NUCLEAR

INSTRUMENTS
Nuclear Instruments and Methods in Physics Research A 331 (1993) 223-227 & METHODS
North-Holland IN PHYSICS
RESEARCH
Section A

A 2 to 4 n m high p o w e r F E L o n the S L A C linac *

C. Pellegrini and J. Rosenzweig
UCLA Department of Physics, Los Angeles, California 90024, USA

H.-D. Nuhn, P. Pianetta, R. Tatchyn and H. Winick

Stanford Synchrotron Radiation Laboratory, Stanford, California, USA

K. Bane, P. Morton, T. Raubenheimer and J. Seeman

Stanford Linear Accelerator Center, Stanford, California, USA

K. Halbach and K.-J. Kim

Lawrence Berkeley Laboratory, Berkeley, California, USA

J. Kirz
State University of New York at Stony Brook, Stony Brook, New York, USA

We report the results of preliminary studies of a 2 to 4 nm SASE FEL, using a photoinjector to produce the electron beam, and
the SLAC linac to accelerate it to an energy up to 10 GeV. Longitudinal bunch compression is used to increase ten fold the peak
current to 2.5 kA, while reducing the bunch length to the subpicosecond range. The saturated output power is in the multi-gigawatt
range, producing about 1014 coherent photons within a bandwidth of about 0.2% rms, in a pulse of several millijoules. At 120 Hz
repetition rate the average power is about 1 W. The system is optimized for X-ray microscopy in the water window around 2 to 4
urn, and will permit imaging a biological sample in a single subpicosecond pulse.

1. Introduction until recently difficult to satisfy. The recent develop-

m e n t of high-brightness photocathode .electron guns
The self-amplified spontaneous emission m o d e of [3], and the expected availability of the S L A C linac,
an F E L has b e e n proposed [1] and analyzed [2] as a open the possibility to make this major extension of
source of tunable, coherent, high peak power soft F E L operation, f r o m the shortest wavelength yet
X-rays, capable of producing a radiation b e a m with a achieved (240 nm) to 2 - 4 nm, the wavelength range
brightness eight orders of magnitude larger than syn- suitable for biological imaging and other applications.
chrotron light sources, and a pulse duration shorter The recognition of this possibility was one of the
than 1 ps. Such a large increase in radiation flux, and main conclusions of the Workshop on Fourth Genera-
the short pulse duration, will open new exciting and tion Light Sources held at S S R L / S L A C on February
unique research possibilities in physics, biology and 24-27, 1992. The workshop report [4] contains many
other sciences. contributions relevant to linac-driven short wavelength
The S A S E approach produces lasing in a single pass FELs. In particular, the use of the S L A C linac for this
of a high peak current electron b e a m through a long purpose is discussed by Pellegrini [5], Kim [6], and by
undulator, eliminating the n e e d for optical cavities, Barletta, Sessler and Y u [7]. The advantage of using
difficult to build at these short wavelengths. The re- the S L A C linac is that its properties are extensively
quirements on the electron b e a m peak current, emit- characterized, because of its use as a linear collider
tance, and energy spread are very stringent [2], and [8-121.
The wavelength range of 2 to 4 nm, corresponding
to the " w a t e r window", is most suitable for imaging
* Work supported by the US Department of Energy, Offices biological materials and other applications in which the
of Basic Energy Sciences, and High Energy Physics. short pulse and high coherent power are important. In

0168-9002/93/$06.00 © 1993 - Elsevier Science Publishers B.V. All rights reserved V. PROJECTS/PROPOSALS
224 C. Pellegrini et al. / A 2-4 nm high power FEL

particular we consider the use of this system for X-ray Table 1

microscopy and holography of biological samples, with FEL characteristics
images obtained in a single pulse [13]. This requires an Electron beam properties: gun exit
energy in a single photon pulse of the order of 300 Energy [MeV] 10 10
m J / c m 2, and a subpicosecond pulse duration. Under Emittance, normalized rms
this condition an image is obtained for a sample "in [ram mrad] 3 3
vivo" before it has time to change, or be affected by Pulse duration, rms [psi 1.6 1.6
the radiation. Relative energy spread, rms [%] 0.15 0.15
The system for the 2 - 4 nm F E L consists of: a 10 Peak current [A] 250 250
MeV, S-band photoinjector; part of the SLAC linac to Electron beam properties: high energy
accelerate the beam up to 10 GeV; two longitudinal Energy [GeV] 7 3.5
bunch compressors, to increase the peak current to Emittance, normalized, rms
2500 A and reduce the bunch length to 0.16 ps; and the [mm mrad] 3 3
undulator; optical beam transport lines and experimen- Peak current [A] 2500 2500
tal areas. These components will be discussed in the Uncorrelated energy spread,
next sections. rms [%] 0.04 0.07
Correlated energy spread,
It is interesting to notice that, with an improvement
rms [%] 0.1 0.1
of present photocathode gun technology, it appears
possible to reduce in the near future the normalized Undulator properties
emittance by a factor of about 3 below that used for Period [cm] 8.3 5
the 4 nm FEL. This would open the possibility to Magnetic field [T] 0.78 0.8
construct a 0.1 nm F E L using the entire SLAC linac, Undulator parameter 6 3.7
Betatron wavelength [m] 62.8 31.4
permitting acceleration up to 55 GeV.
FEL properties
Wavelength [nm] 4 4
2, The photoinjector Field gain length [m] 6.9 5.3
Undulator saturation length [m] 60 48
The electron source is a critical component, since it Peak power at saturation [GW] 28 10
Pulse duration, rms [ps] 0.16 0.16
defines the electron beam properties. The require-
Line width, rms, including
ments for the 2 -4 nm F E L are given in table 1. A
chirping [%] 0.2 0.2
preliminary study indicates that an electron beam of Photons per pulse 2.7 x 10 i4 1 × 1014
characteristics approaching the design goal can be ob- Energy per pulse [mJ] 11 4
tained from an rf photocathode gun. This source con- Peak brightness
sists of a three and a half cell q-mode standing wave [Ph/mm2/mrad 2] 5.3 X 10 31 2.0 X 10 31
accelerating structure, excited to 100 M V / m peak field Repetition rate [Hz] 120 120
on axis, with a metal photocathode illuminated by a 2 Average power [W] 1.2 0.5
ps laser pulse. Average brightness
[ph/mm2/mrad 2] 1.0× 102I 3.7 × 1020
Using a simple analytical model [14], we estimate
that for a bunch with radial size ~ = 3 ram, and
longitudinal size ~rz = 0.54 mm, the emittance growth
due to the space charge force is 2.5 mm mrad. Addi- required design value of 3 mm mrad. Calculation using
tional emittance growth due to the time dependent rf a particle-in-cell model for our gun are in progress.
field is small, 0.7 mm mrad, because of the size of the The longitudinal phase space has an rms pulse
electron beam. length of ~rI = 0.5 mm, a peak current of 250 A, and an
The beam envelope is controlled without additional rms momentum spread 1.8 × 10 3. The uncorrelated
solenoidal or quadrupole focusing using alternating momentum spread which appears towards the rear of
gradient (ponderomotive) focusing effects [15] of the the bunch is due to the longitudinal space charge field
high gradient rf field. of the bunch and its image near the cathode.
The normalized, rms, according to Parmela simula-
tion, is 6 mm mrad for a bunch charge of 1 nC. This is
larger than the actual emittance obtainable from the 3. Beam transport, acceleration and compression
source, as numerical noise in the space charge calcula-
tion done by Parmela typically inflates the emittance We consider the acceleration and compression of
by a factor of two when compared to experimental the bunch produced by the photoinjector. We initially
results and to calculation using particle-in-ceil codes. consider the case of acceleration to 7 GeV, with a
On this basis we assume that the gun will produce the longitudinal compression by a factor of 10. At the end
 C. Pellegrini et al. / A 2 - 4 nm high power F E L 225

of the linac, we require a peak current of 2.5 kA and a

peak-to-peak energy spread A e < 0.2%.
In deciding at what energy to compress we need to
125
. . . .

~ t . . . . i , , , ' i ,

/ ' , ' i , , , , i ,

o,s

consider the longitudinal and transverse wakefields ~00 £

and rf deflections in the linac [8-12]. The first will \\ 0,6
increase the beam's energy spread and is harder to
compensate for short bunches; the last will increase the
transverse emittance and are more severe for long
bunches. In addition we have to consider the character-
istics of the bunch compressors, which limit the com-
@ 75
8

s
~\\\\/X 0,4 ~-~--°'~#c'

25 ~ . . . . . o 0,2
pression factor we can expect to achieve in a single
stage of compression.
We compress in two stages, once at 70 MeV to 0 , , , , I,, ,, I , , , , I , , , , I, ,, , I ~ 0

achieve the shortest bunch length, 200 ~zm, consistent 0 iO0 200 500 400 500
with correcting the correlated energy spread to the Linac bunch lenglh(p.m)
level of 0.2%, and again at 7 GeV to 30 ~m, to achieve Fig. 2. Transverse emittance dilution (solid line) a~d peak-to-
the desired high peak current. To study the develop- peak energy spread (dashed line) as a function of the bunch
ment of longitudinal phase space we use a computer length. The points are the values calculated and the lines are
program that considers the effects of both the longitu- for guidance.
dinal wakefields and the curvature of the rf wave.
After the initial compression the beam shape is still
very similar to a Gaussian. After the second compres- of random errors, are plotted in fig. 2, along with the
sion, the beam distribution is more sharply spiked and final peak-to-peak energy spread, as a function of the
has long tails, as shown in fig. 1. We note that the peak bunch length in the linac.
current and the final energy spread satisfy our require- The apparent knee in the energy spread occurs
ments. because one cannot use the curvature of the rf to fully
Next, we calculate the transverse emittance dilution cancel the longitudinal wakefield for bunches shorter
due to the transverse wakefields, rf deflections, and than roughly 200 txm. At a bunch length of 200 Ixm, we
dispersive errors. To model the SLAC linac, we assume find 25% emittance growth along the linac.
150 ~m rms random misalignments of the quadrupoles
and BPMs, 300 txm rms random misalignments of the
accelerator structures, and a random transverse-longi- 4. FEL performance
tudinal coupling grins = 2 × 10 4 for the rf deflections.
Finally, we assume a transverse beam jitter equal to The F E L design goal is to produce i0 ~4 photons in
the rms beam size. The results, averaged from ten sets a subpicosecond pulse for biological imaging. To reach
this goal we optimize the FEL for maximum laser
power. There are two possible design strategies: a)
3 t [ I i i I I [ I 0,2. large electron energy, in the range of 5 to 10 GeV, and
a long undulator to obtain the required F EL power
directly at saturation; b) a lower electron energy, 2 to 3
Z //f~\ 04 GeV, with the option of using a tapered wiggler after
the SASE saturation, to increase the F E L power fur-
ther when necessary.
We will investigate both options to optimize the
H I: //// b system design. To obtain an initial estimate of the F EL
performance and explore the parameter space we have
-0,1 been using the analytical form of the F EL gain ob-
tained by Chin, Kim, and Xie [16]. We have also used a
program developed by Ben-Zvi and Yu [17], and the
p I . ,-01.t_ I ~ '1.. ,01,1 I 0/ -0.2 simulation code T D A [18]. These theories and codes
-0,2 0 ,2

z(mm}
include energy spread, three dimensional effects, and a
Fig. 1. Current distribution after the final compression (solid general betatron focusing. The results are shown in
line) and a Gaussian fit to the core (dots); the head of the table 1.
bunch is to the left. The energy variation correlated with We have considered one case with a high energy
longitudinal position is shown on the right axis; the uncorre- beam, option a). This is optimized using the largest
lated energy spread is 0.04%. undulator parameter, K = 6, and minimum beam en-

V. PROJECTS/PROPOSALS
226 C. Pellegrini et al. / A 2 - 4 nm high power FEL

ergy, 7 GeV, compatible with a simple undulator de- 6. X-ray beam lines and optics
sign, and an output power larger than 10 GW. The
FEL characteristics for this case are given in table 1,
Extracting and processing the FEL output radiation
column 2. In the same table, in column 3, we show the presents challenging optical engineering problems, be-
case for a reduced beam energy, and a smaller undula-
cause of the unprecedented peak power and brevity of
tor parameter, K = 3.7. When comparing to the 7 GeV
the anticipated radiation pulses. The main tasks will be
case one should also notice that the betatron wave-
to: 1) deflect the output radiation out of the
length has been reduced as the energy ratio, so that bremsstrahlung cone produced by the electrons on the
the beam transverse radius is the same in the two residual gas in the FEL and the upstream collimators;
cases. Note that at smaller energy the uncorrelated
2) to further monochromatize the radiation, if needed.
energy spread is larger, affecting the FEL gain, while
At the power densities expected at normal incidence,
the correlated component remains the same. The gain of the order of 1016 W / c m 2, it is easy to assess that
length is shorter at 3.5 GeV than at 7 GeV. The
energy in excess of 1 e V / a t o m will be deposited in the
saturated power is lower than in option a, but is still
irradiated volume and that ablation of the irradiated
acceptable. surface will become probable due to the comparatively
Another possibility is the use of low field, long
long time constants of alternative energy-removal
period undulator, to obtain a large value of the undula-
channels. This precludes the use of solid state optics,
tor period in a simple electromagnetic structure [19-
such as multilayers, that work at large angles of inci-
21]. dence, leaving multifaceted optics operating at extreme
grazing incidence as perhaps the only viable choice.
An alternative approach that could mitigate the
damage problem would be to configure a series of gas
5. Undulator jets with density gradients.

Our initial discussion of the undulator is done for

the parameters given in table 1, column 2. A simple References
undulator'design can be based on an iron free perma-
nent magnet undulator. For a period of 8.3 cm and a
[1] C. Pellegrini, J. Opt. Soc. of Amer. B2 (1985) 259.
peak magnetic field of 0.78 T, we can select a gap of
[2] K.-J. Kim et al., Nucl. Instr. and Meth. A239 (1985) 54;
1:5 cm. This is a conveniently large gap, that would
K.J. Kim, Phys. Rev. Lett. 57 (1986) 1871;
allow us to install instrumentation for beam diagnostic C. Pellegrini, Nucl. Instr. and Meth. A272 (1988) 364;
all along the undulator. L.-H. Yu and S. Krynsky, Nucl. Instr. and Meth. A285
The natural betatron wavelength for a planar undu- (1989) 119.
lator with these characteristics is h/3 = 2 1 / 2 A w y / K = [3] R.L. Sheffield, Photocathode RF Guns, in Physics of
273 m. We add additional focusing, to obtain the Particle Accelerators, AIP vol. 184, eds. M. Month and
required betatron wavelength of 62 m, using a FODO M. Dienes eds. (1989) p. 1500.
type quadrupole system. To obtain the 62 m betatron [4] Workshop on Fourth Generation Light Sources, SSRL
wavelength at 7 GeV we can use 40 cm long Report 92/02, eds. M. Cornaechia and H. Winick.
[5] C. Pellegrini, ibid., p. 364.
quadrupoles separated by 40 cm of drift, with a gradi-
[6] K.-J. Kim, ibid., p. 315.
ent of 14 T / m . The phase shift per cell is 9°, and the
[7] W. Barletta, A. Sessler and L. Yu, ibid:, p. 376.
modulation of the betatron wavelength is quite small, [8] K. Bane, AIP Conf. Proc., vol. 153 (1987) p. 971.
with a ratio of maximum to minimum values of only [9] J. Seeman et al., Proc. US Particle Accelerator Conf.,
1.1. IEEE Conf. Proc. 91CH3038-7 (1991) p. 2064.
Because of the choice of an iron free undulator the [10] M. Ross et al., ibid., p. 1202.
quadrupoles can be placed around the undulator. Their [11] J. Seeman et al., ibid., p. 3210.
radius is 12 cm, corresponding to a pole tip field of [12] T. Raubenheimer, SLAg-Report 387 (1991).
0.84 T. Having the quadrupoles around the undulator [13] J.C. Solem and G.C. Baldwin, Science 218 (1982) 229;
has the advantage that no interruption is required in M. Howells and J. Kirz, in: Free Electron Generation of
the undulator itself, avoiding phase matching prob- Extreme UV Coherent Radiation, eds. J.M.J. Madey and
C. Pellegrini, AIP vol. 118 (1984) p. 85.
lems. For the same reason it is possible to position
[141 K.-J. Kim, Nucl. Instr. and Meth. A275 (1989) 201.
beam steering magnets and other correctors around [15] C. Hartman and J.B. Rosenzweig, submitted to Phys.
the undulator in between quadrupoles. Rev. A (1992).
If the selected site at SLAC could contain a 220 m [16] Y.H. Chin, K.-J. Kim, and M. Xie, LBL Rep. 32329
insertion device, one could consider using the weak- (1992), submitted to Phys. Rev. A.
field structure discussed in [19-21]. [17] I. Ben Zvi and L.-H. Yu, private communication.
 C. Pellegrini et al. / A 2 - 4 nm high power F E L 227

[18] T.M. Tran and J.S. Wurtele, Comp. Plays. Commun. 54 [20] R. Tatchyn, T. Cremer and P. Csonka, Nucl. Instr. and
(1989) 263. Meth. A308 (1991) 152.
[19] R. Tatchyn, Proc. Workshop on Fourth Generation Light [21] R. Tatchyn, Proc. Workshop on Fourth Generation Light
Sources, SSRL Report 92/02, eds. M. Cornacchia and H. Sources, SSRL Report 92/02, eds. M. Cornacchia and H.
Winick (1992) p. 605. Winick (1992) p. 417.

V. PROJECTS/PROPOSALS