Properties of an Ultra-Short Gain Length, Saturated, Self-Amplified

Spontaneous Emission FEL

A. Murokh, R. Agustsson, P. Frigola, C. Pellegrini, S. Reiche, J.

Rosenzweig, A. Tremaine
UCLA Department of Physics and Astronomy, 405 Hilgard Ave, Los
Angeles, CA 90095-1547

M. Babzien, I. Ben-Zvi, E. Johnson, R. Malone, G. Rakowsky, J.

Skaritka, X. Wang, V. Yakimenko
Brookhaven National Laboratory, Upton, NY 11973

L. Bertolini, J.M. Hill, G.P. Le Sage, M. Libkind, A. Toor, K.A. Van

Bibber
Lawrence Livermore National Laboratory, Livermore, CA 94551

R. Carr, M. Cornacchia, L. Klaisner, H-D. Nuhn, R. Ruland

Stanford Linear Accelerator Center, Stanford, CA 94309

Abstract

The VISA experiment, conducted at the BNL Accelerator Test Facility (ATF), has studied the
properties of SASE FEL process in a saturating system. The experiment utilized a high brightness electron
beam and a strong focusing undulator. Saturated gain greater than 108, with a power gain length below 18
cm, was obtained at 840 nm. Measurements of FEL gain, spectral and angular properties of SASE
radiation are reported, and the results are compared to theory and to start-to-end simulations of the system.
FEL performance was found critically dependent on the compression of the electron beam.

PACS numbers: 41.60.Cr, 41.60.Ap, 41.85.Ja

 Since the concept of an unseeded, single pass, self-amplified
spontaneous emission free-electron laser (SASE FEL) was introduced
[1], significant progress has been made in theoretical, computational,
and experimental programs directed towards the design and
construction of X-ray FELs [2]. A series of successful experiments
demonstrated high gain SASE operation in the IR [3,4], visible [ 5,6],
and ultraviolet [7] spectral ranges. Yet the electron beam quality
achieved so far in operating SASE FELs has been significantly lower
than required for the Linac Coherent Light Source (LCLS) X-ray FEL
[2]. The VISA SASE FEL, reported here, was designed to fill this gap
by achieving high gain and saturation in a short undulator [8], taking
maximum advantage of the low emittance electron beam produced by
the ATF injector [9].

VISA reached saturation in 3.6-3.8 m at a central wavelength of

840 nm. The properties of SASE radiation, as a function of distance
through the 4-m undulator magnet have been measured and analyzed.
For the first time, observation and analysis of the statistical intensity
fluctuations of SASE radiation at saturation is reported, and compared
to the measurements obtained during exponential growth. The results
are compared to the start-to-end numerical model of the experiment,
which follows the electron beam dynamics from the emission at the
photocathode, through acceleration, transport and into the undulator.
The focus of this work is on the results obtained at the fundamental
FEL wavelength. Other results, including measurements on the FEL
harmonics radiation and electron beam microbunching are reported
elsewhere [10].

The ATF injector, which consists of a 1.6-cell photocathode S-

band gun and two consecutive SLAC-type linac structures, accelerates
the beam to near 71 MeV [11]. A 20° double-bend transport section
directs the beam into ATF Beamline III, where the VISA undulator is
located. This in-vacuum undulator [12] is a Halbach array of 220
periods, each 1.8 cm long. The undulator pole magnets provide an
on-axis peak field, B0 , of approximately 0.75 T. An additional periodic

quadrupole lattice provides intra-undulator strong focusing, reducing

an average electron beam b-function down to 30 cm in both planes.
The undulator structure is constructed out of 4 individual sections,
each 0.99 m long. Magnets alignment and shimming, performed at
BNL in tandem with the pulsed-wire measurements on each pair of the
consecutive undulator sections [13], provided a desired central
trajectory straightness of better than 40 µm in a single gain length
[14]. To align all 4 individual sections, an interferometric alignment
procedure is implemented [15]. The undulator vacuum chamber
provides an independent support of the magnet sections and allows
them to be manipulated while closed under vacuum [16]. A CCD
based optical monitoring system is used [17] to trace the relative
positions of the undulator sections throughout the pumping cycle with
an accuracy of 10-µm.

The undulator and vacuum chamber design includes 8 diagnostic

ports, 50 cm apart, starting 25 cm into the undulator. Their function
is to measure SASE radiation properties along the undulator, and to
characterize the electron beam position and envelope profile at 3
points throughout every betatron period. In-vacuum diagnostic probes
intercept the beam path inside the undulator. The active area of the
probe — a double-sided silicon mirror with l / 10 surface finish — could
be positioned to either reflect FEL light towards SASE diagnostics, or to
generate OTR for electron beam imaging [18]. Even though the
optical system allows measuring a beam position with an accuracy of
20-µm, the overall trajectory resolution is about 100 µm due to the
electron beam jitter. A relatively low intensity of the OTR, coupled
with the high optical noise environment inside the undulator, places a
limitation on the dynamic range of the beam profile monitors. An
alternative scheme of measuring beam profile with much brighter
YAG:Ce scintillators was ruled out due to the systematic image
blooming observed during the initial tests [19].

The electron beam characteristics are initially measured at the

linac exit. A projected emittance, e n , is determined with a quadrupole

scanning technique. To find thebeam current, the 2nd linac section is

re-phased to induce a known linear energy chirp to the electron beam;
allowing determination of the bunch length from the dispersion-
dominated horizontal beam profile after the 20° bend. FEL
measurements, however, indicate a gain considerably larger than one
would expect using the electron beam parameters measured after the
linac (Table 1). In addition, the condition of maximum gain is
accompanied by changes in the SASE radiation spectrum, where the
spiky structure is replaced with a smooth peak (Fig. 1). Lower
numbers of SASE spectral spikes points towards shorter bunch length
[20], at least when the linac RF phase is detuned to optimize SASE
gain. A possibility of bunch compression was recognized during the
preliminary analysis of the longitudinal beam dynamics in Beamline III
[21].

To characterize the electron beam pulse compression during

transport, a coherent transition radiation (CTR) diagnostic was
implemented. A single bolometer for sub-mm wavelengths (Golay
cell) was used to measure the CTR intensity emitted as the electron
beam passes through a 45°-mirror. The spectral energy of the CTR
signal is related to the Fourier transform of the current profile,
E CTR (w ) µ Ib 2 (w ) . Hence, the measured signal intensity is inversely

proportional to the electron beam RMS bunch length s t [22], with the

sharp cut-off at s t ~ 2 ps due to the Golay cell sensitivity curve.

Before being focused into the detector window, a collimated CTR beam
is reflected off a vertical wire-grid polarizer to eliminate both optical
noise and horizontally polarized coherent undulator radiation. Strong
compression was observed, as the measured CTR intensity (Fig. 2)
displayed a sharp peak within a narrow (<2°) window of the linac RF
phase. In addition, an insertable low-pass filter with a known
frequency roll-off was employed to allow quantitative assessment of
the high frequency CTR component. Measurements with and without
the filter gave a minimum signal ratio of 0.68, indicating a short
bunch-length (FWHM<0.4 ps) and providing a quantitative benchmark
for the numerical modeling of the system. The results indicate that
electron beam can be compressed by as much as a factor of 5, yielding
the peak current of 250 A, consistent with the SASE spectra
measurements. The high efficiency of the compression is due to a
very large (negative) second order coefficient T566 in the dispersive
section of the Beamline III, which enhances the linear compression as
the linac RF phase is detuned by changing the effective R56 [23].

Since the non-linear effects during the beam transport are

critical for understanding the electron beam properties in the
undulator, a full cathode-to-undulator simulation of the experiment is
needed to model the SASE results. Starting from the photoemission, a
UCLA version of PARMELA [24] is used to model acceleration and
emittance compensation through the gun and linac sections. When
PARMELA results match the beam properties measured after the linac,
the code ELEGANT [25] is used to model the beam dynamics and
longitudinal bunch compression in the dispersive section, including the
effects of coherent synchrotron radiation (CSR) in the bends. The
beam compression was simulated as a function of the linac RF phase
detuning; having ELEGANT output tested against the Golay cell
measurements. Good agreement between a model and the data was
achieved over a wide range of system parameters (Fig. 2). One
critical result of the ELEGANT simulations is the emittance growth in
the dispersive section. It is found that the slice horizontal emittance
growth due to CSR is not significant, i.e., D e CSR
n ª 0.3 µm-rad.

However, at maximum compression point, a beam energy offset

generates a significant residual dispersion term, h ~ 5 cm. Combined

with a large slice energy spread in the compressed beam, dispersion

dominates the emittance growth, increasing the effective horizontal
slice emittance up to e neff ª 3.5 µm-rad.

Due to the strong sensitivity of the compression to small

changes in the linac phase, the SASE signal fluctuations are
dominated by the system RF jitter. To enable data reduction for the
FEL measurements, a collimator was used in the dispersive section to
trace a linac RF phase evolution, by monitoring charge loss, which can
be well correlated with the RF phase. Figure 3 shows typical shot-to-
shot SASE intensity distribution as a function of charge in the presence
of the collimator. On the right hand side of the plot, the linac operates
near nominal phase, and all the charge is transmitted through the
collimator; however, as there is no compression the overall gain is
small. As the linac phase advances, the electron beam undergoes
compression, which improves the FEL gain, while the charge loss in the
collimator becomes significant. At the optimal operating point the
charge loss is around 30%; yet the gain does not degrade, as most of
the scraping removes a non-lasing tail of the beam. Further linac
phase increase reduces the SASE gain due to the stronger residual
dispersion, overcompression, and the charge loss at the collimator. In
the case of uncollimated electron beam, the shot-to-shot charge
fluctuations out of the gun are well contained within 5%.

With the shot-to-shot tracking of the linac phase through charge,

it is possible to separate SASE statistics from the system fluctuations,
by selecting the data points within a narrow charge window (see Fig.
3). In addition, the same data selection criteria enable valid
comparison of different data sets, allowing a systematic evaluation of
SASE gain along the undulator length. For these measurements, a
periodic lens array is used to multiplex between the diagnostic ports.
The SASE radiation beam from all 8 ports can be directed into a
factory calibrated detector equipped with a set of ND filters to increase
the system dynamic range. The intensity of the SASE signal
noticeably exceeds the one of the broadband spontaneous emission
(SE) starting from port 3 (1.25 m into the undulator).

Figure 4 shows SASE energy measurements along the undulator

length. The data points on the plot at a given location represent the
RMS value of the energy distribution selected within the charge
window 125-155 pC. The measurements demonstrate a uniform
exponential amplification with a power gain length L˜g = 17.9 cm. The

two final points deviate from the exponential behavior, indicating

saturation. More evidence of saturation comes from the statistical
properties of the SASE signal (Fig. 5). Shot-to-shot measurements of
the SASE-FEL energy inside the undulator can be fit with a negative
exponential distribution associated with single spike lasing [26]. At
the undulator exit the measured statistical distribution differs
dramatically, and is dominated by shot-to-shot fluctuations of the
saturation length, which can be as large as few gain lengths [20] for a
single spike lasing. Hence, instead of the characteristic flattening of
the intensity curve obtained in a single simulation of SASE process,
one still sees a significant growth of multi-shot RMS energy in the final
25 cm of the undulator. The RMS measure is chosen as a robust
representation of a negative exponential distribution observed during
the exponential gain; it becomes less meaningful at saturation.

Based on the energy measurements, given a single spike

temporal radiation profile with the line bandwidth (D w w ) f ª 1.2% , it is

possible to estimate SASE power at saturation Pf ª 80 MW . This is in

good agreement with the theoretical prediction [1,27]

Psat ª Pbeamr˜ = 85 MW where the 3-D FEL parameter r˜ = 0.0046 is obtained

from the measured gain length. The theoretical estimate of the FEL
saturation length, Lsat ª lu r˜ = 3.9 m , is also in agreement with the

experimental data. The final state SASE spectral power gain

G ≡ ( dP dw ) f (dP dw )noise can be evaluated based on the effective shot
noise power treatment given in [28],

2
È ˘
-1 1 Ê Dw ˆ I p ekr Í K[ JJ ] 2 L˜g ˙
G ª Qg
Î (
Pf Ë w ¯ f e0 Í 2 + K 2 l u) ˙
˚

where Q ~ 1.3 mrad is the effective angular spread of a single FEL

guided mode, K = 1.26 is the undulator parameter, and the peak
current I p ª 250 A is obtained from the CTR data and PARMELA-
8
ELEGANT simulations. The result is G = 1.8 ¥ 10 .

To simulate the SASE process numerically, a PARMELA-ELEGANT

output file is converted into the input file for the 3-D time dependent
code GENESIS 1.3 [29], preserving all 6-D properties of the electron
beam phase space. Once detuning of the linac RF phase and bunch
compression is included into the ELEGANT model, GENESIS output
accurately reproduces the measured FEL gain evolution (Fig. 4), as
well as spectral and angular properties of the radiation. On Figure 6
the measured angular profile of SASE intensity is shown along with the
far field distribution generated from a GENESIS output. The hollow
profile of the SASE radiation cone, repeatedly observed
experimentally, is linked to strong asymmetry in the energy-correlated
horizontal phase space distribution of the compressed electron beam.
Reproduction of such an unusual result with GENESIS is an important
benchmark in development of a start-to-end computational model of
the experiment.

In conclusion, the VISA experiment presents an extensive study

of the dynamics of a high gain SASE FEL driven by a high brightness
electron beam in a strong focusing, compact undulator. A gain of
2¥108 and saturation were obtained, and the statistical properties of
SASE signal at saturation were measured. To explain the performance
of the system, it was necessary to investigate, both experimentally
and through numerical modeling, the dynamics of the electron beam in
the entire injector-transport-FEL system. A novel bunch compression
mechanism was identified and studied. For the first time a start-to-
end integrated numerical model was developed to compare high gain
SASE data with theory. The results of simulations are in excellent
agreement with the measurements, and give strong insights into the
physical mechanisms of the complex phenomena present in the
experimental system.
 CAPTIONS

TABLE 1. Basic properties of the electron beam at VISA.

FIG. 1. Comparison of the FEL radiation spectra measured at low and high gain. SASE
gain improvement was accompanied by transformation of characteristic multi-spike
pattern into a red-shifted single spectral spike.

FIG. 2. Measured Golay cell CTR signal as a function of linac phase, compared to the
calculated value derived from PARMELA/ELEGANT model, including Golay cell
response.

FIG. 3. Shot-to-shot measurement of SASE radiation intensity versus charge Q, 25 cm

before the undulator exit, in the presence of the collimator. Within the narrow charge
window (dashed lines) the RF phase is quasi-stable and SASE statistical analysis is
possible.

FIG. 4. RMS SASE energy along the undulator length, for a charge window centered at
140 pC, compared to the GENESIS simulations (RMS fluctuations boundaries of
simulated values due to a shot-noise are shown with gray lines). The exponential part of
the experimental curve yields a power gain lengths of 17.9 cm. The two final points
indicate system saturation near the undulator exit.

FIG. 5. Statistical properties of the SASE radiation at diagnostic port #7 (3.25 m into the
undulator) and at the exit. Measurements within the undulator are fitted with a negative
exponential, as expected from a random single lasing spike (short bunch limit). The
intensity distribution changes at the undulator exit, at saturation.

FIG. 6. Far field angular distribution of SASE signal at high gain, but not in saturation (a)
measured with the CCD camera and (b) simulated by GENESIS.
Electron Beam Energy ~ 71 MeV
Beam Charge at Linac Exit, Q 200 – 300 pC
Normalized Emittance at Linac Exit, e n 1.5 – 2.3 µm-
Peak Current at the Linac 55 ±rad
5 Amp
Peak Current at the Undulator, I p > 250 Amp
RMS Beam Energy Chirp at High 0.17 ± .03 %
Gain, D g / g
Table 1
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
[1] R. Bonifacio, C. Pellegrini, and L. Narducci, Opt. Commun. 50, 373 (1984).

[2] Linear Collider Conceptual Design Report, DESY 97-48 (1997); LCLS Design Study Report,
SLAC-R-521 (1998).

[3] M. Hogan et al., Phys. Rev. Lett. 80, 289 (1998); 81, 4867 (1998); D.C. Nguyen et al., ibid.
81, 810 (1998)

[4] L.H. Yu et al., Nucl. Instrum. Methods A 445, 301 (1999); Proceeds. FEL 2000, Duke
University (2000).

[5] M. Babzien et al., Phys. Rev. E 57, 6093 (1998).

[6] S. V. Milton et al., Phys. Rev. Lett. 85, 988 (2000); Science 292, 2037 (2001).

[7] J. Andruszkow et al., Phys. Rev. Lett. 85, 3825 (2000); more recently this group has achieved
SASE saturation at 98 nm (private communication).

[8] A. Tremaine, presented at FEL 2001.

[9] X. Wang, X. Qiu, I. Ben-Zvi, Phys. Rev. E 54, 3121 (1996).

[10] A. Tremaine et al., submitted to Phys. Rev. Lett.

[11] X. Wang et al., Proceeds. PAC 99, p. 3495; V. Yakimenko et al, presented at the FEL 2001.

[12] R. Carr et al., Phys. Rev. ST AB 4, 122402 (2001).

[13] G. Rakowsky et al., Proceeds. PAC 99, 2698 (1999).

[14] P. Emma, H.D. Nuhn, Proceeds. FEL 98 II, 35 (1999).

[15] R. Ruland et al., Proceeds. PAC 99, 1390 (1999).

[16] M. Libkind et al., Proceeds. PAC 99, 2477 (1999).

[17] A. Tremaine, A. Murokh, and X.J. Wang, BNL Report No. 68170, 2001 (unpublished).

[18] A. Murokh et al., Proceeds. FEL 2000 (2001).

[19] A. Murokh et al., presented at PAC 2001.

[20] R. Bonifacio et al., Phys. Rev. Lett. 73, 70 (1994).

[21] J.B. Rosenzweig (private communication).

[22] A. Murokh et al., Nucl. Instrum. Methods A 410, 452 (1998).

[23] J.B. Rosenzweig, E. Colby, and N. Barov, IEEE Trans. Plasma Sci. 24, 409 (1996).

[24] E.R. Colby, Ph.D. thesis, University of California, Los Angeles, 1997.

[25] M. Borland, Advanced Photon Source Report LS-287, 2000 (unpublished).

[26] E.L. Saldin, E.A. Shneidmiller and M.V. Yurkov, Opt. Comm. 148, 383 (1998).

[27] K.J. Kim, Phys. Rev. Lett. 57, 1871 (1986).

[28] L.H. Yu and S. Krinsky, Nucl. Instrum. Methods A 285, 119 (1989).

[29] S. Reiche, Nucl. Instrum. Methods. A 429, 243 (1999).