2D Impurity Flow Imaging on MAST
using Coherence Imaging
Scott Silburn
Centre for Advanced Instrumentation, Durham University, UK
James Harrison (CCFE), Clive Micahel (ANU), Ray Sharples (Supervisor, Durham), John
Howard (ANU), Hendrik Meyer (Supervisor, CCFE), Kieran Gibson (York)
FUSENET PhD Event, York, 24th June 2013
Outline
• Introduction
• Coherence Imaging Technique
• The MAST Coherence Imaging Diagnostic
• Initial Data
Introduction – Flow Measurements
• Understanding plasma flow is important for the successful design
and operation of magnetic fusion devices, e.g. it impacts impurity
transport, exhaust & plasma confinement. High quality flow data is
required to develop our understanding and could help benchmark
modelling.
• Doppler spectroscopy of ion line emission is commonly used to
measure impurity flows, but provides a limited number of lineintegrated spectra. Complex or spatially extended flow patterns are
hard to interpret from this.
• Coherence imaging is a technique which can be used for high
spatial and time resolution Doppler shift imaging, developed at
Australian National University and demonstrated on DIII-D, TEXTOR,
KSTAR.
Introduction - MAST
•
•
•
•
Mega Amp Spherical Tokamak - CCFE, Oxfordshire, UK
R = 0.85m, r = 0.65m
Plasma current typically 400 – 900 kA, pulse length up to 700ms
Up-down symmetric
• Open vacuum vessel geometry with many ports offers excellent access
for optical diagnostics
Coherence Imaging Technique
Coherence Imaging Technique
• Doppler shift causes a tiny change in fringe frequency
• This causes a phase shift between the shifted and unshifted interferograms after many wave cycles
• Doppler shift can be determined from the phase
shift at a known value of N
Coherence Imaging & MAST System Design
Camera lenses
Back-to-back
Delay plate Savart plate
N wave delay, Fringes
Camera
lens
Plasma image with
interferogram superimposed
From
Plasma
Linear polarisers, axes
at 45° to plates
Narrow (~3nm)
band pass filter
Coherence Imaging on MAST: Specs
• Targeted ion species:
– C2+ (465nm)
– C+ (514nm)
– He+ (468nm)
• Field of view 9° - 40°
(up to 1.4m x 1.4m of
plasma cross-section)
• Detector: 1024x1024 pixels (but ~10px spatial resolution across fringes), up
to 3kHz full frame.
• Flow resolution ~1 – 4 km/s
• Can view the lower divertor, or radially or tangentially at the midplane
Initial Data
• Data has been obtained with time resolution of 1 – 20ms in all 3 species
• Various flow phenomena observed, including oppositely directed toroidal Carbon
flows above & below the midplane (at the high field side) at early times during the
shot. Quantitative analysis & interpretation is now required.
Centre column
Initial Data
• Initial divertor data appears to show C III flows towards both divertor targets (Also
seen in divertor coherence imaging on DIII-D).
• The divertor data contains enough information to enable tomographic inversion of
flow profiles….
Initial Data – Tomographic Inversion
• Assuming toroidal symmetry and flow mainly along B, obtaining intensity and flow
cross-sectional profiles is a large least-squares problem.
• This is solved using a simple Simultaneous Algebraic Reconstruction Technique
implementation in MATLAB, which is under testing & development with divertor data.
• Fits so far look reasonable (but more work required), time to process a raw image
frame to inverted flow profile is ~ 3min on a laptop.
Line integrated flow (km/s)
C III, 28841 at 362ms (H-Mode)
Flow data
Best fit
Conclusions
• Wide angle flow imaging of multiple impurity species in both
the Scrape-off-Layer & Divertor of MAST has been
demonstrated using coherence imaging.
• Flows observed include reversal about the midplane during
startup in Carbon, and quantitative analysis and interpretation
of the data is now required.
• Initial divertor data shows C III flowing towards the divertor
targets (as on DIII-D)
• Tomographic inversion is being developed to obtain divertor
flow profile cross-sections
Additional Slides
Coherence Imaging & MAST
Remove horizontal
fringes using FFT
Raw data image
Extract local fringe
contrast using FFT
divide out
instrument function
Extract local fringe phase
using FFT, subtract
instrument function.
Choice of Fixed Delay
•
Due to the multiplet structure of the carbon lines, the contrast exhibits a beating pattern
with delay. The delay must be chosen to optimise the fringe contrast and satisfy T << Tc for
the tomography.
Fringe Contrast
Fringe Contrast
•
Left: Calculations of
fringe contrast as a
function of waveplate
thickness and plasma
temperature for C III
(top) and C II (bottom),
based on measured
MAST divertor spectra.
We will have multiple interchangable plate options to allow optimisation for different
measurements.
Position Registration & Available Views
• Line-of-sight calibration done in two stages:
• Calibration of camera properties (focal length, distortion etc) done with test
pattern photos in the lab
• Calibration of position and pointing on MAST done with flashlamp-illuminated
vessel photos and feature matching, given the camera properties.
Test pattern image
Radial midplane view
Tangential midplane view
Position Registration & Available Views
• Expected coil locations line up well with vessel photos
• EFIT lines up reasonably with plasma images
Calibration Setup
Cd spectral
lamp
Small (~20cm)
integrating sphere
Calibration Setup
• Monitoring of calibration drifts when in use:
fibre feed of radial midplane light in to
image..
• Designed to provide a zero flow reference in
part of the image….but doesn’t work yet.
Tomographic Inversion
• With so many line-of-sight measurements, it’s possible to use the data to
reconstruct poloidal profiles of plasma quantities, under some
assumptions e.g. toroidal symmetry.
• Ray trace the pixel lines-of-sight through the machine, and project on to
reconstruction grid in the R-Z plane.
Camera data
Calculated grid cell
/ pixel weightings
Noise etc
Plasma emissivity
profile
Tomographic Inversion
Line-integrated intensity
Recovered R-Z emission profile
what happened to Ti?
• But it also means we can’t measure ion temperature.
Tomographic Inversion
Simulated
line integrated image
• Test cases reconstructed with flow errors of ~1 – 5km/s
– But these cases will be easier than real data! Room for improvement?