SLAC-PUB-8962
LCLS-TN-01-05
17 August 2001
Photoinjector design for the LCLS∗
P.R. Boltona, J.E. Clendenina , D.H. Dowellb, M. Ferrarioc, A.S. Fishera,
S.M. Giermana, R.E. Kirbya, P. Krejcika, C.G. Limborga, G.A. Mulhollana,
D. Nguyend, D.T. Palmera, J.B. Rosenzweige, J.F. Schmergea, L.
Serafinif, X.-J. Wangg
a
SLAC, Stanford, CA 95309, USA
Boeing Physical Sciences Research Center, Seattle, WA 981124, USA
c
INFN-LNF, 00044 Frascati (Roma), IT
d
LANL, Los Alamos, NM 87545, USA
e
UCLA, Los Angeles, CA 90095, USA
f
INFN-MI, 20133 Milano, IT
g
BNL, Upton, NY 11973, USA
b
Corresponding author: J.E. Clendenin, P.O. Box 4349, Stanford, CA
94309, USA. Tel.: +650-926-2962; fax: +650-926-8533. E-mail address:
clen@slac.stanford.edu.
Abstract
The design of the Linac Coherent Light Source assumes that a low-emittance, 1-nC, 10-ps
beam will be available for injection into the 15-GeV linac. The proposed rf photocathode
injector that will provide a 150-MeV beam with rms normalized emittances of 1 µm in
both the transverse and longitudinal dimensions is based on a 1.6-cell S-band rf gun that is
equipped with an emittance compensating solenoid. The booster accelerator is positioned
at the beam waist coinciding with the first emittance maximum and is provided with an
accelerating gradient of ~25 MeV/m, i.e., the "new working point." The uv pulses required
for cathode excitation will be generated by tripling the output of a Ti:sapphire laser system
consisting of a highly stable cw mode-locked oscillator and two bow-tie amplifiers pumped
by a pair of Q-switched Nd:YAG lasers. The large bandwidth of the Ti:sapphire system
accommodates the desired temporal pulse shaping. Details of the design and the supporting
simulations are presented.
PACS Codes: 29.25.Bx, 29.27.Ac, 41.60.Cr, 41.85.Ar
Keywords: Photoinjectors; Low Emittance; Free Electron Lasers
Contributed to
The 23rd International Free Electron Laser Conference
Darmstadt, Germany
20-24 August 2001
∗ Work supported by Department of Energy contract DE-AC03-76SF00515.
1
1. Introduction
The proposed Linac Coherent Light Source (LCLS) is an x-ray free electron laser
(FEL) that will use the final third of the SLAC 3-km linac for the drive beam. The
performance of the LCLS in the 1.5 Å regime is predicated on the availability of a 1-nC,
100-A beam at the 150-MeV point with normalised rms transverse emittance of 1 µm. A
low energy spread, both integrated and slice, and low charge and timing jitters are also
required. The initial injector design made use of a low-gradient booster as was then the
standard [1]. An improved design that better meets the LCLS requirements is the subject
of this paper.
The basic layout of the injector is shown in Fig. 1. This layout is consistent with the
concept of a “new working point” that was introduced in 1999 [2]. The basic difference
seen in this design compared to the earlier version is that the drift distance between the
gun and booster is greater and also the booster accelerating gradient is much higher. Note
that a relatively weak solenoidal field is now used around the first booster section instead
of discrete focusing coils between sections.
2. Gun and cathode
The LCLS gun will be a modified version of the BNL/SLAC/UCLA symmetrized
1.6-cell rf photocathode gun [3]. The principal differences are the addition of a load-lock
to better ensure cathode performance and improved cooling to allow operation at 120 Hz.
In addition, the rf power for the LCLS gun will be fed symmetrically into the full cell in
order to eliminate higher order modes. To increase its heat capacity, the standard 1.6-cell
2
gun will be modified by increasing the number and size of the water cooling channels and
pushing them closer to the iris and cathode surface without compromising structural
integrity [4].
A copper photocathode is chosen because the entire end plate of the half cell can be
formed in the standard manner of rf cavities, permitting operation at the highest field
values. The photoelectric response time of metal cathodes is on the sub-picosecond level,
thus imposing no limitation on the desired temporal pulse shaping. Since the source is
not required to produce multiple microbunches within each pulse, the lower quantum
efficiency (QE) of metal cathodes compared to alkali and semiconductor photocathodes
is not a major concern. The QE for Cu illuminated with uv light depends on surface
preparation, but 10-5 for normal incidence at 266 nm (4.7 eV) in a non-load-locked gun
is achievable [5,6,7]. Much better QE is available from copper installed through a loadlock. This is illustrated by the data of Fig. 2 for which a carefully prepared
polycrystalline Cu photocathode was inserted into high vacuum surface diagnostic
chamber using a load-lock (no baking) and the QE spectrum measured at low voltage [8].
The figure shows a QE of 2.3×10-4 at 266 nm.
At 266 nm, an optical pulse of 500 µJ on the cathode is required to produce 1 nC of
charge when the QE is 10–5. A laser system to meet this requirement is described in
Section 3.
Copper photocathodes made from single-crystal Cu have proven to have not a only
high QE but also low dark current [9]. In addition, the QE uniformity across the cathode
surface appears to be superior to polycrystalline Cu. Single-crystal copper boules with a
3
diameter sufficient to fabricate the back plane of an S-band gun are available, thus the use
of a cathode plug can be avoided even for single-crystal Cu cathodes.
3. Laser
The laser system for illuminating the cathode is essentially unchanged from that
shown in reference [1]. A cw mode-locked Ti:sapphire oscillator, pumped by a diodepumped Nd:YVO4 laser, delivers a stable, continuous pulse train of 12-nJ, 100-fs pulses
at a repetition rate of 79.33 MHz. This frequency, the 36th subharmonic of the linac’s
2856-MHz rf, locks the timing of the laser pulses to the phase of the rf in the linac and rf
gun. The wavelength is tuned to 798 nm and tripled to 266 nm after amplification to
provide a suitable wavelength for the photocathode.
A Pockels cell and polarizer are used to gate single pulses, at 120 Hz, from the 79.33MHz pulse train. The energy of such selected pulses is increased in a two-stage
Ti:sapphire rod amplifier system using a four-pass bow-tie configuration for each stage
[10]. Both are pumped by a pair of Q-switched, doubled Nd:YAG lasers that fire in
alternation, each producing a 60-Hz train of 3 to 10-ns pulses. Relay imaging is used to
maintain a good transverse mode while efficiently filling the pumped volume of the
Ti:sapphire rods.
The large oscillator bandwidth (of order 1%) enables transformed-limited ultrashort
pulse-width capability and use of the well-established chirped pulse amplification [11]
technique to reduce peak power levels in the amplifier stages. With this scheme, pulse
stretching to hundreds of picoseconds is accomplished by imposing a positive chirp on
4
the waveform of the pulse. Following amplification the stretched pulses are compressed
by reversing the imposed chirp.
The temporal pulse shape is determined by modifying (under computer control) the
oscillator output spectrum. Additional pulse-width flexibility is afforded by partial
compression of the stretched, amplified pulses.
After the second amplifier, the transverse shape of the pulse is modified from
Gaussian to uniform to better match the requirements for obtaining a low emittance from
the gun. Next, two crystals triple the frequency of the light to a wavelength of 266 nm.
The spatially flattened pulse also improves efficiency and uniformity in this harmonicgeneration process.
Finally, the beam is transported through an evacuated tube to an optics platform next
to the gun. Since the Fourier-relay image plane that follows the long transport tube has a
spot size that is too small for the photocathode, the spot is magnified and imaged onto a
circular aperture that slightly trims the edge of the beam. This aperture is in turn imaged
onto the photocathode, so that the illuminated region of the photocathode is precisely
defined without jitter. The imaging includes compensation for the temporal and spatial
distortion that results if grazing incidence on the photocathode is used.
The net energy transmission of the laser system is as follows: transmission through
the spatial flattener is about 50%, through the compressor 50%, through the frequency
tripling stage 25%, and through the optical transport to the gun 50%. If grazing incidence
is used, the final steering optic will be a uv grating. Appropriate choice of coating,
polarization, blazing angle and groove density of the grating can yield first order uv
5
diffraction efficiencies in excess of 90%. Consequently, starting with 18 mJ after the
second amplifier, the required 500 µJ is delivered to the cathode.
4. Simulations
Using the semi-analytic code HOMDYN [12], a wide range of injector parameters
was investigated earlier under the constraints imposed by the invariant-envelope [13]
matching condition: injection into the matched accelerating gradient of the booster at a
laminar waist. As a result, it was found that by increasing the gun focusing solenoid
strength so that the waist also occurs when the emittance has its relative maximum, the
second emittance maximum can be shifted to higher energy with a lower final emittance
value than previously achieved [2,14]. This new configuration is here referred to as the
“new working point.”
The LCLS injector has been designed using version 3 of the LANL-maintained code
PARMELA to establish the details of the “new working point.” The electric field map of
the gun was obtained with SUPERFISH and directly used in PARMELA. SUPERFISH
was also used to simulate the fields in the traveling-wave accelerating sections, and space
harmonics were calculated to be used in PARMELA. RF fields were assumed to be
cylindrically symmetric. A magnetic field map for the emittance compensating solenoid
at the gun was produced using POISSON and passed to PARMELA. The magnetic field
for the air core solenoid around the first accelerating section was modeled in PARMELA
using single coils each with appropriate strength to represent the field.
6
Using only the gun, solenoid, and the immediately following drift space (i.e., no
booster), the first emittance minimum after the solenoid was optimized by varying the
solenoidal field and the beam radius at the cathode. A value of Bz = 3.15 kG and hardedge radius of 1 mm was found to be optimum. The emittance minimum very nearly
coincides with the “new working point.” A slightly larger value of BZ is found here than
with HOMDYN, consistent with the solenoid being displaced somewhat downstream
because of the physical interference of the gun structure.
Next the position of the booster with an accelerating gradient of 25 MV/m was
optimized followed by the position and field of the linac solenoid. The injector
parameters determined in this manner are summarized in Table 1. A thermal emittance of
0.26 µm has been added to the PARMELA deck.
For the parameters of Table 1, the emittance and beam size for a risetime of 0.5 ps is
shown in Fig. 3 as a function of distance from the cathode. The normalized transverse
phase space at the exit of the booster for 100K particles is shown in Fig. 4. The upper left
plot is a normalized x-y scatter plot, with xn and yn amplitudes in units of rms beam size.
The normalized xn − xn′ phase space is shown in the upper right with the rms emittance
ellipse given by the circle of unity radius in the center. The density of particles in the
center of the beam is not evident in these plots. The normalized rms slice emittance in x
and y, as a function of axial distance along the bunch, is shown in the lower left. The
projected value is shown by the horizontal line. On the lower right, the beta-mismatch
amplitude, ζ, is shown as a function of ∆z. The beta-mismatch amplitude, which is
normalized such that ζ LQGLFDWHV WKH GHJUHH RI PLVPDWFK EHWZHHQ WKH VOLFH 7ZLVV
7
parameters and the projected Twiss parameters. Transporting the beam through a followon accelerating channel may be difficult if the variation in ζ is large [15].
The same PARMELA simulations yield an integrated (slice) energy spread at the
booster exit that is within 0.1 (0.005)%. However, initial results of an ongoing
comparison of PARMELA with PIC code simulations near the photocathode indicate that
the energy spread predicted by PARMELA may be overly optimistic [16].
5. Conclusions
An rf photoinjector for the LCLS based on the “new working point” configuration
has been described. The injector will utilize a 1.6-cell rf gun designed for 120 Hz and
equipped with a Cu photocathode and load-lock. The cathode is illuminated with 500 µJ
of uv light provided by a tripled-Ti:sapphire laser. At the output of the injector, the 150
MeV beam has a transverse emittance that is well below the 1-µm requirement and the
slice emittance is about 25% (relative) lower except for the head and tail. The integrated
(slice) energy spread is also very low.
8
References
[1] R. Alley et al., Nucl. Instrum. and Meth. A 429 (1999)
324.
[2] M. Ferrario et al., in The Physics of High Brightness
Beams, eds. J. Rosenzweig, L. Serafini, World Scientific
(2000), p. 534.
[3] D.T. Palmer et al., SPIE 2522 (1995) 514.
[4] X.J. Wang et al., “High-Rep Rate Photocathode Injector
for LCLS,” contributed to the 2001 Particle Accelerator
Conference, June 18-22, 2001, Chicago, IL.
[5] T. Srinivasan-Rao et al., J. Appl. Phys. 69 (1991)
3291.
[6] P. Davis et al., in Proc. of the 1993 Particle
Accelerator Conference, p. 2976.
[7] E. Chevallay et al., Nucl. Instrum. and Meth. A 340
(1994) 146.
[8] G. Mulhollan, "Common Sense Copper and RF Guns," LCLSTN-99-9 (July 1999), unpublished.
[9] P.R. Bolton et al., “Transverse and longitudinal
emittance measurements on an S-band photocathode rf
electron gun,” this conference. See also D.T. Palmer et
al., in The Physics of High Brightness Beams, eds. J.
Rosenzweig, L. Serafini, World Scientific (2000), p. 439.
[10] The gain per pass in Ti:sapphire amplifiers is
sufficient to get from 1 nJ to a few mJ in eight (8)
passes. For example, see S. Backus et al., Opt. Lett. 20
(1995) 2000.
[11] P. Maine et al., IEEE J. Quantum Electron. QE-24
(1988) 398.
[12] M. Ferrario et al., Part. Acc. 52 (1996).
[13] L. Serafini, J. Rosenzweig, Phys. Rev. E 55 (1997)
7565.
[14] M. Ferrario et al., Proc. of the 7th European Particle
Accelerator Conference, Vienna (2000) 1642.
[15] P. Emma, private communication.
[16] V. Ivanov, E. Colby, C. Limborg, private
communication.
9
Table 1. PARMELA parameters and results.
Parameter
Value
Bunch charge at cathode/at booster exit
1.0/1.0 nC
Bunch shape at cathode spatial/temporal
Uniform/uniform
Bunch radius at cathode
0.71 mm rms
Bunch length at cathode/at booster exit
2.9/2.9 ps rms
Peak rf field at gun (extraction phase)
140 MV/m (32°)
Gun solenoid axial field
3.15 kG
Cathode to booster-entrance distance
1.4 m
Booster accelerating gradient (phase L01/L02)
25 MV/m (-2.5/+2.6°)
Linac solenoid axial field/length
-1.5 kG/1.0 m
Energy at booster exit
150 MeV
Integrated (slice) energy spread, σ γ γ 0
0.10 (0.005) % rms
Normalized transverse emittance:
0.5 (1.0) ps rise time, εn,th included, 100 K particles
10
0.80 (0.95) µm rms
Figure Captions
Fig. 1. Schematic layout (not to scale) showing only the principal beamline elements, the
location of the diagnostics, and the rf distribution system. In the figure are shown the rf
gun (G), the emittance compensating solenoid (S1), charge coupled devices (CCD),
klystrons (K), the focusing solenoid (S2) around the first 3-m accelerating section (L0-1)
of the booster.
Fig. 2. QE of copper as a function of quantum energy measured with low (22 V) dc bias
with the surface untreated after installation in the analysis system using a load-lock.
Fig. 3. Transverse normalized rms emittance as a function of distance from the cathode
for 100K particles. A rise time of 0.5 ps is assumed. A normalized rms thermal emittance
of 0.26 µm is included.
Fig. 4. Normalized transverse phase space at the exit of booster for 100K particles. (1)
Distribution of particles in the beam (upper plots). The scales are derived from the
right-hand figure in which the rms emittance ellipse in the x-x’ plane (only) is normalized
to a circle having a radius of unity. (2) Transverse normalized slice emittances (lower left
plot) in both planes and mismatch parameter, ζ, (lower right plot) in both planes along
the bunch z-axis. The bunch head is at the right.
11
Figure 1
K
K
S1
G
S2
K
LO-1
e–
LO-2
CCD
Legend
8-2001
8612A1
12
Solenoid
Toroid
x and y Corrector
Viewing Screen
Faraday Cup
BPM
Figure 2
Quantum Yield (electrons/photon)
10–2
10–3
10–4
10–5
6.0
4-2001
8560A98
5.5
5.0
Energy (eV)
13
4.5
2.5
2.0
2.0
1.6
1.5
1.2
Emittance
1.0
0.5
0
0.8
0.4
Beam Size
0
200
σx,y (mm)
γ εx,y (µm)
Figure 3
400
z (cm)
14
600
800
0
3-2001
8560A76
Figure 4
10
x'n
yn
10
0
–10
–10
0
xn
–10
–10
10
0
xn
10
1.0
0.4
ζx,y (mm)
0.8
γεx,y (µm)
0
γεx
0.5
ζx
γεy
0
3-2001
8560A88
–1
ζy
0
∆z (mm)
0
1
15
–1
0
∆z (mm)
1