Energy Loss of a High Charge Bunched Electron Beam in Plasma

N. Barov, J.B. Rosenzweig and M.C. Thompson

UCLA Department of Physics and Astronomy,

405 Hilgard Ave., Los Angeles, CA 90095-1547

Abstract

There has been much interest in the regime of plasma wake-field acceleration (PWFA)
having ultra-high fields, and associated nonlinear plasma motion. With an exact analytical
approach, we examine here a fundamental limit of PWFA excitation, by an infinitesimally
short relativistic bunched beam. The beam energy loss in this case is shown to be linear in
charge even for nonlinear plasma response, where a normalized, unitless charge exceeds
unity. The physical basis for this effect is discussed, as are deviations from linear behavior
observed in simulations with finite length beams.

PACS Codes: 52.40.Mj, 52.75.Di, 29.17.+w, 29.27.-a

Submitted to Physical Review Letters

 The transfer of energy from short, intense electron beams to collective electron plasma
oscillations, known as plasma wake-fields, has been discussed in the context of the
plasma wake-field accelerator (PWFA), and short wavelength light sources. While the
original proposal for the PWFA and related concepts was in the linear regime[1,2], where
the plasma oscillations can be considered small perturbations about equilibrium, highly
nonlinear regimes have also been studied, and found to have advantages[3]. For example,
in the “blow-out” regime[4], the plasma electrons are ejected from the channel of the
intense driving electron beam, resulting in an an electron-rarefied region with excellent
quality electrostatic focusing fields, as well as longitudinal electromagnetic fields, which
can, in tandem, stably accelerate and contain a trailing electron beam. While the beam
dynamics in this regime are linear and analytically tractable, the plasma dynamics are not,
and most predictions concerning plasma behavior in this regime have been deduced from
numerical models.

Despite the lack of analytical models, it has been noted in a number of studies[5-8]
that the energy loss of a bunched beam in nonlinearly-excited PWFA obeys the scaling
usually associated with the interaction of charged particles with linear media[9]. This
scaling predicts that the energy loss is proportional to the square of the plasma
frequency[9]. As the efficient excitation of an oscillatory system by a pulse occurs when
the pulse in short compared with the period of the oscillator[5-8], this scaling further
implies that the energy loss rate of the PWFA drive beam is proportional to the inverse
square of the driving beam pulse length. This prediction has led to a number of
experiments which employ bunch compressors in order to shorten the bunch length, and
thus dramatically increase the transfer of beam energy to the plasma. In recent measure-
ments with compressed beam at FNAL[10], the trailing portion of a 5 nC, 14 MeV beam
pulse was nearly stopped in 8 cm of plasma with density n0 ≅ 1014 cm-3 , a deceleration

rate of over 150 MeV/m. The large collective field, observed in the context of nonlinear
plasma electron motion, in this as well as other recent PWFA experiments[10,11], have
re-opened issue of the validity of wake-field scaling. In addition, there has been a recent
proposal to use ultra-short, high charge beams to drive PWFA in the tens of GeV/m range,
for creation of an ultra-high energy plasma accelerator[8,12]. This paper is intended
address the underlying physics of wake-field scaling of relativistic beam energy loss in
plasma, and to study potential deviations from this scaling.

We are undertaking an analysis of the motion of cold plasma electrons of equilibrium

density n0 (equal and opposite in charge density to the assumed stationary ions). The
r r r
state of the motion is described in terms of the velocity v and momentum p = mev ,

[ r 2
]
−1 / 2
where the Lorentz factor = 1 − (v / c ) . The necessary relations for describing the

cold plasma electrodynamic response with are the Maxwell equations

r r
r r 4 r 1 E r r 1 H
∇×H= J+ , and ∇ × E = − , [1]
c c t c t

the Lorentz force equation in convective form,

r
p r r r  r 1 r r
t
( )
+ v ⋅ ∇ p = −e  E + v × H ,
 c 
[2]

and the equation of continuity for charge and current density

n r r
+ ∇ ⋅ (nv ) = 0 . [3]
t

The results of our analysis will be made more transparent by the adoption of unitless
variables. The natural variables used in discussing a cold plasma problem parameterize

time and space in terms of the plasma frequency = 4 e 2 n0 / me , and wave-number

p
r r
kp = p / c , respectively, densities in terms of n0 , and the amplitudes of the E and H

fields in terms of the commonly termed "wave-breaking limit" EW B = me c p / e . In

addition, all velocities and momenta are normalized to c and me c , respectively. We thus

write the spatiotemporal variables, charge and current density, and field components as

r˜ = k p r , = p (t − z / c ) , v˜i = v i /c , ˜pi = pi /me c , [4]

 n˜ = ne / n0 , J˜i = Ji / n0 c [5]

E˜ i = Ei / EW B, H˜ i = Hi / EW B . [6]

With these variables, we may write a general equation for the azimuthal component of H˜

H˜ 1 H˜ H˜
2
J˜ J˜
2 + − 2 = r+ z. [7]
r˜ r˜ r˜ ˜r r˜

In addition to the governing equation for H˜ , we will have use for the following
relationships between fields and current sources,

E˜ z ˜
˜r
= J r and ( E˜ − H˜ ) = − J˜ .
r r [8]

In this analysis the induced E˜ z is found most directly by determining the transverse

current, as is customary in media-stimulated radiation calculations (cf. Jackson, Ref. 9).

Equation 7 is nonlinear, but may be simplified by assuming small amplitude response,

in which the normalized densities and fields are small compared to unity. In fact, to place
our results in perspective, we must begin with a short review of previous work in the
linear regime[1,2]. From the viewpoint of the fluid equations, this most importantly
r r
implies that the plasma electron response is non-relativistic, i.e. v = p / me . Also, in order

to illustrate a problematic logarithmic dependence of decelerating field on transverse beam

size, we assume a disk-like beam, uniform up to radius a˜ = k p a <<1, and -function in

longitudinal dependence . We are interested in the instananeous response to the plasma,

and integrate the second of Eqs. 8 over the region near =0 to obtain H˜ = E˜ r before,

during, and immediately after beam passage, which we use to further find

H˜ 1 H˜ H˜
2
Q˜
2 + − 2 − H˜ = 2 ( ) (r˜ − a˜ ). [9]
r˜ ˜r r˜ r˜ a˜

Here we have introduced a normalized beam charge

Q˜ = 4 k p re N b [10]
which, when small compared to unity, indicates that the response of the system is linear.
It should be noted in this regard that the experiments of Refs. 10 and 11 have beam-
plasma systems yielding Q˜ values between 2 and 4.

Equation 9 has a temporal -function dependence which we integrate over, to

obtain the inhomogenous modified Bessel equation in r˜

2
H 1 H H Q˜
2 + − 2 − H = 2 (r˜ − a˜) , [11]
r˜ r˜ r˜ r˜ a˜
+ +
where H = ∫ H˜ d = ∫ E˜ d
r . We interpret H as the radial momentum impulse ˜pr , which
− −

in the non-relativistic limit is also approximately equal to J˜r immediately behind the beam

pulse. The solution to this Eq. 11 is given by

Q˜ K1 (a˜ )I1 (r˜ ) (r˜ < a˜)

H( r˜) =  [12]
a˜ K1 (˜r)I1 (˜a) (˜r > a˜),

where I1 and K1 are modified Bessel functions.

We are ultimately interested in E˜ z directly behind the beam, which is found by

integrating Eq. 12

r˜
Q˜ 1− a˜ K1 ( a˜) I0 (r˜ )
E˜ z (r˜ ) = +
= ∫ H( ˜r′)dr˜′ = 
a˜ 2  a˜ I1 ( a˜)K 0 ( r˜)
. [13]
∞

For a˜ <<1, the field inside of the disk is nearly constant, and given by

Q˜ Q˜   2  
E˜ z (r˜ ) ≅ [1− ˜
a K ( ˜
a ) ] ≅ ln − 0.577... , [14]
2   a˜  
= + 2 1
a˜

The induced normalized field is to leading order proportional to Q˜ /2 . In physical units

we may write Eq. 14 as

 1.123 
eEz ≅ 2e 2 k 2p N b ln . [15]
= +
 kp a 

Several comments arise from inspection of Eq. 15. The first is that the scaling of E z with
respect to wavenumber is dominated by the factor of k 2p that is typical of Cerenkov

radiation[9], if we interpret k p as the maximum allowable radiated wavenumber. The

second comment is that the linear result is ill-behaved in the limit of k p a<<1, as Eq. 15

predicts a logarithmic divergence in E z . This pathology is a result of allowing J˜r (through

H ) to diverge as r−1 . Previous analyses by Jackson[9] and Chen, et al.,[1] have attempted
to mitigate this problem by introducing a lower bound on r (or impact parameter b), and
have chosen the Debye length1. This ad hoc way of removing the logarithmic divergence
of an ultra-relativistic electron’s energy loss in plasma has a dubious physical basis,
however. Debye shielding places the scale of maximum distance that a charged particle’s
macroscopic field can be observed in the plasma after an equilibrium is established
through the action of thermal motion of the plasma electrons. On the other hand, we are
concerned here with the minimum distance that the fluid analysis is valid for describing
the plasma electron response to an extremely fast transient, a particle with velocity much
higher than the plasma electron thermal velocity. We will return to this point below.

As Q˜ is raised, we must consider the plasma electrons’ relativistic response. The

relativistically correct motion of an electron in crossed electric and magnetic fields of equal
magnitude has been discussed in Ref. 13. Adapting this analysis to our case, we find that,
as in the non-relativistic case, that the radial impulse ˜pr imparted by the beam is

˜pr = H , [16]

while the longitudinal momentum impulse is given by

˜pz = 12 p˜ r2 . [17]

For large H , the plasma electrons obtain large forward momentum impulse, and can have

1It is not quite clear from Jackson’s discussion if the Debye length is to be applied as the lower bound to b
only in the case to non-relativistic particles (cf. Ex. 13.3, Ref. 9).
relativistic longitudinal velocities just after passage of the beam2. The transverse velocity
of the plasma electrons then becomes

H H
v˜r = = . [18]
1+ H 2 + 14 H 2 1+ 12 H 2

In order to relate this v˜r to J˜r we must multiply by the density, which due to the change

in v˜z directly after passage of the beam, is predicted with the aid of Eq. 17 to be

n˜ = (1− v˜z ) = 1+ 12 H 2 .
−1
[19]

Thus we are led to the remarkable result that the exact, relativistically correct induced
radial current is identical to the approximate, linear, non-relativistic expression,

J˜r = n˜ v˜r = (1+ 12 H 2 ) ⋅

H
= H. [20]
1+ 12 H 2

Since the induced J˜r is unchanged from the linear case, the analysis of the decelerating

field E˜ z leading to Eq. 13 remains valid. Thus we have shown that the scaling observed

in simulations of short pulse electron beam-excited plasma wake-fields has an analytical

basis. This result arises from two effects which cancel each other: the induced v˜r

saturates (at a value well below 1), yet the density enhancement due to longitudinal
motion — a “snow-plowing” of the plasma electrons by the electromagnetic pressure —
exactly makes up for this saturation, and the induced J˜r remains linear in Q˜ .

An additional check on the validity of Eqs. 9-20 is in order. We can calculate the
total (field and kinetic ) energy per unit length deposited in the plasma by the beam as

∫[ ]
∞ ∞
dU
=2 1 + p˜ 2r + p˜ 2z − 1 r˜ dr˜ + 2 ∫ E˜ z2 (r˜ )r˜dr˜
1
2
dz a˜ a˜

2 If one chooses to derive Eqs. 15 and 16 in the laboratory frame, the solution of the “impulse” equations of
motion must take into account a lengthened interaction time (a finite value evaluated in the δ-function limit) as
the plasma elections obtain longitudinal velocity, by multiplying by the factor (1− v z ) .
−1
 Q˜ 2
2 [1− aK1 (a) I0 (a)] = 2 Q E z (r )
1 ˜ ˜
= ˜ ˜ ˜ ˜ [21]
2 a˜ = +

Thus we see from energy balance that the average decelerating field experienced by the
beam in the limit of an infinitesimally thin disk is one-half of that given in Eq. 13. This
factor of two is familiar from the linear theory of beam wake-fields in media and metallic
structures[2]. The validation from this cross-check on our results is notable because the
induced motion of the plasma electrons is explicitly relativistic.

As the results of Eqs. 16-20 concern beams of vanishing length, they are applicable, in
the limit a˜ → 0, to the case of a single particle. The effects of nonlinear plasma electron
response do not, as might have been hoped, remove the logarithmic divergence noted in
Eq. 14, due to the balancing of the limit on transverse velocity with the enhancement of
the plasma density. Noting that the logarithmic term in Eq. 15 corresponds to the familiar
Coulomb logarithm[8], with argument that is the ratio of the maximum to minimum b,
ln(bmax /bmin ) . We deduce that the upper limit of the impact parameter is bmax = 2/ k p ,

while the lower limit in the analysis is a. The value of a cannot be drawn towards zero
without violating several assumptions of our analysis, however. Modeling the plasma
electrons as a continuous fluid introduces errors not in the average energy loss, but in the
fluctuations of this quantity. For ultra-relativistic particles, quantum mechanical effects
h 2
constrain the minimum impact parameter[8] to bmin ≅ through the uncertainty
mec

principle, however. Thus we write the energy loss rate for a point particle of charge q as

dU  mec   
≅ q 2 k 2p ln 0.794  ≅ q 2 k 2p ln 5 p
, [22]
dz  k ph   c 

where p = 2 / k p and c are the plasma and Compton wavelengths, respectively.

Note that both limits of the Coulomb logarithm in Eq. 22 can be viewed quantum
mechanically, as the minimum quantum of energy loss (emission of a plasmon) in the
plasma is in fact h p , as has been verified experimentally for very thin foils[14].
 While the infinitesimally short beam limit is relevant to the point-charge case, it is
not of practical interest in bunched beams, as it has often been argued that one should set
kp z ≅ 1 (where z is rms bunch length) to optimize drive beam energy loss in plasma

wake-field experiments[2-6,10,11]. In order to explore the respect roles of saturating v r

and the snow-plowing of the density with a such a finite length beam, we have performed
a series of simulations using a fully relativistic particle-in-cell code, MAGIC. Taking the

beam of radius a˜ = 0.2, and Gaussian charge profile b ( z) ∝ exp(−z2 /2 2

z ), with rms

length k p z = 1.1, we have scanned the charge from linear to very nonlinear response,

Q˜ = 0.02 to 200. In order to connect with the point beam limit, and to accurately quantify
the energy imparted to the plasma, we compare the average on-axis beam energy loss

(2 ) ∫ eEz ( z) r= 0 exp(−z 2 /2 2z )dz ,

−1
rate, z for these cases with linear theory. This

integral can be evaluated in the linear limit, and the result is simply to multiply Eq. 15 by

the factor 12 exp − (k p[ z ) ].

2

The results of this parametric scan, as well as the analytical results of linear
theory, are shown in Fig. 1. It can be seen that for the low amplitude cases Q˜ << 1 that
the linear theory predicts the average energy loss well. On the other hand, for Q˜ >> 1, the
energy loss is significantly smaller than that predicted by linear theory, by one order of
magnitude at Q˜ = 200. Note also that the energy loss rate does not grow as Q˜ is
increased from 60 to 200. Similar results can be deduced from other published simulations
[12]. From this behavior, it can be deduced that the relativistic saturation of v r is

dominant over the snow-plow of the density for finite length beams. For completeness
we also plot the peak accelerating field excited behind the driving beam. The peak in this
field is localized in a very narrow spike, which is a small region, not terribly useful for
acceleration. This spike is greatly enhanced in the nonlinear case, thus partly explaining
why field saturation was not noted in previous simulation scans[4,8,12]. Even with this
masking effect, the accelerating peak still displays saturation when Q˜ >> 1.
 In conclusion, we restate the most surprising of our results, that the fully
relativistic response of a plasma to the passage of an ultra-short beam is identical to the
linear result. The interplay between the nonlinear effects which cancel for infinitesimal,
but not for finite length beams must be studied in more detail by simulation. Even without
this study, however, we may clearly state that the previously proposed scaling of wake-
−2
field amplitudes as linear with k 2p (or z for constant k p z and k p a) is violated when

the normalized charge Q˜ exceeds unity. In fact, our simulations indicate that when Q˜
exceeds 100, that the coupling of the beam to the plasma does not notably grow as the
beam charge is increased. It should be noted in this regard that current proposals for
PWFA experiments using highly compressed beams at Stanford imply beam-plasma
systems with Q˜ >100.

This work supported by U.S. Dept. of Energy grant DE-FG03-92ER40693.

 References

1. Pisin Chen, J.M. Dawson, Robert Huff, and T. Katsouleas, Phys. Rev. Lett. 54, 693
(1985), and Phys. Rev. Lett. 55, 1537 (1985).

2. R. Ruth, R. D. Ruth and A. W. Chao, in Laser Acceleration of Particles. AIP Conf.

Proc. 91, Ed. P. Channell (AIP, New York, 1982).

3. J.B. Rosenzweig, Phys. Rev. Lett. 58, 555 (1987).

4. J.B. Rosenzweig, B. Breizman, T. Katsouleas and J.J. Su, Phys.Rev.A 44, R6189
(1991).

5. J.B. Rosenzweig, in Proc. of the 1992 Linear Accelerator Conf., (AECL-10728,

Chalk River, 1993).

6. J.B. Rosenzweig, et al., Nuclear Instruments and Methods A 410 532 (1998).

7. N. Barov, J.B. Rosenzweig, M.E. Conde, W. Gai, and J.G. Power, Phys. Rev.
Special Topics – Accel. Beams 3 011301 (2000).

8. S. Lee, T. Katsouleas, R. Hemker and W. Mori, Phys. Rev. E 61, 7012 (2000).

9. J.D. Jackson, Classical Electrodynamics, 2nd Ed, (Wiley, New York, 1975).

10. N. Barov, et al., to be published.

11. M. Hogan, et al., Phys. Plasmas 7, 224 (2000).

12. D. Bruhwiler, et al., Proc. 2000 European Part. Accel. Conf., 877 (Austrian Acad.
Sci. Press, 2000).

13. L.D.Landau and E.M. Lifschitz, The Classical Theory of Fields, (Addison-Wesley,
Reading, Mass., 1971).

14. H. Raether, Springer Tracts in Modern Physics, 38, Ed. G. Hohler (Springer-
Verlag, Berlin, 1965), pp. 84-157.
 10

0.1
~
F

0.01
Sim. energy loss
Linear energy loss
0 Sim. peak acceleration
2
0.01 0.1 1 10 10
~
Q
Figure 1. The average normalized energy loss rate of F˜dec = e E z /me c p of an electron
beam with k p z = 1.1, k p a = 0.2, as a function of Q˜ , from linear theory and self-consistent
PIC simulation; also, the peak excited accelerating field, F˜max = e E z,max / mec p .