PHYSICAL REVIEW LETTERS week ending

PRL 95, 195002 (2005) 4 NOVEMBER 2005

Effects of Ion Motion in Intense Beam-Driven Plasma Wakefield Accelerators

J. B. Rosenzweig,1 A. M. Cook,1 A. Scott,1,2 M. C. Thompson,1 and R. B. Yoder1,3
1
Department of Physics and Astronomy, University of California Los Angeles, Los Angeles, California 90095, USA
2
Department of Physics, University of California Santa Barbara, Santa Barbara, California 93016, USA
3
Department of Physics, Manhattan College, 4513 Manhattan College Parkway, Riverdale, New York 10471, USA
(Received 27 June 2005; published 31 October 2005)
Recent proposals for using plasma wakefield accelerators (PWFA) as a component of a linear collider
have included intense electron beams with densities many times in excess of the plasma density. The
beam’s electric fields expel the plasma electrons from the beam path to many beam radii in this regime.
We analyze here the motion of plasma ions under the beam fields, and find for a proposed PWFA collider
scenario that the ions completely collapse inside of the beam. Simulations of ion collapse are presented.
Implications of ion motion on the feasibility of the PWFA-based colliders are discussed.

DOI: 10.1103/PhysRevLett.95.195002 PACS numbers: 52.40.Mj, 29.17.+w, 29.27.
a, 52.75.Di

The plasma wakefield accelerator (PWFA), driven in the nificantly during the beam passage. We will see below that
blowout regime [1] where the beam is denser than the this is indeed the case for the current proposal, known as an
ambient plasma (nb > n0 ), has been the subject of much ‘‘afterburner,’’ which uses a PWFA as a component in a
recent experimental and conceptual investigation [2–5] in LC. This proposal, which has been initiated by S. Lee,
the context of its application to a future high-energy linear et al. [6], uses a single stage PWFA deployed at the end of a
collider [6,7] (LC). In this regime, the plasma electrons are conventional high-energy linear accelerator. In this stage, a
ejected from the path of the intense driving electron beam, portion of the beam charge is used to drive the PWFA,
resulting in an electron-rarefied region. This region, con- allowing a trailing part of the beam to be doubled (or more)
taining only ions, possesses linear (in radius r) electrostatic in energy before use in collisions. The afterburner idea was
focusing fields that allow high quality propagation of both recently explored in the context of present LC complex
driving and accelerating beams. In addition, this electron- designs by Raubenheimer [7]. For the parameters in
rarefied region has superimposed upon it electromagnetic Ref. [7], which are partially based on the afterburner
fields, which, because the phase velocity of the axisym- scenario discussed in Ref. [6], our analysis will show
metric wake is nearly c, have longitudinal electric fields violent collapse of the ions. This collapse, which we also
essentially independent of r. This wake may accelerate a illustrate in particle-in-cell (PIC) simulations, has serious
trailing electron beam just as a traveling wave linear ac- implications for the preservation of the accelerating beam
celerator, with strong, linear transverse focusing conven- emittance, effectively negating the assumed advantage of
iently supplied by the plasma ions. linear transport in the blowout regime [1].
The existence of linear focusing in both the PWFA and Most previous analysis of the PWFA has been carried
in the related final-focusing underdense plasma lens [8] is out under the assumption of cylindrical symmetry in the
predicated on the assumption that the ions do not move, beam, and thus in the plasma response. We begin our
and thus maintain spatial uniformity. Without uniform ion discussion on this familiar ground, assuming (consistent
density ni , the strong ion-derived focusing fields will not with Refs. [6,7]), that the drive beam is indeed axisym-
give attractive beam transport characteristics. On the con- metric. For the purposes of analysis, we invoke some well
trary, strong nonlinear (in r) or time-dependent (changing accepted approximations to give the forms of the electro-
with longitudinal position in the beam,
 z
ct) fields static fields that serve to focus the beams and the ions. The
give certain degradation of the beam’s transverse phase first approximation is that the net transverse force on the
space density, which in the case of a LC must be of beam arises only from the ions’ electrostatic fields, which
unprecedented quality. Thus the issue of ion motion is of is a defining characteristic of the blowout regime [1]. The
critical importance in evaluating the viability of using the second is that the ions move predominately under the
PWFA in a collider. influence of the beam electrons’ transverse electric field.
The condition nb > n0 is inherent in the blowout regime. As the moving ions remain nonrelativistic, they are negli-
Indeed, for many scenarios of current interest, the self- gibly affected by the beam-derived magnetic field. These
consistent driving beam density nb , as well as that of the approximations are useful in both 2D and 3D analysis of
accelerating beam, exceeds n0 by orders of magnitude. the beam-plasma-ion interaction, and allow us to estimate
Under these circumstances, the beam’s electric field is the degree to which the ions move due to intense beam
high enough to produce relativistic plasma electrons fields in the PWFA.
[9,10], resulting in their ejection to radii large compared We analyze the cases of the driving and accelerating
to the rms transverse beam size x . If the beam fields are beam in different ways. The driver, as it is not used directly
large enough they may also induce the ions to move sig- in the LC experiment, may be taken as axisymmetric, as

0031-9007=05=95(19)=195002(4)$23.00 195002-1 © 2005 The American Physical Society

 PHYSICAL REVIEW LETTERS week ending
PRL 95, 195002 (2005) 4 NOVEMBER 2005

noted above. On the other hand, the accelerating beam beam-derived electric field is predominantly radial, and
must, because of the demands of the final focus and can be evaluated using Gauss’ law. In the beam core (r <
beam-beam interaction (e.g., beamstrahlung mitigation, x and
< z , where nb ’ nb;0 ) Er is nearly linear in r
crab-crossing), have asymmetric emittances and beam and independent of
,
sizes. We make use of the known form of the self-electric
eNb p













field inside such beams [11], and concentrate on the rele- Er ’
2enb;0 r 
r n r: (3)
vant vertical ion motion. For the example PWFA parame- n;x z e 0
ters, we assume the values of n0 , the charge and bunch
The movement of the nonrelativistic plasma ions is driven
lengths for the driving and accelerating beams to be those
mainly by Er , with approximate equation of motion de-
quoted in Ref. [7] for the 1 TeV afterburner based on a
rived from Eq. (3),
superconducting LC. These parameters, given in Table I,
are similar to those of Ref. [6], which contains analysis and d2 r ZeEr Ze2 Nb p













simulations underpinning the physical PWFA model in [7]. ’ 
r n r: (4)
dt2 Ama Ama n;x z e 0
While the afterburner design in [7] does not specifically
address the emittance of the drive beam, for definiteness Here Z is the ion charge state, and A the atomic mass in
we take the transverse normalized emittance of the axi- amu. In terms of the distance measured in the beam
symmetric drive beam to be the geometric mean of the Galilean frame
, Eq. (4) can be recast as
p















accelerating beam emittances, n  n;x n;y . We have
d2 r Zr N p













also assumed that the ionized species is hydrogen, to avoid
2

a b re n0 r 
k2i r; (5)
multiple ionization state and uncontrolled plasma forma- d
An;x z
tion inside of the beam [12]. where ra  1:55  10
18 m is the classical radius of a
Examining first the axisymmetric beam case, we note singly charged ion of mass 1 amu.
the matched  function due to the ion focusing is [13] Under the stated assumptions, Eq. (5) describes a driven,
q





















simple harmonic oscillator with spatial wave number ki 
eq  =2re n0 : (1) q
































Zra Nb =An;x z re n0 1=4 . For static initial conditions its
The equilibrium transverse rms beam size x associated
q

















solution is r  r0 coski
, with r0 the initial radial offset of
with this scenario is x;eq  eq n =, where   1 is the ion. To account for the variation of
with
, we take
nb;0
p






the beam energy normalized to me c2 (the beam is highly the beam effective length as

 2z , and the total
relativistic, with velocity v ’ c), and re is the classical phase advance of the ion motion in the beam’s field is
radius of the electron. In cases relevant to the afterburner, s

























nb  n0 , and blowout proceeds very quickly, giving 2Zra z Nb

 ’ ki

 re n0 1=4 : (6)
plasma electron rarefaction over nearly the full longitudi- An;x
nal beam extent. We may thus take, in the absence of ion
motion, x  x;eq over the whole beam. Upon insertion of drive beam parameters from Table I into
Assuming a bi-Gaussian (in r and
 z
ct) density Eq. (6), we obtain
 ’ 6:45. As total collapse of the ions,
distribution of Nb beam electrons, the peak, on-axis beam accompanied by a large near-axis spike in ion density ni ,
density is occurs for
  =2, this phase advance is an order of
magnitude too large for the assumption of uniform unper-
Nb Nb p












turbed ion density to hold. This result could be anticipated
nb;0   re n0 ; (2) p


























23=2 2x z 2n;x z by noting that the ratio ki =kp is ni me =2n0 mi . The exis-
tence of a beam with density nb;0  1:5  105 n0 , as in our
where z is the rms beam extent in
. With nb  n0 , the example, indicates that in a beam with kp z  1—appro-
priate for a PWFA driver [1,6]—and mi =me  1800, large
TABLE I. Beam and plasma parameters for linear collider amounts of ion motion (ki z  1) are expected. Possible
afterburner, derived from Ref. [7]. ways to mitigate this problem are discussed below.
In order to illustrate the severity of ion collapse, as well
Nb (drive, accelerating) 1:5  1010 , :5  1010
as aspects of the nonlinearity in the ion motion, we show
rms bunch length z 35 m
the results of axisymmetric PIC simulations performed
 (drive, accelerating)  1  106 ,  2  106
Accelerated beam "n;x;y 4  10
6 , 9:6  10
6 m rad
with the code OOPIC [12,14]. The beam parameters in
Drive beam "n;x 6:2  10
7 m rad this calculation are the same as in Table I (drive beam).
Initial ion (electron) density n0 0:9  1016 cm
3 Figure 1 shows ni in and near the beam region. It can be
Ion charge state Z 1 (hydrogen) seen, as expected for the case of such a large
, that the
Matched  function eq 3.1 cm ions indeed collapse very quickly. The ion density gener-
Normalized beam density nb =n0 1:5  105 ally rises as the beam current grows, increasing by a factor
of over 200 in the beam core. In this regime (
  1) the
195002-2
 PHYSICAL REVIEW LETTERS week ending
PRL 95, 195002 (2005) 4 NOVEMBER 2005

ions are released from the beam potential by a combination ent. Assuming the beam has elliptical symmetry, the trans-
of its time dependence, nonlinearity, and self-repulsion of verse electric fields (Ex and Ey ) are equal at the beam edges
the accumulating ions. (y  y , x  x
Ry ) [11]. Thus the ion motion com-
With such a large increase in ni , the self-consistent beam ponent contributing to the ion density perturbation is pre-
size should be reduced further. This process would take a dominantly vertical.
distance z > eq to establish, which is much longer than The vertical field inside of the beam core is, in the linear
our simulation length —the time-step needed in our approximation,
present calculation is 8 attoseconds. We have analyzed p













4enb;0 2eNb re n0 
the establishment of new equilibria under the joint evolu- Ey 
y
y (7)
tion of nb and ni in Ref. [15]; a computational approach to 1  R "n;y z 1  R
understanding this problem will be undertaken in the fu- s















s








ture. It should be noted, however, that there is violent 2eNb re n0  "n;x

y; R  1: (8)
transverse emittance growth associated with this process, z "n;y "n;x "n;y
due to both the
dependence and nonlinear r dependence
of the ion-derived fields. Indeed, the growth rate observed The linearized (for ions initially inside y < y , x < x )
in the simulation of Fig. 1 was disastrously high, vertical equation of motion is
d"n;x =dz ’ 6  10
4 m rad=m, giving 100% growth in s
















only 1 mm of propagation length. 2Zra Nb re n0 
y00 
y 
k2i;y y: (9)
We note that Ref. [6] deals with a case similar to that Az "n;y "n;x
given in Table I, but with even higher charge. There the
The focusing strength (k2i ) is a factor of 2 larger than in the
beam is also assumed round, with x  25 m, whereas
round beam case, if we assume for comparison that the flat
in our self-consistent example we have determined that p
















beam "n;y "n;x is equivalent to "n;x in the round beam—
x  140 nm. Thus our value of the ion focusing wave
number ki is over 100 times larger than that deduced from this is equivalent to requiring that nb;0 is the same in the
the assumed (not derived from a consistent set of beam round and flat beam cases. This factor arises in the flat
parameters) case of Ref. [6]. Even with
 0:1 deduced beam because the field varies strongly only in the vertical
for the case of [6], the ion motion is not negligible, and dimension. Thus in the flat beam scenario the ion collapse
occurs at a level that is also relevant to the accelerating is inherently faster, and even for the smaller Nb and higher
beam. As the ions move further after drive beam passage, energy (  2  106 ) accelerating beam in the Ref. [7],
the ion perturbation due to the drive beam will be stronger the maximum phase advance is
yet inside the trailing beam. v
u












































s

















u
t4Zra Nb z re n0 
The situation is more constrained for the accelerating
y  6:26: (10)
beams, which have emittance and charge requirements set A "n;y "n;x
by the luminosity of the collider. For beams inside of a
This is again unacceptably large, and should be mitigated
cylindrically symmetric ion channel that is preformed by
by over an order of magnitude in order to preserve the
the drive beam, one may assume that the equilibrium x
accelerating beam quality.
and y are equal and given by Eq. (1). Thus for the case in
One may ask if it is possible to choose parameters that
Ref. [7], the beam sizes x;y are a factor of R  10 differ-
ameliorate the ion motion problem, in either the drive or
the accelerating beam. The parameters n0 and z are not
p





actually independent, as n0 / kp / z ; also  is dictated
by the collider design. Thus the only feasible approach
would be to use smaller Nb and larger "n , as these effects
reduce
, if only as a square-root. One may not give up
Nb in the drive beam, however, without losing acceleration
gradient. On the other hand, the drive beam "n may be
made significantly larger, at the expense of ease in ma-
nipulating the beam; for example, if the emittance is too
large, one may not easily compress the beam to shorter
lengths. One must also then solve the problem of creating
the large emittance driver in the presence of a low emit-
tance trailing beam. The constraints of using the beam in
the collider interaction point are much more serious for the
accelerating beam, however. As one may not arbitrarily
choose Nb , "n;x , or R in the trailing beam, it is not likely
FIG. 1 (color). Surface plot of ion density distribution in 
; r, that the afterburner case discussed in Ref. [7] can be made
as simulated by OOPIC for drive beam conditions of Table I. feasible.
195002-3
 PHYSICAL REVIEW LETTERS week ending
PRL 95, 195002 (2005) 4 NOVEMBER 2005

Thus there seem to be two options that may be pursued. erators; in accelerator terminology, it is a generalized
The first is a complete redesign of the LC beam format to ’’transverse wake’’ problem.
accommodate the ion motion problem, in which case the Future work planned includes analysis of the emittance
afterburner concept changes from a possible post-design growth caused by nonuniform ion densities. This study
feature to an inherent design constraint. This option is still requires more in-depth simulations, which we intend to
extremely challenging, however, given the severity of the extend to three-dimensions to examine ion motion in flat
ion collapse scaling we have deduced here, and the con- beams. Further investigations are also planned to search for
straints of collider luminosity. A more radical solution collider design parameters which may be more compatible
would be to eliminate the ions altogether, using a hollow with ion motion than current afterburner-inspired schemes.
plasma capillary. This has already been proposed in the Such parameters are far from those presently under con-
context of accelerating positrons [6], where the transverse sideration; use of smaller Nb and z bunches seem most
wake is defocusing when plasma is allowed in the beam promising. It is clear in this regard that more attention
channel. should be paid to development of the hollow capillary
Obviously, the loss of ions in the beam path precludes version of the PWFA, which not only moots the ion motion
ion focusing of the electrons. It thus also presents an issue in the electron beam case, but provides for stable
obstacle to implementing another compelling aspect of acceleration of positrons as well [6]. Finally, we note that
the afterburner proposal—the use of the plasma lens final experimental tests of ion motion would be desirable. As
focusing [8]. Two scenarios exist for plasma lens final such, recent PWFA experiments at Stanford [18] employ
focusing: one in which thin lenses are used, and another the parameters which indicate that
  0:3 has been
in which the beam is adiabatically focused [16] by a achieved.
steadily increasing n0 in z, to avert beam size limitations This work was performed with support from the US
due to synchrotron radiation (the Oide limit [17]). In the Department of Energy under Contract No. DE-FG03-
first case, if one focuses directly after the PWFA, n0 must 92ER40693.
be denser than assumed in the accelerating section. The
denser plasma produces stronger focusing than in the up-
stream equilibrium, and thus yields beam demagnification. [1] J. B. Rosenzweig, B. Breizman, T. Katsouleas, and J. J. Su,
However, this scenario does not provide any mitigation of Phys. Rev. A 44, R6189 (1991).
the ion collapse problem. One is forced to consider signifi- [2] P. Muggli et al., Phys. Rev. Lett. 93, 014802 (2004).
cant beam expansion before the plasma lens to avoid ion [3] M. J. Hogan et al., Phys. Rev. Lett. 90, 205002 (2003).
motion and concomitant aberrations. [4] B. Blue et al., Phys. Rev. Lett. 90, 214801 (2003).
In the adiabatic focusing case, the rise in plasma density [5] N. Barov et al., Proceedings of the 2001 Particle
Accelerator Conference, Chicago, IL, edited by P. Lucas
occurs slowly, with nb increasing as n1=2 0 . The final beam and S. Webber (IEEE, Piscataway, NJ, 2001), pp. 128.
density thus significantly increases, in a scenario where any [6] S. Lee et al., Phys. Rev. ST Accel. Beams 5, 011001
increase at all exacerbates an already unacceptable level of (2002).
ion motion. To illustrate this situation, one may envision [7] Tor O. Raubenheimer, in 11th Advanced Accelerator
achieving minimum x and y at the collider interaction Concepts Workshop, AIP Conf. Proc. No. 737, (AIP,
point of z (limited by the ‘‘hour-glass’’ effect) through New York, 2004), p. 86.
adiabatic plasma lensing. The associated beam in this case [8] J. J. Su et al., Phys. Rev. A 41, 3321 (1990).
[9] N. Barov, J. B. Rosenzweig, M. C. Thompson, and R. B.
has nb  1:6  1024 cm
3 ; ki increases by nearly a factor
Yoder, Phys. Rev. ST Accel. Beams 7, 061301 (2004).
of 104 over the accelerating beam case discussed above. [10] J. B. Rosenzweig, N. Barov, M. C. Thompson, and R. B.
This degree of ion motion negates the utility of adiabatic Yoder, Phys. Rev. ST Accel. Beams 7, 061302 (2004).
plasma focusing. [11] P. M. Lapostolle, IEEE Trans. Nucl. Sci. 18, 1101 (1971).
In conclusion, we have analyzed ion motion in likely [12] D. L. Bruhwiler et al., Phys. Plasmas 10, 2022 (2003).
scenarios where the PWFA is used as an afterburner accel- [13] N. Barov and J. B. Rosenzweig, Phys. Rev. E 49, 4407
erator in a future linear collider, and have found that the (1994).
assumption of stationary ions which underpins the physics [14] J. P. Verboncoeur et al., Comput. Phys. Commun. 87, 199
model of the scheme is strongly violated. The subsequent (1995).
[15] J. B. Rosenzweig, A. M. Cook, M. C. Thompson, and R. B.
ion motion can produce extremely large perturbations in
Yoder, in 11th Advanced Accelerator Concepts Workshop
the ion density, giving rise to transverse fields that disrupt (Ref. [7]) p. 907.
the beam motion. As the ion motion due to the accelerating [16] P. Chen, K. Oide, A. M. Sessler, and S. S. Yu, Phys. Rev.
beam itself is seen to be extremely large for currently Lett. 64, 1231 (1990).
conceived collider beam parameters, one must also exam- [17] K. Oide, Phys. Rev. Lett. 61, 1713 (1988).
ine the issue of ion motion in laser-driven plasma accel- [18] M. J. Hogan et al., Phys. Rev. Lett. 95, 054802 (2005).

195002-4