SRF Cavity Design Optimization for HighCurrent ERL

1
SRF Cavity Design Optimization for High-Current
ERL

• Haipeng Wang
• also Robert Rimmer, Mircea Stirbet, Gary Cheng,
  Genfa Wu and Gigi Ciovati, Frank Marhauser
• Jefferson Lab, Newport News, Virginia 23606, USA

2
Cavity cell shape optimization concept for HG and
HC

• A right design will come to an easy-achieved
  performance. Dont just copy other design for
  High Current solution.
• Elliptical cavity design survey using a
  normalized inner cell shape and parameters.
• Maximize R/QG toward low loss concept but
  maintain the iris size in 140mm diameter for
  750MHz cavity to reduce beam bunch longitudinal
  energy loss and transverse kick.
• Determine ERL (2-pass) CW beam excitation
  frequency spectrum and power deposition rate for
  1A, CW / 0.1A, CW / pulsed beam etc.
• Optimize equator shape into flatter to maintain
  R/QG but keep trapped HOMs (below beam pipe
  cut-off frequency) away from beam excitation
  resonance to avoid huge power deposition. This
  was done by using cavity dispersion curves.
• Using 5-cell structure to avoid trapped modes in
  a long cavity.
• Use same cell shape in whole cavity, but trim end
  half cell equator shorter to get field flat. This
  avoids multi-die design and save cost.
• Optimized shape also has to avoid multipactoring
  barrier.
• Determine BBU threshold based on the ERL optics.
  In second order, like for monopole modes, try to
  keep resonance frequencies away from the high R/Q
  dipole modes.
• Quadruple mode BBU will be next level
  optimization, because threshold current is
  normally higher.

3
Normalized cavity inner cell shapes in different
SRF projects
See our PAC2005 publication
This comparison is wavelength and design b
independent duo to TM010 mode.
4
Design survey by normalized parameters for
different SRF projects
5
Cavity R/Q verse iris radius
JLab High Current
6
Cavity Epeak/Eacc verse iris radius
JLab High Current
7
Cavity geometry factor verse equator radius
JLab High Current
8
Cavity Bpeakb/Eacc verse equator radius
JLab High Current
9
Cavity Low Loss concept mainly attribute small
iris
JLab High Current
10
Beam excitation spectrum variation depending on
operation modes
750MHz laser, 750MHz RF 1A, 1-pass.
Broadband impedance approach, more detail on my
Thursdays talk.
750MHz laser, 750MHz RF 1A, 2-pass, 50.2m pass
length
Narrowband impedance approach, more suitable for
design optimization.
75MHz laser, 750MHz RF 0.1A, 2-pass, 50.2m pass
length
11
Normalized HOM power spectrum of SNS beam in SRF
Linac
Based on San-ho Kims work on SNS and a
correction to his publication formula, an
algorism for ERL beam has been developed.
See JLAB TN-05-047
Higher-order-mode (HOM) power in
elliptical superconducting cavities for intense
pulsed proton accelerators NIM in Physics
Research A 492 (2002) 110
SNS beam time structure
12
Reproduce to S. Kims SNS results by MathCAD
Beam induced voltage in time domain
Time averaged HOM power normalized to R/Q in
frequency domain
13
HOM power spectrum for different ERL beam modes
operation
1A, 750MHz, (1.33nC) CW beam
0.1A, 75MHz laser, (1.33nC) 750MHz RF CW beam
0.1A, 75MHz, (1.33nC) 60Hz/250ms laser pulsed beam
14
Dispersion Curves calculated by MAFIA 2D for
single inner cell cavity
Dispersion curve for periodic RF structure
particle vbC
tan??group velocity
frequency
Light speed cones
tanfphase velocity
Phase advance
0
p
Stop bands
Pass bands
15
HOM frequencies have to avoid beam bunch
excitation peaks
16
Cell shape optimization to push trapped HOMs away
from resonance
17
Calculated HOM power has to be dumped into load
or absorber
18
Trapped modes are more dangerous than traveling
modes
19
MAFIA time domain calculation for broadband
impedance
Wake Field potential
Gaussian beam bunch FFT
FFT cosine window
20
MAFIA time domain calculation for broadband
impedance
Beam Impedance normalization
amplitude and phase of wake potential FFT
Phase folding
21
broadband impedances contribute to power
deposition or BBU below beam pipe cut-off
frequency
bunch length 3cm
Monopole modes excited by on-axis beam bunch
monopole dipole modes excited by off-axis 5cm
beam bunch
22
Broadband HOM power calculation for above cut-off
Calculation formula
8.5 kW
Total 18kW/cavity
2 Amp
5 cell
1.33 nC
can be improved by model measurement or high
power computing
23
Multipactoring simulation by FishPact
unit is in mm
Final JLab High Current cavity cell shape design
24
End cell shape is simple same cell shape,
trimming on equator

• Same inner and end cell shape.
• Trim end half-cell by 8.4 mm.
• R/Q increase 57.
• Only one die design is needed.
• Reduced HOM degeneration from center cells.
• No B field enhancement at end-cell.
• Bead-pull measurement on copper model cavity
  confirms that the field flatness can be tuned
  within 99 with minimum effort.

25
Narrowband HOM impedances calculated by MAFIA2 D

• Complete R/Q table of monopole and dipole modes
  is in the CDR report.
• R/Q values agree with broadband calculation for
  the modes below cutoffs.
• High R/Q modes are away from 1.5, 3.0, 4.5,GHz.
• To be used for RR/QQload calculation to get
  narrowband impedance.

26
Copper model measurement and data fitting
techniques

• S21 from BP to BP.
• Labview automation.
• Ceramic bead-pull on-axis or off-axis.
• End groups staggered 30o or 60o.
• 5, 6, 7-cell assembly.
• Data sets with dummy loads or shorts.
• Rotatable coupling antennas.

• First data set has been fitted by the 5-peaks
  fitting algorism originated from SLAC to get
  freqs and Qs and amps.
• More detail will be present at my Thursdays
  talk.

27
Cavity waveguide fundamental power coupling
calculation using half scale and MWS eigen mode
simulation

• Qext calculation has cross-checked with MAFIA,
  HFSS, Omega-3P.
• MWS uses EM BCs (Balleyguiers method) at
  waveguide port.
• Qext is accurate but not the E/M field in coupler
  section.
• Only using impedance BC on the waveguide port
  can properly simulate the SW and TW in the
  coupler region like frequency domain solver (MWS
  and HFSS).
• Coupler transverse kick is cavitys gradient,
  beam current and phase dependent. Detail study
  is on going.

d336.2 mm
28
Unmatched waveguide transition and window can
alter Qext by more than 25 due to S11 notch
frequency change
Dogleg waveguide taper design avoids ceramic to
the line of sight to electron beam.
Unmatched dogleg taper and window change Qext
from 9.8e5 to 7.2e5
Dogleg shape affects S11 notch frequency
position, not very broadband.
Window ceramic/iris position is sensitive to the
S11.
An optimized waveguide coupler design un-changes
the Qext in order to minimize the ceramic, iris
and wall heating.
29
Cavity coupling external Q bench measurement

• Using TRL calibration and S21 measurement
  technique
• To avoid ghost waveguide mode, adapter
  removal procedure has been specially developed
  for the waveguide coupler measurement using
  Agilent 8753ES ENA.
• Waveguide bumper separation variation
  measurement data agrees with MWS simulation
  prediction.

30
FPC to FPC RF isolation check with MWS
Specification -60dB Design -64.36 dB at 19.362
inch FPC center to FPC center distance
31
HOM load design and benchmark by
MAFIA/HFSS/ANSYS/MWS
Original SLAC B factor HOM load
Benchmarks between codes using scaled SLAC
waveguide load.
MAFIA simulated electric field plot
32
Dielectric measurement on ceramic samples of HOM
load material

• Statistic measurement on SLAC PEP IIs HOM
  tiles.
• More than 5 vendors and gt 20 samples have been
  measured. No particular vendor is preferred.
• Load dimensions and 4kW/load input HOM power for
  1A class, 748.5 MHz cryomodule 160W/load input
  HOM power have been used for thermal and
  mechanical designs.

33
HOM waveguide load MWS designs for 4kW and 160W
160 W design
4kW design
34
ANSYS RF-thermal coupled simulation

• 99.7 of input power converted into heat.
• 99.5 of the RF heat is absorbed in tiles. Only
  0.5 surface heat loss.

Tile brazing OFHC posts
Water inlet temp. 25 oC
?Tmax 62.3K
Water outlet temp. 37 oC
35
Summary

• With a carefully choice of cavity shape, both
  high gradient and high current goals in the terms
  of low loss and avoidance of HOM resonance can be
  achieved.
• For high current ERL, minimize the monopole
  modes power deposition to the cavity wall is the
  first order of optimization. Remember that this
  requires the cavity shape optimization or the
  HOM frequencies QA (tuning) after the fact of
  manufacture is related to the beam excitation
  (bunch time structure).
• Heavy HOM damping by a good waveguide end group
  coupling and well controlled cryogenic cooling is
  next important. A full ANSYS RF-thermal coupled
  3D FE simulation is on going and successful.
  Details of engineering solution have been given.
• Dipole HOMs induced BBU is the second order of
  design optimization. With staggered waveguide
  dampers, up to quadruple modes are well damped.
  The BBU threshold and beam merger scheme are
  under investigation.
• MAFIA 3D time domain simulation is very powerful
  tool to do the HOM impedance calculation both in
  narrow and broad bands. The bench HOM measurement
  data are better understood now, and details will
  be described on my Thursdays talk.
• A team of SRF physicists, mechanical engineers
  has developed a good integration design of the
  Ampere class cryomodule by the utilization of
  computer simulations.
• All tests of single or 5-cell cavity at VTA proof
  that this cavity shape design optimization is
  successful.