LA-UR-14-23753

High Intensity RF Linear

Accelerators
2-1. Preliminaries of high-intensity
beam dynamics
Yuri  Batygin  
Los  Alamos  Na3onal  Laboratory  

U.S.  Par3cle  Accelerator  School  

Albuquerque,  New  Mexico,  23  –  27  June  2014  
1
Content

2
Maxwell’s Equations

3
 Fields in Focusing Channels

Focusing magnets used in acceleratior facilities:

dipole, quadrupole, sextupole.

4
Magnetostatic and Electrostatic Fields

5
6
7
Normal and Skew Magnets

8
Field Components

9
Expressions for Multipole Fields

10
Quadrupole Pole Shapes and Higher Order Harmonics

11
 Hamiltonian dynamics

Hamiltonian of charged particle with charge q and mass m

H=c m 2 c 2 + (Px - qAx)2 + (Py - qAy)2 + (Pz - qAz )2 + q U

x, y, z position in real space
Px, Py, Pz components of canonical momentum
Ax, Ay, Az components of the vector – potential
U(x,y,z) scalar potential of the electromagnetic field

12
 Liouville’s  Theorem  

Conserva)on  of  phase  space  volume  occupied  by  par)cles  in  Hamiltonian  systems.    

Liouville’s  theorem:  if  the  mo)on  of  a  system  of  mechanical  par)cles  obeys  Hamilton’s
 equa)ons,  then  phase  space  density  remains  constant  along  phase  space  trajectories  and
 phase  space  volume  occupied  by  the  par)cles  is  invariant  (Liouville's  Equa)on):  

df = ∂ f + ∂f dx + ∂f dP = 0
dt ∂t ∂x dt ∂P dt 13
Hamiltonian equations of motion

14
Canonical momentum P = (Px, Py, Pz) is related to mechanical momentum
p = (px, py, pz) via the expression:

p=P-qA (1.30)

Note that the deno minator in Eq.(1.29) is actually mc where the relativistic
factor is:
2
( Px - q A x ) + ( Py - q A y )2 + ( Pz - q A z )2
γ = 1+ . (1.31)
m 2c2

Analogously, the equations for the rates of change of the y- and z - positions
of the particle can be derived. So, the set of equations for the rate of change
of the particle’s position is

dx = (Px - q A x ) dy (Py - q A y ) dz = (Pz - q A z )

=
dt mγ , dt mγ ,
dt mγ . (1.32)

15
 Taking partial derivatives of the Hamiltonian with respect to the particle’s
positions, the equations for t he rate of change of the canonical momentum vector
are:
dPx = q [(P - qA ) ∂A x + (P - qA ) ∂A y + (P - qA ) ∂A z ] - q ∂U
m γ
x x y y z z , (1.33)
dt ∂x ∂x ∂x ∂x

dPy q ∂A ∂A y ∂A ∂U
= [(Px - qAx) x + (Py - qAy) + (Pz - qAz) z ] - q (1.34)
dt mγ ∂y ∂y ∂y ∂y

dPz = q [(P - qA ) ∂A x + (P - qA ) ∂A y + (P - qA ) ∂A z ] - q ∂U
x x y y z z . (1.35)
dt m γ ∂z ∂z ∂z ∂z

16
17
 or
    
dx p dp
= = q{E + [vB]}
dt mγ dt
18
 Canonical Transformations

In Hamiltonian mechanics, a canonical transformation is a change of canonical coordinates (q,p,t) →

(Q,P,t) that preserves the form of Hamilton's equations. Hamiltonian equations of motions are!

dq i ∂H dp i ∂H
= =-
dt ∂p i dt ∂q i

New variables also obey canonical equations of motion

dQi ∂H' dPi ∂H'

= , =- (5.1)
dt ∂Pi dt ∂Qi

where H' is a new Hamiltonian. New v ariables can be considered as f unctions of old
variables and time Qi = Qi (pi,qi,t) , Pi = Pi (pi,qi,t) . Transformations from old variables
to new variables, which keep ca nonical structure of the equation of motion (5.1) are
called canonical transformations.

19
From classical mechanics it follows, that both new and old variables obey principle of
least action :

δ (∑ pi dqi - Hdt ) = 0 (5.2)

δ (∑ Pi dQi - H' dt ) = 0 (5.3)

20
Type 1 generating function

To be a total differential, equation (5.5) has to have the following form:

∂F ∂F ∂F
dF = ∑ dqi + ∑ dQi + dt (5.6)
∂qi ∂Qi ∂t

From comparison of equations (5.5) and (5.6) it is clear, that the variables and the new
Hamiltonian have to obey the following equatons:
∂F ∂F ∂F
pi = , Pi = - (H' - H) dt = dt (5.7)
∂qi ∂Qi ∂t

Therefore new Hamiltonian is connected with the old one via relationship

∂F
H' = H + (5.8)
∂t

Equations (5.7) provide canonical transformation from old variables to new variables, if
generating function depends on old and new coordinates:

∂F1 ∂F1
pi = Pi = - F1 = F1 (q, Q, t) (5.9)
∂qi ∂Qi

21
22
Type 3 generating function

To find third canonical transformation, let us add and subtract ∑ qi dpi from eq. (5.5):

dF = ∑ pi dqi - ∑ Pi dQi + ∑ qi dpi - ∑ qi dpi + (H' - H) dt (5.16)

Introducing generating function of the 3rd type

F 3 = F - ∑ pi qi , dF3 = dF - ∑ pi dqi - ∑ qi dpi (5.17)

the eqution for total differential of the generating function is as follow:

dF3 = - ∑ Pi dQi - ∑ qi dpi + (H' - H) dt (5.18)

Last equation forms the canonical transformation of the 3rd type:

∂F3 ∂F3
Pi = - qi = - F3 = F3 (Q, p, t) (5.19)
∂Qi ∂pi

23
Type 4 generating function

Forth canonical transformation is attained via adding and subtracting of the ∑ Qi dPi
from Eq. (5.5):

dF = ∑ pi dqi - ∑ Pi dQi + ∑ qi dpi - ∑ qi dpi + ∑ Qi dPi - ∑ Qi dPi + (H' - H) dt

Generating function of the 4th ype is defined as follow:

F4 = F - ∑ pi qi + ∑ Pi Qi (5.22)

It results in the eqution for total differential of the generating function:

dF4 = - ∑ qi dpi + ∑ Qi dP i + (H' - H) dt (5.23)

Canonical transformation of the 4th type are descibed by equations:

∂F4 ∂F4
qi = - Qi = F4 = F4 (p, P, t) (5.24)
∂pi ∂Pi

24
Example: Canonical transformation from Cartesian to cylindrical coordinates

Very often, particle dynamics in accelerators is described in a cylindrical

system of coordinates (r, , z), because of axial symmetry inherent to
accelerating structures.

25
26
Inverse transformation of Eqs. (1.49) (1.50), (1.52), (1.53) gives

Px = Pr cosθ - Pθ sinθ , (1.56)

r

Py = Pr sinθ + Pθ cosθ , (1.57)

r

Pz = Pz . (1.51)

A x = Ar cosθ - Aθ sinθ , (1.58)

A y = Ar sinθ + Aθ cosθ . (1.59)

A z = Az (1.54)

27
After a canonical transformation, the new Hamiltonian is expressed in
terms of the old one as

K = H + ∂F3 . (1.55)
∂t

Since the generating function, Eq. (1.45), does not depend on time
explicitly, the new Hamiltonian equals the old one, K = H :

(mc) + (Pθ - qAθ ) + (Pr - qAr) + (Pz - qAz) + qU .

2 2 2 2
H=c (1.60)
r

Hamilton’s equations in cylindrical coordinates read

d r = ∂H dθ = ∂H dz = ∂H
dt ∂Pr , dt ∂Pθ
, dt ∂Pz
, (1.61)

dPr = - ∂H dPθ = - ∂H dPz = - ∂H

dt ∂r , dt ∂θ , dt ∂z
. (1.62)

28
Calculating the partial derivatives, Eqs. (1.61), the equations for
particle position are
d r = Pr - qAr
dt mγ , (1.63)

dθ = 1 (Pθ - qAθ )
, (1.64)
dt m γ r r

dz = Pz - qA z
mγ , (1.65)
dt

29
30
 Averaged  Par3cle  Trajectories  

Field gradient G( ), particle trajectory x( ), and beam envelope Rx( )

as functions of longitudinal coordinate = z/L in an alternating-
gradient focusing structure.

31
Consider one-dimensional particle motion in the combination of constant field
U(x) and fast oscillating field

f (x,t) = f1 (x)cos ω t + f2 (x)sin ω t

Fast oscillations means that frequency ω >> T1 , where T is the time period for
particle motion in the constant field U only. Equation of particle motion:
d2x dU
m 2 =− + f1 cos ω t + f2 sin ω t
dt dx

Let us express expected solution is a combination of slow variable X(t) and fast
oscillation ξ(t) :
x(t) = X(t)+ ξ (t)
where ξ (t ) << X(t )
dU
Fields can be expressed as: U(x) = U(X) + ξ
dX
df
f (x) = f (X) + ξ
dX
32
 Substitution of the expected solution into equation of motion gives:
2
dU d U df
mX +mξ = − − ξ 2 + f (X,t) + ξ
dX dX dX

For fast oscillating term: mξ = f (X,t)

f
After integration: ξ=−
mω 2
2π
Let us average all terms over time, where averaging means mean value over period T = ω
T
1
< g(t) > = ∫ g(t)dt
T 0
2
dU d U df
< mX > + < mξ >= − < > − < ξ 2 > + < f (X,t) > + < ξ >
dX dX dX

Average value of ξ(t) at the period of T = 2π is zero, while function X(t) is changing slowly
ω
during that time. Taking into account that
< X > ≈ X < ξ > = 0
33
 dU df dU 1 df
mX = −  + < ξ > = − − < f >
dX dX dX mω 2
dX
df 1 df 2
Taking into account that
< f >= < >
dX 2 dX

df 2
1 df12 df22
< > = ( + )
dX 2 dX dX

dU eff
equation for slow particle motion is mX = −
dX
1
where effective potential is U eff = U + ( f12 + f22 )
4mω 2

34
 3D  Averaging  Method  

 q   
∂U(r )
Equa)ons  of  mo)on:   r = [ E(r,t)  −    ]

m ∂r

Fast  oscilla)ng  field:  

Par)cle  trajectory  (slow  +  fast

 components):   
 q ∂U eff ( R)
Equa)on  for  slow 
R=− 
m ∂r
 component   2 
  q ∞
Ek ( R)
Effec)ve  poten)al:   U eff ( R) = U( R) +  ∑
4m k=1 ω k2

Fast  component:  
35
(Solid line) typical particle trajectory and (dashed line) the
sine approximation to that trajectory.   36
 4Go πz 1 3π z 1 5π z
G(z) = [sin( ) + sin( ) + sin( ) + ...]
π D 3 D 5 D

FD focusing structure and approximation of field gradient.  

37
 Let us keep only first term:   d2x q 4Go πβ c
m 2 =x sin( t)
dt γ π D

Equation of particle motion in fast oscillating field d2x

m 2 = f1 (x)sin ω t
dt

can be substituted by slow motion in an effective potential  U f12 1 q 4Go D 2 2

= = ( ) X
4mω 2 4m γ π 2 β c
eff

dU eff
Equation for slow particle motion mX = −
dX
d2X 1 q 4Go D 2
= − ( ) X
dt 2 2m 2 γ π 2 β c

d2X
2
+ Ω 2r X = 0
dt

tβc
Let us introduce new variable   τ= where for FD structure     L = 2D
L
38
 Equation of motion in new variables     d2X
+ µo2 X = 0
dτ 2

Frequency of smoothed transverse oscillations q 4 2Go D 2

µo =
in the scale of the period of focusing structure   γ m π 2 (β c)2

Taking into account, that 4 2 1 and Go = β cGmagn, frequency can be written as

≈
π2 3 2
1 qGmagn D
µo =
3 mγβ c
Compare  with  matrix  method  for  FODO  period  with  L  =2D  :  
2 2
L 4 D qGmagn D 1 qGmagn D
µo = 1− =
2D 3 L mγβ c 3 mγβ c

f q 4Go πβ c πβ c
Equa)on  for  fast  component:   ξ = − f =x sin( t) ω=
mω 2 γ π D D
q 4Go D 2 πβ c
Solution for fast component: ξ = −x sin( t)
γ m π (β c)
3 2
D
Amplitude of small fast oscillations in FD structure: ξmax 4 3
= 3 µo
X π
39
 Beam Emittance

The general ellipse equation can be written as

γ x 2 + 2 α x x' + β x'2 = ∍

parameters are called Twiss parameters

40
Realistic beam distribution in phase space. 41
  
g( x,  P, t) = x

42
An expression of the form < x n 1 y n 2 z n 3 Pxn 4 Pyn 5 Pzn 6 > is r eferred to as the nth order moment,
M n 1, n2, n 3, n4, n 5, n6, of the distribution function, where n = n1 + n2+ n3+ n4 + n5 + n6:

∞ ∞ ∞ ∞ ∞ ∞

<x n 1 y n 2 z n 3 Pxn 4 Pyn 5 Pzn 6 > = 1 dx dy dz dPx dPy dPz

N -∞ -∞ -∞ -∞ -∞ -∞

x n 1 y n 2 z n 3 Pxn 4 Pyn 5 Pzn 6 f (x, y, z, Px, Py, Pz , t).

43
The following combination of second moments of distribution function is called
the root-mean-square beam emittance:

∍ rms = <x 2 > <x'2 > - <x x'>2

and the normalized root-mean-square beam emittance is given by

ε rms = 1 <x 2 > <Px 2 > - <xPx >2

mc

By the reasons discussed below, beam emittance is adopted as the value, four times
large than rms emittance

∍=4 <x 2 > <x'2 > - <x x'>2

44
45
Let us calculate rms beam parameters and rms beam emittance for an arbitrary function
ρx (x, x'). We begin by changing variables:

x = r cosϕ
x
σx
{ xσx' - x'σ x = rx sinϕ

Now we rewrite it as

x = rx σx cos ϕ
{ x' = r σ cosϕ - r
'
x x
x sinϕ
σx

The absolute value of the Jacobian of transformation gives us the volume

transformation factor of the phase space element:
∂x ∂x
∂rx ∂ϕ
dx dx' = (abs ) drx dϕ = rx drx dϕ
∂x' ∂x'
∂rx ∂ϕ

46
Let us take into account previously introduced expressions:

σ= β

σ'=- α
β
βγ - α 2 = 1
47
Calculation of integrals over gives:

∞
<x 2 > = π βx rx3 ρx (rx2 ) drx
o

∞
<x'2 > = πγ x rx3 ρx (rx2 ) drx
o

∞
<x x'> = - π α x rx3 ρx (rx2 ) drx
o

Therefore, beam emittance is given by

∞
∍ x = 4π rx3 ρx (rx2 ) drx
o

48
 2 2
Rms beam ellipse ( 4 <x' > ) x 2 - 2 (4 <xx'> ) x x' + (4 <x > ) x' 2 = ∍ x
∍x ∍x ∍x

Beam distribution and rms ellipse.

49
 Example: Uniformly populated ellipse
Consider an example, where the beam ellipse has an area o f Ax, and is uniformly populated
by particles. Particle density is constant inside the ellipse rx2 = Ax :
ρx (rx2 ) = 1
πA x
Calculation of the rms value, <x 2 >, gives:

Ax
<x 2 > = π βx rx3 ρx (rx2 ) d rx = A x βx
o 4

50
Uniformly populated ellipse at phase plane (x, x’).
The beam boundary is given by

Rx = A x βx

Radius of the beam represented as a uniformly populated ellipse is equal to twice the
rms beam size:

R = 2 <x 2 >

Rms beam emittance:

Ax
∍x = 4 rx3 drx = Ax
Ax o

Therefore, the area of an ellipse, uniformly populated by particles, coincides with the 4 x
rms beam emittance. This explains the choice of the coefficient 4 in the definition of
rms beam emittance.

51
 Different Particle Distributions in Phase Space

Consider quadratic from of 4-dimensional phase space variables:

x 2 y
I = (σ x x ' − σ x' x)2 + ( ) + (σ y y' − σ y' y)2 + ( )2
σx σy

Consider different distributions f = f(I) in phase space which depend on

quadratic form:
2
Water Bag: ,   I ≤ Fo
π Fo
2 2

f ={
0,    I > Fo

6 I
f= (1− )
Parabolic: π 2 Fo2 Fo

1 I
Gaussian: f= exp(− )
π 2 Fo2 Fo

∞ ∞ ∞ ∞

Normalization: ∫ ∫ ∫ ∫
−∞ −∞ −∞ −∞
f dx dx 'dy dy' = 1
52
Characteristics of Beam Distributions

53
 Projection of distributions on phase plane
∞ ∞
ρx (x,x') = f (x, x', y, y') dy dy'
-∞ -∞
σyy'- σy'y = T cos ψ
T ,ψ
'
Let us change the variables (y,  y ) for new variables y
= T sin ψ
σy

54
 2
,   I = rx2 + T 2  ≤ Fo
Water Bag distribution π Fo
2 2

f ={
0,    I > Fo

is restricted by surface rx2 + T 12 = Fo , T 12 = Fo - rx2

T12
2
' ρx (x, x') = 2 dT 2 = 2 (1 - rx )
Projection of Water Bag distribution on (x, x ) π Fo Fo
π Fo2 o

For Parabolic distribution, projection on x, x' plane is

T12 2
2 2 2
ρx (x, x') = 6 (1 - rx + T ) dT 2 = 3 (1 - rx )
π Fo2 0 Fo π Fo Fo

55
 KVKV   ε max = 4ε rms

Water Bag
Water  Bag   ε max = 6ε rms

Parabolic
Parabolic   ε max = 8ε rms

Gaussian ε max = ∞

Particle
Particle distributions
distributions with equalwith of ε rms values
valuesequal . 56
of .
 Rms emittance of distributions with elliptical symmetry

∞
Four rms beam emittance ∍x = 4π rx3 ρx (rx2) dr x
o

57
58
 Fraction of particles residing within a specific emittance

2π ∍ ∍
η = N (∍) = ρx (rx2 ) dx dx ' = ρx (rx2 ) rx drx dϕ = π ρx (rx2 ) drx2
No o o o

59
Distributions on phase plane are:

r2
Water bag ρx(rx2) = 4 (1 - 2 x )
3π ∍x 3 ∍x
2
r x2
Parabolic ρx(rx2) = 3 (1 - )
2π ∍x 2 ∍x

r 2
Gaussian ρx(rx ) =
2 2 x
exp( - 2 )
π ∍x ∍x

60
Fraction of particles versus phase space area for different
particle distributions. 61
 Self-Consistent Particle Dynamics

Example of self-consistent dynamics: two - body problem

In classical mechanics, the two-body problem is to determine the motion of two point particles that interact only !
with each other. !

62
Self-Consistent Approach to N – Particle Dynamics

63
2. 63
Field created by the beam is described by Maxwell's equations:

∞ ∞ ∞

space charge density ρ=q f dPx dPy dPz

-∞ -∞ -∞

∞ ∞ ∞

beam current density j=q v f dPx dPy dPz

-∞ -∞ -∞

o = 8.85 x 10-12 F/m is the electric permittivity

-7
o = 4 10 H/m is the magnetic permeability of free space 64
2. 64
 Laboratory and moving systems of coordinates

Consider system of coordinates, which moves with the average beam velocity . We will
'
denote all values in this frame by prime symbol. Potentials U ' , A are connected with that in
laboratory system, U, A , by Lorentz transformation
β
Az = γ (Az' + U ')
c
U = γ (U ' + β cA z' )

A x = A x' , A y = A y'

65
In the moving system of coordinates, particles are static, therefore, vector potential of the
'
beam equals to zero, A b = 0. According to Lorentz transformations, components of vector
potential of the beam are converted into laboratory system of coordinates as follow

A xb = 0 , A yb = 0 , A z b = β Ub
c

Equation for unknown potential of the beam together with Vlasov’s equation
for beam distribution function constitute self-consistent system of equations
describing beam evolution in the field created by the beam itself.

66
 Applicability of Vlasov's equation to particle dynamics

Vlasov's equation describes behavior of non-interactive particles in given external field.

Charged particles within the beam interact between themselves:
(i) interaction of large number of particles resulted in smoothed collective charge
density and current density distribution
(ii) individual particle - particle collisions, when particles approach to each other
at the distance, much smaller than the average distance between particles.

First type of interaction results in generation of smoothed electromagnetic field,

which, being added to the field of external sources, act at the beam as an external field.
The second type of interaction has a meaning of particle collisions resulting in
appearance of additional fluctuating electromagnetic fields.

Using Vlasov's eqauiton, we formally expand it to dynamics of interacting charged

particles, assuming that the total electromagnetic filed of the structure (U, A )
U = Uext + Ub
A = A ext + A b
Uext, A ext, external field
Ub, A b field created by the beam

and neglecting individual particle-particle interactions.

67
68
 λD = ε o kT
Radius of Debye shielding in plasma :
q 2n

Combining all equation one gets:

r << 2π λ D or N D >> 1, or N D = (2π )3/2 nλ D3

where ND is the number of particles within Debye sphere.

Individual particle-particle collisions can be neglected if number of particles w ithin

Debye sphere is much larger than unity (or average distance between particles is much
smaller than D).

Particle density within uniformly charged cylindrical beam of radius R, with current I,
propagating with longitudinal velocity c, is

n= I
π q βc R2

69