Realistic Modeling of 4-Rod RFQs with CST Studio

Proceedings of IPAC2014, Dresden, Germany
S.S. Kurennoy, Y.K. Batygin, E.R. Olivas, and L.J. Rybarcyk, LANL, Los Alamos, NM, USA
RFQ accelerators are usually designed and modeled
with standard codes based on electrostatic field
approximations. There are recent examples when this
approach fails to predict the RFQ performance accurately:
for 4-rod RFQs 3D effects near the vane ends can
noticeably influence the beam dynamics. The same
applies to any RFQ where the quadrupole symmetry is
broken, e.g., 4-vane RFQ with windows. We analyzed
two 201.25-MHz 4-rod RFQs – one recently
commissioned at FNAL and a new design for LANL –
using 3D modeling with CST Studio. In both cases the
manufacturer CAD RFQ model was imported into CST.
The EM analysis with MicroWave Studio (MWS) was
followed by beam dynamics modeling with Particle
Studio (PS). For the LANL RFQ with duty factor up to
15%, a thermal-stress analysis with ANSYS was also
performed. The simulation results for FNAL RFQ helped
our Fermilab colleagues fix the low output beam energy.
The LANL RFQ design was modified after CST
simulations indicated insufficient tuning range and
incorrect output energy; the modified version satisfies the
design requirements. Our PS results were confirmed by
multi-particle beam-dynamics codes that used the MWScalculated RF fields.
Radio-frequency quadrupole (RFQ) accelerators are
now common in front ends of all modern ion linacs. Usual
RFQ design codes, e.g., Parmteq [1], rely on electrostatic
field approximations that are justified for classical 4-vane
RFQs with perfect quadrupole symmetry. Many modern
RFQs contain elements that break this symmetry. RFQ
vane modulations introduce symmetry perturbations to
create a longitudinal accelerating field, which is the basic
idea of the RFQ structure. Their effects are accounted in
analytical field representation in the codes or can be
calculated by solving an electrostatic problem. However,
additional field effects can be introduced by asymmetric
elements like vane windows in split-coax designs or stem
supports in 4-rod RFQs. These effects are more
complicated and can’t be easily taken into account in
electrostatic calculations, even in 3D, but can influence
beam dynamics in some cases. We discuss such RF
effects in two 4-rod RFQs that we have recently analyzed
in details [2, 3] with CST Studio [4].
leaving only the resonator vacuum volume in the CST
model. The outer boundaries are assumed perfectconducting for EM simulations. The resulting model for
the new LANL RFQ is shown in Fig. 1. Here the RFQ
vacuum vessel, in light-blue, is 175-cm long.
Figure 1: CST RFQ model (top) and its side view.
The tuners between the adjacent stems (bottom) are at
different heights. They are adjusted in the CST model to
make the inter-vane voltage flat, see Ref. [3].
We studied the RFQ models using the CST MicroWave
Studio (MWS). The mode frequencies and RF fields were
calculated by the AKS eigensolver that provides more
accurate surface approximations. In the latest CST version
of 2014 the CAD import has been improved, which may
allow using its more efficient tetrahedral eigensolver. The
RF fields have some interesting features. One is the
longitudinal electric field in the end gaps that separate the
vane ends from the RFQ box walls (end-gap bumps) as
illustrated in Fig. 2 for the FNAL 4-rod RFQ, see in [5].
It is important to have an accurate model of the RFQ
cavity for its EM analysis. In both cases under
consideration a CAD model from the manufacturer Kress
GmbH was imported into CST. The model was further
simplified by removing details unessential for EM
calculations. The RFQ cavity walls were also removed,
Figure 2: On-axis longitudinal field in FNAL RFQ.
Two RFQ models were studied [2]: with wide beam
pipes attached to the RFQ cavity (A) and with narrow
ones (B). The RFQ accelerating field – the oscillating part
in the curves in Fig. 2 – is produced by the vane
04 Hadron Accelerators
ISBN 978-3-95450-132-8
A08 Linear Accelerators
Copyright © 2014 CC-BY-3.0 and by the respective authors
Proceedings of IPAC2014, Dresden, Germany
Copyright © 2014 CC-BY-3.0 and by the respective authors
modulation; the longitudinal field, however, also has two
peaks, near the entrance and exit. The RFQ cavity extends
from x = -60 cm to x = 60 cm. Note that the RF fields
extend into the beam pipes. The end-gap longitudinal
field exists because the quadrupole symmetry is broken
near the RFQ ends; it would vanish in a perfectly
symmetric structure. The end-gap bumps depend on the
diameter of beam pipes attached to the RFQ, cf. Fig. 2,
while the accelerating fields coincide in the two models.
In model B, the narrow beam pipes trap the fields in the
gaps, so the peaks are higher and shorter. In model A
without vane modulation (green curve) there is no
accelerating field, as expected, but the end-gap bumps are
the same as with modulation. The exit-end bump can
change the output beam energy depending on the beam
transit-time factor. This was the reason for the low output
energy in the FNAL RFQ in its original configuration (B).
Opening the exit beam pipe by removing an end-wall plug
(change from B to A) restored the design energy [5].
One should mention that 3-D electrostatic computations
of 4-vane quadrupole structure do not exhibit bumps since
the static field remains quadrupole-symmetric, cf. Fig. 3.
Even in 4-vane RFQs with vane cuts that distort the
symmetry, e.g., the split-coaxial structure [6], the end-gap
peaks are present, though in the particular case [6] they do
not influence beam dynamics as drastically as here.
Figure 3: On-axis longitudinal field: RF vs. electrostatic.
Another feature typical for 4-rod RFQs is a small
transverse horizontal (parallel to the RFQ ground plane)
field component on its geometrical axis. It is illustrated in
Fig. 4 that plots LANL RFQ on-axis field components:
longitudinal El, horizontal Eh, and vertical Ev (along the
stems) versus the longitudinal coordinate s. Here s = -x
because in the RFQ model of Fig. 1 the beam travels from
right to left. This effect is due to the fact that the
transverse electric fields between two upper vanes are
stronger than between two lower ones. It results in the
displacement of the center of the transverse quadrupole
field down, to the ground plane, from the RFQ axis by
0.45 mm, to be compared to the 4-mm vane aperture. The
effect is known in higher-frequency 4-rod RFQs [7]. We
found no noticeable influence of this feature on beam
dynamics in the two 4-rod RFQs we studied [2, 3].
Figure 4: On-axis field components in LANL RFQ.
The end-gap bumps of the longitudinal field can also be
seen in Fig. 4; their effect was taken into account in beam
dynamics simulations. The initial design of the LANL
RFQ was modified based on our simulation results to
adjust the frequency tuning range and the final energy [3].
Multi-particle beam dynamics modeling based on the
MWS calculated RF fields can be performed with either
the CST Particle Studio (PS) particle-in-cell (PIC) solver
or other codes. We cross-checked our PS results for the
new LANL RFQ with two well-known dynamics codes,
Parmela [1] and Beampath [8], and the results agree well.
This comparison allowed us to find that the final
transverse emittances reported in [3] were incorrect: they
were calculated in the rotated coordinates of the CAD
model of Fig. 1, so the horizontal and vertical emittances
were mixed. The corrected results and more details on PS
simulations for the LANL RFQ can be found in [9].
Initial “matched” distributions were generated using
LAACG codes [1]. Matched CW beams of 10K macroparticles, one RF period long, were generated for several
different currents using matched-beam Twiss parameters
at the RFQ vane entrance and tracing particles back to the
beam-pipe entrance. The resulting distributions were
converted to CST PS format and repeatedly injected into
RFQ during 10 - 100 RF periods. As an illustration, Fig. 5
shows macro-particles in the PS model of the LANL RFQ
with 12-mA current at t = 405 ns (81.5TRF) after the
injection of 10x10K particles started. The total number of
remaining macro-particles at this moment is about 97.4K,
and most of them are densely packed in bunches. The
particle energy is indicated by color; the energy scale is
overlapped on the right. The particles propagate from
right to left. More than 10 bunches are formed, due to the
longitudinal space-charge push at the train ends; a few
leading bunches visible in Fig. 5 are already accelerated
and approaching the exit. Trailing low-energy particles
are not captured in bunches. To exclude effects of space
charge in the bunch-train head and tail, we analyze the
average energy W, transverse normalized r.m.s. emittance
İt, longitudinal emittance İl, and current transmission only
for the bunches near the train center.
ISBN 978-3-95450-132-8
04 Hadron Accelerators
A08 Linear Accelerators
Proceedings of IPAC2014, Dresden, Germany
The heat flux calculated in post-processing the MWScomputed RF fields was used for thermal-stress analysis
with ANSYS [10]. The dissipated power is 77 kW at
100% duty for ideal copper surfaces. Cooling is provided
by water running through cooling channels in stems and
vanes. The temperature distribution in a structure slice
near the RFQ exit is shown in Fig. 6 for the duty factor of
18% (to account for realistic surface conductivity). The
temperature range is less than 40° C even in this extreme
case. The maximal structure deformations were also
found acceptable [3].
Figure 5: Particles in the LANL RFQ model for 12 mA.
Our PS results are summarized in Table 1. The initial
distribution had a transverse normalized r.m.s. emittance
0.2 ʌ mm·mrad in both transverse planes. The transverse
emittance values are corrected compared to [3], cf. [9].
Table 1: Results for Different Currents in LANL RFQ
Detailed 3D modeling of RFQs is now possible with
CST Studio. It includes importing or building a 3D CAD
model followed by its EM analysis and multi-particle
beam dynamics PIC simulations plus thermal analysis for
high-duty RFQs. We successfully used this approach for
two 4-rod RFQs with CST Studio running on a PC. Our
results revealed 3D effects that were not found by
standard design and analysis but significantly influence
the RFQ performance. Other groups have used different
tools, see Refs. [6, 11], for building RFQ models, their
EM analysis, and beam dynamics studies. Independent of
a particular choice of tools, this modern approach predicts
the RFQ performance more reliably compared to the
standard RFQ design codes.
W, keV
İt, ʌ
PS simulations of RFQs can be significantly sped up if
the RFQ structure is cut transversely to a narrow region
around the vane aperture for particle runs. In fact, other
codes [1, 8] do the same by restricting the simulation
volume only to the beam aperture region. We compared
results of PS runs with full and cut volumes and found no
difference in the output beam parameters for the two 4rod RFQs studied. This trick allowed us to run the initial
beams up to 500 RF periods long on a PC, see in [9].
The LANL RFQ is designed to operate at noticeable
duty factors, up to 15%, so the structure thermal and
stress analysis should be performed.
Figure 6: Temperature distribution from ANSYS [3].
The authors would like to thank R. Garnett, J. O’Hara,
and T. Wangler (LANL), C.-Y. Tan (FNAL), A. Schempp
and J. Schmidt (IAP, Frankfurt), and J. Haeuser (Kress
GmbH) for support, useful information, and stimulating
[1] Los Alamos Accelerator Code Group,
[2] S.S. Kurennoy et al., “EM and Multi-particle Beam
Dynamics Modeling of 4-Rod RFQs,” IPAC2013,
Shanghai, China, p. 3978 (2013);
[3] S.S. Kurennoy et al., “Design Analysis of the New
LANL 4-Rod RFQ,” PAC2013, Pasadena, CA USA,
p. 333 (2013);
[4] CST Studio Suite,
[5] J.S. Schmidt et al., Phys. Rev. ST Accel. Beams 17,
030102 (2014).
[6] B. Mustapha et al., Phys. Rev. ST Accel. Beams 16,
120101 (2013).
[7] B. Kubek et al., PAC11, New York, China, p. 1888
[8] Y.K. Batygin, Nucl. Instr. Meth. A539, 455 (2005).
[9] S.S. Kurennoy, “EM and Beam-Dynamics Modeling
of the New LANL RFQ with CST Studio,” report
LA-UR-13-28693, Los Alamos (2013).
[10] ANSYS Inc.,
[11] S. Jolly at al., Nucl. Instr. Meth. A735, 240 (2014).
04 Hadron Accelerators
ISBN 978-3-95450-132-8
A08 Linear Accelerators
Copyright © 2014 CC-BY-3.0 and by the respective authors
I, mA