Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12

Design considerations for a SASE X-ray FEL$

Claudio Pellegrini*
Department of Physics, University of California at Los Angeles, 405 Hilgard Avenue, Los Angeles, CA 90095, USA

Abstract

The well developed theory of short wavelength SASE-FELs is now being used to design two X-ray lasers, LCLS and
Tesla-FEL. However, the physics and technology of these projects present some unique challenges, related to the very
high peak current of the electron beam, the very long undulator needed to reach saturation, and the importance of
preserving the beam phase-space density even in the presence of large wake-field effects. In the first part of this paper,
we review the basic elements of the theory, the scaling laws for an X-ray SASE-FEL, and the status of the experimental
verification of the theory. We then discuss some of the most important issues for the design of these systems, including
wake-field effects in the undulator, and the choice of undulator type and beam parameters. r 2001 Elsevier Science
B.V. All rights reserved.

PACS: 41.60.Cr; 52.59.Rz; 42.55.Vc

Keywords: Free-electron laser; X-ray laser; Radiation from relativistic electrons

1. Introduction radiation field necessary to start this instability is

Undulator radiation is particularly useful be-
cause of its small line width and high brightness, is
being used in many, if not all, synchrotron
radiation sources built around the world, and
provides at present the brightest source of X-rays.
For long undulators, the free-electron laser (FEL)
collective instability gives the possibility of much
larger X-ray intensity and brightness. The instabil-
ity produces an exponential growth of the radia-
tion intensity, together with a modulation of the
electron density at the radiation wavelength. The
$
Work supported by grant DOE-DE-FG03-92ER409693.
*Corresponding author. Tel.: 310-206-1677; fax: 310-206-
5251.
E-mail address: pellegrini@physics.ucla.edu (C. Pellegrini).
the spontaneous radiation field, or a combination
of the spontaneous radiation field and an external
field. In the first case, we call the FEL a Self
Amplified Spontaneous Emission (SASE) FEL. If
the external field is dominant we speak of an FEL
amplifier.
The existence of an exponentially growing
solution for the FEL has been studied in Refs.
[1–13]. This work led to the first proposals [14–17]
to operate a SASE-FEL at short wavelength,
without using an optical cavity, difficult to build in
the Soft X-ray or X-ray spectral region.
The analysis of a SASE-FEL in the one-
dimensional (1-D) case has led to a simple theory
of the free-electron laser collective instability,
describing all of the free-electron laser physics
with one single quantity, the FEL parameter r [6],

0168-9002/01/$ - see front matter r 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 9 0 0 2 ( 0 1 ) 0 1 5 2 7 - 3
2 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12
a function of the electron beam density and
energy, and of the undulator period and magnetic
field. The extension of the FEL theory to three
dimension, including diffraction effects [18–22] has
been another important step toward a full under-
standing of the physics of this system. From the
collective instability theory, we can obtain a
scaling law [23] of a SASE-FEL with wavelength,
showing a weak dependence of the gain on
wavelength, and pointing to the possibility of
using this system to reach the X-ray spectral
region. This analysis shows that to reach short
wavelengths one needs a large electron beam six-
dimensional phase-space density, a condition until
recently difficult to satisfy.
The development of radio frequency photo-
cathode electron guns [24], and the emittance
compensation method [25,26] has changed this
situation. At the same time, the work on linear
colliders has opened the possibility to accelerate
and time compress electron beams without spoil-
ing their brightness [27–30]. At a Workshop on IV
Generation Light Sources held at SSRL in 1992 it
was shown [31] that these new developments make
possible to build an X-ray SASE-FEL. This work
led to further studies [32,33], and to two major
proposals, one at SLAC [34], the other at DESY
[35], for X-ray SASE-FELs in the 1 A ( region, with
peak power of the order of tens of GW, pulse
length of about 100 fs or shorter, full transverse
coherence, peak brightness about ten orders of
magnitude larger than that of III generation
synchrotron radiation sources.
While the theory of the SASE-FEL has been
developed starting from the 1980s, a comparison
with experimental data has become possible only
during the last few years, initially in the infrared
to visible region of the spectrum [36–46]. These
experimental data agree with our theoretical
model, in particular the predicted exponential
growth, the dependence of the gain length on
electron beam parameters, and the intensity
fluctuations. A recent experiment at DESY [47]
has demonstrated gain of about 1000 at the
shortest wavelength ever reached by an FEL,
80 nm. These results give us confidence that we
can use the present theory to design an X-ray
SASE-FEL.
2. Free-electron laser physics
The physical process on which a FEL is based is
the emission of radiation from one relativistic
electron propagating through an undulator. We
consider for simplicity only the case of a helical
undulator, and refer the reader to other books or
papers, for instance Refs. [48,49], for a more
complete discussion.
Let us consider the emission of coherent
radiation from Ne electrons, that is the radiation
at the wavelength
lu
l¼ ð1 þ a2u Þ ð1Þ
2g2
within the coherent solid angle
2pl
py2c ¼ ð2Þ
lu N u
and line width
Do 1
¼ : ð3Þ
o Nu
In Eq. (1) Ebeam ¼ gmc2 is the beam energy,
and au ¼ eBu lu =2pmc2 is the undulator para-
meter.
When there is no correlation between the field
generated by each electron, as in the case of
spontaneous radiation, the total number of coher-
ent photons emitted is Nph ¼ paNe a2u =ð1 þ a2u Þ;
where a is the fine structure constant. Hence, the
number of coherent photons is about 1% of the
number of electrons. If all electrons were within a
radiation wavelength, the number of photons
would increase by a factor Ne. Even when this is
not the case, and the electron distribution on the
scale of l is initially random, the number of
photons per electron can be increased by the FEL
collective in-stability [6].
The instability produces an exponential growth
of the field intensity and of the bunching
parameter
1 X Ne
B¼ expð2pizk =lÞ
Ne k¼1
where zk is the longitudinal position of electron
k. The growth saturates when the bunching
parameter is of the order of one. For a
 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12
long undulator the intensity is approximately
given by
I0
IE expðz=LG Þ ð4Þ
9
where LG is the exponential growth rate, called
the gain length, and I0 is the spontaneous co-
herent undulator radiation intensity for an un-
dulator with a length LG, and is proportional to
the square of the initial value of the bunching
factor, |B0|2.
The instability growth rate, or gain length, is
given, in a simple 1-D model by
lu
LG E pffiffiffi ð5Þ
4 3pr
where r is the free-electron laser parameter [6],
 
au Op 2=3
r¼ ð6Þ
4g ou
ou ¼ 2pc=lu is the frequency associated to the
undulator periodicity, and Op ¼ ð4pe c2 ne =gÞ1=2 ; is
the beam plasma frequency, ne is the electron
density, and re is the classical electron radius.
A similar exponential growth, with a differ-
ent coefficient, occurs if there is an initial input
field, and no noise in the beam (uniform beam,
B0=0), i.e. amplified stimulated emission. In the
SASE case, saturation occurs after about 20 gain
lengths, and the radiated energy at saturation is
about rNe Ebeam : The number of photons per
electron at saturation is then Nsat ¼ rEbeam =Eph :
In a case of interest to us, an X-ray FEL with
Eph E104 eV; EE15 GeV; rE103 ; we obtain
Nph E103 ; i.e. an increase of almost 5 orders of
magnitude in the number of photons produced per
electron.
The instability can develop only if the undulator
length is larger than the gain length, and some
other conditions are satisfied:
(a) Beam emittance smaller than the wavelength:
l order of the gain length. We use the emittance
eo : ð7Þ
4p
(b) Beam energy spread smaller than the free-
electron laser parameter:
sE or: ð8Þ
3
(c) Gain length shorter than the radiation Raleigh
range:
LG oLR ð9Þ
where the Raleigh range is defined in terms of the
radiation beam radius, o0 by po20 ¼ lLR :
Condition (a) says that for the instability to
occur, the electron beam must match the trans-
verse phase-space characteristics of the radiation.
Condition (b) limits the beam energy spread.
Condition (c) requires that more radiation is
produced by the beam than what is lost through
diffraction.
Conditions (a) and (c) depend on the beam
radius and the radiation wavelength, and are
not independent. If they are satisfied, we can
use with good approximation the 1-D model.
If they are not satisfied and the gain length
deviates from the one-dimensional value
(5)Fas in the LCLS case where the emittance
is about 3 times larger than l=4pFit is con-
venient to introduce an effective FEL parameter,
defined as
lu
reff ¼ pffiffiffi ð10Þ
4 3pLG3D
where LG3D is the three-dimensional gain length
obtained from numerical simulations, including
the effects of diffraction, energy spread, and
emittance. This quantity is a measure of the
three-dimensional effects present in the FEL, and
can be used to obtain more realistic information
on the system.
2.1. Scaling laws
Analyzing Eqs. (6)–(8), one obtains the
scaling law for a SASE-FEL at a given wave-
length. We assume that the beam is focused by
the undulator and an additional focusing struc-
ture, to provide a focusing function bF of the
e ¼ s2T =bF and the longitudinal brightness
BL ¼ ecNe =2psL gsg ; where sL and gsg are the
bunch length and the bunch absolute energy
spread, to describe the bunch density and
energy spread. The result is that the FEL r
parameter scales like the beam longitudinal
4 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12
brightness [23]
pffiffiffiffiffiffi a a2 BL
u
rB 2p
b 1 þ a2u IA
where we have assumed sE ¼ ar; e ¼ bl=4p; and
IA ¼ ec=re is the Alfven current. In the LCLS case,
we have aC1=4; bC4; K=3.7 requiring
BL > 100 A for r to be of the order of 0.001. Since
this discussion does not consider explicitly gain
losses due to diffraction, undulator errors and
beam misalignment, we need in practice a larger
value of BL :
To obtain an emittance which satisfies condition
(7) at about 0.1 nm, using a photo-cathode gun
and a linac, we need a large beam energy, of the
order of several GeVs, to reduce the emittance by
adiabatic damping. The beam longitudinal bright-
ness is determined by the electron source. Wake-
fields in the linac can, however, reduce it con-
siderably. For LCLS the photo-cathode gun gives
a slice emittance e ¼ 6  108 m rad; slice long-
itudinal brightness BL=8000 A at 10 MeV. Accel-
eration and compression in the SLAC linac then
gives eC41011 m rad, BL C1500 A at 15 GeV,
good enough to produce lasing, even considering
the gain losses due to diffraction, imperfections
and misalignment.
2.2. Slippage, fluctuations and time structure
When propagating in vacuum, the radiation
field is faster than the electron beam, and it moves
forward, ‘‘slips’’, by one wavelength l for each
undulator period. The slippage in one gain length
defines the ‘‘cooperation length’’ [55],
l
Lc ¼ LG :
lu
For the SASE case the radiation field is
proportional to IðoÞ; the Fourier component at
o ¼ 2pc=l of the initial bunching factor B0, and
the intensity to jIðoÞj2 If the bunch length, LB is
such that LB bl; and the beam is generated from a
thermionic cathode or photo-cathode, the initial
bunching and its Fourier component IðoÞ are
random quantities. The initial value of B0 is
different for each beam section of length l; and
has a random distribution. The average values are
/IðoÞSB/B0 S ¼ 0; and /jIðoÞj2 SB/jB0 j2 >S
BNe :
ð11Þ As the beam and the radiation propagate
through the undulator, the FEL interaction
introduces a correlation on the scale length of Lc,
producing spikes in the radiation pulse, with a
length of the order of Lc, and a random intensity
distribution. The number of spikes is [55,56] M ¼
LB =4pLc : The total intensity distribution is a
Gamma distribution function
I M1
PðIÞ ¼ M M expðMI=/ISÞ ð13Þ
/ISM GðMÞ
where /IS is the average intensity.pThe
ffiffiffiffiffi standard
deviation of this distribution is 1= M : The line
width is approximately the same as for the
spontaneous radiation, Do=oE1=Nu :
3. Experimental results on SASE-FELs
A SASE-FEL is characterized by LG and the
intensity fluctuations, the distribution of |B0|2.
Very large gain in the SASE mode has so far been
observed in the centimeter [36–38] to millimeter
wavelength. Gain between about 1000% and
100% has been observed at Orsay [39] and UCLA
[40] in the infrared, and at Brookhaven [43] in the
visible. Larger gain in the infrared has also been
observed at Los Alamos [41], and gain as large as
3  105 at 12 mm has been measured by a UCLA-
LANL-RRIK collaboration [42]. The intensity
distribution function has been previously mea-
sured for spontaneous undulator radiation [57],
with no amplification, and long bunches, and more
recently for amplified radiation, and a short bunch
ð12Þ length [40,42].
A BNL group [44], has demonstrated high gain
harmonic generation seeding the FEL with a
10.6 mm external laser and producing a 5.3 mm
FEL output, with intensity 2  107 larger than
spontaneous radiation. The VISA group is com-
missioning a 0.8–0.6 mm experiment, using a 4 m
long undulator with distributed strong focusing
quadrupoles. Initial results have shown a gain
of about 100 [45]. An experiment at Argonne uses
the APS injector, with an energy of 220–
444 MeV, wavelength 500–20 nm, and an 18 m
 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12 5

long undulator. SASE amplification as a function

of undulator length has been demonstrated re-
cently at 530 nm [46]. A DESY group is using the
TESLA Test Facility superconducting linac. In
Phase 1 the electron beam energy can reach up to
390 MeV, and a wavelength of 42 nm. Phase 2 will
reach 1000 MeV, and 6 nm, with an undulator
length of 30 m. Initial results have shown a gain of
about 103 over a wavelength range from 180 to
80 nm, the shortest wavelength ever obtained in an
FEL [47]. Similar experimental programs on
SASE-FELs are being prepared also in Japan, at
Spring 8 and other laboratories, and in China.
Fig. 2. Intensity distribution over many events for the same
The main results of the UCLA-LANL-RRIK-
experiment. The experimental data are fitted with a Gamma
SSRL experiment are shown in Figs. 1 and 2. function distribution [42].
Fig. 1 shows an increase in output intensity by
more than 104, when changing the electron charge
by a factor of seven. The bunch radius, energy
spread, and length change with the charge, making described by a Gamma function with the M
impossible to have a simple analytical model to parameter evaluated from the experimental data,
evaluate the intensity. The experimental data and and is in agreement with the theory.
the theory have been compared using the simula-
tion code Ginger [58], and the measured values of
all bunch parameters. The results are plotted in 4. LCLS: an X-ray SASE-FEL
Fig. 1, and, within experimental errors, agree with
the data. The intensity measured at a charge of The first proposal for an X-ray SASE-FEL was
2.2 nC corresponds to a gain of 3  105, the largest made in 1992 [31], it was then developed by a study
measured until now in the infrared. The measured group until 1996 and by a design group that has
intensity fluctuations, shown in Fig. 2 are well prepared the LCLS design report [34]. The LCLS
parameters are given in Table 1. The average
brightness, /BS, and peak brightness Bp, are
measured in photons/s/mm2/mrad2/0.1% band-
100
width.
The LCLS experimental setup is shown in
10
Fig. 3. The electron beam is produced in a
photo-cathode gun, developed by a BNL-SLAC-
Energy [nJ]

1 UCLA collaboration [50], and producing a bunch

with a charge of 1 nC/bunch, normalized emit-
0.1 tance, rms, 1 mm mrad; pulse length, rms, 3.3 ps
[51]. The beam is then accelerated to 14.3 GeV and
0.01 compressed to a peak current of 3400 A in the
SLAC linac. During acceleration and compression
0.001 the transverse and longitudinal phase-space den-
0 50 100 150 200 sities are increased by space charge, longitudinal
I [Amps] and transverse wake-fields, RF-curvature, coher-
Fig. 1. Measured values of the mean FEL intensity vs beam ent synchrotron radiation effects. The acceleration
current, compared with a Ginger simulation for the UCLA- and compression system has been designed to
LANL-RRIK-SSRL 12 mm SASE-FEL [42]. minimize all these effects simultaneously, and it
6 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12

limits the transverse emittance dilution to about long, separated by 23.5 cm straight sections [53].
10% or less. Since the natural undulator focusing is weak at
The planar hybrid LCLS undulator has vana- the LCLS energy, additional focusing is provided
dium permendur poles, Nd–FeB magnets, and by permanent magnet quadrupoles located in the
K=3.7 [52]. It is built in sections about 3 m straight sections. Optimum gain is obtained
for a horizontal and vertical beta function of
18 m, giving a transverse beam radius of 30 mm,
Table 1 radiation Raleigh range of 20 m, twice the
LCLS electron beam, undulator, and FEL parameters
field gain length, making diffraction effects small.
LCLS Electron beam parameters The FEL gain is sensitive to errors in the
Electron energy, GeV 14.3 undulator magnetic field, and to deviation in the
Peak current, kA 3.4 beam trajectory. Simulations of these effects,
Normalized emittance, mm mrad 1.5
Energy spread, %, at undulator entrance 0.006
including beam position monitors and steering
Bunch length, fs 67 magnets along the undulator to correct the
trajectory, show that the field error tolerance is
LCLS undulator parameters 0.1%, and the beam trajectory error tolerance is
Undulator period, cm 3 about 2 mm [54].
Undulator length, m 100
Undulator field, T 1.32 LCLS generates coherent radiation at lC
Undulator K 3.7 1.5 nm and its harmonics [59]. It also generates
Undulator gap, mm 6 incoherent radiation, which, at 14.3 GeV, has a
spectrum extending to about 500 keV, and a peak
LCLS FEL parameters power density on axis of 1013 W/cm2. The power
Radiation wavelength, nm 0.15
FEL parameter, r 5  104 density of the coherent first harmonic is about
Field gain length, m 11.7 2  1014 W/cm2, and the peak electric field is
Effective FEL parameter, reff 2.3  104 about 4  1010 V/m. Filtering and focusing the
Pulses/s, 120 radiation and transporting it to the experi-
Peak coherent power, GW 9
mental areas is a challenge. A normal incidence
Peak brightness 1033
Average brightness 4  1022
mirror at 100 m would see an energy flux of
Cooperation length, nm 51 about 1 J/cm2, about 1 eV/atom, large enough to
Intensity fluctuation, % 8 damage exposed materials. The LCLS large
Linewidth 2  104 power density will push the optical elements and
Total synchrotron radiation energy loss, GW 90 instrumentation into a new strong field regime,
Energy spread due to synchrotron radiation 2  104
emission but offers also new opportunities for scientific
research.

Fig. 3. LCLS experimental setup.

 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12
5. Effects of undulator wake-fields and spontaneous
radiation
The emission of spontaneous radiation by the
electrons in the undulator has two main effects, a
decrease of the electron energy, WeR, and an
increase of energy spread, sl;R : Both effects can
reduce the gain if the conditions WeR =E5reff ;
sg;R 5reff ; are not satisfied. The two quantities
WeR ; sg;R ; have been evaluated in Ref. [60]. For
the LCLS case we have We =Ebeam C1.8  103>
reff ; sE;R C1:5104 Creff ; and both effects have
to be considered, even if the effect on the gain is
not large. The average energy loss can be
compensated by tapering the wiggler. The energy
spread could be reduced using a shorter undulator.
For an X-ray FEL, with a large peak current,
and a long undulator, the wake-fields in the
undulator vacuum pipe can have an important
effect on the lasing process, and can reduce the
output power and change the temporal structure
of the X-ray pulse. To evaluate these effects, we
use a model which considers the effects of the
vacuum pipe resistivity and roughness. The
resistive longitudinal wake-field is [61]
 pffiffiffi
4ce2 Z0 1 t=t
Wz ðtÞ ¼  e cosð 3t=tÞ
pR2 3
pffiffiffi Z #
2 N x2 x2 t=t
 e dx ð14Þ
p 0 x6 þ 8
where t measures the longitudinal position of the
test particle respect to the particle generating the
field, Z0 is the vacuum impedance, t ¼
ð2R2 =Z0 sÞ1=3 =c; s is the conductivity of the
material, and R the pipe radius.
The effect of the pipe roughness has been
evaluated by several authors. The first models of
roughness impedance, based on a random dis-
tribution of surface bumps, were developed by
Bane, Ng and Chao [62–64], and confirmed by
Stupakov [66]. They give an inductive impedance
proportional to 1/R, depending on the ratio of
bumps height to length. If this ratio is about one,
and the field wavelength is larger than the bump
height and width, then the rough surface can
support the propagation of a wave synchronous
7
with the beam. In this case the wake-field is rather
strong, and the tolerance for LCLS is a bump
height of about 40 nm. For a bump length much
larger than the height, a different model, due to
Stupakov, applies and the effect is much weaker.
Electron microscope observations of the surface of
a metal similar to that a vacuum pipe, reported in
this paper [65], support this case.
In another model [67,68] the roughness is
considered equivalent to a thin dielectric layer on
the surface of the pipe, and the pipe can support a
wave synchronous with the beam, giving a wake-
field
ce2 Z0
Wz ðtÞ ¼  cosðk0 tÞ ð15Þ
pR2
where d is the thicknesspof the layer, ffi Z0 the
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vacuum impedance, k0 ¼ 2e=ðRdðe  1ÞÞ; and it
is assumed eB2:
To have no gain reduction from the wake-fields,
we must satisfy the condition that the variation in
energy that they induce be small compared to the
gain bandwidth, ðDE=EÞwake oreff : In the LCLS
case this gives the condition Wzo30 KV/m.
6. Options for the choice of undulator and beam
characteristics
The LCLS design shows the feasibility of an
X-ray FEL. It is, however, possible to optimize
the system by reducing the undulator saturation
length; reducing in the ratio of total spontaneous
synchrotron radiation to amplified coherent radia-
tion; choosing electron beam parameters to reduce
wake-field effects; controlling the X-ray pulse
output power, pulse length, line-width.
The undulator saturation length is controlled by
the FEL parameter r (6) (5), by the ratio e=l (7),
and by the electron energy spread (8). A reduction
of the beam charge and emittance, keeping their
ratio constant, leaves the 1-D gain length un-
changed, and reduces the ratio e=l: For systems,
such as LCLS, where this ratio is larger than 1, this
reduces the 3-D gain length. Reducing the charge
can also reduce the FEL intensity, giving a way to
control the output power [69], and reduces wake-
field effects in the linac [70] and undulator.
8 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12

Table 2
Parameters for helical undulator and low charge cases

LCLS A B C D

Beam energy, GeV 14.3 14.7 14.7 12 14.7

Bunch charge, nC 1.0 1.0 1.0 1.0 0.2
Normalized emittance, mm mrad 1.1 1.1 1.1 1.1 0.3
Peak current, kA 3.4 3.4 3.4 3.4 1.17
Energy spread, rms, % 0.006 0.006 0.006 0.008 0.006
Undulator type Planar Helical Helical Helical Helical
Undulator period, cm 3 3 3 4 3
Undulator parameter, K 3.7 2.7 2.7 1.8 2.7
Undulator gap, mm 6 8.5 8.5 7.5 8.5
Focusing beta function in undulator 18 17.7 73 20.5 5
Total synchrotron radiation loss, GW 90 50 50 11.6 10
Gain length, m 4.2 2.8 3.4 4.2 1.8

400

surface roughness

200
current [a.u.]
∆Ε/ ∆z [keV/m]

resistive wall
-200

total wake
-400

-600
-40 -20 0 20 40
ct [µm]

Fig. 4. Resistive and roughness wake-field along the electron bunch for LCLS [71].

An optimization of a SASE-FEL has been done
in Ref. [71], where the five cases shown in Table 2
are discussed. One case is the LCLS. Cases A, B,
D, use permanent magnet helical undulators with
large gap and large field. For A, we use additional
FODO focusing, while B uses only the natural
undulator focusing; case D uses a lower beam
charge, emittance and peak current. The beam
parameter for this case have been obtained using
the photo-cathode gun scaling laws [72], and
discussed later in this sections. Case C uses a
lower field helical undulator. The FEL power
growth along the undulator for the 5 case has been
evaluated using the numerical simulation code
Genesis, and including the effect of synchrotron
radiation emission, and of the resistive (14) and
roughness wake-fields (15) in the undulator
vacuum pipe. The total wake-field for the LCLS
case is shown in Fig. 4. The wake-field violates the
condition Wzo30 KV/m by almost a factor of ten.
Even if we consider only the resistive wake-field
this condition is violated in part of the bunch.
 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12 9

1012 1012
LCLS high field
1010 1010

108 108

P [W]

P [W]
6
10 106
4
10 104

102 102
0 20 40 60 80 100 0 20 40 60 80 100
z [m] z [m]

1012 1012
high field + nat. foc. low field
1010 1010

108 108

P [W]
P [W]

106 106

104 104

102 102
20 40 60 80 100 0 20 40 60 80 100
z [m] z [m]

Fig. 5. SASE-FEL power, in W, versus undulator length, in m, for LCLS and cases A, B, D. The upper curve describes the ideal case
of no wake-fields, the intermediate includes the resistive wall effect, and the lower curve includes resistivity and roughness [71].

40

30
P [GW]

20

10

0
-40 -20 0 20 40
t [fs]

Fig. 6. Power distribution along the bunch length for the LCLS and case A, when wake-fields are included in Genesis [71].

The results from Genesis show that the that have a small energy loss in traversing the
undulator wake-fields produce an order of undulator show gain, and this electrons are in a
magnitude reduction in output power for the position in the bunch were the wake-field is near
LCLS case, Fig. 5, and a smaller reduction zero.
in cases A, B, D. The reason for this reduc- The power loss is less in case B, because of
tion can be seen clearly in Fig. 6: only the electrons the larger undulator gap and of the smaller
10 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12
12
10
10
P [W] 10
8
10
6
10
4
10
0 10
Fig. 7. Power as a function of undulator length for case C; full line, no wake-fields; dotted line, including wake-fields [71].
undulator length. Wake-field effects are negligible
in case C, Fig. 7, with a small peak current
and undulator length. Due to the larger value
of r; we have in case C the same output power
as in the standard LCLS case, about 10 GWatt,
while the spontaneous synchrotron radiat-
ion power is reduced from 90 to about
10 Gwatt.
Two methods can be considered to reduce
the charge and the emittance. One controls the
laser intensity, spot size and phase on photo-
cathode gun to minimize the emittance
as a function of charge [72,73]. The scaling
laws, neglecting the effect of thermal emittance,
are
eN ¼ 1:45ð0:38Q4=3 þ 0:095Q8=3 Þ1=2 ;
mm mrad; Q in nC;
sL ¼ 6:3104 Q1=3 ; m; Q in nC:
Another approach [74] is to produce
a 1 nC bunch and then reduce the emittance
and charge by collimation. As an example,
with a collimator to beam rms radius ratio
of 1.5, one can reduce the charge by a factor of
2.5 and the emittance by a factor of 5. The effect
20 30 40
z [m]
of the collimator wake-field on the emittance
has also been studied and found to be small.
7. Conclusions
The possibility of large amplification of the
spontaneous undulator radiation has been demon-
strated in the recent SASE-FELs experiments in
the infrared, visible, and UV spectral regions. The
experimental results on the gain length and the
intensity fluctuation distribution are in good
agreement with the FEL collective instability
theory. Gain as large as 3  105 have been
observed in the infrared, bringing us near the
saturation level, and large gain has been measured
at a wavelength of 80 nm. Experiments over a
ð16Þ
range of wavelengths will continue, to study
saturation, and the spectral, temporal, and angular
ð17Þ properties of the SASE radiation, and completely
characterize the FEL. These results, and the
continued progress in the production, acceleration,
measurements, and wake-field control of high
brightness electron beams, together with the
construction of high quality planar and helical
undulators, will lead to a successfully operation of
X-ray SASE-FEL in the next few years.
 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12
Acknowledgements
I wish to thank all the members of the LCLS
group, and I. Ben Zvi, M. Cornacchia, D. Nguyen,
J. Rosenzweig, R. Sheffield, A. Varfolomeev, H.
Winick, the UCLA students and Postdocs, who
have made this work possible.
References
[1] N.M. Kroll, Phys. Quantum Electron. 5 (1978) 115.
[2] P. Sprangle, R.A. Smith, NRL Memo-Report 4033,
1979.
[3] D.B. McDermott, T.C. Marshall, Phys. Quantum Elec-
tron. 7 (1980) 509.
[4] A. Gover, P. Sprangle, IEEE J. Quantum Electron. QE-17
(1981) 1196.
[5] G. Dattoli, et al., IEEE J. Quantum Electron. QE-17
(1981) 1371.
[6] R. Bonifacio, C. Pellegrini, L. Narducci, Opt. Comm. 50
(1984) 373.
[7] J.B. Murphy, C. Pellegrini, R. Bonifacio, Opt. Comm. 53
(1985) 197.
[8] J. Gea-Banacloche, G.T. Moore, M. Scully, SPIE 453
(1984) 393.
[9] P. Sprangle, C.M. Tang, C.W. Roberson, Nucl. Instr. and
Meth. A 239 (1985) 1.
[10] E. Jerby, A. Gover, IEEE J. Quantum Electron. QE-21
(1985) 1041.
[11] K.-J. Kim, Nucl. Instr. and Meth. A 250 (1986) 396.
[12] J.-M. Wang, L.-H. Yu, Nucl. Instr. and Meth. A 250
(1986) 484.
[13] R. Bonifacio, F. Casagrande, C. Pellegrini, Opt. Comm. 61
(1987) 55.
[14] J.B. Murphy, C. Pellegrini, Nucl. Instr. and Meth. A 237
(1985) 159.
[15] J.B. Murphy, C. Pellegrini, J. Opt. Soc. Am. B 2 (1985)
259.
[16] Ya.S. Derbenev, A.M. Kondratenko, E.L. Saldin, Nucl.
Instr. and Meth. A 193 (1982) 415.
[17] K.J. Kim, et al., Nucl. Instr. and Meth. A 239 (1985) 54.
[18] G.T. Moore, Nucl. Instr. and Meth. A 239 (1985) 19.
[19] E.T. Scharlemann, A.M. Sessler, J.S. Wurtele, Phys. Rev.
Lett. 54 (1985) 1925.
[20] M. Xie, D.A.G. Deacon, Nucl. Instr. and Meth. A 250
(1986) 426.
[21] K.-J. Kim, Phys. Rev. Lett. 57 (1986) 1871.
[22] L.-H. Yu, S. Krinsky, R. Gluckstern, Phys. Rev. Lett. 64
(1990) 3011.
[23] C. Pellegrini, Nucl. Instr. and Meth. A 272 (1988) 364.
[24] J.S. Fraser, R.L. Sheffield, E.R. Gray, Nucl. Instr. and
Meth. 250 (1986) 71.
[25] B.E. Carlsten, Nucl. Instr. and Meth. Phys. Res. A 285
(1989) 313.
11
[26] X. Qiu, K. Batchelor, I. Ben-zvi, X-J. Wang, Phys. Rev.
Lett. 76 (1996) 3723.
[27] K. Bane, Wakefield effects in a linear collider, AIP
Conference Proceedings, Vol. 153, 1987, p. 971.
[28] J. Seeman, et al., Summary of emittance control in the SLC
linac, US Particle Accelerator Conference, IEEE Con-
ference Proceedings 91CH3038-7, 1991, p. 2064.
[29] J. Seeman, et al., Multibunch energy and spectrum control
in the SLC high energy linac, US Particle Accelerator
Conference, IEEE Conference Proceedings 91CH3038-7,
1991, p. 3210.
[30] T. Raubenheimer, The generation and acceleration of low
emittance flat beams for future linear colliders, SLAC-
Report 387, 1991.
[31] C. Pellegrini, A 4 to 0.1 nm FEL based on the SLAC
Linac, in: M. Cornacchia, H. Winick (Eds.), Proceedings
of the Workshop on IV Generation Light Sources, SSRL-
SLAC report 92/02, (1992), p. 364.
[32] C. Pellegrini, et al., Nucl. Instr. and Meth. A 341 (1994)
326.
[33] G. Travish, et al., Nucl. Instr. and Meth. A 358 (1995) 60.
[34] LCLS Design Study Report, Report SLAC-R-521, 1998.
[35] J. Rossbach, et al., Nucl. Instr. and Meth. A 375 (1996)
269.
[36] T. Orzechowski, et al., Phys. Rev. Lett. 54 (1985) 889.
[37] D. Kinfrared, K. Patrick, Nucl. Instr. and Meth. A 285
(1989) 43.
[38] J. Gardelle, J. Labrouch, J.L. Rullier, Phys. Rev. Lett. 76
(1996) 4532.
[39] R. Prazeres, et al., Phys. Rev. Lett. 78 (1997) 2124.
[40] M. Hogan, et al., Phys. Rev. Lett. 80 (1998) 289.
[41] D.C. Nguyen, et al., Phys. Rev. Lett. 81 (1998) 810.
[42] M. Hogan, et al., Phys. Rev. Lett. 81 (1998) 4867.
[43] M. Babzien, et al., Phys. Rev. E 57 (1998) 6093.
[44] L.-H. Yu, et al., Nucl. Instr. and Meth. A 445 (1999) 301,
and Proceedings of this Conference, p. II-41.
[45] A. Tremaine, et al., Initial gain measurements of a 800 nm
SASE-FEL, VISA, Proceedings of this Conference,
p. II-35.
[46] S.V. Milton, et al., Phys. Rev. Lett. 85 (2000) 988, Nucl.
Intr. and Meth. A 475 (2001) 28, these proceedings.
[47] J. Rossbach, et al., Nucl. Instr. and Meth. A 475 (2001) 13,
these proceedings.
[48] J.B. Murphy, C. Pellegrini, Introduction to the physics of
free-electron lasers, in: W.B. Colson, C. Pellegrini, A.
Renieri (Eds.), Laser Handbook, Vol. 6, North Holland,
Amsterdam, 1990.
[49] R. Bonifacio, et al., La Rivista del Nuovo Cimento 13
(1990) 9.
[50] D.T. Palmer, et al., Emittance studies of the BNL-SLAC-
UCLA 1.6 Cell photocathode RF Gun, Proceedings of
IEEE 1997 Particle Accelerator Conference, 1998, pp.
2687–2690.
[51] R. Alley, et al., Nucl. Instr. and Meth. A 429 (1999)
324.M. Ferrario, et al., New design study and related
experimental program for the LCLS RF photoinjector,
LCLS-TN-00-9, SLAC, 2000.
12 C. Pellegrini / Nuclear Instruments and Methods in Physics Research A 475 (2001) 1–12
[52] R. Carr, Design of an X-ray free electron laser undulator,
Proceedings of US Synchrotron Radiation Instrumenta-
tion Conference, Ithaca, NY 1997, Rev. Sci. Instr.
November 1997, SLAC pub 7651.
[53] E. Gluskin, et al., Nucl. Instr. and Meth. A 475 (2001) 323,
these proceedings.
[54] P. Emma, Electron trajectory in an undulator with dipole
field and BPM errors, LCLS-TN-99-4 SLAC, 1999.
[55] R. Bonifacio, et al., Phys. Rev. Lett. 73 (1994) 70.
[56] E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov, Statis-
tical properties of radiation from VUV and X-ray free
electron laser, DESY rep. TESLA-free-electron laser
97-02, 1997.
[57] M.C. Teich, T. Tanabe, T.C. Marshall, J. Galayda, Phys.
Rev. Lett. 65 (1990) 3393.
[58] W. Fawley, private communication.
[59] K.-J. Kim, private communication, and proceedings of this
Conference, p. II-15.
[60] E.L. Saldin, et al., Nucl. Instr. and Meth. A 381 (1996)
545.
[61] K. Bane, SLAC report AP-87, 1991.
[62] C.-K. Ng, Phys. Rev. D 42 (1990) 1819.
[63] K. Bane, C.-K. Ng, A. Chao, Estimate of the impedance
due to wall surface roughness, SLAC-PUB-7514, 1997.
[64] K.L. Bane, A.V. Novokhatskii, SLAC, Technical Report
SLAC-AP-117, 1999.
[65] G.L. Stupakov, et al., Phys. Rev. ST 2 (1999) 60701/1.
[66] G. Stupakov, Surface impedance and Synchronous modes,
SLAC, SLAC-PUB-8208, 1999.
[67] A.V. Burov, A.V. Novokhatskii, Budker Institute of
Nuclear Physics, Technical Report BudkerINP-90-28, 1990.
[68] A.V. Novokhatskii, A. Mosnier, Proceedings of the
1997 Particle Accelerator Conference, IEEE, New York,
1997, pp. 1661–1663.
[69] C. Pellegrini, X. Ding, J. Rosenzweig, Output Power
Control in an X-ray FEL, Proceedings of the IEEE
Particle Accelerator Conference, New York, 1999, p. 2504.
[70] P. Emma, LCLS Accelerator Parameters and Tolerances
for Low Change Operations, LCLS-TN-99-3, SLAC, 1999.
[71] C. Pellegrini, et al., Nucl. Instr. and Meth. A 475 (2001)
328, these proceedings.
[72] J.B. Rosenzweig, E. Colby, in AIP Conference Proceed-
ings, Vol. 335, 1995, p. 724.
[73] J.B. Rosenzweig, et al., Optimal scaled photoinjector
designs for FEL applications, Proceedings of IEEE
Particle Accelerator Conference, 1999, p. 2045.
[74] C. Schroeder, C. Pellegrini, H.-D. Nuhn, Proceedings of
this Conference.