The Gamma Ray Imager (GRI) is a pinhole camera providing 2D imaging of MeV hard x-ray (HXR) bremsstrahlung emission from runaway electrons (REs) over the poloidal cross section of the DIII-D tokamak. We report a series of upgrades to the GRI expanding the access to RE scenarios from the diagnosis of a trace amount of REs to high flux HXR measurements during the RE plateau phase. We present the implementation of novel gamma ray detectors based on LYSO and YAP crystals coupled to multi-pixel photon counters, enabling a count rate in excess of 1 MHz. Finally, we highlight new insights into the RE physics discovered during the current quench and RE plateau phase experiments as the result of these upgrades.

Runaway electrons (REs) generated in tokamak disruptions can reach energies of tens of MeV and carry the majority of the pre-disruption current. In the case of a localized loss to the wall, they can cause serious damage. Melting of wall tiles addressed to a RE strike has been already observed in existing tokamaks.1,2 Since the RE current and total RE energy content are expected to be much higher in ITER and future high-current tokamaks, REs must be avoided or mitigated in them.3,4 To evaluate the effectiveness of disruption mitigation techniques, study RE physics, and validate existing RE models, diagnosis of REs in existing tokamaks is necessary. In this paper, we describe measurements of the hard x-ray (HXR) bremsstrahlung radiation from REs on the DIII-D tokamak using the Gamma Ray Imager (GRI). We report how the continuous upgrades to the GRI expanded its operational space and provided an access to new RE physics. Since the HXR energy measured by the GRI is in the range from sub-MeV to tens MeV, we use the words “HXRs” and “gamma particles” interchangeably throughout the paper.

The GRI is a pinhole camera imaging the 2D bremsstrahlung emission from REs on DIII-D. It consists of a lead body, a 3 in. long lead collimator, and an array of gamma ray detectors (up to 123) acting as individual pixels of the camera as shown in Fig. 1. A unique feature of the GRI is the ability to span the almost entire DIII-D poloidal cross section as shown in Fig. 2(b). This is achieved due to its close proximity to the plasma—the GRI is installed at the DIII-D mid-plane, inside a re-entrant port between toroidal magnetic field and error field correction coils—and the use of tangential sightlines as pictured in Fig. 2(a). The initial GRI design is described in detail in Ref. 5.

FIG. 1.

The GRI on DIII-D (cut in the vertical plane). It has an irregular shape with approximate dimensions of 35 × 32 × 32 cm3 (W × H × D).

FIG. 1.

The GRI on DIII-D (cut in the vertical plane). It has an irregular shape with approximate dimensions of 35 × 32 × 32 cm3 (W × H × D).

Close modal
FIG. 2.

(a) Toroidal cross section of DIII-D showing the GRI location and sightlines (excluding the sightlines observing the centerpost) and (b) GRI sightlines in the DIII-D poloidal cross section.

FIG. 2.

(a) Toroidal cross section of DIII-D showing the GRI location and sightlines (excluding the sightlines observing the centerpost) and (b) GRI sightlines in the DIII-D poloidal cross section.

Close modal

The GRI went through a series of upgrades in the past few years. Among them is the doubling of the number of sightlines equipped with gamma ray detectors [from 30 to 56 BGO detectors, see Figs. 3(a) and 3(b)]; replacement of the machined aluminum housing using threads to secure the detectors with a 3D-printed non-conductive plastic holder, providing quick installation of the detectors and electrical insulation from the vessel [see Figs. 3(c) and 3(d)]; and implementation of a movable rear shielding (consisting of 5 cm thick lead blocks, with the thickness limited by the shielding weight of 215 kg safely supported by tokamak structures) reducing the flux of non-collimated back-scattered by tokamak hall structures HXRs by a factor of 10 [see Figs. 3(e) and 3(f)].

FIG. 3.

(a) and (b) The sightlines populated with detectors in experiments before and after the number of detectors was doubled (the numbers in rounds refer to digitizers’ channels), (c) and (d) old aluminum detector housing and new 3D-printed non-conductive plastic holder, respectively, and (e) and (f) the GRI before and after the rear lead shielding was installed.

FIG. 3.

(a) and (b) The sightlines populated with detectors in experiments before and after the number of detectors was doubled (the numbers in rounds refer to digitizers’ channels), (c) and (d) old aluminum detector housing and new 3D-printed non-conductive plastic holder, respectively, and (e) and (f) the GRI before and after the rear lead shielding was installed.

Close modal

The initial and most basic gamma ray detectors of the GRI consist of a scintillation BGO (Bismuth Germanate) crystal coupled to a PIN diode [Fig. 4(a)], manufactured by Scionix Holland BV. These detectors feature a compact design and the ability to operate in strong magnetic fields (>1 T). The BGO crystal has a pulse decay time of about 300 ns, but the PIN diode, read-out electronics, and long signal cables (>50 m) result in a total pulse decay time of about 60 µs (by 1/e), which translates to the maximum count rate of about 20 kHz. The BGO + PIN detector (referred to as “BGO detector” below) in the GRI is sensitive to HXRs with energy >0.5 MeV and has an energy resolution of about 30% at 1 MeV. Since the GRI is located close to DIII-D magnetic coils, it is subject to strong electromagnetic pick-up noise. To reduce this noise, a pair of detectors exhibiting a similar noise pattern (evaluated using the standard deviation) is found, and their signals are subtracted from each other, which accentuates gamma pulses. To further reduce the noise, a soft smoothing filter is applied (reducing the high-frequency component), and a pulse shaping filter (Digital Ramp to Gaussian Shaper, DRGS) is used to reduce the low-frequency component and shorten the long pulse tails as shown in Fig. 4(b). The amplitudes and timestamps of these processed pulses are found via pulse height analysis (PHA), and HXR spectra are produced by analyzing the statistics of pulses. The details of the PHA, as well as results of the initial application of the BGO detectors on DIII-D, can be found in Ref. 6. The inversion of the HXR energy spectra to RE energy spectra is made using a synthetic diagnostic and applying an onion-peel method (by going from high to low HXR energies and calculating the RE energy required to produce such HXRs). It takes into account the geometry and material of the scintillation crystal based on the results of Monte Carlo simulations. The synthetic diagnostic and inversion technique are presented in detail in Refs. 6 and 7, respectively.

FIG. 4.

(a) A GRI BGO detector and (b) a raw signal by the BGO detector showing gamma pulses and the output of the digital ramp to Gaussian shaper filter.

FIG. 4.

(a) A GRI BGO detector and (b) a raw signal by the BGO detector showing gamma pulses and the output of the digital ramp to Gaussian shaper filter.

Close modal

The BGO detectors reliably measure the HXR radiation from a trace amount of REs. This scenario is typically achieved during an ohmic flattop discharge when the plasma density decreases to such a small value where the acceleration of electrons by the electric field overcomes the collisional drag and RE generation can be observed. In this so-called Quiescent Runaway Electron (QRE) regime, the majority of the plasma current is still thermal; however, the magnitude of signals from REs is high enough, and their temporal dynamics is slow enough to allow their studying using the BGO detectors. In such QRE experiments, the RE energy distribution, its evolution in the course of time, and the effects of collisional damping and synchrotron radiation have been studied for the first time using HXR measurements by the GRI.8 It has been found that the RE distribution is non-monotonic and has a bump at mid-energies (about 7 MeV) as shown in Figs. 5(a) and 5(b), which was addressed to an interplay between RE acceleration by the electric field and deceleration via collisions and synchrotron emission. Later, these experiments and comparisons with the modeling were discussed in more detail in Ref. 7. These studies were followed up by the observation of kinetic instabilities driven by the RE beam in the QRE regime.9 

FIG. 5.

(a) and (b) HXR and RE spectra measured by the BGO detectors in the QRE experiment.8 Reproduced with permission from Paz-Soldan et al., Phys. Rev. Lett. 118, 255002 (2017). Copyright 2017 American Physical Society. (c) saturation of the BGO detector during the RE plateau experiment.

FIG. 5.

(a) and (b) HXR and RE spectra measured by the BGO detectors in the QRE experiment.8 Reproduced with permission from Paz-Soldan et al., Phys. Rev. Lett. 118, 255002 (2017). Copyright 2017 American Physical Society. (c) saturation of the BGO detector during the RE plateau experiment.

Close modal

In disruption experiments with the formation of a non-trace RE current amount, the BGO detectors show the limitation of their count rate of 20 kHz. In all analyzed RE plateau scenarios, BGO detectors experience very high pile-up, leading to a complete lack of distinguishable HXR pulses as shown in Fig. 5(c). Attempts to reduce the incident HXR flux by using tungsten inserts to the GRI collimator show an insufficient attenuation level; since to keep the collimation ratio above at least a factor of 3–5, the inserts should attenuate the HXR flux by no more than a factor of 10 [see Fig. 16(a) for reference]. An application of another BGO detector, also made by Scionix, where the PIN diode is replaced by a multi-pixel photon counter (which allowed increasing the maximum count rate up to 500 kHz), showed very quick degradation of the detectors’ electronics, up to the total lack of operability, in the harsh neutron and gamma DIII-D environment. To overcome the limitation of the relatively low count rate, completely new faster HXR detectors were required, as will be discussed in Sec. IV.

The LYSO + MPPC detectors (referred to as “LYSO detectors” below) have been developed at the University of Milan-Bicocca10 and installed on the DIII-D tokamak in order to increase the GRI count rate and extend its capability to the RE plateau phase. These detectors consist of an LYSO:Ce (Cerium doped Lutetium Yttrium Orthosilicate) crystal coupled to a multi-pixel photon counter (MPPC). Many properties of an LYSO:Ce crystal are comparable to the properties of a BGO crystal (see Table I), but the pulse decay of the LYSO:Ce crystal is much shorter (35 vs 300 ns, respectively). Due to the use of an MPPC and optimized read-out electronics, the decay time of the LYSO detector is only about 50 ns, which is by a factor of 1000 faster than for the BGO detector. This allows achieving the count rate in excess of 1 MHz. The LYSO detector is also less sensitive to electromagnetic pick-up, has a noise floor of 0.4 MeV, and has an energy resolution of about 8% at 1 MeV.10 Comparison of HXR pulses recorded by the LYSO and BGO detectors is shown in Fig. 6(a). A zoomed-in single pulse by the LYSO detector is shown in Fig. 6(b). The signals from the LYSO detectors are transmitted by about 50 m to Annex, where they are continuously digitized at 200 MHz for up to 5 s (previously by GaGe CSE1642 and presently by AlazarTech ATS9637 digitizers). This results in 2 GB of data generated per detector per discharge and requires special handling to read and process the data. This is presently implemented offline in IDL and is based on the processing of segmented data using the same method as described in Sec. III but with adjusted coefficients. The read-out electronics is chosen to be relatively simple and handle only power and signal since a prototype of the LYSO detector having a built-in high-voltage generator and capable of in situ MPPC gain correction has been found to pick up very high electromagnetic noise (up to above 1 V, while the detector response to gammas was about 10 mV/MeV) in the DIII-D environment. The LYSO detector and its main components can be seen in Figs. 7(a)7(c).

TABLE I.

(top) Properties of scintillation crystals and (bottom) parameters of the HXR detectors on DIII-D (MPPC and read-out electronics are different for every detector).

CrystalDensity (g/cm3)Light yield (photons/keV)Decay time (ns)Energy resolution at 662 keV (%)Attenuation length for 511 keV (cm)
BGO 7.1 300 12 
LYSO:Ce (Lu1.8Y0.2SiO5:Ce) 7.1 33 36 1.2 
YAP:Ce (YAlO3:Ce) 5.4 25 25 ≈5 2.7 
CrystalDensity (g/cm3)Light yield (photons/keV)Decay time (ns)Energy resolution at 662 keV (%)Attenuation length for 511 keV (cm)
BGO 7.1 300 12 
LYSO:Ce (Lu1.8Y0.2SiO5:Ce) 7.1 33 36 1.2 
YAP:Ce (YAlO3:Ce) 5.4 25 25 ≈5 2.7 
DetectormV/MeVDecay time by 1/e (ns)Detectable count rate (1/s)
BGO + PIN 40 60 000 104 
BGO + MPPC 25 2 000 105 
LYSO + MPPC 10 50 >106 
YAP + MPPC 20 55 >106 
DetectormV/MeVDecay time by 1/e (ns)Detectable count rate (1/s)
BGO + PIN 40 60 000 104 
BGO + MPPC 25 2 000 105 
LYSO + MPPC 10 50 >106 
YAP + MPPC 20 55 >106 
FIG. 6.

(a) Comparison of HXR pulses recorded by the BGO and LYSO detectors and (b) single HXR pulse by the LYSO detector.

FIG. 6.

(a) Comparison of HXR pulses recorded by the BGO and LYSO detectors and (b) single HXR pulse by the LYSO detector.

Close modal
FIG. 7.

(a) LYSO crystal (50 ×□13 mm), (b) MPPC and read-out base, and (c) the GRI LYSO detector assembled.

FIG. 7.

(a) LYSO crystal (50 ×□13 mm), (b) MPPC and read-out base, and (c) the GRI LYSO detector assembled.

Close modal

The implementation of the LYSO detectors to the GRI provides a count rate exceeding 1 MHz. Typically, a factor absolutely limiting the HXR measurements is the pile-up level high enough to prevent the identification of all or the most of HXR pulses. However, for the LYSO detector, the practical limit is the pile-up level low enough to allow identification of the most of HXR pulses. This is because the pile-up always results in worse statistics since not every pulse can be discriminated, and also, the amplitude of recognized pulses can deviate in both directions from the true value. Such effects are particularly undesirable for the LYSO detector due to the non-linear response of the MPPC.10 At first, the MPPC gain decreases as the HXR flux increases since the increasing current reduces the MPPC overvoltage; for example, the relative gain shift exceeds 10% as the count rate of 1 MeV mono-energetic HXRs exceeds 1 MHz. Secondly, the MPPC gain also decreases as the HXR energy increases since the increasing amount of light increases the probability of two or more photons interacting with the same MPPC avalanche photodiode cell at the same time; for example, a HXR with an energy of 10 MeV would be detected as a HXR with an energy of 7.5 MeV. This non-linear response can be compensated during the PHA using the calibration curves as the ones obtained in Ref. 10. However, this requires the LYSO operation at a low pile-up level in order to avoid a significant number of missed HXR pulses and incorrectly measured HXR energies. The comparison of LYSO pile-up levels at different simulated HXR count rates is shown in Fig. 8. It can be seen that the increasing pile-up leads to the signal shifting away from the zero baseline. Experimentally, we limit the region of analysis by the pile-up between very low and low levels, which corresponds to the count rate of about 4 MHz. Since the PHA cannot resolve two pulses if their peaks are within a time window of about 35 ns, the number of pulses missed at 4 MHz is about 13%.

FIG. 8.

(a) Very low pile-up level at the simulated HXR count rate of 2.5 · 106 s−1, (b) low pile-up at the HXR count rate of 1.2 · 107 s−1, and (c) high pile-up at the HXR count rate of 6.5 · 107 s−1. Assuming that the PHA cannot discriminate two pulses with peaks within less than 35 ns and taking this as “dead time,” 8% of all HXRs would be missed in case (a), 35% in case (b), and 90% in case (c).

FIG. 8.

(a) Very low pile-up level at the simulated HXR count rate of 2.5 · 106 s−1, (b) low pile-up at the HXR count rate of 1.2 · 107 s−1, and (c) high pile-up at the HXR count rate of 6.5 · 107 s−1. Assuming that the PHA cannot discriminate two pulses with peaks within less than 35 ns and taking this as “dead time,” 8% of all HXRs would be missed in case (a), 35% in case (b), and 90% in case (c).

Close modal

An example of the LYSO measurements’ validity region in the real RE experiment11 is shown in Fig. 9. In this experiment, a disruption with the generation of the RE beam is deliberately caused by the injection of an argon pellet. Later in the discharge, to study RE dynamics in the deuterium background, argon is purged by massive D2 injection.12 

FIG. 9.

(a) Raw signal by the LYSO detector in the RE experiment11 and (b) HXR count rate obtained via PHA. Adapted with permission from Lvovskiy et al., Nucl. Fusion 60, 056008 (2020). Copyright 2020 IAEA, Vienna.

FIG. 9.

(a) Raw signal by the LYSO detector in the RE experiment11 and (b) HXR count rate obtained via PHA. Adapted with permission from Lvovskiy et al., Nucl. Fusion 60, 056008 (2020). Copyright 2020 IAEA, Vienna.

Close modal

While the RE beam is in the argon plasma, the moving minimum of the raw signal shifts far above the baseline, defined by the signal before the disruption; however, the HXR count rate found via PHA is unexpectedly low. This clearly indicates a very high pile-up and deficit of discriminated HXRs during the argon phase. Once argon is replaced by deuterium, the bremsstrahlung radiation decreases (its power roughly scales as nZ2, where n is the impurity density and Z is the impurity atomic number), the moving minimum returns to the baseline, and the resolved HXR count rate increases and reaches 4.3 · 106 s−1. Such regions are considered suitable for analysis.

Commissioning of the LYSO detector allowed first-time measurements of the HXR spectra during the RE plateau phase and current quench (CQ) on DIII-D. Here we briefly present the main results achieved. The words “RE plateau phase” and “RE beam” used below describe the same phenomenon of sustained RE current after the tokamak disruption.

1. RE-driven kinetic instabilities during the CQ

The LYSO detector is capable of obtaining HXR spectra during the CQ quench (about 5–10 ms on DIII-D) while the RE current increases and replaces the thermal current. These measurements were used to revise the scenario with RE beam generation using argon massive gas injection.13 Empirically, it is known that a greater amount of argon injected leads to a higher probability of sustained RE beam on DIII-D. Using radio frequency measurements of toroidal magnetic fluctuations, it was found that both cases with and without RE beam exhibit MHz magnetic fluctuations during the CQ. These fluctuations are longer in time, extend to higher frequencies, and are more powerful in the case without sustained RE beam as shown in Figs. 10(a) and 10(b). The fluctuations also correlate with increased RE loss and are observed when the maximum energy of REs exceeds about 3 MeV. The HXR spectra, obtained during the CQ, indicate that the RE population is more energetic in disruptions without sustained RE beams as shown in Fig. 10(c). However, the increasing amount of impurity injected leads to less energetic REs, smaller power and energy of instabilities, and eventually sustained RE beam. It was proposed that REs drive Compressional Alfvén Eigenmodes (CAEs) during the CQ, which increases the radial transport of REs and reduces avalanche gain preventing the generation of sustained RE beams. The increasing amount of argon presumably increases collisional drag and reduces the CQ time available for RE acceleration by the induced electric field, which reduces the energy of REs and causes weaker CAEs. This experiment is discussed in more detail in Refs. 13 and 14.

FIG. 10.

(a) and (b) Fluctuations of the toroidal magnetic field, plasma current, and RE loss measured by the distant HXR detector in discharges without and with sustained RE beam respectively,13 (c) HXR spectra measured by the LYSO detector during the CQ for different amounts of argon injected to trigger the disruption. The integration window is the same for all cases shown, equal to the first 10 ms after the disruption. Adapted with permission from Lvovskiy et al., Plasma Phys. Controlled Fusion 60, 124003 (2018). Copyright 2018 IOP Publishing Ltd.

FIG. 10.

(a) and (b) Fluctuations of the toroidal magnetic field, plasma current, and RE loss measured by the distant HXR detector in discharges without and with sustained RE beam respectively,13 (c) HXR spectra measured by the LYSO detector during the CQ for different amounts of argon injected to trigger the disruption. The integration window is the same for all cases shown, equal to the first 10 ms after the disruption. Adapted with permission from Lvovskiy et al., Plasma Phys. Controlled Fusion 60, 124003 (2018). Copyright 2018 IOP Publishing Ltd.

Close modal

2. Reconstruction of RE current profile

To assess the MHD stability of the RE beam, the RE current profile was reconstructed for the first time using HXR measurements.11 It was made using the single central GRI sightline equipped with the LYSO detector during the slow (300 ms) downward motion of the RE beam as shown in Fig. 11(a). The measured HXR spectra were inverted to RE spectra, which provided local current density. This current density was supplied as constraints to the equilibrium reconstruction code EFIT, which calculated the current profile and the safety factor profile as shown in Figs. 11(b) and 11(c), respectively. The reconstructed current profile was found to peak with internal inductance li = 1.13. However, this relatively low-current RE beam (Ip = 180 kA) was observed MHD stable, which can be explained by the elevated safety factor profile. Only RE beams with higher current and/or smaller cross section were found MHD unstable and driving internal and external kink instabilities.11,15,16

FIG. 11.

(a) Scenario of the RE current profile reconstruction via HXR measurements using the LYSO detector,11 (b) RE current profile reconstructed using HXR measurements as constraints to EFIT (flat impurity profile assumed, hollow impurity profiles are discussed in Ref. 11), and (c) safety factor profile. Adapted with permission from Lvovskiy et al., Nucl. Fusion 60, 056008 (2020). Copyright 2020 IAEA, Vienna.

FIG. 11.

(a) Scenario of the RE current profile reconstruction via HXR measurements using the LYSO detector,11 (b) RE current profile reconstructed using HXR measurements as constraints to EFIT (flat impurity profile assumed, hollow impurity profiles are discussed in Ref. 11), and (c) safety factor profile. Adapted with permission from Lvovskiy et al., Nucl. Fusion 60, 056008 (2020). Copyright 2020 IAEA, Vienna.

Close modal

3. RE-driven kinetic instabilities during the RE plateau phase

The LYSO detector also played an important role in the study of RE-driven frequency chirping instabilities. These instabilities were observed for the first time on DIII-D when a large decelerating loop voltage was applied to the post-disruption RE beam in deuterium plasma.17 In the MHz range, chirping instabilities correlated with increased RE loss from the plasma. Using the LYSO detector, the RE spectrum was obtained before the chirping was observed, and this revealed a pronounced bump at the RE distribution as shown in Fig. 12(c). Such a non-monotonic feature of the energy distribution function presents a potential source of free energy capable to drive kinetic instabilities via energy exchange in wave-particle resonant interactions. Fast relaxation of the RE distribution function was directly measured in this experiment, as shown in Figs. 12(d)12(g), which can explain the chirping nature of observed instabilities.

FIG. 12.

(a) Signals by the GRI (confined REs) and HXR distant (lost REs) detectors, (b) frequency chirping instabilities in the MHz range (shaded area indicates the chirping analyzed below), (c) RE energy distribution function measured before the chirping observed (integration time 5 ms), (d) and (e) relaxation of the HXR distribution function during a single chirping event (the whole process taking about 1 ms is shown in two stages of 0.5 ms, integration intervals are every 0.1 ms), (f) and (g) relaxation of the RE distribution function.. RE spectra present absolute numbers of electrons per second calculated via the inversion process described in Ref. 7. Adapted with permission from Lvovskiy et al., Nucl. Fusion 59, 124004 (2019). Copyright 2019 IAEA, Vienna.

FIG. 12.

(a) Signals by the GRI (confined REs) and HXR distant (lost REs) detectors, (b) frequency chirping instabilities in the MHz range (shaded area indicates the chirping analyzed below), (c) RE energy distribution function measured before the chirping observed (integration time 5 ms), (d) and (e) relaxation of the HXR distribution function during a single chirping event (the whole process taking about 1 ms is shown in two stages of 0.5 ms, integration intervals are every 0.1 ms), (f) and (g) relaxation of the RE distribution function.. RE spectra present absolute numbers of electrons per second calculated via the inversion process described in Ref. 7. Adapted with permission from Lvovskiy et al., Nucl. Fusion 59, 124004 (2019). Copyright 2019 IAEA, Vienna.

Close modal

Implementation of the LYSO detector to the GRI was a great advancement, providing access to HXR measurements in new RE scenarios and new RE physics. The LYSO detector was found to operate reliably during the CQ when the RE current is still replacing the thermal current and during the RE plateau phase at relatively low RE current (about 200 kA) in low-Z (deuterium) plasma. Measurements at higher RE current or in high-Z background plasma (such as argon) are very challenging due to strong bremsstrahlung radiation, leading to high pile-up. To overcome this limitation, novel detectors have been developed at the University of Milan-Bicocca18 and commissioned on DIII-D as discussed in Sec. V.

These novel detectors consist of a YAP:Ce (cerium doped yttrium aluminum perovskite) crystal coupled to an MPPC (referred to as “YAP detector” below). Compared to the LYSO crystal, the YAP crystal has a shorter pulse decay time (25 ns vs 36 ns); however, due to delays introduced by the MPPC and read-out base, the pulses produced by both detectors have about the same decay time of 50–55 ns as shown in Fig. 13(a). Despite the YAP detector is not faster than the LYSO detector in real experiments, it allows HXR measurements at significantly higher incident gamma flux due to two key factors. At first, the YAP crystal has a lower density compared to the LYSO crystal, which results in almost twice a longer attenuation length at 511 keV as listed in Table I. Taking into account five times shorter crystal length [see Fig. 16(b) for comparison of the crystal absorption] and two times smaller crystal cross-sectional area, it makes the YAP detector about ten times less sensitive to the HXR total flux, which effectively reduces the observed pile-up. Secondly, the YAP detector has a built-in LED, which can be used to monitor the MPPC gain shift throughout the discharge. The LED produces reference pulses all having the same amplitude if no HXRs observed; thus, if the MPPC gain changes due to variations in the gamma flux or gamma energy, this is clearly seen by the changing amplitude of LED pulses measured by the detector. This approach provides a way to compensate for the MPPC shift without relying on measurements of the gamma count rate and gamma energy unobscured by the pile-up and application of the calibration curves as is required for the LYSO detector. Because of that, the pile-up level tolerable by the YAP detector can be higher. The CAD model of the YAP detector presenting its main components is shown in Fig. 13(b).

FIG. 13.

(a) Comparison of HXR pulses measured by the LYSO and YAP detectors and (b) CAD model of the YAP detector presenting its main components.

FIG. 13.

(a) Comparison of HXR pulses measured by the LYSO and YAP detectors and (b) CAD model of the YAP detector presenting its main components.

Close modal

Another novel feature of the YAP detector is a built-in temperature sensor. Since the MPPC gain is temperature-dependent, annual (Δt = 10 °C, daily-averaged) and daily (Δt = 5 °C) variations of the temperature in the DIII-D hall can change the gain by up to 20%. This sensor can be used together with a power supply providing temperature compensated output voltage to adjust the bias voltage of the MPPC and ensure constant MPPC gain.

It is important to mention that the YAP detector has a casing with optimized dimensions compatible with the GRI detector housing, which will allow simultaneous use of an array of YAP detectors in the GRI (it is planned to begin with an array of ten detectors coarsely resolving the plasma core and edges).

A comparison of the YAP vs LYSO detector performance in a high HXR flux scenario is shown in Fig. 14. In this experiment, the RE beam is maintained in argon plasma from 710 to 830 ms, and its current gradually decreases from 400 kA. The YAP and LYSO detectors observe plasma through mid-plane sightlines Nos. 43 and 62, respectively [see Fig. 2(b) for reference]. Since the runaway plasma is circular, relatively small, and shifted to the inboard [see, for example, Fig. 11(a)], both detectors look close to the core of the RE beam. There are only very few pulses found via PHA for the LYSO detector until the RE current decreases to about tens kA and the pile-up becomes small enough to observe more pulses. In contrast to that, the YAP detector provides many more pulses discovered via PHA. In addition, the LED pulses can be easily observed above the HXR pulses. While the RE current decreases, the amplitude of LED pulses increases, which indicates that the MPPC gain recovers to its initial value and, moreover, that the gain shift can be compensated, and true HXR amplitudes can be obtained offline using LED pulses as a reference. These data show the capability of the YAP detector to operate in high flux RE scenarios.

FIG. 14.

(a) Plasma current, (b) HXR and LED pulses found via PHA for the YAP detector, and (c) HXR pulses found via PHA for the LYSO detector. The background signal is negligible and ignored.

FIG. 14.

(a) Plasma current, (b) HXR and LED pulses found via PHA for the YAP detector, and (c) HXR pulses found via PHA for the LYSO detector. The background signal is negligible and ignored.

Close modal

Since the YAP detector provides previously elusive access to HXR measurements of RE beam distribution in argon plasma, it is interesting to compare HXR spectra in deuterium (low-Z) and argon (high-Z) plasmas. Such an experiment is shown in Fig. 15(a). The RE current of about 100–150 kA is maintained initially in argon and then in deuterium. The HXR spectra, measured during the RE plateau, are shown in Fig. 15(c). It can be seen that the RE population is relatively highly energetic in the argon plasma with the maximum RE energy exceeding 20 MeV. While the argon is replaced by deuterium, the HXR spectra become less energetic, and the maximum energy of REs decreases to about 8 MeV. These are the first results obtained with the YAP detector in such a RE scenario. They still need to be analyzed and compared with the modeling. Preliminarily, this change in the energy of the RE population can be explained by strong collisional damping of REs in the deuterium plasma, since there are 20 times more particles introduced by the deuterium injection compared to the argon injection, and no signs of electric field drop (explaining RE deceleration) can be seen in the plot of loop voltage in Fig. 15(b).

FIG. 15.

(a) and (b) Plasma current and loop voltage during the RE beam phase in argon and later in deuterium background plasmas and (c) HXR spectra obtained by the YAP detector at the times shown by the color lines in panels (a) and (b).

FIG. 15.

(a) and (b) Plasma current and loop voltage during the RE beam phase in argon and later in deuterium background plasmas and (c) HXR spectra obtained by the YAP detector at the times shown by the color lines in panels (a) and (b).

Close modal

The GRI is unique diagnostic imaging of the entire poloidal cross section of DIII-D and provides measurements of the HXR bremsstrahlung radiation from REs in the MeV range. Due to significant upgrades to the GRI, including the implementation of novel ultra-fast gamma ray detectors consisting of LYSO and YAP crystals coupled to MPPCs and enabling access to HXR measurements exceeding 1 MHz count rate and operation at a non-negligible pile-up level, the reliable diagnosis of RE scenarios was greatly expanded: from measurements of a trace amount of REs to the observation of RE generation during the current quench and finally to high flux HXR measurements during the RE plateau phase. This allowed studying for the first time RE-driven kinetic instabilities during the current quench and RE plateau, constraining the RE current profile via HXR measurements and diagnosis of HXR spectra from the RE beam in post-disruption deuterium and argon background plasmas.

Despite the GRI having been significantly improved in the series of continuous upgrades and many new insights into RE physics have been made, there are still many plans for future work. The main tasks planned for the upcoming years are (1) commission 10 YAP + MPPC detectors on DIII-D to provide simultaneous spatial HXR measurements of the RE beam distribution; (2) implement pulse fitting technique to improve pulse discrimination and energy resolution; (3) implement RE spectrum deconvolution using an expectation–maximization algorithm; (4) study RE seed birth location during thermal and current quenches; (5) study RE energy distribution in high-Z RE plateau experiments; (6) measure RE spatial distribution in RE plateau experiments; and (7) implement a collimator with a greater collimation ratio to access even higher RE current scenarios by using stronger attenuators.

The authors are grateful for the consultations and support of D. Pace, R. Tompkins, G. Campbell, K. Thome, G. Degrandchamp, D. Shiraki, A. Dvorak, D. Piglowski, I. Anyanetu, S. Beavers, J. Kulchar, K. Agustin, D. Ayala, D. Pierce, M. Watkins, W. Carrig, C. Pawley, R. Lee, K. Holtrop, and E. Gonzales.

This work was supported by the U.S. Department of Energy, Office of Science, Office of Fusion Energy Sciences, using the DIII-D National Fusion Facility, a DOE Office of Science user facility, under Award No. DE-FC02-04ER54698.

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of the authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

The authors have no conflicts to disclose.

A. Lvovskiy: Investigation (lead); Writing – original draft (lead). C. Paz-Soldan: Investigation (equal); Supervision (lead). N. Eidietis: Project administration (lead); Supervision (supporting). A. Dal Molin: Investigation (equal); Resources (equal); Validation (equal). M. Nocente: Investigation (supporting); Project administration (equal); Resources (equal); Validation (equal). C. Cooper: Investigation (equal); Software (equal). D. Rigamonti: Investigation (supporting); Resources (equal). M. Tardocchi: Project administration (supporting). D. Taussig: Resources (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

See Table 1. 

See Fig. 16.

FIG. 16.

(a) Mass attenuation coefficient for lead and tungsten. Since the minimum attenuation is observed for 3.5 MeV gammas, attenuation of such gammas is used as a figure of merit in the evaluation of different shielding and collimation lengths. The mass attenuation for stainless steel SS316L is given for comparison. (b) Absorption efficiency of scintillation crystals. This plot shows the percent of gammas absorbed by a crystal considered as a semi-infinite plate with a certain thickness. Solid lines indicate the length of crystals used in the GRI detectors. For example, an LYSO crystal of 5 cm would absorb 75% of 10 MeV gammas vs a YAP crystal of 1 cm absorbing only 14%.

FIG. 16.

(a) Mass attenuation coefficient for lead and tungsten. Since the minimum attenuation is observed for 3.5 MeV gammas, attenuation of such gammas is used as a figure of merit in the evaluation of different shielding and collimation lengths. The mass attenuation for stainless steel SS316L is given for comparison. (b) Absorption efficiency of scintillation crystals. This plot shows the percent of gammas absorbed by a crystal considered as a semi-infinite plate with a certain thickness. Solid lines indicate the length of crystals used in the GRI detectors. For example, an LYSO crystal of 5 cm would absorb 75% of 10 MeV gammas vs a YAP crystal of 1 cm absorbing only 14%.

Close modal
1.
I.
Jepu
 et al, “
Beryllium melting and erosion on the upper dump plates in JET during three ITER-like wall campaigns
,”
Nucl. Fusion
59
,
086009
(
2019
).
2.
M.
Diez
 et al,
West Team
, “
In situ observation of tungsten plasma-facing components after the first phase of operation of the WEST tokamak
,”
Nucl. Fusion
61
,
106011
(
2021
).
3.
M.
Lehnen
 et al, “
Disruptions in ITER and strategies for their control and mitigation
,”
J. Nucl. Mater.
463
,
39
48
(
2015
).
4.
R.
Sweeney
 et al, “
MHD stability and disruptions in the SPARC tokamak
,”
J. Plasma Phys.
86
,
865860507
(
2020
).
5.
D. C.
Pace
 et al, “
Gamma ray imager on the DIII-D tokamak
,”
Rev. Sci. Instrum.
87
,
043507
(
2016
).
6.
C. M.
Cooper
 et al, “
Applying the new gamma ray imager diagnostic to measurements of runaway electron Bremsstrahlung radiation in the DIII-D Tokamak (invited)
,”
Rev. Sci. Instrum.
87
,
11E602
(
2016
).
7.
C.
Paz-Soldan
 et al, “
Resolving runaway electron distributions in space, time, and energy
,”
Phys. Plasmas
25
,
056105
(
2018
).
8.
C.
Paz-Soldan
 et al, “
Spatiotemporal evolution of runaway electron momentum distributions in tokamaks
,”
Phys. Rev. Lett.
118
,
255002
(
2017
).
9.
D. A.
Spong
 et al, “
First direct observation of runaway-electron-driven whistler waves in tokamaks
,”
Phys. Rev. Lett.
120
,
155002
(
2018
).
10.
A.
Dal Molin
 et al, “
Development of a new compact gamma-ray spectrometer optimised for runaway electron measurements
,”
Rev. Sci. Instrum.
89
,
10I134
(
2018
).
11.
A.
Lvovskiy
 et al, “
Runaway electron beam dynamics at low plasma density in DIII-D: Energy distribution, current profile, and internal instability
,”
Nucl. Fusion
60
,
056008
(
2020
).
12.
E. M.
Hollmann
 et al, “
Study of argon expulsion from the post-disruption runaway electron plateau following low-Z massive gas injection in DIII-D
,”
Phys. Plasmas
27
,
042515
(
2020
).
13.
A.
Lvovskiy
 et al, “
The role of kinetic instabilities in formation of the runaway electron current after argon injection in DIII-D
,”
Plasma Phys. Controlled Fusion
60
,
124003
(
2018
).
14.
C.
Liu
 et al, “
Compressional Alfvén eigenmodes excited by runaway electrons
,”
Nucl. Fusion
61
,
036011
(
2021
); arXiv:2009.11801.
15.
C.
Paz-Soldan
 et al, “
Kink instabilities of the post-disruption runaway electron beam at low safety factor
,”
Plasma Phys. Controlled Fusion
61
,
054001
(
2019
).
16.
C.
Paz-Soldan
 et al,
DIII-D Team
, “
A novel path to runaway electron mitigation via deuterium injection and current-driven MHD instability
,”
Nucl. Fusion
61
,
116058
(
2021
).
17.
A.
Lvovskiy
 et al, “
Observation of rapid frequency chirping instabilities driven by runaway electrons in a tokamak
,”
Nucl. Fusion
59
,
124004
(
2019
).
18.
A.
Dal Molin
 et al, “
Novel compact hard x-ray spectrometer with MCps counting rate capabilities for runaway electron measurements on DIII-D
,”
Rev. Sci. Instrum.
92
,
043517
(
2021
).