A SLAB-SYMMETRIC DIELECTRIC-LOADED STRUCTURE FOR

HIGH-GRADIENT ACCELERATION AT THZ∗

R. B. Yoder† and J. B. Rosenzweig

Dept. of Physics and Astronomy, UCLA, Los Angeles, CA 90095-1547

Abstract the relatively low Q (100–1000) of a planar structure, al-

low the production of large gradients (hundreds of MeV/m)
We present a design for a slab-symmetric accelerating
before breakdown limitations are reached.
structure to be resonantly excited at terahertz frequencies.
The device, consisting of a vacuum gap between dielectric-
lined walls, combines the advantages of a slab geometry Structure Overview
(including strong suppression of transverse beam wake- Fundamentally, our device consists of a pair of parallel
fields and low power density) with the existence of a reso- dielectric planes, separated by a narrow vacuum gap and
nant mode having phase synchronism with relativistic elec- bounded above and below by a thin conductive layer. Pe-
trons. Accelerating fields of hundreds of MeV/m are pre- riodic slots in the conductor provide a means for coupling
dicted when the structure is powered by a high-power FIR radiation into the gap and also enforce longitudinal peri-
radiation source in development at UCLA. Simulation of odicity in the structure fields. When the correct resonant
the structure fields is described and compared with theory, geometry is achieved, the mode pattern is dominated by a
and an experimental program is discussed. longitudinal standing wave having a phase velocity exactly
equal to the speed of light. The concept is illustrated in
Fig. 1.
INTRODUCTION
conductor y
In all laser-based acceleration concepts, the large trans-
laser light
verse fields of a laser pulse must be converted to useful lon- z
λ
gitudinal fields, a problem solved in the microwave regime x
by coupling radiation into resonant structures. While this ebeam ε
dielectric a b
acceleration method is venerable, conceptually simple, and
capable of generating very large axial fields, the well-
known physical and technical limitations of scaling such
structures to near-optical wavelengths have strengthened (a) (b)
the cases for plasma-based and far-field optical accelera-
tors, despite the considerable challenges of these latter ap- Figure 1: (a) Schematic drawing of the slab-symmetric
proaches. However, many of the harmful effects produced structure; (b) a cross-section in x, showing the parameter
by the presence of field-shaping boundaries near the accel- definitions. The ‘infinite’ dimension (x) extends into the
erating particles are mitigated if the system is allowed to page; electron beam propagation is in +z, with radiation
become very large in one transverse dimension, in effect (also polarized in z) impinging on the structure from the
producing a two-dimensional device. This paper extends +y direction.
and refines previous work on laser-based acceleration in a
“slab-symmetric” device, and introduces a new design for a A number of versions of this concept have been pro-
structure driven resonantly at 340 µm, which is to be exper- posed [2, 3, 4], each of which took a slightly different
imentally investigated at UCLA using a novel high-power approach to generating the correct longitudinal periodic-
terahertz source. ity in the structure properties. Periodic slots, as used here,
The case for slab structures rests on the inherent transla- represent the optimum choice for coupling power preferen-
tional symmetry of their geometry, which forces the fields tially into the accelerating mode while satisfying the speed-
of a speed-of-light wave to be invariant in the small trans- of-light resonance condition. Numerical modeling of the
verse dimension; this leads in turn to the suppression structure, discussed below, is used to optimize the coupling
of transverse (dipole-mode) wakefields, which otherwise slots for best performance.
would dominate the beam-structure interaction. A slab-
beam can then contain very high charge, acting in effect Structure Modes
as a large number of parallel beamlines. The use of dielec-
tric materials (which can withstand surface fields of a few Mode analyses for this class of slab-symmetric struc-
GV/m for pulses of a few picoseconds) [1], coupled with tures have been outlined in several recent papers [3, 5];
the solutions hold for any periodically-modulated struc-
∗ Supported by US DoE grant no. DE-FG03-92ER40693 ture, regardless of the mechanism used for the modulation.
† [email protected] We summarize the results briefly here for the fundamental
speed-of-light mode, which is the one of interest for accel- is clearly a very strong field in the coupling slot itself; how-
eration. Using the axes and structure parameters defined ever, in this design the slot acts as a quarter-wavelength
in Fig. 1, and assuming initially that the structure is truly transmission line, causing the field to fall to zero at the in-
infinite in the x direction, we obtain an axial electric field ner aperture and hardly perturb the accelerating mode. The

 large slot field indicates that these slots can cause a substan-
E0 tial change to the structure resonant frequency; it also limits
Ez (y, z) = √ cos(kz z)eiωt
AE0 cos[
− 1 kz (b − y)] the amount of power that can be introduced. Eliminating
(1) the quarter-wave condition on the slot length, while mis-
where the upper line holds within the vacuum gap (|y| < tuning the structure to compensate for the frequency shift,
√ (a < |y| < b), and
a) and the lower√within the dielectric may overcome this problem.
where A = csc[
− 1(b − a)kz ]/
− 1. The transverse
fields are proportional to y and zero on axis. Imposition
of boundary conditions at metal and dielectric boundaries
gives a resonance condition which specifies allowed values
of a, b and
for a given kz = 2π/λ0 , where λ0 is the
free-space laser wavelength.

E z /E 0
The true translational symmetry of the structure is of
course broken in any physical device with finite extent. The
effect of a non-infinite structure on the accelerating mode
has been addressed in [5], where field variation in the x
direction is imposed by the addition of conducting walls
at x = ±L, where L  a. It is shown that the axial z (µm)
y (µm)
field in the gap gains a cosh(k⊥ y) dependence, but with
sufficiently large structure aspect ratio (and hence small
wavenumber k⊥ ), the deviation from flatness of the field Figure 2: Surface plot showing time snapshot of axial
can easily be kept below 1%. fields, normalized to the amplitude of the incident wave.
The field along the z axis shows the accelerating mode,
with a field spike in the coupling slot and a plane wave in
Optimized Parameters the space above the structure.
Starting from a wavelength of 340 µm, we choose
= 3
(close to values for e.g. silicon) and find that the lowest res- The growth of the accelerating fields in time is shown in
onance occurs for a = 115 µm, b = 145 µm, or in other Fig. 4, where we see that after a relatively long fill time (on
words a vacuum gap of full width 230 µm, bounded by a the order of 500 ps) the accelerating field amplitude within
dielectric layer 30 µm thick. To optimize the coupling of the structure reaches a value near 15 times that of the inci-
external power into the structure, we can adjust the width dent wave. The same figure compares the amplitude of the
of the coupling slots as well as the thickness, or depth, of accelerating mode with that of its most important competi-
the conducting layer itself. One must bear in mind that tor, the so-called ‘zero-mode’, with fields that are constant
since the drive radiation is not cut off in the coupling slots in z. As the figure shows, the zero-mode component of
(they are very long in the x direction), the slots are in ef- the fields remains less than 1% of the accelerating compo-
fect waveguide sections which can perturb the resonance nent. The fill time of 600 ps suggests a loaded Q = ωτ for
significantly. The high-Q results presented below are ob- the structure on the order of 3000, a relatively high value.
tained for a slot of width 5 µm and 69 µm deep. For experiments driven with pulsed lasers, it will be impor-
tant to decrease this fill time while maintaining a reason-
able shunt impedance. Nevertheless, the design exhibited
NUMERICAL RESULTS
here would produce an accelerating gradient of 400 MeV/m
Field Modeling when illuminated at 100 MW/cm2 .

Extensive modeling of the ideal slab geometry has been

Wakefield Stability
carried out using a two-dimensional finite difference solver
to verify the resonance condition and to optimize the cou- The advantageous transverse wakefield properties of the
pling of external fields into the structure. In this code, slab-symmetric geometry are one of the reasons for interest
a single structure period is illuminated by a plane wave, in these structures, and our assertion that damaging trans-
with symmetry enforced at the midplane; periodic bound- verse wakefields vanish for very wide beams has been veri-
ary conditions are used in z. Typical output is shown in fied by analysis and simulation. Ref. [5] contains an analyt-
Fig. 2. Fig. 3 shows the field within the gap in more detail, ical calculation of the wake potentials in a slab-symmetric
where it is clear that the accelerating field is nearly invari- structure, and OOPIC simulations of particle wakes in this
ant in y, as predicted by (1). The transverse field is zero at 340 µm structure were presented in an earlier paper [7]. In
the location of the peak accelerating field; see Fig. 3. There both cases, the transverse electric and magnetic forces in
 ation wavelengths. As mentioned earlier, the accelerator
is to be driven with a new terahertz source under devel-
opment at UCLA. The large resonant vacuum gap in this
wavelength range—more than 0.2 mm—makes it feasible
to inject the 11–14 MeV beam from the Neptune photoin-
jector, with normalized transverse emittance in the range of
6–10 π mm mrad, into the structure successfully.
We expect to obtain multimegawatt laser radiation at 340
µm using a difference frequency generation scheme: two
frequencies from the Neptune terawatt CO2 laser will be
mixed at high power in a gallium arsenide crystal, with con-
version efficiency into the difference frequency near 1%.[6]
The two input frequencies, as well as the output radiation,
are non-collinear in order to maintain synchronism over
a relatively large (several centimeter) interaction length.
Output power levels in excess of 100 MW are projected;
experimental work is currently in progress.
Diagnosis of the structure can be carried out through
cold testing as well as by acceleration experiments and
the possible detection of wake radiation. Transmitted light
from the lower coupling slots, as shown in Fig. 1, can be
used to measure the filling of the structure. Since the de-
sign is easily scaleable to other wavelength regimes, it may
also be useful to evaluate prototype structures using con-
ventional laser radiation, e.g. at 10.6 µm.

Figure 3: Contour plots showing Ez (above) and Ey (be- CONCLUSION

low) over one structure period. The shaded area represents
the conductor (with coupling slot), and the dark line shows We have described a slab-symmetric dielectric-loaded
the dielectric/vacuum boundary. structure which serves as a resonant laser-driven accel-
erator with advantageous transverse stability for high-
16
charge beams. When such an accelerator is powered by
14 a submillimeter-wave source at 340 µm, the structure di-
Accelerating mode
mensions become ample for acceleration of a slab elec-
12
tron beam with achievable transverse size, and with real-
10 istic input power levels we calculate acceleration gradients
of 400 MeV/m. Experimental investigation of these struc-
E z/E 0

tures, including cold testing in several wavelength regimes,

6
is planned at the UCLA Neptune facility.
4

2
REFERENCES
Fabry-Perot mode x 10

0 [1] A. C. Tien, et al., Phys. Rev. Lett. 82, 3883 (1999).

0 200 400 600 800 1000 1200 1400
time (ps) [2] J. Rosenzweig, A. Murokh and C. Pellegrini, Phys. Rev. Lett.
74, 2467 (1995).
Figure 4: Amplitude of axial electric fields within the vac- [3] J. B. Rosenzweig and P. V. Schoessow, in Advanced Acceler-
uum gap vs. time, normalized to amplitude of the incoming ator Concepts, Eighth Workshop, 1998, pp. 693–700.
wave. The accelerating mode has longitudinal periodicity; [4] P. V. Schoessow and J. B. Rosenzweig, Proceedings of the
the non-accelerating Fabry-Perot mode (multiplied by 10 1999 Particle Accelerator Conference, pp. 3624–3626.
here) does not.
[5] A. Tremaine, J. Rosenzweig and P. Schoessow, Phys. Rev. E
56, 7204 (1997).
the vacuum gap canceled within machine precision. [6] S. Tochitsky, private communication.
[7] R. B. Yoder and J. B. Rosenzweig, in Advanced Accelerator
EXPERIMENT Concepts, Tenth Workshop, 2002, pp. 331–340.

The geometry of the experiment which is planned for

the UCLA Neptune facility is dictated by available radi-