You are on page 1of 4

Nuclear Instruments and Methods in Physics Research A 393 (1997)376-379

Secton A

Space-charge oscillations in a self-modulated electron beam in

multi-undulator free-electron lasers
J. Rosenzweig*, C. Pellegrini, L. Serafini’, C. Ternieden, G. Travish
UCLA Department of Physics and Astronomy, 405 Hilgard Ave., LOS Angeles, CA 90095-1547, UsA

We examine here the oscillation of electron-beam density perturbations (longitudinal plasma oscillations) produced at
the exit of a high-gain free-electron laser (FEL) by the action of the FEL instability. These oscillations, which are
analyzed in the case of both a free-space drift and a dispersive section, can degrade the bunching of the beam in the drift
between undulator sections in multi-stage FELs. The impact of these oscillations on the gain of an FEL in an undulator
following such a drift, as well as the case of an optical klystron is studied.

1. Longitudinal space-charge short wavelength limit, the electric field associated with
a pure harmonic density perturbation n, = n,, - nbD =
In space-charge dominated beams derived from RF 6nb cos(k,[) of wavelength 1, = 2n/k,, with nbOthe unper-
photoinjectors, such as typically moderate energy driven turbed beam density, and < = z - vbot (distance meas-
(few tens of MeV) free-electron lasers (FELs), the collec- ured in the Galillean frame moving with the beam) is
tive transverse motion is characterized by single-com-
ponent plasma behavior. The free expansion due to the 47ce6nb
repulsive self-forces is controlled by the externally im- E,,r - ~ sm(k,i).
posed focusing lattice, producing beam envelope, or sur- k,
face plasma oscillations. In contrast, for many beams, the
longitudinal space charge forces do not produce signifi- For a finite, constant density bunch distribution with
cant debunching, and small efforts (running slightly off of a hard edge at radius a, the longitudinal electric field
RF crest) are necessary to compensate the energy slew inside of the beam is given by
introduced by the space-charge forces. This suppression
of longitudinal forces is due to the fact that, in the beam
rest frame, the bunch length is much larger than the
radius y60,>>~,., and the longitudinal surface plasma
oscillation frequency 01, ,,is much smaller than the trans-
verse frequency O,~. In the limit that the beam radius is large compared to the
For short-wavelength longitudinal density perturba- oscillation wavelength in the beam rest frame (k,a>>y,),
tions (microbunching) such as are introduced by the FEL the correction factor to Eq. (1) found in Eq. (2) ap-
process, however, the field is no longer transverse, as the proaches unity. In the case of the FEL the radiation
longitudinal density gradients may have a rest frame wavenumber has a strong dependence on the normalized
scale length shorter than the beam width. In the ultra- energy, k, cc Yg, and so this situation, where the longitu-
dinal space-charge force attains its one-dimensional limi-
ting strength, is to be expected for any beam of moderate
energy or above (i.e. for the UCLA IR FEL k,cr, z yo,
while for the TTF-FEL k,o,>>y,). We shall therefore
* Corresponding author. be concerned for the remainder of this work with one-
’ Permanent address: INFN-Milano, Via Celoria 16, 20133 dimensional plasma oscillations in FEL-microbunched
Milan, Italy beams.

0168-9002/97/$17.00 Copyright 0 1997 Published by Elsevier Science B.V. All rights reserved
PII SO168-9002(97)00516-O
J. Rosenzweig et al. /Mud. Instr. and Meth. in Phy. Res. A 393 (1997) 376-374 377

2. Longitudinal plasma oscillations sional theory of the SASE FEL process [2],

The derivation of the longitudinal plasma oscillation Pl 2 1’1

equations in the linear, small amplitude (In, I<<nbO) limit

p, = ?o -
= PFEL~~
is straightforward, and is given elsewhere [l]. In this where b, = 6nb/nb0 is the bunching factor, or fractional
model, we consider the beam to be initially a cold plasma modulation of the beam due to the action of the pon-
_ a continuous, infinitely wide electron beam with aver-
deromotive potential, which has a periodicity very close
age velocity vbOand density nbO. Under this assumption, to the radiation wavelength 2x/k,. Here, also
the thermal effects of the uncorrelated energy spread in pFEL = [uwk,/4k,]2’3 is the FEL [2] or Pierce parameter,
the beam are negligible; the validity of this assumption k, is the wavenumber of the undulator field. and a,,. is its
for FEL beams is discussed below. The effect of the FEL normalized strength. The configuration space bunching
instability on the cold plasma distribution is, in addition is a quarter wavelength out of phase with the velocity
to the density perturbation, a self-consistent small per- space bunching.
turbation in the beam velocity vi s v,, - vbO, where we
again have 1v1I<<v,,,,. nl(O, <) = nbOhl sin(k,;). (8)
In the absence of the radiation field, the equation for
We therefore have
plasma density oscillations is
n i (0.4 ) = n,,k, PFELbl sin(k,[),
..2 (9)
and the plasma oscillations described by Eq. (5) are given,
or -d72 n, + kzn, = 0. in the case of a beam bunched by a SASE FEL, with hi is
& the bunching factor at the end of the undulator.
where the plasma frequency is given by
nl(z, <) = hlnbo sin(k,<) cos(k,z) + i F sin(k,=)
4ne2nb0 4xe2nb0 P ;‘(I 1
WP _- k2=
2 _

or P 3’ (4)
Tim,, ?JO%&
It should be noted that Eq. (4) does not contain deriva- The expression given in Eq. (10) shows that the bunch-
tives with respect to the spatial variable [ = z - v,et, and ing is enhanced for propagation distances less than one-
so a small amplitude disturbance in the beam density is quarter of a plasma wavelength. The maximum in
stationary in the rest frame position - each beam “slice” the beam density occurs within this distance
is an independent oscillator. This is always the case (%X = tan-‘(k,p,,,/k,,~~)), and is given by
for electrostatic plasma oscillations in a cold, uniform
The general solution for a density oscillation described (11)
by Eq. (4) is

ni(O,i) . The squared term in Eq. (11) represents the additional

n,(c. <) = nl(O, i)cos(k,z) + - sm(k,z), (5) bunching due to the initial velocity distribution at the
end of the undulator. In the limit of a beam with neg-
where the prime indicates a derivative with respect to z. ligibly small space-charge, this number is much larger
The two apparently independent constants associated than unity, which corresponds to the tight bunching one
with this solution are in fact not, as they are related by expects from a velocity modulated beam with no repul-
the equation of continuity, which gives sive self-force active. It should also be noted that this
extra bunching term is proportional to the plasma period
=I!!? =-- = ---nb0avl (6) over the gain length, with no explicit dependence on the
vbo dt VbO & ;=O vbO ai r=O'
beam energy. This term is greater than unity for beams
with high gain. For the UCLA IR FEL case [3] this
number is approximately 2.3, and so the beam bunches
3. Longitudinal dynamics in between FEL undulator more tightly by an additional factor of 150%. This occurs
sections very quickly after the undulator, because in this case
kp 1 = 21 cm, with the maximum beam density at
In order to use Eqs. (5) and (6) to analyze the longitudi- z g 3rr/8k, = 25 cm. The oscillation proceeds towards
nal plasma oscillations occurring after an undulator sec- smaller bunching after this point, with the null occurr-
tion, we must have a model for the microbunching which ing near z = 7x/8k, = 30.2cm. After the null, the
occurs in the undulator. We have, from the one-dimen- bunching reverses (to the extent that the assumption of

378 J. Rosenzweig et al. /Nucl. Instr. and Meth. in Phys. Rex A 393 (1997) 376-379

one-dimensional laminar flow holds), at maximizes again we average the dispersion over the chicane section,
after an additional one-quarter plasma period. We plan we can assign an average value to it in Eq. (13)
to investigate these phenomena at the UCLA IR FEL ( - q,/R) 2 af,,/2yi z 8’12, where ach>>l is the equi-
using the coherent transition radiation-based micro- valent undulator parameter, and Q is the maximum bend
bunching diagnostic we have developed, with the angle in the chicane section. Taking this average value of
transition radiator placed at a number of positions the dispersion, we can again integrate the oscillation
downstream of the undulator [4]. The experimental equations to obtain the equivalent of Eq. (lo),
setup for this measurement is discussed in Ref. [S].
For the TTF-FEL in its initial stage [6], we take for
%(z, i) = hub0 Wd) cos(kp,,bz)
the sake of example the energy as 500 MeV, the rms
bunch length beam cZ = 50 urn, and the RMS beam size
rr, = 50 urn. In this case, we have lipi = 300 cm, which is
a relatively short length considering
beam. The additional
the energy of the
bunching factor for these TTF-
I k, db/kX.
1 (14)

FEL parameters is k,/k, pFEL/yi = 4.5 (twice that of the where kp,ch = cop,c&+,O. The maximum enhancement of
UCLA case, as pFEL is four times smaller), and the bunch- the bunching is now approximately larger by a factor of
ing factor increases in a inter-undulator module drift by a,,/& from the case of a pure drift, the same factor by
approximately this factor if the drift length is chosen to which the plasma frequency is raised.
be approximately z = 3n/7k, = 400 cm. One is likely to Because the plasma oscillation wavelength 2n/k,,,, is
have shorter drifts in practice, which produce smaller shortened, in the case of the UCLA IR FEL it may be
density enhancements. One possible way to circumvent difficult to attempt chicane bunching after the initial
this is to use an optical klystron configuration, in which 60 cm undulator section. For the TTF-FEL example,
a short dispersive section allows enhanced pulse com- however, there may be much to be gained by employing
pression in a shorter propagation distance. a chicane section in the so-called optical klystron mode
[7]. Assuming a chicane section 66 cm long, which with
a, = 10 corresponds to one-quarter of a plasma oscilla-
tion length, the bunching enhancement is then raised to
4. Space-charge oscillations in dispersive sections: the approximately 31 at the exit of the chicane.
optical klystron

The use of a dispersive section at the end of an FEL

5. Future work
undulator serves to enhance the phase bunching of the
electron distribution by introducing a change in the rela-
It is interesting to note that the results obtained above
tionship (see Eq. (7)) between the longitudinal velocity
are independent of microbunching amplitude as long as
and the momentum,
i%i<<~bO> i.e. the wave is of small amplitude. For inter-
undulator sections of a multi-undulator array in an FEL
(12) amplifier, this condition is not violated, as saturation
( llzl / z nbO)is only exponentially approached in the final
undulator section. Even so, the results of this paper can
For high energy electrons, the second, dispersive path-
length term on the right side of Eq. (12) is dominant in be extended to include large amplitude oscillations in
cases of interest. A chicane, which is basically one period a straightforward way. It is well known that the density
of a long wavelength undulator, is typically used to give profile of a large amplitude one-dimensional plasma
an average horizontal dispersion which is negative, and wave can be expanded in a power series of Fourier ampli-
so the bunching proceeds in the same direction as pure tudes [S]. The harmonics in the density wave associated
velocity bunching. with the large amplitude dynamics can then be analyzed
The introduction of this additional effect changes the separately, including two-dimensional corrections.
plasma oscillation equation (and expressions derived An integrated theory of the interplay between space-
from it), by a redefinition of the plasma frequency, charge and FEL action, including undulators, drifts,
and dispersive sections, based on the linearized
SASE theory of Bonifacio, Narducci and Pellegrini [2],
is now under development. Preliminary numerical results
verify the conclusions based on the linearized fluid
theory presented above. It is hoped that features of
both the theoretical and computational approaches
This frequency has an explicit dependence on z, which to this problem will be verified soon at the UCLA
makes the oscillation equation not easily integrable. If IR FEL experiment [3,4].
J. Rosenzweig et al. /hhcl. Ins~. and Meth. in Phys. Res. A 393 (1997) 376-379 379

It should also be pointed out that the study of the [3] G. Travish, M. Hogan, C. Pellegrini. J. Rosenzweig, Nucl.
general subject of space-charge oscillations in relativistic Instr. and Meth. A 358 (1995) 75.
beams in both drifts and chicanes occurs outside the [4] J. Rosenzweig, G. Travish, A. Tremaine. Nucl. Instr. and
context of FELs. An example of current interest is the Meth. A 365 (1995) 255.
so-called plasma klystron [9], which has been proposed [S] J. Rosenzweig. A. Tremaine, G. Travish. Presented at 18th
FEL Conf., Rome, Italy, 1996.
as an injector for ultra-short wavelength acceleration
163 J. Rossbach et al., Conceptual design report of the TESLA-
schemes such as the plasma beatwave accelerator.
FEL, in preparation.
[7] J.C. Gallardo. C. Pellegrini, Optical klystron configuration
References for a high gain X-ray free-electron laser, Opt. Commun. 77
(1990) 45.
[l] J. Rosenzweig et al., Nucl. Instr. and Meth. A, submitted. [8] J.B. Rosenzweig, Phys. Rev. A 38 (1989) 3634.
[2] R. Bonifacio. C. Peliegrini, L. Narducci. opt. Commun. 50 [9] T. Katsouleas et al., IEEE Trans. Plasma Sci. 24
(1984) 313. (1996).