Proceedings of 2005 Particle Accelerator Conference, Knoxville, Tennessee

HIGH ENERGY GAIN IFEL AT UCLA NEPTUNE LABORATORY

P. Musumeci∗ , S. Ya. Tochitsky, S. Boucher, A. Doyuran, R. J. England,

C. Joshi, C. Pellegrini, J. Ralph, J. B. Rosenzweig, C. Sung, G. Travish, R. Yoder,
University of California at Los Angeles, CA 90095, USA
A. Varfolomeev, S. Tolmachev, A. Varfolomeev Jr., T. Yarovoi, RRCKI, Moscow, Russia
Abstract Inverse Free Electron Laser has in fact experimentally
demonstrated microbunching [5], phase-dependent accel-
We report on the observation of energy gain in excess of
eration of electrons [6], phase locking and multi-stage ac-
20 MeV at the Inverse Free Electron Laser Accelerator ex-
celeration [7] and control of final energy spread [8]. De-
periment at the Neptune Laboratory at UCLA. A 14.5 MeV
spite such significant progress on a phase-controlled accel-
electron beam is injected in a 50 cm long undulator strongly
eration, in all these experiments the energy gain was mod-
tapered both in period and field amplitude. A CO2 10.6 µm
est and the measured acceleration gradient was well below
laser with power > 400 GW is used as the IFEL driver. The
the gradient attributed to the conventional rf accelerator.
Rayleigh range of the laser (∼ 1.8 cm) is shorter than the
It is of great importance to demonstrate both high en-
undulator length so that the interaction is diffraction dom-
ergy gain and acceleration gradient in a laser accelerator.
inated. Few per cent of the injected particles are trapped
This can be achieved not only by increasing the coupled
in stable accelerating buckets and electrons with energies
laser power, but also by increasing the effective interaction
up to 35 MeV are detected on the magnetic spectrometer.
length. The latter is obtained a) by extending the region
Three dimensional simulations are in good agreement with
where the laser is most intense (diffraction effects) and b)
the electron energy spectrums observed in the experiment
by increasing the length over which the particles and the
and indicate that substantial energy exchange between laser
laser wave maintain synchronism (tapering).
and electron beam only occurs in the first 25-30 cm of the
The inverse free-electron laser experiment at the Nep-
undulator. An energy gradient of > 70 MeV is inferred. In
tune laboratory at UCLA accelerated electrons from 14.5
the second section of the undulator higher harmonic IFEL
MeV up to more than 35 MeV utilizing a CO2 laser beam
interaction is observed.
with a peak power ( ∼ 0.4 TW ), one order of magnitude
greater than any other previous IFEL experiment had used
INTRODUCTION [16]. To maintain the resonant condition with the acceler-
Laser acceleration of charged particles in vacuum is of ating electrons, the 50 cm long undulator is strongly ta-
great interest since the electric field of a laser at its focus pered both in period and magnetic field amplitude. An
can exceed by many orders of magnitude the average ac- important point of the Neptune IFEL configuration is that
celeration gradient for a conventional radiofrequency ac- the Rayleigh range of the laser beam is much shorter than
celerator (e.g. 50 MeV/m for a S-band accelerator), which the undulator length and the IFEL interaction is diffraction
is fundamentally limited by the field emission on the cop- dominated. We also report about the observation of higher
per walls. In addition laser acceleration allows control of harmonic IFEL interaction manifested in a modulation of
the electron bunch length on the scale of the laser wave- the output electron energy spectrum.
length λ. Among all proposed concepts for laser accelera-
tion, the Inverse Free Electron Laser (IFEL) scheme is one EXPERIMENTAL SETUP
of the most mature. In an IFEL [1, 2], relativistic particles
In Table 1 we report the design parameters of the IFEL
are moving through an undulator magnet; a laser beam is
experiment at the Neptune Laboratory.
propagating parallel to the electron beam. The undulator
In Fig. 1 it is shown the experimental layout for the IFEL
produces a small transverse velocity in a direction paral-
experiment. An electron beam of 300 pC at 14.5 MeV is
lel to the electric vector of the electromagnetic wave so
that energy can be transferred between the particles and
the wave. Efficient energy exchange takes place when the Final focus
To streak camera
electron phase slippage is such that its transverse velocity Phosphorous screen
quadrupoles
CO2 beam
changes sign synchronously with the laser field (resonant NaCl lens
condition).
Recent successful proof-of-principle IFEL experiments E-beam
have shown that along with acceleration [3, 4] this scheme
offers the possibility to manipulate and control the longi- Kurchatov undulator mirror with hole
tudinal phase space of the output beam at the laser wave- Electron spectrometer
length. First among other laser accelerator schemes, the
0 Now at INFN - Universita’ di Roma La Sapienza Figure 1: Layout of Neptune IFEL experiment.

c
0-7803-8859-3/05/$20.00 2005 IEEE 500
 Proceedings of 2005 Particle Accelerator Conference, Knoxville, Tennessee

24

IFEL Output energy (MeV)

Table 1: Electron beam and CO2 Laser parameters at the 45
Neptune Laboratory

Amplitude (a. u.)

22
Electron beam
30
energy 14.5 MeV 20
charge 0.3 nC
emittance 5 mm-mrad 18 15
pulse length (rms) 6 ps
σrms at focus 120 µm 16
CO2 laser 0
-200 -100 0 100
power 400 GW
Relative timing (ps)
wavelength 10.6 µm
pulse length (FWHM) 240 ps
spot size (1/e2 ) 240 µm Figure 2: IFEL output energy vs. measured delay on the
streak camera between reference e-beam pulse and CO2
delivered to the experimental region by the Neptune rf pho- laser. Also shown is a typical streak of the CO2 laser pulse.
toinjector [9]. Final focus quadrupoles with large aperture
to avoid clipping of the copropagating laser beam were in-
stalled on the beamline and focused the electron beam to than the laser beam. The accelerator is virtually turned on
the nominal spot size of 120 µm rms in the middle of the only at the time the laser pulse goes through the undulator.
undulator. A TW-class CO2 laser system [10] is used to The ponderomotive gradient that accelerates the electrons
drive the IFEL. The high power laser beam is brought in depends on the instantaneous power felt by the electrons.
vacuum through a NaCl lens that has both the function of At the Neptune laboratory, the electron pulse ( ∼ 15 ps
producing the correct focusing geometry and serves as a FWHM) and the longer CO2 laser pulse (∼ 240 ps FWHM)
vacuum window. The laser is made collinear to the e-beam have been synchronized in the past with a cross-correlation
utilizing a plane copper mirror with a hole. The beams are method that was based on the e-beam controlled transmis-
aligned on a detection screen in the midplane of the undu- sion of CO2 through a Germanium sample. This technique
lator with an accuracy of < 100 µm and the angular mis- [15] constituted the first step of the synchronization proce-
alignment is kept below 1 mrad using the screens located dure also in the IFEL experiment. On the other hand, the
before and after the undulator. After the interaction region, cross-correlation measurement is conducted with the un-
the e-beam is energy analyzed by the magnetic spectrome- amplified laser pulse propagating through the final triple
ter and the laser beam is sent to a streak camera for timing pass, 2.5 m long multiatmosphere CO2 amplifier with no
measurements. inversion of population and has an intrinsic systematic er-
The IFEL hybrid planar undulator KIAE-2p [11, 12] was ror due to the different group velocity of the laser pulse
designed and built at the Kurchatov Institute. This magnet within the inverted medium of the final amplifier in com-
was optimized for a 400 GW, 3.6 cm Rayleigh range driver parison with no-gain conditions. Moreover, fluctuations in
laser beam. It has a very strong variation of the wiggling laser power (gain in the final amplifier) ± 50 % cause fluc-
period and magnetic field amplitude along the axis (see Ta- tuations in the time of arrival of ± 50 ps. This jitter is
ble 2), carefully tailored to maintain the resonance condi- intrinsic in the laser amplification system and could not be
tion of the IFEL interaction between the CO2 photons and eliminated.
the quickly accelerating electrons. Because the Rayleigh In order to get very accurate information on the relative
range is much shorter than the undulator length, the IFEL- timing between the electrons and the amplified pulse on
driving laser field along the undulator varies rapidly, and each shot, we set up a new streak camera based timing di-
it becomes important to properly consider the Guoy phase agnostics. The cathode of the streak camera is not sensitive
shift experienced by the laser near the focal point [13]. In to 10.6 µm photons so an optical gating technique, based
order to maintain synchronism through the focal region, on the Kerr effect, was used to gate a long red laser diode
two undulator sections were built with the field in the cen- with the CO2 pulse. This gated portion of red pulse is sent
tral region adjusted using a corrector Vanadium-Permandur on the entrance slit of a 10 ps resolution streak camera to-
plate. The undulator period and field strength were speci- gether with a small fraction of the photocathode driver laser
fied using simulation results [11] obtained with the three
dimensional code TREDI [14]. Table 2: RRCKI IFEL undulator parameters
Synchronization between electrons and photons is the
initial final
key for every laser accelerator with an externally injected
Undulator period 1.5 cm 5 cm
electron beam and it is even more delicate in an IFEL ex-
Magnetic field amplitude 0.16 T 0.65 T
periment. In this kind of interaction in fact, there is no
K 0.2 2.8
resonant medium (or cavity) where the accelerating elec-
Resonant energy 14.5 MeV 52 MeV
tromagnetic wave is resonating and can live independently

501 c
0-7803-8859-3/05/$20.00 2005 IEEE
 Proceedings of 2005 Particle Accelerator Conference, Knoxville, Tennessee

as reference for the electron beam timing. In Fig. 2, the

final maximum energy of the IFEL accelerator is shown
as a function of delay between the e-beam reference pulse
and the peak of the CO2 pulse, as measured on the streak
camera. A lineout of a typical CO2 pulse is also displayed
on the same time scale. Even for very small laser power
some interaction is observed (+150 ps data point). On the
other hand, the output energy is clearly maximized when
the electrons sample the peak laser intensity. This diagnos-
tic allowed us to determine for each laser shot the pulse
length –and thus the peak power– of the CO2 beam and
which part of the laser pulse intensity profile the electron
beam sampled with an accuracy of ± 10 ps. This measure-
ment was used to optimize the injection time of the elec-
trons in the IFEL accelerator.
The optical geometry used in the experiment to focus
and control transversely the laser beam size is of particular Figure 3: Measured and simulated output energy spectrums
importance. In the original design, a 2.56 m focal length for the Neptune IFEL accelerator experiment for two differ-
NaCl lens focused the laser in the middle of the undulator ent laser focus positions. The simulated spectra have been
to a spot size of 340 µm with a Rayleigh range of 3.6 cm to normalized to the same area as the measured ones.
increase as much as possible the extent of the region where
the beam is more intense. The resulting peak intensity is stream from the optimum point, the final energy decreases
2 · 1014 W/cm2 in the laser focus, about two orders of mag- because the effective interaction length is shortened.
nitude more than any previous IFEL experiment used [7, 8].
Dealing with this very high laser intensities on the other EXPERIMENTAL RESULTS AND
hand had some disadvantages. Experimentally, in fact, we
were limited by damage threshold on the last optical ele-
SIMULATIONS
ments of the CO2 transport line and we could not increase The electrons were detected by a phosphorous screen ob-
the f-number of the optical geometry as we planned. For served by CCD cameras. The screen was attached to the
our typical pulse lengths of 200 ps, we observed damage thin mylar window at the exit slit of the magnetic broad-
on the copper mirrors for fluences above 3J/cm2 and on the band spectrometer. Different neutral density filters were
single crystal NaCl optics for fluences above 2J/cm2 . In applied to each cameras to get unsaturated images of the
the end, we measured a spot size of 240 µm and Rayleigh output slit. A postprocessing application that takes into ac-
range of 1.8 cm that, respect to the original design geome- count the different filters and scales the horizontal axis of
try, implied a stronger variation of the beam size along the the images with energy, reconstructed the single shot spec-
interaction region, and with the same nominal focus po- trum of the electron beam out of the IFEL accelerator for
sition at the mid-point of the undulator, a larger and less each laser shot. In Fig. 3 we show two measured energy
intense beam at the entrance and exit of the undulator. spectra (histograms in Fig. 3) obtained with the laser fo-
To trap and accelerate particles along the design reso- cus at two different position, respectively 5 cm and 2 cm
nant orbit, the ponderomotive IFEL gradient generated by upstream of the undulator midpoint. The laser power for
the laser electric field has to match the designed tapering these shots was 350 GW and 400 GW, respectively. As dis-
gradient. If this is not the case, no trapping or acceleration cussed earlier, when focusing the laser beam earlier along
are possible. Because of the differences between nominal the undulator, less power is required to trap and acceler-
and measured Rayleigh range, it was found out necessary ate the electrons, but their final energy is lower because the
to move the laser focus upstream from the nominal position interaction length is reduced by the detrapping occurring
to increase the intensity at the undulator entrance above the when the laser intensity falls below the trapping threshold
trapping threshold and start the acceleration process early value.
in the undulator. Of course that would also cause the par- The energy spectrum in Fig. 3b shows more than 5 %
ticles to fall out of resonance soon after the mid point of of particles trapped and accelerated up to 35 MeV with
the undulator because of lack of enough ponderomotive a 150% energy gain. The dispersion δ = ∆x/∆p is not
force to sustain the acceleration along the designed orbit. constant along the measurement screen, and it grows from
Positioning the laser focus 2 cm (∼ one Rayleigh range) 1 cm/MeV for the lower energies end to 4 cm/MeV for
upstream of the undulator midpoint was found to give the higher energies. The resolution is sufficient to fully distin-
maximum final energy. When moving the focus towards guish different peaks in the high energy part of the spectra.
the undulator center, the intensity at the magnet entrance The camera noise level sets a minimum detectable charge
drops below the trapping threshold and there is no resonant density. Converting this threshold into the graph units, we
IFEL interaction. If the laser focus is moved further up- obtain a minimum detection threshold below 10 counts.

c
0-7803-8859-3/05/$20.00 2005 IEEE 502
 Proceedings of 2005 Particle Accelerator Conference, Knoxville, Tennessee

0 20 a. u. 40 60 damental radiation frequency but also with its harmonics

a) b) [18]. From an alternative point of view, particles of differ-
100 n=1 35 Experimental data
3D simulations
1D simulations
ent energies γr,n can interact with the same laser frequency,
80
30 because they see the EM wave as a higher harmonic of the
n=2 fundamental frequency which they are resonant with. In
γ

60
n=3 25 other words, for a given laser wavelength and undulator
40
magnet, there are multiple resonant energies
20
0.0 0.1 0.2 0.3 0.4 0.5
20
0.43976 0.43980 0.43984


Distance along the undulator (m) z(m) λw (1 + K 2 /2)
γr,n = (1)
2nλ
Figure 4: Simulations of IFEL experiment. (a) Maximum where λw is the undulator period, K is the normalized
energy particle evolution along the undulator. The first, undulator amplitude and λ the radiation wavelength. The
second and third harmonic resonant curved are also shown. higher harmonic IFEL interaction can be viewed as the nat-
(b) Longitudinal phase space at 45 cm with the energy ural consequence of the fact that the spectrum of the spon-
modulation due to the higher harmonic interaction taneous radiation emitted by a relativistic electron passing
through a planar undulator magnet presents peaks at differ-
Adjusting the laser intensity distribution along the un- ent frequencies that are harmonics of a fundamental reso-
dulator to the experimental configuration, the simulation nant frequency. In a quantum mechanic description of the
results agree quite well with the experiment. The code interaction, because of the presence of strong resonant lines
that was used to simulate the experiment is a 4th-order in the spontaneous radiation spectrum, stimulated emission
Runge-Kutta integrator that solved the Lorentz equation for (FEL emission at higher harmonics) or absorption (higher
the particle motion in the combined fields of the undulator harmonic IFEL) of radiation at these frequencies is possi-
magnet and laser beam (TREDI[14]). ble.
In Fig. 4a we show the evolution of the maximum e- To explain the structure of the high energy side of the ex-
beam energy along the accelerator for the case where the perimental spectrum in the Neptune IFEL experiment, we
laser focus is located 2 cm upstream of the undulator mid- consider the evolution of the maximum-energy particle in
point (solid red line in Fig. 4a). The IFEL acceleration the simulations represented in Fig. 4a. A few cm after the
mostly takes place in the first section of the undulator (first undulator center, the particles fall off the resonance curve
25-30 cm). This gives an accelerating gradient larger than (first harmonic resonance, n= 1 dashed curve in Fig. 4a)
70 MeV/m in the first half of the undulator. Few cm after because the driving laser intensity has decreased below the
the mid-point the laser intensity has decreased below the trapping threshold.
√ Some distance later, their energy is a
trapping threshold and the designed tapering is too strong factor of 2 less than the first harmonic resonant energy
for the particles too follow. at that point of the undulator, and therefore the electrons
Fig. 4b shows the simulated longitudinal phase-space at are resonant with the n=2 dotted curve and can exchange
45 cm along the undulator. The solid red line on the left energy with the 10.6 µm photons via the second harmonic
of the graph is the projection of the phase-space on the en- IFEL interaction. Later on, the electrons have energies such
ergy axis. Also shown is the measured experimental spec- that energy exchange with the laser can start even via the
trum (cyan histogram). The energy modulation appearing stronger third harmonic coupling (n = 3 dash-dotted curve).
in both the simulated and experimental energy spectra is It is known that the on-axis spectral content of the pla-
particularly interesting because the spectrum predicted by nar undulator radiation presents peaks only at the odd har-
the solution of the 1D FEL-like equations does not show monics. The even harmonics in fact, are suppressed by the
any peaks (blue dotted line). The high energy side of the symmetry of the radiation emitting source. Only off-axis,
3D simulation spectrum, a few cm after the midpoint, de- the even harmonic radiation is present. Correspondingly,
velops a peaked structure similar to that observed in the in the IFEL interaction, the strength of the interaction be-
data. Experimentally, this structure was reproducible shot tween the relativistic electrons and the even-harmonics ra-
to shot, ruling out the possibility of being caused by micro- diation also vanishes for a perfectly symmetric interaction.
structures present in the e-beam or in the laser beam. Different factors contribute to the second harmonic cou-
The IFEL resonance condition is ordinarily understood pling observed in the Neptune IFEL experiment. This ex-
to mean that efficient energy exchange between the trans- periment was characterized by the fact that the laser beam
verse EM wave and the electrons can only take place at was coupled in the undulator with a Rayleigh range short
electron energies such that, in the electron rest frame, the compared to the undulator length. Thus three-dimensional
wiggling induced by the laser has the same frequency as effects dominated the dynamics of the IFEL interaction. In
the wiggling induced by the undulator. However, when the particular, the laser waves have an intrinsically large angu-
motion in the electron rest frame is not a simple dipole os- lar spread characterized by the diffraction angle. Unavoid-
cillation, resonance can also occur if the laser frequency is able angular misalignments, as well as trajectory offsets of
a multiple of the undulator wiggling frequency, that is, elec- the electron and the laser beams, possibly contributed to
trons of a fixed energy may interact not only with the fun- the second harmonic interaction coefficient. Moreover, the

503 c
0-7803-8859-3/05/$20.00 2005 IEEE
 Proceedings of 2005 Particle Accelerator Conference, Knoxville, Tennessee

laser mode used in the experiment is not a purely Gaus- P freely propagating in vacuum we can write
sian TEM00 mode, as it was assumed in the design and in 
 P 1
the simulations. After passing in the final amplifier, the ∂γ  
∝
 λzr
·  2  (2)
transverse profile of the CO2 pulse experimentally shows a ∂z z− L2u
1+ 1+ zr
supersaturated Gaussian profile, with a content of different
higher order Hermite-Gaussian modes.
For a constant K IFEL, the optimum Rayleigh range is
The higher harmonic interaction observed in the Nep-
found by imposing
tune IFEL experiment is weak, but the energy exchange
is significant because the laser is still somewhat intense Lu  1/2
∂ P 1
(> 1010 W/cm2 ) and the wiggling parameter K is large ·   Lu 2  dz = 0 (3)
∂zr 0 λzr z− 2
at the end of the undulator. A small energy exchange takes 1+ 1+ zr
place, enough to modulate the final energy distribution and
this corresponds to the peaked microstructure observed re- whose solution (numerically obtained) is
producibly in the experimental spectrum. zr
= 0.15. (4)
Lu
If K grows along the undulator as it would be the case
CONCLUSION in a strongly tapered undulator, we have to solve
Lu  1/2
We report on the observation of > 20 MeV energy gain ∂ P K(z)
(150 %) at the Inverse Free Electron Laser experiment ·   Lu 2  dz = 0 (5)
∂zr 0 λ · zr z− 2
at the Neptune Laboratory. An energy gradient of > 70 1+ 1+ zr
MeV/m is inferred. The fraction of self-trapped particles
exceeded 5 % of the injected bunch. The acceleration gain and the optimum ratio depends on how fast the undulator is
reported is to date the highest obtained with an IFEL ac- tapered. For K(z) growing polinomially along the undula-
celerator. Self-trapping of particles in a stable accelerat- tor the optimum ratio is in the range 0.15-0.25.
ing bucket from a not-prebunched initial distribution was The conclusion is that for given input power P , the best
demonstrated. The effects of the laser diffraction were an- coupling of a diffraction dominated laser beam and an elec-
alyzed in the design phase and studied experimentally. Fi- tron beam in a IFEL accelerator is given by a choice of the
nally, for the first time higher harmonic IFEL (HH-IFEL) Rayleigh range about one fifth of the undulator length.
interaction was observed in the second section of the un-
dulator. The HH-IFEL adds a degree of freedom (the har- REFERENCES
monic coupling number n) in the design of magnetic sys-
[1] R.B. Palmer. J. Applied Physics, 43:3014, 1972.
tems capable of coupling lasers and electron beams. This
work is supported by U.S. Dept. of Energy grant DE-FG03- [2] E. D. Courant, C. Pellegrini, and W. Zakowicz. Phys. Rev.
92ER40693 A, 32:2813, 1985.
[3] I. Wernick and T. C. Marshall. Phys. Rev. A, 46:3566, 1992.
[4] A. Van Steenbergen et al. Phys. Rev. Lett., 77:2690, 1996.
APPENDIX [5] Y. Liu et al. Phys. Rev. Lett., 80:4418, 1998.
[6] R. B. Yoder, T. C. Marshall, and J. L. Hirshfield. Phys. Rev.
In the Neptune IFEL experiment, the choice of the opti- Lett., 86:1765, 2001.
cal scheme was constrained by the space available in the [7] W. Kimura et al. Phys. Rev. Lett., 86:4041, 2001.
experimental hall. In theory one can try to answer the [8] W. Kimura et al. Phys. Rev. Lett., 92:054801, 2004.
question: what is the optimum Rayleigh range for an in- [9] S. G. Anderson et al. AIP Conf. Proc., 569:487, 2000.
verse free-electron laser accelerator given a certain amount
[10] S. Ya. Tochitsky et al. Opt. Lett., 24:1717, 1999.
of laser power available? What is the best optical scheme
to maximize the exchange of energy between the photons [11] A. A. Varfolomeev et al., NIM A, 483:377, 2002.
and the electrons? [12] S. Tolmachev et al. Proc. of 2004 FEL conference, Trieste,
Italy, 504 (2004)
The optimum point is a result of the compromise of fo-
cusing tighter to increase the intensity and keep the beam [13] P. Musumeci et al. AIP Conf. Proc., 647:278, 2003.
size uniform along the undulator to minimize the diffrac- [14] L. Giannessi, P. Musumeci, M. Quattromini Nucl. Instr.
tion effects. To find it, keeping the focal point of the laser Meth. A, 436:443, 1999.
in the center of the undulator (zw = Lu /2), we should [15] S. Ya. Tochitsky et al. Phys. of Plasmas, 11:2875, 2004.
maximize the available gradient integrated along the undu- [16] P. Musumeci et al. Phys. Rev. Lett., 94:154801, 2005.
lator. The available gradient is proportional to kKl (z)K [17] M. J. Schmidt and C. J. Elliot Phys. Rev. A, 34:4843, 1986.
where Kl is the normalized vector potential of the electro-
[18] P. Musumeci et al. 2005. accepted for publication in PRE.
magnetic wave, and for a gaussian laser beam with power

c
0-7803-8859-3/05/$20.00 2005 IEEE 504