Using the Relativistic Two-Stream Instability for the Generation of Soft-X-Ray

Attosecond Radiation Pulses

A. Marinelli,
1
E. Hemsing,
2
and J. B. Rosenzweig
1
1
Department of Physics and Astronomy, University of California Los Angeles, Los Angeles, California 90095, USA
2
SLAC National Accelerator Laboratory, Menlo Park, California 94025, USA
(Received 25 September 2012; published 6 February 2013)
In this Letter we discuss a novel method for generating ultrashort radiation pulses using a broadband
two-stream instability in an intense relativistic electron beam. This method relies on an electron beam
having two distinct two-energy bands. The use of this new high brightness electron beam scenario, in
combination with ultrashort soft x-ray pulses from high harmonic generation in gas, allows the production
of high power attosecond pulses for ultrafast pump and probe experiments.
DOI: 10.1103/PhysRevLett.110.064804 PACS numbers: 41.60.Cr, 41.50.+h, 41.75.Ht
The successful operation of the Linac Coherent Light
Source [1] and other free-electron laser (FEL) facilities
around the world [2,3] has established the FEL as by far the
most brilliant current source of coherent x rays. In a
high-gain FEL [4], a high-brightness electron beam travels
in an undulator magnet and amplies to saturation a cop-
ropagating resonant radiation pulse. The main features of
FEL light sources are the very high power (up to several
tens of gigawatts [1]), transverse coherence [5], narrow
bandwidth, and tunability over a continuous range of wave-
lengths (see, e.g., Ref. [2]). The generation of coherent
hard x rays enables new methods, such as diffraction
imaging, that may examine atomic and molecular systems
at their characteristic length scale (angstrom). Further, FEL
pulse lengths are now obtained at the femtosecond level,
thus resolving much of the dynamics of such systems.
While this is an impressive achievement, there is demand
for generating yet shorter x-ray pulses for pump-probe
experiments. In this case the narrow bandwidth of an
XFEL, a highly desirable feature for many applications,
limits the capability of the FEL to achieve amplication in
ultrashort pulses.
Thus, in this Letter we discuss an alternative amplica-
tion scheme based on a relativistic two-stream instability
driven by longitudinal space-charge forces. As we shall
see, the two stream instability is a broadband exponential
amplication process and may represent an important
alternative to the FEL amplier in cases in which broad-
band operation is needed. Indeed, it may allow the genera-
tion and amplication of few cycle pulses at x-ray
wavelengths.
The two-stream instability is a well-known physical
effect in the context of fusion plasmas, space plasmas,
and high-energy accelerators. The instability is driven by
the longitudinal Coulomb eld generated by a plasma with
two distinct peaks in the longitudinal velocity distribution
(see, e.g., Ref. [6]). This type of velocity distribution can
present itself in a wide variety of forms, such as, for
example, a particle beam being injected in a fusion plasma
[7] or an ion beam propagating in the presence of a
background plasma that is employed for transverse focus-
ing and stabilization [8].
In the current case of interest, we study the the
two-stream instability in a relativistic electron beam which
has an energy distribution with two narrow peaks, repre-
senting a beam with two distinct energy strata or bands.
This type of scenario was examined in a different context
having much different physical goals by Beke and Jacobs
[9], to explore use of the two-stream instability to enhance
the gain and efciency of low-energy, mm-wave and
sub-mm-wave FELs. In contrast, in this Letter we discuss
the exploitation of the broadband nature of the two-stream
instability in a modern FEL in the vacuum ultraviolet
(VUV) and soft x-ray regions, in order to allow the
generation and amplication of attosecond pulses.
Figure 1 shows a schematic layout of the proposed
amplication scheme. A premodulated (meaning weakly
microbunched by inverse FEL-derived seeding or shot
noise) electron beam with two distinct energy bands prop-
agates in a focusing channel. This bifurcated energy dis-
tribution may be created in many beam pulse compression
processes (e.g., Ref. [10]), as is discussed further below.
The two-stream instability then serves to amplify the initial
density modulation. After saturation of the instability, the
beam is sent to a broadband radiator such as a short
undulator or a transition radiation screen and the strong
microbunching obtained induces the emission of coherent
Two energy level
pre-modulated
electron beam
Broad-band
undulator
+

Two-stream instability
microbunching
Radiation
pulse
Focusing
Channel
FIG. 1 (color online). Layout of the two-stream amplier.
PRL 110, 064804 (2013)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 FEBRUARY 2013
0031-9007,13,110(6),064804(5) 064804-1 2013 American Physical Society
radiation. Contrary to the FEL, in which the undulator eld
couples the electron motion to the radiation eld, thus
driving a collective instability, in this case the undulator
only serves the purpose of extracting energy from the
microbunched beam in the form of electromagnetic radia-
tion (super-radiant emission). The exponential growth of
the microbunching, instead, is driven by the longitudinal
space-charge forces outside of the undulator.
The two-stream instability for a relativistic electron
beam with two energy levels can be well described by a
simple one-dimensional model. We assume a coasting
(nonaccelerating) beam with an average energy of mc
2
.
The beam is described by a distribution function in longi-
tudinal phase-space ]z, ,, r where z is the longitudinal
coordinate along the electron beam with respect to a
reference particle traveling at the normalized speed

1
2
_
, , u, is the relative energy devia-
tion with respect to the mean beam energy, and r ci
where i is the time variable and c is the speed of light. The
collective longitudinal beam dynamics is described by the
Vlasov equation in the relevant two-dimensional phase
space. We expand the distribution function to rst order
in perturbation theory: ] ]
0
]
1
with j]
1
j (]
0
.
Furthermore, we assume that the lowest order distribution
function is factorable as ]
0
n
0
]

, where n
0
is
the beam volume density and ]

is the beams energy

distribution function. The resulting linearized Vlasov
equation is:
]
1
r

,

2
]
1
z

eE
z
mc
2
]
0
,
0 (1)
where e is the electron charge and E
z
is the longitudinal
electric eld, which can be computed by solving the
one-dimensional Poisson equation,

z
E
z

e
e
0
_
]
1
J,. (2)
It is convenient to solve Eqs. (1) and (2) in the Laplace-
Fourier domain. We give the following denitions:
^
]
1

_
]
1
e
|kz
Jz; and
~
^
]
1

_
1
0
^
]
1
e
|
w
c
r
Jr. With these de-
nitions, the Fourier-Laplace transforms of Eqs. (1) and (2)
yield
|
w
c
~
^
]
1

^
]
1

r0

|k,

2
~
^
]
1

e
mc
2
n
0
~
^
E
z
]

,
0 (3)
and
~
^
E
z

|
k
e
e
0
_
~
^
]
1
J,. (4)
After some algebraic manipulation (see, e.g., Ref. [11]), it
can be shown that the phase-space perturbation ]
1
can be
expressed as
~
^
]
1

1
|
w
c

|k,

2
_
_^
]
1
j
r0

1
e

w
2

c
2
]

2
|k
_ ^
]
1
j
r0
|
w
c
|k
,
0

2
J,
0
_
_
,
(5)
where e

is the beams plasma dielectric function, given by

e

1
w
2

c
2

2
|k
_
]

,
|
w
c

|k,

2
J,, (6)
and w
2

e
2
n
0
e
0
m
3
being the relativistic beam plasma
frequency.
The Laplace transform in Eq. (5) can be inverted by
using the residue theorem. In doing so we will only
consider the poles associated to the zeros of the dielectric
function e

since those are the poles that describe the

collective response of the electrons.
The two-streamenergy structure can be modeled with the
following energy distribution function ]

2r
p
a
,

e
,,
2
,2a
2
,
e
,,
2
,2a
2
,
where , (1. We intro-
duce the following dimensionless variables: K
kca
,

2
w

is
the energy spread parameter,
kc,

2
w

is the normalized
energy energy separation, and
w
w

is the normalized
Laplace variable. The resulting plasma dielectric function is
e

1
1
4K
2
_
Z
0
_

2
p
K
_
Z
0
_

2
p
K
__
, (7)
where Z
0
is the complex derivative of the plasma dispersion
function dened as: Z[
1

r
p
_
~ c
Jx
e
x
2
x[
, where ~ c is the
Landau contour which runs in the complex plane from
1 to 1 and below the singularity at x [.
To study the stability of the system we will focus on the
cold beam limit, i.e., the limit for vanishing energy spread:
K !0. In this limit the dispersion equation for the system
reduces to
1
1
2
2

1
2
2
0. (8)
Equation (8) has the following solutions:

,

1

2
p

2
2
1

1 8
2
_
_
. (9)
For <1 the root
,
|
|

2
p

18
2
p
2
2
1
_
is
purely imaginary with a positive imaginary part, leading to
an exponential growth of the phase-space perturbation
^
]
1
as a function of time. In analogy with the theory of free-
electron lasers, we dene as the gain parameter and the
gain length l
g

c
w

. Figure 2 shows the unstable root as a

function of for the for the cold beam limit and for a warm
beam with different values of K. The gain parameter has an
optimum value for
opt


3
p
,2

2
p
corresponding to

opt
1,2

2
p
. In terms of physical units, the streaming
PRL 110, 064804 (2013)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 FEBRUARY 2013
064804-2
instability provides broadband amplication mechanisms
for beam microbunching at wavelengths larger than

th

2
with an optimum gain length l
g,opt


2
p

,r
(where

2rc,w

is the plasma period) at

opt

2

2
p

3
p

2
.
To quantify the spectral properties of the amplica-
tion process, we dene the bunching factor as lk
1
N
_
^
]
1
J,, where N is the number of particles in the
beam. In the cold beam limit, explicit inversion of the
Laplace transform in Eq. (5) gives, in dimensionless units,
l
l
0

2
3
2
e
T
(10)
where T
w

r
c
and l
0
is the bunching factor at T 0.
To study the behavior of the amplication spectrum
around the optimum energy separation we note that, to
second order,
1
2

2
p

3

2
p
4

opt

2
, giving jlj
2
/
e
T,

2
p

opt

2
,2a
2

, with a relative root-mean-square (rms)

power amplication bandwidth of
a

opt

a
k
k

2
5,4
3

T
p
. The
amplication bandwidth has a rather weak dependence
on the normalized interaction time T, yielding a broad
amplication bandwidth for most cases of practical inter-
est. Note that, from basic Fourier analysis, the shortest rms
pulse duration achievable with the amplier is given by
a
z,min

1
2a
k

opt
k
4ra
k

opt
3

T
p
4r2
5,4
. It follows that the
two-stream instability can amplify few optical cycle pulses
without signicant lengthening, unlike the free-electron
laser instability in which the radiation slippage limits the
minimum pulse duration to a cooperation length [12].
For example, assuming T 14 (corresponding to roughly
ve gain lengths) we have
a
k
k
20%, giving a minimum
rms pulse length of a
z,min
0.4
opt
. This feature makes
the two-stream amplier an attractive method for the
generation of attosecond pulses.
A central challenge in the operation of a two-stream
amplier lies in the generation of intense electron beams
with a two-energy-band structure. This problem may be
addressed in several ways and a detailed discussion of the
generation of two-stream beams will be left for future
publications. However, we mention here that this type of
phase-space structure can be generated by illuminating a
phot-cathode with a pulse train and imparting a z-energy
correlation by accelerating the resulting multibunch beam
off-crest. At this point the micropulses can be overlapped
in time with using either velocity bunching or, again,
magnetic compression. It is also worth mentioning that
the peak current amplication induced by nonlinear wave
breaking, demonstrated in Ref. [13], could greatly enhance
this scheme. Alternatively, an ultrashort two-stream struc-
ture could be induced with an E-SASE compression
scheme [14] by adding a beam mask in the center of the
magnetic chicane.
The longitudinal space-charge microbunching instabil-
ity has also recently been proposed as an amplier for the
generation of broadband radiation pulses, in a scheme
known as the longitudinal space-charge amplier (LSCA)
[15,16]. In a LSCA, an electron beam travels in a focusing
channel and the collective longitudinal space-charge elds,
generated by shot noise or by a pre-existing density modu-
lation, induce an energy modulation. After the focusing
channel, the energy modulation is transferred into density
modulation with a magnetic chicane, generating a bunch-
ing factor that is much greater than its starting value. One
obvious advantage of the two-stream instability with
respect to the LSCA is the absence of a magnetic chicane
in the amplication process. Another key advantage of the
two-stream amplier over the LSCA is the exibility in the
wavelength tuning. The central amplication wavelength
of a LSCA is given by the condition k
opt
a
x
, 1, where
a
x
is the rms transverse size of the electron beam. This
condition is difcult to attain at very short wavelengths and
requires drastic changes in the beam transport in order to
be tuned over a wide range of wavelengths. On the other
hand, the two-stream amplier can be simply tuned by
varying the energy separation of the two beamlets.
As pointed out in Ref. [15], broadband ampliers based
on relativistic electrons have a wide range of applications,
due to their robustness to beam stability requirements and
their unique spectral properties. In particular, the broad-
band nature of the two-stream instability makes it an
attractive method for the generation and amplication of
tunable ultrashort pulses. The generation of intense atto-
second VUV radiation pulses from high-harmonic genera-
tion in gas has recently been demonstrated [17]. Such
radiation pulses cannot be amplied in a conventional
seeded FEL due to the effect of slippage, which limits
the time duration of an FEL pulse to a cooperation length
(which is typically several radiation periods). Few-cycle
pulses can be amplied with a broadband amplier, such as
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4

K=0
K=0.1
K=0.25
FIG. 2 (color online). Imaginary part of the normalized
unstable root as a function of the normalized energy separation
for three different values of the energy-spread parameter K.
PRL 110, 064804 (2013)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 FEBRUARY 2013
064804-3
the two-stream amplier discussed in this Letter. We
discuss an example corresponding to the generation of an
ultrashort soft x-ray pulse starting from high harmonic
generation (HHG) in gas. We assume a seed pulse at
s

30 nm with a power of 1 MW and a rms duration of 150
attoseconds. To quantify the amplication process, we
dene the density perturbation ~ n
_
J,]
1
,n
0
. The seed
interacts with a resonant electron beam of energy
E 750 MeV in a magnetic undulator. Assuming an
undulator with two periods of length
v
5 cm, an undu-
lator parameter of K
v
1.3 and a beam uncorrelated
energy spread of a
,
0.75 10
4
, the resonant inter-
action, followed by a magnetic chicane with longitudinal
dispersion R
56
z,, 25 m, we can assume a
third harmonic density perturbation of the type ~ n
~ n
max
coskze
kz
2
,2a
2
z

at 2r,k 10 nm, with

~ n
max
2 10
3
and a
z
350 asec c. The induced en-
ergy modulation is smaller than the assumed uncorrelated
energy spread, which means that to generate strong micro-
bunching, amplication through the two-stream instability
is needed. Due to the two-energy structure, the two beam-
lets will shift with respect to each other as a result of
dispersion in the bunching chicane, leading to dephasing
of the microbunches and to an overall lengthening of the
microbunching structure. For an efcient short-pulse seed-
ing the chicane dispersion has to fulll the following
conditions: 2R
56
, n, which means that the relative
shift of the two beamlets has to be a multiple of the
microbunching wavelength, and 2R
56
, <a
z
which
ensures that the relative shift of the two streams does not
wash out the ultrashort pulse structure. The nal rms length
of the microbunched structure is estimated by adding in
quadrature the length of the radiation pulse ( 66 nm for
the eld distribution), the undulator slippage (2
r

60 nm) and the relative shift of the two beamlets in the
dispersive section (2,R
56
60 nm). Alternatively, one
could choose a larger R
56
to generate two separate pulses
with a shorter length. With the parameters chosen, assum-
ing R
56
220 m, the relative shift of the beamlets is
2,R
56
,c 1760 asec, which results in two isolated
microbunching pulses of amplitude ~ n
max
1.5 10
3
and duration a
z
300 asec c.
We assume a beam current of I 500A, a transverse
radius of r
x
11 m with an energy separation of ,
0.0012. With these parameters, the gain length is optimized
at
opt
10 nm with l
g
5.2 m. Figure 3 shows the lon-
gitudinal density perturbation ~ n
_
J,]
1
,n
0
as a func-
tion of the position along the electron bunch, at saturation
for the case of a single pulse (roughly after 6 gain lengths)
and for the double pulse case (roughly after 6.5 gain
lengths). Note that the attosecond structure is not washed
out during the exponential gain due to the broadband
nature of the instability. Note also that for a bunching
factor larger than ~ n 50%, nonlinear effects become
important, leading to saturation of the exponential gain.
In this nonlinear regime a numerical particle tracking code
would be needed to describe this phenomenon.
After the amplication process, the microbunched elec-
tron beam can be sent into a broadband radiator for the
emission of coherent radiation. The radiator can be a
broadband undulator (i.e., an undulator with few periods)
or a metal foil, which causes the emission of coherent
transition radiation. The undulator radiation mechanism
induces a pulse-lengthening process due to the slippage
of the radiation over the electrons. The condition to pre-
serve the short pulse structure in the undulator is that the
slippage length, dened as the resonant wavelength times
the number of undulator periods, be much shorter than
the length of the density perturbation in the beam, i.e.,
N
v

r
<a
z
. The peak power emitted by a prebunched
electron beam in a helical undulator can be estimated
with the following expression, derived in Ref. [18]:
P W
l
r
2
l
2
m
2
I
I
o
K
2
v
1 K
2
v
N
v
E
^
N, (11)
where
^
N kr
2
x
,4L
v
, with L
v
being the undulator length,
E
^
N
2
r
arctan
1
2
^
N

^
Nln
4
^
N
2
14
^
N
2
, W
l
mc
2
I,e is
the beam power, l
m
is the peak bunching factor, I
o

17 kA is the Alfven current, K
v
is the undulator parameter,
and N
v
is the number of undulator periods. For the ex-
amples chosen, a four-period undulator with a period of

v
2 cm and K
v
1.15 yields a peak power of W
15 MW, with a pulse energy of U 10 nJ, for the single
pulse case, and U 8.6 nJ per pulse in the double pulse
case, with negligible pulse lengthening. The nal radiation
pulse duration is a
z
,c

2
p
250 asec for the single pulse
1500 1000 500 0 500 1000 1500
1
0.5
0
0.5
1
t (attosec )
n
~
2000 1500 1000 500 0 500 1000 1500 2000
1
0.5
0
0.5
1
t (attosec )
n ~
FIG. 3 (color online). Longitudinal density perturbation at
saturation for the single pulse (upper plot) and double pulse
(bottom plot) congurations.
PRL 110, 064804 (2013)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 FEBRUARY 2013
064804-4
and a
z
,c

2
p
214 asec for each of the two pulses in the
double pulse case.
In conclusion, in this Letter we have discussed the gen-
eration of attosecond, short wavelength electromagnetic
pulses using the relativistic two-stream instability. The
relativistic two-stream amplier is a tunable broadband
amplier, which holds great promise for the generation of
ultrashort VUV and soft x-ray pulses. The two-stream
amplier relies on the formation of a beamwith two distinct
energy levels and is continuously tunable due to the exi-
bility in the choice of the energy separation ,. We have
discussed a one-dimensional model of the two-stream
amplier and identied simple scaling laws for the optimal
wavelength, gain length and for the coupling to an initial
bunching factor. Finally, we have discussed an example
corresponding to the generation of an attosecond pulse at
10 nm, thus illustrating the practical application of this
novel amplication scheme in the context of creating an
unprecedented 250 attosecond, coherent soft-x-ray pulse.
The authors would like to acknowledge P. Musumeci, G.
Andonian, and S. Stagira for useful discussions and sug-
gestions. This research is supported by grants from US
DOE Contracts No. DE-FG02-07ER46272 and No. DE-
FG03-92ER40693, Ofce of Naval Research Contract
No. N00014-06-1-0925, DARPA Contract No. N66001-
11-4197.
[1] P. Emma et al., Nat. Photonics 4, 641 (2010).
[2] W. Ackermann et al., Nat. Photonics 1, 336 (2007).
[3] T. Ishikawa et al., Nat. Photonics 6, 540 (2012).
[4] R. Bonifacio, C. Pellegrini, and L. Narducci, Opt.
Commun. 50, 373 (1984).
[5] R. Ischebeck et al., Nucl. Instrum. Methods Phys. Res.,
Sect. A 507, 175 (2003).
[6] P. M. Bellan, Fundamentals of Plasma Physics
(Cambridge University Press, Cambridge, England,
2006).
[7] L. E. Thode and R. N. Sudan, Phys. Rev. Lett. 30, 732
(1973).
[8] R. C. Davidson, H. Qin, and T.-S. F. Wang, Phys. Lett. A
252, 213 (1999).
[9] G. Beke and K. D. Jacobs, J. Appl. Phys. 53, 4113
(1982).
[10] G. Andonian et al., Phys. Rev. ST Accel. Beams 12,
030701 (2009).
[11] A. Marinelli and J. B. Rosenzweig, Phys. Rev. ST Accel.
Beams 13, 110703 (2010).
[12] R. Bonifacio, L. De Salvo, P. Pierini, N. Piovella, and
C. Pellegrini, Phys. Rev. Lett. 73, 70 (1994).
[13] P. Musumeci, R. K. Li, and A. Marinelli, Phys. Rev. Lett.
106, 184801 (2011).
[14] A. A. Zholents and W. M. Fawley, Phys. Rev. Lett. 92,
224801 (2004).
[15] E. A. Schneidmiller and M. V. Yurkov, Phys. Rev. ST
Accel. Beams 13, 110701 (2010).
[16] M. Dohlus, E. A. Schneidmiller, and M. V. Yurkov, Phys.
Rev. ST Accel. Beams 14, 090702 (2011).
[17] F. Ferrari, F. Calegari, M. Lucchini, C. Vozzi, S. Stagira,
G. Sansone, and M. Nisoli, Nat. Photonics 4, 875
(2010).
[18] E. Saldin, E. Schneidmiller, and M. Yurkov, Nucl.
Instrum. Methods Phys. Res., Sect. A 539, 499 (2005).
PRL 110, 064804 (2013)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 FEBRUARY 2013
064804-5