Work supported in part by US Department of Energy contract DE-AC02-76SF00515.
First
demonstration of the echo-enabled harmonic generation technique for
short-wavelength seeded free electron lasers
D. Xiang, E. Colby, M. Dunning, S. Gilevich, C. Hast, K. Jobe, D. McCormick, J. Nelson,
T.O. Raubenheimer, K. Soong, G. Stupakov, Z. Szalata, D. Walz, S. Weathersby, M. Woodley
SLAC National Accelerator Laboratory, Menlo Park, CA, 94025, USA
P-L. Pernet
´
Ecole Polytechnique F´ed´erale de Lausanne, Lausanne, Switzerland
(Dated: July 17, 2010)
We report the first experimental demonstration of the echo-enabled harmonic generation (EEHG)
technique which holds great promise for generation of high power, fully coherent short-wavelength
radiation. In this experiment, coherent radiation at the 3rd and 4th harmonic of the second seed
laser is generated from the so-called b eam echo effect. The experiment confirms the physics behind
this technique and paves the way for applying the EEHG technique for seeded x-ray free electron
lasers.
PACS numbers: 41.60.Cr
Free electron lasers (FELs) can provide high-intensity
coherent short-wavelength radiation which is essential for
the studies of molecular and atomic dynamics. In the x-
ray wavelength, the two leading concepts are the self-
amplified spontaneous emission (SASE) configuration
[1, 2] and the high-gain harmonic generation (HGHG)
scheme [3, 4]. One FEL working in the SASE mode
has been successfully operated at hard x-ray wavelengths
[5]. While the radiation from a SASE FEL has excellent
transverse coherence, it typically has rather limited tem-
poral coherence because a SASE FEL starts from elec-
tron beam shot noise. Alternatively, the HGHG scheme
allows the generation of fully coherent radiation by up-
converting the frequency of a high-p ower seed laser. How-
ever, due to the relatively low up-frequency conversion
efficiency, multiple stages of HGHG FELs are needed [6]
in order to generate coherent x-rays from a UV laser.
The up-frequency conversion efficiency can be greatly
improved with the recently proposed echo-enabled har-
monic generation (EEHG) technique [7, 8]. In the EEHG
scheme, the beam is energy modulated by a laser with
wave number k
1
in the first modulator and then sent
through a chicane with strong dispersion after which the
modulation obtained in the first modulator is macroscop-
ically smeared. Simultaneously, complicated fine struc-
tures are introduced into the phase space of the beam.
A second laser with wave number k
2
is used to further
modulate the beam energy in the second modulator. Af-
ter passing through a second chicane the echo signal then
occurs at the wave number k
E
= nk
1
+ mk
2
as a reco-
herence effect, where n and m are integers. The main
advantage of EEHG is that the bunching factor is a very
slowly decaying function of the harmonic numb er, thus
allowing the generation of coherent soft x-ray radiation
directly from a UV seed laser in a single stage.
The remarkable up-frequency conversion efficiency of
the EEHG technique has stimulated world-wide interest
in using EEHG to achieve fully coherent radiation in the
x-ray wavelength from UV seed lasers [9–11]. While sig-
nificantly relaxing the requirements on laser power and
beam slice energy spread as compared to the HGHG
scheme, EEHG requires more challenging control of the
beam dynamics as the beam goes through the undulators
and chicanes, because it involves a long-term memory of
the beam phase space correlations.
In this Letter we report the first exp erimental demon-
stration of the EEHG technique at the Next Linear Col-
lider Test Accelerator (NLCTA) at SLAC. The experi-
ment at NLCTA [12] is a generic EEHG proof-of-principle
experiment that uses two different lasers (the first laser
wavelength is λ
1
= 795 nm and the second is λ
2
= 1590
nm) in the modulators. The echo signals generated by the
interplay of the two lasers at the 3rd and 4th harmonic
of the second laser were observed and clearly differen-
tiated from those generated by the lasers individually.
The dependence of the echo signal wavelength as a func-
tion of beam energy chirp was measured and found to be
in go od agreement with theory. The experiment verifies
the EEHG theory and indicates that scaling to short-
wavelength echo-seeded x-ray FELs may be possible.
The design considerations, commissioning and the ex-
perimental details at the NLCTA are discussed in [13, 14].
The layout of the experiment is schematically shown in
Fig. 1. The experiment is based on the existing NLCTA
facilities where the electron beam is generated in a 1.6
cell S-band photocathode rf gun with a 0.5 ps frequency-
tripled (266 nm) Ti:Sapphire laser and is boosted to 60
MeV in an x-band linac (X1). A flexible chicane (C-1)
which can provide either positive or negative momentum
compaction is available to manipulate the beam’s longitu-
dinal phase space. In this experiment it is set to provide
a p ositive momentum compaction of R
56
= 76 mm. The
facility can provide 20∼40 pC electron bunches with a
normalized emittance of about 8 mm·mrad.
In order to conduct the EEHG experiment in the
NLCTA tunnel, another x-band structure (X2) was
added downstream of C-1 to further accelerate the beam
to 120 MeV. The main components of the EEHG beam
SLAC-PUB-14199
2
line consist of 3 chicanes (C0, C1 and C2), 3 undulators
(U1, U2 and U3), several insertable optical transition
radiation (OTR) screens for electron and laser beam po-
sition and size measurements, and quadrupoles for beam
matching and focusing.
The mini-chicane (C0) is used to generate an orbit
bump to allow laser injection into the first undulator
(U1). The beam is modulated by the 795 nm laser (1
ps FWHM, Ti:Sapphire laser) in U1 (10 periods with a
period of 3.3 cm and a K value of 1.82). After pass-
ing through C1 complicated fine structures are intro-
duced into the beam longitudinal phase space. The
beam is again modulated by the 1590 nm laser (0.7 ps
FWHM, produced by an optical parametric amplifier sys-
tem pumped with the 795 nm laser) in the second undu-
lator U2 (10 periods with a period of 5.5 cm and a K
value of 2.09). Finally density modulations at shorter
wavelength are generated after passage through C2. The
density-modulated beam will produce coherent radiation
in U3 (10 periods with a period length of 2 cm) which is
presently tuned to 318 nm. The radiation generated in
U3 is reflected by a downstream OTR screen to a spec-
trometer. The undulators U1 and U2 were built by the
STI Optronics company and U3 was built by LBNL.
FIG. 1: Schematic of the EEHG experiment at SLAC
The beam-laser interaction is achieved when electron
and laser beam overlap both spatially and temporally in
the modulators. The spatial overlap is achieved by steer-
ing the laser to the same position as the beam on the
OTR screens upstream and downstream of the undula-
tors. In the experiment the laser size is about two times
larger than that of the electron beam [14] to provide uni-
form modulation in transverse direction. An OTR screen
immediately downstream of each undulator can be in-
serted to reflect out the laser and undulator radiation for
temporal synchronization. The radiation is detected by a
fast photodiode (2 GHz bandwidth) and is analyzed with
an oscilloscope. By referencing the signals to an exter-
nal trigger, the laser and beam can be synchronized to
within approximately 30 ps. More precise timing is done
by using a scanning delay stage and measuring the co-
herent radiation enhancement that is produced when the
beam is energy modulated by each modulator and further
bunched by the subsequent dispersive section [14].
After setting the delay stages such that the two lasers
interact with the electron beam simultaneously, the
EEHG experiment was set up to generate the 3rd and 4th
harmonics of the second laser. The beam was accelerated
about 10 degrees off crest in X1 so that a positive energy
chirp is imparted in X1 and the beam is decompressed to
about 2.5 ps rms after C-1. The momentum compaction
of chicane C1 was set to R
(1)
56
= 4.9 mm and that of chi-
cane C2 was set to R
(2)
56
= 2.8 mm. The rf phase of X2
was adjusted to cancel the energy chirp imparted in X1
so that the beam had a minimal global energy spread at
the entrance to C0. This also minimized the energy chirp
in the beam. The radiation generated in U3 is reflected
out by the OTR screen and guided into a spectrometer
which consists of a 300 lines/mm grating, an optical lens
and a CCD detector. The wavelength calibration in the
measurement is achieved using the incoherent undulator
radiation which has a continuous spectrum, together with
two bandpass filters (one at 395 nm and the other at 531
nm; both have a bandwidth of 11 nm). With this setup
the radiation wavelength from 350 nm to about 600 nm
can be measured in a single shot. While the resolving
power of the grating is about 1-2 nm, the relatively large
source size limits the resolution of our measurement sys-
tem to about 4 nm.
FIG. 2: Spectrum of the radiation at the exit of U3 when
b eam energy chirp is minimized:(a) only 1590 nm laser is on;
(b) only 795 nm laser is on; (c) both lasers are on.
Results are shown in Fig. 2 when the beam energy chirp
is minimized. The spectrum of the coherent radiation is
broadened in our experiment due to the relatively large
residual chirp from the X-band linac. Henceforth, only
the center wavelength of the coherent radiation is quoted.
In the experiment the 1590 nm laser was first turned on
and its 3rd and 4th harmonics were observed (not shown
in Fig. 2) at the CCD when the laser power is high enough
to generate an energy modulation amplitude much larger
than the beam slice energy spread. The 1590 nm laser
power was then reduced until its 3rd and 4th harmonics
disappeared. The measured radiation spectrum is shown
in Fig. 2a. Then the 795 nm laser was also turned on and
its power was set to provide a peak energy modulation
of about 20-30 keV in U1. After finely adjusting the
795 nm laser timing, the 3rd harmonic of the 1590 nm
laser with a center wavelength at about 530 nm (E0 in
Fig. 2c) was observed (see Fig. 2c). Note the radiation at
530 nm can not be generated by the 795 nm laser alone,
because it is not its harmonic. It is not generated by
3
the 1590 nm laser alone either, because the laser power
is significantly reduced to provide a very small energy
modulation. Therefore, the radiation E0 at about 530
nm when both lasers are turned on is believed to be the
echo signal generated by the interplay of the two lasers
with the set n = −1, m = 5. To further confirm this, the
1590 nm laser was turned off, and the measured radiation
spectrum in Fig. 2b shows only the wavelength around
400 nm (H2), which is the 2nd harmonic of the 795 nm
laser, is enhanced.
It should be pointed out that an echo signal (n =
−1, m = 6) at about 400 nm might be generated in this
setup as well. Since it has the same wavelength as the
H2 signal, they both appear at the same position on the
CCD and are difficult to separate. In order to distin-
guish them, we repeated the exp eriment with an energy-
chirped beam. Analysis shows that with sufficient energy
chirp in the beam, the echo signal and that generated by
the lasers individually will shift to different wavelengths
and may be separated with the spectrometer.
It has been shown in [15] that with an energy chirp in
the beam, the harmonic radiation wavelength generated
by a single laser will shift to λ = λ
0
/C, where λ
0
is
the radiation wavelength when the beam has a vanishing
energy chirp, C = 1/(1 +hR
56
) is the compression factor
of the chicane and h = dδ/dz is the energy chirp factor at
the entrance to the chicane. The EEHG theory predicts
a quite different wavelength dependence [16]. With an
energy chirp h, the wave number of the echo signal will
shift to,
k
E
(h) =
nk
1
+ (1 + hR
(1)
56
)mk
2
1 + h(R
(1)
56
+ R
(2)
56
)
. (1)
To verify this prediction, we adjusted the phase of X2
to imprint considerable positive energy chirp on the elec-
tron beam. As shown in Fig. 3a, due to the energy chirp
and the bunch decompression in C2, the 4th harmonic of
the 1590 nm laser (H4) shifted from 397.5 nm to about
430 nm, and the 3rd harmonic of the 1590 nm laser (H3)
shifted from 530 nm to about 570 nm. Here the 1590 nm
laser power is increased to provide an energy modulation
amplitude of about 10 keV in the beam so that the har-
monic radiation generated by the 1590 nm laser alone can
be clearly seen. The radiation is measured downstream
of U3, so the compression factor for the modulation gen-
erated in U2 is C
U 2
= (1 + hR
(1)
56
)/(1 + h(R
(1)
56
+ R
(2)
56
)),
where h is the chirp at the entrance to chicane C1. The
compression factor is C
U 2
= 0.9244 and the chirp is in-
ferred to be h = 33.4 m
−1
.
The modulation generated in U1 will be decompressed
in both C1 and C2, and the corresponding compression
factor is C
U 1
= 1/(1 + h(R
(1)
56
+ R
(2)
56
)). With this chirp
the compression factor for the modulation induced by
the 795 nm laser is found to be C
U 1
= 0.7937. Accord-
ingly, the wavelength for the 2nd harmonic of the 795 nm
laser (H2) should shift from 397.5 nm to 501 nm while it
was measured at about 499 nm, which is well within the
FIG. 3: Spectrum of the radiation at the exit of U3 when
b eam has considerable energy chirp:(a) only 1590 nm laser is
on; (b) only 795 nm laser is on; (c) both lasers are on.
experimental resolution (Fig. 3b).
350 400 450 500 550 600
0
0.02
0.04
0.06
0.08
λ (nm)
bunching
E1
E2
M1
H4
M2
E3
H2
H3
FIG. 4: Bunching at various wavelengths from simulation.
When both lasers are turned on, 3 additional echo sig-
nals with different wavelengths E1, E2 and E3 are ob-
served (Fig. 3c). The simulated bunching factors for var-
ious wavelengths with this chirp are shown in Fig. 4. The
beam typically has a slice energy spread of about 1 keV
and was decompressed by about a factor of 4 in chicane
C-1. Consequently, a slice energy spread of 0.25 keV
is assumed in the simulation. The energy modulation
amplitudes in U1 and U2 are assumed to be 21 and 10
keV, which is consistent with the experimental values.
With these specific parameters, simulation predicts that
enhanced coherent radiation at 8 wavelengths may be
generated in our experiment.
Analysis shows that the echo signal E1 in Fig. 3 and
Fig. 4 is generated with the set n = −5, m = 4. It follows
from Eq. (1) that with this chirp the wavelength of E1
should shift from 397.5 nm to 374 nm and experimentally
it was shifted to about 373 nm. Similarly the echo signal
E2 is found to be generated with the set n = −1, m = 6.
The wavelength of E2 should shift from 397.5 nm to 402
4
nm and was measured at 405 nm. E3 is generated with
the set n = 1, m = 2. The wavelength of E3 should shift
from 530 nm to 462 nm and was measured at about 458
nm. Considering the resolution of the spectromter and
some uncertainties of the beam chirp, the experimental
results are in good agreement with the EEHG theory.
It should be pointed out that the simulated signals M1
(414 nm) and M2 (444 nm), are however, missing in our
experiment. From simulation we found that the strength
of the spectral lines are sensitive functions of modulation
amplitudes, which are not well known in the experiment;
this may account for the discrepancy. Nevertheless, all
the observed spectral lines in the experiment can be well
explained with the EEHG theory.
30 35 40 45 50 55
400
450
500
550
Energy chirp h (m
−1
)
Wavelength (nm)
FIG. 5: Radiation wavelength vs beam energy chirp for the
echo signal E2 (red), E3 (magenta) and that from the 795 nm
laser alone H2 (blue).
To further benchmark the EEHG theory, the beam en-
ergy chirp is varied and the wavelengths of the harmonic
radiation for E2, E3, H2 and H4 are measured and shown
in Fig. 5 (the E1 signal, being close to the cut-off wave-
length of the diagnostic system, was not observed in ev-
ery shot, and is not included in Fig. 5). The chirp factor
is inferred from the wavelength shift of H4. The blue
dots are the measured wavelength for the 2nd harmonic
of the 795 nm laser (H2) and the blue line is the the-
oretical prediction. The red and magenta dots are the
measured wavelength for the echo signal E2 and E3, and
the red and magenta lines are the prediction in Eq. (1).
The experimental results are in good agreement with the
theory. The results confirm that the harmonic radiation
wavelength generated by a single laser scales as the com-
pression factor and the sensitivity of the echo signal to
the beam chirp depends on n and m. Specifically the
echo signal generated with the set n = −1, m = 6 is
insensitive to the beam chirp.
It is worth mentioning that echo signals at even shorter
wavelength could be generated in this experiment setup
as well. However, the current diagnostics limit the mea-
surement to optical wavelengths because the transmis-
sion of the spectrometer lens drops rapidly below 380
nm. Currently the spectrometer is being upgraded with
a UV grade lens to explore the EEHG technique in the
shorter wavelength range.
In summary, we presented the first experimental
demonstration of the EEHG technique which allows far
higher harmonic numbers to be accessed as compared
to the normal limit arising from the beam slice energy
spread. The good agreement between the EEHG theory
and our exp erimental results confirms the physics behind
this technique and paves the way for applying the EEHG
technique for future seeded x-ray FELs.
We thank C. Adolphsen, K. Bane, J. Byrd, A. Chao,
J. Corlett, J. Cruz, Y. Ding, P. Emma, W. Fawley, J.
Frederico, J. Frisch, G. Hays, Z. Huang, C. Hudspeth, H.
Loos, H.-D. Nuhn, G. Penn, S. Prestemon, J. Qiang, M.
Racine, R. Schlueter, M. Venturini, W. Wan, F. Wang,
X.-J. Wang, W. White, J. Wu and A. Zholents for help-
ful discussions, comments and commissioning assistance.
This work was supported by the US DOE Office of Basic
Energy Sciences using the NLCTA facility which is partly
supported by US DOE Office of High Energy Physics un-
der Contract No. DE-AC02-76SF00515.
[1] A. Kondratenko and E. Saldin, Part. Accel. 10, 207
(1980).
[2] R. Bonifacio, C. Pellegrini, and L.M. Narducci, Opt.
Commun. 50, 373, (1984).
[3] L.-H. Yu, Phys. Rev. A, 44, 5178 (1991).
[4] L.-H. Yu, et al., Science, 289, 932 (2000).
[5] P. Emma, et al., Nature Photon., to be published, (2010).
[6] J. Wu and L. H. Yu, Nucl. Instrum. Methods Phys. Res.,
Sect. A 475, 104 (2001).
[7] G. Stupakov, Phys. Rev. Lett, 102, 074801 (2010).
[8] D. Xiang and G. Stupakov, Phys. Rev. ST-AB, 12,
030702 (2009).
[9] D. Xiang and G. Stupakov, in Proceedings of PAC 09
(IEEE, Vancouver, 2009).
[10] E. Allaria, et al., in Proceedings of FEL 09 (Liverpool,
2009), p.39.
[11] S. Reiche et al., in Proceedings of FEL 09 (Liverpool,
2009), p.51.
[12] “NLC test accelerator conceptual design report”, SLAC-
Rep ort-411, (1993).
[13] M. Dunning et al., in Proceedings of IPAC 10 (IEEE,
Kyoto, 2010), p.2293.
[14] D. Xiang et al., in Proceedings of IPAC 10 (IEEE, Kyoto,
2010), p.2299.
[15] T. Shaftan and L.-H. Yu, Phys. Rev. E 71, 046501 (2005).
[16] Z. Huang et al., in Proceedings of FEL 09 (Liverpool,
2009), p.127.
