CURRENT-ENHANCED SASE USING AN OPTICAL LASER AND ITS

APPLICATION TO THE LCLS

A.A. Zholents, W.M. Fawleyy , LBNL, Berkeley, CA 94720-8211, USA

P. Emma, Z. Huang, G. Stupakov, SLAC, Stanford, CA 94309, USA
S. Reiche, UCLA, Los Angeles, CA 90095-1547, USA

Abstract sion from the remainder of the e-beam if the exponential

gain lengths in the current spikes have been reduced by
We propose a significant enhancement of the electron
peak current entering a SASE undulator by inducing an en-
25% or greater.
The ESASE technique shares with other recent propos-
ergy modulation in an upstream wiggler magnet via res-
als [5, 6, 7] the common feature of employing an optical
onant interaction with an optical laser, followed by mi-
laser to induce energy modulation to produce short du-
ration (P 500 as) output pulses from an FEL together
crobunching of the energy-modulated electrons at the ac-
celerator exit. This current enhancement allows a reduc-
with absolute synchronization of the x-ray probe pulse to
tion of the FEL gain length. The x-ray output consists of
the laser pump pulse, thus permitting pump-probe experi-
a series of uniformly spaced spikes, each spike being tem-
ments. In principle, it is also related to harmonic cascade
porally coherent. The duration of this series is controlled
configurations [8]. However, via the use of a dispersion
by the laser pulse and in principle can be narrowed down
to just a single,  200
-attosecond spike. Given potentially
section to strongly enhance the current, ESASE has the
additional advantage of significantly reduced gain lengths.
absolute temporal synchronization of the x-ray spikes to
This may allow possible relaxation of the required electron
the energy-modulating laser pulse, this scheme naturally
beam emittance for saturation in the ”standard” LCLS un-
makes pump-probe experiments available to SASE FEL’s.
dulator or (for the same base emittance) saturation at x-ray
We also study various detrimental effects related to the high
wavelengths shorter than 0.15 nm.
electron peak current .
We now present initial “start-to-end” numerical simula-
tion results, investigate some potential detrimental effects
INTRODUCTION due to the high peak currents, and then conclude with some
remarks concerning possible experimental applications.
SASE-based x-ray FEL projects continue to gather mo-
mentum at SLAC[1] and DESY[2]. Recently, Ref. [3]
has proposed a technique (which we call current-Enhanced
ACCELERATOR SIMULATIONS
SASE or ESASE for short) to shorten the exponential gain We initiated “start-to-end” ESASE simulations with 2 
length and provide temporal synchronization of the x-ray 105 macro-particles in the Parmela[9] code for the RF
pulse to a master laser. Figure 1 shows a schematic of photo-cathode gun and low-energy injector systems, up to
this technique as applied to the LCLS. An external, short an energy of 135 MeV. After this point, the macroparti-
pulse (30-100 fs) from a high power (1-10 GW) laser 10
cle number is increased , retaining all second moments
at wavelength L 2.2 m (e.g.TOPAS laser [4]) overlaps and coordinate correlations, and then transported with the
a short e-beam section at moderate energy (4.5 GeV) Elegant[10] code through the main linac, up to the entrance
in a short, appropriately-tuned wiggler (e.g. u 30 cm, of the FEL undulator. These simulations included 2nd-
Bu 1.7 T). This induces a periodic energy modulation order optics, accelerating structure wakefields, a model for
with relative amplitude B 

= 0  5-10, where the laser/wiggler energy modulator (EM) at 4.54 GeV, and
 0 is the uncorrelated energy spread, but with minimal a line-charge (1D) model of coherent synchrotron radiation
microbunching. Following final acceleration the electron (CSR) in and after the bends.
beam passes through a dog-leg bend whose chromatic dis-
persion produces a periodic enhancement of the peak cur-
rent to 20 kA. Finally, the electrons enter a long undu-
lator and emit x-rays via the standard SASE process but
with first saturation occurring by z 45-60 m. The ra-
diation output from the current-enhanced portions will be
a train of spikes at uniform spacing L whose total du-
ration is directly related to that of the upstream, energy-
modulating laser. This emission will dominate SASE emis-
 Work supported in part by the Office of Science, U.S. Dept. of Energy
under Contracts DE-AC03-76SF0098 and DE-AC02-76SF00515. Figure 1: A schematic of ESASE as applied to the LCLS.
y WMFawley@lbl.gov
 1 20
4 1

γεx,y (µm)

γεx,y (µm)
Ipk (kA)

(kA)
γεx (solid)
0.5 10
0.5 γε (solid)
2 γε (dash)

pk
x
γεy (dash)

I
y

0 0 0 0
−2 0 2 −2 0 2 −2 0 2 −2 0 2
z (µm) z (µm) z (µm) z (µm)
0.02 0.06
0.1 0.1

∆E/E (%)
∆E/E (%)
σ /E (%)

σ /E (%)
0.04
0.05 0

0
0
0.01
0

0
0.02
E

E
0
−0.1
0 −0.05 0
−2 0 2 −2 0 2 −2 0 2 −2 0 2
z (µm) z (µm) z (µm) z (µm)

Figure 2: ELEGANT tracking results for a short section Figure 3: Same plots as in Fig. 2, but at a z -position imme-
of 0.8-m energy-modulated e bunch at 14 GeV, with in- diately following the DL2 beamline. Strong current modu-
stantaneous current, transverse emittances, rms relative en- lation and CSR effects are now apparent.
ergy spread, and relative energy centroid modulation plot-
ted versus bunch coordinate z .
particles and narrow bin width (6 nm) due to the short
ESASE bunching length (30 nm rms). This spike
The EM consists of a resonant laser-electron interaction length still greatly exceeds the CSR point-charge wake
in an 2.5-m long planar wiggler located immediately after limit of R= 3  0.01 nm at 14 GeV for bends with radius
the second bunch compressor chicane where the FWHM R 300 m [11]. It is worth noting that the CSR in the DL2
bunch length is 65m. The EM can easily be installed in dipoles will be partially suppressed by a combination of a
place of one existing 3-m long accelerating structure. Ta- non-zero R51 component and a relatively large transverse
ble 1 gives the EM and the final buncher parameters. beam size. The projected spike length in the bends along
Figure 2 displays the resulting current profile, transverse the axis of propagation will broaden to xjj  400 nm,
slice emittance, rms energy spread, and centroid energy where x  45m is the rms horizontal beam size at the
modulation induced by an 0.8-m wavelength EM for a 05
bend entrance and   : Æ is the bend angle. This longer
short core-section of the beam at 14.1 GeV. The current effective length, the low charge in each spike, and the weak
profile shows weak density modulation due to the wiggler’s bend angles all minimize CSR effects, although a 20% hor-
small momentum compaction (opposite in sign to the final izontal emittance growth and some phase space distortion
buncher). Similar weak modulation of E is also apparent. is still evident in Fig. 3. Nonetheless, a fully 3D CSR cal-
Following final acceleration, the beam passes through a culation is needed to verify these 1D results.
‘dog-leg’ transport line (DL2) which is used to produce
a 1.25-m jog in the beamline to the south, in order to al-
low energy measurement and collimation. The DL2 beam- FEL SIMULATION RESULTS
line nominally has a very small momentum compaction of
R56 0.1 mm, but it can be easily increased, or even re-
The 6D macroparticle distributions resulting from the
ELEGANT tracking studies were directly imported as
versed in sign, by small adjustments in the gradients of
input to the FEL simulation codes GINGER[12] and
three nearby quadrupole magnets. For the energy modu-
=
lation shown in Fig. 2, R56 0.30 mm is optimum. Fig. 3
GENESIS[13]. We performed a number of fully time-


dependent ( tsli e 10 as) SASE runs at two different
shows the beam characteristics immediately following this
energy modulation wavelengths (0.8 and 2.2 m) and two
bend system, including CSR effects.
undulator focusing lattices (h i 26 and 12 m) with the=
CSR calculations require both a large number of macro-
=
standard LCLS u 30 mm undulator period. For these
Table 1: Energy modulator (EM) at 4.54 GeV for B =5 and calculations wakefield effects were neglected. Table 2 sum-
=
two laser wavelengths (L 0.8 m and 2.2 m).
Parameter sym 0.8 m 2.2 m unit Table 2: Simulation results from GINGER and GENESIS.
N wiggler periods Nw 8 8 — Param. h i = 26 m h i = 12 m unit
period of wiggler w 25 30 cm L STD 0.8 2.2 0.8 2.2 m
peak laser power Ppk 9.7 10.7 GW Lsat 70 58 57 44 45 m
laser rms waist r 0.25 0.25 mm hP i 13 2.0 3.0 2.9 7.6 GW
modulation amp.
14 14 — Pspike 240 17 65 40 160 GW
buncher R56 R56 0.30 0.78 mm !=
! 1500 550 660 660 790 –
 uniformly-spaced spikes with effective widths 0.5 fs up
to and about one-to-two gain lengths beyond Lsat . Be-
tween the individual spikes, the instantaneous SASE power
is smaller by three orders of magnitude or greater. As pre-
dicted by Ref. [14] for SASE in the short pulse limit, the in-
dividual spikes support a single coherent longitudinal mode
and their bandwidth is approximately transform-limited for
z  Lsat (presuming one has matched L =B to approxi-
mately one coherence length). Fig. 5 shows P t snap-()
22
shots at various z -locations for a : -m laser-modulated
pulse with h i=12 m. At z =57 m, approximately three gain
lengths beyond saturation, some spikes begin to display
multi-peak structure and pulse widths approaching 1-fs or
greater due to slippage effects.

Figure 4: hP i plotted versus z for ESASE and standard

LCLS configurations. The solid lines and boxes represent COLLECTIVE EFFECTS
GINGER and GENESIS results, respectively.
The high peak current in the ESASE spikes raises con-
cerns that impedance effects may induce sufficient energy
spread to seriously degrade FEL performance. We now es-
timate the effects from three different sources.

Wakefield Effects
The longitudinal resistive wall wake of high-current
spikes will be partially suppressed by the large catch up
distance z associated with the short spike length. Indeed,
for spike =30 nm and undulator pipe radius a=2.5 mm,
= 2
we find z a2= spike  100 m, which exceeds the ex-
pected FEL saturation length. Due to the finite beam , the
()
Figure 5: P t snapshots at 5 different z -locations for a
2.2 m-energy-modulated ESASE pulse with h i 12 m,= extension of the image charge induced on the wall by a sin-
gle spike, a= 100 nm, is larger than spike . This some-
plotted with staggered offsets of 1.5 fs in time and 15 GW what reduces z . However, even without this reduction, the
=
in power. For legibility, the z 37 m data has been multi- wake induced by a single spike inside itself is expected to
plied by a factor of 2.0. be small due to the small charge of the spike ( O (10)pC).
marizes various output parameters from the simulations.
If we take for the wake the value w = Z0 =a2 (the
limit of the steady state wake for very short bunches; see,
For simplicity, we have defined the saturation point as that
where the normalized inverse spectral bandwidth != !

e.g.Ref. [15]) where Z0 =377 is the vacuum impedance,

we estimate the induced energy spread in the spike due to
reaches a maximum and Pspike as the average peak power
of the spikes measured at z = Lsat. The ESASE results this wake will be ÆE=E   4 10 4 at z =100 m. This is
smaller than the intrinsic energy spread E of the ESASE
show significantly reduced saturation lengths, albeit with
spike and should only slightly affect FEL performance.
somewhat smaller average powers and inverse bandwidths
than the “standard” LCLS configuration. In particular, if it The total charge in the spikes is a relatively small frac-
were possible to achieve an h i=12 m or less, Lsat is re- tion of the total charge of the bunch. This means that the
duced to less than 50 m. We believe the enhanced gain from average wake of multiple spikes will be dominated by the
such an ESASE configuration would allow the LCLS to op- averaged smooth current distribution in the bunch. This
erate at a somewhat larger normalized emittance than the wake is not supposed to generate excessive energy spread
nominal 1.2 mm-mrad or at somewhat shorter wavelengths in the beam.
than the nominal 0.15 nm. The time-averaged powers for
both the ESASE cases and standard LCLS parameter cases Coherent Undulator Radiation
(“STD”) are shown in Fig. 4. Because GENESIS unlike
GINGER includes non-axisymmetric modes, in the early Coherent undulator radiation (CUR) at wavelengths
regions of exponential gain it will tend to show a signifi- comparable to the spike length (30 nm) but much
cantly higher power. longer than the resonant FEL wavelength r =2
=kr
One important feature of the ESASE runs is that the (=0.15 nm) can be an important impedance source. The
instantaneous power is essentially composed of simple, CUR impedance per unit length at wavenumber k for a long
pencil beam is [16, 17] CONCLUDING REMARKS
 
Z0 K2k
1D
ZCUR (k) = 4 2 8 2 + 4i ln kkr ; (1) The ESASE technique allows obtaining electron beam
micro-bunches with a peak current unthinkable from a stan-
where K is the undulator parameter. However, long- dard electron bunch compressor. Our preliminary inves-
wavelength undulator p radiation is emitted primarily off- tigations suggest that the high peak currents induce only
axis at an angle   2 1
=u  mrad. Consequently, limited beam quality degradation due to collective effects
such as CSR, wakefields, and space charge; nonetheless
the effective radiation impedance taking into account the
transverse size of the bunch is modified to more detailed studies are needed for greater certainty. Al-

= ZCUR
though the exponential ESASE gain length is only some-
3D
ZCUR  ZCUR
1D
e k22 x2 1D
e 2kkux2
: (2) what shortened relative to an unmodulated beam for the
2
with ku  =u . For an infinite train of ESASE bunches standard LCLS undulator with  =26 m, there is a much
spaced by laser wavelength L , the modulated electron cur- shorter gain length and more rapid saturation in z for
rent is (see e.g., Ref. [8])  =12 m. For this latter case, detailed “start-to-end” sim-
 1
X
ulations show that the output SASE x-ray pulse will be
I (z ) = I0 1+2 Jn (nkLR56BÆ ) dominated by periodic spikes of a few hundred attosecond
n=1 or less duration. Consequently, the duration of the overall
 output envelope can be controlled by adjusting the modu-
 exp 2 os(nkLz)
 
n2 kL2 R256Æ2 = ; (3) lating laser pulse. One can also obtain absolute synchro-
nization between these two pulses which can be used in the
where I0 =3.4 kA is the unmodulated beam current, Jn is
=2
pump-probe experiments for ultra-fast dynamics studies at
the Bessel function of order n, kL =L , z is the lon- the femtosecond time scale and beyond. We plan further
gitudinal position along the bunch, and Æ   0 = is the studies to explore more fully the potential of ESASE tech-
normalized rms energy spread. At the optimal compression
1
niques at the LCLS.
kL R56BÆ  , the energy loss per unit length is
d(
) 2I0 K 2 X
1 
n2

REFERENCES
ds
 I 2 Jn(n)exp 2B 2 2knku x 2
A n=1 [1] LCLS Conceptual Design Report, SLAC-R-593 (2002).
   
kn kn kr
 4 os (2n 
z) ln
2 kn sin(2 nz) ;(4)
[2] TESLA Technical Design Report, DESY TESLA-FEL
2001-05 (2001).

where IA  17 kA is the Alfvén current, kn  nkL [3] Zholents, A.A., submitted to Phys. Rev. Lett. (2004)
and z  z=L . For B =5, L =0.8m, and = 2:8  [4] http://www.lightcon.com/lc/scientific/topas.htm
104 and x=33 m in the undulator, Eq. (4) yields max-
imum d(
)=ds  0:003 m 1, more than two orders of
[5] Zholents, A.A. and Fawley, W.M., Phys. Rev. Lett. 92,
224801 (2004).
magnitude smaller than that obtained using the 1-D CUR
[6] Saldin, E.L, Schneidmiller, E.A., and Yurkov, M.V., Opt.
impedance (i.e., Eq. (1)). Comm, 237, 153 (2004)
[7] Saldin, E.L, Schneidmiller, E.A., and Yurkov, M.V., Opt.
Longitudinal Space Charge Effects Comm., in press, (2004)
Another high-frequency impedance is the longitudinal
1
[8] L.H. Yu, Phys. Rev. A, 44, 5178 (1991).
space charge (LSC). Since knx =  (pencil beam in
1
the rest frame) and kna=  for all n except n at =1 [9] J. Billen, LANL Report LA-UR-96-1835 (1996).

L =0.8 m, we use the free-space LSC impedance per unit [10] M. Borland, ANL report APS LS-287 (2000).
length (see, e.g., Ref. [18]): [11] E.L. Saldin, E.A. Schneidmiller, and M.V. Yurkov, Nucl.
  Inst. Meth. A, 398, 373 (1997).
Z0 k
ZLSC (k)  i
4 2 1 + 2ln k (x + y) : (5) [12] W.M. Fawley, LBNL-49625 (2002).
[13] S. Reiche, Nucl. Inst. Meth. A, 429, 243 (1999).
For B =5, L =0.8 m, and x =33 m, we find that
the maximum d(
)=ds  0.1 m 1. For L =2.2 m, [14] R. Bonifacio et al., Phys. Rev. Lett. 73, 70 (1994).
d(
)=ds 0.05 m 1 . For the 200-m long diagnostic [15] K.L.F. Bane and M. Sands, SLAC-PUB-95-7074 (1995).
section between DL2 and the undulator, the space-charge-


[16] E.L. Saldin, E.A. Schneidmiller and M.V. Yurkov, Nucl. In-
induced growth in to O(20) is large enough for concern. strum. Meth. A, 417, 158 (1998).
If necessary, the current bunching could be delayed by
[17] J. Wu, T. Raubenheimer, G. Stupakov, Phys. Rev. ST Accel.
making the DL2 dogleg isochronous together with instal- Beams, 6, 040701 (2003).
lation of a weak chicane just before the undulator. Over the
[18] A. Chao, Physics of Collective Beam Instabilities in High
100-m length of the undulator, the space-charge-induced
Energy Accelerators (Wiley, New York, 1993).
energy spread increase is less than that of the spikes’s ini-
tial value E .