Transcript T e

Study of the pedestal dynamics and
stability during the ELM cycle
A. Burckhart
Advisor: Dr. E. Wolfrum
Academic advisor: Prof. Dr. H. Zohm
MPI für Plasmaphysik, EURATOM Association
A. Burckhart, PhD network, September 30th, 2010
Overview
• Motivation: why study ELMs?
• Overview on peeling ballooning theory and ELMs
• Evolution of Te, ne and pe during the ELM cycle
• Stability calculations
• Conclusions
A. Burckhart, PhD network, September 30th, 2010
2/30
Overview
• Motivation: why study ELMs?
• Overview on peeling ballooning theory and ELMs
• Evolution of Te, ne and pe during the ELM cycle
• Stability calculations
• Conclusions
A. Burckhart, PhD network, September 30th, 2010
3/30
Motivation
• Highest performance is found in H-mode plasmas featuring
type-I ELMs  Forseen operation scenario for ITER
• Edge localized modes (ELMs) are instabilities of the plasma
edge, leading to a periodic deterioration of the pedestal profiles
• They cause massive power loads on the divertor, up to 10% of the confined
energy in less than 1ms
• They cause high energy and particle losses
• They “clean” the plasma by expelling impurities
•They are not yet fully understood
• Investigating the dynamic behavior of pedestal profiles can lead to a better
understanding of ELMs, and maybe ultimately to their control
• Stability calculations are usually performed for one time point directly before
the ELM crash, what about the temporal evolution of the stability?
A. Burckhart, PhD network, September 30th, 2010
4/30
Overview
• Motivation: why study ELMs?
• Overview on peeling ballooning theory and ELMs
• Evolution of Te, ne and pe during the ELM cycle
• Stability calculations
• Conclusions
A. Burckhart, PhD network, September 30th, 2010
5/30
Plasma stability
• We consider a plama equilibrium with the total potential energy W, and calculate
the effect of an arbitrary small displacement ξ
• If δW > 0, the system is stable, with δW < 0 it is unstable
δW > 0
δW < 0
ξ
ξ
stable (oszillation)
A. Burckhart, PhD network, September 30th, 2010
unstable
6/30
Edge stability
Instability: The change of potential energy is negative
W  W plasma  Wvacuum
Wvacuum 
Stabilising:
B1
1
dr

2 vac
0
2
(vacuum field line bending, always stabilising)
Field line
bending
Compression of
magnetic field
lines
Plasma
compression
 B1, B 2
1
2
0
W plasma 





2



 p0   




2 plasma   0
0
 2   p0       j||    b  B1,
Destabilising:
Pressure driven
modes
2

Current driven
modes
[Saarelma, JET 13.09.2010]
Stability code: Find  that minimises W
A. Burckhart, PhD network, September 30th, 2010
7/30
Peeling modes
• Current driven
instabilities that are
localized radially at
the plasma edge and
poloidally near the Xpoint.
• Usually low-n
n=7
Mode localized near
the x-points.
[Saarelma, JET 13.09.2010]
A. Burckhart, PhD network, September 30th, 2010
8/30
Edge ballooning modes
• Driven by the pressure
gradient.
• localized on the low
field side.
• Radially more extended
than peeling modes.
• Usually high-n.
n=20
Mode localized
at the bad
curvature region
[Saarelma, JET 13.09.2010]
A. Burckhart, PhD network, September 30th, 2010
9/30
ELMs are thought to be combined
peeling-ballooning modes
• Driven by both the current
and the pressure gradient.
• Intermediate-n
• Mode structure has
features from both modes.
Peeling component
Ballooning component
[Saarelma, JET 13.09.2010]
A. Burckhart, PhD network, September 30th, 2010
10/30
ELMs expel particles from the plasma
• Glow: mostly H-alpha,
cold and dense
• Filaments expelled,
travelling along field
lines
• Highest intensity lasts
less than 1ms
[T. Lunt]
A. Burckhart, PhD network, September 30th, 2010
11/30
Te and ne pedestals relax due to ELMs
• Ipolsola: currents in the divertor, used as
ELM indicator
• Te relaxes quickly during ELM crash, then
slowly recovers
• ne also relaxes, but evolves around
„pivot point“: ne inside drops, ne in the
SOL increases
A. Burckhart, PhD network, September 30th, 2010
12/30
Overview
• Motivation: why study ELMs?
• Overview on peeling ballooning theory and ELMs
• Evolution of Te, ne and pe during the ELM cycle
• Stability calculations
• Conclusions
A. Burckhart, PhD network, September 30th, 2010
13/30
Te recovery shows several distinct phases
Max(Te) mapped
relative to t=0
during ELM
Te is small
initial recovery
of Te
Te recovery
stalls
Te exhibits
large fluctuations
Te recovery
continues
A. Burckhart, PhD network, September 30th, 2010
14/30
ne recovers differently than Te
Te pedestal takes around
7ms to fully recover,
ne pedestal only 4ms
Only one short recovery
phase ~ 2-3ms
Last phase is stationary,
but large scale fluctuations
Overshoot at the
end of the recovery
A. Burckhart, PhD network, September 30th, 2010
15/30
Interplay between ne and Te buildup
• In the phase where Te recovery
stalls, ne rapidly recovers
• In the following phase where Te
recovers rapidly, ne stays
constant
• This is observed in all analyzed
discharges, at all gas fueling
levels
• Reminiscent of critical value of
he=Ln/LT above which Te is
clamped: possibly ETGs are
responsible for pause in the
recovery of the Te profile
A. Burckhart, PhD network, September 30th, 2010
16/30
New laser triggering method will allow for a
higher density of TS data points
• Next steps:
• Improve characterization of
edge Te profile recovery using
TS data, which does not have
the shine through problem
(no accurate ECE data outside
ρpol=0.995 )
 long discharges necessary
TS data point
• Thomson scattering diagnostic:
• Six 20Hz Nd-Yag lasers  One profile every 8.3ms
• Or: burst mode, fire all lasers in a given time interval
Result:
All data points from
the TS system lie in
time interval of the
ELM cycle that we
want to analyze
• Next campaign:
ELM
triggers laser
optical signal
detected
by XVR
Delay
A + n*x
A. Burckhart, PhD network, September 30th, 2010
17/30
Next steps concerning pedestal dynamics
• Test reproducibility of interplay between Te and ne in the recovery of the
pedestal profiles on other devices, starting with JET
•
Very good Thomson scattering profiles available, but only 20Hz
 long discharges and coherent data selection necessary
• Fast ECE data available, but low resolution in the pedestal region, and
shine through effect often more pronounced than on AUG
• Reflectometry measurements, Li-beam and DCN also available but not
suited for calculating gradients and/or temporal resolution not sufficient
• No method that combines the data from different diagnostics to one joint
profile (c.f. integrated data analysis at AUG)
• Start with analyzing TS and ECE data of „old“ discharges
• Then analyze dependencies of the pedestal dynamics on collisionality
and on fueling level
A. Burckhart, PhD network, September 30th, 2010
18/30
„Steady state“ phase dependent on fueling level
AUG
JET
Fueling
[E. Giovannozzi, EPS 2009]
• Length of the last, quasi stationary phase, is dependent on fueling level
• Length of initial recovery phase barely changes with fueling
• Next Step: trying to find a solid dependence of the different phases in the ELM
cycle on the fueling level (AUG and JET)
A. Burckhart, PhD network, September 30th, 2010
19/30
Reaching the pe limit does not necessarily
lead to an ELM
‚fast‘ ELM frequency
‚slow‘ ELM frequency
• Clearly, a limit to max(pe) exists, but reaching it does not automatically trigger
an ELM
• Some ELMs seem to be triggered by reaching max(pe)
(consistent with JETTO simulation [J. Lönnroth, PPCF 2004])
• Others can sit at the pressure limit for several ms before ELM happens
(seen before on AUG [T. Kass, NF 1998] and DIII-D, [R. Groebner, NF 2009])
A. Burckhart, PhD network, September 30th, 2010
20/30
Current diffusion cannot explain delayed ELM
[J. Connor, Phys.Plasmas 1998]
• ELM thought to be driven by combination of pressure and current gradient
• Pressure gradient limited by ballooning mode
• Edge current gradient limited by peeling (=very edge localized kink) mode
• Bootstrap current roughly proportional to pe, but delayed by about 1ms because
of current diffusion  cannot explain 6-7ms observed between reaching final
pe and occurrence of the next ELM
A. Burckhart, PhD network, September 30th, 2010
21/30
Pressure further inside might have an effect on
ballooning stability
‚slow‘ ELM frequency
‚fast‘ ELM frequency
Pressure further inside
continues to increase
Pedestal slope and
pedestal top stay constant
• Pressure further inside might influence peeling-ballooning stability
• Next step: stability calculations
A. Burckhart, PhD network, September 30th, 2010
22/30
Overview
• Motivation: why study ELMs?
• Overview on peeling ballooning theory and ELMs
• Evolution of Te, ne and pe during the ELM cycle
• Stability calculations
• Conclusions
A. Burckhart, PhD network, September 30th, 2010
23/30
Stability calculations performed in several steps
ELM-coherent
averaged profiles
and magnetic data
CLISTE
equilibrium
(with kinetic
profiles)
HELENA
High resolution
equilibrium
A. Burckhart, PhD network, September 30th, 2010
ILSA
stability
calculations
24/30
Experimental data is ELM synchronized
ELM-coherent
averaged profiles
and magnetic data
CLISTE
equilibrium
(with kinetic
profiles)
HELENA
High resolution
equilibrium
ILSA
stability
calculations
ne (1019 m-3)
Te (eV)
[M. Dunne]
A. Burckhart, PhD network, September 30th, 2010
25/30
Grad-Shafranov equation is solved using CLISTE
CLISTE
equilibrium
(with kinetic
profiles)
ELM-coherent
averaged profiles
and magnetic data
HELENA
High resolution
equilibrium
ILSA
stability
calculations
Output
Inputs for CLISTE
– Pressure data (10%)
– Sauter Formula (1%)
– MAC currents (<1%)
– Magnetics (~90%)
– cBoot (1%)
– [SXR]
– [CES/CEZ]
Pressure
Current
density
bootstrap
A. Burckhart, PhD network, September 30th, 2010
26/30
Different p - j combinations are tested for stability
CLISTE
equilibrium
(with kinetic
profiles)
ELM-coherent
averaged profiles
and magnetic data
HELENA
High resolution
equilibrium
ILSA
stability
calculations
Reminder:
 B1, B 2
1
2
0
W plasma 





2



 p0   




2 plasma   0
0

 

 2   p0       j||    b  B1,
2

Destabilising:
• edge current density
• pressure gradient
Code changes pedestal profiles to vary
p and j, while keeping β and Ip constant
A. Burckhart, PhD network, September 30th, 2010
[C. Maggi, C. Konz]
27/30
First stability calculations performed,
but need to be repeated
•First calculations would
suggest a pedestal far away
from stability limit, but:
•CLISTE output had too
low gradients
•Need to run CLISTE and
ILSA again, carefully
comparing CLISTE results
with experimental data
[C. Konz]
A. Burckhart, PhD network, September 30th, 2010
28/30
Overview
• Motivation: why study ELMs?
• Overview on peeling ballooning theory and ELMs
• Evolution of Te, ne and pe during the ELM cycle
• Stability calculations
• Conclusions
A. Burckhart, PhD network, September 30th, 2010
29/30
Conclusions and to-do’s
Study of ELM cycle with high temporal and spatial resolution sheds new
light on both transport and stability of the pedestal
• Interplay between recovery of Te and ne observed on all analyzed AUG
discharges, still to be studied on JET
• Better characterization of Te recovery might be possible with TS
(no ECE shine through)
• Maximum of pe and jb can be reached well before ELM onset
• While peeling ballooning model is consistent with limits to pedestal pressure,
there is still a physics ingredient missing to explain the ELM trigger
- Influence from pressure further inside to be assessed
- Temporal evolution of the stability to be characterized, comparison
between stability of slow and fast ELM-cycles
A. Burckhart, PhD network, September 30th, 2010
30/30