- Email: [email protected]
Interaction of plasma transport and turbulence on particle fuelling
Journal of Nuclear Materials 438 (2013) S148–S154
Contents lists available at SciVerse ScienceDirect
Journal of Nuclear Materials journal homepage: www.elsevier.com/locate/jnucmat
Interaction of plasma transport and turbulence on particle fuelling P. Tamain a,⇑,1, G. Bonhomme b, F. Brochard b, F. Clairet a, C. Gil a, J. Gunn a, P. Hennequin c, G. Hornung a, J.L. Segui a, L. Vermare c, Ph. Ghendrih a, Tore Supra Team a a
CEA, IRFM, F-13108 Saint-Paul-lez-Durance, France Institut Jean Lamour, Université de Lorraine, Boulevard des Aiguillettes, BP 239, F-54506 Vandoeuvre-lès-Nancy Cedex, France c Ecole Polytechnique, LPP, CNRS UMR 7648, 91128 Palaiseau, France b
a r t i c l e
i n f o
Article history: Available online 11 January 2013
a b s t r a c t We report the results of an experimental investigation of the impact of Supersonic Molecular Beam Injection in the Tore Supra tokamak. Several diagnostics were synchronised with the injection to extract a global picture of the physics at play from the time scale of turbulence (10 ls) to the full-recovery time (1 s). As previously reported, a strong impact of the injection on density and temperature profiles is observed. Both fields exhibit a complex dynamic response involving different phases and time scales. In particular, we show that the effective particle fuelling efficiency is determined by a period of degraded confinement that follows the injection, during which the edge density collapses, in some cases, lower than the initial one. This phase is characterised by a dramatic change in the turbulent transport, with a drop of the frequency spectrum and the observation of large coherent structures as opposed to small intermittent fluctuations before the injection. Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction Particle fuelling is essential to control the density in tokamak plasmas. In ITER, a fuel ion flux to the core of up to 100 Pa m3 s1 will be required in steady state [1]. These particles will have to penetrate through the pedestal (Te = 3–5 keV at the top) without affecting the H-mode confinement. Three main methods of particle fuelling are routinely used in current machines: gas puffing, supersonic molecular beam injection (SMBI) and pellet injection (neutral beam injection can also contribute in a non negligible way to particle fuelling). The associated fuelling efficiencies, as found on Tore Supra, range from 5% to 10% for gas puffing up to 90% for pellet injection, through 40–50% for SMBI. In ITER, due to the expected pedestal conditions, these figures should be lower, so that only high-field side pellet injection is considered as being able to fulfil the required particle fuelling rate. The choice of high-field side with respect to low-field side injection is motivated by the experimental observation of a better fuelling efficiency, usually explained as resulting from drift effects on the ablation cloud surrounding the pellet [2]. Thus, the interaction between the injected particles and the local plasma transport appears as a major player in the physics at play during particle fuelling [3]. However, in the absence of reliable models for the particle transport and redistribution after an injection, extrapolations
⇑ Corresponding author. 1
E-mail address: [email protected] (P. Tamain). Presenting author.
0022-3115/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jnucmat.2013.01.023
to ITER remain speculative and can exhibit very different results depending on the chosen transport parameters [4]. We present here the results of an experimental investigation of the interaction between particle fuelling and plasma transport during SMBI in the Tore Supra tokamak. Particular effort was put into providing exhaustive diagnostic coverage to observe the interaction between the plasma and the injection. The experimental setup is described in Section 2. We then focus on the response of the plasma temperature and density to demonstrate the existence of a degraded confinement phase that determines the fuelling efficiency (Section 3). Turbulent-resolved measurements detailed in Section 4 back up this explanation by showing a dramatic change of the fluctuations characteristics during this period. These results are finally discussed and summarised in Section 5.
2. Experimental setup Tore Supra is equipped with three supersonic gas injectors, one located in the low-field side (LFS) mid-plane and the two others in the high-field side (HFS) one (see Fig. 1). They are able to inject gas at a Mach number of the order of 4 with a flow rate of 200 Pa m3 s1 for a duration of 2 ms, which represents an influx 5 1020 particles per pulse [5,6]. In this paper, we focus on results obtained with the LFS injector. The injections were performed in limited ohmic plasmas with HFS contact point on the inner bumpers. The toroidal floor limiter was 2 cm away from the Last Closed Flux Surface (LCFS) and the outboard poloidal limiter located in sector 3 was advanced 2–4 cm away from the LCFS in order to allow for pecker
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
S149
from one injection to the other in order to get highly-resolved measurements on longer time scales. In the remainder of this paper, the t = 0 time reference is, unless explicitly stated, the time tSMBI of the electric trigger to the supersonic injector valve. For LFS injection, a repeatable time delay of 23 ms is observed between the trigger and the first observable perturbation of the plasma. This is most likely due to the inertia of the injector and the flight time of neutrals from the nozzle to the plasma. 3. Impact of SMBI on the plasma confinement and consequences on the fuelling efficiency
Fig. 1. Top view of the tore supra tokamak showing the distribution of the diagnostics used. The angles refer to the poloidal/toroidal positions (poloidal: 0° = LFS mid-plane, increasing in electron diamagnetic direction; toroidal: 0° = plane S1/S6, increasing counter Ip). The three thick arrows are the 3 supersonic injectors, poloidally located in the mid-plane. APL is the outboard antenna protection limiter.
probes measurements up to the LCFS (see below). The plasma current Ip was varied from shot to shot from 0.8 MA to 1.5 MA while keeping the volume-averaged density around 2.4 1019 m3. Depending on Ip, two to three injections were performed per shot, separated by enough time to let the plasma density and temperature fully recover from one to the following. The diagnostic coverage used in this study is shown in Fig. 1. These diagnostics are the following: ECE system: electron temperature measurements were acquired with the Tore Supra 32-channel heterodyne electron cyclotron emission radiometer [7,8]. Fast acquisitions at 1 GHz were synchronised with the SMBIs. Far infrared interferometer: the interferometer is located in sector 3. It comprises 10 chords distributed in the poloidal plane, as represented in Fig. 5. The standard time resolution is 1 ms but fast acquisition windows with a time resolution of 10 ls were synchronised with the injection to capture the fast dynamics [8]. Ultrafast sweeping reflectometer: this profile reflectometer is located in sector 3 just beside the outboard limiter housing the pecker probes. It can measure density profiles in the edge of the plasma with a time resolution up to 3 ls per profile for a time window of 10,000 profiles [9]. Fast visible imaging: the Tore Supra fast visible camera [10,11] was used to monitor edge plasma fluctuations. In order to get sufficiently time-resolved movies, a resolution of 128 160 was chosen, allowing for a frame rate of 57 kHz with 10 ls exposure time. Pecker probes: Tore Supra has recently been equipped with reciprocating probes located in the outboard poloidal limiter in the vicinity of the LFS mid-plane [12], where the radial turbulent flux was shown to be ballooning in character [11]. The system is composed of two magnetically driven manipulators, each of which can insert up to 3 cm in front of the limiter two tunnel probes [13] positioned back to back in a Mach probe configuration. These allow the measurement of the ion saturation current jsat at a sampling rate of 1 MHz, as well as an insight on the electron temperature Te and on the parallel velocity of the plasma. The total duration of a plunge is of the order of 100 ms. For technical reasons, only the top probe, located 19° above the equatorial plane, could be used in the considered discharges. Particular effort was put in synchronising the data acquisition windows with the injections. The exact synchronisation was varied
Let us first focus on the global response of the plasma. One of the most widely reported impacts of a supersonic injection is a drop in the temperature of the plasma. This effect is believed to be fundamental in explaining the increased efficiency of SMBI with respect to gas puffing, since a cold edge plasma both increases the parallel loss time of particles to the wall and favours the neutrals penetration [4]. Such behaviour is found also in Tore Supra, as shown in Fig. 2 which gives the time traces of the electron temperature measured by ECE at radii varying from the LCFS to the centre of the plasma in a shot at Ip = 1.2 MA. A sudden cooling of the edge plasma from 200 eV to 120 eV on a time scale of 4 ms is observed, propagating up to the centre where the temperature decreases from 2.1 keV down to 1.7 keV on a longer time scale (of the order of 100 ms). These cold conditions remain for a period of 40–60 ms on the outermost ECE channels (RLCFSR < 20 cm) before the temperature starts relaxing back to the initial one. For the central channels as well as the outermost one, this relaxation occurs on a slow and unique time scale of the order of 0.8 s (about 3 times the pre-SMBI energy confinement time). For intermediate channels (2.5 cm < RLCFSR < 20 cm), this slow relaxation is preceded by a faster temporary rise. The corresponding characteristic time depends on the considered radius, ranging from 55 ms 2.7 cm inside the LCFS to more than 200 ms 20 cm inside the plasma. The current scan showed that the plasma current Ip has a strong influence on these dynamics, the duration of the intermediate cold phase as well as the penetration depth of the cold front increasing with decreasing Ip. The density also exhibits distinct phases in its response to the injection. Fig. 3 shows the time traces of the volume-averaged density measured by the interferometer after each LFS SMBI pulse in our database. The volume-averaged density is calculated from the density profile which is itself reconstructed from the 10 chords of the interferometer using a least-mean square method and
Fig. 2. Electron temperature time traces measured by 28 channels (28 radial positions; two lines of each colour corresponding to the 2 different radial positions indicated in the right-hand side legend) of the ECE radiometer during a supersonic injection in a discharge with Ip = 1.2 MA plasma current.
S150
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
Fig. 3. Volume-averaged density traces as a function of time for SMBIs at various plasma currents. The right panels are a magnified version of the left ones. The four response phases identified in the text are indicated approximately (they may differ slightly for each SMBI).
assuming that the density is constant along each flux surface. The data are repeatable for each given Ip value and four distinct phases can systematically be observed. In the first phase (phase 1), the density sharply rises on a time scale of 4 ± 2 ms to reach about the double of the initial density at time t12. A rapid decay phase (phase 2) follows, during which the density signal is intermittent, until a time t23 when the density reaches a local minimum n23 in almost all cases (3 out of 8 of the Ip = 1.5 MA SMBIs just showed
a flattening of the curve, but no local minimum). In a third phase (phase 3), the density trace gets smooth and starts rising again, or at least remained flat in the 3 cases when no local minimum was observed. A local maximum is then reached at a time t34 and the density decays slowly back to the initial one with a decay time of the order of 0.5 s (phase 4). A dramatic impact of the plasma current Ip can be observed on the relative importance of the various phases. This is evidenced in Fig. 4 where the transition times tij between the four phases as well as the corresponding densities are plotted as a function of Ip. If phases 1 and 4 do not seem to depend on Ip, the durations of phases 2 and 3 increase for lower Ip. As a consequence, the depth of the local minimum at t23 is lower the lower Ip is, as well as the local maximum reached at t34. The density reached at t23 is even 10% lower than the pre-SMBI one for the Ip = 0.8 MA shots. More details on the behaviour of the different interferometer chords are given in Fig. 5. Although the exact times at which the transitions from one phase to the other occur can vary by up to 20 ms from one chord to the other, the existence of phases in the density response to the injection can be observed on most of the chords. The succession of phases 2, 3 and 4 is particularly visible on the tangential edge chord, which probes the Scrape-Off Layer (SOL) and the edge plasma up to 4 cm inside the LCFS. The lineintegrated density measured on this chord at the end of phase 2 is even lower than the initial one by up to 50% for all the discharges at Ip 6 1.2 MA. It is also interesting to note that all the chords do not show exactly the same behaviour, suggesting a poloidal asymmetry of the density response. For the first 4 ms (phase 1), all the chords rise more or less at the same rate (2 1021 m2 s1). It is during phase 2 that they mainly differ. While horizontal chords measure an immediate flattening of the line-averaged density evolution (it still increases but at a much slower rate 4 1019 m2 s1 until the end of phase 2), vertical chords measure a marked overshoot of the signal, peaking about 25% higher than the lineaveraged density observed at the end of phase 1 before dropping in a very intermittent manner until the end of phase 2. In low Ip shots (Ip 6 1 MA), horizontal chords exhibit the same kind of overshoot, but at a later time in the middle of phase 2 suggesting some propagation from the zones of the plasma probed by the vertical chords and those probed by the horizontal ones. After this overshoot phase, the density rises much slower and smoothly until each chord reaches a maximum around the end of phase 3 (the maxima do not occur precisely at the same time for each chord). The interpretation of this poloidally asymmetric behaviour is not straightforward. One may speculate that the toroidal limiter,
Fig. 4. Left: dependence on the plasma current Ip of the transition times between the four response phases observed on the volume-averaged density signal. Right: corresponding volume-averaged density variation with respect to pre-SMBI density. The indices for the time and density refer to the phases described in the text. For example, t12 and n12 are respectively the time and the density at which the transition between phase 1 and phase 2 occurs.
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
S151
Fig. 5. Time traces of the line-integrated density from the 10 chords of the interferometer for four supersonic injections at different plasma currents. The positions of the various chords are represented in the left hand-side poloidal plot which also gives the Last Closed Flux Surface and the positions of plasma facing components. The chord with a line-integrated density lower than the others is the edge tangential one, numbered 10 in the left plot. The dashed lines represent the transition times between the four response phases identified in the text and the boxed sub-graphs are magnifications.
located at the bottom of the machine just 1 cm away from the LCFS in the considered shots, influences transport processes in the edge, both through the boundary conditions it imposes to the plasma and through the recycling of particles around it. Further experiments with smaller plasmas or plasmas shifted upward would be necessary to test this hypothesis. Insight on the radial dynamics of the density relaxation is given by fast reflectometry measurements. Fig. 6 shows the plasma response to a supersonic injection in an Ip = 1.2 MA discharge. The left hand side plot shows successive profiles. However, for a reason yet not identified, a large error bar of the order of 5 cm in the absolute position of every subsequent profile (solid shift of the profile by ±2.5 cm) was observed in the considered shots so that it is more convincing to look at the gradient lengths. The right-hand side figure hence shows, as a function of time, the radial distance between iso-density lines and a reference density in the SOL, i.e. the distance between the curves is proportional to the gradient lengths. The first observable impact of the injection is visible at t tSMBI = 23 ms. Once again, the observed phenomenology is consistent with the existence of the four relaxation phases observed with the interferometer. For the first 4 ms, a large density gradient builds up in
the SOL and propagates inwards. One then observes a relaxation of the profiles, during which the SOL first extends radially outwards before emptying itself while the gradient further inside decreases. At t tSMBI 60 ms, corresponding to the end of phase 2, a transition occurs and the gradient starts increasing in the vicinity of the LCFS (associated with an increase of the density) while it carries on decreasing further inside until the end of phase 3. Phase 4 then corresponds to a steady and uniform relaxation of the profile. This way, ECE, interferometry and reflectometry data consistently show a response of the plasma involving four distinct phases. A way to explain these observations is to interpret phase 2 as a temporary phase of degraded confinement triggered by the SMBI. Confinement progressively goes back to pre-injection value during phase 3 and the density rises in the edge plasma due to strong recycling from the walls. Phase 4 is then simply the slow decay of the density back to the equilibrium one at the rate of the particle confinement time. We can hence define the effective fuelling efficiency g of the supersonic injection according to the density reached at the end of phase 3. Considering an influx of 5 1020 particles per SMBI and a plasma volume of 23 m3, g varies from 0% to 40% in the studied
Fig. 6. Fast reflectometry data after a supersonic injection at Ip = 1.2 MA. Left: successive density profiles in the edge plasma. Right: radial position of iso-density lines as a function of time, plotted with respect to the position of the n0e ¼ 0:3 1019 m3 density; i.e. DRðne ; tÞ ¼ Rðne ; tÞ Rðn0e ; tÞ. The time intervals corresponding to the four response phases identified on the interferometry data are indicated by vertical dashed lines.
S152
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
Fig. 7. Frames of the fast visible imaging system during a SMBI at Ip = 1.5 MA (shot #47760). Time labels are with respect to the SMBI trigger time. For a better visibility, the colour scales are different for each frame, normalised to the brightest pixel. The left plot shows the time evolution of the luminosity of the brightest pixel, with the dashed vertical lines indicating the times of the frames. The black horizontal stripe is due to several rows of dead pixels on the sensor.
Fig. 8. Time traces of the ion saturation current measured by the pecker probe during a supersonic injection in a shot at 1.2 MA plasma current (#47766). (a) Overall time traces for two plunges, one centred at t tSMBI = 30 ms, the other at t tSMBI = 110 ms. The smooth curves give the radial position of the probe, RLFS being the projection along flux surface in the low field side mid-plane. (b–d) Zooms on the fluctuations before the injection, during and after. Note that the scales (time and current) of plots (b and d) are the same. The y-scale of plot (c) is half of that of plots (b and d). LFS (HFS) refers to the tunnel collector magnetically connected to the Low (High) Field Side mid-plane.
shots, increasing with Ip. Since this dependence of g with Ip is correlated with an increasing importance of the degraded confinement phase, understanding the fuelling efficiency therefore requires one to understand the impact of the SMBI on the plasma transport. In the following, we analyse closer the first three phases of the relaxation to isolate the mechanisms at play.
4. Impact of SMBI on turbulence A clear correlation can be established between the various phases of the plasma response and the phenomenology observed
with the diagnostics focusing on turbulence. This is in particular the case for the movies recorded by the fast visible imaging system. Fig. 7 shows snapshots of the visible emission of the plasma just after a supersonic injection. The observations are very similar for different plasma currents. Due to the limited sensitivity of the sensor combined with the short exposure time imposed by the frame rate necessary to capture fluctuations, only phase 1 and the very beginning of phase 2 provide enough light intensity to be observed. At t tSMBI = 23 ms, the luminosity measured by the camera increases extremely fast (on a time scale of the order of 60 ls) and peaks for 2.5 ms before decaying exponentially with a fall time of 1 ms. During the first millisecond of the peak, the plasma exhibits
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
S153
auto-correlation half-width increases from 33 ls before the injection to 110 ls 27 ms after. The normalised cross-correlation between the signals measured by the two sides of the pecker probe also exhibits important modifications (Fig. 9): peaking at 0.33 with a half-width of 52 ls before the injection, it increases strongly just after the injection to reach a maximum of 0.78 with a half-width of 130 ls. Data measured during a second probe plunge delayed with respect to the injection time (Fig. 8 and plots at t tSMBI = 80 and 130 ms in Fig. 9) shows that the fluctuations characteristics go back to normal 60–70 ms after the injection, hence at the middle of phase 3. Such behaviour supports the observations made with the fast camera by suggesting an increase of the size of the turbulent structures and hence a modification of the basic properties of the turbulent transport coincident with the second phase of the density response. 5. Conclusion Fig. 9. Temporal correlation functions of the ion saturation current signals measured by the pecker probe around times t tSMBI = 0, 50, 80 and 130 ms in time windows of 10 ms width in a shot at 1.2 MA plasma current (#47766). The initial data are the same as those plotted in Fig. 8 and the times were chosen such that the probe was at the same radial location (RLFS = 3.005 m). Top: autocorrelation of the signal on the side of the probe pointing towards the low field side mid-plane. Bottom: cross-correlation between the signals on each side of the probe.
filamentary (aligned along the magnetic field) fluctuations with small relative amplitude of the order of 10%. Their poloidal correlation length is 5 pixels, which corresponds roughly to 1 cm. In the following 1.5 ms, larger amplitude fluctuations (still fieldaligned) progressively appear in the radiative layer. A rapid relaxation then occurs, with the turbulent structures propagating outwards, leading to a radial expansion of the radiative zone into the outer region in 1–2 ms. The time at which this relaxation occurs corresponds to the transition time between phases 1 and 2. The extended radiative layer then smears out and one can observe remnant turbulent structures at the initial location. However, the characteristics of these structures are dramatically different from the initial ones. They have a relative fluctuation level of 20– 40%, a poloidal correlation length of 11 pixels (2 cm) and are quasi-coherent (the cross-correlation function between 2 poloidally separated points exhibits oscillations). This way, fast visible camera measurements suggest a strong change in the turbulence characteristics at the transition from phase 1 to phase 2. Such observation is backed-up by tunnel probes data, whether around the outboard mid-plane with the pecker probes or at the top of the machine with the reciprocating probe. As shown in Fig. 8, the SMBI leads to an increase by 200% of the fluctuationaveraged ion saturation current hjsati in a 4 ms time. This peak relaxes rapidly to recover a level of hjsati about 60% higher than before the injection. Although it is impossible to decouple the time evolution from the spatial one due to the movement of the probe, one can notice that the decay of hjsati on the way back of the probe is much quicker than the increase on the way in, which suggests either a steepening of the edge radial decay or a rapid drop of the density with time. Zooming on smaller time scales, one notices a dramatic impact of the injection on the characteristics of fluctuations. The fast and extremely intermittent signal observed before the injection is replaced by much slower and coherent bursts. The fluctuation level collapses from 30% down to less than 10% for the side of the probe magnetically connected to the LFS mid-plane, and from 15% to 7% for the side connected to the HFS. The modification of the temporal characteristics of the fluctuations can be quantified by looking at the auto-correlation of the ion saturation current signal (Fig. 9). At the same radial location (3 cm outside the LCFS), the
We have investigated the impact on the plasma of Supersonic Molecular Beam Injection (SMBI) using extended diagnostics coverage. Four distinct phases are systematically observed in the plasma response. After a first peak of 4 ms (phase 1), the volumeaveraged density decays rapidly with an intermittent behaviour, until it reaches a local minimum (phase 2). A temporary increase follows, with a much smoother behaviour, until the density reaches a local maximum, lower than the initial peak (phase 3). In a fourth phase, the density relaxes back to the initial one on a much longer time scale of the order of the particle confinement time. The fuelling efficiency g can hence be defined with respect to the density reached at the end of the third phase. Plasma currents scans show that g drops when decreasing Ip, which is correlated to an increase of the duration of the second response phase. We interpret the latter as a temporary period of degraded confinement triggered by the injection. Such conclusion is backed-up by fast camera and Langmuir probe measurements that both show a dramatic modification of the properties of turbulence after the supersonic injection, with slower dynamics and larger coherent structures. Note that Ip was used in these experiments as the scanning parameter, but one cannot exclude a possible dependence on the density (or altogether a dependence on the Greenwald fraction) which would require more data to be demonstrated. These results shed new light on the understanding of the fuelling efficiency by SMBI but also potentially pellets. They demonstrate that a poloidally localised particle injection can have a strong impact on local transport properties which, in feed-back, impacts the efficiency of the particle deposition. Such a strong non-linear plasma response cannot be captured by diagnostics or simulations set for quasi-steady state analysis so that the understanding of particle fuelling mechanisms requires tools addressing transport issues consistently. We also want to draw the reader’s attention on the fact that local gas injections, often used to increase the contrast of fast visible imaging measurements, are susceptible to change the properties of turbulence so that great caution must be taken in the interpretation of such experiments [14]. Acknowledgements The authors wish to thank N. Fedorczak for fruitful discussions concerning the interpretation of the data presented here. This work, supported by the European Communities under the contract of Association between EURATOM and CEA, was carried out within the framework of the European Fusion Development Agreement. The views and opinions expressed herein do not necessarily reflect those of the European Commission.
S154
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
References [1] [2] [3] [4] [5]
G.W. Pacher et al., Nucl. Fusion 48 (2008) 105003. P.T. Lang et al., Phys. Rev. Lett. 79 (1997) 1487–1490. V. Rozhansky et al., Nucl. Fusion 46 (2006) 367. P. Tamain et al., J. Nucl. Mater. 363–365 (2007) 844–848. J. Bucalossi et al., Particle fueling for long pulse with standard gas puff and supersonic pulsed gas injection, in: Proceedings of 19th IAEA Fusion Energy Conferences, Lyon, October 2002, EX/P4-04.
[6] [7] [8] [9] [10] [11] [12] [13] [14]
B. Pégourié et al., J. Nucl. Mater. 313–316 (2003) 539–542. J.L. Ségui et al., Rev. Sci. Instrum. 76 (2005) 123501. C. Gil et al., Fusion Sci. Technol. 56 (2009) 1219. F. Clairet et al., Rev. Sci. Instrum. 81 (2010) 10D903. A. Géraud et al., Rev. Sci. Instrum. 80 (2009) 033504. N. Fedorczak et al., J. Nucl. Mater. 415 (2011) S467–S470. J.P. Gunn, J.-Y. Pascal, Rev. Sci. Instrum. 82 (2011) 123505. J.P. Gunn et al., Rev. Sci. Instrum. 75 (2004) 4328. Y. Marrandet et al., This conference, 2012.
Contents lists available at SciVerse ScienceDirect
Journal of Nuclear Materials journal homepage: www.elsevier.com/locate/jnucmat
Interaction of plasma transport and turbulence on particle fuelling P. Tamain a,⇑,1, G. Bonhomme b, F. Brochard b, F. Clairet a, C. Gil a, J. Gunn a, P. Hennequin c, G. Hornung a, J.L. Segui a, L. Vermare c, Ph. Ghendrih a, Tore Supra Team a a
CEA, IRFM, F-13108 Saint-Paul-lez-Durance, France Institut Jean Lamour, Université de Lorraine, Boulevard des Aiguillettes, BP 239, F-54506 Vandoeuvre-lès-Nancy Cedex, France c Ecole Polytechnique, LPP, CNRS UMR 7648, 91128 Palaiseau, France b
a r t i c l e
i n f o
Article history: Available online 11 January 2013
a b s t r a c t We report the results of an experimental investigation of the impact of Supersonic Molecular Beam Injection in the Tore Supra tokamak. Several diagnostics were synchronised with the injection to extract a global picture of the physics at play from the time scale of turbulence (10 ls) to the full-recovery time (1 s). As previously reported, a strong impact of the injection on density and temperature profiles is observed. Both fields exhibit a complex dynamic response involving different phases and time scales. In particular, we show that the effective particle fuelling efficiency is determined by a period of degraded confinement that follows the injection, during which the edge density collapses, in some cases, lower than the initial one. This phase is characterised by a dramatic change in the turbulent transport, with a drop of the frequency spectrum and the observation of large coherent structures as opposed to small intermittent fluctuations before the injection. Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction Particle fuelling is essential to control the density in tokamak plasmas. In ITER, a fuel ion flux to the core of up to 100 Pa m3 s1 will be required in steady state [1]. These particles will have to penetrate through the pedestal (Te = 3–5 keV at the top) without affecting the H-mode confinement. Three main methods of particle fuelling are routinely used in current machines: gas puffing, supersonic molecular beam injection (SMBI) and pellet injection (neutral beam injection can also contribute in a non negligible way to particle fuelling). The associated fuelling efficiencies, as found on Tore Supra, range from 5% to 10% for gas puffing up to 90% for pellet injection, through 40–50% for SMBI. In ITER, due to the expected pedestal conditions, these figures should be lower, so that only high-field side pellet injection is considered as being able to fulfil the required particle fuelling rate. The choice of high-field side with respect to low-field side injection is motivated by the experimental observation of a better fuelling efficiency, usually explained as resulting from drift effects on the ablation cloud surrounding the pellet [2]. Thus, the interaction between the injected particles and the local plasma transport appears as a major player in the physics at play during particle fuelling [3]. However, in the absence of reliable models for the particle transport and redistribution after an injection, extrapolations
⇑ Corresponding author. 1
E-mail address: [email protected] (P. Tamain). Presenting author.
0022-3115/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jnucmat.2013.01.023
to ITER remain speculative and can exhibit very different results depending on the chosen transport parameters [4]. We present here the results of an experimental investigation of the interaction between particle fuelling and plasma transport during SMBI in the Tore Supra tokamak. Particular effort was put into providing exhaustive diagnostic coverage to observe the interaction between the plasma and the injection. The experimental setup is described in Section 2. We then focus on the response of the plasma temperature and density to demonstrate the existence of a degraded confinement phase that determines the fuelling efficiency (Section 3). Turbulent-resolved measurements detailed in Section 4 back up this explanation by showing a dramatic change of the fluctuations characteristics during this period. These results are finally discussed and summarised in Section 5.
2. Experimental setup Tore Supra is equipped with three supersonic gas injectors, one located in the low-field side (LFS) mid-plane and the two others in the high-field side (HFS) one (see Fig. 1). They are able to inject gas at a Mach number of the order of 4 with a flow rate of 200 Pa m3 s1 for a duration of 2 ms, which represents an influx 5 1020 particles per pulse [5,6]. In this paper, we focus on results obtained with the LFS injector. The injections were performed in limited ohmic plasmas with HFS contact point on the inner bumpers. The toroidal floor limiter was 2 cm away from the Last Closed Flux Surface (LCFS) and the outboard poloidal limiter located in sector 3 was advanced 2–4 cm away from the LCFS in order to allow for pecker
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
S149
from one injection to the other in order to get highly-resolved measurements on longer time scales. In the remainder of this paper, the t = 0 time reference is, unless explicitly stated, the time tSMBI of the electric trigger to the supersonic injector valve. For LFS injection, a repeatable time delay of 23 ms is observed between the trigger and the first observable perturbation of the plasma. This is most likely due to the inertia of the injector and the flight time of neutrals from the nozzle to the plasma. 3. Impact of SMBI on the plasma confinement and consequences on the fuelling efficiency
Fig. 1. Top view of the tore supra tokamak showing the distribution of the diagnostics used. The angles refer to the poloidal/toroidal positions (poloidal: 0° = LFS mid-plane, increasing in electron diamagnetic direction; toroidal: 0° = plane S1/S6, increasing counter Ip). The three thick arrows are the 3 supersonic injectors, poloidally located in the mid-plane. APL is the outboard antenna protection limiter.
probes measurements up to the LCFS (see below). The plasma current Ip was varied from shot to shot from 0.8 MA to 1.5 MA while keeping the volume-averaged density around 2.4 1019 m3. Depending on Ip, two to three injections were performed per shot, separated by enough time to let the plasma density and temperature fully recover from one to the following. The diagnostic coverage used in this study is shown in Fig. 1. These diagnostics are the following: ECE system: electron temperature measurements were acquired with the Tore Supra 32-channel heterodyne electron cyclotron emission radiometer [7,8]. Fast acquisitions at 1 GHz were synchronised with the SMBIs. Far infrared interferometer: the interferometer is located in sector 3. It comprises 10 chords distributed in the poloidal plane, as represented in Fig. 5. The standard time resolution is 1 ms but fast acquisition windows with a time resolution of 10 ls were synchronised with the injection to capture the fast dynamics [8]. Ultrafast sweeping reflectometer: this profile reflectometer is located in sector 3 just beside the outboard limiter housing the pecker probes. It can measure density profiles in the edge of the plasma with a time resolution up to 3 ls per profile for a time window of 10,000 profiles [9]. Fast visible imaging: the Tore Supra fast visible camera [10,11] was used to monitor edge plasma fluctuations. In order to get sufficiently time-resolved movies, a resolution of 128 160 was chosen, allowing for a frame rate of 57 kHz with 10 ls exposure time. Pecker probes: Tore Supra has recently been equipped with reciprocating probes located in the outboard poloidal limiter in the vicinity of the LFS mid-plane [12], where the radial turbulent flux was shown to be ballooning in character [11]. The system is composed of two magnetically driven manipulators, each of which can insert up to 3 cm in front of the limiter two tunnel probes [13] positioned back to back in a Mach probe configuration. These allow the measurement of the ion saturation current jsat at a sampling rate of 1 MHz, as well as an insight on the electron temperature Te and on the parallel velocity of the plasma. The total duration of a plunge is of the order of 100 ms. For technical reasons, only the top probe, located 19° above the equatorial plane, could be used in the considered discharges. Particular effort was put in synchronising the data acquisition windows with the injections. The exact synchronisation was varied
Let us first focus on the global response of the plasma. One of the most widely reported impacts of a supersonic injection is a drop in the temperature of the plasma. This effect is believed to be fundamental in explaining the increased efficiency of SMBI with respect to gas puffing, since a cold edge plasma both increases the parallel loss time of particles to the wall and favours the neutrals penetration [4]. Such behaviour is found also in Tore Supra, as shown in Fig. 2 which gives the time traces of the electron temperature measured by ECE at radii varying from the LCFS to the centre of the plasma in a shot at Ip = 1.2 MA. A sudden cooling of the edge plasma from 200 eV to 120 eV on a time scale of 4 ms is observed, propagating up to the centre where the temperature decreases from 2.1 keV down to 1.7 keV on a longer time scale (of the order of 100 ms). These cold conditions remain for a period of 40–60 ms on the outermost ECE channels (RLCFSR < 20 cm) before the temperature starts relaxing back to the initial one. For the central channels as well as the outermost one, this relaxation occurs on a slow and unique time scale of the order of 0.8 s (about 3 times the pre-SMBI energy confinement time). For intermediate channels (2.5 cm < RLCFSR < 20 cm), this slow relaxation is preceded by a faster temporary rise. The corresponding characteristic time depends on the considered radius, ranging from 55 ms 2.7 cm inside the LCFS to more than 200 ms 20 cm inside the plasma. The current scan showed that the plasma current Ip has a strong influence on these dynamics, the duration of the intermediate cold phase as well as the penetration depth of the cold front increasing with decreasing Ip. The density also exhibits distinct phases in its response to the injection. Fig. 3 shows the time traces of the volume-averaged density measured by the interferometer after each LFS SMBI pulse in our database. The volume-averaged density is calculated from the density profile which is itself reconstructed from the 10 chords of the interferometer using a least-mean square method and
Fig. 2. Electron temperature time traces measured by 28 channels (28 radial positions; two lines of each colour corresponding to the 2 different radial positions indicated in the right-hand side legend) of the ECE radiometer during a supersonic injection in a discharge with Ip = 1.2 MA plasma current.
S150
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
Fig. 3. Volume-averaged density traces as a function of time for SMBIs at various plasma currents. The right panels are a magnified version of the left ones. The four response phases identified in the text are indicated approximately (they may differ slightly for each SMBI).
assuming that the density is constant along each flux surface. The data are repeatable for each given Ip value and four distinct phases can systematically be observed. In the first phase (phase 1), the density sharply rises on a time scale of 4 ± 2 ms to reach about the double of the initial density at time t12. A rapid decay phase (phase 2) follows, during which the density signal is intermittent, until a time t23 when the density reaches a local minimum n23 in almost all cases (3 out of 8 of the Ip = 1.5 MA SMBIs just showed
a flattening of the curve, but no local minimum). In a third phase (phase 3), the density trace gets smooth and starts rising again, or at least remained flat in the 3 cases when no local minimum was observed. A local maximum is then reached at a time t34 and the density decays slowly back to the initial one with a decay time of the order of 0.5 s (phase 4). A dramatic impact of the plasma current Ip can be observed on the relative importance of the various phases. This is evidenced in Fig. 4 where the transition times tij between the four phases as well as the corresponding densities are plotted as a function of Ip. If phases 1 and 4 do not seem to depend on Ip, the durations of phases 2 and 3 increase for lower Ip. As a consequence, the depth of the local minimum at t23 is lower the lower Ip is, as well as the local maximum reached at t34. The density reached at t23 is even 10% lower than the pre-SMBI one for the Ip = 0.8 MA shots. More details on the behaviour of the different interferometer chords are given in Fig. 5. Although the exact times at which the transitions from one phase to the other occur can vary by up to 20 ms from one chord to the other, the existence of phases in the density response to the injection can be observed on most of the chords. The succession of phases 2, 3 and 4 is particularly visible on the tangential edge chord, which probes the Scrape-Off Layer (SOL) and the edge plasma up to 4 cm inside the LCFS. The lineintegrated density measured on this chord at the end of phase 2 is even lower than the initial one by up to 50% for all the discharges at Ip 6 1.2 MA. It is also interesting to note that all the chords do not show exactly the same behaviour, suggesting a poloidal asymmetry of the density response. For the first 4 ms (phase 1), all the chords rise more or less at the same rate (2 1021 m2 s1). It is during phase 2 that they mainly differ. While horizontal chords measure an immediate flattening of the line-averaged density evolution (it still increases but at a much slower rate 4 1019 m2 s1 until the end of phase 2), vertical chords measure a marked overshoot of the signal, peaking about 25% higher than the lineaveraged density observed at the end of phase 1 before dropping in a very intermittent manner until the end of phase 2. In low Ip shots (Ip 6 1 MA), horizontal chords exhibit the same kind of overshoot, but at a later time in the middle of phase 2 suggesting some propagation from the zones of the plasma probed by the vertical chords and those probed by the horizontal ones. After this overshoot phase, the density rises much slower and smoothly until each chord reaches a maximum around the end of phase 3 (the maxima do not occur precisely at the same time for each chord). The interpretation of this poloidally asymmetric behaviour is not straightforward. One may speculate that the toroidal limiter,
Fig. 4. Left: dependence on the plasma current Ip of the transition times between the four response phases observed on the volume-averaged density signal. Right: corresponding volume-averaged density variation with respect to pre-SMBI density. The indices for the time and density refer to the phases described in the text. For example, t12 and n12 are respectively the time and the density at which the transition between phase 1 and phase 2 occurs.
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
S151
Fig. 5. Time traces of the line-integrated density from the 10 chords of the interferometer for four supersonic injections at different plasma currents. The positions of the various chords are represented in the left hand-side poloidal plot which also gives the Last Closed Flux Surface and the positions of plasma facing components. The chord with a line-integrated density lower than the others is the edge tangential one, numbered 10 in the left plot. The dashed lines represent the transition times between the four response phases identified in the text and the boxed sub-graphs are magnifications.
located at the bottom of the machine just 1 cm away from the LCFS in the considered shots, influences transport processes in the edge, both through the boundary conditions it imposes to the plasma and through the recycling of particles around it. Further experiments with smaller plasmas or plasmas shifted upward would be necessary to test this hypothesis. Insight on the radial dynamics of the density relaxation is given by fast reflectometry measurements. Fig. 6 shows the plasma response to a supersonic injection in an Ip = 1.2 MA discharge. The left hand side plot shows successive profiles. However, for a reason yet not identified, a large error bar of the order of 5 cm in the absolute position of every subsequent profile (solid shift of the profile by ±2.5 cm) was observed in the considered shots so that it is more convincing to look at the gradient lengths. The right-hand side figure hence shows, as a function of time, the radial distance between iso-density lines and a reference density in the SOL, i.e. the distance between the curves is proportional to the gradient lengths. The first observable impact of the injection is visible at t tSMBI = 23 ms. Once again, the observed phenomenology is consistent with the existence of the four relaxation phases observed with the interferometer. For the first 4 ms, a large density gradient builds up in
the SOL and propagates inwards. One then observes a relaxation of the profiles, during which the SOL first extends radially outwards before emptying itself while the gradient further inside decreases. At t tSMBI 60 ms, corresponding to the end of phase 2, a transition occurs and the gradient starts increasing in the vicinity of the LCFS (associated with an increase of the density) while it carries on decreasing further inside until the end of phase 3. Phase 4 then corresponds to a steady and uniform relaxation of the profile. This way, ECE, interferometry and reflectometry data consistently show a response of the plasma involving four distinct phases. A way to explain these observations is to interpret phase 2 as a temporary phase of degraded confinement triggered by the SMBI. Confinement progressively goes back to pre-injection value during phase 3 and the density rises in the edge plasma due to strong recycling from the walls. Phase 4 is then simply the slow decay of the density back to the equilibrium one at the rate of the particle confinement time. We can hence define the effective fuelling efficiency g of the supersonic injection according to the density reached at the end of phase 3. Considering an influx of 5 1020 particles per SMBI and a plasma volume of 23 m3, g varies from 0% to 40% in the studied
Fig. 6. Fast reflectometry data after a supersonic injection at Ip = 1.2 MA. Left: successive density profiles in the edge plasma. Right: radial position of iso-density lines as a function of time, plotted with respect to the position of the n0e ¼ 0:3 1019 m3 density; i.e. DRðne ; tÞ ¼ Rðne ; tÞ Rðn0e ; tÞ. The time intervals corresponding to the four response phases identified on the interferometry data are indicated by vertical dashed lines.
S152
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
Fig. 7. Frames of the fast visible imaging system during a SMBI at Ip = 1.5 MA (shot #47760). Time labels are with respect to the SMBI trigger time. For a better visibility, the colour scales are different for each frame, normalised to the brightest pixel. The left plot shows the time evolution of the luminosity of the brightest pixel, with the dashed vertical lines indicating the times of the frames. The black horizontal stripe is due to several rows of dead pixels on the sensor.
Fig. 8. Time traces of the ion saturation current measured by the pecker probe during a supersonic injection in a shot at 1.2 MA plasma current (#47766). (a) Overall time traces for two plunges, one centred at t tSMBI = 30 ms, the other at t tSMBI = 110 ms. The smooth curves give the radial position of the probe, RLFS being the projection along flux surface in the low field side mid-plane. (b–d) Zooms on the fluctuations before the injection, during and after. Note that the scales (time and current) of plots (b and d) are the same. The y-scale of plot (c) is half of that of plots (b and d). LFS (HFS) refers to the tunnel collector magnetically connected to the Low (High) Field Side mid-plane.
shots, increasing with Ip. Since this dependence of g with Ip is correlated with an increasing importance of the degraded confinement phase, understanding the fuelling efficiency therefore requires one to understand the impact of the SMBI on the plasma transport. In the following, we analyse closer the first three phases of the relaxation to isolate the mechanisms at play.
4. Impact of SMBI on turbulence A clear correlation can be established between the various phases of the plasma response and the phenomenology observed
with the diagnostics focusing on turbulence. This is in particular the case for the movies recorded by the fast visible imaging system. Fig. 7 shows snapshots of the visible emission of the plasma just after a supersonic injection. The observations are very similar for different plasma currents. Due to the limited sensitivity of the sensor combined with the short exposure time imposed by the frame rate necessary to capture fluctuations, only phase 1 and the very beginning of phase 2 provide enough light intensity to be observed. At t tSMBI = 23 ms, the luminosity measured by the camera increases extremely fast (on a time scale of the order of 60 ls) and peaks for 2.5 ms before decaying exponentially with a fall time of 1 ms. During the first millisecond of the peak, the plasma exhibits
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
S153
auto-correlation half-width increases from 33 ls before the injection to 110 ls 27 ms after. The normalised cross-correlation between the signals measured by the two sides of the pecker probe also exhibits important modifications (Fig. 9): peaking at 0.33 with a half-width of 52 ls before the injection, it increases strongly just after the injection to reach a maximum of 0.78 with a half-width of 130 ls. Data measured during a second probe plunge delayed with respect to the injection time (Fig. 8 and plots at t tSMBI = 80 and 130 ms in Fig. 9) shows that the fluctuations characteristics go back to normal 60–70 ms after the injection, hence at the middle of phase 3. Such behaviour supports the observations made with the fast camera by suggesting an increase of the size of the turbulent structures and hence a modification of the basic properties of the turbulent transport coincident with the second phase of the density response. 5. Conclusion Fig. 9. Temporal correlation functions of the ion saturation current signals measured by the pecker probe around times t tSMBI = 0, 50, 80 and 130 ms in time windows of 10 ms width in a shot at 1.2 MA plasma current (#47766). The initial data are the same as those plotted in Fig. 8 and the times were chosen such that the probe was at the same radial location (RLFS = 3.005 m). Top: autocorrelation of the signal on the side of the probe pointing towards the low field side mid-plane. Bottom: cross-correlation between the signals on each side of the probe.
filamentary (aligned along the magnetic field) fluctuations with small relative amplitude of the order of 10%. Their poloidal correlation length is 5 pixels, which corresponds roughly to 1 cm. In the following 1.5 ms, larger amplitude fluctuations (still fieldaligned) progressively appear in the radiative layer. A rapid relaxation then occurs, with the turbulent structures propagating outwards, leading to a radial expansion of the radiative zone into the outer region in 1–2 ms. The time at which this relaxation occurs corresponds to the transition time between phases 1 and 2. The extended radiative layer then smears out and one can observe remnant turbulent structures at the initial location. However, the characteristics of these structures are dramatically different from the initial ones. They have a relative fluctuation level of 20– 40%, a poloidal correlation length of 11 pixels (2 cm) and are quasi-coherent (the cross-correlation function between 2 poloidally separated points exhibits oscillations). This way, fast visible camera measurements suggest a strong change in the turbulence characteristics at the transition from phase 1 to phase 2. Such observation is backed-up by tunnel probes data, whether around the outboard mid-plane with the pecker probes or at the top of the machine with the reciprocating probe. As shown in Fig. 8, the SMBI leads to an increase by 200% of the fluctuationaveraged ion saturation current hjsati in a 4 ms time. This peak relaxes rapidly to recover a level of hjsati about 60% higher than before the injection. Although it is impossible to decouple the time evolution from the spatial one due to the movement of the probe, one can notice that the decay of hjsati on the way back of the probe is much quicker than the increase on the way in, which suggests either a steepening of the edge radial decay or a rapid drop of the density with time. Zooming on smaller time scales, one notices a dramatic impact of the injection on the characteristics of fluctuations. The fast and extremely intermittent signal observed before the injection is replaced by much slower and coherent bursts. The fluctuation level collapses from 30% down to less than 10% for the side of the probe magnetically connected to the LFS mid-plane, and from 15% to 7% for the side connected to the HFS. The modification of the temporal characteristics of the fluctuations can be quantified by looking at the auto-correlation of the ion saturation current signal (Fig. 9). At the same radial location (3 cm outside the LCFS), the
We have investigated the impact on the plasma of Supersonic Molecular Beam Injection (SMBI) using extended diagnostics coverage. Four distinct phases are systematically observed in the plasma response. After a first peak of 4 ms (phase 1), the volumeaveraged density decays rapidly with an intermittent behaviour, until it reaches a local minimum (phase 2). A temporary increase follows, with a much smoother behaviour, until the density reaches a local maximum, lower than the initial peak (phase 3). In a fourth phase, the density relaxes back to the initial one on a much longer time scale of the order of the particle confinement time. The fuelling efficiency g can hence be defined with respect to the density reached at the end of the third phase. Plasma currents scans show that g drops when decreasing Ip, which is correlated to an increase of the duration of the second response phase. We interpret the latter as a temporary period of degraded confinement triggered by the injection. Such conclusion is backed-up by fast camera and Langmuir probe measurements that both show a dramatic modification of the properties of turbulence after the supersonic injection, with slower dynamics and larger coherent structures. Note that Ip was used in these experiments as the scanning parameter, but one cannot exclude a possible dependence on the density (or altogether a dependence on the Greenwald fraction) which would require more data to be demonstrated. These results shed new light on the understanding of the fuelling efficiency by SMBI but also potentially pellets. They demonstrate that a poloidally localised particle injection can have a strong impact on local transport properties which, in feed-back, impacts the efficiency of the particle deposition. Such a strong non-linear plasma response cannot be captured by diagnostics or simulations set for quasi-steady state analysis so that the understanding of particle fuelling mechanisms requires tools addressing transport issues consistently. We also want to draw the reader’s attention on the fact that local gas injections, often used to increase the contrast of fast visible imaging measurements, are susceptible to change the properties of turbulence so that great caution must be taken in the interpretation of such experiments [14]. Acknowledgements The authors wish to thank N. Fedorczak for fruitful discussions concerning the interpretation of the data presented here. This work, supported by the European Communities under the contract of Association between EURATOM and CEA, was carried out within the framework of the European Fusion Development Agreement. The views and opinions expressed herein do not necessarily reflect those of the European Commission.
S154
P. Tamain et al. / Journal of Nuclear Materials 438 (2013) S148–S154
References [1] [2] [3] [4] [5]
G.W. Pacher et al., Nucl. Fusion 48 (2008) 105003. P.T. Lang et al., Phys. Rev. Lett. 79 (1997) 1487–1490. V. Rozhansky et al., Nucl. Fusion 46 (2006) 367. P. Tamain et al., J. Nucl. Mater. 363–365 (2007) 844–848. J. Bucalossi et al., Particle fueling for long pulse with standard gas puff and supersonic pulsed gas injection, in: Proceedings of 19th IAEA Fusion Energy Conferences, Lyon, October 2002, EX/P4-04.
[6] [7] [8] [9] [10] [11] [12] [13] [14]
B. Pégourié et al., J. Nucl. Mater. 313–316 (2003) 539–542. J.L. Ségui et al., Rev. Sci. Instrum. 76 (2005) 123501. C. Gil et al., Fusion Sci. Technol. 56 (2009) 1219. F. Clairet et al., Rev. Sci. Instrum. 81 (2010) 10D903. A. Géraud et al., Rev. Sci. Instrum. 80 (2009) 033504. N. Fedorczak et al., J. Nucl. Mater. 415 (2011) S467–S470. J.P. Gunn, J.-Y. Pascal, Rev. Sci. Instrum. 82 (2011) 123505. J.P. Gunn et al., Rev. Sci. Instrum. 75 (2004) 4328. Y. Marrandet et al., This conference, 2012.