Characterization of off-axis fishbones

Plasma Physics and Controlled Fusion Characterization of off-axis fishbones To cite this article: W W Heidbrink et al 2011 Plasma Phys. Control. Fusion 53 085028 View the article online for updates and enhancements. Related content ‘Beam-emission spectroscopy’ diagnostics also measure edge fast-ion light - Beam-ion confinement for different injection geometries - Central flattening of the fast-ion profile in reversed-shear DIII-D discharges - Recent citations Mechanisms of energetic-particle transport in magnetically confined plasmas W. W. Heidbrink and R. B. White - Origin of ion cyclotron emission at the proton cyclotron frequency from the core of deuterium plasmas in the ASDEXUpgrade tokamak B Chapman et al - Comparing theory and simulation of ion cyclotron emission from energetic ion populations with spherical shell and ringbeam distributions in velocity-space B Chapman et al - This content was downloaded from IP address 87.218.223.151 on 18/08/2020 at 11:42 IOP PUBLISHING PLASMA PHYSICS AND CONTROLLED FUSION Plasma Phys. Control. Fusion 53 (2011) 085028 (23pp) doi:10.1088/0741-3335/53/8/085028 Characterization of off-axis fishbones W W Heidbrink1,MEAustin2, R K Fisher3, M Garc´ ıa-Mu˜ noz4, G Matsunaga5, G R McKee6, R A Moyer7, C M Muscatello1, M Okabayashi8,DCPace 9, K Shinohara5, W M Solomon8, E J Strait3, M A Van Zeeland3and Y B Zhu1 1Department of Physics and Astronomy, University of California, Irvine, CA, USA 2Institute of Fusion Studies, University of Texas at Austin, Austin, TX, USA 3General Atomics, San Diego, CA, USA 4Max-Planck-Institut f¨ ur Plasmaphysik, Garching, Germany 5Japan Atomic Energy Agency, Naka City, Ibaraki, Japan 6Department of Engineering Physics, University of Wisconsin at Madison, Madison, WI, USA 7Center for Energy Research, University of California, San Diego, La Jolla, CA, USA 8Princeton Plasma Physics Laboratory, Princeton, NJ, USA 9Oak Ridge Institute for Science and Education, Oak Ridge, TN, USA Received 28 February 2011, in final form 13 June 2011 Published 7 July 2011 Online at stacks.iop.org/PPCF/53/085028 Abstract Repetitive bursting instabilities with strong frequency chirping occur in highbeta, beam-heated plasmas with safety factor q>1 in the DIII-D tokamak. Although the mode structures differ, in many ways, the off-axis fishbones are similar to the q=1 fishbones first observed on the Poloidal Divertor Experiment (PDX). The modes are driven by energetic trapped ions at the fastion precession frequency. During a burst, the frequency changes most rapidly as the mode reaches its maximum amplitude. Larger amplitude bursts have larger growth rates and frequency chirps. Unlike PDX fishbones, the decay phase is highly variable and is usually shorter than the growth phase. Also, the waveform is highly distorted by higher harmonics during the latter portion of a burst. The radial mode structure alters its shape during the burst. Like PDX fishbones, the modes expel trapped ions in a ‘beacon’ with a definite phase relationship relative to the mode. Seven types of loss detectors measure the beacon. The losses scale linearly with mode amplitude. The neutron rate changes most rapidly at maximum mode amplitude but, depending on the loss diagnostic, the losses often peak a few cycles later. The non-ambipolar fast-ion losses cause a sudden change in toroidal rotation frequency across the entire plasma. In addition to an overall drop, the neutron signal oscillates in response to the wave. Unlike the beacon of lost particles, which maintains a fixed phase relative to the mode, the phase of the neutron oscillations steadily increases throughout the burst, with the greatest phase slippage occurring in the highly nonlinear phase near maximum mode amplitude. (Some figures in this article are in colour only in the electronic version) 0741-3335/11/085028+23$33.00 © 2011 IOP Publishing Ltd Printed in the UK & the USA 1 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 1. Introduction The fishbone instability was first observed [1] during deuterium near-perpendicular neutral beam injection into the Poloidal Divertor Experiment (PDX). The mode had an (m, n) =(1,1) structure in the plasma, as expected for an internal kink. (Here mand ndenote the poloidal and toroidal mode numbers, respectively.) The fishbones occurred in periodic bursts. During each burst, the mode frequency chirped down in frequency from (typically) 20 to 12 kHz. The neutron emission dropped at each burst, with the rate of change peaking near the time of peak mode amplitude [2]. Neutral particle analyzers (NPAs) detected a ‘beacon’ of expelled trapped fast ions, a burst of signal with a fixed phase relative to the magnetic fluctuation on each mode cycle [3,4]. The transport of fast ions was explained as convective radial transport caused by resonance between the fast-ion precession motion and the kink mode [5]. The instability was identified as a new branch of the internal kink associated with the large energetic ion population [6]. Subsequently, q=1 fishbone instabilities were observed on many other devices. (A comprehensive review of both the PDX fishbones and subsequent measurements on other tokamaks, including DIII-D, appears in [7].) Both an MHD branch [8] and an energeticparticle branch are theoretically predicted [9] and observed [7]. A comprehensive picture emerged: at modest fast-ion density, the trapped ions help stabilize the fluid branch but, at large fast-ion density, they destabilize the energetic-particle branch [10]. Fishbone-like modes in plasmas with central safety factor well above unity were first observed on JET [11]. More recently, fishbone-like modes were studied in JT-60U [12,13] and DIII-D [14,15] plasmas with similar qprofiles as the JET plasmas but higher beta. While q=1 fishbones have eigenfunctions whose maximum is inside the q=1 surface, the mode structure for the off-axis fishbones peaks near the q=2 surface [11,15]. Some authors have suggested that the off-axis fishbones are internal kinks [11], while others suggest that they are related to an external kink [15]. In either case, it is likely that the off-axis fishbone is an energetic-particle branch of the fluid internal or external kink mode. A theoretical study showed that a modest fast-ion population can stabilize an internal double kink mode but a large population destabilizes a new fishbone branch [16], just as it does for the 1/1 internal kink. On the other hand, the observation that the instabilities occur in plasmas that are susceptible to resistive wall modes suggests that the off-axis fishbones may be an energetic-particle branch of the MHD external kink [15]. (Resistive wall modes occur in plasmas that would be unstable to external kinks in the absence of a conducting wall [17].) A recent paper [15] shows that the JT-60U and DIII-D modes have the same phenomenology and discusses their relationship to the resistive wall mode. The goal of this paper is to provide a detailed description of the off-axis fishbone in DIII-D. The data are compared and contrasted to PDX fishbones throughout the paper. The nonlinear evolution of the off-axis fishbone bursts displays novel features not previously reported for any fast-ion driven instability. The paper begins with a description of the plasma conditions and diagnostics (section 2). Next, the mode is characterized (section 3.1) and the loss (section 3.2) and confined (section 3.3) fast-ion data are described. Speculation about the reasons for the differences between PDX fishbones and off-axis fishbones appears in section 4, together with the conclusion. 2. Apparatus All of the data in this paper are from DIII-D. Most of the discharges described in this paper were discussed in the DIII-D portion of the paper by Okabayashi et al [15]. A database of 513 bursts 2 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 R (m) -1.0 -0.5 0.0 0.5 1.0 z (m) BILD BES #141089 2315 ms q=2 1.7 2.5 3.0 3.5 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 150o BILD 210o 330o30o BES BES (b) (a) FILD FILD NPA NPA FIDA FIDA FIDA NEUTRONS NEUTRONS ICE ICE ISAT ISAT MIRNOV MIRNOV CER CER ECE ECE Figure 1. (a) Elevation of the DIII-D vacuum vessel, showing the locations of the various measurements. (The BES, NPA and FIDA signals depend on the edge neutral density profile, so the indicated positions are estimates.) The flux surfaces for a typical equilibrium are labeled by their qvalue; q0=1.5 on axis and q95 =4.4. (b) Plan view. The lines represent the center of the beamlines. The toroidal field (plasma current) is in the clockwise (counter-clockwise) direction. was compiled that contains all of the n=1 off-axis fishbone bursts observed during two days of operation. The plasma conditions are similar for nearly all of the bursts: plasma current Ip=0.95–1.14 MA; toroidal field BT=1.72; minor radius a=0.60 m; normalized beta βN=βT/(Ip/aBT)=2.1–2.7; line-average plasma density ¯ne=(3.4–4.2)×1019 m−3. The beam power was the actuator for feedback control of βN, so it varies between 5.7 and 14.9 MW for the bursts in the database, with a typical average power of 10 MW . The beams inject both in the co-Ipdirection (⩽15 MW ) and in the counter-Ipdirection (⩽5 MW ); both neartangential (tangency radius Rtan =1.15 m) and near-perpendicular (Rtan =0.76 m) angles are employed. The average injection energies for the co-tangential, co-perpendicular, countertangential, and counter-perpendicular beams are 66–80 keV, 60–73 keV, 80–81 keV and 60– 75 keV, respectively. Typically, the beam beta is ∼1/6 of the total beta. In many discharges, ∼2.7 MW of electron cyclotron current drive power is injected to avoid neoclassical tearing modes [14]. The plasma is deuterium, the neutral beams inject deuterium atoms, and the primary impurity is carbon from the graphite walls (Zeff ≃2). All of the discharges have the shape shown in figure 1, a lower single null divertor configuration. The ∇Bdrift is downward toward the divertor and all of the plasmas are in H-mode. The equilibrium is based on EFIT [18] solutions of the Grad–Shafranov equation that use motional Stark effect [19] measurements of the internal magnetic field. The central safety factor is q0=1.15–1.78, the magnetic shear is relatively weak near the magnetic axis, crosses the q=2 surface near a major radius of R=2.1 m and rises rapidly to a value of q95 =4.3–4.7 at the edge. (q95 is the safety factor at the surface that encloses 95% of the poloidal flux.) Typical central electron and ion temperatures are Te=5 and Ti=6 keV. 3 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al Figure 1shows the locations of the principal diagnostics. Toroidal and poloidal arrays of Mirnov coils [20] are the primary fluctuation diagnostic. In this paper, all of the Mirnov coil traces are from a coil that is located at a toroidal angle of 307◦. At this magnetic field and density, the electron cyclotron emission (ECE) diagnostic [21] measures optically thick, third harmonic ECE emission for the spatial positions shown in the figure. The charge-exchange recombination (CER) diagnostic [22] measures the temperature, toroidal rotation, and density of carbon ions; on some discharges, the CCD camera acquired data in 0.5 ms time bins to detect rapid changes in toroidal rotation. The primary diagnostic for the confined fast ions is a neutron scintillator that measures 2.5 MeV neutrons produced in deuterium–deuterium fusion reactions. Two complementary scintillators with identical electronics are mounted beside a vacuum flange just outside the vacuum vessel: a plastic scintillator and a ZnS(6Li) scintillator. The plastic scintillator can resolve oscillations in neutron flux with frequencies 10 kHz, while the ZnS scintillator has an effective bandwidth of ∼3 kHz [23]. Consistent with the expected instrumental response, the fluctuations in neutron flux discussed in section 3.3 are much weaker on the ZnS signal than on the plastic signal. All neutron data in this paper are from the plastic scintillator. Seven different diagnostics detect escaping fast ions. The beam-ion loss detector (BILD) is a Faraday cup that is mounted on the edge of a vacuum port 12 cm below the midplane [24]; in this paper, the signals are from the foil that detects co-circulating orbits. The fast-ion loss detector (FILD) is a scintillator-based diagnostic that resolves the gyroradius and pitch of the escaping fast ions in an image that is recorded on a CCD camera [25]. For higher bandwidth measurements, a photomultiplier detects light from a portion of the scintillator. A reciprocating Langmuir probe with robust graphite tips [26] measures fluctuations in ion saturation current (ISAT); at the time of the measurements, the probe was in the scrape-off region 6.1 cm from the last-closed flux surface and 5.6 cm from the outer wall. A solid-state NPA operated in current mode measures fluctuations in charge-exchange reactions when fast ions escape to the high neutral density region at the plasma edge. The fast-ion D-alpha (FIDA) diagnostic [27,28] measures Doppler-shifted light emitted by fast ions after they undergo a charge-exchange reaction. For the rapidly changing background conditions associated with these fishbone bursts, the temporal resolution of the spectroscopic FIDA diagnostics proved inadequate to detect changes in core fast-ion density; however, bursts of edge FIDA emission are observed by the filter-based f-FIDA diagnostic [28] on both ‘active’ sightlines that view an injected beam and ‘passive’ sightlines that do not intersect an injected beam. The beam-emission spectroscopy (BES) diagnostic [29] suffers from contamination by edge FIDA light for these fishbone bursts [30], complicating use of the data for internal eigenfunction measurements of density fluctuations. However, for one discharge where the neutral beams viewed by the BES diagnostic were off, the signals are entirely from passive FIDA light produced when fast ions are expelled to the high neutral density region at the plasma edge [30]. The seventh edge loss diagnostic is indirect. When fast ions are expelled into the scrape-off region, they usually destabilize ion cyclotron emission (ICE) [7]. These waves are thought to be magnetoacoustic waves excited by the anisotropic ‘bump-on-tail’ distribution function caused by the losses [31]. For these experiments, the signal from a high-bandwidth toroidal coil [32] was filtered by two bandpass filters, rectified and digitized. The ‘lower pass’ filter includes the first deuterium cyclotron harmonic at the plasma edge, while the ‘higher pass’ filter spans the second and third harmonics. All fishbone bursts that meet certain criteria are included in the database. Figure 2shows examples of included and excluded bursts. Like the first burst in figure 2(a), all of the included bursts are n=1 modes that rotate in the direction of the plasma current. Roughly 3% of the 4 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al -150 -100 -50 0 50 100 150 2313 2314 2315 2316 2317 0 1 2 3 4 5 6 TIME (ms) #141092 -200 -100 0 100 200 1950 1952 1954 1956 1958 0 1 2 3 4 5 6 n=4 n=1 TIME (ms) MIRNOV (T/s) BILD (V) #141086 (a) (c) (b) (d) Figure 2. Sample Mirnov coil (a), (b) and BILD (c), (d) signals for bursts that are excluded from the database. (a), (c) The n=1 burst at 1951 ms in (a) is included but the n=4 burst at 1957 ms is excluded. (b), (d) Bursts with appreciable beating due to a steady tearing mode (in this case, an n=3 mode) are also excluded. Note that the BILD signals saturate at 5 V. observed bursts have higher toroidal mode numbers (usually n=3orn=4). Like the n=1 modes, these higher frequency modes chirp down in frequency and cause losses that have a definite phase relative to the wave; however, these higher frequency modes are not discussed further here. There are also many examples of fishbone bursts that occur in plasmas with steady tearing modes (figure 2(b)). Phenomenologically, these bursts appear identical to the retained bursts but, because of difficulties in accurately analyzing the Mirnov coil signal in the presence of strong beating, they are excluded from our database. 3. Data 3.1. Mode properties The off-axis fishbone bursts typically last for 2–3 ms and occur about every 20 ms. Figure 3 shows a typical Mirnov coil signal. Initially, the signal is nearly sinusoidal during the growth phase. The amplitude grows approximately exponentially. At maximum amplitude, the waveform becomes strongly distorted from a sine wave. Data from the toroidal array show that this distortion is associated with the appearance of higher nharmonics: in other words, the entire structure becomes distorted. The mode period visibly lengthens as the burst passes through its maximum amplitude. Figure 3also illustrates analysis of the signal for inclusion in the database. First, the maxima and minima are identified (figure 3(a)). Then the rectified amplitude of the maxima and minima are plotted (figure 3(b)). Note that, during the rise phase, the amplitudes of the maxima and minima have nearly the same value but, during the decay phase, the maxima are larger than the minima. The amplitude data are fit to exponentials to obtain the growth and decay rates. In the third step, the time difference between the maxima and minima is graphed (figure 3(c)). The initial mode frequency is obtained from a fit to the first several cycles, while the final mode frequency is obtained from the last two cycles. The frequency chirping rate is obtained from a linear fit to the change in period near peak mode amplitude. The mode distortion is derived from the difference in period between half-cycles, T1/2/¯ T1/2. 5 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al -400 -200 0 200 400 0.00 0.10 0.20 1837.5 1838.0 1838.5 1839.0 1839.5 1840.0 0 100 200 300 400 (a) Mirnov (T/s) Rise TIME (ms) #141076 (b) Amplitude (T/s) Fall Initial Final Chirp (c) Half-period (ms) Distortion 0 5 10 15 20 25 30 Half-c y cle # Figure 3. (a) Mirnov coil signal for a typical fishbone burst. The symbols indicate the maxima and minima used for calculation of the half-period. (b) Amplitude of the maxima and minima for the same burst, together with exponential fits to the growth and decay phases. (c) Half-period in milliseconds for each half-cycle throughout the burst. The initial frequency is derived from a horizontal fit at low amplitude. The frequency chirping rate is obtained from a linear fit near maximum amplitude. (The dashed vertical line indicates the time of maximum amplitude.) The final frequency is an average over the last two cycles. Linear fits to the long half-periods and the short half-periods are used to measure the waveform distortion. (This quantity is zero for a sine wave.) This definition of distortion gives similar results to alternative definitions [15] based on the amplitude or the deviation from a sine wave. The rate of change of the distortion is also recorded in the database. The observed rise time increases with the maximum amplitude of the mode (figure 4(a)). The correlation coefficient ris 0.72 for these data. (Figure 4(a) shows the correlation with d˜ B/dt; the correlation with the peak amplitude ˜ Bis slightly higher (r=0.76).) PDX fishbones had a similar rise time and dependence on mode amplitude [33]. In contrast, as shown in figure 4(b), the decay rate is uncorrelated with mode amplitude (r=0.05). Unlike PDX fishbones, the decay rate is generally faster than the growth rate. An example of the wide variation in behavior during the decay phase for a pair of otherwise similar bursts is shown in figure 5. This example illustrates the following general features: (1) the evolution of the bursts is quite reproducible during the growth phase; (2) frequency chirping and mode distortion always occur near maximum amplitude; (3) the decay phase is highly variable. Sometimes the burst decays in a few cycles, while other times it persists for many. Often the mode distortion remains large but sometimes nearly sinusoidal successor oscillations persist at small amplitude following the primary, highly distorted, crash. The magnitude of the frequency chirping scales with mode amplitude (figure 4(c)) (r=0.77). The rate of change of the chirping rate also scales with mode amplitude (r=0.55). The frequency chirping also increased with mode amplitude for PDX fishbones [2]. Within a given burst, the chirping rate is largest in the period just after maximum amplitude. Similarly, 6 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 0 2 4 6 8 0 5 10 15 0 50 100 150 200 250 PDX Peak Mirnov Amplitude (T/s) 0 1 2 3 4 5 RISE γ / ω (%)DROP γ / ω (%) (a) (b) (c) FREQ. CHIRP (kHz) Figure 4. (a) Growth rate, (b) decay rate and (c) change in frequency fi−ffinal versus peak mode amplitude for all of the bursts in the database. The growth and decay rates are normalized to the mode frequency in the plasma frame, ω=2π(fi−frot), where frot is from the R=212 cm CER channel. The lines are linear fits to the data. The error bars are the values observed for typical cases in PDX [33]. -1 0 1 2 -300 -200 -100 0 100 200 300 400 RELATIVE TIME (ms) MIRNOV (T/s) #141092 1801 ms 1981 ms Figure 5. Example of two similar bursts with strikingly different behavior in the decay phase. 7 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 50 100 150 200 250 0.0 0.2 0.4 0.6 0.8 1.0 Distortion Peak Mirnov Amplitude (T/s) Figure 6. Distortion of the waveform T1/2/¯ T1/2near maximum mode amplitude versus peak Mirnov amplitude for all of the bursts in the database. Illustrative examples of waveforms with distortions of 0.32 and 0.76 are included. PDX fishbones [33] and high frequency chirping instabilities in DIII-D [34] chirp the fastest near maximum mode amplitude. All of the fishbones exhibit mode distortion (figure 6). Most bursts exhibit very strong distortion but even the most weakly distorted waveforms deviate significantly from a sine wave. Analysis of the poloidal Mirnov coil data [15] shows that the sinusoidal portion of the waveform has an m=3 structure but the distorted portion has an m≃2 structure. Strong mode distortion was not observed for PDX fishbones [33]. Figure 7shows the radial mode structure during the early phase as measured by the ECE diagnostic. The δTefluctuations are largest near the q=2 radius (R≃2.1 m). There is a ∼60◦phase variation between the weak fluctuations observed near the magnetic axis and the stronger fluctuations observed near the q=2 surface. In the theory of energetic-particle instabilities, it is important to determine if the eigenfunction changes shape or remains constant during a burst. (If the eigenfunction changes shape, proper treatment of nonlinear effects requires a ‘non-perturbative’ model.) To investigate the temporal constancy of the mode structure, we compare the relative amplitude of the various ECE channels to the Mirnov coil as the burst evolves. For the example in figure 8(a), the ratio of two ECE signals remains nearly constant during the growth phase but deviates near maximum amplitude. During the growth phase, the ECE signals grow more slowly than the integrated Mirnov signal; after the maximum the ECE signals decay more slowly than δB. Figure 8(b) shows the average behavior for a set of similar bursts. Although the uncertainties associated with incoherent Tefluctuations and with conditional averaging of an ensemble of bursts are substantial, the data appear to follow the same trend as the example in figure 8(a): δTegradually decreases relative to δB during the growth phase, then rises in the decay phase. Due to the large variability of behavior in the decay phase, the standard deviations are large but examination of individual bursts shows that there are substantial differences in the evolution of the various channels during this phase. We conclude that the mode structure varies, as expected for a strongly driven non-perturbative energetic-particle mode. The initial mode frequency is determined by the fast-ion precession frequency. Figure 9 shows calculations by an orbit-following Monte Carlo (OFMC) code [35] of the precession frequencies fpre,0of fast ions deposited by the four different angles of beam injection in a typical discharge. These calculations neglect the radial electric field but analytical theory [36] and separate calculations show that inclusion of the electric field increases the actual precession 8 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 0 50 100 150 200 250 Peak Mirnov Amplitude (T/s) 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0123456 Frequency Chirp (kHz) 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Peak beam-ion loss rate (ms-1)Peak beam-ion loss rate (ms-1) (a) (b) Figure 14. Maximum neutron loss rate versus (a) maximum Mirnov amplitude and (b) change in mode frequency, fi−ffinal for all of the bursts in the database with constant beam power during the burst. Linear (solid lines) and quadratic (dashed line) fits to the data are included. last-closed flux surface and 5.6 cm from the vessel wall. This location minimizes contributions to the ion saturation current from thermal ions, while still allowing fast ions to strike the collector tip. High-bandwidth FILD measurements became available at the end of the experimental campaign. For this measurement, the photomultiplier imaged a portion of the scintillator centered on a lower pitch angle (∼50◦) than the primary fishbone loss peak at ∼70◦. Nevertheless, evidence of a coherent beacon is also observed by this diagnostic (figure 15(c)). Two NPA channels made high-bandwidth measurements throughout the campaign. The NPA sightlines originate from a lower port, rise up to intersect the counter beams near the midplane, then strike the inner wall above the midplane; toroidally, the sightlines are normal to the plasma. Whether the viewed counter beams are off or on, fishbone modes cause beaconlike bursts on the NPA detectors (figure 15(c)). Presumably, these bursts are caused by fast ions that are expelled to the high neutral density region at large major radius. On PDX, the signal on the vertically viewing NPA that viewed the outer edge of the plasma was two orders of magnitude larger than the NPA that viewed the inner edge of the plasma [3,40]. There also are some indications of reductions in active signal caused by the fishbone burst. Normally, the signal for the BES diagnostic is dominated by beam emission from the active injected beam but, in the case of large losses of fast ions, FIDA light from the edge plasma can also be appreciable [30]. In DIII-D, the BES filters accept Doppler-shifted D-alpha light produced by fast ions on escaping trapped orbits. To avoid the complication of concurrent beam emission, the data in figure 15(d) are from a discharge where the active beams were both off. Bursts of light are observed. In the case of the BES diagnostic, two bursts per cycle are often observed because the BES sightline intercepts the plasma edge at two different toroidal 15 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 0.40.0-0.4 120-1-2 4 2 -4 -2 0 #141089 2319 ms -0.8 -4 -2 0 2 4 6 -0.5 0.0 0.5 1.0 -0.4 -0.2 0.0 0.2 0.4 #142124 1618 ms -400 -200 0 200 400 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 -400 -200 0 200 400 600 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 0.0 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 RELATIVE TIME (ms) RELATIVE TIME (ms) RELATIVE TIME (ms) RELATIVE TIME (ms) RELATIVE TIME (ms) RELATIVE TIME (ms) T/s T/s Volts Volts (a) (b) (d) (f) (e) (c) #141069 1497 ms #142125 1614 ms #141069 1339 ms #142126 1662 ms BILD ISAT BES FILD NPA Lower Band Reference Active Higher Band ICE FIDA Figure 15. Loss-detector data during a fishbone burst from (a) the BILD foil, (b) the Langmuir probe Isat tip, (c) the FILD scintillator and SSNPA detector, (d) a central BES channel on a discharge without any beam emission, (e) active and passive f-FIDA channels and (f) ICE passband signals that include the fundamental cyclotron resonance (lower band) or the first few harmonics (higher band). The diamonds represent peaks used to measure the phase relative to the Mirnov signal. locations. The sightline is angled slightly downward. The first peak corresponds to a poloidal position below the midplane, while the second peak corresponds to a poloidal position above the midplane. Generally, the first peak (lower poloidal position—in the direction of the ∇B drift) appears earlier in the evolution of the mode and is larger in amplitude than the second peak, although often the two peaks are comparable during the decay phase. Like the other edge diagnostics, both peaks have a fixed phase relative to the mode. 16 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 0 50 100 150 200 250 300 0 50 100 150 200 250 PHASE LAG (º) ISAT ICE SSNPA FILD BILD BES BES FIDA FIDA TOROIDAL ANGLE (º) Figure 16. Measured phase lag relative to the Mirnov signal versus toroidal angle of the diagnostic. The vertical error bar represents the standard deviation over several cycles and the horizontal error bar is an estimate of the uncertainty in angle. The solid lines show the expected variation for an n=1 mode. Bursts of edge FIDA light are also observed by a dedicated FIDA diagnostic (figure 15(e)). The sightlines for this diagnostic look down at a canted toroidal angle from an upper port toward the vessel floor [28]. One channel views one toroidal direction to intersect a heating beam (for active FIDA measurements), while the other channel views in the opposite toroidal direction to provide a reference signal. The bandpass filter for these measurements accepts Doppler-shifted light from co-going ions in the case of the active view and from counter-going ions for the reference view. Both sightlines intersect the edge region near the top and bottom of the vessel. Both channels observe beacon-like bursts of signal that presumably are caused by expulsion of fast ions by the fishbones. Because the ∇Bdrift is downward, the bursts are probably caused by fast ions that charge exchange with neutrals near the bottom of the vessel. Beacon-like bursts of signal are also measured by the ICE diagnostic (figure 15(f)). Both the higher band and lower band filtered signals measure bursts with a definite phase relative to the mode, although the variation of the phase is somewhat larger than for the other loss diagnostics. The higher band signal correlates more reliably with the Mirnov signal than the lower band signal. For the higher band, like the maximum neutron loss rate, the signal peaks at the same time as the maximum Mirnov amplitude (within statistical error). The magnitude of the ICE signal summed over the burst also correlates strongly with the maximum mode amplitude (r=0.52). Given that the ICE radiation is produced through a two-stage process (the fishbones expel fast ions, which then excite magnetoacoustic waves), it is rather surprising that the ICE signal provides such a reliable measurement of the beacon and of the timing and magnitude of the losses. The phase of all of the beacon measurements is summarized in figure 16. For diagnostics that measure the same quantity, such as the two FIDA measurements and the two BES peaks, the phase difference between the measurements agrees with the toroidal separation of the detection points, as expected for an n=1 mode. For the larger dataset, although each diagnostic measures a well-defined phase relative to the mode, there is no clear pattern for the entire set of measurements. Upon reflection, this is not surprising. The fishbone mode rotates at the relatively slow precession frequency. On the other hand, the phase of any particular diagnostic signal depends on the final leg of the orbit of the fast ions that produce that signal; this final leg occurs at the much higher bounce frequency. On this final leg, a loss orbit such as the one shown in figure 12(b) moves over 90◦toroidally. Since the different diagnostics 17 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al -0.01 0 0.01 0.02 0.03 0.04 0.05 125 130 135 140 110 115 120 125 130 76 78 80 82 84 86 88 90 -4 -2 0 2 4 36 38 40 42 44 46 48 50 -4 -2 0 2 4 28 30 32 34 36 38 40 42 -4 -2 0 2 4 14 16 18 20 22 24 26 -4 -2 0 2 4 6 8 10 12 14 16 18 20 RELATIVE TIME (ms) RELATIVE TIME (ms) RELATIVE TIME (ms) RELATIVE TIME (ms) TOROIDAL ROTATION (km/s) NEUTRON LOSS RATE (ms-1) (a) (b) 177 cm (c) 187 cm (d) 193 cm (h) 219 cm (g) 212 cm(f) 204 cm(e) 199 cm Figure 17. Time evolution of (a) the fast-ion loss rate inferred from the drop in neutron signal and (b)–(h) the toroidal rotation for 7 CER channels. The data are conditionally averaged over an ensemble of relatively large (130 <B max <180 T s−1) fishbone bursts. The error bar in (a) represents the standard deviation of the peak loss rate. The error bars in (b)–(h) represent the standard deviation of the relative change between −0.75 and −0.25 ms. measure different types of loss orbits, the final orbit leg effectively scrambles the toroidal phase of the various measurements. A similar result was obtained for PDX fishbones. Both the NPA [3] and a silicon detector that directly measured lost fast ions [41] observed a beacon but the two diagnostics did not observe the same phase relative to the mode [33]. Since the fast-ion losses are non-ambipolar, they should transiently alter the radial electric field Er. Because the plasma must preserve radial force balance, a changing electric field alters the toroidal rotation. To detect this effect, CER measurements of toroidal rotation are acquired in 0.5 ms time bins. To obtain adequate statistics, data from 16 similar fishbone bursts are conditionally averaged. The results appear in figure 17. All CER channels measure an abrupt drop in toroidal rotation of ∼−10 km s−1(1 kHz) when the neutron loss rate is greatest. This suggests that fast-ion losses occur throughout the plasma. Comparing channels, the drop persists longer at large major radius than it does in the core. The magnitude of the drop is comparable to the ∼+10 km s−1jump in toroidal rotation that is observed when a single neutral beam source injects for 10 ms [42]. Evidently, the fishbone burst acts like a ‘negative beam blip.’ A 10 ms beam blip injects about +0.5 C of charge. From the ∼10% drop in neutron rate for these bursts, the estimated charge of expelled fast ions is ∼−0.3 C, so the ‘negative beam blip’ associated with the fishbone should reduce the toroidal rotation about as much as a positive beam blip increases it, just as is observed. 3.3. Confined fast ions In addition to the drop in signal associated with fast-ion losses, the neutron signal also contains oscillations (figure 18). After filtering out high-frequency noise, these oscillations are visible 18 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al -400 -200 0 200 400 1837 1838 1839 1840 0.80 0.85 0.90 0.95 1.00 1.05 1838.0 1838.5 1839.0 1839.5 1840.0 (a) Mirnov (T/s) (b) Neutrons (a.u.) TIME (ms) (c) #141076 0 5 10 15 20 25 30 (d) Half-cycle # 0 0.4 0.8 Distortion Neutron Phase 0 400 800 Figure 18. Time evolution of (a) the Mirnov coil signal and (b) the neutron signal during a fishbone burst. (c) Overlay of the oscillations in the Mirnov signal and the neutron signal. An error function fit to the drop in the neutron rate is used to detrend the neutron signal. (d) Evolution of the mode distortion and neutron phase (degrees) versus half-period number. The dashed vertical line indicates the time of maximum amplitude. in the total signal (figure 18(b)) but they become particularly evident after detrending the signal to remove the overall drop (figure 18(c)). In contrast to all of the lost-ion measurements, the phase of these neutron oscillations does not remain constant throughout the fishbone burst. Instead, the phase steadily slips relative to the Mirnov trace as the burst evolves (figure 18(c)). The phase slippage is largest near maximum amplitude, which is also when the mode distortion 19 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 0.0 0.5 1.0 1.5 2.0 -8-6-4-20246 -100 0 100 200 300 400 4 5 6 7 8 9 PERIOD # (relative to Mirnov peak) 0 200 200 150 150 100 100 50 50 MIRNOV AMP. (T/s) MIRNOV FREQ(kHz) (c) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 50 100 150 200 250 Peak Mirnov Amplitude (T/s) 0 100 200 300 400 NEUT. AMP. (%) NEUT. AMP. (%) NEUT. PHASE (º) PHASE SLIP (º) (a) (b) (d) Figure 19. Average neutron fluctuation (a) amplitude and (c) phase versus period number for 40 relatively large (140 <B max <162 T s−1) fishbone bursts. The evolution of the mode (a) amplitude and (c) frequency for the same bursts is also shown. The error bars represent the standard deviation for this ensemble. (b) Neutron fluctuation amplitude and (d) phase slippage versus maximum Mirnov amplitude for all bursts with constant beam power during the burst. The lines are linear fits to the data. The amplitude is an average over cycles −3–0; the phase slippage is the average over cycles −5—3 subtracted from the average over cycles 1–3; the utilized cycles are highlighted by boxes in (a) and (c), respectively. and frequency chirping rapidly increase (figure 18(d)). For this particular case, in the course of the burst, the phase slips over 360◦. Because it measures (predominately) virgin neutrons, the plastic scintillator signal is sensitive to the geometrical position of the confined fast ions that produce the neutron reactions. Neutron oscillations of similar amplitude were measured for PDX fishbones [2]. Strachan et al [2] showed that these oscillations were caused by the helical distortion of the neutronemitting core. Using the expected geometrical efficiency of the plastic scintillator and the measured displacement of the internal kink mode, they showed that the predicted oscillations agree in both amplitude and phase with the observations. No phase slippage was reported. Strachan’s model cannot account for the phase slippage associated with off-axis fishbones. Presumably, as for PDX fishbones, the oscillations are associated with motion of confined fast ions toward and away from the detector. (Motion relative to the thermal deuterium density gradient may also contribute.) But the phase slippage indicates that the oscillations of the majority of confined fast ions decouples from the mode itself, even while the expelled ions retain their phase. This decoupling accelerates when the mode is large, distorting and changing rapidly in frequency. This phenomenon occurs for all off-axis fishbone bursts. Figures 19(a) and (c) show the average behavior for a set of similar bursts. The neutron fluctuation amplitude usually peaks slightly before maximum Mirnov amplitude. (For an average of all of the bursts in the database, the peak occurs two cycles prior to the Mirnov maximum.) The phase changes most rapidly when the frequency changes. Note that a fixed temporal delay relative to the mode would produce phase changes of the opposite sign. For the entire database, the neutron fluctuation amplitude scales linearly with maximum Mirnov amplitude (figure 19(b)) (r=0.80). The phase slippage also tends to increase with maximum Mirnov amplitude (figure 19(d)) (r=0.55). In this latter case, the relatively large uncertainty in determination of the phase slip may account for some of the scatter. 20 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al 4. Discussion and conclusion Many features of DIII-D off-axis fishbones are similar to PDX q=1 fishbones. •The modes occur in high-beta plasmas near an MHD stability limit. •They are n=1 modes driven by neutral beam ions. •The initial mode frequency matches the precession frequency of full-energy trapped ions; also, the radial profile of the precession frequency is fairly flat, so fast ions can stay in resonance while being transported radially. •The frequency chirp is large (f /fi≃0.4) and increases with increasing mode amplitude. Also, during a burst, the frequency chirp rate is largest near maximum mode amplitude. •The growth rate is γ/ω ≃5%. •The modes appear in repetitive bursts. (The burst cycle was more regular for PDX fishbones but this difference is probably caused by the strong beam modulation in the DIII-D experiments.) •The modes cause losses of fast ions that peak near the time of maximum mode amplitude. •The magnitude of the losses scales approximately linearly with mode amplitude. •Fast ions are expelled to the outside of the plasma in a ‘beacon’ (with a fixed phase relative to the mode). •Higher nmodes with similar phenomenology are occasionally observed, as they were on PDX [2]. In addition, the loss of fast ions causes a sudden drop in toroidal rotation. This effect was not measured on PDX but probably occurred. On the other hand, several features of DIII-D off-axis fishbones differ from PDX fishbones. •On PDX, the decay rate was smaller than the growth rate but the opposite is true for off-axis fishbones. Also, the decay phase is highly variable for off-axis fishbones but, in PDX, except when a sawtooth crash occurred, the decay rate was regular. •On PDX, the waveform remained nearly sinusoidal throughout the burst. In contrast, for off-axis fishbones, the waveform always becomes highly distorted near maximum mode amplitude. •On PDX, the fluctuations in the neutron signal maintained a fixed phase relative to the mode. For off-axis fishbones, the neutron phase slips continuously when the amplitude is large and the mode frequency and waveform shape are changing rapidly. In addition, for off-axis fishbones, the radial eigenfunction changes shape during the evolution of a burst. This information is not available for PDX fishbones. The many similarities between PDX fishbones and off-axis fishbones supports the notion advanced by Okabayashi et al [15] that off-axis fishbones are an energetic-particle branch of a kink mode, just as PDX fishbones are an energetic-particle branch of the 1/1 internal kink. In this paradigm, the burst cycle resembles classic predator–prey relaxation oscillations because the increase in fast-ion density associated with beam fueling pushes a marginally stable mode across the stability boundary, then the expulsion of the fast ions stabilizes the mode [6,43]. In contrast to a resistive wall mode (or other normal mode of the background plasma), the mode frequency is determined by the precession frequency of the fast ions. The large frequency chirping and the non-perturbative nature of the eigenfunction are consistent with the idea that the off-axis fishbone is a beam mode that exists due to the intense fast-ion population. What accounts for the differences in nonlinear behavior of PDX and off-axis fishbones? One possible explanation is that the off-axis fishbone is an external mode instead of an internal mode. (The response of the instability to feedback is strong evidence for the external nature of 21 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al the off-axis fishbones [15].) In this hypothesis, the mode distortion is caused by the dissipation associated with the conducting wall. Another difference between PDX fishbones and off-axis fishbones is the beam fueling. In PDX, all of the sources injected at the same angle, so the fast-ion distribution function was concentrated at a single pitch angle. In contrast, the DIII-D experiments employ many angles of injection. (In addition, Teis sufficiently high that the pitch-angle scattering rate is comparable to the slowing-down rate, which further isotropizes the distribution function.) Perhaps the response of confined fast ions from other sources is responsible for the phase slippage in the neutron oscillations and also contributes to the distortion of the mode. A key goal of the energetic-particle community is to develop a predictive capability for fast-ion driven instabilities. The fascinating nonlinear evolution of off-axis fishbones is a ripe topic for theoretical study. Acknowledgments We thank John deGrassie, Boris Breizman, Liu Chen, and the referees for helpful insights and gratefully acknowledge the invaluable contributions of the DIII-D team. This work was funded by the US Department of Energy under SC-G903402 and DE-FC02-04ER54698 DE-FG02-07ER54917. References [1] Mcguire K et al 1983 Phys. Rev. Lett. 50 891 [2] Strachan J D et al 1985 Nucl. Fusion 25 863 [3] Beiersdorfer P, Kaita R and Goldston R J 1984 Nucl. Fusion 24 487 [4] Goldston R J et al 1987 Nucl. Fusion 27 921 [5] White R B et al 1983 Phys. Fluids 26 2958 [6] Chen L, White R B and Rosenbluth M N 1984 Phys. Rev. Lett. 52 1122 [7] Heidbrink W W and Sadler G J 1994 Nucl. Fusion 34 535 [8] Coppi B and Porcelli F 1986 Phys. Rev. Lett. 57 2272 [9] Chen L 1994 Phys. Plasma 11519 [10] Porcelli F 1991 Plasma Phys. Control. Fusion 33 1601 [11] HuysmansGTAet al 1999 Nucl. Fusion 39 1489 [12] Matsunaga G et al 2009 Phys. Rev. Lett. 103 045001 [13] Matsunaga G et al 2010 Nucl. Fusion 50 084003 [14] Okabayashi M et al 2009 Nucl. Fusion 49 125003 [15] Okabayashi M et al 2011 Phys. Plasma 18 056112 [16] Helander P, Gimblett C G, Hastie R J and Mcclements K G 1997 Phys. Plasma 42181 [17] Freidberg J P 1987 Ideal Magnetohydrodynamics (New York: Plenum) section 9.4.6 [18] Lao L L, St John H, Stambaugh R D, Kellman A G and Pfeiffer W 1985 Nucl. Fusion 25 1611 [19] Rice B W, Nilson D G and Wroblewski D 1995 Rev. Sci. Instrum. 66 373 [20] Strait E J 2006 Rev. Sci. Instrum. 77 023502 [21] Austin M E and Lohr J 2003 Rev. Sci. Instrum. 74 1457 [22] Gohil P, Burrell K H, Groebner R J and Seraydarian R P 1990 Rev. Sci. Instrum. 61 2949 [23] Heidbrink W W 1986 Rev. Sci. Instrum. 57 1769 [24] Zhu Y B, Heidbrink W W and Pickering L D 2010 Nucl. Fusion 50 084024 [25] Fisher R K et al 2010 Rev. Sci. Instrum. 81 10D307 [26] Watkins J G et al 1992 Rev. Sci. Instrum. 63 4728 [27] Luo Y, Heidbrink W W, Burrell K H, Gohil P and Kaplan D 2007 Rev. Sci. Instrum. 78 033505 [28] Muscatello C M, Heidbrink W W, Taussig D and Burrell K H 2010 Rev. Sci. Instrum. 81 10D316 [29] Gupta D K, Fonck R J, Mckee G R, Schlossberg D J and Shafer M W 2004 Rev. Sci. Instrum. 75 3493 [30] Heidbrink W W, Mckee G R, Smith D R and Bortolon A 2011 Plasma Phys. Control. Fusion 53 085007 [31] Gorelenkov N N and Cheng C Z 1995 Nucl. Fusion 35 1743 [32] Watson G W and Heidbrink W W 2003 Rev. Sci. Instrum. 74 1605 22 Plasma Phys. Control. Fusion 53 (2011) 085028 W W Heidbrink et al [33] BuchenauerDAJ1985 Fast ion effects on magnetic instabilities in the PDX tokamak PhD Thesis Princeton University [34] Heidbrink W W 1995 Plasma Phys. Control. Fusion 37 937 [35] Shinohara K et al 2007 Nucl. Fusion 47 997 [36] Graves J P, Hastie R J and Hopcraft K I 2000 Plasma Phys. Control. Fusion 42 1049 [37] White R B 2001 The Theory of Toroidally Confined Plasmas 2nd edn (London: Imperial College Press) [38] Van Zeeland M A et al 2011 Phys. Plasma 18 056114 [39] Budny R V 1994 Nucl. Fusion 34 1247 and references therein [40] Heidbrink W W and Beiersdorfer P 1987 Nucl. Fusion 27 608 [41] Buchenauer D, Heidbrink W W, Roquemore L and Mcguire K 1987 Rev. Sci. Instrum. 58 2264 [42] Degrassie J S, Groebner R J and Burrell K H 2006 Phys. Plasma 13 112507 [43] Heidbrink W W et al 1993 Phys. Fluids B52176 23