APPLIED PHYSICS LETTERS 98, 202901 共2011兲

Resonant excitation of coherent Cerenkov radiation in dielectric lined

waveguides
G. Andonian,1,a兲 O. Williams,1 X. Wei,1 P. Niknejadi,1 E. Hemsing,1 J. B. Rosenzweig,1
P. Muggli,2 M. Babzien,3 M. Fedurin,3 K. Kusche,3 R. Malone,3 and V. Yakimenko3
1
Department of Physics and Astronomy, University of California, Los Angeles, Los Angeles,
California 90095, USA
2
University of Southern California, Los Angeles, California 90089, USA
3
Accelerator Test Facility, Brookhaven National Laboratory, Upton, New York 11973, USA
共Received 16 March 2011; accepted 27 April 2011; published online 18 May 2011兲
We report the observation of coherent Cerenkov radiation in the terahertz regime emitted by a
relativistic electron pulse train passing through a dielectric lined cylindrical waveguide. We describe
the beam manipulations and measurements involved in repetitive pulse train creation including
comb collimation and nonlinear optics corrections. With this technique, modes beyond the
fundamental are selectively excited by use of the appropriate frequency train. The spectral
characterization of the structure shows preferential excitation of the fundamental and of a higher
longitudinal mode. © 2011 American Institute of Physics. 关doi:10.1063/1.3592579兴

Dielectric lined cylindrical waveguides have long been
studied for use in advanced acceleration applications, as
they provide the potential high accelerating gradients for
the dielectric wakefield accelerator 共DWA兲, in a relatively
compact form in the difficult to access terahertz 共THz兲 spec-
tral region.1,2 Initial experiments using dielectric loaded
waveguides, at longer wavelengths, measured the wake po-
tentials of a driving bunch for different materials,3 while
more recent THz excitation experiments using single, ex-
tremely short, very high charge electron pulses demonstrated
⬎GV/ m fields before structure breakdown.4 Follow-on ex-
periments have demonstrated their applicability as sources of
narrow-band, coherent THz radiation.5 The characteristics of
the emitted radiation depend on the dielectric tube geometry,
dielectric material choice, inner/outer dielectric tube bound-
ary, as well as driving electron beam properties such as
charge, and bunch length or time structure. In preliminary
experiments, Cook et al.5 demonstrated single-mode excita-
tion of the fundamental mode 共⬃0.3 THz兲 limited by the
moderate single bunch length 共⬃1 ps兲 and choice of wave-
guide parameters 共restricted by the beam width at the lower
energy of 11 MeV兲.
Further applications of coherent radiation sources are
accessible using the higher energies and pulse train produc-
tion scheme employed at the Brookhaven National Labora-
tory Accelerator Test Facility 共BNL ATF兲.6 This pulse train
scheme employs a rigid mask inserted at a high-dispersion
point along the beamline, that introduces a periodic
temporal-transverse position correlation along the beam. Af-
ter the bend system, the transverse offsets are removed, leav-
ing a train 共typically three to six beamlets兲 with variable
spacing and bunch lengths. Additionally, the high energy,
low emittance beams at the BNL ATF allow for operations
with highly focusable, intense beams, resulting in high
power, high frequency beam spectral components from small
transverse size DWA structures.
Pulse trains are important in wakefield schemes due to
the need to introduce multibunch beam loading for high av-
erage energy efficiency. It is also critical in coherent Ceren-
kov radiation 共CCR兲 production, as it can allow selectivity in
resonant mode excitation by pulse train tuning. In extending
operation to pulse trains, one must also examine potential
limitations due to beam break-up 共BBU兲 instabilities medi-
ated by dipole modes.
Critical aspects of this array of physical effects are stud-
ied in the experiments reported here through examination of
CCR, including selection of the longitudinal oscillating
mode 共TM0n兲. The beam properties have direct effect on the
parameters of the emitted radiation, in particular in enabling
resonant excitation at different frequencies. Here the number
of beamlets and the consistency of beamlet bunch spacing
are of high importance. As such, in addition to CCR mea-
surements, an independent investigation of the pulse train
properties was carried out, with coherent transition radiation
共CTR兲 used as a diagnostic, and a linear and nonlinear beam
transport used to tune the pulse train.

Interferometer
Bolometer FIG. 1. 共Color online兲 Experimental layout at the BNL
ATF. The beam 共dotted line兲 is generated with appropri-
ate chirp in the photoinjector and linac 共not sketched兲
CTR CCR Collimating
Spectrometer and transported through a dispersive line where the
Optics
Dipole pulse train is generated by the mask. Two interferom-
Mask eter setups measure the generated CTR and CCR 共green
Insertable DWA solid line兲. The beam energy is measured with a dipole
Mirror
e-beam CTR Foil Structure CCD spectrometer at the beam dump.
with
with Horn
hole

a兲
Electronic mail: gerard@physics.ucla.edu.

0003-6951/2011/98共20兲/202901/3/$30.00 98, 202901-1 © 2011 American Institute of Physics

Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
202901-2 Andonian et al.
98.4
98.2
98.0
p (mec)
97.8
97.6
97.4
97.2
1.0 2.0 3.0 4.0 5.0 1.0 2.0 3.0 4.0 5.0
t (ps) t (ps)
FIG. 2. 共Color online兲 Longitudinal phase space of a simulated 4-pulse train
multibunch beam after transport through the dispersive line without correc-
tion for nonlinear dispersion effects. The beamlet spacing is not consistent
from pulse to pulse 共left兲. The beamlet spacing is consistent from pulse to
pulse when sextupole correctors are engaged in the dispersive line 共right兲.
The experiment was carried out at the BNL ATF 共Fig. 1兲.
For these measurements, the nominal electron beam energy
was 60 MeV, the charge 300 pC with a normalized emittance
of 1 mm-mrad. The pulse train is generated by the insertion
of a periodic mask at a high dispersion point along the trans-
port, and its periodicity is adjustable from 150– 600 ␮m by
varying the phase of the rf accelerating wave with respect to
the bunch timing. For example, running a pulse train of three
beamlets with period Lb = 495 ␮m requires operating at a
phase 3° forward of crest. The photoinjector/linac phase is
stable up to 0.5° due to the feedback loop, providing the
needed reproducibility in pulse train performance as required
by the interferometric measurements of the bunch train and
of the radiation.
Forward of crest running conditions generate a beam
with positive chirp 共beam head with higher energy than the
tail兲; conversely, backward of crest operation yields a nega-
tive chirp. This distinction is important not only for generat-
ing the appropriate bunch spacing for the pulse train but also
eventually for direct observation of energy gain and loss with
the dipole spectrometer at the end of the beam line. Unam-
biguous energy gain observation requires that the beam must
have a negative chirp for the spectrometer to resolve wake-
field acceleration of a trailing beamlet. Due to the low charge
of the trailing beamlet and the relatively short DWA structure
length, energy gain was at the limit of the spectrometer res-
olution.
The bunch spacing is measured using an insertable
foil that generates CTR. The radiation is transported to a
Michelson-type interferometer with a LHe-cooled Si-
bolometer detector. The resultant autocorrelation trace is
Fourier transformed to yield spectral content, thus character-
izing the pulse train properties.7 The peak position of the
Fourier transformed signal indicates the bunch spacing while
its width indicates the consistency of the periodicity. It was
observed that the width of the signal is diminished by tuning
two sextupole correctors placed at high dispersion points
along the dispersive line. The sextupole correctors serve to
mitigate the second order dispersion terms 共T566兲,8 and con-
sequently yield a more uniform periodicity. For the uncor-
rected case at the BNL ATF, ELEGANT9 simulations show
that T566 = 6.73 m whereas when sextupoles are appropri-
ately engaged, T566 is reduced to near zero and the beam
bunch spacing is consistent 共Fig. 2兲 in agreement with ex-
perimental observations.
After tuning, the beam is transported to and focused
through a 1cm long DWA structure-a hollow dielectric
共SiO2兲 cylinder coated with a thin layer of aluminum. The
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
Appl. Phys. Lett. 98, 202901 共2011兲
3.5 3.0
3.0 2.5
Signal [a.u.]
2.5 2.0
2.0 1.5
1.5 1.0
0.5
1.0
0.0205 0.0210 0.0215 0 200 400 600 800
Step[m] Wavelength [μm]
FIG. 3. 共Color online兲 CCR autocorrelation curve for the fundamental mode
共left兲 and Fourier transform of the interferogram 共right兲. The peak in the
spectrum is observed at 495 ␮m.
DWA tube has an inner radius, a = 100 ␮m and an outer
radius, b = 170 ␮m. Calculations show that the induced peak
accelerating field for this structure is 55 MV/m. The beam
traversing the dielectric structure emits CCR, which is ex-
tracted by an in-vacuum flat copper mirror with a 5 mm hole.
The hole allows the electron beam to continue to the beam
dump, which incorporates a dipole magnet spectrometer for
energy measurements. The CCR is launched from the DWA
structure through a cylindrical horn and extracted out of the
vacuum chamber through a TPX window. The radiation is
collimated using short focal length off-axis paraboloid mir-
rors to minimize exposure to humid air, which has strong
absorption bands in the THz frequency range. The collimated
CCR is transported to a second Michelson-interferometer
共also with a Si-bolometer detector兲 where autocorrelation
measurements yield the spectrum of the radiation pulse.
Initially, to excite the fundamental longitudinal mode
共TM01兲 of the DWA structure a 3-pulse train 共rms pulse
length ␴z ⯝ 55 ␮m, and charge per pulse ⬃30 pC兲 was
tuned to a beamlet spacing of Lb ⯝ 495 ␮m. The electron
bunch train was then transported through the DWA, where
the CCR was generated and characterized. With this train
spacing the Fourier spectrum of the radiation autocorrelation
trace showed a definitive peak at a wavelength of ⬃495 ␮m,
indicating resonant excitation at the fundamental TM01 mode
共Fig. 3兲.
The second resonant longitudinal mode 共TM02兲 of the
structure occurs at a wavelength of 195 ␮m. For this case, a
4-pulse train 共␴z ⯝ 25 ␮m兲 was tuned using the CTR/
sextupole method to a bunch spacing of Lb = 195 ␮m 共Fig. 4,
top right兲, and the beam transported to the DWA tube. The
emitted CCR was again autocorrelated 共Fig. 4, bottom left兲
and the spectral analysis indeed displayed a peak near a
wavelength of ⬃200 ␮m 共Fig. 4, bottom right兲. In order to
have the second mode dominate over the fundamental, the
beam tuning had to be performed carefully, because the cou-
pling to the fundamental mode is stronger. In addition, the
group velocity of the second mode is greater due to the
shorter wavelength in the DWA guide, and thus a relatively
lesser total signal is expected; the total energy in the nth
harmonic is related to its power, Pn, and time of propagation,
␶ p,n = L共1 − vg,n / c兲 / 共vg,n兲, in the structure as Un = Pn␶ p,n,
where L is the structure length and vg,n is the group velocity
of the nth harmonic.
As can be seen in Fig. 4, the beam tuning to give Lb
= 195 ␮m suppressed the signal of the TM01 mode to just
above the noise floor, and allowed the second mode to domi-
nate 共Fig. 4, bottom right兲. However, the spectrum also
showed a secondary peak near ␭ ⯝ 300 ␮m, which may be
attributed to spectral artifacts due to the rapid growth of the
202901-3 Andonian et al. Appl. Phys. Lett. 98, 202901 共2011兲

1.2 electric lined waveguide, with the preferred mode excitation

3.0 1.0 due to the periodicity of the relativistic bunch train. The
Signal [a.u.]

0.8 precise bunch train period adjustability was accomplished by

2.5 0.6 the careful manipulation of nonlinearities via sextupole cor-
2.0 0.4 rectors during the dispersive transport. Both electron beam
0.2 and radiation properties were modeled with start-to-end
0.0045 0.0050 0.0055 0 200 400 600 800 simulation codes and experimentally benchmarked. The se-
Step[m] Bunch Spacing [μm] lective tuning of bunch spacing allowed one to tune the CCR
1.2 output from fundamental, at a wavelength of 495 ␮m, to the
3.5
1.0 second resonant mode, at 195 ␮m, thus giving another
Signal [a.u.]

0.8 method–beyond DWA tube substitution–for CCR tuning, and

3.0
0.6
2.5
0.4
2.0 0.2
0.020 0.021 0.022 0 200
Step[m]
FIG. 4. 共Color online兲 CTR autocorrelation 共top left兲 and Fourier transform
共top right兲 demonstrating bunch spacing of ⯝200 ␮m. CCR interferogram
共bottom left兲 for second resonant mode and Fourier transform 共bottom
right兲. The peak in the CCR spectra is observed at ⯝195 ␮m. Excitation of
the fundamental mode is heavily suppressed.
Cerenkov radiation during the passage of the four-beamlet
bunch structure, or to the presence of a transverse dipole
mode. Indeed, the structure geometry is expected to yield
strong coupling to the second dipole mode 共HEM12兲 which
has a frequency of ⬃0.9 THz, close to the observed second-
ary peak. Dipole mode characterization is important in the
context of BBU instabilities afflicting pulse train applications
and will be pursued in follow up experiments.
These measurements demonstrate the ability of preferen-
tial wavelength selection using resonant excitation of a
bunch train with as few as three to four beamlets. Simula-
tions using OOPIC 共object-oriented particle-in-cell 2D simu-
lation code兲 共Fig. 5兲 confirm the experimental results and
also demonstrate the suppression of the fundamental at ap-
propriate bunch spacing.
In conclusion, we have shown the ability to select reso-
nant modes of coherent THz Cerenkov radiation from a di-
Power Spectrum
(3 bunches, 495μm spacing)
150 200
Power (a.u.)
thus introducing needed experimental flexibility for applica-
tions of CCR.
As such issues are critical for DWA applications, studies
400 600 800 are planned in future experiments having longer interaction
Wavelength [μm] lengths. We are initiating similar studies in the context of
slab-symmetric DWA experiments, where the geometry
yields smaller transverse wakes and thus diminished BBU.
Additionally, future studies on dipole modes 共HEM1n兲 will
be executed by manipulation of the beam transverse 共posi-
tion, angle offsets兲 properties. Other experiments following
in this series 共Refs. 4 and 5兲 will employ a single, very short
beam pulse to excite a large number of resonant modes,
which can then be selectively filtered to give a wider choice
of high power, narrow-band THz frequencies. Further, we
plan to obtain smaller radius beams by using permanent
magnet quadrupole focusing. This permits smaller aperture
tubes with higher DWA fields; these tubes will also include
those fabricated from materials such as chemical-vapor
deposition diamond. Higher fields and operation on correct
side of rf crest will permit unambiguous measurement of the
pulse train energy loss, and with a modified beam mask that
appropriately rephases a trailing beamlet in the wakefield,
acceleration.
Work supported by U.S. Department of Energy under
Grant Nos. DE-FG02-04ER41294, DE-AC02-98CH10886,
DE-FG03-92ER40695, and DE-FG02-92ER40745.
1
A. Tremaine, J. Rosenzweig, and P. Schoessow, Phys. Rev. E 56, 7204
共1997兲.
2
Power Spectrum M. Rosing and W. Gai, Phys. Rev. D 42, 1829 共1990兲.
3
W. Gai, P. Schoessow, B. Cole, R. Konecny, J. Norem, J. Rosenzweig, and
(4 bunches, 195μm spacing)
J. Simpson, Phys. Rev. Lett. 61, 2756 共1988兲.
4
M. Thompson, H. Badakov, A. Cook, J. Rosenzweig, R. Tikhoplav, G.

150
Travish, I. Blumenfeld, M. Hogan, R. Ischebeck, N. Kirby, R. Siemann, D.
100 Walz, P. Muggli, A. Scott, and R. Yoder, Phys. Rev. Lett. 100, 214801
100 共2008兲.
50 5
A. M. Cook, R. Tikhoplav, S. Y. Tochitsky, G. Travish, O. B. Williams,
50
and J. B. Rosenzweig, Phys. Rev. Lett. 103, 095003 共2009兲.
0 6
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 P. Muggli, B. Allen, V. E. Yakimenko, J. Park, M. Babzien, K. P. Kusche,
f (THz) f (THz) and W. D. Kimura, Phys. Rev. ST Accel. Beams 13, 052803 共2010兲.
7
P. Muggli, V. Yakimenko, M. Babzien, E. Kallos, and K. P. Kusche, Phys.
FIG. 5. 共Color online兲 OOPIC simulations for BNL ATF parameters with Rev. Lett. 101, 054801 共2008兲.
three bunches at 495 ␮m spacing showing enhancement of the fundamental 8
R. J. England, J. B. Rosenzweig, G. Andonian, P. Musumeci, G. Travish,
共TM01兲 mode 共left兲 and with four bunches spaced at 195 ␮m displaying and R. Yoder, Phys. Rev. ST Accel. Beams 8, 012801 共2005兲.
enhancement of TM02 and suppression of fundamental 共right兲. 9
M. Borland, APS Technical Report No. LS-287, 2000 共unpublished兲.

Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp