See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/369198928

Effects of the transversely nonuniform plasma density in a blowout regime of a

plasma wakefield accelerator

Preprint · March 2023

DOI: 10.48550/arXiv.2303.07253

CITATIONS READS

0 12

1 author:

Stanislav Baturin
ITMO University
72 PUBLICATIONS   522 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

UNCD field emitters View project

Wakefields in dielectric loaded waveduides View project

All content following this page was uploaded by Stanislav Baturin on 14 March 2023.

The user has requested enhancement of the downloaded file.

 Effects of the transversely nonuniform plasma density in a blowout regime of a plasma
wakefield accelerator
S. S. Baturin∗
School of Physics and Engineering, ITMO University, St. Petersburg, Russia 197101
(Dated: March 14, 2023)
We present an analytical study on the effects of the transverse plasma gradient in the blowout
regime of a plasma wakefield accelerator (PWA). The analysis departs from a simple ballistic model
of plasma electrons that allows us to derive a complete analytic solution for the pseudopotential
and, consequently, for the wakefield. We demonstrate that the transverse plasma gradient modifies
the bubble shape and affects the wakefield. Namely, the dipole plasma gradient results in a dipole
component of the wakefield. Analysis suggests that, despite the asymmetry, the instability due to
arXiv:2303.07253v1 [physics.acc-ph] 13 Mar 2023

the fixed transverse plasma gradient is unlikely, as the total wakefield has a single stable point inside
the bubble. The only effect that occurs is the shift of the electromagnetic center. We point out that
random fluctuation of the transverse plasma gradient could become an issue.

I. INTRODUCTION In contrast to the Ref.[24], where it was demonstrated

Beam breakup (BBU) and emittance degradation are
one of the main challenges on the way to the high lumi-
nosity collider based on a wakefield accelerator technol-
ogy. [1–6]. Among others, [1–3] plasma-based wakefield
accelerators (PWA) are considered the most promising
candidate for the particle accelerators of the future [7–
12] as it possesses intrinsic focusing mechanism provided
by the ions and extreme accelerating gradients at the
same time [13, 14]. In particular, in a so-called bubble
regime, when plasma electrons are expelled, due to the
high intensity of either laser or electron driver longitu-
dinal (accelerating) electric field could be ∼ 50 GV/m
[15, 16]. Accelerating gradient is tightly connected with
the transverse wakefield by the means of the Panofsky-
Wenzel theorem [17]. When the longitudinal wake is
axisymmetric, the transverse wake is zero, but even a
tiny asymmetry is enough to seed the instability, that is
caused by the transverse wake. The latter is connected
with the fact that projected transverse beam emittance
grows exponentially due to the BBU [18]. Moreover, even
a short noise (a small random force or random asymme-
try in the beam distribution) triggers the BBU.
Despite such a strong limitation, it tums out that
BBU could still be controlled. Many mechanisms and
approaches, including ion motion [19], BNS damping
[20, 21], and other methods of instability suppression
have been investigated extensively. It was demonstrated
that BBU is not just suppressed but eliminated in a cer-
tain parameter range.
In the present study, we investigate how the local
transverse asymmetry of the plasma gradient affects the
wake and analyze these results from a beam dynamics
perspective. The analysis is based on the ballistic model
of the plasma electrons introduced in Ref.[22, 23]. To
an extent, present calculations echo the analysis of the
Ref.[24], where a flat bubble formation was investigated.
that the wakefield is insensitive to the transverse plasma
gradient, we show that in a commonly considered round
bubble, transverse plasma gradient results in a transverse
wake. Within the considered approximation we analyze
plasma flow, derive an analytic expression for the pseu-
dopotential, and provide a discussion of the possible con-
sequences.
The paper is organized as follows. In Sec.II, for
the sake of convenience, we reproduce basic formulas
and equations provided in Refs.[22–24] that we will use
throughout the paper. In Sec.III we derive the expres-
sion for the electromagnetic shock wave produced by the
driver, and in Sec.IV we present an expression that de-
scribes the bubble shape within ballistic approximation.
In Sec.VI plasma electron density is derived, and the out-
come of this section is then used in Sec.VII to get to the
main result of the paper - the expression for the pseu-
dopotential.
II. BASIC EQUATIONS
We start from the set of equations derived in Ref.[23].
We use the same convention as the Ref.[23] and we use di-
mensionless variables: time is normalized to ωp−1 , length
to kp−1 , velocities to the speed of light c, and momenta
to mc. We also normalize fields to mcωp /e, forces to
mcωp , potentials to mc2 /e, the charge density to n0 e,
the plasma density to n0 , and the current density to en0 c.
With e being the elementary charge, e > 0.
The equation of motion for the plasma electrons could
be written as
dp dr p
= ∇ψ − v × B, = . (1)
dt dt γ

Here
p p is the momentum of the plasma electrons, γ =
1 + p2 is the relativistic gamma factor of the plasma
∗ s.s.baturin@gmail.com electrons, v = p/γ is the velocity and ψ = φ − Az is the
 2

pseudopotential that defines the wake field as localized in an infinitesimally thin layer, i.e. have a delta-
function discontinuity
∂ψ
Ez = , F⊥ = −∇⊥ ψ, (2)
∂ξ
E⊥ = Dδ(ξ),
and ∇ = (∂x , ∂y , −∂ξ ). Here F⊥ is the transverse part of B⊥ = ẑ × Dδ(ξ). (13)
the Lorentz force per unit charge of the test particle and
ξ = t − z. The transverse profile of these fields is defined by the
Eqs.(1) have the following integral of motion 2D vector D.
To solve for the shock wave at ξ = 0, we assume that
γ − pz − ψ = 1, (3) the plasma density in front of the moving driver has a
linear gradient
as a consequence we have
n0 = 1 + gx, (14)
1+ψ
1 − vz = . (4)
γ where the uniform part of the density is 1, g is a constant,
and x is the transverse coordinate. We assume a small
In a quasi-static picture, it is convenient to replace the gradient,
derivative by time t with the derivative by ξ. We use the
fact that g  1, (15)
dξ
= 1 − vz , (5) and use the perturbation theory.
dt We consider equation for the vector D that according
consequently for an arbitrary function f (ξ) we have. to Ref.[23] reads
n0
df df dξ df 1 + ψ df ∆⊥ D = D. (16)
= = (1 − vz ) = . (6) γ0
dt dξ dt dξ γ dξ
If we split D into r and φ component then Eq.(16) could
Since in the quasi-static picture momentum of the plasma be written in expanded form as
electron is a function of ξ Eqs.(1) with Eq.(6) are reduced
to Dr 2 ∂Dφ n0
∆⊥ D r − 2
− 2 = Dr ,
dp⊥ γ Bz r r ∂φ γ0
= [∇⊥ ψ + ẑ × B⊥ ] − p⊥ × ẑ. (7) Dφ 2 ∂Dr n0
dξ 1+ψ 1+ψ ∆⊥ Dφ − 2 + 2 = Dφ , (17)
r r ∂φ γ0
Equation for the pseudopotential reads
with the Laplace operator given by
∆⊥ ψ = (1 − vz )ne − ni (x), (8)
1 ∂2
 
1 ∂ ∂
∆⊥ = r + 2 2. (18)
here ne is the plasma electron density and ni (x) is the r ∂r ∂r r ∂φ
ion density that depends on x. In what follows, we will
assume that If we assume that n0 is a constant then due to the ax-
ial symmetry we have D = Dr (r)r and Dφ = 0. This
ni (x) = 1 + gx, (9) immediately results in the equation for D(r) in a form
 
with g  1. Equations for the magnetic field are 1 ∂ ∂Dr 1 n0
r − 2 Dr = Dr . (19)
r ∂r ∂r r γ0
∆⊥ Bz = ẑ · (∇⊥ × ne v⊥ ), (10)
With the unmodified plasma density, initial gamma set
∆⊥ B⊥ = −ẑ × ∇⊥ ne vz − ẑ × ∂ξ ne v⊥ . (11)
to unity (n0 = 1, γ0 = 1) and boundary conditions
The continuity equation reads Dr (∞) = 0, Dr (r → 0) = 2ν/r we have

∂ξ [ne (1 − vz )] + ∇⊥ · ne v ⊥ = 0. (12) Dr (r, φ) = 2νK1 (r). (20)

Next, we consider n0 as given by Eq.(14). We apply

perturbation theory and seek a solution of the Eq.(17) in
III. SHOCK WAVE
a form

To calculate the field distribution that is produced by Dr = Dr(0) + Dr(1) , (21)

the point driver that travels through plasma, we follow (0) (1)
Refs.[22, 23]. Namely, we assume that driver fields are Dφ = Dφ + Dφ .
 3

(0) (0) (1)

With the Dr = 2νK1 (r), Dφ = 0 and Dr,φ ∼ g - small Inverse Hankel transform gives
corrections. Substituting Eqs.(21) and Eq.(14) into the
k2
 
Eqs.(17), equating terms of the same order and account- −1 νg
Ψ = −4νgH2 2 3
= − r2 K0 (r),
ing for x = r cos(φ) we arrive at a set of equations for (1 + k ) 2
corrections in the form
 
1 νg
Φ = −4νgH0−1 = − r2 K2 (r). (30)
(1) (1) (1 + k 2 )3 2
Dr 2 ∂Dφ
∆⊥ Dr(1) − 2 − 2 = Dr(1) + 2νgK1 (r)r cos(φ),
r r ∂φ D1r and D1φ are recovered as
(1) (1)
(1) Dφ 2 ∂Dr (1)
∆⊥ Dφ − + = Dφ . (22) Ψ+Φ
r2 r2 ∂φ D1r = , (31)
2
(1) (1) Ψ−Φ
Next we decompose Dr and Dφ in a Fourier series D1φ = , (32)
2
X
Dr(1) (r, φ) = D̃nr (r) cos(nφ), (23) and reads
(1) νg 2
X
Dφ (r, φ) = D̃nφ (r) sin(nφ), (24) D1r = − r [K0 (r) + K2 (r)] , (33)
4
and substitute this decomposition into Eqs.(22). Equat- νg
D1φ = − r2 [K0 (r) − K2 (r)] . (34)
ing amplitudes of corresponding cosines and sines we have 4

1 ∂

∂D1r

2 2 Finally, first order corrections could be written as
r − 2 D1r − 2 D1φ = D1r + 2νgK1 (r)r,
r ∂r ∂r r r νg 2
" # Dr(1) = − r [K0 (r) + K2 (r)] cos(φ), (35)
1 ∂ ∂D1φ 2 2 4
r − 2 D1φ − 2 D1r = D1φ . (25) Dφ
(1) νg
= − r2 [K0 (r) − K2 (r)] sin(φ). (36)
r ∂r ∂r r r 4
Next we introduce new functions Ψ = D1r + D1φ and Φ =
D1r − D1φ . Adding and subtracting Eqs.(25) we get IV. SHAPE MODIFICATION OF THE PLASMA
BUBBLE
 
1 ∂ ∂Ψ 4
r − 2 Ψ = Ψ + 2νgK1 (r)r,
r ∂r ∂r r We neglect the effect of the plasma self-fields on the
1 ∂

∂Φ
 trajectories of the plasma electrons. This is a “ballis-
r = Φ + 2νgK1 (r)r. (26) tic” regime of plasma motion introduced in Ref.[22]; it
r ∂r ∂r
assumes that the plasma electrons are moving with con-
A general solution to the equations above is zero as none stant velocities.
of the functions fulfill boundary conditions Ψ(∞) → 0, We assume plasma electrons to be non-relativistic and
Φ(∞ → 0) and Ψ(0) < ∞, Φ(0) < ∞. A specific solution vz0 ≈ 1 and ν  1 as well as ν  r < 1. With help of
on the other hand does not vanish and could be found Eq.(20) and Eq.(35) we may write equations of motion
via Hankel transformation. Forward and inverse Hankel for the plasma electrons as
transformation of the order n of some function f (r) are
given by dx 2ν ν
≈ cos(φ0 ) − g ,
dξ r0 2
Z∞
dy 2ν
fˆ(k) = Hn [f (r)] ≡ rf (r)Jn (kr)dr, (27) ≈ sin(φ0 ). (37)
dξ r0
0
h i Z∞ Solution to the equations above gives electron trajectories
f (r) = Hn−1 fˆ(k) ≡ k fˆ(k)Jn (kr)dk. (28)  
cos(φ0 ) g
0 x = r0 cos φ0 + 2νξ − ,
r0 4
Here Jn (kr) is the Bessel function of the fist kind of the sin(φ0 )
order n. y = r0 sin φ0 + 2νξ . (38)
r0
We apply H2 to the fist equation and H0 to the second
equation and get From Eqs.(38) one can deduce (see Appendix A) an equa-
2
tion of an envelope surface that defines the boundary of
k the bubble in the ballistic approximation in a form
Ψ̃ = −4νg ,
(1 + k 2 )2
 2
1 1
Φ̃ = −4νg . (29) x + νξg + y 2 = 8νξ. (39)
(1 + k 2 )2 2
 4

3 3
Here as before v ⊥ is the vector of the transverse velocity
2 2 of the plasma electrons that according to the Eq.(37)
reads
1 1
ϕ=0 ϕ=π/8
 T
0 0 2ν ν 2ν
v⊥ = cos(φ0 ) − g , sin(φ0 ) . (44)
-1 -1 r0 2 r0
-2 -2 We notice that
-3 -3 ∂ ∂
0.5 1 0.5 1
= ,
ξ ξ ∂ x̃ ∂x
3 3 ∂ ∂
= . (45)
∂ ỹ ∂y
2 2

1
ϕ=3π/8 1
ϕ=π/2
Next, if we assume that n = n(x̃, ỹ, ξ) then Eq.(43)
should be modified as
0 0
ν
-1 -1 ∂ξ ne + g ∂x̃ ne + vx ∂x̃ ne + vy ∂ỹ ne
2
-2 -2 + ne ∂x̃ vx + ne ∂ỹ vy = 0. (46)
-3 -3
0.5 1 0.5 1  T
Introducing ṽ⊥ = 2ν
cos(φ0 ), 2ν ˜⊥ =
, ∇
ξ ξ r0 r0 sin(φ0 )
(∂x̃ , ∂ỹ ) and noting that
Figure 1. Plasma flow in the ballistic approximation for the
case of ν = 1 and g = 0.8. The numbers are far from realistic ∂x̃ vx = ∂x̃ ṽx , (47)
and are chosen to emphasise the effect visually. The bubble
is ”bended” towards lower plasma density. we have
˜ ⊥ · ne ṽ⊥ = 0.
∂ξ ne + ∇ (48)
It could be seen from the equation above that the circular
cross sections of the bubble in x, y plane are shifted in We notice that in the new coordinates the plasma flow
the x direction by the distance νξg/2 that is linearly (in particular velocity field) has a cylindrical symmetry
increasing with ξ. and thus continuity equation has a cylindrical symmetry
We introduce
p the coordinates x̃ = x + gνξ/2, ỹ = y as well and could be written as
and r̃ = x̃2 + ỹ 2 and the angle φ̃ = arccos x̃/r̃. Then
Eqs.(38) can be written as 1 ∂
∂ξ ne + r̃ne ṽ⊥ = 0. (49)
r̃ ∂ r̃
ξ
r̃ = r0 + 2ν , φ̃ = φ0 . (40) 2ν
r0 with ṽ⊥ = r0 .

Condition dr̃/dr0 = 0 gives equation for the bubble

boundary VI. PLASMA DENSITY IN THE BALLISTIC
p APPROXIMATION
r̃b (ξ) = 2 2νξ, (41)
in full agreement with Ref.[22]. Switching back to the Immediately behind the driver, at ξ = 0+ , the plasma
(x, y) we arrive at the final expression for the bubble density n0 is given by Eq.(14). If we assume that electron
boundary in the form trajectories are known then from the continuity of the
p νξ plasma flow we conclude that
rb (ξ, φ) = 2 2νξ − g cos (φ) . (42)
2 n(x, y, ξ)dS = n0 (x0 , y0 )dS0 , (50)

from which it follows that

V. AXIAL SYMMETRY OF THE CONTINUITY
dS0
EQUATION IN THE BALLISTIC n(x, y, ξ) = n0 (x0 , y0 ) . (51)
APPROXIMATION dS
the ration dS/dS0 is calculated though the Jacobian
First, we refer to the continuity equation (12) that in  
the ballistics approximation takes the form dS  ∂x ∂x 
 ∂x0 ∂y0 
=  ∂y ∂y  . (52)
∂ξ ne + ∇⊥ · ne v ⊥ = 0. (43) dS0  ∂x0 ∂y0 
 5

Near the bubble

√ boundary, as far as maximum bubble ra- As expected the first term in Eq.(58) coincides with the
dius rbm ∼ ν and ν  1, we can use electron trajecto- electron density in the uniform case given in Ref.[22]
ries as defined by Eq.(40). Accounting for the cylindrical while the second term provides correction that accounts
symmetry in the (x̃, ỹ) coordinates we get for the initial transverse plasma gradient.
VII. PSEUDOPOTENTIAL
r0 dr0
n(r̃, φ̃, ξ) =n0 (r0 , φ0 ) =
r̃dr̃
Equation for the pseudopotential under the assump-
r3 tion of the non-relativistic plasma flow (vz  1) reads
n0 (r0 , φ0 ) 2 0 , (53)
r̃|r0 − 2νξ|

where we have used Eq.(40) to calculate dr̃/dr0 . ∆⊥ ψ = n e − n i , (59)

The initial radius r0 in this equation should be ex-
pressed through r and ξ from Eq. (40): here ne is the plasma electron density and ni is the ion
r density that depends on x. In what follows, we will as-
± 1 1 2 1 √
sume that
r0 = r̃ ± r̃ − 2νξ = r̃ 1 ± 1 − t , (54)
2 4 2
ni (x) = 1 + gx. (60)
where
8νξ r̃b2 We assume r0  p1 and r  1 switch to x̃ = x + gνξ/2,
t= = < 1, (55)
r̃2 r̃2 ỹ = y and r̃ = x̃2 + ỹ 2 and the angle φ̃ = arccos x̃/r̃,
√ account for the Eq.(57) and Eq.(60) and rewrite Eq.(59)
with r̃b ≡ 2 2νξ as given by Eq.(41). in the extended form
Two solutions correspond to two trajectories that ar-
rive from different initial radii r0 , φ0 to a given point 1 ∂

∂ψ

1 ∂2ψ 2r̃2 − r̃b2
r̃, φ̃, ξ. One of these trajectories arrives before and the r̃ + 2 = p θ(r̃ − r̃b ) − 1 (61)
r̃ ∂ r̃ ∂ r̃ r̃ ∂ φ̃2 2r̃ r̃2 − r̃b2
other one after it touches the envelope. Correspondingly, !
at a given ξ, r̃, φ̃, we need to sum the two densities for 4r̃2 − 3r̃b2
both trajectories. + gr̃ cos φ̃ p θ(r̃ − r̃b ) − 1 ,
4r̃ r̃2 − r̃b2
Substituting (54) into (53) we obtain
√
3
1 1± 1−t Next we decompose ψ in in a Fourier series
±
n± (r̃, φ̃, ξ) =  √  (1 + gr0 cos φ0 ),
2  1 ± 1 − t 2 − t


X  
(56) ψ(r̃, φ̃) = ψ̃n (r̃) cos nφ̃ . (62)

where we have used Eq. (14) for n0 . In this formula we Consequently we have
have to express r0 through r̃ using Eq. (54) and also use
φ̃ = φ0 . For the total density, after some simplifications, " #
we find 1 ∂ ∂ ψ̃0 2r̃2 − r̃b2
r̃ = p θ(r̃ − r̃b ) − 1, (63)
r̃ ∂ r̃ ∂ r̃ 2r̃ r̃2 − r̃b2
n(r̃, φ̃, ξ) = n+ (r̃, φ̃, ξ) + n− (r̃, φ̃, ξ)
1 2−t 4 − 3t and
= √ + gr̃ √ cos φ̃
2 1−t 4 1−t
" #
2r̃2 − r̃b2 4r̃2 − 3r̃b2 1 ∂ ∂ ψ̃1 ψ̃1
= p + g p cos φ̃. (57) r̃ − 2 = (64)
2r̃ r̃2 − r̃b2 4 r̃2 − r̃b2 r̃ ∂ r̃ ∂ r̃ r̃
!
As a consequence of the singularity in the shock wave, 4r̃2 − 3r̃b2
gr̃ θ(r̃ − r̃b ) − 1 .
plasma density has a square root singularity at the
p
4r̃ r̃2 − r̃b2
boundary of the bubble.
We switch back to the initial coordinates r, φ, ξ and
keep only terms of the order g: We introduce new normalized radius κ = r̃/r̃b and
rewrite Eq.(63) and Eq.(64) in the universal form
2r2 − r̃b2
n(r, φ, ξ) = p " #
2r r2 − r̃b2 2κ2 − 1
 
1 ∂ ∂ ψ̃0
κ = r̃b2 √ θ(κ − 1) − 1 , (65)
r̃b6 κ ∂κ ∂κ
 
cos φ 2 2

2 2

2κ κ2 − 1
+g 3/2
4r − 3r̃b r − r̃b − 2 . (58)
4 (r2 − r̃b2 ) 8r
 6

and

" #
4κ2 − 3
 
1 ∂ ∂ ψ̃1 ψ̃1 3
κ − 2 = r̃b gκ √ θ(κ − 1) − 1 . (66)
κ ∂κ ∂κ κ 4κ κ2 − 1

Equations above could be integrated and the solutions read

1 2 2

ψ̃0 (κ) = 1 2
√a1 + a2 log κ − 14 r̃2b κ κ < 1√
, (67)
a3 + a4 log κ + 4 r̃b κ( κ − 1 − κ) − 4 r̃b log κ + κ2 − 1 1 < κ  ∞
2

and
b2 1
( 3 3
ψ̃1 (κ) = n b1 κ + κ√− 8 gr̃b κ κ h< 1√ io . (68)
b4 1 3 κ+√κ2 −1
b3 κ + κ − 16 r̃b gκ 2κ2 − 2κ κ2 − 1 + log κ− 2
κ −1
1<κ∞

Here ai and bi are constants that could be found from the condition ψ(r = 0) < ∞ for ξ 6= 0 and continuity of the
pseudopotential and its derivative at the bubble
 boundary.
First, we consider the monopole part ψ̃0 r̃r̃b Eq.(67). Condition ψ̃0 (r̃ = 0) < ∞ for ξ 6= 0 leads to a2 = 0,
continuity of the potential gives a3 = a1 and continuity of the derivative requires a4 = 0. Consequently we arrive at
 2
   a1 − r̃4 r̃ < r̃b
r̃  r 
2
!
ψ̃0 = r̃
p
2 2 r̃b2 r̃ r̃ . (69)
r̃b  a1 + 4 ( r̃ − r̃b − r̃) − 4 log r̃b +
 r̃b −1 r̃b < r̃  ∞

 
Next, we consider the dipole part ψ̃1 r̃r̃b Eq.(68). Condition ψ̃1 (r̃ = 0) < ∞ for ξ 6= 0 leads to b2 = 0 continuity
of the potential gives b1 = b3 + b4 and continuity of the derivative requires b1 = b3 − b4 . Consequently b4 = 0, b3 = b1
and we arrive at
b1 r̃r̃b − 81 gr̃3 r̃ < r̃b

  
r̃  √ 
ψ̃1 = r̃ 1
p r̃+ r̃ 2 −r̃b . (70)
r̃b 2 2 2
 b1 r̃b − 16 gr̃ 2r̃ − 2r̃ r̃ − r̃b + log √ r̃b < r̃  ∞
r̃− 2
r̃ −r̃b

We note that at large r̃  r̃b particular solution for the turbed, and plasma electrons screen the field that arises
r̃ 2 from the bubble. This, in turn, results in the vanishing
ψ̃0 diverges as ∼ − 4b log r̃ and particular solution for the
r̃ 2 g of the pseudopotential. To account for this and estimate
ψ̃1 diverges ∼ − b8 r̃ log r̃ (See Appendix B for the de- the remaining unknown constants a1 and b1 we request
tails). This is connected with the fact that we neglected pseudopotential to be zero at r = 1.
screening effects in the considered approximation. Fist, we notice that at r  r̃b expressions for the ψ̃0
We switch back to the original coordinates r and φ. and ψ̃1 reduces to
Noticing that cos φ = cos φ̃ + O[g] and accounting for the
r̃b2 r̃b2
     
fact that ψ̃(κ) ∼ O[g] one may write r r r̃b
ψ̃0 = a1 − − log 2 +O , (72)
    r̃b 8 4 r̃b r
r r̃b r
+ g ψ̃00 gr̃b2 gr̃b2
     
ψ(r, φ, ξ) =ψ̃0 cos φ r r r r̃b
r̃b 16 r̃b ψ̃1 = b1 − r− r log 2 +O .
  r̃b r̃b 16 8 r̃b r
r
+ψ̃1 cos φ + O[g 2 ]. (71) Next, keeping only divergent terms and setting r = 1 we
r̃b
arrive at
We observe that, as expected, pseudopotential consists r̃b2
 
2
of two a1 ≈ log ,
 parts: monopole - that corresponds to the term 4 r̃b
ψ̃0 r̃rb and dipole - that is a combination of the total gr̃3
 
2
b1 ≈ b log . (73)
by κ of the monopole term ψ̃00 (κ) and a cor-
derivative  8 r̃b
rection ψ̃1 r̃rb . Following Eqs.(69) and (70) with Eq.(71) and Eq.(73)
At distances, r ∼ 1 plasma density should be unper- we arrive at the final expression for the pseudopotential
 7

inside the bubble in the form � �

r̃b2 r2
 
2 0.22 0.42 ξ=0.01
ψ(r, φ, ξ) ≈ log −
4 r̃b 4
0.487
� �

��

��
0.3
2
r3
    0.46
r̃b cos φ r 2
−g − r log + 2 . (74)
8 4 r̃b r̃b 0.22
0.38

For the analysis it is convenient to normalize pseudopo- -� -�

tential to 1/r̃b2 . We recall the definition of r̃b given by -� � � -� � �
Eq.(41) and write the expression for the normalised pseu- �� ��
dopotential
� -0.05
�
ψ 1 x2 + yn2
2 = − log (2νξ) − n 0.05 ξ=0.1
r̃b 8 4 0.15
√  
2νξ 1 1 �
0.2
�

��

��
+ log (2νξ) + x2n + yn2 ,
0 -0.05
−g xn (75)
4 4 2 0.18
0.1

with xn = x/rb and yn = y/rb . -�

-� -0.05

-� � � -� � �
VIII. ANALYSIS �� ��

With the help of the Eq.(2) and Eq.(74) transverse � �

-0.1

part of the Lorentz force per unit charge of the negatively -0.27 ξ=0.4
-0.022
charged test particle could be evaluated as
-0.022 0.025
� �
��

��
0.01 0
r̃b2 1 3x2 + y 2
   
x 2 0.02

Fx = − − g − log + ,
2 8 4 r̃b r̃b2 -0.05
-0.03 -0.27

y xy -�
Fy = − − g . (76) -�
2 4
-� � � -� � �
It is convenient to normalize it to 1/r̃b and present in �� ��
terms of xn = x/r̃b and yn = y/r̃b
√ Figure 2. Contour plot for the normalized pseudopotential
ψ/r̃b2 given by Eq.(75) (left column) and normalized trans-
 
Fx xn 2νξ 1 1 2 2
=− −g + log (2νξ) + 3xn + yn , verse wakefield F⊥ /r̃b vector field given by Eq.(77) (right col-
r̃b 2 4 4 2
umn) for three different values of the longitudinal coordinate
Fy yn p xn yn ξ and ν = 1, g = 0.8. Red dots indicate fixed points of the
= − − g 2νξ . (77)
r̃b 2 2 vector field given by Eq.(78). We note that xn = x/r̃b and
yn = y/r̃b .
We note that pseudopotential has a cubic term in xn that
naturally leads to two fixed points of the vector field (one
stable and one unstable). By setting Fx = 0 and Fy = 0
one may find fixed points of the transverse wakefield by fixed points of the vector field Eq.(78) for three differ-
solving the corresponding algebraic system that follows ent values of the longitudinal coordinate ξ. We chose ex-
from Eq.(77): treme (and probably unreachable in practice) parameters
of g = 0.8 and ν = 1 to emphasize the effect. We observe
ynF = 0, that at small values of ξ, where the model is directly
√ applicable, only one stable fixed point exists within the
g νξ
xSn = − √ [1 + 2 log (2νξ)] , (78) bubble cross-section. Transverse gradient shifts the elec-
4 2 tromagnetic origin towards the higher densities of the ion
√ √
uS 2 g νξ column, but the net effect remains focusing albeit asym-
xn = − √ + √ [1 + 2 log (2νξ)] . metric. Further increase in ξ does not change the picture.
3g νξ 4 2
The asymmetry in the focusing grows, but the structure
To simplify the final formula, we kept only termsof the of the wake remains the same. Interestingly, if we spec-
order g, i.e. we disregarded terms of the order O g 2 in ulate and go beyond the formal applicability of the con-
Taylor decomposition. sidered model. We may observe the situation when both
In Fig.2 we show level sets of the pseudopotential given stable and unstable fixed points are located inside the
by Eq.(75), transverse wakefield vector field Eq.(77) and bubble cross-section. We point out, that despite the com-
 8

plex structure of the pseudopotential, the stable region
(the region where the beam is attracted to the stable
fixed point) occupies more than half of the bubble cross-
section even in this unrealistic scenario. The latter in-
dicates that most likely fixed transverse plasma gradient
(a transverse plasma gradient that does not change in z)
should not affect the driver dynamics (at least within con-
sidered approximation) and only results in some asym-
metric distortion of the bubble shape and wake.
It is worth mentioning that the plasma gradient may
fluctuate randomly due to the random fluctuations of the
plasma density. Such random fluctuation will result in a
random kick. It is well known (see Refs.[25, 26]) that ran-
dom kicks may lead to emittance growth and potentially
may lead to driver instability. Indeed, in 1D emittance
growth (see for instance Ref.[27]) due to the random kick
reads
δε hx2 ihFx2 i
= . (79)
δs ε
Here ε is the beam emittance and τ is the characteristic
time of the fluctuation. Following Eq.(77) we can write
hFx2 i ∼ hg 2 i ∼ hn2 i − n20 . (80)
Consequently, the dispersion of the density fluctuation
sets the growth rate for the emittance. This observation
motivates further studies in more realistic scenarios by
either applying Lu model [28] or proper extension of a
numerical simulation [29–34].
Despite the restrictions outlined above, the model pre-
sented is still useful, as it is complementary to the Lu
model of the plasma bubble [28]. The model presented
could be ”merged” with the Lu model such that the re-
sults of the ballistic model may serve as an initial condi-
tion for the Lu equation. A combined model will be free
of the empiric parameters and cover the whole range of
the driver beam intensities.
The equation derived in the present paper could be
used as a crude estimate for the transverse emittance
growth due to the random fluctuations of the plasma den-
sity.
Appendix A: Envelope surfaces for the ballistic
trajectories
First, we notice that if g = 0, then as was shown in
Ref. [22], for a given ξ, the map Eq.(38) when r0 varies
from 0 to ∞ and φ √0 varies from 0 to 2π leaves and empty
circle of radius 2 2νξ centered at x = y = 0. With a
nonzero g, we can move the term −νξg/2 from the right
to the left-hand side. We then see that this empty circle
is shifted by −νξg/2 along x, and hence its equation is
Eq. (39).
Another approach is to consider an arbitrary ballistic
trajectory as given by Eq.(38). This trajectory could be
represented in a vector form in xyξ space as
T
r = (x, y, ξ) (A1)

IX. CONCLUSIONS with x and y given by Eq.(38). Next we consider a trans-

formation of the xyξ space along ξ axis given by a rota-
tion matrix
We have presented a detailed analysis of the wakefield
in the presence of the transverse plasma gradient. A sim-
 
cos φ0 sin φ0 0
ple ballistic model from Refs.[22, 23] was updated to ac- Rξ =  − sin φ0 cos φ0 0  (A2)
count for the linear transverse inhomogeneity in plasma. 0 0 1
As a result, we provide final analytic expressions for the
pseudopotential and the transverse wakefield. We note and apply to Eq.(A1). With Eq.(38) we have
that as in the flat bubble regime considered previously
in Ref.[24], the bubble shape in the present study shows Rξ r =
similar distortion. Namely, a small perturbation to the  T
plasma density results in the “bending” of the bubble to- 2ξν ξgν ξgν
r0 + − cos φ0 , sin φ0 , ξ . (A3)
ward a lower plasma gradient. However, in contrast to r0 2 2
the flat bubble regime, in the round bubble transverse
wake does not vanish. Next we consider rotation along x axis
We point out that random fluctuation of the plasma  
density, which naturally occurs, may lead to emittance 1 0 0
growth and potentially become a challenge. Conse- Rx =  0 cos θ − sin θ  (A4)
quently, further developments in this direction are in or- 0 sin θ cos θ
der. νg
We note that the numerical examples provided in the such that tan(θ) = 2 sin(φ0 ). Combining Eq.(A4) with
paper are synthetic, and we choose parameters for these Eq.(A3) we get
examples to emphasize corresponding effects. In reality,
Rx Rξ r = (A5)
the parameter g— a transverse plasma gradient should be !T
on the order of 1% or less as well as ζ  1. We emphasize
r i2
2ξν ξgν h gν
that the whole analysis is applicable only when plasma r0 + − cos φ0 , 0, ξ 1+ sin φ0 .
r0 2 2
electrons are nonrelativistic.
 9
√ √
It follows from Eq.(A5) that xyξ space could be rotated If now x ≡ 2 2νξ cos φ0 − ξgν
2 and y ≡ 2 2νξ sin φ0 then
with the help of Rx Rξ transformation such that after the one can arrive at the Eq.(39).
combiner rotation any given trajectory will always lay in The analysis presented above has another important
the Oxξ plane. Combiner rotation Rx Rξ depends only on consequence. As far as each φ0 family of the trajectories
the initial polar angle φ0 of the trajectory starting point forms a separate and independent of others envelope line,
and is independent of r0 . Consequently, we conclude that plasma density that results from electron blowout could
all trajectories with starting points with the same initial be derived exactly the same way for each family in the
angle φ0 are transformed by the rotation Rx Rξ to the transformed plane.
Oxξ plane as well. It is worth reiterating that all trajec- Appendix B: Divergence of the unscreened
tories that start at the same initial angle φ0 will stay in pseudopotential for the large values of κ
the same plane and, with the help of two rotations Rξ
and Rx , could be always translated into Oxξ plane. As We consider an asymptotic at κ  1 of the right hand
far as the transformation Rx Rξ is not degenerate and has side of the Eq.(66)
only one stationary point (0, 0, 0) trajectories that start
from different angles φ0 never cross as different φ0 defines
4κ2 − 3
 
different rotations of the xyξ space. This, in turn, effec- rb3 gκ √ θ(κ − 1) − 1 ≈
tively reduces the initial problem of finding an envelope 4κ κ2 − 1
 
surface for all trajectories to a problem of finding an en- θ(κ − 1)
rb3 gκ θ(κ − 1) − 1 − . (B1)
velope curve for families of the trajectories with the same 4κ2
angle φ0 transformed to the Oxξ plane with the help of
Rx Rξ rotation. With this for κ > 1 Eq.(66) reduces to
If x̃ ≡ r0 + 2ξν ξgν
r0 − 2 cos(φ0 ) then an envelope curve " #
for each family according to Eq.(A5) could be written as 1 ∂ ∂ ψ̃1 ψ̃1 r3 g
κ − 2 ≈− b . (B2)
p ξgν κ ∂κ ∂κ κ 4κ
x̃e = 2 2νξ − cos(φ0 ). (A6)
2
Particular solution has the form
Thus, points on an envelope curve in a transformed plane
have the coordinates rb3 gκ
ψ̃1 ≈ (1 − 2 log κ) (B3)
16
r̃e = (A7)
r i2
!T accounting for the log κ  1 we get
p ξgν h gν
2 2νξ − cos φ0 , 0, ξ 1+ sin φ0 .
2 2 rb3 g
ψ̃1 ≈ − κ log κ. (B4)
8
Inverse transformation Rξ−1 Rx−1
applied to the Eq.(A7)
gives a set of points that resemble envelope curve that
results from the trajectories that have fixed polar angle
ACKNOWLEDGMENTS
φ0 for the starting points.

Rξ−1 Rx−1 r̃e = (A8) The author is grateful to G. Stupakov for fruitful dis-
 T cussions. The work was supported by the ”BASIS” foun-
p ξgν p dation grant #22-1-2-47-17 and ITMO Fellowship and
2 2νξ cos φ0 − , 2 2νξ sin φ0 , ξ .
2 Professorship program.

[1] Chunguang Jing. Dielectric wakefield accelerators. Re-
views of Accelerator Science and Technology, 09:127–149,
2016.
[2] A. Siy, N. Behdad, J. Booske, M. Fedurin, W. Jansma,
K. Kusche, S. Lee, G. Mouravieff, A. Nassiri, S. Oliphant,
S. Sorsher, K. Suthar, E. Trakhtenberg, G. Waldschmidt,
and A. Zholents. Fabrication and testing of corrugated
waveguides for a collinear wakefield accelerator. Phys.
Rev. Accel. Beams, 25:021302, Feb 2022.
[3] A. Zholents, S. Baturin, S. Doran, W. Jansma, M. Kasa,
R. Kustom, A. Nassiri, J. Power, K. Suthar, E. Trakht-
enberg, I. Vasserman, G. Waldschmidt, and J. Xu. A
compact wakefield accelerator for a high repetition rate
multi user x-ray free-electron laser facility. In High-
Brightness Sources and Light-driven Interactions, page
EW3B.1. Optica Publishing Group, 2018.
[4] David H. Whittum, William M. Sharp, Simon S. Yu,
Martin Lampe, and Glenn Joyce. Electron-hose instabil-
ity in the ion-focused regime. Phys. Rev. Lett., 67:991–
994, Aug 1991.
[5] C. Li, W. Gai, C. Jing, J. G. Power, C. X. Tang, and
A. Zholents. High gradient limits due to single bunch
beam breakup in a collinear dielectric wakefield acceler-
ator. Phys. Rev. ST Accel. Beams, 17:091302, Sep 2014.

[6] S. S. Baturin and A. Zholents. Stability condition for the
drive bunch in a collinear wakefield accelerator. Phys.
Rev. Accel. Beams, 21:031301, Mar 2018.
[7] Spencer Gessner, Erik Adli, Weiming An, Sebastien
Corde, Richard D’Arcy, Eric Esaray, Anna Grassellino,
Bernhard Hidding, Mark Hogan, Ahmad Fahim Habib,
Axel Heubl, Chan Joshi, Wim Leemans, R. Lehe,
Carl Lindstrøm, Michael Litos, Wei Lu, Warren Mori,
Sergei Nagaitsev, Brendan O’Shea, Jens Osterhoff, Hasan
Padamesee, Michael Peskin, Sam Posen, John Power,
Tor Raubenheimer, James Rosenzweig, Marc Ross,
Carl Schroeder, Paul Scherkl, Navid Vafaei-Najafabadi,
Jean-Luc Vay, Glen White, and Vitaly Yakimenko.
Path towards a beam-driven plasma linear collider.
SNOWMASS-21, LOI, 2020.
[8] L.K. Len. Report of the doe advanced accelerator con-
cepts research roadmap workshop. DOE, Gaithersburg,
MD, 2016.
[9] ALEGRO collaboration. Towards an advanced linear in-
ternational collider. arXiv, 1901.10370, 2019.
[10] Erik Adli. Plasma wakefield linear colliders – opportu-
nities and challenges. Philosophical Transactions of the
Royal Society A: Mathematical, Physical and Engineer-
ing Sciences, 377(2151):20180419, 2019.
[11] Alex Murokh, Pietro Musumeci, Alexander Zholents,
and Stephen Webb. Towards a compact high efficiency
fel for industrial applications. In OSA High-brightness
Sources and Light-driven Interactions Congress 2020
(EUVXRAY, HILAS, MICS), page EF1A.3. Optica Pub-
lishing Group, 2020.
[12] J B Rosenzweig, N Majernik, R R Robles, G Andonian,
O Camacho, A Fukasawa, A Kogar, G Lawler, Jianwei
Miao, P Musumeci, B Naranjo, Y Sakai, R Candler,
B Pound, C Pellegrini, C Emma, A Halavanau, J Hast-
ings, Z Li, M Nasr, S Tantawi, P. Anisimov, B Carl-
sten, F Krawczyk, E Simakov, L Faillace, M Ferrario,
B Spataro, S Karkare, J Maxson, Y Ma, J Wurtele,
A Murokh, A Zholents, A Cianchi, D Cocco, and S B
van der Geer. An ultra-compact x-ray free-electron laser.
New Journal of Physics, 22(9):093067, sep 2020.
[13] J. B. Rosenzweig. Nonlinear plasma dynamics in the
plasma wake-field accelerator. Phys. Rev. Lett., 58:555–
558, Feb 1987.
[14] J. B. Rosenzweig, B. Breizman, T. Katsouleas, and
J. J. Su. Acceleration and focusing of electrons in two-
dimensional nonlinear plasma wake fields. Phys. Rev. A,
44:R6189–R6192, Nov 1991.
[15] W. P. Leemans, B. Nagler, A. J. Gonsalves, Cs. Tóth,
K. Nakamura, C. G. R. Geddes, E. Esarey, C. B.
Schroeder, and S. M. Hooker. Gev electron beams from a
centimetre-scale accelerator. Nature Physics, 2(10):696–
699, 2006.
[16] Ian Blumenfeld, Christopher E. Clayton, Franz-Josef
Decker, Mark J. Hogan, Chengkun Huang, Rasmus
Ischebeck, Richard Iverson, Chandrashekhar Joshi,
Thomas Katsouleas, Neil Kirby, Wei Lu, Kenneth A.
Marsh, Warren B. Mori, Patric Muggli, Erdem Oz,
Robert H. Siemann, Dieter Walz, and Miaomiao Zhou.
Energy doubling of 42 gev electrons in a metre-scale
plasma wakefield accelerator. Nature, 445(7129):741–744,
2007.
[17] W. K. H. Panofsky and W. A. Wenzel. Some consid-
erations concerning the transverse deflection of charged
particles in radio-frequency fields. Review of Scientific
10
Instruments, 27(11):967–967, 1956.
[18] Alex Chao. Physics of Collective Beam Instabilities in
High Energy Accelerators. Wiley and Sons, New York,
1993.
[19] T. J. Mehrling, C. Benedetti, C. B. Schroeder, E. Esarey,
and W. P. Leemans. Suppression of beam hosing in
plasma accelerators with ion motion. Phys. Rev. Lett.,
121:264802, Dec 2018.
[20] R. Lehe, C. B. Schroeder, J.-L. Vay, E. Esarey, and W. P.
Leemans. Saturation of the hosing instability in quasi-
linear plasma accelerators. Phys. Rev. Lett., 119:244801,
Dec 2017.
[21] T. J. Mehrling, R. A. Fonseca, A. Martinez de la
Ossa, and J. Vieira. Mitigation of the hose instabil-
ity in plasma-wakefield accelerators. Phys. Rev. Lett.,
118:174801, Apr 2017.
[22] G. Stupakov, B. Breizman, V. Khudik, and G. Shvets.
Wake excited in plasma by an ultrarelativistic pointlike
bunch. Phys. Rev. Accel. Beams, 19:101302, Oct 2016.
[23] G. Stupakov. Short-range wakefields generated in the
blowout regime of plasma-wakefield acceleration. Phys.
Rev. Accel. Beams, 21:041301, Apr 2018.
[24] S. S. Baturin. Flat bubble regime and laminar plasma
flow in a plasma wakefield accelerator. Phys. Rev. Accel.
Beams, 25:081301, Aug 2022.
[25] R.L. Gluckstern, F. Neri, and R.K. Cooper. Cumula-
tive beam breakup with randomly fluctuating parame-
ters. Particle Accelerators, 23:37–51, 1988.
[26] J. R. Delayen. Cumulative beam breakup in linear accel-
erators with random displacement of cavities and focus-
ing elements. Phys. Rev. ST Accel. Beams, 7:074402, Jul
2004.
[27] Ji Qiang. Emittance growth due to random force er-
ror. Nuclear Instruments and Methods in Physics Re-
search Section A: Accelerators, Spectrometers, Detectors
and Associated Equipment, 948:162844, 2019.
[28] W. Lu, C. Huang, M. Zhou, W. B. Mori, and T. Kat-
souleas. Nonlinear theory for relativistic plasma wake-
fields in the blowout regime. Phys. Rev. Lett., 96:165002,
Apr 2006.
[29] C. Huang, V.K. Decyk, C. Ren, M. Zhou, W. Lu, W.B.
Mori, J.H. Cooley, T.M. Antonsen, and T. Katsouleas.
Quickpic: A highly efficient particle-in-cell code for mod-
eling wakefield acceleration in plasmas. Journal of Com-
putational Physics, 217(2):658–679, 2006.
[30] J-L Vay, D P Grote, R H Cohen, and A Friedman.
Novel methods in the particle-in-cell accelerator code-
framework warp. Computational Science & Discovery,
5(1):014019, dec 2012.
[31] R. A. Fonseca, L. O. Silva, F. S. Tsung, V. K. Decyk,
W. Lu, C. Ren, W. B. Mori, S. Deng, S. Lee, T. Kat-
souleas, and J. C. Adam. Osiris: A three-dimensional,
fully relativistic particle in cell code for modeling plasma
based accelerators. In Peter M. A. Sloot, Alfons G. Hoek-
stra, C. J. Kenneth Tan, and Jack J. Dongarra, editors,
Computational Science — ICCS 2002, pages 342–351,
Berlin, Heidelberg, 2002. Springer Berlin Heidelberg.
[32] T D Arber, K Bennett, C S Brady, A Lawrence-Douglas,
M G Ramsay, N J Sircombe, P Gillies, R G Evans,
H Schmitz, A R Bell, and C P Ridgers. Contem-
porary particle-in-cell approach to laser-plasma mod-
elling. Plasma Physics and Controlled Fusion, 57(11):1–
26, November 2015.

[33] M. Bussmann, H. Burau, T. E. Cowan, A. Debus,
A. Huebl, G. Juckeland, T. Kluge, W. E. Nagel,
R. Pausch, F. Schmitt, U. Schramm, J. Schuchart, and
R. Widera. Radiative signatures of the relativistic kelvin-
helmholtz instability. In Proceedings of the International
Conference on High Performance Computing, Network-
View publication stats
11
ing, Storage and Analysis, SC ’13, pages 5:1–5:12, New
York, NY, USA, 2013. ACM.
[34] Rémi Lehe, Manuel Kirchen, Igor A. Andriyash, Bren-
dan B. Godfrey, and Jean-Luc Vay. A spectral, quasi-
cylindrical and dispersion-free particle-in-cell algorithm.
Computer Physics Communications, 203:66–82, 2016.