• Gas Puff Imaging (GPI) diagnostic• Summary of C-Mod GPI results• GPI diagnostic set-up in NSTX• GPI data from NSTX ‘01 run • Interpretation of GPI signals• Tentative conclusions • Plans for 2002 run

Edge Turbulence Imaging on NSTX and Alcator C-Mod

S.J. Zweben1, R.A. Maqueda2, J.L. Terry3, B. Bai3, C.J. Boswell3, C. E. Bush4, D. D’Ippolito5, E.D. Fredrickson1, M. Greenwald3, K. Hallatschek6, S. Kaye1, B. LaBombard3, R. Maingi4,

J. Myra5, W.M. Nevins7, B.N. Rogers8, D. Stotler1, and X.Q. Xu7

1 Princeton Plasma Physics Laboratory 2Los Alamos National Laboratory 3MIT Plasma Science and Fusion Center 4Oak Ridge National Laboratory

5 Lodestar Research Corp 6 Max-Plank-Institut fuer Plasmaphysik 7Lawrence Livermore Laboratory 8 Dartmouth University

Motivations for Studying Edge Turbulence

• Very likely determines radial power and particle flow through SOL onto first wall (“scrape-off length”)

• Probably controls L / H transition (turbulence suppressed ?)

• Maybe determines density limit (turbulence increased ?)

• Maybe affects RF power coupling through edge

• Good place to study physics of tokamak turbulence, since both measurements and modeling are relatively simple

Gas puffnozzle

GPI telescope

Diodetelescope

B

Gas Puff Imaging Diagnostic

• Gas puff imaging (GPI) telescope views neutral line emission from He or D2 gas puff ≈ along B field at the plasma edge (like BES but uses neutral gas instead of NBI)

S(photons/cm3) = no f(ne,Te) A

where the radiative decay rate is A >> 107 sec-1 for these lines.

• Space and time variation of neutral light emission is measured with fast- fast gated cameras and PMs or PDs on discrete chords to determine edge turbulence structure (assumes kII << k)

• Gas puff changes plasma density by ≈ 1% in C-Mod and ≤10% in NSTX, but this is unlikely (based on previous experience) to perturb edge turbulence

Summary of C-Mod GPI Results

• Edge turbulence imaged over ≈ 6 cm poloidally x 3 cm radially at outer midplane

• Poloidal k-spectra from GPI compared with DBM code and BOUT for one case

• Videos at 250,000 frames/sec often show “blobs” moving radially and poloidally

GPI Diagnostic Set-up in NSTX

Gas manifold

Side-viewing re-entrant window

Bay B

Local magnetic

field

• Similar to C-Mod system but using re-entrant port instead of coherent fiberbundle and elongated gas manifold instead of single-point gas nozzle

• Generally have used He puffs in NSTX and Deuterium puffs in C-Mod (although results from both are similar in each machine)

Location of GPI View (2001 Run)

#105710 0.200 sec 0.215 sec

0.216 sec 0.230 sec

• Kodak camera views ≈ 30 cm x 30 cm area just above outer midplane

• Fast chords view 2 cm diameter “spots” in 7 channel radial array with PM tubes(bandwidth ≈ 100 kHz, digitized at 500 kHz)

12 cm

GPIview

NSTX GPI Diagnostic Issues

• He puff itself is unlikely to affect the local edge turbulence- see similar edge turbulence in natural recycling light - see same turbulence in GPI over a range of x5 in He puff level- in C-Mod, see same frequency spectrum in LP (with or without puff) - search for such an effect in PBX-M was negative (Pedrosa, PoP ‘95)- need to compare GPI turbulence signals with LP signals in NSTX

• Spatial resolution due to angle between GPI sightlines and local turbulence filaments probably ≈ 2-3 cm (est).

• Time resolution set by HeI radiative decay time ≈ 0.014 µsec

• Relative effect of ne and Te fluctuations on GPI not yet known- in theory these should be in phase and have a similar spectrum, with (Te/Te) / (ne/ne) ≤ Ln / LT, where “ < “ may be due to parallel electron heat conduction (e ≈ 1/kIIve ≈ 20 µsec), depending on magnetic flutter [BOUT got (Te/Te) / (ne/ne) ≈ 4 in SOL and ≈ 2 in edge of NSTX DN]3

GPI Results from NSTX in 2001

• Typical 2-D GPI images using LANL Kodak Camera

• Typical signals from fast chords (#105637, 0.7 MA, 3.0 T)

• GPI spectra, fluctuation levels, correlation lengths, etc.

• H-L transitions from ELM-free case (#105710, 0.7 MA, 4.5 T)

• Results from XP#34 (SN vs. DN, Ip / BT scans) - LATER

Typical 2-D Images vs. Time from Kodak

#105724(0.7 MA, 3 kG)density scan

#105710(0.7 MA, 4.5 kG)

H-L transition

#105711(0.7 MA, 4.5 kG)

no-H mode

see mpeg files

Typical Shot with GPI data

• Moderate density shot from Stan’s density scan 7/3/01 (no NBI or ICRH)

Time (sec)Time (sec)

Typical Signals from Fast GPI Chords

• Signals from 7 radial chords digitized for 0.128 msec @ 500 kHz

• Near outer wall see “intermittant” fluctuations with ~ 100% modulation

• Nearer center, see “Gaussian” fluctuations with ≈ 20% modulation

outer

inner

1 msec100 msec

Comparison of GPI and Mirnov Data

• GPI power spectrum is broadband with most of power below ≈ 50 kHz• Broadband GPI spectrum exists without coherent MHD present • Coherent MHD or ST cause small modulation in GPI signals

Time Time

Frequency spectrum of GPI Frequency spectrum of Mirnov

E. Fredrickson

250 kHz

no puff

Statistics of Inner GPI Chord

• Looks like the usual “Gaussian” plasma turbulence

Statistics of Outer GPI Chord

no puff

• Looks like the usual “intermittent” SOL turbulence

Radial Profile of GPI Fluctuations

0

0.2

0.4

0.6

0.8

1

0.8 0.85 0.9 0.95 1 1.05 1.1

624625627630632637/.18

r/a at midplane from EFIT

-0.5

0

0.5

1

1.42 1.44 1.46 1.48 1.5 1.52 1.54

624625627630632637/.18

Major radius R (m)

0

0.05

0.1

0.15

1.42 1.44 1.46 1.48 1.5 1.52 1.54

624625627630632637/.18

Major radius R (m)

• Relative fluctuation level goes from ≈10% @ r/a≈0.85 to ≈100% outside separatrix

• Autocorrelation times ≈10-20 µsec @ r/a≈0.85 to ≈30-100 µsec outside separatrix

• Radial correlation lengths ≈ 4-7 cm in all cases (FWHM)

=> These are all at least qualitatively similar to edge turbulence measured elsewhere

Relative fluctuation level Autocorrelation time Cross-correlation w/#4

Fluctuations vs. Time (shot #105637)

• Relative fluctuation levels, autocorrelation times, and and cross-correlation coefficients are approx. constant over 0.16-0.20 sec

• GPI signal levels rise by ≈ x5 during this time due as He gas enters, while plasma density rises by only ≈ 10% • Discharge changes after ≈ 0.2 sec as ne increases and Te drops (maybe in part due to MHD after this time)

GPI Signal levels Relative fluctuation level Autocorrelation time

0

1

2

3

4

5Sig. 1sig. 3sig. 4sig. 5sig. 7

0.16 0.18 0.2 0.22 0.24 0.26

Time (sec)

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08auto 1auto 3auto 4auto 5auto 7

0.16 0.18 0.2 0.22 0.24 0.26

Time (sec)

0

0.2

0.4

0.6

0.8

1

1.2

0.16 0.18 0.2 0.22 0.24 0.26

rel fluct 1rel fluct 3rel fluct 4rel fluct 5rel fluct 7

Time (sec)

Radial Propagation of Edge Turbulence

• Often (not always) a radially outward group velocity, especially in the outer channels, at Vg ≈ 105 cm/sec (seen with LP in many machines)

• Also consistent with radially outward “blob” motion seen in GPI in C-Mod SOL

Raw data Cross-correlation functions

Outer

Inner

Outer

Inner

Edge Turbulence Signals at L-H Transition

Time (sec) Time (sec)

• L-H transition occurs within ≈ 1 msec on most channels

• Perhaps preceeded by large “bursts” in outermost channels

• Data quality for L-H transitions is poor due to low signal levels

H-mode L-modeOuter

Inner

Outer

Inner

L-modeL-H

He puff

Edge Turbulence Signals at H-L Transition

Time (sec) Time (sec)

• H-L transitions occurs within ≈20 µsec in most channels

• Transition seems to propagate outward at ≈106 cm/sec

• Transition seems to start with coherent mode (ELM ?)

H-mode L-mode

ELM ?

Outer

Inner

Outer

Inner

0

50

100

150

200

250

300

350

400

Te

ne

#105710 @ 0.210 sec

(H-Mode)

0

50

100

150

200

250

300

350

400

142 144 146 148 150 152 154

Major radius R (cm)

ne

Te

#105710 @ 0.226 sec

(L-Mode)

0

0.2

0.4

0.6

0.8

1

1.42 1.44 1.46 1.48 1.5 1.52 1.54

710/.20710/.21710/.218710/.22

Major radius R (cm)

L-mode

H-Mode

Fluctuation Profiles in H- vs. L-mode• Fluctuation level seems to be lower in H-mode (but non-zero) in #105710 etc,

as expected from appearance of “flat” H-mode emission in images

• Only one Thomson Scattering point within range of GPI data

Te=13.3 eVne=1.1x1012 cm-3

Te=8.9 eVne=2.2x1012 cm-3

sep.

DEGAS-2 Simulation of GPI in NSTX

• Average HeI emission cloud can be simulated given edge profiles

• For Te ≈ 10 - 50 eV, HeI line emission ne0.5 to1.0 and Te

1.5 to -0.5

• H-mode emission is narrower due to narrower edge profiles

L-modeH-mode

D. Stotler

Tentative Conclusions

• Edge turbulence as seen by GPI in NSTX looks qualitatively similar to that in tokamaks, at least near outer midplane

• Most dramatic variation comes with H-mode, where edgeturbulence level seems to be reduced by x2 or so

• GPI fluctuation levels need to be corrected for the variation ofHeI light emission vs. ne and Te derived from edge profile measurements and neutral modeling (similar to BES)

• Autocorrelation times and correlation lengths from GPI areprobably close to correct, since the atomic physics justmultiplies small fluctuations by ≈ some constant

Plans for 2002

1. Add fast channels to make simultaneous 7 radial x 7 poloidal array DONE

2. Set up new Phantom fast camera for GPI imaging @ 1000 frames/sec (more sensitive and more pixels than Kodak camera) DONE

3. Swap Phantom with Princeton Scientific Instruments camera @ 1 million frames/sec for about 1-2 months (SBIR with PSI) EST. MID-APRIL

4. Get GPI data for the “Edge Characterization” XP along with good edge profiles (TS, LP, reflectometry), and cross-check GPI data with other edge fluctuation data (LP, reflectometer, USXR, etc.) MAY ? 5. Feed edge profiles to BOUT and DEGAS 2 to calculate expected GPI signals from theoretical simulations and compare with GPI data

6. Repeat (4)-(5) for other conditions, especially L/H mode and density limit