TdeV-URG.html

WiZeus > Biography | Education | Scientific | Archive (Demjanjuk, Suslensky, Harasymiw, etc.) >

Centre canadien de fusion magnetique
Tokamak de Varennes

Une entreprise conjointe d'Hydro-Quebec, de I'Energie atomique du Canada limitee et de l'Institut national de la recherche scientifique.

CCFM RI 375e

URG EMISSION FROM THE TOKAMAK DE VARENNES

W.W. Zuzak¹, B.L. Stansfield¹, C. Janicki², P. Couture³ J.L. Lachambre³, G. Ratel³, G. Abel¹

Avril 1992

¹INRS-Energie, Varennes, Quebec, Canada
²Hydro-Quebec, Varennes, Quebec, Canada
³Les Technologies MPB, Dorval, Quebec, Canada

1804 montee Ste-Julie, Varennes, Quebec, Canada J3X 1S1 Telex No.05 267486 Telephone (514) 652-8700 Fax (514) 652-8625

Abstract:

Continuum radiation, predominantly in the IR but extending into the visible, has been observed from low density plasmas in the Tokamak de Varennes (TdeV). Between 510 and 870 nm the spectrum is well represented by an exponential. This radiation (hereafter referred to by the mnemonic URG) is highly collimated in the direction of the electron drift and depends upon the presence of runaway electrons in the discharge as deduced by the generation of hard X-rays. The URG emission is sharply maximized at a critical electron density, n(crit.) , which for TdeV occurs at ω_pe/ω_ce = 0.57.

This critical density defines a discontinuous transition from an unstable to a stable regime in the characteristics of the Tokamak discharge. Below n(crit.) the Tokamak discharge enters an unstable regime typified by hard X-ray "bursts" of erratic duration, periodicity and directionality and by decreases in soft X-ray emission (factor 2) and electron temperature deduced therefrom (ΔT_e = 300 eV). In the unstable regime the URG emission decreases drastically or even disappears. The stable regime above n(crit.) is typified by hard X-ray "spikes" of about 100 μsec duration associated with the collapse of regular sawteeth and preferentially emitted into the hard X-ray cone from the limiters.

Color and black/white video pictures of the URG emission generally show an elliptical spatial distribution about 180 mm in height and with the maximum intensity situated about 100 mm inside the major plasma radius. Crescents, visible on the b/w images early in the discharge, indicate that the URG emission commences on the inversion radius, which occurs at the q=l surface and that the emission from this region is emitted in a "rabbit-ear" cone of 27° half-angle relative to the direction of the magnetic field lines.

Normal synchrotron radiation does not appear to be an appropriate explanation of the observed emission. Interaction of the runaway electrons with electrostatic fluctuations appears to be a more promising avenue for theoretical effort.

¹INRS-Energie, 1650 Montee Ste-Julie, Varennes, Quebec J3X 1S2
²MPB Technologies, 1725 Trans Canada Hwy, Dorval, Quebec H9P 1J1
³IREQ, 1804 Montee Ste-Julie, Varennes, Quebec J3X 1S1

C:\WP51\CCFM>CCFM6URG.PUB; 1992-02-12

URG EMISSION FROM THE TOKAMAK DE VARENNES
W.W. ZUZAK, B.L. Stansfield, C. Janicki, P. Couture
J.L. Lachambre, G. Ratelle, G. Abel

Centre Canadien de Fusion Magnetique, Varennes, Quebec, Canada

Abstract:

1.   INTRODUCTION
2.   EXPERIMENTAL SETUP
2.1 Tokamak de Varennes
2.2 TV cameras
2.3    Photodiodes
2 .4   Interferometry
2.5    X-rays
3.   ANALYSIS OF DATA IN POOLS A TO I
3.1. The Critical Density: INT3(crit.)
3.2 X-rays in Stable and Unstable Regimes
3.3 Plasma Parameters as Function of Electron Density
3.3.1 Loop voltage, V_L
3.3.2 Sawtooth period, Δτ
3.3.3 Inversion radius, r_inv
3.3.4 Electron temperature, T_e
3.4 Comparison of #5040 Stable URG with #5044 Unstable URG
3.4.1 Electron Temperature from Pulse Height Analyzer
3.4.2 URG1 and URG2 photodiode signals
3.4.3 Initial decrease of URG2 signals
3.4.4 Growth rates of URG emission
3.5 Other Phenomena
3.5.1 Polarization
3.5.2 83% of URG emission is above 715 nm
3.5.3 URG emission is inversely proportional to R_MAJ
3.5.4 Step-jumps in URG emission
3.5.5 "Late" URGs
4.   ANALYSIS OF THE URG SPECTRUM
5.   SPATIAL DISTRIBUTION OF URG EMISSION
5.1 Shot #1396
5.2 Shot #5040
5.3 Shot #2326
5.4 Anomalous shot #2256
5.5 General Results
6.   TOTAL ENERGY OF URG EMISSION
7.   DISCUSSION
7.l Energy of runaway electrons
7.2 Implication of 27^o half angle cone
7.3 ω_pe and ω_ce at n(crit.)
7.4 Synchrotron radiation
7.5 Scattering from fluctuations
8.   CONCLUSIONS

ACKNOWLEDGEMENTS
REFERENCES
FIGURE CAPTIONS

C:\WP51\CCFM>CCFM6URG.LST; 1992-02-12

1. INTRODUCTION
The Unidentified Red Glows (URGs) described in this paper were first observed and recorded on 8 October, 1987 (shots #1396, 1397, 1402) with the b/w camera looking tangentially upstream the runaway electrons. The mnemonic arose (eventually) from the fact that the glow appeared mysteriously (like a UFO) for eleven video fields (167 msec) near the centre of the discharge and then disappeared. Eventually, the phenomenon, which occurred haphazardly during shots characterized by low electron densities and large X-ray emission, was also observed with the color camera, with photodiodes and with a monochromator plus intensified camera combination. By the end of the first phase of the TdeV experimental program on 8 December, 1988, over 250 URG discharges had been observed and recorded. The relevant experimental apparatus is described in Section 2.

Thereafter, eight pools of data A to H incorporating shots #5519 to # 3654 were created and analyzed in turn [1] . Three particularly relevant earlier shots are included in pool I (#2253-#2356) . Each pool contains the archived data of the relevant diagnostics which were on-line at the time. The highlights of this analysis are reported in Section 3.

The image analysis system called XICAS3 allowed digitization, archival and analysis of the video-tape images. The URG spectrum (corresponding to shots #5395-#5414 in pool A) was recorded with the monochromator-intensified camera combination. The results of image analysis are presented in Section 4.

Further image analysis of six selected shots from #1396 to #5040 allowed us to determine the spatial distribution of the URG emission and to highlight particular features of this distribution as the plasma current, I_p, was increased from 85 to 189 KA. These results are presented in Section 5.

Energy calculations are carried out in Section 6 and, finally, some of the possible explanations for the URG phenomenon are briefly discussed in Section 7.
2

2. EXPERIMENTAL SETUP
2.1 Tokamak de Varennes
TdeV is a medium-sized Tokamak of major radius, R = 0.866 m; minor radius, a = 0.24 m (adjustable; normally defined by the vertical limiters, LM1 and LM3) and a toroidal magnetic field, B_T = 1.42 Tesla. The duration of a plasma discharge is typically 800 msec. Although plasma currents of 280 KA and electron densities exceeding 4.5xl0¹⁹ m^-3 have been attained, URG emission has typically been observed at I_p = 185 KA (from 85 to 195 KA) and a line averaged electron density, INT3 = 0.97xl0¹⁹ m^-3 (from 0.95 to 1.28xl0¹⁹ m^-3) . A more detailed description of TdeV and its operational characteristics during the first phase of operations are provided elsewhere [2] .

A top view of TdeV showing the relevant diagnostics is given in Fig. 1. Sixteen toroidal field coils divide the torus naturally into 16 bays. Their numbering proceeds counter-clockwise from Bay 1 in which the graphite limiters are located to Bay 16. We note in passing that the electron drift and the toroidal magnetic field are both in the clockwise direction. Windows in Bays 2, 5 and 13 allowed direct observation of the URG emission by means of cameras and photodiodes.

2.2 TV Cameras
Two TV cameras have been used to observe the plasma, generally looking toroidally either upstream or downstream (with respect to the electron drift direction). The location of the cameras for the measurements described herein are indicated in Fig. l.

A General Electric 4TN2505-B2 CID solid state camera is always located in Bay 13 looking upstream toward Bay 2. The solid state CID silicon chip (388H x 242V pixels) is sensitive from 400 to 1050 nm with relatively flat response between 500 and 700 nm.

A Hitachi KP-C100U color camera was located either in Bay 2 looking toward Bay 7 or in Bay 5 looking toward Bay 10. A 4 m long Reichert-Jung fibre optic image guide, placed between the zoom lens and camera, was used for all shots. The CCD silicon chip (492H x 512V pixels) with the appropriate RGB mosaic filters limit the sensitivity to the 400 to 700 nm wavelength region. The color images of three particularly relevant shots #2256, #2326 and #5040 are shown in Fig. 2 (a) to (c) . They are analyzed in detail in Section 5.

To measure the URG spectrum, an intensified TV camera was placed some 30 cm from the exit plane of a 0.5 m Jarrell-Ash monochromator (1180 lines/mm grating, 200 μ entrance slit) as indicated in insert (a) of Fig. 1. Two lenses were placed near the exit plane and experimentally adjusted to maximize the spectral coverage (58 nm). The camera (XYBION Electronic Systems) consists of a single multi channel plate intensifier fitted onto a GE 4TN2505-B2 camera as described above. The image intensifier increases the luminous sensitivity by a factor of 18,000 and the S-20 extended red photocathode has a spectral sensitivity from 400 to about 900 nm with peak sensitivity around 650 nm.

A 2.5 m long quartz fibre optic cable was used to transfer the
3

light from the Tokamak to the monochromator. A lens, placed in front of Bay 2 and oriented toward Bay 7, focused the URG emission onto the 4 mm dia. entrance surface of the fibre optic cable. The 2 mm x 5 mm exit surface was placed flush against the bottom portion of the entrance slit of the monochromator. An identical fibre optic cable carried a reference spectrum of He, Ne, or Ar to the top portion of the entrance slit. A detailed analysis of the spectrum results is given in Section 4.

2.3 Photodiodes
Several silicon photodiodes are mounted in telescopes and used to monitor the plasma emission; they are placed at various locations around the machine and can be oriented in any direction. Their final configuration consists of a Hammamatsu S1722-02 silicon photodiode (4.1 mm dia., sensitivity from 400 to 1100 nm with peak at 900 nm) operating at +45 volts reverse bias and fitted with a R_L = 100 Kohm load resistor. A glass lens (13 mm effective aperture, f = 114 mm) at one end of a 1"OD x 118 mm long aluminum tube focuses the light onto the photodiode surface.

Typical signals illustrating URG emission are shown in Figs. 3 (a) and (b) . As indicated in Fig. 1, URGl refers to the signal from the photodiode situated in Bay 13 looking upstream the runaway electrons toward Bay 2 and URG2 refers to the photodiode situated in Bay 2 looking downstream towards Bay 13.

In order to increase their sensitivity (by a factor of about 40) and allow the observation of weaker spectral lines by using interference filters, four of the photodiodes were fitted with LH0032 amplifiers. Unfortunately, these photoamplifiers were particularly sensitive to X-rays which swamped the signal during shots exhibiting URG emission. This is illustrated by the PA6 and PA4 signals in Figs. 8 and 9 discussed later in the text.

2.4 Interferometry
The electron density measurements are made with a 6 channel FIR laser interferometer situated in Bay 10 and operated at a wavelength of 214 μm [3] . The 6 laser beams pass vertically through the plasma, and the horizontal spacing of the channels is typically 50 mm. The phase shifts from the 6 chords are used to calculate effective line-average densities (INTl - INT6) using a nominal plasma diameter of 0.48 m for each of the lines of sight. Density profiles are obtained from fitting a model profile represented by a fourth order polynomial to the chord-averaged phase shift measurements. This allows the radial plasma displacement, the ellipticity and the triangularity to be calculated.

For earlier shots preceding #2356, the line-average density measurement (DENS) is obtained from the phase shift produced in a microwave interferometer, assuming a plasma diameter of 0.48 m. The microwave source is an Extended Interaction Oscillator operating at 140 GHz, and the beam passes through the centre of the plasma column (R = 866 mm) in Bay 6.
4

2.5 X-rays
Soft X-rays in the energy range from 2-14 KeV are measured by an SiLi detector operating in the Pulse Height Analysis mode and situated in Bay 9. An energy spectrum (corrected for filter transmission and detector response) can be obtained every 50 ms, from which the central electron temperature can be calculated.

The poloidal distribution of the soft X-ray emission at one toroidal location in Bay 12 is measured using an array of 80 surface-barrier detectors mounted in 5 separate cameras (SX100 to SX600 arrays with 16 detectors each) providing 5 complete poloidal projections [4]. The lower energy limit of the detector response is determined by the 25 μm Be window used with the arrays, whereas the upper limit is determined by the thickness of the active region of the detector. The entire system is compatible with a data acquisition rate of 100 KHz. This system is particularly useful in providing accurate measurements of the inversion radius for the sawteeth.

Hard X-rays, which were invariably present during URG emission, are presumably the result of high energy runaway electrons striking the limiters and/or the walls of the Tokamak vessel. These were normally monitored by means of 4 hard X-ray detectors positioned in the hard X-ray cone (opposite Bay 14), and aligned to look toroidally toward the limiters in Bay 1. The X-rays produce visible photons in BGO scintillators of different thicknesses, which are optically coupled to photomultiplier detectors. Absorbers of Cu and Pb are placed in front of the scintillator, between two lead collimators. The energy ranges covered by the 4 detectors are determined by a combination of collimator and scintillator thicknesses. As typical results, we show in Fig. 9 the signal from the detector channel XDUR3, which measures X-rays in the range 0.1-5 MeV.

We have observed, however, that many of the other diagnostics pick up the hard X-rays as a noise signal superposed on their regular signal. In addition to the photodiodes (URG1, URG2) and the photoamplifiers (PA2-PA6) described above, the signals from the SX300 horizontal and SX600 vertical arrays of the soft X-ray tomography experiment and the BREM1-BREM7 signals of the Z-eff metre are particularly useful in monitoring the presence of hard X-ray spikes.

Direct reading dosimeters [Dosimeter Corp #862, 0-200 mrem] were used to measure the integrated X-radiation in front of and behind 10 cm of lead shielding the color camera situated at R = 4 .6 m, h = -1.26 m in front of Bay 2. Within a factor 2, typical values were 30 and 2 mrem respectively during a strong URG shot.

The CCD chip of the color camera and the MCP of the intensified camera are sensitive to X-rays which appear as white flecks on the video images. The 4 m long image guide physically removed the color camera from the region of high X-rays. In addition, 10 cm of lead shielding further reduced the problem on both the color and intensified cameras. We note, however, that the CID chip of the b/w camera appears to be immune to X-rays, despite being directly in the hard X-ray cone from the limiters.
5

3 . ANALYSIS OF DATA IN POOLS A TO I
The first two data pools A and B illustrate most of the important phenomena of strong stable URG emission. The data in the other pools confirm these results and also provide illustrations of less "standard" URG emission profiles.

In this section we first define the concept of a critical density, the stable and unstable regimes and the hard X-ray "spikes" and "bursts" in these regimes. Thereafter, we discuss the problem of X-ray noise and the general variation of plasma parameters with electron density, before examining the various aspects of the URG phenomenon in detail.

3.1. The Critical Density: INT3(crit.)
INT3 is the line averaged electron density at R=0.830 m. It is normally the largest of the 6 SMM interferometer signals (INTl-INT6) from which the electron density profiles are calculated [3] . In fact, because of uncertainties and inaccuracies in the deconvolution process, it has turned out to be the more stable parameter to utilize as a representation of the electron density during URG shots.

A plot of the peak URG amplitude versus INT3 for pool B is shown in Fig . 4 . The error bars represent the minimum and maximum values of INT3 between 200 and 700 msec; the crosses, x, represent an average or typical value between 200 and 350 msec before rapid growth of the URG emission commences. We have found this to be the most appropriate value of INT3 to use.

We define the critical density, INT3(crit.), to be the value of INT3 at which the URG emission is a maximum. For densities greater than INT3(crit.), the URG amplitude decreases very rapidly in an exponential fashion. This is defined as the stable regime. At densities below INT3(crit.), the peak URG amplitude drops discontinuously and can even be undetectable. Since the character of the discharge changes, this is called the unstable regime.

Although four of the nine pools indicated INT3(crit.) = 0.97xl0¹⁹ m^-3, the other pools indicated values ranging from 0.95 to 1.28xl0¹⁹ m^-3. This variation may perhaps be partially attributed to a weak inverse dependence on the major plasma radius. The lowest critical density of 0.95xl0¹⁹ m^-3 (pool B, Fig. 4) corresponded to R_MAJ = 869±7 mm; whereas the highest value of 1.28xl0¹⁹ m^-3 was associated with "late" URGs in pool D with R_MAJ = 844 ±3 mm.

3 .2 X-rays in Stable and Unstable Regimes
One of the most striking differences between the stable and unstable regimes defined above is the character of the hard X-radiation emitted during the discharge. The stable regime is characterized by hard X-ray "spikes" associated with the collapse of sawteeth and preferentially emitted into the hard X-ray "cone" from the limiters. Presumably, the sawtooth crash mechanism deflects the orbits of some of the runaway electrons into the limiters. Invariably, the onset of strong stable URG emission is accompanied by a large increase in the amplitude of these hard X-ray spikes. Secondly, the peak in the hard X-ray emission always
6

occurs before the peak in the URG emission.

The unstable URG regime below the critical density is always characterized by hard X-ray "bursts" which are defined as hard X-ray emission of erratic duration, periodicity and directionality. There is also a drastic decrease or disappearance of URG emission, a weak diffuse spatial distribution as seen on the camera video pictures and a sharp decrease in the electron temperature of some 300 eV, which we shall discuss in detail later.

In Fig. 5, we show the soft X-ray signal SX309 of three consecutive shots #4432-#4434, which illustrate a stable URG, an unstable URG and a transition from the unstable regime back into the stable regime. The hard X-ray signals (which are superposed on the normal soft X-ray signals) shown on the expanded scale from 640 to 720 msec are an excellent illustration of the difference between "spikes" and "bursts". We note in passing that the hard X-ray signals picked up by the horizontal SX300 array is about 10 times larger than that picked up by the vertical SX600 array (compare to Fig. 9(a) and (b)).

Hard X-ray bursts are not normally observed for the first 200 msec when the current and density are being ramped up to steady state values. Normally, if the steady state density is below the critical density, a hard X-ray burst around 220 msec will signal the transition into the unstable regime where the discharge will remain (as in #4433) unless the density is increased above the critical density. Analysis with a Fast Fourier Transform routine indicates that there is no hidden periodicity in the hard X-ray burst signals associated with the unstable regime.

If the electron density is increased above INT3(crit.) (as in #4432), the discharge will make a transition back into the stable regime and after 100 msec or longer URG emission may commence (if the electron density is not too high).

3.3 Plasma Parameters as Function of Electron Density
The range of densities covered in pools A and B was from INT3 = 0.7 to 3.5xl0¹⁹ m^-3 at a constant plasma current, I_p = 185 KA. This allowed us to plot the loop voltage, V_L, the sawtooth period, Δτ, the inversion radius, r_inv and the electron temperature, T_e as a function of INT3. These plots, shown in Figs. 6(a) to (d) act as a standard against which values from other pools can be compared.
We note that there appears to be a definite discontinuity in the plasma parameters on either side of the critical density INT3(crit.) = 0.97xl0¹⁹ m^-3 and that the regular sawteeth and thus the inversion radius disappear below the critical density. Although we have drawn straight lines to represent the general variation, there is a substantial scatter from 10 to 20 % of the experimental points from these lines.

We shall briefly discuss each of these figures in turn:

3.3.1 Loop voltage, V_L
The plot shows a sharp discontinuity at the critical density (INT3 = 0.97xl0¹⁹ m^-3, V_L = 1.58). Obviously, the effective plasma resistance in the unstable regime decreases much more rapidly
7

(2.0x10^-19 V/m^-3) below the critical density than its variation in the stable regime (0.22x10^-19 V/m^-3) above the critical density. As observed on the XDUR3 signal and the dosimeter measurements, production of hard X-rays generally increases below the critical density.

3.3.2 Sawtooth period, Δτ
There appear to be significant variations in sawtooth period between the various pools as well as during a single discharge. For example, in shot #5055 the period increased suddenly from 1.09 to 1.79 msec for no apparent reason.

3.3.3 Inversion radius, r_inv
The apparent inversion radius for the sawteeth is easily observed on the soft X-ray tomography signals SX601-SX616 (vertical) and SX301-SX316 (horizontal). Inside the inversion radius the sawtooth slope is positive between crashes, whereas outside the inversion radius the slope is negative. As seen in Fig. 6(c), the apparent inversion radius increases from 63 mm at INT3 = 0.97xl0¹⁹ m^-3 to 83 mm at 2.5xl0¹⁹ m^-3 at a plasma current I_p =185 KA.

However, based on an Abel inversion of the 16 SX600 signals, the real inversion radius at these two density values increases from 85 to 101 mm. Although Abel inversion of the SX300 signals in the horizontal plane gives similar values at high densities, it was not possible to confirm the calculations at the critical density because of the hard X-ray noise superposed on the signals.

For TdeV in the limiter mode, the inversion radius appears to depend mainly on the current profile and only weakly on the density according to r_inv ~ I_p^0.73. Measurement of the total vertical extent of the URG emission observed on the b/w video images before saturation occurs ranges from 53 mm at 85 KA to 145±15 mm near 185 KA. If the two phenomena are directly related, this would suggest a rather stronger current dependence of 1.3 rather than 0.73 in the exponent given above.

According to Kadomtsev [25], the inversion radius is directly related to the q=l magnetic field surface. Later in the text, in struggling to explain the spatial distribution of the URG emission, a possible interpretation suggests that, at the critical density, the q=l surface falls significantly inside the inversion radius.

3.3.4 Electron temperature, T_e
The data plotted for pool B in Fig. 6 (d) indicates that the electron temperature is greatest at the critical density (1350 eV at 0.97xl0¹⁹ m^-3) and decreases with density with a slope of -200xl0^-19 eV/m^-3.
Below the critical density, the electron temperature drops discontinuously by about 350 eV to 1000 eV. This phenomenon is treated in more detail below.

During the course of the URG measurements over a period of several months, the PHA diagnostic was being debugged and calibrated, such that quantitative values vary significantly. For example, for pool A the electron temperature seems to be higher by
8

some 200 eV. Values from other pools generally fall between these two extremes. Nevertheless, although the quantitative values vary somewhat, the qualitative results are consistent.

We should also point out that the electron temperature is taken from the slope of the photon count versus energy curves within the range 2.58 to 5.53 KeV which could be associated with a high energy tail in the electron distribution rather than the bulk electrons representing the real electron temperature.

3.4 Comparison of #5040 Stable URG with #5044 Unstable URG
Shots #5040 and #5044 in pool B have been chosen as the most representative examples of strong stable URGs and unstable URGs. The various diagnostic signals associated with these shots illustrate most of the features which we wish to highlight. In this subsection we highlight four of these features:

3.4.1 Electron Temperature from Pulse Height Analyzer
A typical PHA signal is shown in Figs. 7(a) for shot #5044. For the stable URG #5040 the photon count was reasonably steady around 210 photons per unit time interval from about 125 to 725 msec. For the unstable URG (#5044) however, there is a definite drop in the photon count from about 220 to 110 at 375 msec, which is precisely the time when the transition into the unstable regime occurred.

With 100% consistency for all the unstable URGs observed, the soft X-ray emission in the unstable regime is about a factor of 2 lower than in the stable regime near the critical density. This even holds true for shots which go unstable and then recover into the stable regime.

We should note that the soft X-ray tomography signals SX309, SX609 indicate a similar difference of soft X-ray emission in the two regimes.

With the same 100% consistency, the electron temperature calculated from the slope of the photon count between 2.58 and 5.53 KeV is about 30% lower in the unstable regime relative to the stable regime. For example, the stable URG #5040 shows an electron temperature of 1352±61 eV throughout the whole shot. However, as shown in Fig. 7 (b) , the unstable URG #5044 has an electron temperature of 1300±79 eV before 375 msec, and 977±74 eV for the unstable portion thereafter.

3.4.2 URG1 and URG2 photodiode signals
The URG1 and URG2 signals for shots #5040 and #5044 were previously shown in Figs. 3(a) and (b) respectively. In both these figures, the URG emission on the URG1 signal clearly stands out above the background URG2 signal.

On both shots rapid onset of URG emission appears to start at about 380 msec, but for shot #5044 the growth rate is so rapid that the discharge goes unstable and the URG emission stops increasing. The telltale hard X-ray bursts are clearly visible on the URG1 signal.
9

3.4.3 Initial decrease of URG2 signals
For the stable URGs in pool A, we find that between 110 to 250 msec the URG2 signal consistently decreases by about 10% to a minimum value of 23.3 mV. This decrease is even larger (30 to 40%) for unstable URGs.

This contrasts with the URG1 signal which between 110 to 250 msec tends to increase by 29% (from 20±1 mV to 25.8±0.0 mV for stable URGs) . This phenomenon can be seen in Figs. 3 (a) and (b) for shot #5040 and #5044.

We note that the URG2 photodiode, which measures bremsstrahlung continuum and line radiation between 400 and 1100 nm is looking downstream the runaway electrons. Thus, this 10% dip in the URG2 signal may presumably be related to a decrease in the bremsstrahlung continuum as a substantial fraction of the electrons start running away in the opposite direction (see [24], Fig. 1).

3.4.4 Growth rates of URG emission
For most of the stable URG shots (especially in pools A and B) the evolution can be divided into at least two parts: the initial slow growth rate section before 380 msec, and a fast growth rate after this time when the URG emission rapidly approaches its saturation value. Deciding whether the growth of the URG emission is best described by linear or exponential fitted curves is not straight forward especially in the early slow growth region. The evolution of the URG emission varies from shot to shot and is not exactly reproducible even if the final amplitude is the same.

For shot #5040 illustrated in Fig. 3 (a), a linear fit results in a slow growth rate of 0.114 mV/ms before 385 ms and a fast growth rate of 4.016 mV/ms thereafter to a maximum value of 0.6 volts. On the other hand, on a semilog plot, one can distinguish three different slopes suggesting three different growth rates with e-folding times of 0.0886 s before 257 ms, 0.167 s before 360 ms and 0.0354 s thereafter. Although such three-slope curves are typical of all strong URGs in pool B, the stable URGs in the pool A series are more varied and exhibit 3, 2 or even 1 slope. These observations appear to indicate that if the initial growth rate is too large, it cannot be maintained and it slows down by a factor of 2 or even stays constant in amplitude. Nevertheless, by about 380 ms conditions are such that the growth rate can increase dramatically.

3.5 Other Phenomena
The diagnostic signals in Figs. 8 and 9 illustrate certain other phenomena which we wish to highlight.

First of all in Fig. 8 for #5040, the PA6 photoamplifier fitted with an H_α filter shows a spike at 405 msec which corresponds to a gas puff which in turn is reflected in a momentary increase in the INT3 signal. Gas puffs of such short duration do not appear to affect the URG mechanism. Also note the spike in the plasma current, I_p at 387 msec which corresponds to the ohmic heating current being ramped down through zero. As far as we can determine, any correlation with the onset of the fast growth of the
10

URG emission is purely coincidental.

In Fig. 9 (a) and (b) for shot #5413, we compare the signals URGl, SX609, PA4 and XDUR3 . Although the PA4 photoamplifier was fitted with a CI 9095 interference filter, we have finally concluded that it is picking up mostly "intermediate" X-rays in the 20 to 200 KeV region. (A large fraction of the PAG (H_α) signal in Fig. 8 is also presumably due to hard X-rays.) The reason for our conclusion is that it tends to appear after the peak in the soft X-ray signal, SX609, but before the peak in the hard X-ray signal, XDUR3.

Secondly, on the expanded scale in Fig. 9 (b) , we note that the hard X-ray spikes on the URGl, SX609 and XDUR3 signals are closely correlated in amplitude to each other but not to the spikes in the PA4 signal. The detectors for the first three signals are all in the hard X-ray cone from the limiters in Bay 1; whereas the PA4 photoamplifier is situated in Bay 2 and is thus not in the hard X-ray cone.

There are five other phenomena associated with URG emission which should be highlighted:

3.5.1 Polarization
There is no detectable linear polarization of the URG emission. One would expect normal synchrotron radiation to be horizontally polarized.
We were not equipped to check for circular polarization.

3.5.2 83% of URG emission is above 715 nm
By placing a Schott 715 nm cutoff filter in front of a detector on succeeding URG shots #4653-#4654 and #4905-#4906 in pool C, it was determined that 85 and 81% respectively of the URG emission is in the infrared above 715 nm. Both sets of shots indicate that before the URG emission commences, only 27% of the background signal is above 715 nm. We note that H_α 656.3 nm contributes 13% of the total background radiation.

3.5.3 URG emission is inversely proportional to R_MAJ
Throughout all our shots containing strong stable URGs we find that slow 30 msec oscillations in the URGl amplitude are inversely correlated with similar oscillations in the major radius R_MAJ. In fact, for 19 shots in pool A, we find that the % change in URGl amplitude per 1 mm change in R_MAJ is -2.4±0.6 %/mm. (For an average URG amplitude of 0.27 V, this corresponds to -6 mV/mm.) Such oscillations are illustrated on the URGl and R_MAJ traces in Fig. 8.

This observation is compatible with our video camera pictures which indicate that the bulk of the URG emission is emitted from regions inside the major radius where the toroidal magnetic field is larger (B = constant/R). It is also compatible with the idea that the motion of the plasma inwards allows the survival of higher energy runaway electrons whose orbits are normally shifted outwards.

Secondly, both on the camera pictures as well as on the URGl signals, we sometimes noticed a "flash" or "peak" in URG emission
11

just before disruption. These appear to be very closely correlated to dips in R_MAJ which often occur just after rampdown of the plasma current begins.

3.5.4 Step-jumps in URG emission
A sudden increase in the URG emission of up to 20% is visible on many of the strong stable URGs as illustrated in Fig. 9(b) (shot #5413 at 547 msec). With only one observed exception, this "step-jump" coincides in time with a hard X-ray spike. Normally, the time period between successive sawteeth increases after the step-jump but decays back to its normal value after several time periods.

For several shots in pool H, two photodiodes were set up in Bay 2 looking toward Bay 7. One photodiode was oriented toward the region of maximum URG emission; the other was inadvertently rotated +8.4° ccw toward the inside wall (i.e toward the inside edge of the URG emission) . We noted that the amplitude of the step-jumps at the inner edge were about twice as large as at the centre. The implication is that these step-jumps occur primarily in the edge region of the URG as more plasma enters a state where URG emission can occur.

3.5.5 "Late" URGs
We arbitrarily define "late" URGs to appear after 520 msec on the URG1 signals. Normally, these are much more difficult to analyze because they are seldom very intense and are often buried in background noise (especially for the shots where the limiters start glowing brightly toward the end of the discharge). Excluding pools A and B, about 25% of the shots were classified as late.

Most of these late URGs are of two types:
One type starts in the unstable regime at low densities and makes a transition into the stable regime as the density is increased (e.g. #4432, Fig. 5(a)). The growth of the URG mechanism is thus delayed by some 200 msec and they seldom have sufficient time to reach saturation before disruption occurs. They often exhibit a peak if R_MAJ decreases just before disruption.

The other type starts substantially above the critical density and the density either remains constant or decreases during the discharge. The growth mechanism for URG emission is obviously much slower at these higher densities and may perhaps be related to the balance between acceleration and destruction of runaway electrons. Once again saturation of URG emission is seldom reached and a peak just before disruption, as R_MAJ decreases, is often observed.

Many other "late" URG shots are not easily categorized in the above two types. These must be examined individually.
12

4. ANALYSIS OF THE URG SPECTRUM
The XYBION intensified camera plus monochromator combination described in Section 2.2 was utilized to record the URG spectrum from 510 to 914 nm at 50 nm intervals. The relevant data was digitized (320H x 240V), archived and later analyzed with the XICAS3 image analysis system.

The intensity (0-255 grey scale) inside six "windows" (10 pixels high by 17 pixels wide) spanning the spectrum every 10 nm was measured for each video field between about 350 and 700 msec. The results for shot #5410 are shown in Fig. 10. In this figure the topmost curve corresponds to a digitized version of the URGl signal (from which the background signal has not been subtracted) and the other six curves represent the grey scale intensities within the windows from 713 down to 663 nm. We note that, except for a touch of saturation above 240 grey scale, the shape of the intensity curves are very similar to that of the photodiode URGl signal.

A similar procedure with the black body calibration spectrum recorded a month earlier allows us to normalize the above curves and, by combining the results of all 15 shots between 450 and 700 msec, to obtain the URG spectrum shown in Fig. 11. The spectrum is best represented by an exponential fit of the form y ~ exp(bx) with b = 5.80x10^-3 nm^-1.

Several things should be noted in Fig. 11. The contribution of the C II 514.0 nm triplet and the Hα 656.3 nm line can be seen on the curve. Every sixth point corresponding to the sixth window at each wavelength setting is too high. This is due to an inexact cancellation of the instrumental profiles resulting from a 23 pixel rotation of the camera between the time the calibration and URG spectra were recorded. Because of saturation, the values for wavelengths greater than 770 nm may be too low. Above 870 nm the sensitivity of the intensified camera is so low that calibration is extremely inaccurate such that the results go off scale and are omitted.

Finally, we should point out that the above results are valid for the saturated portion of the URG emission between about 500 and 700 msec. Also, the XYBION camera plus monochromator combination was viewing the most intense region of the URG emission. Thus, the spectrum may be different during the early stages of the URG or at the edge of the URG emission.
13

5. SPATIAL DISTRIBUTION OF URG EMISSION
Over a period of one year, as the plasma current was increased from 85 to 190 KA, the spatial distribution of the URG emission as observed with the TV cameras changed significantly. It evolved from a small ellipse approximating a beam, into a larger hollow oval within which fine structure could be resolved, and finally into the fuller, brighter double ellipse structure typical of the shots analyzed in detail above. Shots #1396, 2326 and 5040 representing these three stages are treated below. The anomalous shot #2256, the brightest URG observed, is treated separately.

The color video images of shots #2256, 2326, 5040 in Fig. 2 give a good overall indication of the spatial distribution of the URG emission and have ideal intensities for image analysis. However, the corresponding b/w images are saturated during peak URG emission, such that structure can only be seen during the early stages of URG development.

Interpretation of these pictures is not straight forward for two reasons. First of all, the radiation is rather sharply peaked in the direction of motion of the runaway electrons. Thus, since the radiation is not isotropic, the position of the observed image depends upon the relative position of the observing camera. Secondly, the two-dimensional image represents a transparent three-dimensional object, such that the pixel position on the image cannot be easily related to an object position along the line of sight inside the vacuum chamber. The reader is cautioned to keep these problems in mind when reading the following sections.

5.1 Shot #1396
This shot was characterized by a plasma current of 85 KA, a minor radius of 200 mm and hard X-ray emission of 45 mrem as measured on the dosimeter. Data on the major radius and electron density are not available. The URG mysteriously appeared on 11 video fields of the b/w camera between 350 and 517 msec and then disappeared.

The contours of the digitized image, shown in Fig. 12(a), are elliptical in shape with no sign of any fine structure. The FWHH (full width at half height) dimensions as defined by the 96 grey scale intensity contour is about 106 mm horizontal by 61 mm vertical (using an estimated R_MAJ = 816 mm and a camera-object distance of 1631 mm).

5.2 Shot #5040
This shot was characterized by a plasma current of 182 KA, a constant plasma density of INT3 = 0.95xl0¹⁹ m^-3, minor plasma radius of 240 mm, R_MAJ = 868 mm and hard X-ray emission of 37 mrem.

The contour plots of the color image of shot #5040, as depicted in Fig. 12 (b) , show the double elliptical structure to which we have referred previously. Superposed thereon are circles corresponding to the minor plasma radius, a = 240 mm, and the inversion radius suggested in section 3.3.3, r_inv = 85 mm. Their slight ellipticity is due to the vertical scaling being about 15% larger than the horizontal scaling.
14

The crescent at the top of the image, which was clearly defined on earlier shots (see below), is now fuller and appears to meld into the emission from the midplane. The vertical extent of the URG emission clearly falls below the rinv = 85 mm surface. The bulk of the URG emission, defined by the 48 grey scale contour, is situated slightly above the midplane and 111 mm inside the major radius.

The second weaker ellipse descends far below the midplane and the r_inv = 85 mm surface, but appears to be confined within the 240 mm minor radius of the plasma. (The fine structure visible at the bottom is associated with time/date generator numbers superposed on the image.)

5.3 Shot #2326
This shot was characterized by a plasma current of 168 KA, a constant plasma density DENS = 0.86xl0¹⁹ m^-3 (not directly comparable to INT3), minor plasma radius of 240 mm, R_MAJ = 823 mm (estimated) and hard X-ray emission of 36 mrem. For this current, we calculate an inversion radius of 79 mm (see Section 3.3.3).

The contour plots of the b/w digitized image in Fig. 12(c) is at t = 417 msec (videoframe 03.90 sec) just before saturation occurs; whereas the color image in Fig. 12(d) corresponds to t = 533 msec (videoframe 04.01 sec) when the URG emission is a maximum. On these contour plots, we have located the plasma centre at R_MAJ = 823 mm and drawn r_q = 65 mm (rather than r_inv =79 mm) circles to help interpret the spatial features observed on the images.

Also superposed on these plots are three windows (5 pixels high by 11 pixels wide) labelled from left to right Bl, B2, B3 for the b/w contours and Cl, C2, C3 for the color contours. These windows are situated on regions of enhanced URG emission: B1, C1 are both exactly on the midplane; B2, C2 are both below the midplane but B2 is on a sharp horizontal crescent, whereas C2 is in a region of diffused emission; B3, C3 are both above the midplane and both are situated on a sharp crescent near the top of the image. We emphasize that the crescents associated with windows B2, B3 and C3 are usually extremely peaked, indicating a structure as fine as 1 pixel or 2 mm in width.

We have confirmed that the time variation of the average grey scale intensities inside these windows for the color images closely follows that of the URG1 photodiode signals (comparable to Fig.10). However, the intensity in C1 on the midplane is consistently more than twice as large as that in C2, C3 on the edges. There is no evidence that different regions of the URG evolve differently.

For the b/w contours in Fig. 12(c), the camera was situated 125 mm below the midplane looking upward at an angle of 4.38°. The spatial distribution of the URG emission at 417 msec appears as a crescent situated to the left of the plasma major radius at (x,y = 179, 133) and opening to the right. (The increasing intensity to the right of the centreline is due to recycling from the outboard limiter, LM2, rather than to URG emission.)

In Fig. 12 (c) , the crescents in windows B2 and B3 do not appear to be located on the r_q = 65 mm surface. However, in a
15

perspective view, it becomes immediately apparent that B3 and the top crescent, in general, could be associated with the top edge of this surface situated from about +11 to +23° (i.e. further back behind the perpendicular midplane). Window B2 appears to correspond to emission from the bottom of this surface situated at about -20° (i.e. in front of the perpendicular midplane).

Similarly, for the contours in Fig. 12 (d) , where the color camera was situated 53 mm above the midplane looking downward at an angle of -2.08°, window C3 is clearly associated with the top edge of the r_q surface situated at +23°. However, window C2 is clearly below this surface and no sharp crescents are visible in this region.

At the q=l surface the runaway electrons presumably follow the magnetic field lines and make one poloidal revolution for every toroidal revolution. If our r_q surface actually corresponds to the q=l surface then the angle of these magnetic field lines relative to the toroidal direction is thus tan^-1 (r_q/R_MAJ) = 4.5° with a left-handed pitch. Combining this with the angles between B2, B3, C3 and the cameras, the angle between the B field lines and the camera can be readily (though not very accurately) determined to be about 27° for both the B3, C3 and B2 windows.

The implication is that from the r_q surface the URG emission is being radiated into a "rabbit-ear" cone of half angle θ(max) = 27°.

For the b/w contours (at t = 417 msec) , B1 on the midplane appears to be situated on the edge of the r_inv surface (toward the centre of the Tokamak) , the URG emission is relatively weak and does not penetrate inside the r_inv volume. On the other hand, on the color contours (at t = 533 msec) , C1 is also situated at the r_inv surface, but the emission is strong and clearly penetrates within this volume. Because the vertical position of the emission in this region appears to be exactly on the midplane, where the magnetic field lines are pointing directly at the color camera, the implication is that the emission from this region is very peaked (which is contrary to our hypothesis for B2, B3) .

Should these conjectures prove correct, the further implication (see Section 7.2 below) would be that the emission at the r_q surface (B2, B3, C3) may be associated with relatively low-energy runaway electrons, whereas the emission from Cl (at the r_inv surface) would be associated with highly relativistic runaway electrons.

It is, of course, entirely possible that our hypothesis that the crescents are associated with an r_q surface significantly inside the inversion radius is incorrect. In that case, an alternate explanation for the observed crescents must be found.

5.4 Anomalous shot #2256
This shot was characterized by an average plasma current of 174 KA (decreasing slowly), an average plasma density DENS = 1.61xl0¹⁹ m^-3 at the time of maximum URG emission at 560 msec (increasing from 0.52 to 1.75xl0¹⁹ m^-3 between 200 and 500 msec, then decreasing slowly) and hard X-ray emission of 49 mrem. The
16

unusually small value of the major radius R_MAJ = 759 mm (decreasing from 816 mm at 275 msec to 751 mm at 560 msec) resulted in a minor radius of 200 mm (since the inboard limiter was at 559 mm). We also note that the average loop voltage was unusually high at 1.89 volts (compared to 1.56 volts for #2326).

Although the image of shot #2256 was about 2.2 times brighter than #5040, the overall spatial distributions for the two shots as shown in Fig. 2 appear quite similar. However, the region of maximum URG emission is located about 109 mm further inside for #2256 (i.e. 648 mm compared to 757 mm for shot #5040) .

Thus, the relevant conditions which appear to be conducive to enhancing the URG emission are the low initial density to allow acceleration of large numbers of runaway electrons, an increase of the electron density above the critical density to allow growth of the URG mechanism and finally a decrease of the major radius to force the plasma inwards into regions of higher magnetic field.

5.5 General Results
In addition to the specific results described above, we list several other results which appear to be relevant:
(1) As seen on the b/w images, the URG emission commences as a hollow oval (presumably at an r_q = 65 mm surface) which progressively fills up inward with time.
(2) For stable URGs, the peak in the URG emission is always to the inside of the major radius: 99±12 mm for window C1 and 95±4 mm for window B1 (for shots #2253, 2326, 5040).
(3) For unstable URGs, the emission is displaced even further inside (142 mm for #5043 compared to 111 mm for #5040) and is, of course, much weaker.
(4) As observed with the b/w camera situated at h = -125 mm, the URG emission seldom extends below h = -88 mm and never below h = -125 mm.
(5) As observed with the color camera situated at h = +5 mm for shot #5040, the URG emission does not extend above 86 mm. For shot #2326, camera at h = +53 mm, the cut off is at about 96 mm. On the other hand below the midplane, the URG emission appears to extend down towards h = -240 mm.

In the final analysis, it is not clear whether the asymmetry of the spatial distribution as observed with the two cameras is real or is the result of the non-isotropic angular emission as suggested in Section 5.3.
17

6. TOTAL ENERGY OF URG EMISSION
At the critical density, the background from 400 to 1100 nm on the URG1 signal is about 20 mV of which H_α typically comprises 13%. Spectral lines of carbon and oxygen impurities dominate the background spectrum and substantially less than half of the background signal is due to bremsstrahlung radiation. We note that the 600 mV URG emission on shot #5040 is about 200 times brighter than the H_α emission viewed tangentially.

Since the measured URG1 signal amplitude is an integral of the product of the radiant sensitivity of the S1722-02 photodiode and the exponential URG spectrum determined in Section 4, it is possible to determine the URG energy as a function of wavelength. Thus, for a 180 mV URG1 signal (corresponding to a just-saturated 255 grey scale image on the b/w camera, a 22 grey scale image on the color camera and a 155 grey scale on the observed URG spectrum between about 650 and 750 nm) , we find that the URG power as a function of wavelength, X, is given by E(X) =1.98x10^-4 exp(5.80x10^-3 X)

Thus, at the H_α wavelength the power of URG emission is E(656.3 nm) = 8.92x10^-3 W/ (str.m².nm) and that the integrated power up to 1100 nm is 20.2 W/(str.m²).

The value of 8.92x10^-3 W/(str.m².nm) at the Hα wavelength obtained above was (perhaps fortuitously) confirmed to an accuracy of 5% by a direct comparison of the grey scale intensities between a tungsten ribbon lamp and the URG spectrum as recorded with the XYBION camera plus monochromator combination.

On the other hand, analysis of the b/w video images indicates a power flux in the Hα wavelength region a factor of 2.2 larger than that obtained from the photodiode calculations. Although this could be due to an inaccurate value chosen for the pixel saturation current at 255 grey scale, the possibility remains that the URG power above 900 nm is not as large as suggested by the exponential variation determined in Section 4.

Determining the total power radiated as URG emission from the whole Tokamak is not straightforward because the emission is not isotropic. As suggested in Section 5.2 and Section 7.2 below, the radiation pattern could be in a hollow cone ranging from 2. 9° to 27° half-angle with corresponding solid angles of 0.008 to 0.68 steradians.

The typical URG has a toroidal volume of 0.121 m³. The photodiode gathers light from a cylindrical volume of 2.13x10^-3 m³. We thus estimate that the total radiated power corresponding to the 20.2 W/str.m² calculated above, ranges from 0.02 to 2 watts. This is very small compared to the total input power of 300 KW.

We note, however, that the energy flux further in the infrared above 1100 nm could be much greater than that measured here.
18

7. DISCUSSION
In a recent publication, continuum radiation in the infrared has been reported in TEXTOR [11]. The IR radiation observed from 3 to 6 μm is consistent with synchrotron radiation from runaway electrons of 25 - 30 MeV energy. Nothing could be seen with a normal CCD camera which was sensitive up to 1.2 μm. It is interesting to note that no emission was observed from the inboard side of the torus; which is exactly opposite to our case where no emission was observed from the outboard side of the torus. Many of the details presented are intriguingly similar to the URG emission described in this paper. Nevertheless, we feel that the radiation we observe is not due to synchrotron radiation, although it is certainly related to the presence of runaway electrons. In addition to the lack of horizontal polarization expected for synchrotron radiation, this conclusion is based on the inconsistency between the relatively short wavelength of the radiation we see, and the maximum energy of the runaway electrons which can theoretically be confined in TdeV.

7.1 Energy of runaway electrons
In their 1979 review paper Knoepfel and Spong [5] (hereafter referred to as K&S) summarize the experimental and theoretical understanding concerning runaway electrons in toroidal discharges. We shall calculate various parameters relevant to our conditions using formulas presented in this paper.

At the critical density, INT3(crit.) = 0.97x10¹⁹ m^-3, the loop voltage, V_L = 1.58 volts, such that the electric field accelerating the electrons is E = V_L/2πR = 0.290 V/m. This is substantially below the critical field as defined by Dreicer [6]
E_c = 0.267x10^-16 n(lnΛ)/T_e =2.58 V/m (n in m^-3, T_e in eV, lnΛ = 15)
at which most of the 1300 eV thermal electrons would run away. Thus, only electrons in the high energy tail with parallel energies greater than the critical runaway energy, W_c = (E_c/E) (kT_e/2) =5.8 KeV will run away. Experimental evidence for this is seen in Fig. 7 (b) .

As these electrons gain energy, their drift surfaces shift outward, by an amount depending on the energy. K&S give a rough estimate for the maximum confined energy as that which produces a radial shift of a/2; in our case this gives a maximum energy of 25.6 MeV. For the particular case of shot #5040 with R_MAJ = 868 mm, R_LM2 = 1152 mm, we calculate that the energy of the runaways causing the URG emission at R_URG = 757 mm cannot exceed 19.2 MeV.

Wong [21] has calculated the outward drift of the accelerating runaway electrons due to the toroidal electric field, from which we can calculate the maximum energy of the electrons from the time for the electron to drift outward to the wall. Using our experimental parameters, we calculate the maximum energy to be 15.6 MeV.

In addition, we have a clue to the electron energy from the time delay between the onset of the soft X-ray signal and that of the hard X-ray signal. For a loop voltage of 1.58 volts in TdeV and assuming free acceleration, the minimum time required to increase
19

the energy of the runaway electrons by 5.11 MeV (γ = 10) is 58.7 msec. We note that both the PHA signal (Fig. 7(a)) and the soft X-ray signals (SX309, Fig. 5; SX609, Fig. 9(a)) in the energy range from 2.58 to about 10 KeV approach their maximum values by t = 160 msec. This obviously implies that substantial quantities of high energy electrons above 5.8 KeV exist at this time. On the logical assumption that most of the runaway electrons are drawn from this pool starting at 160 msec, their energy by 380 msec (when the XDUR3 signal is maximum) would be 19.2 MeV. Although by 525 msec (when the URG emission approaches saturation) their energy could theoretically be as high as 31.8 MeV, they should already have been lost by the above-mentioned processes.

Examination of the stainless steel support structure 5 cm behind the outboard graphite limiter have shown the presence of transmuted nuclei of Ni, Cr and Mn requiring threshold energies of up to 12.2 MeV [2] . Nuclei with transmutation energies greater than 19.6 MeV [7] were not detected. Thus, there are good reasons for believing that the maximum electron energy is somewhat below 19 MeV.

On the other hand, it is not at all clear that the URG emission is associated with runaway electrons at these high energies. Measurements on PLT [7] indicate an exponentially decreasing runaway electron distribution, N = (1x10⁶ cm^-3) exp (-E/3.2 MeV), with a logarithmic slope of [3.2 MeV]^-1. If a similar relation were valid for TdeV, there would be 120 times as many runaway electrons at 5.11 MeV (γ = 10) than at 20.44 MeV (γ = 40). For a decrease in energy to 186 KeV, there would be a further increase by a factor 4.7. The dilemma, of course, is whether the URG emission is associated with a relatively large number of low energy runaway electrons , or with a very small number of high energy (γ = 40) runaway electrons.

7 .2 Implication of 27° half angle cone
The conjecture in Section 5.3 that URG emission is being radiated from an r_q = 65 mm surface into a cone of half angle 27° may presumably be related to the energy of the runaway electrons associated with the URG emission. Jackson [8] , describing the angular distribution of radiation emitted by an accelerated charge, gives an equation to determine the angle for which the intensity is a maximum:
θ(max) = cos^-1(((l + l5β²)^0.5-l)/(3β))
where β =v/c is the relativistic factor and it is assumed that the velocity and acceleration are parallel.

For θ(max) = 27°, we find β = 0.681 which corresponds to weakly-relativistic 186 KeV electrons.

Conversely, for relativistic 5.11 MeV electrons (γ = 10), θ(max) --> 1/(2*γ) --> 2.9°
and Jackson points out that the maximum intensity varies as γ. Such relativistic electrons could conceivably be associated with the URG emission emanating from window C1 in Fig. 12(d).

This highly speculative scenario, thus, implies that the whole runaway spectrum from 186 KeV to 20 MeV may be involved in the URG
20

emission. However, Jackson [see eq'n 14.29] warns that radiation losses for linear acceleration are normally extremely small; they are completely negligible in linear accelerators.

7.3 ω_pe and ω_ce at n(crit.)
Based on experimental observations, K&S identify three regimes of Tokamak operation as (1) nearly thermal, (2) intermediate and (3a) slide-away/(3b) runaway beam. Normal high density discharges are in the nearly thermal regime where ω_pe/ω_ce > 1, whereas in the other two regimes ω_pe/ω_ce < 1.

To help classify these regimes further, K&S define the parameter ξ as the ratio of the electron drift velocity to the electron thermal velocity. Transition from the intermediate to the slide-away regime defines a critical value of ξ called ξ_c. The value of ξ_c appears to be somewhat machine dependent: for Alcator, ξ_c = 0.2 - 0.4 (in hydrogen); whereas for TM3, typically ξ_c = 0.1.

For shot #5040, with I_p = 185 KA, a = 0.24 m, n_e = 0.84xl0¹⁹ m^-3 and T_e =1300 eV, we calculate a drift velocity of 7.6xl0⁵ m/s and a thermal velocity of 2.14xl0⁷ m/s, such that ξ = 0.036.

Based on these criteria the discharges associated with URG emission clearly fall into the intermediate regime and n(crit.) perhaps signals a transition into the slide-away regime. K&S claim that the intermediate regime is characterized by enhanced emission at ω_pe and ω_ce, which would indicate the presence of strong fluctuations. This regime presumably incorporates the steady runaway regime in which "runaway electrons are formed continuously in the hot central part of a normal, moderately low density discharge and gradually build up a high-energy runaway tail in the electron distribution function. The steady-state nature of this regime implies a loss of runaways above a certain energy (e.g. due to uncontained orbits) counterbalanced by a source of low-energy runaways."

A more recent review paper on current-driven turbulence by de Kluiver, Perepelkin and Hirose [9] (hereafter referred to as KPH) provides a slightly different classification scheme in terms of ω_pe/ω_ce and E/E_c as proposed by the URAGAN 2 stellarator team [10] . In this classification scheme the URG mechanism would presumably fall between the "accelerative regime" (0.02 < E/E_c < 0.1, ω_pe/ω_ce < 1) and the "restrictive runaway regime" (0.1 < E/E_c < l, ω_pe/ω_ce < 1) . Of particular relevance to us is the observation of KPH that high energy runaway electrons are preferentially observed at relatively low values of E/E_c because at high values the runaway electrons are dissipated by instabilities.

It is highly likely that the observed critical density, n(crit.), corresponding to INT3 = 0.97xl0¹⁹ m^-3 is related to the electron plasma frequency, ω_pe, and the electron cyclotron frequency, ω_ce. For shot #5040, this relationship is illustrated in Fig. 13, where we sketch the electron density profile, n_e, the URG profile and the magnetic field as a function of major radius.

The n_e profile (with enforced cylindrical symmetry) is obtained from a deconvolution of the six INT1 to INT6 signals of
21

the laser interferometer measuring the line averaged densities between 688 to 1048 mm major radius [3] . Although we have confirmed that the density profile at these low densities is generally peaked, the pronounced dip at 750 mm in Fig. 13 is unlikely to be real and appears to be due to consistently low values of INT2 not being counterbalanced by consistently high values of INT5. It is perhaps more than coincidence that the position of the dip coincides with the approximate position of maximum URG emission.

At any rate, using a value of n_e = 0.84xl0¹⁹ m^-3, the plasma frequency is ω_pe = e (n_e/ε_em_e)^0.5 = 1.63xl0¹¹ rad/sec. The magnetic field varies inversely with R giving B = (868/757) *l .42 Tesla = 1.63 Tesla at 757 mm, such that the cyclotron frequency, ω_ce = eB/m_e = 2.87xl0¹¹ rad/sec. Thus, at the critical density, ω_pe/ω_ce = 0.57.

We are aware of three other relevant cases where similar ratios of ω_pe/ω_ce were observed. In TEXTOR (R = 1.75 m, B_T = 2.0 T, n_e = 1xl0¹⁹ m^-3) infrared emission was observed [11, see above] at a major radius of about 1.975 m where B_T = 1.77 Tesla. We thus calculate ω_pe/ω_ce = 1.78x10¹¹/3.11xl0¹¹ = 0.57, which is exactly the same as in our case.

In the second case, involving TEXT (R = 1.0m, B_T = 2.0 T, n_e = 2xl0¹⁹ m^-3) enhanced microturbulence was observed at -8 cm relative to the major plasma radius [12] . Thus, for R = 0.92 m, B_T = 2.17 T, we calculate ω_pe/ω_ce = 2.52x10¹¹/3.82xl0¹¹ = 0.66. This is within the error bounds associated with our value of 0.57 (from 0.5 to 0.8) .

In the third case involving URAGAN 2 (R=1.10 m, B_T = 0.7-2.0 T, n_e = 0.2-0.6xl0¹⁹ m^-3) , enhanced fluctuation intensities at ω_pi and ω_pe as well as X-rays from probes placed at rational q surfaces were observed for n_e = 0.2xl0¹⁹ m^-3, ω_pe/ω_ce = 0.625 [13]. KPH state: "This observation clearly confirms the tendency of runaway electron concentration on magnetic rational surfaces and may have important implications for runaway dynamics in a sheared magnetic field." Similar results were observed by Cheetham [22] in LT-4. As noted previously, URG emission in TdeV appears to commence on the q=l surface.

In a fourth case involving TFR (R = 0.98 m, B_T = 3-4 T, n_e = 0.5-l.0xl0¹⁹ m^-3) , a marked change in the character of the hard X-ray emission on either side of a critical density was observed [23]. The published hard X-ray signals appear very similar to those illustrated in Fig. 5 and discussed in Section 3.2. However, their value of ω_pe/ω_ce = 0.25 is more than a factor 2 smaller than in our case.

Theoretical non-relativistic calculations of runaway tails in magnetized plasmas [14] "have identified four regions in parameter space (ω_pe/ω_ce, E/E_c) for the existence of runaway-induced instabilities. The authors conclude that if E < 0.1E_c and ω_pe < ω_ce, the runaway tail should be stable" [KPH]. On Fig. 6 of their paper ω_pe/ω_ce = 0.57 corresponds to E/E_c = 0.06 and signals a transition from a marginally stable regime (region IV) into region III where the instability is triggered via the anomalous Doppler effect. Although in section 7.1 we have calculated a value of E/E_c = 0.290/2.58 = 0.11 which is almost twice as large as that obtained above, this may be due to the non-relativistic treatment or,
22

perhaps, to the factor of two discrepancy between the Dreicer field and the critical field as discussed by K&S.

7.4 Synchrotron radiation
Synchrotron radiation is often emitted from toroidal devices in which electrons are accelerated to relativistic energies. This is especially true of betatron and synchrotron devices [15,16,17] .

A theoretical semilog plot of the synchrotron intensity as a function of normalized wavelength, X/X_c, is shown in Fig. 14, curve (a) (multiplied by a factor 100 for visual convenience) . At the shortest wavelengths the variation is nearly exponential but flattens out at longer wavelengths. The critical wavelength, X_c, originally defined by Jackson [8] in terms of a critical frequency,ω_c, above which there is negligible synchrotron radiation, is given by
X_c = 2πc/ω_c = 2πρ/(3γ³)
where ρ is the radius of curvature of the orbit and γ is the relativistic factor (1-v²/c²)^-0.5. For normal synchrotron radiation one would choose ρ = 0.866 m (the major plasma radius) such that
X_c = 1.814γ^-3.

There are two conflicting criteria which restrict the location of the exponential spectrum determined in Section 4 on curve (a) of Fig. 14. These are the maximum energy of the runaway electrons which can theoretically be confined in TdeV (25 MeV, according to Section 7.1) and the maximum power which can conceivably be emitted as synchrotron radiation (300 KW, ohmic power). For example, the dotted straight line superposed on curve (a) , which has been arbitrarily fitted at X/X_c = 0.1667 corresponding to γ = 49, 25 MeV electrons, limits the maximum position to the right. However, if we take the power radiated below 1100 nm for a typical URG to be 0.3 watts (see Section 6) , we find that the total radiated power integrated over all wavelengths to be of the order of 10¹¹ Watts. In order to reconcile this energy criterion, runaway energies greater than 31 MeV are required.

Obviously, in TdeV synchrotron radiation in the normal sense cannot be used as an explanation for the URG emission. Finken et al. [11] claim that in TEXTOR the radius of curvature of the runaway electron orbits is on the average smaller than the major radius (by as much as a factor of 2) for runaways which have a significant perpendicular velocity component. Even this artifice appears insufficient in our case, particularly since the URG emission is emitted on the inboard side of the major radius rather than on the outboard side as in TEXTOR.

It is perhaps conceivable that synchrotron-like ω_ce radiation could be emitted as a result of the parallel velocity component of the runaway electrons being transformed into a perpendicular component by some mechanism such as the anomalous Doppler instability. However, it is not clear how such a mechanism could be reconciled with the observed spatial distribution of the URG emission.
23
7.5 Scattering from fluctuations
In recent years many workers have proposed strong turbulence as a source of the anomalous diffusion or transport across the magnetic field lines in Tokamaks. Their theories are developed in terms of solitons, Langmuir wave packets and/or vortices [18,19,20] .

Recent work on relativistic electron beam-plasma experiments [19] in which pulsed 1 MeV electron beams are scattered from magnetized plasmas (1 - l0xl0¹⁹ m^-3, 50-90 eV, 0.1 T) indicate the presence of localized soliton-like wave packets such as predicted by theories of strong turbulence. According to the theory, the relatively large wave packet becomes decoupled from the background turbulence and continues collapsing until, at a diameter of about 16 Debye lengths (a = (16±5)λ_D), it becomes relatively stable. The dipole axis of the wave packet is aligned with the magnetic field and spectroscopic measurements of Stark shifts indicate electric fields within the soliton of the order of 10⁶-10⁷ V/m.

A solution for the radiation power and frequency spectrum of a relativistic electron beam passing through a strongly turbulent plasma has been developed by Weatherall [18]. Regions of intense, localized electrostatic fields in the plasma are characterized as dipolar "solitons" or other electrostatic fluctuations. It is further suggested that beam bunching would shift emission of coherent radiation to frequencies higher than the dipole oscillation frequency.

This "fluctuation" spectrum obtained by Weatherall is shown in Fig. 14 as curves (b). "Perpendicular" and "parallel" refer to the orientation of the dipole moment of the soliton relative to the velocity vector of the runaway electron. The wavelength is normalized according to
X_c = 2πc/ω_c = πaγ^-2
where a is the diameter of the soliton and γ is the relativistic factor as before. Since the Debye length for our plasma is given by λ_D = 7430 (T_e/n_e)^0.5 = 9.24x10^-5 m (T_e = 1300 eV, n_e = 0.84xl0¹⁹ m^-3) , we expect a soliton diameter a = 16λ_D = 1.48x10^-3 m, such that X_c = 4.65x10^-3γ^-2 .

The dotted straight line on the "parallel" curve (b) of Fig. 14 represents the URG spectrum of Section 4 arbitrarily fitted at X/ X_c = 0.25. This particular point corresponds to γ = 18.35 or 9.4 MeV electrons. In the range of X/X_c from 0.222 to 0.333, the energy of the runaway electrons varies from 7 to 15 MeV, which is within the range attainable in TdeV.
24

8. CONCLUSIONS
Highly directional, visible continuum radiation has been observed from low density discharges in TdeV. The experimental observations have clearly demonstrated:
(1) the existence of a critical density at which the URG emission is maximized; for densities above and below this point the discharge changes its characteristics,
(2) the exponential nature of the continuous emission spectrum maximizing somewhere in the infrared above 900 nm, and
(3) the emission is spatially limited to the inboard side of the major radius of TdeV and appears to commence at the r_q = 65 mm magnetic surface.

Normal synchrotron radiation does not appear to be an appropriate explanation of the observed emission. We propose the following general scenario for the mechanism generating URG emission. The runaway electrons are continuously accelerated by the toroidal electric field. Near the critical plasma density, ω_pe/ω_ce = 0.57, an instability (such as the anomalous Doppler instability) generates electrostatic fluctuations. These appear to start at the q=l magnetic surface. "Bunching" of the runaway electrons is a distinct possibility. The electrons which scatter from these fluctuations give rise to "bremsstrahlung" or "synchrotron" radiation. It is not clear whether the radiation is from high energy (19 MeV) or weakly-relativistic (186 KeV) electrons.

The "scattered" runaway electrons either strike the limiters, giving rise to enhanced hard X-ray spikes, or are thermalized, giving rise to a high energy tail in the electron velocity distribution and an increased electron temperature. Finally, these electrons transfer their energy to the ions via collisions.

Theoretical work to explain the details of the URG mechanism is required. On the experimental side, two issues are crucial: the wavelength at which the URG emission is maximum, the energy and orbits of the runaway electrons responsible for the emission. Finally, efforts to detect URG emission on other Tokamaks should be made.

ACKNOWLEDGEMENTS
The authors would like to thank the operating and support staff of TdeV for making the measurements possible. The principle author is especially indebted to Louis Pelletier for maintaining the camera systems and to Dr. Magdi Shoucri for theoretical discussions.

The Centre canadien de fusion magnetique is a joint venture of Hydro-Quebec, Atomic Energy of Canada Limited, the Institute national de la recherche scientific, Universite de Montreal, MPB Technologies Inc. and Canatom Inc. It is principally funded by AECL, Hydro-Quebec and INRS-Energie.
25

REFERENCES
[I] ZUZAK, W.W., URG files and report thereon in possession of principal author.
[2] BOLTON, R.A., and TdeV Team, IAEA-CN-50/A-VII-15 (1989) 495
[3] LACHAMBRE, J.L., private communication
[4] DECOSTE, R., NOEL, P., SPIE 661 Optical Testing and Metrology (1986) 50
[5] KNOEPFEL, H. SPONG, D.A., Nuclear Fusion 19 (1979) 785
[6] DREICER, H., Phys. Rev. 115 (1959) 238; 117 (1960) 329
[7] BARNES, C.W., STAVELY, J.M., STRACHAN, J.D., Nuclear Fusion 21 (1981) 1469
[8] JACKSON, J.D., Classical Electrodynamics, Wiley, New York (1975) 663, eq'n 14.40
[9] KLUIVER, H. de, PEREPELKIN, N.F., HIROSE, A., Phys. Reports 199 (1991) 281
[10] VOLKOV, E.D., PEREPELKIN, N.F., SUPRUNENKO, V.A., SUKHOMLIN, E.A., Collective Phenomena in Current-Carrying Plasmas (Naukova Dumka, Kiev, 1979) (English transl. Gordon and Beach, New York, 1985)
[11] FINKEN, K.H., WATKINS, J.G., RUSBULDT, D., CORBETT, W.J., DIPPEL, K.H., GOEBEL, D.M., MOYER, R.A., Nuclear Fusion 30, 859 (1990)
[12] BROWER, D.L., YU, C.X., LI, W.L., PEEBLES, W.A., LUHMANN, N.C., Bull. Am. Phys. Soc. 34 (1989) 2153
[13] PETVIASHVILI, V.I., PEREPELKIN, N.F., SUPRUNENKO, V.A., VASIL'EV, M.P. and KULAGA, A.E., Sov. Phys. JETP 52 (1980) 421
[14] MOGHADDAM-TAAHERI, E., VLAHOS, L., ROWLAND, H.L., PAPADOPOULOS, K., Phys. Fluids 28 (1985) 3356
[15] IWANENKO, D., POMERANCHUK, I., Phys. Rev. 65, (1944) 343
[16] ELDER, F.R., LANGMUIR, R.V., POLLOCK, H.C., Phys. Rev. 74, (1948) 52
[17] TOMBOULIAN, D.H., HARTMAN, P.L., Phys. Rev. 102, (1956) 1423
[18] WEATHERALL, J.C., Phys. Rev. Lett. 60 (1988) 1302
[19] ROBINSON, P.A., NEWMAN, D.L., Phys. Fluids B 2 (1990) 3120
[20] JOVANOVIC, D., SHUKLA, P.K., ANGELIS, U. de, HORTON, W., Phys. Fluids B 3 (1991) 45
[21] WONG, K.L., Leakage of Runaway Electrons from Tokamaks, PPPL 1875 (February 1982)
[22] CHEETHAM, A.D., HOW, J.A., HOGG, G.R., KUWAHARA, H., NORTON, A.H., Nuclear Fusion 23 (1983) 1694
[23] EQUIPE TFR, Nuclear Fusion 16 (1976) 473
[24] VON GOELER, S., STEVENS, J., BERNABEI, S., Nuclear Fusion 25 (1985) 1515
[25] KADOMTSEV, B. B., Sov. J. Plasma Phys. 1 (1975) 389
26

Figure Captions
FIG. 1. Top view of TdeV indicating position of cameras, photodiodes and other diagnostics: 32 = PHA, 33 = soft X-rays (SX309, SX609), 34 = hard X-rays (XDUR3), 24 = microwave interferometer (DENS), 56 = SMM interferometer (INT1 - INT6). Note the clockwise direction of B_T and the electron drift.

FIG. 2. Color video prints of 3 URGs:
(a) Shots #2256 and (b) #2326: Camera in Bay 2 (h = +53 mm) looking toward Bay 7 . Note the beam dump for ω_ce on inside wall opposite Bay 3 on left of image and magnetic coils at outside wall radius in Bay 6 on right of image. The black spots in the URG image are associated with broken fibres in the image guide, (c) Shot #5040: Camera in Bay 5 (h = +5 mm) looking toward Bay 10.

FIG. 3. Emission looking upstream (URG1) and downstream (URG2) the runaway electrons for (a) stable URG #5040 and (b) unstable URG #5044.

FIG. 4. Peak amplitude of URG1 signals plotted as function of integrated line intensity at 830 mm, INT3, for pool B (shots #5033 to #5058) . The vertical dashed line defines the critical density, INT3(crit.) = 0.95x10¹⁹ m^-3 at which URG1 (#5040) = 0.6 volts is a maximum.

FIG. 5. Hard X-rays superposed on soft X-ray signals, SX309, indicating (a) transition from unstable to stable regime (#4432),
(b) unstable regime (#4433), (c) stable regime (#4434) and on expanded scale from 640 to 720 ms illustrating hard X-ray (d) "bursts" (#4433) and (e) "spikes" (#4434).
[FIG. 5(a)-(c)] [FIG. (d)-(e)]

FIG. 6. Plasma parameters as function of integrated line intensity, INT3.
(a) Loop voltage, V_L = (0.22x10^-19 V/m^-3) INT3 + 1.366 V for INT3 > 0.97x10¹⁹ m^-3 and V_L = (2.0x10^-19 V/m^-3) INT3 -0.360 V for INT3 < 0.97x10¹⁹ m^-3.
(b) Sawtooth period, Δτ = (0 .166x10^-19 ms/m^-3)INT3 + 1.319 ms
(c) Inversion radius, r_inv: apparent (x) versus deconvolved (*...*)
(d) Electron temperature, T_e = (-200x10^-19 eV/m^-3) INT3 + 1544 eV

FIG. 7. PHA data for unstable URG #5044.
(a) Photon count as function of time. Note transition into unstable regime at just before 400 ms.
(b) Electron temperature in stable (1300±79 eV) and unstable regimes (977±74 eV)

FIG. 8. Typical diagnostic signals between 200 and 700 ms: URG1 (URG emission), PA6 (photoamplifier measuring H_α 656.3 nm but also picking up intermediate X-rays), INT3 (line averaged electron density at R = 830 mm), R.MAJ (major radius, R_MAJ, as measured with magnetic flux coils), IP (plasma current, I_p) .
27

FIG. 9. Typical diagnostic signals illustrating X-ray signals for shot #5413. (a) URGl (URG emission), SX609 (soft X-rays in vertical direction) , PA4 (photoamplifier fitted with C1 909.5 nm filter but picking up "intermediate" X-rays), XDUR3 (hard X-ray signal). (b) Expanded scale from 500 to 580 ms illustrating "step-jumps" and hard X-ray spikes.
[FIG. 9(a)] [FIG. 9(b)]

FIG. 10. Comparison of URGl photodiode signal with URG intensity (0 - 255 grey scale) obtained with XICAS3 image analysis system using 6 windows on the URG spectrum for shot #5410, 7200 angs. setting.

FIG. 11. Exponential fit to URG spectrum from 500 to 900 nm obtained by normalizing the intensities as in Fig. 10 with a calibration black body spectrum.

FIG. 12. Contour plots of URG emission for camera positions and parameters tabulated below. In (b) , the r_inv = 85 mm and the a = 240 mm plasma radius are superposed. In (c) and (d) the r_q = 65 mm
surface and 3 windows are superposed.
***********************************************************************
Fig Shot#    Lens Bay    h   R R_MAJ D   deg./pix.    mm/pix.
12          (mm) (mm) (mm) (mm) (mm) hor. vert. hor. vert.
***********************************************************************
(a) #1396 b/w 25 B13.B    -125 1824    816    1631    0.0524    0.0624 1.49    1.78
(b) #5040 c 12.5 B5C      +5 1466    868    1181    0.0888    0.1005 1.83    2.07
(c) #2326 b/w 25 B13.B    -125 1824    823    1628    0.0524    0.0624 1.49    1.77
(d) #2326 c 12.5 B2C.LU +53 1674 823 1458 0.0876 0.1002 2.23 2.55
***********************************************************************
[FIG. 12(a)] [FIG. 12(b)] [FIG. 12(c)] [FIG. 12(d)]

FIG. 13. Profiles of URG intensity, B_T and n_e across plasma cross section for shot #5040 with R_MAJ = 868 mm. Note ω_pe = 1.63X10¹¹ rad/sec and ω_ce = 2.87x10¹¹ rad/sec for the URG maximum at R_URG = 757 mm.

FIG. 14. Theoretical spectra fitted to experiment.
(a) Synchrotron emission: abscissa normalized to 3γ³X/2πρ
(b) Weatherall theory: abscissa normalized to γ²X/πa The dotted straight lines correspond to the exponential URG spectrum determined in Section 4 .
28

[Figures 1 to 14 were scanned at 200 dots/inch and saved as pdf files TdeV-URG31.pdf to TdeV-URG50.pdf . Links thereto in the Figure Captions page and in the text above are provided.]

[Will Zuzak; 01Dec2013: The above report was scanned page-by-page using "ABBYY FineReader 5.0 Sprint Plus" OCR software to create a Microsoft Word 97 .rtf file; copied and pasted into a Microsoft Notepad .txt file (to remove the wasteful formatting); then copied and pasted into KompoZer software to create this .html file. A Greek keyboard was added and utilized to insert the Greeek symbols. Subscripts and superscripts were added manually to the source code using <sub>...</sub> and <sup>...</sup>. Although most of the OCR conversion errors and spelling errors in the original text have been removed, it is highly likely that many errors remain. Apologies to the reader.]

Centre canadien de fusion magnetique Tokamak de Varennes

Centre canadien de fusion magnetique
Tokamak de Varennes