Search

Prediction and likelihood identification of high-beta disruption in JT-60U has been discussed by means of feature extraction based on sparse modeling. In disruption prediction studies using machine learning, the selection of input parameters is an essential issue. A disruption predictor has been developed by using a linear support vector machine with input parameters selected through an exhaustive search, which is one idea of sparse modeling. The investigated dataset includes not only global plasma parameters but also local parameters such as ion temperature and plasma rotation. As a result of the exhaustive search, five physical parameters, i.e., normalized beta βN, plasma elongation κ, ion temperature Ti and magnetic shear s at the q = 2 rational surface, have been extracted as key parameters of high-beta disruption. The boundary between the disruptive and the non-disruptive zones in multidimensional space has been defined as the power law expression with these key parameters. Consequently, the disruption likelihood has been quantified in terms of probability based on this boundary expression. Careful deliberation of the expression of the disruption likelihood, which is derived with machine learning, could lead to the elucidation of the underlying physics behind disruptions.

The advanced tokamak (AT) scenario with the strong reversed magnetic shear is an attractive candidate of the steady state tokamak because the strong internal transport barrier leads to the high bootstrap current fraction, resulting in the reduction of the cost of the fusion reactor. In this paper, the causes of the collapses during or just after plasma current ramp-up of the experimental campaign of the AT scenario in 2007 and 2008 are investigated and the initial results are reported. As the observations are consistent with characteristics of the stability on the resistive wall mode (RWM) and the results of MARG2D code, the RWM is suggested as the candidate of the cause of the collapses in the analyzed AT scenario.

In JT-60U, simultaneous excitation of n = 1 helical cores (HCs) and m/n = 2/1 Tearing Modes (TMs) was observed [T. Bando et al., Plasma Phys. Control. Fusion 61 115014 (2019)]. In this paper, we have investigated the excitation mechanism of n = 1 HCs with m/n = 2/1 TMs based on the experimental observations and a simple quasi-linear MHD model. In the previous study, it was reported that a "coupling" on the phase of the MHD mode is observed between n = 1 HCs and m/n = 2/1 TMs. In this study, it is found that the coupling is observed with the mode frequency from several Hz to 6 kHz. This indicates that the resistive wall and the plasma control system do not induce the coupling because the both time scales are different from the mode frequency. In addition, n = 1 HCs appear to be the non-resonant mode from the two observations: n = 1 HCs do not rotate with the plasma around the q = 1 surface in the core and the coupling is also observed even when qmin > 1. It is also observed that the electron fluctuation due to an n = 1 HC in the core region disappears with the stabilization of an m/n = 2/1 neoclassical tearing mode by electron cyclotron current drive, implying that n = 1 HCs are driven by m/n = 2/1 TMs. This perspective, n = 1 HCs are driven by m/n = 2/1 TMs, is supported by the observation that the saturated amplitude of the m/n = 1/1 component of the radial displacement in the core is smaller than that of the m/n = 2/1 component. Finally, we revisit a quasi-linear MHD model where the m/n = 1/1 HC is induced directly by the sideband of the current for the m/n = 2/1 TM, which allows to excite the non-resonant m/n = 1/1 mode. The model also describes the characteristic of the coupling, fm/n=1/1(HC) = 2fm/n=2/1(TM).

In present day devices, the external kink ideal MHD instability establishes hard operational boundaries for both the tokamak and the Reversed Field Pinch (RFP) configurations. An interesting feature of it is that its growth rate critically depends on the device passive boundary characteristics and this can slow it down to time scales accessible to modern real time feedback control systems, normally using external active coils as actuators. 3D passive structures and external fields play a key role in determining physics and control of this instability. This is in particular true for equilibria with multimodal unstable RWM spectra where modes can couple to specific 3D features of passive and active magnetic boundary. In the paper we will present recent data and simulations from RFX-mod, a medium size (R = 2 m, a = 0.459 m) device able to confine RFP and tokamak plasmas with currents up to 2 MA and 120 kA, respectively. Successful quantitative modeling of multimodal RWM control experiments performed using different actuator configurations will be presented and commented.

In order to validate some current decay models during the current quench, the plasma current decay time was studied using the experimental plasma resistance and inductance in high poloidal beta, βp, disruptions in JT-60U. The plasma resistance and inductance were evaluated from an equilibrium calculation code and the measurement value of a magnetic sensor, the electron temperature evaluated by using ECE measurement and the electron density measured by FIR interferometer. In high βp disruptions, it was found that the electron temperature at the plasma center just after current quench starts was approximately 1-4 keV under almost the same current decay time observed during the initial phase of current quench. This result indicates that the electron temperature itself plays no major role in the determination of the current decay time in the initial phase of current quench. Moreover, the current decay time predicted by a modified L/R model [Y. Shibata et al., Nucl. Fusion 50, 025015 (2010)], in which the time derivative of plasma inductance was considered, was in good agreement with the experimental current decay time, while the values obtained from the conventional L/R model were two orders of magnitude larger than the experimental results.

This paper describes the dependence of the rotation frequency of an m/n = 2/1 neoclassical tearing mode (NTM) on its magnetic island width in JT-60U (m and n are the poloidal and toroidal mode numbers, respectively). Throughout the experiments, the island width is actively controlled by changing the location of the electron cyclotron current drive (ECCD), based on the fact that aligned ECCD and misaligned ECCD make the NTM islands shrink and grow, respectively. The NTM frequency was found to gradually decrease with the increasing island width. While the frequency is 5.5 kHz for zero island width, it is 3 kHz for the full island width of about 15% of the plasma minor radius. Also, the mode frequency suddenly decreases (typically by 80%) shortly after the misaligned ECCD, during which the increment in the island width is only less than 10%. A sudden increase in the mode frequency is observed shortly after the ECCD turnoff, which is nearly the reverse process of the sudden decrease, but with a hysteresis in the island width at which downward and upward frequency changes occur. The downward frequency change occurs when the frequency decreases to about half of that without the NTM.

We have carried out experiments to clarify the stabilizing effect of a plasma rotation on the resistive wall mode (RWM) that could limit the achievable-βN in high-βN plasmas above the no-wall ideal βN-limit. On JT-60U plasma rotations are controlled using neutral beams with varying combinations of net torque input while keeping βN constant. The RWM is destabilized as the plasma rotation is being reduced. Detailed measurements of the mode structure revealed that the RWM has a global structure that rotates with the order of the resistive wall time. In these experiments, it is found that the critical toroidal rotation speed for the RWM stabilization is less than 1% of the Alfven speed. Moreover, the critical rotation does not strongly depend on βN. The results suggest that high-βN operation up to the ideal wall βN-limit could be possible by suppressing the RWM with a slow plasma rotation in fusion reactors.

Electron cyclotron range of frequency system of the JT-60U finished operation at the end of August 2008, and improvements toward JT-60SA have been started. In the last two years stable gyrotron oscillation at an output power of 1.5 MW for 1 s was demonstrated, for the first time, using the 110 GHz gyrotron. It was verified that the heat load on the cavity was at an acceptable level with continuous oscillations at 1.5 MW. The absorption power of the collector was also at an acceptable level for the longer pulse oscillation of 5 s. A power modulation technique based on anode voltage modulation was also developed in order to study the effects of modulated Electron Cyclotron Current Drive (ECCD) on Neoclassical Tearing Mode (NTM) stabilization. Modulation frequencies of up to 7 kHz were achieved at output power of 0.8 MW exceeding the previous limit of 3 kHz. Modulated ECCD experiments in synchronization with the NTM were successfully performed with a modulation frequency of around 5 kHz. Development of an accurate synchronization system played an essential role in the experiments that needed a maintained phase between the magnetic probe signal and modulated ECCD in real time. The results provide significant information for further developments that will enhance the overall performance of ECRF systems in the near future.

The roles of momentum transport on the toroidal rotation velocity (Vt) profile and the properties of transport coefficients are found by transient momentum transport analysis. The perturbation technique enables us to evaluate the momentum diffusivity (χφ) and the convection velocity (Vconv), and to calculate Vt profiles driven by external torque input by neutral beams (NBs). The measured Vt profiles with and without the external torque input are almost reproduced by χφ and Vconv in low-β (βN < 0.4) plasmas. At higher β, the local pressure gradient plays a role in determining the local value of intrinsic rotation velocity. Concerning the momentum transport, χφ increases with increasing heating power, and decreases with increasing plasma current (Ip). In H-mode plasmas, χφ is smaller than that in L-mode plasmas under similar experimental conditions. It is found that χφ, which is separated from the non-diffusive term increases with increasing heat diffusivity (χi), χφ/χi ∼ 1-3, and -Vconv increases with increasing χφ, Vconv/χφ ∼ -2.5 to -0.7 m-1 , in H-mode plasmas.

Alternating magnetic perturbations, generated by external electrodes, are applied to a neutral-beam-heated plasma in the Compact Helical System, to excite stable Alfven eigenmodes (AEs). Resonant peaks in the frequency range of the toroidal Alfven eigenmode (TAE) have been observed in a transfer function defined by the ratio between the magnetic probe signal and the alternating electrode current. The resonant frequency fo agrees well with that of the TAE gap located near the plasma edge, where the strength of energetic ion drive of AEs will be sufficiently low. The damping rates γ derived from the width of the resonant peak are fairly large, i.e., γ/(2π fo) ∼ 20%.