointly fitted the instantaneous experimental data from multiple diagnostics to construct time-resolved kinetic profiles. This avoids using profiles from a single steady state snapshot. The four slices mark the fishbone growth phase (), the nonlinear saturation (), the sawtooth ITB decay (), and the sawtooth ITB recovery (). The sawtooth crash develops faster than the time resolution of the CXRS and Thomson scattering (TS) diagnostics. The profiles reconstructed at and carry larger uncertainty than those at and . The fishbone-driven ITB collapse in this period is additionally constrained by the experimental measurements (SXR tomography and Mirnov spectrograms) in Section 2. Equilibrium, sources, and transport are advanced in sequence by EFIT [38], ONETWO [39], and TGYRO [40]. Inside TGYRO the turbulent fluxes are computed with the Trapped Gyro-Landau-Fluid model (TGLF [36]) using the SAT2 saturation rule [37] on a radial grid of 21 points over , and the neoclassical coefficients with NEO [43, 44]. Once the transport iteration converges, TGLF-scan [36] maps the linear growth rate against the normalized poloidal wavenumber and sweeps selected profile quantities one dimension at a time. Table 1 lists the discharge parameters and compares the OMFIT-simulated plasma parameters against the experimental measurements, including the confinement factor [45] and the Greenwald density fraction [46]. Figure 5 shows the fitted kinetic profiles at the four time slices.

Figure 5. Fitted kinetic profiles at the four time slices, reconstructed from multi-diagnostic synthesis and used as initial inputs to the OMFIT integrated workflow. The rows correspond to electron density , electron temperature , ion temperature , and ion rotation frequency (top to bottom), and the columns to the four time slices , , , and (left to right). The colored lines are the fits and the markers are the measured points, with error bars giving the experimental uncertainty; the fits stay inside them. The color and marker identify the time slice: (blue, circles), (red, diamonds), (yellow, asterisks), and (black, upward triangles). The radial coordinate is .
| Physical parameter | (Exp./Sim.) | (Exp./Sim.) | (Exp./Sim.) | (Exp./Sim.) |
| (m) | 1.78/1.78 | 1.78/1.78 | 1.78/1.78 | 1.78/1.78 |
| (m) | 0.589/0.589 | 0.586/0.586 | 0.589/0.589 | 0.585/0.585 |
| (MA) | 0.609/0.609 | 0.616/0.616 | 0.621/0.621 | 0.621/0.621 |
| (T) | 1.67/1.67 | 1.67/1.67 | 1.67/1.67 | 1.66/1.66 |
| (MW) | 1.9/1.9 | 2.5/2.5 | 2.5/2.5 | 2.5/2.5 |
| ( m) | 1.07/1.11 | 1.37/1.37 | 1.32/1.32 | 1.40/1.41 |
| (MJ) | 0.128/0.130 | 0.184/0.205 | 0.184/0.202 | 0.191/0.188 |
| 0.76/0.84 | 1.17/1.27 | 1.21/1.29 | 1.27/1.24 | |
| –/0.62 | –/0.99 | –/0.85 | –/0.94 | |
| –/1.08 | –/1.04 | –/0.96 | –/0.94 | |
| –/4.82 | –/5.33 | –/4.76 | –/4.51 | |
| (%) | –/8 | –/17 | –/13 | –/14 |
| (%) | –/20 | –/24 | –/23 | –/24 |
Table 1. Discharge parameters and OMFIT-simulated plasma parameters for HL-3 shot #11972 at the four time slices. Each entry gives the experimental and simulated values as Exp./Sim.; the discharge parameters (, , , , ) are common to the measurement and the simulation, and dashes mark quantities without a directly measured experimental value (the bootstrap fraction has no experimental counterpart).
The radial electric field comes from the radial force balance, whose three terms are the diamagnetic, toroidal-rotation, and poloidal-rotation pieces [13, 26]. Figure 6 shows that the toroidal rotation term dominates at every time slice, exceeding the diamagnetic term by up to a factor of five. The corresponding shearing rate is given by the Waltz-Miller formula [36]
Figure 6. Pressure and radial electric field profiles at the four time slices from the OMFIT workflow. (a)–(d) Total pressure (black) and fast ion pressure (green) on the left axis, with their gradients (yellow) and (blue) on the right axis. (e)–(h) Split of the radial electric field into diamagnetic (yellow), toroidal rotation (red), and total (blue) parts.
In the toroidal-rotation-dominated limit this becomes . The rotation profile and its radial derivative fix the shearing rate. Once this shearing rate exceeds the drift-wave growth rate it quenches the turbulence. It contributes to the ITG stabilization analyzed in § 5.
Panels (e)–(h) of Figure 6 break the radial electric field into the terms that together build the shear layer. The diamagnetic piece never exceeds 20 kV/m in magnitude. The toroidal-rotation piece carries essentially all of the field. At (fishbone nonlinear saturation, strongest ITB) it is five times the diamagnetic piece. The peak (most negative) values rank (−95 kV/m) > (−70 kV/m) > (−55 kV/m) > (−40 kV/m), following the rotation history in Figure 4b. Rotation supplies most of the radial electric field and, through it, the shear that suppresses turbulence. The peak of lies at –0.4 at both and , coinciding with the iITB (Figure 4a). Inside the ITB the elevated temperature ratio is the main stabilizing contribution, and the rotation-driven shear grows to comparable importance at the barrier edge, where the temperature ratio falls.
The sharply peaked fast-ion pressure profile built by the core NBI deposition ( at , Figure 8d) supplies the free energy that drives the , fishbone instability [21]. The mode growth and nonlinear saturation set up a cross-scale coupling loop in which the fishbone drives fast-ion transport that relaxes . Figure 7 shows the evolution of the current-density and radial deposition profiles. The HL-3 ITB develops on a core with negative magnetic shear, and low-density operation permits it to form as the two NBI beams step up to 1.2 MW and then 1.3 MW during the main heating phase. At every time slice the current density peaks away from the axis, within . The Ohmic component provides most of the current: 0.44 MA, 0.39 MA, 0.40 MA, and 0.43 MA, that is 73%, 63%, 65%, and 70% of the total. As NBI continues, the bootstrap contribution climbs (0.05 MA, 0.11 MA, 0.08 MA, 0.09 MA), reaching 8%, 17%, 13%, and 14% in Figure 7(a)–(Figure 7d). The beam-driven current falls off slowly.
Figure 7. Current, safety factor, and shear profiles at the four time slices from the OMFIT workflow. (a)–(d) Current density: total (blue), Ohmic (green), bootstrap (red), and beam-driven (cyan). (e)–(h) Safety factor (yellow) and magnetic shear (magenta). The dashed lines mark , 1.5, and 2, with symbols at each radial crossing of , , , and .
Figure 8. Radial profiles from ONETWO/NUBEAM and TGLF-scan at the four time slices. (a) NBI power density to electrons (dashed) and ions (solid, left axis), and the ion torque density after TGYRO iteration (dash-dotted, right axis). (b) Fast ion density (solid, left axis) and fast-ion-to-electron density ratio (dashed, right axis, %). (c) Electron beta (dashed) and thermal ion beta (solid). (d) Fast ion beta (dashed) and total beta (solid). Blue, red, yellow, and black mark , , , and .
Figure 7(e)–(Figure 7h) plots the simulated safety factor and magnetic shear profiles at the four time slices. The minimum of is at ms, 1.04 at ms, 0.96 at ms, and 0.94 at ms. The profile follows the same trend, remaining weakly sheared over at ms and at ms (Figure 7(e)–(Figure 7f)). Both match the sawtooth oscillations [35] recorded in the soft X-ray channel (Figure 1d) and the current deposition profiles of Figure 7(a)–(Figure 7d). The high core ion temperature and its long-lived ITB are maintained jointly by low density at high current, weak shear, shearing rate and strong NBI momentum input.
Across the four time slices, the radial location of moves steadily inward, from at ms (early ITB formation) to at ms (fishbone nonlinear saturation). It then drops to at ms (post-sawtooth crash) and stays there at ms (ITB recovery). The current density profile changes from an off-axis distribution during early ITB formation to an on-axis peaked distribution. During its active phase the fishbone mode holds slightly above unity through its , internal kink resonance ( at , 1.04 at ). This condition sustains the off-axis current deposition profile and the weak (near-zero) magnetic shear (). It reduces the ITG drive and gives a favorable magnetic topology for ITB formation. The fishbone mode helps ITB establishment (holding and weak shear) and harms nonlinear saturation (redistributing fast ions radially, as analyzed in § 4). As drops below unity, the fishbone mode excites the sawtooth crash. During the cyclic process of sawtooth crashes and ITB rebuilding with fishbone mode reappearance, oscillates around .
In JT-60U’s strong reversed-shear discharges (, , ), the barrier is trapped at the surface [9, 47]. HL-3 operates near zero shear, which suppresses the ITG growth rate only mildly. Turbulence suppression must come from the shear, and whether it overlaps the ITB (Figure 5) determines the outcome. Larger devices with stronger heating shift the ITB-formation physics away from shear reversal and toward rotation-driven flow shear.
NBI deposition, beta, and transport simulation results
Figure 8a shows the NBI power deposition and fast ion profiles computed by ONETWO with the NUBEAM code [34]. The NBI heating in this discharge couples mainly and strongly to the ion channel. In the core region (), the ion heating power density exceeds the electron heating power density by a factor of four to six at all four time slices. This ion-dominated heating follows from the collisional slowing-down of the injected beam ions. A fast ion loses energy through Coulomb collisions with both the electrons and the thermal ions. The share deposited on each species is set by the critical energy [48], the energy at which the electron and ion drag rates are equal. Above the beam heats mainly the electrons, and below it heats mainly the ions. The critical energy is , where is the mass number of the beam ions, is the electron temperature in keV, and and are the charge and mass numbers of the background thermal ions. For a deuterium beam () injected into a deuterium plasma (, ) at keV, this gives keV. The NBI beam energy (45 keV) lies below . The beam ions transfer most of their energy to thermal ions via collisional drag rather than to electrons. This ion-biased power partition is the main cause of the extreme ratio that marks the HIM. It also ensures that the NBI torque mainly enters the ion rotation channel.
At ms (early ITB formation), the still-low core density gives a fairly high NBI deposition. Performance peaks at ms (fishbone nonlinear saturation, strongest ITB).
Figure 8b shows the fast ion density and its ratio to the electron density. These reflect the cross-scale coupling between fast ion transport and the fishbone mode. At ms (early ITB formation), the core density stays low. The NBI fast ions deposit readily in the core, and the fast ion-to-electron density ratio is centrally peaked. At ms (fishbone nonlinear saturation), the fishbone mode drives resonant fast ions outward through wave-particle interaction. Trapped fast ions move from the core to . These fast ions carry the NBI torque. This redistribution moves the toroidal momentum to outer radii and decouples the torque source from the core plasma. This shifts the peak outward and reduces the on-axis value. At ms (post-sawtooth crash), the outer-region fast ion population stays similar to that at , and the core value recovers. This improves collisional energy transfer to the thermal ions and raises the NBI heating efficiency to its maximum. The fishbone-driven fast ion expulsion decouples the torque source from the core. The sawtooth crash that the fishbone triggers dominates the drop in toroidal rotation from above 100 krad/s to 60 krad/s at 1100 ms. Panels (c) and (d) of Figure 8 show how the core beta is apportioned among the species, as computed by TGLF-scan. The leading core-beta component switches between fast and thermal ions across the four times. Fast-ion beta wins at ms (0.81%) and ms (1.14%). Thermal-ion beta wins at ms and ms, reaching 2.24% at . At the fast-ion beta (0.98%) is close to the electron beta (1.08%). The total beta peaks at 4.26% at , drops to 3.03% at ms as the post-sawtooth ion temperature relaxes and falls, then recovers to 3.69% at ms.
Figure 9 shows the turbulent and neoclassical coefficients and Gyro-Bohm-normalized fluxes from TGYRO. Inside the fast-ion suppression, the ITB, and the strong shear jointly push the ion turbulent transport down to the neoclassical level at –. At it remains far above the neoclassical level. Electron turbulent transport stays about ten times above neoclassical, and the electron scale channel is not fully quenched by the shear. In the outer region beyond the core ITB (), turbulent transport still dominates. The electron and ion turbulent transport coefficients and Gyro-Bohm-normalized energy fluxes there are of comparable magnitude across all four time slices, with the normalized ion heat flux () staying stiffness-locked within 10–100. The ion neoclassical transport coefficient and energy flux exceed the corresponding electron values by more than an order of magnitude. The neoclassical transport is carried almost entirely by the ions. The total transport in this outer region is carried jointly by the ion and electron turbulent channels.
Figure 9. Turbulent and neoclassical transport at the four time slices. (a),(b) Ion and electron thermal diffusion coefficients from TGLF (turbulent) and NEO (neoclassical). (c),(d) Corresponding Gyro-Bohm energy fluxes. Filled markers are turbulent (tur) and open markers neoclassical (neo). Blue, red, yellow, and black mark the four time slices –.
The ion heating power density (Figure 8a) rises from to as the total NBI power grows from 1.9 to 2.5 MW at fixed beam energy, and by a further 30% from to . The fishbone-driven outward migration of the fast ions empties the core phase space, and the continuously injected high-energy beam ions deposit their power more readily in the core. The fast ion density (Figure 8b) directly records the fishbone-driven transport. At the core () drops below its value and the outer region rises, and at continued injection raises the density across by a factor of 1.2. The density peak shifts from at and to at , when the core ion temperature and ITB strength peak, marking the outward redistribution that relaxes the core fast-ion pressure gradient.
The ion torque density (Figure 8a, right axis) varies most strongly. Toroidal momentum conservation forces the diffusive momentum flux to vanish on axis, concentrating the effective torque there. This torque-source relocation, not any loss of total NBI momentum, decouples the NBI torque from the core. The fishbone-driven outward shift of the fast-ion density peak (Figure 8b) moves the point where the torque enters the thermal channel, and the ensuing sawtooth crash collapses the core rotation from above 100 to 60 krad/s. The NBI torque drives the core rotation and shear that sustain the ITB. The same fast-ion population supplies the steep that excites the , fishbone. The fishbone redistributes the fast ions, relaxes the shear, and reduces core confinement. MHD excitation, fast-ion redistribution, and microscale transport bifurcation form one cross-scale mechanism.
Micro-turbulence instability analysis
The linear spectra span two wavenumber ranges: the ion-scale branch (, the ITG and TEM modes) and the electron-scale branch (, the ETG mode). The stabilization mechanisms central to this work (the high temperature ratio, fast-ion dilution, and shear) act on the ion-scale branch, which is therefore the focus here; the electron-scale ETG mode is identified in the spectra but is not quenched by these ion-scale mechanisms and lies outside the cross-scale coupling of Section 4. Figure 10 maps, for each of the four time slices, the linear instability spectrum versus normalized poloidal wavenumber at several radii, as computed with TGLF-scan. The spectra show that in the intermediate radial region , the electron temperature gradient (ETG) mode dominates at all four time slices. It propagates in the electron diamagnetic drift direction. At ms, during the early nonlinear growth phase of the fishbone, the trapped electron mode (TEM) appears in the outer region . At ms, the fishbone nonlinear saturation phase, the TEM weakens. The instability turns to the ion temperature gradient (ITG) mode in the ion diamagnetic drift direction. At ms (post-sawtooth crash), the redistribution of core fast ions improves the core microturbulence stability.
Figure 10. Drift wave turbulence instability spectra of the most unstable eigenmode from TGLF-scan simulations as functions of the normalized poloidal wavenumber and radial position . The columns correspond to the four time slices , , , and ; the first row (a1)–(a4) shows the cases with fast ions and the second row (b1)–(b4) the cases without fast ions. Blue indicates the ion diamagnetic drift direction, red the electron diamagnetic drift direction, and darker coloring higher instability growth rates.
The dominance of the ion direction ITG over the electron direction TEM in follows from the asymmetry of the drive sources there. This region lies outside the ion temperature ITB, where the ion temperature gradient drops sharply (the row of Figure 5) and weakens the direct ITG drive. The electron density and temperature gradients remain substantial (the and rows of Figure 5) and keep driving the electron direction TEM. Without the fast ions, the TEM dominates. Once the fast ions are added, their gradual outward diffusion at and (Figure 3) stabilizes the TEM to a degree that varies with the local fast-ion population. The TEM is stabilized and the ion direction ITG becomes the dominant unstable branch. The fast ions stabilize the TEM preferentially. The TEM is driven by trapped electrons [16], and the fast ions occupy the same trapped phase space. Their precessional dynamics couple directly to the trapped-electron resonance and damp the electron direction mode [17, 33]. The ITG, driven by the thermal-ion temperature gradient, is affected only through the weaker dilution channel.
To quantify the fast ion effect on core microturbulence, we compared the TGLF-scan maximum linear growth rate and the shearing rate in the core () at the four time slices, with and without fast ions (Figure 11). The fast ion population consistently lowers across all four slices, confirming that fast ions stabilize ion scale drift-wave turbulence rather than acting as an inert background, in agreement with the improved core stability at after fast ion redistribution. Part of this stabilization is dilution: the fast ions replace thermal ions in the ITG drive, weakening the instability for a given ion pressure gradient. This dilution channel complements the temperature-ratio and shear stabilization discussed below. SXR tomography (Figure 3) shows that the fishbone mode continuously moves resonant fast ions outward through its nonlinear saturation cycle. The TGLF spectra (Figure 10) confirm the microturbulence response. At ms the fast ions moved to leave the TEM weakened and the ion direction ITG appearing in the outer region (). At ms the post-sawtooth re-accumulation of core fast ions restores the stabilization.

Figure 11. TGLF-scan comparison of the maximum linear growth rate and the flow shearing rate in the core region () at four time slices: (a) , (b) , (c) , and (d) . In each panel, diamonds denote the case without fast ions (blue) and circles the case with fast ions (red); filled symbols mark and open symbols mark .
The effect of the hot ion temperature ratio on turbulent transport is isolated with a TGLF two-dimensional scan at ms and , sweeping the normalized ion temperature gradient and the ratio over 0 to 3 times their experimental values (Figure 12). The particle, energy, and momentum fluxes of all three species (electron, thermal deuterium, and impurity carbon) fall steadily as increases, most steeply between 1 and 2 times its experimental value. The high of the H1M is inherently favorable to high confinement, with no anomalous high-flux band in the scanned range. In the low-ratio region ( below its experimental value), Figure 12 show a direct
Figure 12. TGLF two-dimensional flux scan of the Gyro-Bohm-normalized particle (), energy (), and momentum () fluxes (columns) for the electron, thermal deuterium, and impurity carbon ion species (rows) at ms and . The horizontal axis is the ion-to-electron temperature ratio , and the vertical axis is the normalized ion temperature gradient , both scanned over the range 0 to 3 times their experimental values. Dashed lines mark the experimental operating point.
relation between and . A larger is needed to reach a given flux as rises, for all three transport channels alike. The flux falls as rises. This raises the ITG excitation threshold. A larger temperature ratio demands a steeper to drive the same flux. Once the gradient exceeds the raised threshold, the growth rate and flux rise linearly with . This mapping between the temperature ratio and the required gradient gives a model-based constraint for future HIM operation.
At , the scan also locates the ITG excitation threshold. When both and lie below their experimental values, the deuterium particle flux does not rise with , indicating that sits near the threshold. This threshold rises linearly with . The high temperature ratio moves the operating point toward the strongly-suppressed regime. A complementary one-dimensional TGLF eigenvalue scan at ms and (Figure 13) holds the ion-to-electron temperature ratio fixed at 0.3, 1.0, and 3.0 times its initial value (, 1.49, and 4.77, one per column) and sweeps over 0 to 3 times its initial value. At the lowest ratio () and low gradient (), the low-wavenumber drift waves () propagate in the electron diamagnetic direction and are identified as trapped electron modes (TEM). As rises they turn into ion direction ion temperature gradient (ITG) modes, and over the scanned range the growth rate rises with . At the experimental ratio () the raised ITG threshold breaks this trend. The low-wavenumber growth rate no longer rises monotonically with but first falls and then recovers, and even at 3 times the initial gradient it stays below its value at 0.3 times the initial gradient. The growth rate in this column is everywhere below that of the 0.3 times column. At the highest ratio () every low-wavenumber mode () propagates in the ion direction, and ITG alone is unstable. The growth rate again falls and then recovers as rises, and even at 3 times the initial gradient it stays below its value at 0.3 times the initial gradient. The ion direction ITG dominates at the highest ratio, where the ITG threshold is largest. This appears at first to contradict the threshold formula. A raised threshold should make excitation harder, and it suppresses the electron direction TEM more strongly than the ion direction ITG. ITG is harder to excite in absolute terms, but it wins the competition against the TEM and emerges as the sole unstable mode. The non-monotonic dependence reflects a balance of the ion diamagnetic drive , which grows linearly with , against finite-Larmor-radius (FLR) stabilization, which grows quadratically as and weakens the drive through the gyroaveraging factor . Ion Landau damping strengthens as the parallel phase velocity approaches the ion thermal velocity , with the Maxwellian resonant rate
Figure 13. TGLF linear eigenvalue spectra with fast ions at ms and : growth rate (top row) and real frequency (bottom row), both in units of , as functions of the normalized poloidal wavenumber , for three fixed ion-to-electron temperature ratios , 1.0, and 3.0 times the initial value, corresponding to , 1.49, and 4.77 (columns). Within each column, the rainbow-colored curves correspond to a one-dimensional scan of the normalized ion temperature gradient over 0 to 3 times its initial value, with the horizontal colorbar beneath each column indicating the scanned value. The vertical dashed line marks , and the horizontal dashed line in the bottom row separates the electron direction () from the ion direction ().
Both stabilization terms grow faster than the drive with wavenumber and overtake it at high [16, 49], shifting the most unstable mode to shorter wavelength while it stays in the ion direction.
The flux decline with in Figure 12 and Figure 14 is a direct consequence of the kinetic ITG threshold. In the near-flat-density and strong-ballooning limits, the toroidal ITG branch obeys [13, 16, 49, 50]
Figure 14. TGLF two-dimensional flux scan identical to Figure 12 but with the fast-ion population removed.
which rises linearly with . The large of the HIM lifts the critical gradient and curbs ion scale transport. An explicit fast-ion species raises the threshold further through an extra stabilizing term [18]. The fishbone erodes this term by moving the resonant fast ions outward. The threshold falls and the ion direction ITG reappears in the fast-ion depleted outer region (Figure 10), linking the macroscopic fishbone to the microscopic stability threshold.
To separate the fast-ion contribution from the temperature-ratio effect, we repeated the two-dimensional scan with the fast ions removed (Figure 14). Relative to Figure 12, the fluxes of all three species rise markedly at high once the fast ions are absent. Fast ions suppress electron, thermal-ion, and impurity transport alike, most strongly at high . This corroborates the instability-level result of Figure 11, where fast ions reduce . The local correspondence between fast ion density and the dominant instability closes the cross-scale coupling of Section 4. Fast ions drive the macroscopic fishbone via and suppress microscopic ITG via linear wave-particle interaction. Their outward redistribution is a self-regulating feedback that depletes the stabilizer, lowers the ITG threshold, and raises turbulent transport. The core confinement degradation from to ( from 9.8 to 6.4 keV and from 0.99 to 0.85) proceeds through this core ITG re-excitation. Once the rotation collapse relaxes the shear and fast-ion depletion lowers the threshold, the core ITG reappears and erodes the ITB. The near-neoclassical ion transport at
(Figure 9) is confined to the deep core (), where fast ions re-accumulate after the crash. The re-excited ITG appears in the fast-ion-depleted outer region () and drives the degradation.
Repeating the eigenvalue scan with the fast ions removed (Figure 15) leaves the qualitative picture unchanged and raises the growth rate at every temperature ratio. The electron direction TEM, the ion direction ITG, and their non-monotonic dependence all persist. The overall growth rate is systematically higher than in the fast-ion-inclusive case of Figure 13. The suppression is largest at high and high , where the fast-ion-inclusive scan is lower by the widest margin. The elevated temperature ratio and the fast-ion population suppress drift-wave turbulence together, neither alone achieving the depth of suppression they deliver jointly. This cooperative effect underlies the favorable confinement of the HIM and points to the HIM with its high fast-ion fraction as a promising route to high-parameter plasmas, in which the fishbone-driven fast-ion population suppresses drift-wave anomalous transport and improves confinement.
Figure 15. TGLF linear eigenvalue spectra identical to Figure 13 but with the fast-ion population removed, at ms and .
Figure 16 sweeps the shear rate and the normalized ion temperature gradient at for ms. As the shear rate rises, the particle, energy, and momentum fluxes are strongly suppressed, and above (twice the experimental value) they fall to a very low level. The radial transport is effectively controlled.
Figure 16. TGLF two-dimensional scan of the Gyro-Bohm-normalized particle (), energy (), and momentum () fluxes (columns) for the electron, thermal deuterium, and impurity carbon ion species (rows) at ms and . The horizontal axis is the shear rate (in units of ), and the vertical axis is the normalized ion temperature gradient , both scanned over the range 0 to 3 times their experimental values. Dashed lines mark the experimental operating point.
The three stabilizing channels combine near the threshold, where the maximum linear growth rate rises linearly with the temperature gradient [49, 50]. The eigenvalue scan of Figure 13 confirms this rise, reaching at the upper end of the scanned range. The shear quenches the mode once [30] (the boundary between filled and open symbols in Figure 11), so a stronger shear demands a steeper gradient to keep the mode above the quench threshold. The kinetic threshold of Eq. (3) rises with , fast-ion dilution raises it further [18], and the shear quench supplies a third, rate-based condition. Raising therefore suppresses the fluxes until, above , they collapse (Figure 16). The same reasoning shows why the ITB is self-sustaining: the steep is itself an ITG drive, but the NBI-torque-driven shear, the fast-ion dilution, and the elevated temperature ratio together hold the operating point below threshold.
The radial balance of the three channels follows from the measured profiles. Inside the ITB the elevated dominates the ITG suppression, and the fast-ion dilution strengthens it further. The shear is smaller in the deep core and becomes the main stabilizing contribution where drops steeply at the barrier edge, keeping the operating point sub-critical inside the barrier. The sawtooth-affected slices and reverse this: the ITB collapse flattens the core , and the fast-ion transport flattens the shear profile, so the ion direction ITG reappears over a broad range. A single fishbone mode thus erodes, channel by channel, the stabilization that sustains the ion temperature ITB.
Summary and discussion
Combining OMFIT integrated modeling with TGLF quasilinear gyro-Landau-fluid simulations, this study traces how the fishbone mode couples toroidal rotation, NBI torque, and microturbulence across scales in the HL-3 HIM (shot #11972). The HIM (, ) forms under 2.5 MW co-injected NBI. Its ion-biased deposition builds a steep fast-ion pressure gradient () that excites an , fishbone mode (25–45 kHz). This mode drives outward redistribution of resonant fast ions (Figure 3), decoupling the NBI torque from the core. The sawtooth crash it triggers collapses the toroidal rotation from above 100 to 60 krad/s. The same steep that drives the fishbone also directly stabilizes ion scale ITG turbulence. The redistribution erodes the ITB (Figure 17).
Figure 17. Cross-scale coupling mechanism spanning macroscopic MHD, fast-ion transport, and microturbulence. NBI drives two channels. The momentum channel (left) couples the NBI torque deposition to toroidal rotation (above 100 krad/s) and shear, which suppresses ITG/TEM turbulence and sustains the ITB. The fast-ion/MHD channel (right) excites an , fishbone mode (25–45 kHz) through the steep () and drives outward fast-ion redistribution. Three stabilizing mechanisms sustain the ion ITB: the high temperature ratio , the fast-ion dilution, and the shear. The fishbone erodes all three at once. Torque decoupling (rotation collapse) relaxes the shear, and fast-ion depletion removes the dilution stabilizer. Fast ions drive MHD and stabilize ITG turbulence, and a high temperature ratio is intrinsically favorable to confinement.
The SXR tomographic inversion (Figure 3) and the TGLF instability spectra (Figure 10) give two sides of a single self-consistent mechanism. The fishbone-driven fast-ion transport seen in experiment and the core ITG stabilization computed by TGLF match, confirming that fast ions directly damp the ITG growth rate. The results thereby complete a single cross-scale coupling loop that spans macroscopic MHD, fast-ion transport, and microturbulence (Figure 17). In this loop, NBI torque sustains the ITB through rotation and shear, and the fishbone-driven redistribution erodes it. Prior work examined these channels separately. The present study shows that a single fishbone mode is the common driver of both momentum balance and microturbulence stability. Three stabilizing mechanisms sustain the ion temperature ITB: the high temperature ratio, which contributes most inside the barrier; the fast-ion dilution; and the shear, small in the deep core and dominant at the barrier edge. The fishbone erodes all three at once. Its outward fast-ion transport removes the dilution stabilizer, and the sawtooth-induced rotation collapse relaxes the shear and flattens the core . The barrier protection is lost over a broad region.
The temperature-ratio mechanism is confirmed at both the transport and instability levels. The two-dimensional flux scan (Figure 12) shows the fluxes of all three species falling steadily as increases. The one-dimensional eigenvalue scan (Figure 13) gives the microscopic origin. The ion direction ITG is damped and its growth rate falls as rises. This match between the macroscopic flux decline and the microscopic growth-rate reduction shows that the flux suppression comes from first-principles instability dynamics rather than a fitting model. Repeating the eigenvalue scan without fast ions (Figure 15) leaves the mode structure unchanged and raises the growth rate at every temperature ratio, most at high and high . The high temperature ratio and the fast-ion population damp drift-wave turbulence in combination; either mechanism on its own falls short of the suppression they jointly deliver. This pairing gives the HIM its intrinsic confinement advantage and makes the hot-ion, high-fast-ion regime a promising route to high-parameter plasmas, in which the fishbone’s fast-ion population suppresses drift-wave anomalous transport and improves confinement. The two competing effects of fast ions matter directly for burning plasmas. There, alpha particles will contribute to both MHD drive and microturbulence stabilization. The fast-ion density profile in reactor-relevant scenarios must balance the two. On HL-3, the weak magnetic shear makes shear the main ITB sustainment mechanism. The identified coupling loop gives a
predictive framework for fast-ion confinement and rotation sustainment in the upcoming deuterium-tritium phase.
Toward a reactor, a higher fusion triple product needs a high core temperature ratio together with a steep inside the ITB. The two-dimensional scan (Figure 12) shows that the fluxes fall steadily as and rise together, crossing no high-transport region. A strong ion edge transport barrier (ETB) offers another route, raising the pedestal-top ion temperature to control with the core fixed. These pathways highlight both the gains and the built-in challenges of reaching high-performance hot ion H-mode parameters. ITB formation still depends on the weak magnetic shear set up early in the discharge together with strong rotational shear. Suppressing macroscopic MHD instabilities requires , a condition the fishbone mode helps to maintain.
Acknowledgments
The authors gratefully acknowledge the support of the OMFIT integrated simulation platform provided by General Atomics (GA), USA, and express our gratitude to the HL-3 team. This work was supported by the Nuclear Technology Research and Development Program: HJSYF2024(02).
Funding
This work is supported by the National Key R&D Program of China (Grant Nos 2025YFE03000200 and 2025YFE03000204), the National Natural Science Foundation of China (Grants Nos 12125502 and 12535013), the Nature Science Foundation of Sichuan Province (Grant No 2025ZNSFSC0833) and the Innovation Program of Southwestern Institute of Physics (Grant No 202301XWCX001).
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Supplementary data
None.
References
- [1]Kamada Y, Fujita T, Ishida S et al. 2002 Fusion Sci. Technol. 42 185–254
- [2]Han H, Park S J, Sung C et al. 2022 Nature 609 269–275
- [3]Gormezano C, Baranov Y F, Challis C D et al. 1998 Phys. Rev. Lett. 80 5544–5547
- [4]Kishimoto H, Ishida S, Kikuchi M and Ninomiya H 2005 Nucl. Fusion 45 986
- [5]De Vries P, Joffrin E, Brix M et al. 2009 Nucl. Fusion 49 075007
- [6]Lazarus E A, Navratil G A, Greenfield C M et al. 2006 Phys. Rev. Lett. 96 115002
- [7]Gruber O, Wolfr R C, Dux R et al. 1999 Phys. Rev. Lett. 83 1787–1790
- [8]McNamara S A M, Asunta O, Bland J et al. 2023 Nucl. Fusion 63 054002
- [9]Fujita T, Kamada Y, Ishida S et al. 1999 Nucl. Fusion 39 1627–1636
- [10]Li Z J, Chen W, Sun A P et al. 2024 Acta Phys. Sin. 73 065202
- [11]Meneghini O, Smith S, Lao L et al. 2015 Nucl. Fusion 55 083008
- [12]Shu Y K, Wang Z H, Xu X L et al. 2025 Acta Phys. Sin. 74 095201
- [13]Shu Y K, Chen W, Li Z J, Hao B L, Wang Z et al. 2026 Nucl. Fusion 109290
- [14]Zhong W L, Chen W and Ji X Q 2026 The Innovation 7 101167
- [15]Chen W, Zhong W L, Xue L, Ji X Q, Zhang L et al. 2026 Sci. Bulletin 71 1849–1851
- [16]Horton W 1999 Rev. Mod. Phys. 71 735–778
- [17]Na Y S, Hahm T S, Diamond P H et al. 2025 Nature Reviews Physics 7 190–202
- [18]Jung M K, Hahm T S, Na Y S and Yoon E 2026 Physics of Plasmas 33 Fasoli A, Gormezano C, Berkh H et al. 2007 Nucl. Fusion 47 S264
- [20]Heidbrink W, Austin M, Fisher R et al. 2011 Plasma Physics and Controlled Fusion 53 085028
- [21]Chen L, White R and Rosenbluth M 1984 Phys. Rev. Lett. 52 1122
- [22]Coppi B and Porcelli F 1986 Phys. Rev. Lett. 57 2272
- [23]Berk H, Breizman B, Fitzpatrick J et al. 1996 Phys. Plasmas 3 1827–1838
- [24]Fu G, Park W, Strauss H et al. 2006 Phys. Plasmas 13 052517
- [25]Ge W, Wang Z X, Wang F et al. 2023 Nucl. Fusion 63 016007
- [26]Burrell K H 1999 Phys. Plasmas 6 4418–4435
- [27]Terry P 2000 Rev. Mod. Phys. 72 109
- [28]Biglari H, Diamond P and Terry P 1990 Phys. Fluids B 2 1–4
- [29]Garbet X, Mourgues F and Samain A 1990 Plasma Physics and Controlled Fusion 32 917–931
- [30]Hahm T S and Burrell K H 1995 Phys. Plasmas 2 1648–1651
- [31]Itoh K and Itoh S I 1996 Plasma Phys. Control. Fusion 38 1–49
- [32]Yan Q, Li D, Ji X, Chen W, Zhong W et al. 2026 Nucl. Fusion 66 126047
- [33]Di Siena A, Bilato R, Görler T et al. 2021 Phys. Rev. Lett. 127 025002
- [34]Pankin A, McCune D, Andre R et al. 2004 Comput. Phys. Commun. 159 157–184
- [35]von Goeler S, Stodiek W and Sauthoff N 1974 Phys. Rev. Lett. 33 1201
- [36]Waltz R E, Staebler G M, Dorland W et al. 1997 Phys. Plasmas 4 2482–2496
- [37]Staebler G, Kinsey J and Waltz R 2005 Phys. Plasmas 12 102508
- [38]Lao L, St John H, Stambaugh R et al. 1985 Nucl. Fusion 25 1421
- [39]St John H, Taylor T, Lin-Liu Y and Turnbull A 1994 Transport simulation of negative magnetic shear discharges Tech. rep. General Atomics, San Diego, CA (United States)
- [40]Pan C, Staebler G, Lao L et al. 2017 Nucl. Fusion 57 036018
- [41]Belli E and Candy J 2008 Plasma Phys. Control. Fusion 50 095010
- [42]Wagner F, Becker G, Behringer K et al. 1982 Phys. Rev. Lett. 49 1408
- [43]Marin M, Camenen Y, Bourdelle C et al. 2025 Nucl. Fusion 65 036015
- [44]Dudkovskaia A, Connor J, Dickinson D et al. 2023 Nucl. Fusion 63 126040
- [45]ITER Physics Expert Group on Confinement and Transport and Confinement Modelling and Database and ITER Physics Basis Editors 1999 Nucl. Fusion 39 2175–2249
- [46]Greenwald M 2002 Plasma Phys. Control. Fusion 44 R27–R53
- [47]Sakamoto Y, Kamada Y, Ide S et al. 2001 Nucl. Fusion 41 865
- [48]Stix T H 1972 Plasma Phys. 14 367–384
- [49]Romanelli F 1989 Phys. Fluids B 1 1018–1025
- [50]Hahm T S and Tang W M 1989 Physics of Fluids B 1 1185–1192













