RFQ Development for FETS
Simon Jolly
Imperial College
7th January 2010
FETS RFQ Development
• FETS will utilise a 4m-long, 324 MHz four-vane RFQ channel, consisting of four resonantly coupled sections.
• RFQ focuses beam from LEBT and accelerates it to 3MeV, ready for Chopper/MEBT.
• 4-vane cold model showed agreement between CST simulations and bulk RF properties.
• Previous beam dynamics simulations, based on field maps produced with a field approximation code, provide a baseline for the new design.
• Novel design method currently under development to combine CAD and electromagnetic modeling with beam dynamics simulations in GPT.
7/1/10 2Simon Jolly, Imperial College
FETS Integrated RFQ Design
• Would like to have a method of designing RFQ where all steps are integrated:– Engineering design.– EM modelling.– Beam dynamics simulations.
• Integrating design steps allows us to characterise effects of:– Fringe fields and higher order modes.– Particular CNC machining techniques and
options on beam dynamics.
7/1/10 Simon Jolly, Imperial College 3
CST Mesh Density
• A lot of effort on optimising meshing in CST.• Need a field map that gives transmission results
similar to RFQSIM.• Important to quantify whether we can model
Electrostatic field of vanes with enough accuracy in CST to measure beam dynamics.
• Non-trivial: modelling 30mm x 30mm x 4m volume with micron accuracy.
• Changes in beam dynamics MUST be unaffected by coarseness of CST field meshing so we can compare to optimised field.
7/1/10 Simon Jolly, Imperial College 4
First CST Field Map
• Produced with CST, tracked with GPT:– Transmission = 99%
– Mean energy = 1.31 MeV
– Energy rms = 261 keV
7/1/10 Simon Jolly, Imperial College 5
Increase Vane Radius
7/1/10 Simon Jolly, Imperial College 6
An educated guess: increasing tip radius should improve quadrupole field close to beam axis. Try field mesh with larger tip radius…
Vane Tip Field Maps
7/1/10 Simon Jolly, Imperial College 7
3.24 mm 7 mm
Second CST Field Map
• Increase vane radius to 7mm “by hand”:– Transmission = 99.6%
– Mean energy = 3.02 MeV
– Energy rms = 14 keV
7/1/10 Simon Jolly, Imperial College 8
7 mm Vane Tips
• The Good News: we now have a vane shape that produces comparable transmission results to RFQSIM field!
• The Bad News: we can’t actually build it…
• 7 mm vane tip radius exceeds Kilpatrick Limit…
7/1/10 Simon Jolly, Imperial College 9
22/4/09 State Of Play (Previous UKNF)• CAD modelling process now pretty mature: can model vane, rod
and “vod” with parameter adjustment on-the-fly (everything except no. of cells).
• Models import into CST and output to GPT: beam dynamics simulations well understood.
• Definite issues in CST modelling:– We think we see “plateau” at 4,100 points and 0.5mm point
spacing of output file.– Clear discrepancies with Alan’s field map: good transmission
and energy with = 7mm vane, = 3.2mm rod but NOT = 3.2mm vane! It should match…
• Are the discrepancies real or down to faulty method?• Need to ensure we’re not re-inventing the wheel: RFQ’s have been
designed before without this process. • Need to ensure CAM systems will understand our CAD models so
we can manufacture what we’re designing (this is the point…).
7/1/10 Simon Jolly, Imperial College 10
Corrected Input Values
• Couldn’t understand why increasing the radius of curvature made the results closer to Alan’s field map.
• The solution: – Inconsistencies in the units of the RFQSIM
input file.– Reported radius was incorrect, so RFQSIM
actually modelling a larger radius.• Reran RFQSIM with corrected values to get
new modulation parameters.7/1/10 Simon Jolly, Imperial College 11
Third CST Field Map
• Using corrected input values:– Transmission = 93.8%
– Mean energy = 2.77 MeV
– Energy rms = 470 keV
7/1/10 Simon Jolly, Imperial College 12
Incoherent Acceleration
• Better results:– Some particles accelerated to the full 3 MeV.– Still not achieving coherent acceleration, even with
new modulation parameters.– This result published at PAC’09.
• Most obvious way to improve results is to increase mesh density:– Prior studies with 7mm vanes showed current mesh
density was sufficient.– Smaller radius of curvature (3.24 mm) needs tighter
mesh to model vane tip region.– CST at computational limit: split RFQ into 5
sections and mesh separately.7/1/10 Simon Jolly, Imperial College 13
Fourth CST Field Map
• Modelling RFQ in 5 sections:– Transmission = 99.8%
– Mean energy = 1.98 MeV
– Energy rms = 378 keV
7/1/10 Simon Jolly, Imperial College 14
High Mesh Particle Losses• Many possible causes of
problems in fourth CST field map:– Try hexahedral rather than
tetrahedral meshing.– Try open and tangential
boundary conditions.• Better meshing with
tetrahedral mesh and tangential boundary.
• Eventually found the problem was in the reconstruction of the RFQ from the five sections:– one of the sections was using
the parameters from a different section, and so the modulation pattern was incorrect.
– reran with correct reconstruction
7/1/10 Simon Jolly, Imperial College 15
Fifth CST Field Map• Five sections reconstructed
into whole RFQ, high mesh density, tangential boundary:– Transmission = 100%– Mean energy = 3.03 MeV– Energy rms = 12 keV
7/1/10 Simon Jolly, Imperial College 16
Varying CST Mesh (640 points)
7/1/10 Simon Jolly, Imperial College 17
Varying CST Mesh (4700 points)
7/1/10 Simon Jolly, Imperial College 18
Particle Tracking For High/Low Mesh
Conclusions• Finally seeing match between RFQSIM optimised field map and
CST field map from CAD modelling.• Each step has required a lot of effort for small progress, but that
progress has proved extremely valuable.• Advantages of this process already starting to become apparent: for
example, very easy to modify CAD model based on thermal/stress simulations (Scott Lawrie) and measure effect on beam dynamics.
• Next steps:– Output CAD model to Comsol, repeat process from CST to
produce more easily adaptable field map (tighter integration with Inventor and Matlab).
– Compare CST coarse, fine, Comsol and RFQSIM field maps point-by-point to determine whether discrepancies are a result of poor field mapping or more accurate modelling of vane tips.
7/1/10 Simon Jolly, Imperial College 20
Transverse Field Map Comparison
7/1/10 Simon Jolly, Imperial College 21
Transverse Field Map Comparison
RFQ Design Parameters• RFQ parameterised by 3 (+ 1)
parameters:– a and m parameters define
modulation depth.– r0 defines the mean vane
distance from the beam axis and is derived from a and m.
gives the radius of curvature (vane) or mean radius (rod).
– L defines the length of each cell (half sinusoidal period).
• For field approximation method, these values generated for idealised RFQ field.
7/1/10 Simon Jolly, Imperial College 23
L
a
r0 (mm)ma
0 (mm)
L/2
rod axis
beam axis
RFQ Parameters (from TUP066, LINAC06)
7/1/10 Simon Jolly, Imperial College 24
RFQ Design Stages
7/1/10 25Simon Jolly, Imperial College
RFQ CAD Modelling
• Autodesk Inventor CAD package used to model RFQ cold model (and a lot more besides…).
• Inventor can dynamically link to parameters in Excel spreadsheet:– Change spreadsheet
parameters and model updates automatically.
– Use spline to approximate sinusoidal vane shape: only 2% difference.
7/1/10 Simon Jolly, Imperial College 26
CST MicroWave Studio E-field Modelling
• Four vanes from inventor imported by a macro.
• Model cut into 6 sections (5 plus matching section) for ease of modelling and to increase CST mesh density.
• Potentials and boundary conditions defined in the macro.
• Run solver to produce electrostatic field map.
7/1/10 Simon Jolly, Imperial College 27
Beam Dynamics Simulations
• GPT used for beam dynamics simulations.
• Import electrostatic field map from text file produced by CST.
• Integration algorithm traces particle movements through time-varying field.
• Compare results to field map from optimised RFQ field expansion.
7/1/10 Simon Jolly, Imperial College 28