Observations with fast visible cameras in high power Deuterium plasma experiments in the JET ITER-like wall tokamak

High speed visible imaging allows to visualize the 2-dimensional dynamics of fast phenomena in the boundary layer of fusion plasmas. Here we describe the two high speed visible cameras currently operating in the JET tokamak, enabling the simultaneous observation of a large fraction of the JET tokamak vacuum vessel with the appropriate configuration for many different plasma phenomena. As an example we discuss recent observations on the spatial and temporal dynamics of visible emission at the outer wall during Edge Localized Modes (ELMs) and the penetration of Shattered Injected Pellet (SPI) into the plasma.


Introduction
The Joint European Torus (JET) (R = 2.96 m, r = 1, 25 m), operating with an all metallic wall (ITER-like wall (ILW)) [1,2], is the largest tokamak currently in operation.High power Deuterium plasmas, (NBI up to 32 MW + 8 MW of ICRF power) were achieved during the last experimental campaign (called C38), in preparation for Deuterium-Tritium (D-T) plasmas.
More than 30 cameras and pyrometers are installed in the JET.These elements can be grouped in three main categories: First, four operational cameras which provide a real time wide-angle view of the JET vacuum vessel from different toroidal sectors.Second: A system of absolutely calibrated near infrared (IR) CCD cameras for active protection were installed in order to monitor the temperature of wall surfaces during plasma pulses [3].This system combines wide-angle and divertor-view cameras [4,5] which can interrupt the plasma pulse by mean of a feedback system [17].Finally, scientific cameras: four Divertor and wide-angle view IR cameras for thermographic studies, two wide-angle view visible cameras for spectroscopy and two fast visible cameras, that will be described in the present work.The monitoring of visible and IR emission from different regions of the vacuum chamber will be also fundamental in ITER for both protection and scientific purposes [6].
Fast framing cameras are non-perturbing diagnostics used to observe the 2-dimensional dynamics of fast phenomena in fusion plasmas [7][8][9][10][11][12][13]. A first fast visible camera, named as KL8A, was installed in JET [14] near the equatorial port and looking through an endoscope shared with an IR camera [15], providing a wide-angle view of the JET vacuum vessel from octant 8.It has demonstrated its usefulness to observe disruptions [16], pellets, ELMs [21] and the plasma breakdown.More recently, an additional fast visible camera, known as KLDT-E5WE, was installed outside the JET bio-shield in order to avoid radiation damage and allowing its operation during the next D-T campaign.The combination of both systems allows the observation of a large fraction of the vacuum vessel with the appropriate configuration for different plasma phenomena, as seen in the Fig. 1.In this article we describe the two fast visible cameras installed at JET and we explain recent observations on ELMs and SPI pellets.The working principle of fast visible cameras in JET as well as its operation and analysis procedures are explained in the section 2. In section 3 are shown the observation of ELMs in the outer wall and the monitoring of SPI pellet penetration, together with respective discussions.Finally, some conclusions are given.

Experimental set-up
Both KL8A and KLDT-E5WE diagnostics are Photron APX high speed cameras with CMOS sensors (1024 × 1024, 20 μm pixels) with a maximum speed of 250,000 frames per second (fps) and a minimum exposure time of 1 μs for reduced Regions of Interest (ROI) of the sensor.KL8A has a remotely controllable spectroscopic filter wheel with four selectable positions.KLDT-E5WE camera is installed outside the JET biological shield.It receives the image of the JET vacuum vessel by a mirror based optical relay, as described in reference [17].This is possible by two optical systems.Each optical system has individual field of view of the upper and lower half cross-sections of JET vacuum vessel, which are combined on the single sensor.Each optical system consist of in-vessel off-axis parabolic mirror and folding mirror [19].The cameras are triggered by the JET data acquisition systems and a 2 GB maximum size video is stored as JET pre-processed signal.The video length depends on both the ROI size and frame rate, being from ̃0.5 s to ̃10 s and can be divided in separated sub-intervals.One of the most used configuration was the simultaneous observation of the outer wall region in octant 4 (KLDT-E5WE) and the divertor area in octant 8 (KL8A) (Fig. 2).The camera exposure times, τ exp , depend on the expected light flux, which in turn depends on the plasma emission, and must be at least equal to the inverse of the frame rate.In Table 1 are listed typical exposure times used for each type of observations.In Table 1 are shown typical settings to observe regions of the JET vacuum vessel.Finally, the videos can be visualized and analyzed with the software JUVIL [20], specifically created to handle all the cameras of the JET viewing system.Generally, the cameras are operated without filters in order to maximize the light flux onto the sensor and thereby the frame rate.In Fig. 2 are shown the images of JET divertor (2.e) and outer wall (2.a) obtained with the two fast cameras, respectively.As seen in Fig. 2.b, the visible light at the inner divertor follows the time trace of the vertical D α spectroscopic chords (Fig. 2.c) [21], indicating that the visible light at the fast camera is dominated by D α emission.

Results and discussion
Visible light from plasmas: The main source of visible light in the boundary region of hot fusion hydrogenic plasmas is the atomic line emission from Balmer electronic level transitions (known as D α , λ = 656.3nm in the case of Deuterium plasmas) during atomic excitation processes induced by electron collisions [18].
Observation of ELMs with fast visible cameras: During an ELM crash electrons and ions from the pedestal region loose partially their confinement and are expelled towards the wall and divertor through the separatrix by perpendicular and parallel transport, respectively [22 and there in].A fraction of the ion fluxes, F i , that impacts against the facing materials, penetrates into the surface and are thermalized and finally are re-emitted to the plasma as neutral atoms and molecules.This process is known as Recycling.The neutral flux, F 0 , can be directly related to the impinging ion flux, F i , through wall recycling coefficient: F 0 = RF i .On the other hand, the neutral flux entering in the plasma region can be estimated by: F 0 = KS/XBL Dα .K is a geometrical factor, L Dα is the spectral radiance measured by an absolutely calibrated detector and S/ XB is a coefficient which gives the number of ionization events per photon for a given atomic line [23] and is given by the Collisional-Radiative model [24].This is only valid in ionizing conditions, i.e. when recombination can be neglected as a neutral source.This is generally valid at limiters, but not at the divertor in sufficiently low temperature-high density conditions.During the ELM crash, the level of D α radiance measured by fast cameras clearly increases at the outer wall limiters, as seen in Figs. 2 and 3, indicating an increase of the ion flux towards the limiter targets.A relevant level of D α emission is also observed at a region between limiters 4B and 3B, labelled from now as the Gas Puff Imaging (GPI) area (Fig. 2).This emission is originated at the separatrix by the plasma excitation of the puffed neutral Deuterium.In Fig. 3 are shown the time traces of maximum intensity at both the limiter target and the GPI area during one ELM, taken with the fast camera at 30kfps and 33μs of exposure time.The mean value of D α emission during the discharge is higher in the GPI than in the limiter 4D, while its relative variation during the ELM event is larger at the limiter region, as seen in Fig. 3.The maximum peak of emission in the limiter is delayed with respect to the GPI by ̃150 μ.This dynamics was repeatedly observed and its origin is still under discussion.Assuming a radial velocity of ELMs of the order of 1 − 5 km/s [22] and a clearance in the outer wall (distance from the separatrix to the limiter) of ̃8 cm, radial velocities of up to of 1km/s could be accessible with the fast camera at > 10 kfps.
For more intense ELMs, the described time delay is not clear (Fig. 4).The reason for the sudden D α emission in the outer limiter region is the increased ion flux, sustained during the ELM crash, with the subsequent release of recycled neutral fluxes which are then all ionized.On the other hand, the increase of the signal at the puffing is however not attributed to a change of the neutral source, which is constant, but to a modification of the edge/SOL plasma during the ELM crash that modifies the S/XB coefficient.If this is true, than the time delay of the intensity increase at the different regions can be attributed to the "time of flight" of the plasma ELM front.
Finally, the maximum D α emission measured by the fast camera at the outer limiter 4B follows the time trace of the ion saturation current measured by a Langmuir probe installed at the midplane region of the outer limiter (Fig. 5).This allows the study of the filamentary character of ELMs.
Observation of Shattered Injected Pellets: Plasma disruptions lead to strong heat and electromagnetic loads against the wall.An innovative strategy is taken in order to mitigate that loads: the injection of large cryogenic pellets (SPI pellets).The pellets are composed of Deuterium, Neon or a mixture of both, and are propelled by compressed Deuterium gas towards an injection valve located at the top of JET octant 1 [25].KL8A sensor can collect the D α produced by the Deuterium pellet when it      Fig. 6. a) Penetration of SPI large pellet into the plasma during a disruption.Each frame is an average over 7 frames in order to don't show 70 frames.In frame 1) we see the propellant gas entering the vacuum chamber.In frame 7) some parts of the pellet impact the inner wall.In frame 10) the pellet mostly impact the inner wall.
penetrates into the plasma and is ablated as well as the propellant gas at the top of JET octant 1.The high speed (up to 18 kHz with full sensor) allows the observation of the pellet trajectory inside vacuum chamber.Simultaneously, KLDT-E5WE observes the plasma breakdown at octant 4. In the case of Neon pellet, the filter wheel allows the observation of Ne I atomic line (visible at 692,8 nm).Fig. 6 shows a SPI 100% Neon pellet entering into the plasma and recorded with a 50 μs exposure time.

Conclusions
In the present work we have described the two fast visible cameras operating in JET-ILW.We have also shown recent observations made by fast cameras during the last experimental campaign: The use of fast cameras allows the simultaneous observation of a large fraction of the whole tokamak vessel with the adequate settings.The radial propagation of ELMs towards the outer limiters could be resolved with recording speed > 10kfps and with exposure times τ exp > 33 μs.Studies on turbulence in the Scrape-off layer are possible during ELM phases by observing the fluctuations in D α emission with KLDT fast camera at speed > 10 kfps.In addition, the simultaneous use of the two fast imaging systems is crucial in the observation of SPI pellets and their effects.In addition, fast imaging has demonstrated its utility in the study of other topics, not treated here, such as: disruptions, pacing pellets, divertor recycling dynamics, study of of L-mode and ELM filaments, among others.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1. a) Wide-angle view of JET vacuum vessel from KLDT-E5WE during disruption.b) Equivalent view from KL8A.In the center: top schematic view of JET tokamak and viewing systems.

Fig. 2 .
Fig. 2. a) Photograph of JET vacuum vessel from KL14 endoscope and superposed image of the outer wall observed with KLDT-E5WE.In the central figure, the time trace of integrated signal from KLDT-E5WE (b) and the time trace of D α signal measured by spectroscopy diagnostic looking to the same divertor region (c).In d) is shown the visible emission from the divertor (same instant as in a)) observed with KL8A fast camera.

Fig. 3 .
Fig. 3. a) The time trace of the integrated intensity from three different ROIs at the outer wall in KLDT-E5WE.In b) is shown a set of sorted frames of de-averaged intensity during the ELM.

Fig. 4 .
Fig. 4. In a) are shown the time traces of the integrated de-averaged emission from both the outer limiter and the GPI area during one ELM.In b) are represented the time traces of Tungsten (WI: 406 nm) and Beryllium (BeII: 527 nm) from the inner divertor.c) Sorted frames at the outer wall during the same ELM.

Fig. 5 .
Fig. 5. a).Time traces of both fast camera and limiter Langmuir probe.In b) are shown the time traces of D α and Tungsten impurity lines measured by spectroscopic lines at the inner divertor.

Table 1
Table indicating the typical settings used in the observation of different regions of the vacuum vessel.