VOLUME
PHYSICAL REVIEW LETTERS
4$, NUMBER 4
23 Jvtv 1979
Laser Electron Accelerator
Department
T. Tajima and J. M. Dawson
of 1'%ysics, University of California, Los Angeles, California 90024
(Received 9 March 1979)
pulse can create a weak of plasma oscillations through the
action of the nonlinear ponderomotive force. Electrons trapped in the wake can be accelerated to high energy. Existing glass lasers of power density lotsW/cm shone on plasmas of densities 10 8 cm can yield gigaelectronvolts of electron energy per centimeter
of acceleration distance. This acceleration mechanism is demonstrated through computer
simulation. Applications to accelerators and pulsers are examined.
An intense
electromagnetic
Collective plasma accelerators have recently
received considerable theoretical and experimental investigation.
Earlier Fermi' and McMillan' considered cosmic-ray particle acceleration by moving magnetic fields' or electromagnetic waves. ' In terms of the realizable laboratory technology for collective accelerators,
present-day electron beams' yield electric fields
of -10' V/cm and power densities of 10" W/cm'.
On the other hand, the glass laser technology is
capable of delivering a power density of 10"W/
cm', and, as we shall see, an electric field of
10' V/cm. We propose a mechanism for utilizing this high-power electromagnetic radiation
from lasers to accelerate electrons to high energies in a short distance. The details of this
mechanism are examined through the use of
computer simulation. Meanwhile, there have
been a few works for particle acceleration using
lasers. Chan considered electron acceleration
of the order of 40 MeV with comoving relativistic
electron beam and laser light. Palmer' discussed
an electron accelerator with lasers going through
a hebcal magnetic field. Willi. s' proposed a positive-ion accelerator with a relativistic electron
beam modulated by laser light.
A wave packet of electromagnetic radiation
(photons) injected in an underdense plasma ex-
cites an electrostatic wake behind the photons.
The traveling electromagnetic wave packet in a
plasma has a group velocity of v, E~ = c(1- &up'/
~')'t'& c, where &op is the plasma frequency and
~ the photon frequency. The make plasma wave
(plasmon) is excited by the ponderomotive force
created by the photons with the phase velocity of
Vp
= 4&p /kp = vg
= C(1 —(dp /(d )
where k~ is the wave number of the plasma wave.
Such a wake is most effectively generated if the
length of the electromagnetic wave packet is half
'
the wavelength
I., =x
of the plasma waves in the make:
/2=ac/top.
way of exciting the plasmon is to
inject two laser beams with slightly different
frequencies (with frequency difference h~ - cop)
so that the beat distance of the packet becomes
2wc/~p. The mechanism for generating the wakes
An alternative
can be simply seen by the following approximate
treatment. Consider the light wave propagating
in the x direction with the electric field in the
y direction. The light wave sets the electrons
into transverse oscillations. If the intensity is
not so large that the transverse motion does not
become relativistic, then the mean oscillatory
„)/a2mcu' where
energy is (bW~) = m(v, ')/2 = e'(Z—
the angular brackets denote the time average.
In picking up the transverse energy from the
light wave, the electrons must also pick up the
light wave's momentum (hp„) = (bWz, )/c. During
the time the light pulse passes an electron, it is
displaced in x a distance &x = (bv„7), where 7 is
the length of the light pulse. Once the light pulse
has passed, the space charge produced by this
displacement pulls the electron back and a plasma
oscillation is set up. The wake plasmon, which
propagates with phase velocity close to c [Eq.
(1)], can trap electrons. The trapped electrons
which execute trapping oscillations can gain a
large amount of energy when they accelerate
forward, since they largely gain in mass and
only get out of phase with this wave after a long
time.
Let us consider the electron energy gain through
this mechanism. We go to the rest frame of the
photon-induced plasmon. Since the plasma wave
has the phase velocity vp [Eq. (1)], we have P
=vp/c and y = ~/u&p. Note that this frame is also
the rest frame for the photons in the plasma;
in this frame the photons have no momentum.
The Lorentz transformations of the momentum
four-vectors for the photons and the plasmons
1979 The American Physical Society
267
PHYSICAL REVIE%' LETTERS
4), +UMBER 4
VOLUME
are
lPy
y
~-iiiy
k»
f'
0
I'
«~/c
y
(3)
«~y/c)'
'')', /)-("")
~'-'
(4)
where the right-hand side refers to the wave
frame quantities (k~" '=k~/y), k„ is the photon
wave number in the laboratory frame and use
was made of the well-known dispersion relation
for the photon in a plasma &u = (~~'+k„'c')'~'.
Equation (3) is reminiscent' of the relation between the meson and the massless (vacuum) photon: Eq. (3) indicates that the photon in the plasma (dressed photon) has the rest mass ~~/c, because the electromagnetic interaction shielded
by the plasma can reach only the collisionless
skin depth c/~~ just as the nuclear force reaches
the inverse of the meson rest mass. At the same
time, the Lorentz transformation gives the longitudinal electric field associated with the plasmon as invariant (E~
ZI, ).
The critical amplitude of the plasmon is determined by the wave-breaking limit. The oscilin one plasma period
1.ation length by the plasmon
should not exceed the wavelength: k~x~ = 1,
where x~=eZ»/m~~'. ' From Gauss's law the critical longitudinal field attainable is
"'
= ISCMp .
eEg
If follows from the Z~ invariance and Eq. (4) that
the wave potential in the wave frame becomes
=ye+ ~yrec
.
(6)
An electron achieves maximum
energy when it
reverses its acceleration in the wave frame.
Transforming the energy [Eq. (6)] and velocity
in the wave frame to the laboratory frame, we
obtain the maximum
this process as
—i Py
~
y
~
ipy
of 8"
electron energy
yPmc
t
iymc
~
y /
'"=y 'xmc2~2y'mc'.
yiDiix
2~2/~
$™xthrough
2y'pmc
imcy'(1+pc)) '
~
(7)
Therefore, we have
2
t, and length E, for electrons to reach
energies of Eq. (8) may be given by the relation
t, = ~ '"/ceZ z, and l, = ct, or
The time
"
l~=
24&
c/id@,
For the glass laser of 1-p.m wavelength shone on
a plasma of density 10" (10") cm ', it would re26S
23 JUL@ 1979
quire under the present mechanism a power of
10" (10")W/cm' to accelerate electrons to energies W '" of 10' (10") eV over the distance of
1 (30) cm with the longitudinal electric field E~"
of 10' (3 &&10') V/cm.
To demonstrate the present mechanism for
electron acceleration, we have performed computer simulations employing the 1-,'-D (one spatial and three velocity and field dimensions) relativistic electromagnetic code. ' A finite-length
train of electromagnetic radiation with wave number k„ is imposed on an initially uniform thermal
electron plasma. The direction of the photon
propagation, as well as of the allowed spatial
variation, is taken as the x direction. The system has typically the length I.„=5124, the speed
of light c = 5u„ the photon wave number k„=2m/
156, the number of electrons 5120, and the particle size 1~ with a Gaussian shape, and the ions
are fixed and uniform, where 4 and v, are the
unit spatial grid distance and the electron thermal
speed, respectively. In order to effectively
change the ratio ar/~~, we have changed c, keeping the following relations fixed: eE,/m&u = eB,/
m~ =c, L, =wc/&u~ [Eq. (2)], p, =eE,/~, and
~ = (~,'+k„'c')'~', where E, and B, are the pumpwave electric and magnetic field amplitudes and
p, is the corresponding amplitude for the momentum modulatiori. We have run cases with
c=(5, 7.25, 10, and 14.7)a&~A. The initial pump
wave has the form of E, =E, sink„(x -x, ), B,
=B,sink„(x -x, ), and p, =p""" +p, cosk„(x -x, )
for the interval of x = [50&, 81.4&] with x, = 506.
With this assignment, the wave packet has a
spectrum in k with a peak around k =k„and ~
= (v~'+k„'c')'~', and propagates in the forward x
direction approximately retaining the original
polarization. In this series of simulations, a
run with larger c means larger ~, longer photon
train „and stronger E„since ~~ is taken to
be constant.
Figure 1 shows an early stage of the system
The phase-space plot [p, vs x in
development.
Fig. 1(b)] indicates a strong modulation in the p,
distribution within the photon wave-packet location. A kink structure extends behind the packet
ending at the packet starting point (x 506) with
a net motion of the particles in the positive p„
direction. Figure 1(a) shows p„vs x; this should
be compared to Fig. 1(b). The intense longitudinal momentum oscillations are clearly shown,
beginning at the photon wave packet and extending
back to its initial position. This is the wake
plasma wave excited by the photons. Note that
I
-
VOLUME
PHYSICAL REVIEW LETTERS
43, NUMBER 4
-3
-2
-t
0
23 JUx, v 1/79
+1
60-
20-
)0I
I
ttI
I
~I
~I
tt
)00
I~
~
~~
~
I
I
II
~
~
~~
'
~
~~
~
~
~~
~
~
~
00
~~
~
~~
~ t,
~
~~
~
~~ ~~
~ 001
~~~~
~
~~~
~
~ ~
~ ~ 00 ~ ~ ~ ~
~ ~ ~ ~, ~ ~
~ ~
~
10
~~ ~
~ 00 ~ ~ ~
~ 00 ~ ~ ~
~~~ ~~~~
~
~ ~ ~~~ ~ ~~
~
~
~~ ~ ~ ~
~~~
~ ~
~~~~ ~ ~
~~~
~ ~ 00
~~
~~~~
~I ~ ~ I ~ ~ ~ ~ ~
~
0~10
~~ ~
~~~
~ ~~ ~
~
O.'I
~~
~ ~~~
~ ~ ~~~~ ~
~
~
~
~
~ ~ 00 ~ ~ ~ ~ ~ ~
~
0~ ~ ~ ~ 0~
~ ~ ~ ~ ~ 0 ~ ~ ~ ~ ~ 0 ~ 'I 0
~~~
~~~~
0~
~
~~
~~
~~
~~~
~ ~ ~ ~0~
~
~
I'I ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
~ 00 ~ ~ 1100~ ~ 0000 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 10 ~
~~
~
~ ~~~
~~
~~~
~ ~ ~ ~ ~ ~ ~ ~ ~ I~~ ~ ~ ~ ~ ~
I ~
~~~
~~~
~~
~ ~ ~ ~~
~ ~~ ~ ~ ~ ~ ~
~~~
~~
~~
00
~~~
~0
~
~~
~
~~
~
II
l
I I
I
II
It
II
II
I
II
ttII
It
I I I Itt II I I I I
I I It I t I t I I
II I II I IttI I IIt II
I II II I I II I II I
II
I
I
II
II
I
I
I
IIII I
I I
t
I
I
I
I I I I ttI I I I ItI
tt
I I tt II II
II t0t
II I t ttI I I I
It I
III I I III II II II II It I
~,
II
I
II
I
(b)
~ ~ ~
~ ~
~ ~ ~
~ ~
~
~
~
~
~
10 ~ 0 ~
~ ~ ~ ~ ~ ~
~
~
I ~~ ~ ~ ~ ~
ot ~ ~ ~ ~ ~ ~ ~ el ~ ~ ~ ~
I
~
\ ~ ~ ~ ~ t[
~ ~
~ ~ ~
~
~ ~ ~ ~
~ ~ ~ ~ ~
~ ~ ~ ~
~ ~
~ ~
0,4
I
tI
I
~
~
II
~
I
~ ~ 0 ~ ~ ~ ~
~ ~ ~ ~ ~ ~ ~
~ ~
~ ~
I
~ ~ ~ ~
~ ~ ~ ~
~
I
I
~
~ ~
~
~
~
~
~ ~ ~
I
~
~ ~ ~
I ~~ ~ ~ ~~ ~ ~ ~
J
I
~
~ ~
~ ~ ~' ~ ~ ~ ~
~ ~ ~
~ ~ ~ ~
~ ~ ~ ~ \ ~ ~ ~
i ~~~ ~ ~ ~
i
~
0
I
I
~ ~
~ 0 ~
~ ~ ~
~ ~ ~ ~ ~
~ ~ ~
~
~
I
x
0
0
I
IO
I
20
I
I
30
40
(Q/Q)p j
~
II
I
k
1024
0
0
~
~
~
~ ~ ~
~ ~ ~
~ 0 ~
~ ~ ~ ~
~ 0
~
sl ~ ~ ~ ~ ~ ~ ~ ~
~ ~
~ ~ ~ ~ ~ ~ 0, ~ ~
~
~ ~& ~ ~ ~ ~ ~ ~ ~ ~ ~
~ ~
~
~ ~ ~ ~
~ ~ ~
~ ~
~ ~ ~ 0
~ ~ ~ ~
~
~
I
I
~ ~
~ ~ ~ ~
~ ~ ~ ~
~ ~
~
~ ~
I
~
~ ~
~ ~ ~
~ ~ ~ ~
~
~ ~
I 0
~
~
FIG. 2. (a) Spectral intensity of electromagnetic
waves in wave number k. The arrow with 0 indicates
the rough position of the original peak; e indicates 4'
=keek& with k&=Id&/c. (h) Maximum electron energy vs
(Id/Id&)2. The dots are from simulation results and the
solid line from Eq. (8).
~
~
~
~ ~ ~
~
~
I
shaped spectrum evolves into
ture with a roughly equal, but
separation in wave number as
This indicates that the photons
number
- nk~
0
(c)
-OA512
FIG. 1. Wake-plasmon excitation and trapping of
electrons. The head of the photon packet has proceeded
forward to @=310 at t =34~~ . Id/&@~=4. 3. (a) The
momentum (p, =p ~~) vs position (p, -x phase
space) of electrons. (b) p„-x phase space. (c) The longitudinal field E& =E~( vs position.
longitudinal
already a set of electrons is accelerated to large
positive P„momenta by the wake plasmon, exemplified by the long stretching armlike phasespace pattern [Fig. 1(a)] . These arms keep
stretching to large momenta.
The wake plasmon structure is also apparent
in the plot of the longitudinal electric fields [Fig.
1(c)] . The longitudinal field strength reaches
values around ZI, -O 6mc&u~/e .[cf. Eq. (5)] or
0.6 of the theoretical maximum value for a cold
plasma. As time progresses, the photons continue to emit the plasmon wake leaving a longer
and longer plasma wave train. Figure 2(a) shows
the wave-number spectrum of the electromagnetic
pulse at successive times. The original smooth-
a multipeak strucslightly increasing,
0 approaches k~.
with peak wave
= k„
k„decay into the photons with 0'—
and nI'=&a-n~, (n an integer) through successive or multiple forward Raman scattering
instabilities. As co' decreases, the photon group
velocity decreases. This process is simply the
photon deceleration caused by the emission and
drag of the wake plasmons. The possibility of
multiple forward Raman scattering may relax
the condition in Eq. (2), since the forward Raman
instability itself creates wake plasmons. Also,
this possibility of acceleration by the forward
Raman scattering" may raise a serious problem
of high-energy electrons in the laser fusion experiment. An arm of accelerated particles
stretches from each plasma wavelength and constitutes approximately 1'Fp of the total electron
in this simulation. The longitudinal
electron momenta and the total longitudinal electric field energy typically increase linearly in
time until saturation.
For various photon frequencies (nI/re~), we measure the maximum electron energy achieved.
The simulation results are given by the dots in
Fig. 2(b). These points should be compared with
the theoretical prediction, Eq. (8) (solid line).
We find that the (uI/nI~) characteristics of the attainable electron energy by simulations, indeed,
follow very closely Eq. (8). Because of the finite
system size and the periodic boundary conditions,
population
269
VQLUMR
PHYSICAL REVIEW LETTERS
4$, NUMBER 4
the interference of the wake fields becomes significant beyond (~/u&~)'-40. To extend the simulations to really high energy, one needs much
longer system size for the simulation; however,
since the scaling agrees with Eq. (8) we can use
it with some confidence there.
The present mechanism of electron acceleration seems f easible within present-day technology.
Although the pulse lengths of (2n+ 1)mc/~~ are
also allowed, techniques of making short pulses
have to be perfected (e.g. , pulse chopping by
backscattering). Having two laser beams with
hen = ~~ as mentioned above is an alternative.
We may speculate that the present acceleration
process may play a role in such an environment
as a pulsar atmosphere, where the dipole radiation fields can be so large that eE/m&a» c. In
the early life of a pulsar when the blowoff plasma,
still not far from the pulsar, faces these intense
fields, the pulsar plasma can be a strong cosmicray source through this mechanism.
N. Leboeuf, M. AshourWe thank D. Vitkoff,
Abdalla, and C. F. Kennel for discussions. This
work was supported by the U. S. National Science
Foundation Grant No. PHY 79-01319.
E. Fermi, Phys. Rev. 75, 1169 (1949).
E. M. McMillan, Phys. Rev. 79, 498 (1950).
~B. Bernstein and I. Smith, IEEE Trans. Nucl. Science 8, 294 (1978).
Y. W. Chan, Phys. Lett. A35, 805 (1971).
~R. B. Palmer,
Appl. Phys. 48, 3014 (1972).
W. J. Willis, CERN Report No. 75-9, 1975 (unpublished).
~Conditions eE/m~ =c for electromagnetic and eEz/
= c for electrostatic waves change the plasma frem~& —
quencies only slightly, because particles acquire relativistic momenta only at the peak of the oscillations,
except for the trapped electrons for longitudinal oscillations whose population is a fraction of the total. Highly
relativistic cases {eE/m~) c) have magnetic acceleration as well.
H. Yukawa, Proc. Phys. Math. Soc. Jpn. 17, 48
J.
(19S5).
A. T. Lin, J. M. Dawson, and H. Okuda, Phys. Fluids 17, 1995 (1974).
OB. I. Cohen, A. N. Kaufman, and K. M. Watson,
Phys. Bev. Lett. 29, 581 (1972).
J.
Neutral-Beam-Heating
25 JUx. v 1979
Results from the Princeton Large Torus
'
H. Eubank, B. Goldston, V. Arunasalam, M. Bitter, K. Bol, D. Boyd,
N. Bretz, J.-P. Bussac,
S. Cohen, P. Colestock, S. Davis, D. Dimock, H. Dylla, P. Efthimion, L. Grisham, R. Hawryluk,
K. Hill, E. Hinnov, J. Hosea, H. Hsuan, D. Johnson, G. Martin, S. Medley, E. Meservey,
L. Stewart, '&
Schivell, G. Schmidt, F. Stauffer,
Sauthoff, G. Schilling,
J. Strachan, S. Suckewer, H. Takahashi,
W. Stodiek, R. Stooksberry,
G. Tait, @ M. Ulrickson, S. von Goeler, and M. Yamada
Plasma Physics Laboratory, Princeton University, Princeton, New Jersey 08544
¹
'
J.
and
C. Tsai, W. Stirling, W. Dagenhart,
W. Gardner,
Oak Ridge National Laboratory,
Oak Ridge,
M. Menon, and H. Haselton
Tennessee 87830
(Received 1 March 1979)
Experimental results from high-power neutral-beam-injection
experiments on the
Princeton Large Torus tokamak are reported. At the highest beam powers (2.4 MW) and
lowest plasma densities [n, (0) =5&&10 3 cm 3l, ion temperatures of 6.5 keV are achieved.
The ion collisionality v;* drops below 0. 1 over much of the radial profile. Electron heating of AT, /T~ =50% has also been observed, consistent with the gross energy-confinement time of the Ohmically heated plasma, but indicative of enhanced electron-energy
confinement in the core of the plasma.
The purpose of the Princeton Large Torus
experi(PLT) tokamak neutral-beam-injection
ments is to produce collisionless high-temperature tokamak plasmas in which to study ion and
electron thermal transport. In this paper we pre270
exsent data from recent neutral-beam-heating
periments on the PLT tokamak, extending the results of previous injection-heating experiments, ' '
to the better confinement conditions associated
with large tokamaks, and also extending our pre-
1979 The American Physical Society