An Ultra-High Gradient Cerenkov Wakefield

Acceleration Experiment at SLAC FFTB

J.B.Rosenzweig*, S. Hoover*, M.J. Hogan♣, P. Muggliℵ, M. Thompson*,
G. Travish*, R. Yoder*
*
UCLA Dept. of Physics and Astronomy, 405 Hilgard Ave., Los Angeles, California 90095
♣
Stanford Linear Accelerator Center, Stanford California 94309
ℵ
University of Southern California, University Park Campus, Los Angeles, CA 90089

Abstract: The creation of ultra-high current, ultra-short pulse beams (Q=3 nC, σ z= 20 µm ) at the
SLAC FFTB has opened the way for very high gradient plasma wakefield acceleration
experiments. We study here the use of these beams in a proposed Cerenkov wakefield experiment,
where one may excite electromagnetic wakes in a simple dielectric tube with inner diameter of
few100 microns that exceed the GV/m level. We discuss the scaling € of the fields with design
geometric design parameters, and choice of dielectric. We also examine measurable aspects of the
experiment, such as the total coherent Cerenkov radiation energy one may collect, and the
expected aspects of dielectric breakdown at high fields.

With the recent theoretical and experimental interest in laser acceleration, the issue of
using dielectrics to support large accelerating fields has gained prominence. Various
analyses have indicated that GV/m accelerating fields should be possible in dielectric-
based laser accelerators1, as long as illumination times are very short2. Dielectric
accelerators may also be powered directly by charged particle beams, via wake-field
excitation3,4. This mechanism has been studied in depth over the last several years, but
with the maximum fields limited to 10’s of MeV/m by the lack of ultra-short drive
beams. According to scaling laws (described below), however, the recently achieved 20
µm pulse-length beams obtained at the SLAC FFTB SPPS facility may give over GV/m
longitudinal fields. This possibility has led to a preliminary proposal, described in this
paper, to use this beam in ultra-high field dielectric wakefield experiments.
Before discussing the experimental goals, we first present a heuristic analysis of the
wakefields produced by the beam. In this analysis, as with the experiment, we refer to the
schematic drawing of Fig. 1. To estimate the wakefields experienced by the beam, we
note that the radial fields produced at the dielectric beam boundary at r=a are
2N r m c 2
eE r ≅ − b e e exp(−ξ 2 /2σ z2 ) , (1)
2πσ z a
where σ z is the rms bunch length and ξ = z −v b t . The radial field inside of the dielectric
is smaller by ε−1; additionally, we know that the angle of the field lines with respect to r
5
is the Cerenkov angle € , cos(θ c ) = 1 ε , and the ratio of the longitudinal-to-transverse
€ fields inside of the dielectric €
at r=a is E z / E r = tan(θ c ) = ε −1. With continuity of E z at
the dielectric boundary, we have that the maximum decelerating electric field is
€
€
CP737, €
Advanced Accelerator Concepts: Eleventh Workshop, €
edited by Vitaly Yakimenko
© 2004 American Institute of Physics 0-7354-0220-5/04/$22.00
811
Downloaded 10 Oct 2007 to 169.232.154.218. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp
 2N b re me c 2 ε −1
eE zdec ≅ − exp(−ξ 2 /2σ z2 ) . (2)
2πσ z a ε
We note that the decelerating field follows the beam current profile, a result that will be
verified in simulation. This heuristic analysis is expected to give a good estimate as long
as the radiated Cerenkov field does not reflect off of the conducting boundary at r=b, and
superimpose € field within the beam. Much more complete analyses based on traditional
wakefield mode impedances are available in the literature6,7,8,9, but these do not give
numerical values for the field, or functional dependences on design parameters easily. We
note that for the parameters given in Table 1, which are derived from known beam
performance at the SPPS beam line, that the estimated decelerating field, which is nearly
as large as the radial field at the dielectric boundary, is 4.25 GV/m. This is well above
that expected to break down laser-excited structures. We will return to this point below.

FIGURE 1. Geometry of Cerenkov wakefield experiment, using a metal-clad dielectric tube with inner
radius a and outer radius b.

TABLE 1. Nominal parameters for Cerenkov wakefield experiment at SLAC FFTB.

Dielectric inner radius a 125 µm

Dielectric outer radius b 360 µm
Dielectric relative permittivity ε 3
Beam charge Q 3 nC
Rms bunch length σ z 20 µm
Beam energy 30 GeV

In order to validate the€ model for energy loss to Cerenkov wakes, and explore
implications of Eq. 2, we have performed a series of simulations using the code OOPIC10.
An example of the output of such a simulation is shown in Fig. 2, which displays a map
of the longitudinal field induced, from parameters slightly different from nominal, for
aiding visualization. One can see that the decelerating field on the beam indeed is
localized in a pulse that follows the beam current. The wakefield behind the beam
propagates in the dielectric at the Cerenkov angle until it strikes the outer wall, and
returns towards the axis. This wake waveform consists of many frequency components,
which stay nearly coherent, reproducing a pulse structure repetitively behind the beam.
These frequency components can be deduced from the Cerenkov radiation propagation
time in the dielectric, which for one “bounce” can be written as
2(b − a) ε 2(b − a) ε
T≅ = . (3)
c sin(θ c ) c ε −1

€ 812
Downloaded 10 Oct 2007 to 169.232.154.218. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp
 The fundamental mode wavelength λ1 excited in the structure, as well as its harmonics
λn , can be approximated well using time; the fundamental is the difference in the
position of the Cerenkov front after two bounces, and the beam, which travels at nearly
the speed of light,
€
2cT  cos(θ c )  4 (b − a) ε  1  4 (b − a)
€ λn ≅ 1− = 1−  = ε −1 . (4)
n  ε  n  ε n

This harmonic relationship can be seen in the wake waveform shown in Fig. 2, and
examined quantitatively in the example plot of the on-axis value of E z vs. z in Fig. 3.
€
Here the prediction of Eq. 4 is for the distance between the positive E z pulse and the
negative pulse is, for Fig. 2, λ1 /2=2.82 mm, and for the parameters of Fig. 3, λ1 /2=0.68
mm. Both predictions are validated by the simulations. €
€
€ €

FIGURE 2. Example OOPIC simulation-derived false-color map of longitudinal electric field driven by a 2
nC, σ z =20 µm, 30 GeV beam, propagating in a dielectric tube of 300 micron ID, 2.3 mm OD, and ε=3.

In order to evaluate the validity of the expression given in Eq. 2, as well as the
€ acceleration fields behind the beam, we examine both the peak decelerating field inside
of the beam, as well as the accelerating peak which trails, as shown in Fig. 3, which
displays the on-axis value of E z vs. z. The positive and negative peaks were deduced
from such plots; the determination of the field inside of the beam suffers somewhat from
noise.
€

813
Downloaded 10 Oct 2007 to 169.232.154.218. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp
 9
4 10

9
2 10

E (V/m)
0
z

9
-2 10

9
-4 10

9
-6 10
0.5 1 1.5 2 2.5
z (mm)
FIGURE 3. On-axis dependence of longitudinal field from OOPIC simulation- longitudinal electric field
for case with Q=3 nC, σ z =20 µm (centered at z=1.82 mm), 30 GeV beam, a=120 µm, b=360 µm, and ε=3.

Parametric studies of the fields’ dependence on design parameters were performed

with OOPIC,€ scanning over a range of a and b, and investigating the effect of ε choice.
For the geometric scan, we kept constant the beam parameters: Q=3 nC, σ z =20 µm, 30
GeV, and also took ε=3, while varying a and b, together, with b=3a. The results of this
study are given in Fig. 4, in which we show the peak in the decelerating E z,d ,max , the
prediction of our heuristic theory for this quantity, and the first maximum in acceleration
€
field, as a function of a. It can be seen the functional dependence of E z,d ,max ∝ a−1 is well
reproduced in simulations. The numerical prediction for E z,d ,max is quite close for our
€
simple model, and is dominated by uncertainty in extracting this quantity from the
simulation data in the presence of noise. We note that the rms beam size in the FFTB
€
beam line at this point, σ x ≅ 5 µm, can be accommodated by this aperture. As noted in
Table 1, the length of the tube is taken to be 1 cm,€ and so the transverse beam size is not
a problem from the viewpoint of beam interception. It is, on the other hand, not
reasonable to look at operation with the assumed beam and a ≤ 50 µm, as the fields are
€ high at this point.
already excessively
For the scan of permittivity choice, we kept the beam parameters as before, and took
a=120 µm, b/a=3. In Fig. 5, we plot the same quantities as in Fig. 4, but here as a
€
function of ε. The heuristic model predicts that a maximum in decelerating field should
be observed at ε=2, and this is reproduced by the simulations. The maximum is quite
broad, however, and both quartz (ε ≅ 2.15) and sapphire (ε ≅ 3) would be good choices for
maximizing the wakefield response.

€ €

814
Downloaded 10 Oct 2007 to 169.232.154.218. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp
 10
1.5 10

E_d,max
10
E_a,max
1 10 E_d,theory

E (V/m)
z

9
5 10

0
40 60 80 100 120 140 160

a (µm)
FIGURE 4. Dependence of longitudinal fields in Cerenkov wake simulations, with b/a=3, ε=3, and beam
parameters Q=3 nC, σ z =20 µm, energy of 30 GeV.

9
6 10
€

9
5 10

9
4 10
E (V/m)

9
3 10
z

9
2 10

9
E_a,max
1 10 E_d,max
E_d,theory

0
0 2 4 6 8 10 12
ε
FIGURE 5. Dependence of longitudinal fields in Cerenkov wake simulations on ε, with a=120 µm, b/a=3,
and beam parameters as in Figs. 3 and 4.

While one may calculate wakefield responses well in excess of GV/m, in practice this
is the level at which we expect breakdown. Thus a series of experiments may be
conceived of which explore breakdown, by changing either the beam parameters, or the
dielectric tube geometry. At present, we are exploring the use of a “revolver” geometry,
in which 6-8 tubes are loaded into a precision revolving mount. One of the tubes would
be a metallic dummy, used only for alignment purposes. Note that as we are expecting
energy loss rates at ~GeV/m, that the total energy change in the beam will be on the order

815
Downloaded 10 Oct 2007 to 169.232.154.218. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp
 of 100 MeV, which is not measurable, given then that compression to 20 µm pulse length
requires an energy spread of over 1 GeV.
In the proposed experiments, we therefore are interested in two signatures of high
field Cerenkov wakes: detection of material breakdown and measurement of the coherent
Cerenkov radiation produced. This proposed wakefield scenario is unique in that it may
produce fields of interest in breakdown studies. We are particularly interested in learning
the differences between material breakdown in wake-excited (very short pulse — sub-
100 fs — relatively long wavelength) systems, as compared with optically (laser) excited
systems. Quantum absorption of photons to produce free-electrons in the material is
known to be an initiator of avalanche ionization. This mechanism may be mitigated in
the wake experiment, thus allowing higher fields to be tolerated. For beam-derived fields
in the beam geometry use here, the unipolar, large amplitude fields have led to extremely
efficient ionization and plasma creation in the E164X plasma wakefield experiment.
Possible detection methods for dielectric breakdown under present discussion include
visible light from the discharge, and measuring current flow with a voltage bias imposed
on the dielectric tube.
The propagating radiation fields associated with the energy loss of the beam also
represent another robust experimental signal which can be exploited. The amount of
energy lost to creating coherent Cerenkov radiation (CCR) power in our experimental
scenario is approximated as
QE N 2 r m c 2 ε −1
UC( ≅ b z,max Ld ≅ b e e Ld , (5)
2 2πσ z a ε
where Ld is the dielectric length. This radiation is emitted in narrow spectral lines, about
wavelengths given approximately by Eq. 4. To estimate the energy in each wavelength,
assuming that a large number of modes approximately equally spaced in frequency ω n
€
€ ( Δω = πc 2(b − a) ε −1, cf. Eq. 4) contribute to the flux, we note that Cerenkov radiation
is inherently created with efficiency11 proportional to ω n , and also that the coherence12
( 2
)
gives a multiplicative factor N b exp −(ω nσ z c ) . Equating the integral of €the over the
€ mode contributions to the total of Eq. 5, we have6
€  2
 π  3 / 2 nN b2 re me c 2σ z Ld  nπσ z
Un ≅   exp −  , n = 1,2,3... . (5)
 2€ 2a(b − a) 2 ε ε −1   2(b − a) ε −1  
 
The energy in each line is approximately proportional to n, up until a roll-off wavelength
at λn ≅ 2πσ z .
€The overall amplitude of the coherent Cerenkov radiation (CCR) signal, for
N b = 2 ×1010 , and σ z = 20 microns, is 50 mJ, which is easily detectable by a variety
methods, such as pyroelectric detectors and Golay cells. The collection of the radiation
€
may be aided by using an appropriately shaped axicon-like geometry, as shown in Fig. 1,
both to refract the radiation, and use Brewster angle incidence to suppress reflections at
€ €
the boundary. As the overall geometry has two surfaces indicated that can induce
transition radiation, coherent transition radiation (CTR) would be expected to form a
significant background for this CCR measurement. The total CTR energy emitted can be
estimated by looking at the total field energy in the beam outside of r1. This radiation

816
Downloaded 10 Oct 2007 to 169.232.154.218. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp
 (more correctly termed coherent diffraction radiation) energy can thus be estimated to
have total energy
E r2 3 N b2 exp(−z 2 /σ z2 ) N 2r m c 2  r 
UCTR ≅ ∫∫∫ d x ≅ ∫∫∫ 2 2
drdφdz ≅ b e e ln 2  . (7)
4π 2π σ z r πσ z  r1 
The upper limit in the logarithm is a bit arbitrary (and unimportant), as it reflects the
outer radius of the target. The ratio of the CCR to CCR energy, or signal-to-noise (STN),
is thus
€ UC( Ld ε −1
≅ . (8)
UCTR aln(r2 /r1 ) 2ε 2
For r1=125 microns, r2 =2 cm, Ld =1 cm, and ε =3, we have a ratio of 5.2, which is an
encouraging number. Note that because TR has a naturally flat spectrum, however, while
CR grows as ω , STN should be quite improved for the highest frequencies, and high-
€
pass filters should give further enhancement. Also, the CCR should have localized lines,
€ € € €
as opposed to the broadband CTR spectrum, and so use of band-pass filters should allow
a quite large improvement in STN.
€
Preliminary planning of an experiment at the FFTB using test-beam access is now
underway. This test, which should be performed in the first half of 2005, will incorporate
both experimental signatures for very large amplitude, coherent Cerenkov wake
excitation discussed above.

ACKNOWLEGMENTS
This work is supported by U.S. Dept. of Energy grant DE-FG03-92ER40693.

REFERENCES

1
J.B. Rosenzweig, A. Murokh, and C. Pellegrini, Phys. Rev. Letters 74, 2467 (1995).
2
D. Du, et al., Appl. Phys. Lett. 64, 3073 (1994).
3
W. Gai, et al., Phys. Rev. Lett. 61, 2756 (1988).
4
J.G. Power, et al., Phys. Rev. ST-Accel. Beams 3, 101302 (2000)
5
P. Cerenkov, Dokl. Akad. Nauk SSSR 2, 451 (1934).
6
King-Yuen Ng, Phys. Rev. D 42, 1819 (1990).
7
M. Rosing and W. Gai, Phys. Rev. D 42, 1829 (1990).
8
George Dome, in Proceedings of the European Particle Accelerator Conference, Nice, France 1990 (Editions
Frontiers, Paris, 1990), p. 628.
9
T-B. Zhang, J. L. Hirshfield, T. C. Marshall, and B. Hafizi, Phys. Rev. E 56, 4647 (1997).
10
D. L. Bruhwiler, et al. Phys. Rev. ST — Accel. Beams 4, 101302 (2001).
11
J.D. Jackson, Classical Electrodynamics, 2nd Ed, (Wiley, New York, 1975).
12
Y. Shibata, et al., Phys. Rev. E 44, R3449 (1991).

817
Downloaded 10 Oct 2007 to 169.232.154.218. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp