You are on page 1of 16

PSFC/JA-12-71

ICRF in the SOL: Fine Structure Radial Electric


Fields

I. Cziegler*, J. L. Terry**, S J Wukitch**, M L Garrett**, C Lau**, Y Lin**


*
University of California San Diego, La Jolla, CA 92093 USA
**
Plasma Science and Fusion Center, MIT, Cambridge, MA 02139 USA

April 2012

Plasma Science and Fusion Center


Massachusetts Institute of Technology
Cambridge MA 02139 USA

This work was supported by the U.S. Department of Energy Coop.


Agreement DE-FC02-99-ER54512. Reproduction, translation, publication,
use and disposal, in whole or in part, by or for the United States
government is permitted.

Submitted for publication to Plasma Physics and Controlled Fusion


ICRF in the SOL: Fine Structure Radial Electric
Fields
I Cziegler1 , J L Terry2 , S J Wukitch2 , M L Garrett2 , C Lau2 , Y
Lin2
E-mail: icziegler@ucsd.edu
1
University of California San Diego, Center for Momentum Transport and Flow
Organization, Center for Energy Research, 9500 Gilman Drive, La Jolla, CA, 92093
2
Massachusetts Institute of Technology, Plasma Science and Fusion Center, 175
Albany Street, Cambridge, MA, 02139

Abstract. Gas-puff-imaging techniques are utilized to detect radial electric field


structures in the scrape-off-layer (SOL) of the Alcator C-Mod tokamak via the
observation of poloidal motion of advected fluctuations. When the diagnostic’s field of
view is magnetically connected to the Ion-Cyclotron Range of Frequencies (ICRF)
active antennas, large (up to 8km/s) poloidal velocities are observed in a radial
region encompassing both field lines terminating on and those passing in front of the
antennas. The radial electric field switches sign indicating a peak in the potential profile
corresponding to the transition from piercing to passing field lines. The electric field
extends a few centimeters into the scrape off layer and its local magnitude is of order
Er ≈ 20 − 30kV/m. The corresponding plasma potentials scale as the square root of
RF power; the poloidal structure is peaked when the field lines are connected to the top
and bottom of the antenna. This structure is consistent with the presence of potential
structures arising as a consequence of sheath rectification of the RF waves. The most
striking result, however, is that the radial penetration λ⊥ of the potential structures is
an order of magnitude larger than the basic theoretical expectation (λ⊥ ≈ 10δe , where
δe is the skin depth). This substantial broadening is expected to have a strong impact
on RF impurity physics. A (weak) power-dependence observed in the width of the
poloidal velocity features is explained as a competition between the RF induced and
the background potential gradients.

PACS numbers: 52.40.Fd, 52.40.Hf, 52.40.Kh, 52.50.Qt, 52.25.Vy, 52.55.Fa, 52.70.Kz

Submitted to: Plasma Physics and Controlled Fusion


ICRF in the SOL 2

1. Introduction

Although heating with waves in the ion cyclotron range of frequencies (ICRF) has been
demonstrated on high performance plasmas [1], impurity contamination associated with
ICRF antenna operation, while acceptable in machines with plasma facing components
(PFCs) made of low atomic number (Z) materials, has been shown to be more challenging
in experiments with high Z PFCs.[2, 3, 4] In addition, the antenna structures have shown
enhanced localized thermal loads [5] that need to be addressed in long pulse operation.
For nearly three decades, the primary suspect have been an enhancement of the DC
plasma potential [6] induced via sheath rectification of the radiated RF waves. Since
the first experimental verification of the phenomenon [7], such electric fields have been
confirmed to exist in the vicinity of energized RF antennas[8, 9, 10], including recent
measurements on Alcator C-Mod [11].
Nevertheless, there is a caveat in the hypothesis that these electric fields are the
main cause for the increased rates of impurity generation and transport modification
through the formation of convective cells. As described in the original theory of sheath
rectification [12], the plasma facing components will develop a DC potential only in
response to the slow wave, which is evanescent in the antenna region with a decay
length of the plasma skin depth. Since this skin depth is much smaller than the typical
width of the SOL, it is difficult to understand how the transport through the entire
SOL would be affected. There is, however, a growing body of evidence[8, 9, 13, 14, 15]
indicating that the rectified potentials could penetrate substantially deeper than the
skin depth.
The direct experimental investigations of these SOL modifications, and primarily
the electric potential structures, are generally carried out using probes measuring
electron density and local electric potential in a spot or one-dimensional (scanning
probe) measurement. This means the experimentalist faces two issues. Firstly, a proper
description of the geometry of RF sheath rectification is essential for the design of
optimized RF antennas for future devices. Thus, a 2D resolution is minimally required
(perpendicular to the magnetic field), since the currents in the antenna obviously break
the poloidal symmetry. Secondly, the rectification phenomenon will distort the I-V
characteristics of probes themselves. As a result, most recent measurements have
employed emissive probes for more accuracy. Unfortunately, these devices are both
difficult to construct and difficult to operate in RF heated plasmas, and generally have
a short life time.
This paper presents a different technique for the characterization of SOL potentials
via the observation of the poloidal velocities of turbulent structures by Gas-Puff-
Imaging. The technique was used to map out the 2-dimensional structure of sheath
rectified RF potentials around two ICRF antennas in Alcator C-Mod. Results show a
poloidal structure exhibiting an excellent match with simulations, a clear dependence
on the input power, and finally a significantly enhanced penetration depth.
ICRF in the SOL 3
a)
E antenna

D antenna

GPI
view
J antenna

b)
0.4
SOL
0.2
field-
line
J antenna

D antenna
z (m)

E antenna

0
GPI
-0.2 view

-0.4
-180 -100 0 100 180
toroidal angle (deg)

Figure 1. (color online) a) top view of the Alcator C-Mod tokamak with the toroidal
locations of the RF antennas and the diagnostic gas puff marked; b) toroidal layout of
the antenna locations and the gas puff with a magnetic field line in the SOL overlaid
at a standard q95 = 4.6, demonstrating the magnetic mapping.

2. Experiments

In this paper, the SOL response to ICRF waves is examined by measuring the radial
structure of the poloidal dynamics of emission fluctuations during the application of
ICRF power. For the discharges investigated, the ICRF heating scenario is hydrogen
minority with strong single pass absorption (H fraction ∼ 5%). The antennas
are operated in dipole phasing, historically the phasing with the lowest impurity
contamination, at 80 and 78 MHz. A more complete description of the ICRF antennas
can be found in reference [16]. The edge and scrape-off-layer (SOL) fluctuations are
diagnosed using gas-puff-imaging, i.e. a locally injected, room temperature neutral gas
enhances line radiation (He 586nm in this study). On C-Mod, the diagnostic gas puff
enters from a 4-barrel nozzle mounted on the low-field-side limiter, typically ∼ 3cm
outside the last closed flux surface (LCFS), centered 2.54cm below the height of the
midplane. The limiter at the location of the diagnostic gas puff is split in the middle
such that the magnetic field lines connecting the gas puff to the RF antennas do not
graze any limiter at the midplane. The 2D GPI array views the puffed neutral gas
cloud toroidally, perpendicular to the radial-vertical plane at the toroidal location of
the nozzle. The view covers 4cm(radial) × 4.4cm(vertical) at this location. The views
are sampled by avalanche photo-diodes at 2MHz. A more detailed description of the
diagnostic can be found in [17]. For ICRF powers as low as 200kW we find a fine scale
ICRF in the SOL 4

limiter

nozzle

Figure 2. (color online) The cross section of Alcator C-Mod with a representative
lower single null (LSN) magnetic equilibrium, and the location of the gas-puff-nozzles
and the GPI views overlaid. The inset is an enlarged view of the GPI viewing spots.

radial structure in the poloidal phase velocities of SOL fluctuations, very different from
typical profiles of ohmic plasmas.
In the SOL of regular, ohmically heated plasmas, these fluctuations have been
observed to move in the ion diamagnetic drift (IDD) direction [17, 18, 19, 20] (see
Fig. 3 b and c). This is expected simply because in the absence of externally imposed
potentials, the electrostatic potential is completely dominated by the parallel electron
losses to material surfaces in the open field line region of a tokamak, which leads to
φ ≈ 3Te /e [21] for a deuterium plasma. Since in the SOL ∂r Te < 0, we find that
vθ = Er /Bϕ = −∂r φ/Bϕ ≈ 3∂r Te /(eBϕ ), (the same direction as the IDD, vertically
down at the midplane with the normal magnetic field orientation of Bϕ < 0), wherever
the poloidal propagation speed is simply E × B. This trend was demonstrated both in
simulations [22], and experiments reported in a previous paper [17]. In ohmic plasmas
it was shown from independent measurements of the electron temperature that these
turbulent structures do move in the SOL at the E × B velocity, except in the immediate
vicinity (up to ∼ 5mm) of the separatrix (cf. Fig. 3 and reference [17]). In the absence
of independent measurements of the local plasma potential during ICRF heating, it is
assumed that the turbulent structures still move at the E × B velocity. Consequently,
the observed radial profiles shown in Figs. 3d and 6a imply a fine-scale DC Er structure
formed in the presence of ICRF waves.
The Er profile observed during ICRF heating has radial variations on the scale of
∼ 0.5cm. We consider this to be a fine scale since it is smaller than both the typical
SOL width (∼ 2 − 3cm) and the fast wave perpendicular wave-length (∼ 10cm). Even
though as Fig. 1 shows, the GPI observations are in a region toroidally well separated
from the powered antennas (L|| ≈ 2m), the GPI view and some parts of the antennas
are magnetically connected at all times. Therefore it is argued here that these features
ICRF in the SOL 5
a) 500
b) 6
GPI > 0
400 4 GPI < 0
v*
f (kHz)

vθ (km/s)
300 2 E×B

LCFS

IDD : EDD
200 0

100 -2

0 -4
-6 -4 -2 0 2 4 6 -2 -1 0 1 2
kθ(cm-1) Distance into SOL (cm)

c) d)
6 Ohmic RF
4
shot 1110127014, t=1.00-1.05s

shot 1110127011, t=1.10-1.15s


IDD : EDD

2
vθ (km/s)

-2

-4
LCFS
LCFS

J ant

-6

88.0 88.5 89.0 89.5 90.0 90.5 91.0 88.0 88.5 89.0 89.5 90.0 90.5 91.0
R (cm) R (cm)

Figure 3. (color online) a) Conditional (kθ , f )-spectrum from an RF heated SOL –


the dashed lines represent linear fits; b) common radial velocity profile in an ohmic
plasma measured as shown in Fig. a) (reproduced from [17]) with the diamagnetic flow
(at ρ < 0) and E × B (at ρ > 0); c) velocity distribution D(R, vθ ) from an ohmic SOL
and d) from a SOL heated by a single antenna (at J port) whose front face is mapped
to the location marked by the dashed blue line. In b) and d), diamonds (blue) stand
for the velocities of kθ > 0, squares for kθ < 0.

arise as a consequence of sheath rectification.


To demonstrate this effect, the poloidal velocity field in the ∼ 3cm region outside
the separatrix is more closely examined. To define the radial regions in which a certain
poloidal propagation velocity is dominant, while providing a measure of a) how much
overlap there is between potential counter-propagating features and b) how large the
velocity spread is, we calculate a distribution in velocity space by integrating the
conditional spectrum along fixed phase velocity contours, as demonstrated in Fig. 3a,
1 vθ k∑
N /(2π)

D(vθ ) = S(kθ = 2πf /vθ |f ), (1)


N (vθ ) f =0
where the conditional wavenumber-frequency spectrum is defined as
S(kθ , f )
S(kθ |f ) = ∑ , (2)
kθ S(kθ , f )
and is used to extract the experimental dispersion from the k − f distribution. Radial
profiles D(R, vθ ) constructed from these distributions are plotted in Fig. 3 for c) a
standard ohmic discharge and d) an RF heated one. The same linear fits to the lobes
observed in the conditional spectra as the ones that produced Fig. 3b are overlaid in
Fig. 3 partly to show the good match between the two representations of the velocity field
and partly to highlight the inadequacy of the individual linear fits at gauging which of
the two observed velocities is locally dominant. As Fig. 3d shows, in a discharge heated
by a single RF antenna vθ (r) varies from downward (Er > 0) in the first ∼ 1cm outside
ICRF in the SOL 6

0.2
D antenna
X-section
pierces D
0.1

0.0
GPI
z (m)

view

-0.1

pierces J

-0.2 J antenna
X-section
1110127014; t=1.21 s

-0.3
passing
0.70 0.75 0.80 0.85 0.90 0.95 1.00
R (m)

Figure 4. (color online) The geometry of field lines in the GPI field-of-view that
either pass or pierce the RF antennas. Shown are poloidal projections of 1) the GPI
field-of-view (black outline), 2) the lower half of the J antenna box green outline at
bottom, 3) the upper half of the D antenna box (blue outline at top), 4) the GPI-
viewed field line that just passes the J antenna box (solid red line), 5) the GPI-viewed
field line that just pierces the J antenna (dashed red line at bottom), and 6) the field
line that just pierces the D antenna and is just outside the GPI view (dashed red line
at top). The solid (red) field lines pass and connect to the region where the poloidal
velocity is EDD (Er < 0), while the dashed (red) lines pierce the J antenna and, if the
J antenna only is energized, connect to a region where the poloidal velocity is strictly
IDD (Er > 0), as shown in Fig. 3d. When both the D and J antennas are energized
in this non-conformal configuration, the observed velocity profiles exhibit more radial
structure as shown in Fig. 6a.

the separatrix (normal for the ohmic SOL) to upward (Er < 0), to downward (Er > 0)
in the next ∼ 2cm. The sign changes in Er = ∂r φ imply local potential extrema: a
potential minimum at the first change (∼ 89.35cm in Fig. 3d) and a local maximum at
the second (∼ 90.1cm). This local buildup of space charge is predicted at the face of an
active antenna by models [12, 23, 24] which examine RF wave physics in SOL plasmas,
considering the field line sheaths. Since at the measured temporal and spatial scales the
parallel resistivity is negligible, field lines are expected to be equipotential structures
(except at the sheaths). The GPI observation region was therefore magnetically mapped
to the region around any active ICRF antennas.
Since the curvature of the magnetic flux surfaces does not exactly conform to the
front face of the antennas, the traced field lines are classified into two groups: those that
pass in front of an active antenna (“passing”) and those that pierce the face (or side) of
ICRF in the SOL 7

92.0

91.5
R of pot. max.

91.0

90.5

90.0
90.0 90.5 91.0 91.5 92.0
R of active ant. box. boundary

Figure 5. (color online) The major radius of the observed potential maximum (sign
change of poloidal velocity) at the vertical center of the GPI field of view plotted
against the major radius at the same height of the boundary between the field lines
piercing the active antenna and those passing in front of it. The dashed (red) line
represents perfect correlation.

an antenna (“piercing”). By the models of RF sheath rectification [12, 14, 23, 24] the
former are expected to sample a region of positive Er , while the latter, penetrating the
antenna structure, are expected to sample a region of negative Er . In order to clarify
the mapping, Fig. 4 shows the poloidal projections of field lines traced in a counter-
clockwise and down direction from the GPI field-of-view toward the J antenna, as well
as the projection of a clockwise-upward field line that just pierces the D antenna. The
figure illustrates the mapping for a case in which non-conformal mapping leads to GPI-
viewed field lines that are radially inside of both the J and D antennas (“passing”)
and field lines that pierce the J antenna, but are still “passing” with respect to the
D antenna. As will be shown below, this modifies the J-antenna-only velocity profiles
(Fig. 3d) such that an additional region of electron diamagnetic drift (EDD) propagating
turbulence appears, as shown in Fig. 6.
In order to verify that the sign change of the poloidal velocity profile at the implied
potential maximum corresponds to the location of a boundary between “passing” and
“piercing” field lines, the boundaries from a large number of ICRF heated discharges
were mapped to the GPI observation location and compared against the vθ > 0 → vθ < 0
boundary radius. Figure 5 demonstrates an excellent correlation between the two radii,
corroborating the interpretation of our data.
The structure shown in Fig. 3d was found to exist for all antennas individually in the
proper magnetic geometry. Additional evidence supporting this interpretation is found
in the magnetic geometry shown in Fig. 4. By shifting the typical plasma equilibria
down slightly, it was possible to make the boundaries of the two active antennas (J and
D) map to different radial locations. In this case, when both antennas are energized, the
ICRF in the SOL 8

potential structures from both are superimposed with the potential structures radially
separated in the GPI view, and produce a more complex radial profile of the poloidal
E × B velocity in the SOL. The velocity profile measured in such an experiment is
shown in Fig. 6a. With both antennas energized, an additional region of rapid EDD
propagating turbulence appears in the edge of the field of view (at 91cm in Fig. 6a). We
can estimate the part of the radial velocity and potential profile outside of the GPI view
by combining the measured profile of Fig. 6a with one measured in a different discharge
configuration for which the field line just piercing D was near the center of the GPI field-
of-view and only the D antenna was energized. This profile is shown in the Fig. 6b, but
with its radial dimension shifted so that the major radius of its just-piercing field line is
placed at the major radius of the just-piercing field line of Fig. 6a. The D-antenna-only
profile has been shifted by 0.7 cm. Good alignment is found between small portion
of the EDD-feature of Fig. 6a that was formed when D-antenna power was combined
with J-antenna power and the EDD-feature measured for the passing field line region of
D-antenna-only case of Fig. 6b. (The velocities of Fig. 6b were also adjusted slightly to
account for the radial variation in B(r)) consistent with the radial shift.) The estimated
radial profile of the local electric potential is constructed from the composite velocity
profile of Figs. 6a and b. This is shown in the bottom part of the figure, where the
vertical dashed lines mark the boundaries of the passing or piercing field lines for each
antenna. These boundaries quantify the interpretation of the GPI measurements and
the three components of the potential: 1) the φ ≈ 3Te /e gradient due to the always
present sheath drop in the SOL, and the rectified RF fields 2) from the J antenna φJ
and 3) from the D antenna φD .
The heights of the potential peaks in the figure are true to the experiment and are
estimated as
φ = λ⊥ Er = λ⊥ vθ Bϕ , (3)
where λ⊥ is the penetration depth of the rectified potential into the plasma perpendicular
to the magnetic field as mapped to the location of the active antenna, and is a quantity
determined from the GPI measurements. Using stepwise ramps of the ICRF power, we
experimentally tested the power dependence of the potential buildup. The potential
arises from the rectification of the RF parallel electric field component [12], and is
therefore proportional to the potential drop along the field line. The RF power is
proportional to the square modulus of its electric field, so that


= Crect E|| dl ∝ PICRF
1/2
φrect (4)

with a rectification ratio Crect that depends on the details of the mechanism, variously
predicted to be somewhere between 1/π and 1/2 [14]. The measured potentials were
γ
therefore fitted with a power law of the ICRF power of the form φ = A PICRF , for the
two antennas for which we were able to measure enough of the radial profile to allow
estimation of the potentials. Note that the expected curve is required to go to zero
at no input power, when there is no space charge buildup expected or observed. The
ICRF in the SOL 9
a) b)

estimated profile
6

2
vθ (km/s)

shot 1110127014 t=1.05-1.10s


0

experimental profile
-2

-4
LCFS

-6

88.0 88.5 89.0 89.5 90.0 90.5 91.0


c) R (cm)
0.3

D ant
J ant

0.2
φ (kV)

3kBTe/e
0.1 φJ
φD
Σφ
0.0
88 89 90 91 92
Rmap (cm)

Figure 6. (color online) Demonstration of the superposition of φD and φJ ; a) the


measured velocity distribution D(R, vθ ) exhibiting 4 flips in propagation direction when
the non-conformal equilibrium of Fig. 4 is present and both the J and D antennas are
energized (cf. Fig. 3d for J antenna only). The profile is extended with a distribution
b) from a measurement of D(R, vθ ) in a different discharge for which only the D
antenna was energized and which has been shifted radially so that passing–D-piercing
boundary is aligned with that for the configuration of (a). The plot c) below represents
the estimated electric potential profile derived from the estimated velocity profile of
(a) and (b) including an estimated contribution of 3Te /e resulting from the Te profile.
All values for R are mapped to the midplane.

results for the D antenna are plotted in Fig. 7 along with the best fit curve, for which
the exponent of the power law is γ = 0.52, very close to the theoretical expectation of
γexp = 0.5. The measured peak potential in the magnetic configuration plotted in Fig. 7
at around the maximum output (∼ 1.4MW) of the antenna is as much as 380V. The
same data processing for the J antenna gives γ = 0.51 and a somewhat smaller potential
of 220V at a similar power output level. In both cases the potentials are large enough
to induce sputtering by deuterium ions on field lines that intersect the molybdenum
armored plasma facing components in C-Mod.
Additionally, the poloidal structure of the induced potentials is of considerable
interest and significance. It was possible to measure parts of this profile by varying

the safety factor, since specific calculations of | E|| dl| for the geometry of the D and
J antennas showed that with the (0, π) and (0, π, 0, π) phasings of the two antennas,
respectively, the parallel electric fields change very little in the 5◦ − 10◦ range of
misalignment angles between the magnetic field and the axis of the antennas[25]. This
moves the regions at the antennas to which the GPI observation region maps poloidally
up or down. In order to keep the heating scenario the same, this experiment was
carried out using a fixed toroidal magnetic field while varying the plasma current. The
poloidal profile of the measured potentials was shown to fit the results of the calculations
ICRF in the SOL 10
400

300
φmax (V)

200

γ=1
100
γ=0.52
0
0.0 0.5 1.0 1.5
PICRF (MW)

Figure 7. (color online) Estimated peak potential at the antenna boundary against
the power radiated from the antenna (for the D antenna). The inset shows the same
measurement points in a log-log plot. The solid (red) lines in the figures are the fitted
power law curves, with the best fit γ = 0.52. The dotted line represents a linear (γ = 1)
relationship for demonstration purposes.

500
D ant
J ant

D sim. Crecφ/π
J sim. Crecφ/π
400
D GPI exp.
J GPI exp.
300
φ (V)

200

100

0
-30 -20 -10 0 10 20 30
zave (cm)

Figure 8. (color online) Poloidal profile of the GPI measured rectified potentials for
both the 4-strap J and the 2-strap D antennas, with the magnitude of the integrated
parallel electric field from simulation [25] reduced by the theoretically motivated 1/π
rectification ratio and a further Crec residual rectification factor for a least squares best
fit. The J antenna extends to z = ±30cm, while D to z = ±25cm

of the induced parallel electric field to a remarkable degree, as is demonstrated in


Fig. 8. In addition to the theoretical rectification ratio, an additional factor, Crec ,
has been used to scale the values from the simulation of the antennas because the
curves plotted in the figure were generated by a simplified model. With E|| calculated
at the front face of the antenna boxes, the particular values of Crec for the best fit
D J
were not identical for the two antennas: Crec = 0.5 for the D antenna and Crec = 0.27
for the J antenna. The computation was done by a finite element simulation with
the realistic antenna geometries [25] using both the cold-plasma dielectric tensor and
isotropic dielectric chosen to mimic observed antenna loading. The calculations were
simplified by artificially lowering the plasma density, so that the E|| values represent
vacuum fields, and the realistic plasma response is obviously not present.
As is evident from the figure, the effect we are describing is strongest at the
ICRF in the SOL 11

(poloidal) edge of the antennas where cancelations in the parallel electric field due
to symmetry are minimal. Note that even though the antennas end at approximately
z = ±20cm, there is a large potential outside of this z, as the vertical height plotted
is the average z value of the field line in front of the antenna which passes at an angle
5−10 degrees from horizontal. This corresponds well to the results of earlier experiments
[9, 13, 15, 26] in which the particular role of “antenna corners” on RF sheath effects
was already pointed out.

3. Discussion

In the previous section we showed extensive evidence of the time-invariant, radially


alternating poloidal velocity of SOL fluctuations being caused by the rectification of RF
electric fields:

(1) the sign changes in the radial profile of the poloidal E × B velocity are consistent
with local extrema in the electric potential profile if the potential is a superposition
of potentials from energized antennas and a “background” 3Te (r)/e potential;
(2) the locations of the sign changes corresponding to potential maxima were shown
to correlate extremely well with the magnetically mapped front boundary of the
active ICRF antenna, indicating not only that
(3) the field lines are charged up by the (non-local) antennas, but also that
(4) the RF-produced potential profiles are radially maximal at those front boundaries;
(5) the dependence of the potential buildup on the input power was shown to follow a
1/2
PRF scaling, expected since the potential arises from the rectification of the parallel
electric field component of the slow wave; and
(6) the poloidal structure of the potential was shown to fit very well the first principles
calculations of parallel electric field induced in front of the antennas.

The most striking result of the reported measurements, however, is that the
potentials induced by the RF waves are observable in a significant portion of the
SOL, indeed in some shots across its entire depth. This fact can have very important
consequences for impurity physics in the presence of ICRF heating, and yet the
phenomenon has not been thoroughly investigated or explained. For a quantitative
analysis we note that for the results listed above, the local potential-differential was
estimated according to the expression Eq. 3, in which λ⊥ , the radial penetration depth of
the rectified potential, was evaluated from the GPI measurements. Here we address the
question of how the experimental value of this penetration depth relates to expectations.
The parallel electric field in the radiation of the ICRF antennas appears solely due
to the presence of the slow wave (SW), which could be generated directly or by the
fast wave propagating into a nearby wall and coupling into the slow wave to satisfy
boundary conditions[27]. The SW only propagates in a very low density plasma, and
is therefore cut off immediately at the antenna. The wave is then radially evanescent
ICRF in the SOL 12
a) 6
t t=0.90–1.05 (0.00 MW)
1.05–1.10 (0.27 MW)
5 1.10–1.15 (0.50 MW)
1.15–1.20 (0.70 MW)
1.20–1.25 (0.90 MW)
ne (1019 m-3)

4 1.25–1.30 (1.10 MW)


LCFS

1.30–1.35 (1.25 MW)


3

1110218001
1

0
0.00 0.01 0.02 0.03
Distance from LCFS (m)

b) 1.2

1.0

0.8
λ GPI (cm)

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
PICRF (MW)

Figure 9. (color online) a) electron density profiles in a power-ramp experiment, b)


radial width of the velocity-reversal region, from which the penetration depth λ⊥ is
calculated, against RF heating power from the same experiment as a); the dashed line
is the expected GPI observed width when λ⊥ = 1cm and the potential has a Gaussian
shape.

being shielded out by the plasma at the plasma skin depth δpe = c/ωpe , which with the
densities at the antenna in C-Mod is about δpe = 1 − 3mm. This estimate is based on
measured profiles, an example of which is shown in Fig. 9a. The observed λ⊥ widths
are significantly broadened compared to this estimate, as they range from 7mm up to
11mm depending on the input power, as shown in Fig. 9b.
For the measured density profile of Fig. 9, we estimate δpe = 1mm while the
measured λ⊥ at the time of maximum power (1.25MW at t = 1.30 − 1.35s) is 9.5mm.
In the recent literature of RF wave physics, a number of mechanisms have been
proposed that could be invoked to explain this “anomalous” penetration depth of the
DC potential structures into the plasma.
a) In the presence of sheaths the RF fields could propagate along material boundaries.
This possibility was examined in reference [28] where the relevant branch was
termed a “Sheath-Plasma Wave”, with dispersion properties quite different from
the usual Fast Wave and Slow Wave. For the exact amount of broadening, the
sheath parameter Λ = −(∆ϵ|| )/L|| must be self-consistently determined, where ∆
is the sheath width, ϵ|| is the parallel component of the dielectric tensor and L|| the
connection length. The process is somewhat involved, the case of C-Mod, however,
ICRF in the SOL 13

is studied in the reference and it is predicted that the discussed broadening effect
should be significant for sheaths up to ∼ 30cm away from the RF source. While
along “piercing” field lines the private flux region of the antennas are sampled,
and therefore the sheaths are certainly closer than the required range, invoking this
mechanism for the broadening at the front face of the antenna, where the connection
length is ∼ 2m leads to some difficulty.
b) As pointed out in reference [14], the additional broadening can also be explained
if self-consistent exchange of transverse RF current is allowed between neighboring
flux tubes. The amount of broadening at large powers is well approximated by a
linear model [14] of such transverse currents, estimating λ = (ρci L|| /2)1/2 , where ρci
is the ion Larmor radius and L|| is the connection length of the field line in question.
With the real C-Mod values of ρci ≈ 0.1mm and L|| ≈ 2m, the estimation yields
λ = 1cm, consistent with the maximum values measured and shown in Fig. 9b.
Note that the above formula was first compared to probe measurements on Tore
Supra but did not provide such an excellent match to experimental data as on
C-Mod. The linear model of the mechanism proposed in reference [14] does not
predict λ⊥ to depend on PRF .
c) However, once a transverse plasma conductivity is introduced, in addition to the
radial propagation of RF waves as in [14], the radial exchange of DC currents
between neighboring flux tubes could also induce a radial broadening of the DC
plasma potentials. Although its physical nature is still unclear, the existence
of a DC transverse plasma conductivity is attested from both Langmuir probe
measurements in the absence of RF waves [29] and discussed in theory [30]. In
the context of RF sheath rectification this possibility was investigated in [31] and
1/2
revisited more recently in [32]. In the latter, the broadening still scales as L|| via
a DC current diffusion length, but also depends on RF power.
We note, however, that lambda as determined from the location of velocity
direction reversal may also limit the measured values of lambda, and lead to the
appearance of a false dependence of λ⊥ upon power. The mechanism is as follows:
due to the background potential gradient caused by the temperature gradient in the
SOL, the poloidal velocity reversal is only expected above a certain threshold power.
Approximating the exponentially decaying potential away from the antenna as a linear
drop up to λ⊥ , this threshold can be estimated as φthr = vθohmic Bϕ λ⊥ ≈ 50V, and
therefore PRFthr
≈ 50kW from Fig. 7. Since the actual potential drop-off is not linear, this
threshold will not present as a sharp boundary but rather as an increase in λ⊥ around
this value going to saturation at the broadened depth λmax ⊥ , which to first order should
not depend on the input power according to the model of reference [14]. This also means
that the employed method for the measurement of λ⊥ always slightly underestimates
the true value due to the background potential gradient.
While the details of the scaling of λ⊥ with RF power are thus not easily quantified,
the reality of radially and toroidally extended regions with large electric potentials set up
ICRF in the SOL 14

by energized ICRF antennas remains. The existence of these potential structures affect
the particle transport in a relatively large radial region of the SOL ∆r ∼ 1cm – and up
to 2cm! We speculate that this modification may result in large convective cells that
will influence the penetration of impurities and increase the impurity contamination of
plasmas that are strongly ICRF-heated. Consequently, the reported profiles may help
to explain the increased impurity content observed with ICRF heating [4], as a result
of both enhanced sputtering and enhanced transport/impurity penetration across the
SOL. We note that these effects will be important for impurity generation and SOL
transport in regions well away from the antennas.
Due to the excellent match the estimation in reference [14] achieves with the
experimental measurements of the potential broadening, the dependence of λ⊥ on L||
and on the toroidal magnetic field will be addressed in a follow-up paper, along with
whether the enhanced performance of N and Ne seeded plasmas in ICRF heating is
due to the reduction of the reported effects. In addition, a new, rotated antenna
geometry was recently implemented on Alcator C-Mod, in which the straps are aligned
perpendicularly to the equilibrium magnetic field. This antenna is expected to reduce
the rectified potentials significantly. The 2D structure of electric potentials around this
antenna will be also examined.

Acknowledgments

The authors would like to thank Jim Myra, Dan D’Ippolito and Sergei Krashenninikov
for discussions and insightful comments as well as the Alcator C-Mod team of students,
scientists, engineers and technical staff for making the experiments possible. This work
is supported by U.S. Department of Energy Cooperative Agreement No. DE-FC02-
99ER54512

References

[1] D. F. H. Start et al. Nucl. Fusion, 39:321, 1999.


[2] R. Neu et al. Plasma Phys. Control. Fusion, 49:B59 – B70, 2007.
[3] S.J. Wukitch, B. LaBombard, Y. Lin, B. Lipschultz, E. Marmar, M.L. Reinke, and D.G. Whyte.
Journal of Nuclear Materials, 390 - 391(0):951 – 954, 2009. Proceedings of the 18th International
Conference on Plasma-Surface Interactions in Controlled Fusion Device.
[4] S.J. Wukitch, B. Lipschultz, E. Marmar, Y. Lin, A. Parisot, M. Reinke, J. Rice, and J. Terry.
Journal of Nuclear Materials, 363365(0):491 – 497, 2007. Plasma-Surface Interactions-17.
[5] L. Colas, S. Heuraux, S. Bremond, and G. Bosia. Nucl. Fusion, 45:767, 2005.
[6] R. I. Taylor et al. In Proc. 9th Conf. on Plasma Physics and Controlled Nuclear Fusion Research,
Baltimore, 1982, volume 3, page 251. IAEA, 1983.
[7] R. Van Nieuwenhove and G. Van Oost. Journal of Nuclear Materials, 162 - 164(0):288 – 291,
1989.
[8] R Van Nieuwenhove and G Van Oost. Plasma Physics and Controlled Fusion, 34(4):525, 1992.
[9] L Colas, A Ekedahl, M Goniche, J P Gunn, B Nold, Y Corre, V Bobkov, R Dux, F Braun,
J-M Noterdaeme, M-L Mayoral, K Kirov, J Mailloux, S Heuraux, E Faudot, J Ongena,
ICRF in the SOL 15

ASDEX Upgrade Team, and JET-EFDA contributors. Plasma Physics and Controlled Fusion,
49(12B):B35, 2007.
[10] J. P. Gunn, L. Colas, A. Ekedahl, E. Faudot, V. Fuchs, S. Heuraux, M. Goniche, M. Koean,
A. Mendes, A. Ngadjeu, V. Petrzilka, F. Saint-Laurent, and K. Vulliez. In Proc. 22nd IAEA
Fusion Energy Conference, Geneva, pages 6 – 32, 2008.
[11] R. Ochoukov et al. Investigation and control of rf-enhanced plasma potentials on alcator c-mod.
PSI 2012 Conference, Aachen, Germany, to be published in Journal of Nucl. Mat., 2012.
[12] V. A. Godyak and A. A. Kozuvnikov. Fiz. Plazmy, 1:496, 1975.
[13] L. Colas, J.P. Gunn, I. Nanobashvili, V. Petrzilka, M. Goniche, A. Ekedahl, S. Heuraux, E. Joffrin,
F. Saint-Laurent, C. Balorin, C. Lowry, and V. Basiuk. Journal of Nuclear Materials, 363 -
365(0):555 – 559, 2007.
[14] E. Faudot, L. Colas, S. Heuraux, and J. P. Gunn. Phys. Plasmas, 17:042503, 2010.
[15] V. V. Bobkov et al. Nucl. Fusion, 50:035004, 2010.
[16] P. T. Bonoli, R. Parker, S. J. Wukitch, et al. Fusion Science and Technology, 51:401 – 435, 2007.
[17] I. Cziegler, J.L. Terry, J.W. Hughes, and B. LaBombard. Phys. Plasmas, 17:056120, 2010.
[18] D. L. Brower, W. A. Peebles, N. C. Luhmann, and R. L. Savage. Phys. Rev. Lett., 54:689, 1985.
[19] P. C. Liewer. Nucl. Fusion, 25:543, 1985.
[20] S.J. Zweben and R. W. Gould. Nucl. Fusion, 25:171, 1985.
[21] P. C. Stangeby. The Plasma Boundary of Magnetic Fusion Devices. Institute of Physics, Berkshire,
2000.
[22] Y. Sechrest, T. Munsat, D. A. D’Ippolito, R. J. Maqueda, J. R. Myra, D. Russell, and S. J. Zweben.
Phys. Plasmas, 18:012502, 2011.
[23] M. A. Lieberman. IEEE Trans. Plasma Sci., 16:638, 1988.
[24] D. D’Ippolito and J. Myra. Phys Fluids B, 5:3603, 1993.
[25] M. L. Garrett and S.J. Wukitch. Mitigation of radio frequency sheaths through magnetic field-
aligned icrf antenna design. Fusion Eng. Des., 2012. in press.
[26] T. Tanaka, R. Majeski, D. A. Diebold, and N. Nershkowitz. Nucl. Fusion, 36:1609, 1996.
[27] J. R. Myra, D. A. D’Ippolito, and M. Bures. Phys. Plasmas, 1:2890, 1994.
[28] J. R. Myra and D. A. D’Ippolito. Plasma Phys. Control. Fusion, 52:015003, 2010.
[29] A. Carlson et al. In Controlled Fusion and Plasma Physics (Proc. 20th Eur. Conf. Lisbon, 1993),
1993.
[30] V. A. Rozhansky, A. A. Ushakov, and S. P. Voskoboynikov. Nucl. Fusion, 39:613, 1999.
[31] D. A. D’Ippolito, J. R. Myra, et al. Nucl. Fusion, 42:1357 – 1365, 2002.
[32] L. Colas et al. In Proc. 37th EPS conference Dublin, 2010 P5.145, 2010.

You might also like