I
GENERATION OF RF DRIVEN CUR RENTS BY LOWER-IIYBRID
WAVE INJECTION IN THE VERSATOR II TOKAMAK
S.C. Luckhardt, M. Porkolab, S.F. Knowlton, K-I. Chen,
A.S. Fisher, F.S. McDermott, and M. Mayberry
Massachusetts Institute of Technology
Department of Physics, Research Laboratory of Electronics
and Plasma Fusion Center
Cambridge, MA 02139
PFC/JA-81-14
July, 1981
By acceptance or this article, the publisher and/or recipient acknowledges the U.S. Government's right to retain a non-exclusive,
royalty-frec license in and to any copyright covering this paper.
GENERATION OF RF DRIVEN CURRENTS BY LOWER-HYBRID
WAVE INJECTION IN THE VERSATOR II TOKAMAK*
by
S. C. Luckhardt, M. Porkolab, S. F. Knowlton, K-I Chen,
A. S. Fisher, F. S. McDermott, and M. Mayberry
Massachusetts Institute of Technology
Department of Physics, Research Laboratory of Electronics, and
Plasma Fusion Center
Cambridge, MA 02139
ABSTRACT
Experimental verification of lower-hybrid RF current drive in the Versator II tokamak is presented. The
experiments show that efficient current drive exists only in low density discharges (ffe < 6 X 1012 cm- 3 ) in
the "slide-away" regime where a preformed supra-thermal electron tail exists prior to the application of the RF
power.
*Work supported by the U.S. Department of Energy.
1
In tokamak-type plasma confinement devices the confining toroidal current is induced in the plasma by
means of a pulsed ohmic-heating transformer'. From the engineering point of view a pulsed tokamak reactor is
unattractive, and therefore ways must be found to drive the toroidal current continuously by other means, such
as neutral beams2 or radio-frequency (RF) power 3. Lower-hybrid travelling waves, launched by a phased array
of waveguides, have recently been proposed as an attractive means of efficiently generating toroidal currents in
tokamaks4'5 .
At low densities and temperatures in both a linear device6 and a torus 7 it was shown that modest currents
could be generated (a few amps with a few hundred watts of RF power) with unidirectionally launched lowerhybrid waves. In lower-hybrid injection experiments in tokamaks changes in the plasma loop voltage have been
reported and identified with current drive at relatively low densities (1012
_
0'l cm-3)8-
2
. In some of these
experiments the inferred current was attributed to generation of runaway electrons 8 "0 "',1. However, in the JFT2 experiment9 good agreement with the Fisch theory4 was claimed even though the rf pulse length, tRF, was
much shorter than the L/R time, t L/R, so that observation of an RF induced change in the toroidal current was
not possible. Even though in the JFT-2 experiment a weak phase dependence of the loop voltage was found,
changes in the loop voltage due to surface currents, current profile modifications, electron heating, and runaway
production cannot be excluded. Therefore, the agreement with theory4'5 must be regarded as fortuitous rather
than proven. In the present paper we present the first unambiguous experimental demonstration of a lowerhybrid wave driven current in a tokamak (which is possible in the Versator experiment since tL/R
<
tR). It is
also shown that under normal circumstances current generation can be obtained only in the slide-away regime' 2
where a tail on the electron distribution exists even before the RF power is applied13 . Thus the present results
show that the Fisch mechanism of efficient lower-hybrid current drive (i.e., raised plateau formation) is not
the dominant mechanism in these experiments. Furthermore, all the experimental results reported until now
operate in this low density "slide-away" regime; therefore, at high density some mechanism must prevent the
formation of the raised plateau, which is a prerequisite for efficient current generation4 . It is likely that an
explanation of the lower-hybrid current drive results observed so far will require a detailed transport code
which includes both a runaway or slide-away tail, a Fokker-Planck collision calculation, radial transport and a
realistic model of wave absorption and heating in a tokamak discharge.
The experiments were carried out on.the MIT Versator 11 tokamak: R, = 40 cm, a. = 13 cm, Iol =
30 kA, tpe = 30 msec, f, = 2 X 1012 - 3 X 10 "cm,
2
Ti, = 120 eV, T,, = 200 - 350 eV. The time
dependence of the density is controllable by gas puffing. The ohmic heating power is supplied by a 100 kjoule
capacitor bank. The stainless steel vacuum chamber walls are conditoned by low density hydrogen discharges13
and by periodic localized titanium deposition which yields Zeff = 2 discharges.
The RF source is an 800 MHz, 150 kW klystron with a modulator controlled pulse duration of 10 msec,
or 30 msec at reduced power. The power is split into four or six well isolated channels and fed through
continuously variable phase shifters into a four or a six waveguide grill. Present results were obtained using the
six waveguide grill which has 0.8 cm gaps separated by 0.17 cm thickness walls, and 25.4 cm height. Prior to
installation the stainless steel grill was baked in a vacuum oven at 350* C and during operation frequent RF
processing, and periodic titanium coating ensured operation with up to 50 kW of total RF power (0.5 kW/cm 2 )
without RF breakdown or plasma formation in the guides. The grill is normally positioned 1 cm behind
the limiter and changing the relative phase between adjacent waveguides, A#, from 600 to 120' moves the
antenna spectrum peak from Nz = 4 to 11 with ANz/Nz
~ 0.8. If toroidal effects are taken into account16
with Te0 ~ 250eV the waves should always damp on electrons in the regime neo < 1.5 X 1013 cm-
3
Furthermore, collisional absorption on the cold surface plasma accounts for no more than 25% of the injected
power. Scattering on drift waves at the plasma surface does not significantly modify these results' 6 .
The basic effect associated with the injection of lower-hybrid power is shown in Fig. la. Total current and
loop voltage signals are superimposed for two discharges with and without RF power injection. During the RF
pulse there is a large increase in the total current above that due to ohmic heating only. The difference between
the two currents, AIt, exceeds 10 kA at maximum..The decrease in the total current after about 15 msec results
from a loss of radial equilibrium caused by vertical field supply limitations. For times between the electric field
diffusion time TDE (-.5 - 1 msec) and the L/R time TL/R (-10-15 msec) simple circuit analysis indicates that
the quantity Al/At is approximately proportional to the RF current where At is the RF pulse duration.
'The loop voltage signal initially decreases by .15 volts, then shows sudden spikes consisting of a rapid
increase followed by a slower decrease. It can be seen in Fig. la that these spikes occur at the same time
as sudden small decreases in the time derivative of the total'current, dIl/dt . These voltage spikes generally
become larger and more frequent at higher RF power. This voltage behavior has been reported in other lowerhybrid experiments' 0 ,1 . Also evident in Fig. la is an interesting time lag of several miliseconds between
initiation of the RF power pulse and the beginning of the RF driven increase in total current. Although at
present it is not well-understood, it is likely that the time delay results from a slow evolution of the slide-away
3
tail on the electron distribution function, and can usually be eliminated by applying the RF power later in
the discharge. The continuous slow drop in the loop voltage in both discharges is caused by limitations in the
capacitor driven ohmic heating power supply.
When a short RF pulse is applied at the time when the ohmically driven current is beginning to drop, the
total current can be maintained at a constant level, dIt/dt = 0 (Fig. 1b). The single turn loop voltage drops
through zero and is maintained at a small negative level, V
-
-0.1
volt for the duration of the RF pulse.
Interestingly, the loop voltage drop occurs rapidly, in approximately 100 pzsec; an RF current generated near
the magnetic axis is expected to cause loop voltage changes on the classical electric field penetration time scale,
=_ 1 msec. Therefore, it appears that in the case shown, Fig. 1b, the RF driven current extends, at least in part,
to the plasma edge. Alternatively, anomalous electric field penetration could be responsible. Independent of
electric field penetration effects, the observation of negative resistive loop voltage,with dI/dt = 0, eliminates
edge heating as an interpretation of the RF effects; increasing the plasma edge conductivity by heating could
cause a voltage drop, but clearly could not by itself produce a negative loop voltage. Such an explanation would
also be difficult in view of the observation of the large temperature drop (Fig. 3c) in the plasma center and the
observed increase in total current.
The incremental current, AIt, normalized to constant transmitted power shows a strong asymmetrical dependence on A# (see Fig. ic). The largest current increments are generated when A# = -90*
, correspond-
ing to a wave spectrum travelling in the direction of the electron ohmic drift. This value of the array phasing
launches a wave packet with a spectrum peaked at Nz = 8 with a width ANz = 6, and approximately 90% of
the power is directed along the drift direction with 10% going the opposite way. Interestingly, when waves are
launched opposite to the electron ohmic drift, A# > 0, no decrease in total current is observed: instead, only a
smaller current rise occurs.
The results indicated in Fig. 1 are consistent with the generation of a toroidal current in the plasma by
the travelling lower-hybrid waves. However, possible electron heating by the RF could result in conductivity
increases and it may also cause an increase in the toroidal current. To eliminate this possibility it is necessary to
measure the electron temperature with Thomson scattering, Figs. 2a, 3c.
The central electron temperature during RF injection with waveguide phasing optimized for current drive,
AO = -90'
is not found to increase, Fig. 3c; instead, there is a cooling trend due to a decrease in the ohmic
heating power, Fig. 3b. The cooling effect disappears for positive values of A# (anti-current drive injection)
4
and there may be a small temperature increase, Figs. 2a, b. Bulk electron heating is therefore eliminated as an
explanation of the RF induced changes in current and loop voltage. Transport code simulations of the Versator
current drive experiment have predicted a qualitatively similar electron temperature dependence on A014.
The profile changes in the electron temperature can produce, through conductivity increases, a modification
of the current profile and therefore an inductive voltage proportional to IdL/dt . However, such current
profile changes due to heating will occur on the electron energy confinement time scale, of the order 0.51 msec in Versator II. Voltage transients of one millisecond duration can be expected from this process. In
the experiments constant incremental.voltage drops for the duration of the RF pulse, 10-30 msec have been
observed and therefore cannot be explained as profile changes. A model circuit calculation can easily be made
to obtain an approximate value for the RF generated current. Considering the mutual inductive coupling
between the plasma current, RF driven current, and OH transformer a simple expression for the loop voltage
can be obtained: VL
-
R,(IT - IR - IRUN) + "IT where VL is the loop voltage, IT is the total plasma
current, IpR is the RF driven current, IRUN is the current carried by runaway or slide-away electrons, R, is the
Spitzer plasma resistance, and PIt is the inductive voltage component. Using this expression and the data in
Fig. 3a, we find IRp
~ 10 - 15kA; and under optimized conditions IR -
20 - 30kA has been obtained.
We shall now present experimental evidence which shows that RF current can be initiated only in a low
density regime characterized by the presence of an energetic electron tail before RF injection. The dependence
of AIt on the line average electron density at the beginning of the RF pulse is shown in Fig. 2c for a series of
deuterium discharges. The current rise, AIt is approximately independent of density for
Above this density, AI
ie <
6 X 1012 cm-3.
~ 0 and apparently RF current cannot be driven. Also indicated in Fig 2c is an
experimental point obtained with
lie
e 4 X 1012 cm- 3 where AI = 0 . In this type of discharge little or
no RF current is driven, even though the density was low. Discharges of this type, which we call type II, are
obtained by reducing the electric field applied to the plasma during the first millisecond of the shot and by
increasing the initial hydrogen gas feed rate. At the time the RF is applied, the discharge indicated in Fig. 2c
has nominally the same parameters, It, BO,
ie , and
plasma position as the discharges, type I, supporting current
drive; however, the two discharges differ in that type I has a larger preformed supra-thermal electron tail than
type II. Soft x-ray spectroscopy using a Kevex lithium doped silicon Si(Li) detector and pulse height analysis
indicates greatly enhanced soft x-ray emission in the energy range of 2-5 keV, and an elevated tail temperature
of typically 1.7 keV in type I discharges compared to a tail temperature of 900 eV in a type II discharge. Type
5
II discharges generally exhibit fractionally larger loop voltage levels than type I with other conditions of current
and density the same. Therefore, on the basis of these data it appears that a preformed suprathermal electron
tail of medium energy is a necessary prerequisite for efficient RF current generation.
Other observations indicated in Fig. 3 are a sudden decrease in hard x-ray emission during RF current
drive with a subsequent sudden increase at the end of the RF pulse. The line average density (Fig. 3d) shows a
tendency to increase during RF current drive without a significant change in light impurity line radiation. When
RF is injected into type II discharges no current is driven, and there is little or no increase in density. These
phenomena are not well understood at present; however, more efficient recycling and/or ionization may occur
during RF current drive.
We recall that in several past lower-hybrid experiments production of the RF current has been attributed
to RF enhanced runaway production 8 "0 "' . However, such a runaway current cannot account for many of
the present experimental results, particularly the generation of a negative loop voltage in Fig. 1b. Similarly, a
large increase in the number of runaway electrons would be expected to cause an enhancement in the plasma
electron cyclotron radiation at the second harmonic; however, the 2 wce radiation is found to decrease during
RF current drive. The runaway production picture may also be difficult to reconcile with the observed decrease
in the hard x-ray emission. Based on these results, plus the soft x-ray measurements, we conclude that the RF
driven runaway effect alone does not explain the present experimental results. Rf current generation has also
been obtained with a four waveguide antenna with n, < 4. This antenna is expected to interact with higher
energy electrons so that RF induced runaway production may be more important.
In summary, observations in the Versator II current drive experiment can be consistently described if
we hypothesize that efficient absorption of the lower-hybrid power occurs only by damping on a preformed
suprathermal electron tail in the energy range 1-5 keV (f/KTe c- 4 - 20). This range of energies is equivalent
to the "plateau" electron population in the conventional current drive theory4 . Such a slide-away electron tail
is a well known feature of tokamak discharges of sufficiently low density. The tail absorption picture therefore
accounts for the observed density limit, and the different behavior of discharges of types I and II with regard
to current drive efficiency. Discharges with insufficient resonant tail electrons do not support efficient current
generation. Since the slide-away tail is asymmetrically weighted in the direction of the electron drift, waves
launched in the opposite direction to the electron ohmic drift do not lead to efficient RF drive in agreement
with experiment (see Fig. Ic and 2b). At present we do not understand what prevents efficient current gener-
6