Transverse Beam Dynamics

— Betatron Motion

國家同步輻射研究中心
周炳榮 Ping J. Chou
[email protected]

PHYS553000— 02b Last update: 2019-10-7

1
 十年樹木，百年樹人

Important Notes to Students: The sole purpose of this lecture notes is

meant for educational use only. Some photographs and graphic
illustrations are adapted from various reference literatures, which are NOT
to be distributed beyond the classroom use.

[References]
1. S.Y. Lee, Accelerator Physics, 3rd ed. (World Scientific, 2012),
Chap. 1 and 2.
2. P. Schmuser, “basic course on accelerator optics” in CERN 87-10, CAS proceedings
(CERN Yellow Report).
3. I. Percival and D. Richards, Introduction to Dynamics, (Cambridge Univ Press, 1982).
4. H. Wiedemann, Particle Accelerator Physics, 4th ed. (Springer, 2015)
5. CERN/MPS-SI/Int. DL/70/4 (1970), C. Bovet, R. Gouiran et al., “A selection of formulae
and data useful for the design of A.G. synchrotrons”
6. A. Wolski, Beam Dynamics in High Energy Particle Accelerators, (ICP, 2014)
2
Outline
Basic properties of Hill’s eq., transfer matrix
Alternating gradient focusing
General solutions of Hill’s eq., stability condition of betatron motion
Concept of amplitude function (betatron function)
Floquet transformation
Courant-Synder invariant
Concept of phase ellipse, emittance
2nd order O.D.E. with error terms  Inhomogeneous Hill’s eq.
How to introduce field errors to eq. of betatron motion
Off-momentum: dispersion function
Momentum compaction factor
Dipole field error  avoid integer tune
Gradient error due to dp/p  avoid half-integer tune
Beta beating
Natural chromaticity
Sextupole magnet+ dispersion function  chromaticity correction
Rms beam size

3
 
 Homework 7) Show that  p  p  0 as given by Eq.(87)
 Homework 8) Using the transfer matrix M(ss+L) for one period as an explicit
example, show that the transfer matrix M satisfies the symplectic condition as given
by Eq.(94).

 Homework 9) Please show the result as given by Eqs.(56) and (57).

 Homework 10) Please prove the Courant-Synder invariant as shown in Eq.(60).

 Homework 11) Please prove the transformation matrix for the Courant-Synder
parameters as shown in Eq.(115).
 Homework 12) Using the general properties as given in Eqs.(39 & 40), please
finish the detailed derivation of Eq.(124).

 Homework 13) Using the transfer matrix given in Eqs.(30), please show that the
particular solutions for dispersion function and its derivative in a sector dipole magnet
are indeed given by Eq.(131)

 Homework 14) Prove the result as given by Eq.(153).

 Homework 15) Please derive the closed orbit distortion for a misaligned focusing
quadrupole as given by Eq.(162).
4
 Homework 16) Show that the betatron tune shift due to a localized gradient error
is given by Eq.(166).

 Homework 17) If the order of matrix product in Eq.(165) is reversed, i.e.

M(s1+L | s1) = m(s1) M0(s1+L | s1) , will we get a different result for the betatron
tune shift as shown in Eq.(166)? Why?
 Homework 18) Show the relation as given by Eq.(168) for the one-turn transfer
matrix with a gradient error in focusing quadrupole.
 Homework 19) Show the result of beta beating as given by Eq.(169) for a
gradient error in a circular accelerator.
 Homework 20) Show that the phase advance from point 1 to point 2 can be
described by elements of transfer matrix M(21) and amplitude functions at point 1
𝑀12
by ∆y =𝑡𝑎𝑛−1 ( ). [D.Edwards & M.Syphers, Problem 2.15]
𝛽1 𝑀11 −𝛼1 𝑀12

5
 Alternating Gradient Focusing

We can use a magnet sequence which is comprised by quadrupole magnets of

strong alternating gradients |K|>1, and still result in net focusing in both transverse
planes. The advantage of such scheme is a significant reduction in the size of
accelerators. This kind of design is called the strong focusing principle.

In an accelerator of strong focusing design, the general expression of equation of

betatron motion is:
x( s)  K ( s) x( s)  0 (37)
K ( s  L)  K ( s )

where K(s) is a periodic function, is known as a Hill’s equation, because G.W.

Hill first investigated such an equation related to the theory of moon’s motion.
According to the Floquet’s theorem, the general solution for Eq.(37) is:

x(s)  w(s)eiy ( s ) , where w(s)  w(s  L) (38)

[Ref.] J. Mathews & R.L. Walker, Mathematical Methods of Physics, 2nd ed. (1970).
Google Books, search “Hill’s Equation”, W. Magnus & S. Winkler (Dover phoenix edition) 6
The general solution of a second order linear differential equation is

 x( s)  C ( s, s0 ) x0  S ( s, s0 ) x0 C and S: osiculating

 x( s)  C ( s, s ) x  S ( s, s ) x parameters
 0 0 0 0

The solution of transverse motion can be described in matrix form,

~  x( s)   C ( s, s0 ) S ( s, s0 )  ~
x ( s)      x ( s0 )  M ( s, s0 ) ~
x ( s0 )
  
 x ( s) C ( s, s0 ) S ( s, s0 ) (39)

where M(s,s0 ) is called the transfer matrix, and it has the following property:

M (s, s0 )  M (s, s1 )M (s1 , s0 ) , det(M )  1 (40)

Liouville’s theorem

7
 Formal Treatment of Betatron Motion

Recall the general expression of equation of betatron motion in Eq.(37), from the
Floquet’s theorem, the general form of solution is given by Eq.(38) as:

x(s)  w(s)eiy ( s ) , where w(s)  w(s  L)

Expressed in real form,

 x( s)  x0C ( s)  x0 S ( s)
 (41)
 x( s)  x0C ( s)  x0 S ( s)
where C(s) is a cosine-like function,
and S(s) is a sine-like function

In general, the solution of Eq.(37) can be written as:

x(s)  x0 w(s) cosy (s)  x0 w(s) siny (s) (41)

x(s)  A  w(s) cos(y (s)   ) (42)

Let’s differentiate the preceding expression to get explicit forms of x’ and x’’, then
we substitute them back into Eq.(37). 8
The explicit expression of equation of betatron motion is
x  K ( s) x
 A[ w( s) cos(y ( s)   )  w( s) sin(y ( s)   ) y ( s)  w( s) sin(y ( s)   ) y ( s)
 w( s) cos(y ( s)   )  (y ( s)) 2  K ( s) w( s) cos(y ( s)   )
(43)
 w( s) sin(y ( s)   )y ( s)]  0
Further simplification gives us the following relation,
[ w( s)  w( s)(y ( s)) 2  K ( s) w( s)] cos(y ( s)   )
 [2w( s)y ( s)  w( s)y ( s)] sin(y ( s)   )  0 (44)

Since the cosine function and sine function are orthogonal to each other, we obtain
the following relation:

 2w( s)y ( s)  w( s)y ( s)  0 (45)


w( s)  w( s)(y ( s))  K ( s) w( s)  0
2
(46)

From Eq.(45) we have:

wy   2wy 
9
 dy  1 dw 2 1
 y ( s)  (47)
ds y  ds w w2 ( s )
Betatron phase equation

Substituting Eq.(47) to Eq.(46) we get the following result,

1
w( s)  3
 K ( s) w( s)  0 (48)
w ( s)
Betatron envelope equation
maximum b(s) [K(s)] -1/2

The definition of betatron function b(s) is introduced as the following: w2 ( s)  b ( s)

Eqs.(47) and (48) are rewritten as:

ds
s
1

y ( s)  ,y ( s)   (49)
b ( s) s0
b ( s)
b ( s) 1 b ( s) 2
 K ( s)  2 [1  ( ) ]0 (50)
2b ( s) b ( s) 2 10
Therefore, the general solution of Hill’s equation in terms of the betatron function is
given by:
x(s)  A b (s) cos(y (s)   ) (51)

The phase advance of betatron oscillation from point 1 to point 2 is:

s2
ds
Proportional to K1/2 y 12 
s1
b (s)
, then the phase advance of betatron motion in one revolution period is mx.
ds
mx   (52)
b (s)
Qx= mx /2p is the betatron tune. The betatron tune means the number of betatron
oscillation per revolution period. In other words, mx /Qx= 2p, where the periodic
condition has been implicitly embedded in this statement.
x( s )  K ( s ) x( s )  0
Is it possible to convert the Hill’s
equation into the same type of 
equation for a simple harmonic d 2u ( )
oscillator?  const.  u ( )  0
d 2 11
Horizontal betatron phase vs. longitudinal position (TPS)

Qx= 26.16
12
Vertical betatron phase vs. longitudinal position (TPS)

Qy= 14.24
13
 Betatron Motion in Floquet Transformed Coordinates

We will perform a Floquet transformation on the Hill’s equation, i.e. Eq.(37).

We introduce the Floquet transformed variables as defined below:

x( s )
u (s)  (53)
b 1/2 ( s )
y (s)
 (s)  (54)
Q

We can readily find the following results,

d 1 dy 1
 
ds Q ds Q b ( s )

du du d du
   Q b ( s )u ( s )
ds d ds d

1 d 2u
 b u   b u  (55)
Q b d
2 2

14
Using the betatron envelope equation as given by Eq.(50), we obtain the following
result,
x( s )
b (s)u( s)  b (s)u( s)  b 1/ 2 ( s)[ x(s)  K (s) x(s)  ] (56)
b ( s)
2

Substituting Eqs.(37) and (55) to Eq.(56), we get the following results,

d 2u ( )
𝑥 𝑡 + 𝜔2 𝑥 𝑡 = 0  Q 2u ( )  Q 2 b 3 2 ( x  Kx)
d 2
0
(57)

𝜔→𝑄
u ( )  A cos(Q   )  b 1/2 x( s ) (58)
𝑡 →
, where A is a constant which depends on the initial conditions.

Furthermore, we find the following property:

1 du ( ) 2 x 2 (s) 1
u ( )  [
2
] A 
2
 [ ( s ) x( s )  b ( s ) x( s )]2 (59)
Q d b (s) b (s)
b ( s)
, where  ( s)  
2
1 du ( ) 1
  x  b x  Pu ( )   A sin(Q   )
Q d b
15
 A physical quantity is introduced, A2  

Emittance (an invariant of betatron motion)

Eq.(59) is expressed as
x x  bx 2
( )2  ( )  (60)
b b
This relation is called the Courant-Synder invariant. It is a property of betatron
motion in a particle accelerator.

Courant-Synder parameters are defined as,

b ( s) Envelope/amplitude function
b ( s )
 (s)  
2
(61)
1   2 ( s)
 ( s) 
b (s)

[Ref.] C. Bovet et al. “A selection of formulae and data useful for the design of A.G. synchrotrons”, 16
CERN-MPS-SI-INT-DL-70-4
 Ernest Courant
"If you can't do two things together, you just do
one after the other - that's all there is to it!" Ernest
Courant, Brookhaven distinguished scientist
emeritus, was describing how he came to think of
the strong-focusing principle that he, together with
M Stanley Livingston and Hartland Snyder, co-
discovered in 1952.
This article originally appeared in the 4 August
2001 issue of the Brookhaven Bulletin.
[Ref.] http://felix.physics.sunysb.edu/PAhist/ecourant.html

Fig. 30
[Ref.] http://www.pas.rochester.edu/~bodek/dept/courant.gif
[Ref.] http://encyclopedia.vbxml.net/Ernest_Courant

17
 Normalized phase space (a circle)

1 du 1
Eq.(59)  Pu ( ) (y  by)
Q d b

𝜔→𝑄
A 𝑡 → A
Qy Qy
u()
y
b

 :[0, 2p]
Fig. 31 Fig. 32

Q = y (the betatron phase advance from point 1 to point 2 in the longitudinal

coordinate)

The contour of betatron motion in the normalized phase space (after Floquet
transformation) is a circle with a radius A= 1/2. The area of this circle is p. The
phase contour is the same as a simple harmonic oscillator.

18
 Beam Ellipse and Courant-Snyder Invariant

If we follow a particle in the phase space  2

(y, y’), period after period, this particle y’ tan 2q 
(<0) b 
will trace out a closed curve--- an ellipse (/)1/2
for a linear accelerator lattice (Courant- ()1/2
(/b)1/2
Snyder invariant): (/b)1/2
y 2  (y  by) 2  b
A= p q
The area enclosed by this y
invariant contour is p, which (b)1/2
does not change at different
longitudinal location s. But the (/)1/2
shape factors s and bs do
change at different locations. Fig. 33

For a given beam emittance , the size ymax is (b)1/2.

The magnitude and variation of b within the period describes the focusing.
> 0 means that positive y has an average negative slope a converging beam.
< 0 means a diverging beam.
= 0 means that the size, and b, is passing through a maximum or minimum.
 is proportional to the phase space area occupied by the beam ellipse.
The ratio of major axis to minor axis for the beam ellipse is bs. 19
The equation of phase ellipse in (y, y’) can be easily 1
( y  b y )  Pu
changed to a circle by a coordinate transformation: b
y y
( )2  (  b y) 2   2
b b
𝑄𝜑 =y
, where y is the betatron phase advance from 1 y
point 1 to point 2. This circle will have a constant u
b
size, its area p. This is the area occupied by
particles in the phase space.
y Fig. 34
s
1 2
 x2  x 2  [ xb ( s)]2
Then, the propagation of particle from point 1 to point 2
can be described as a rotation of angle  along the  xbx
circular contour in the normalized phase space:
mean-square beam size,
particles lie within one
u2   cosy siny  u1  standard deviation from
 v    siny cosy   v1 
(62) the beam axis
 2 

A circular rotation in the Floquet transformed 20

coordinate system (the normalized phase space)
 We have found the general expression for solutions of Hill’s equation in
terms of the betatron function. What is the general expression of transfer
matrix in terms of the betatron function?

Using the Floquet transformation, we can convert the transfer

matrix expressed in the normalized (u,v) coordinates back to the
regular (y, y’ ) coordinates! Hence, the propagation of particle from
point 1 to point 2 can be described as:

 1 b2 0   y2   cosy siny   1 b1 0   y1 
      
 2 b 2 b 2   y2   siny cosy  1 b1 b1   y1  (63)

A rotation in the new coordinate transformation

coord. System of y, y’
(Floquet transformed) Floquet’s transformation

21
  y2   b2 0   cosy siny   1 b1 0   y1 
 y       
 2    2 b 2 1 b 2   siny cosy  1 b1 b1   y1  (64)

 b2 
 (cosy  1 siny ) b1b 2 siny 
b
M 21   1 
  (1  1 2 ) (   2 ) b1  (65)
 siny  1 cosy (cosy   2 siny )
 b1b 2 b1b 2 b2 
a description of beam
 M ( s2 | s1 ) focusing system
, where y is the betatron phase advance from point 1 to point 2.
Eq.(65) is the general expression of transfer matrix M for the particle trajectory,
i.e. the transformation of (y, y’) from s1 to s2. We can track the particle trajectory
with the help of transfer matrix as shown in Eq.(65) ray tracing
s2
ds
The betatron phase advance from s1 to s2 is given by: y  
s1
b (s) 22
We can also express the betatron phase advance from s1 to s2 in terms of the
elements of transfer matrix and Courant-Snyder parameters,

 M 12 
y  tan 1  
 b1M 11  1M 12 
, where Mij is the element of transfer matrix, b1 and 1 the Courant-Synder
parameters at the starting point s1.

When particles travel one period, the beam shape will repeat. We must have
2= 1, b2= b1 after one period (closure condition).

cos  p   sin  p b sin  p 

M ( s  L | s)  
   sin  p cos  p   sin  p  (66)

, where p is the betatron phase advance per period. If there are N periods in one
revolution period, then the phase advance for one turn is yx= Np. Note that the
betatron tune is Qx= yx /2p.

23
From Eq. (52), the betatron tune is related to the mean radius of accelerator R:
1
Qx  R   (67)
b ( s)
Note that b(s) depends on the total configuration of magnets in a circular accelerator.
The betatron phase advance along a beamline is not a linear function of longitudinal
coordinate.

The particle oscillates inside a

beam envelope [b(s)]1/2. The
envelope changes with the
lattice structure.

Fig. 35 [Ref.] adapted from M. Sands, SLAC-121 (Nov. 1970)

24
The general solution of betatron motion in Eq.(51) is explicitly described as:

y ( s )  b ( s ) cos(y ( s )  y 0 ) (68)

, where 0 is the starting phase for a given (y0, y’0).

The equation of phase ellipse can be expressed by Courant-Snyder parameters as:

 (s) y 2  2 (s) yy  b (s) y2   (69)

(s), b(s), and (s) are also mistakenly called the Twiss parameters. Sometimes
accelerator physicists call them beam optical functions.

F D F
Fig. 36

b(s)

s
y’ y’ y’ y’
Phase space:
y y y y 25
The emittance is invariant everywhere along the accelerator when there is no
acceleration. In a circular accelerator, only the phase ellipse that repeats after one
lattice period survives.

Example: tracking a single particle

y= 2 mm, y’= 0
in the phase space turn by turn

diffusion
y= 1.3 mm, y’= 0

Example: generating a beam

distribution of many particles in
the phase space

26
 Example: beam optics of double mini-by lattice in TPS storage ring
b-functions
30

20

bx [m]
10

0
0 20 40 60 80 100 120 140 160

30

20
by [m]

10

0
0 20 40 60 80 100 120 140 160
s - position [m]
Horizontal Acceptance Ellipse Vertical Acceptance Ellipse
2 2
horizontal vertical
1.5 1.5

1 1

0.5 0.5
y'(mrad)
x'(mrad)

0 0

-0.5 x0= 19 mm, x’0= 0 -0.5

-1 -1

-1.5 -1.5
y0= 3.5 mm, y’0= 0
27
-2 -2
-25 -20 -15 -10 -5 0 5 10 15 20 25 -10 -8 -6 -4 -2 0 2 4 6 8 10
x(mm) y(mm)
Example: to generate a random Gaussian beam in the transverse phase space
J
H
b
x  2 J b cosy
2J
x   (siny   cosy )
b

J   ln R1 , where R1 and R2 are uniform random

y  2p R2 numbers, 0<R1, R2 ≦1

[Ref.] R. Siemann, ARDB Note-284

28
If x1, x2 are random deviates with a joint probability distribution p(x1, x2)dx1dx2, and if
y1, y2 are each functions of all the x’s, then the transformation between two uniform
deviates on (0,1), x1,x2 and two quantities y1, y2 ,
y1   2 ln x1 cos 2px2
y2   2 ln x1 sin 2px2
[ref.] Numerical Recipes, Chap. Random Numbers

Normal Gaussian distribution:

( xm )2
1 
f ( x)  e 2 2
 2p FWHM
= 68.3 %
2= 95.4 %
3= 99.7 %

[ref.] http://en.wikipedia.org/wiki/Gaussian_distribution

FWHM (full width at half maximum)= 2 2 𝑙𝑛2 = 2.35 𝜎

29
 -3 2 2
x 10( 1/2.055) x +2 (-0.0) x xp +2.055 xp -(62.6e-9) = 0 Histogram of xp
2 500

450
1.5
400

1 350

300
0.5
250

200
xp

0
150 
-0.5 100

50
-1
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
-3
-1.5 x 10

Histogram of x
500
-2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 450
x -3
x 10 400

350

rms transverse beam size (no dispersion): 300

250

 x   xbx 200

150

 y   yby 100

50

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
-3
x 10

30
Example: Injected beam: Dxjitter= -3mm, other errors=0
bx= 2.503 m, x= 0, x= 41.6 nm-rad, by= 6.156 m, y= -0.73, y= 53 nm-rad,
Random error activated for injection kickers.

Injected beam: <xinj>= -27.96 mm, <x’inj>= 0 mrad, <yinj>= 0.01 mm, <y’inj>= 0
mrad, 2000 particles.

Configuration space Phase space (horizontal)

31
 Stability of Betatron Motion in Alternate Gradient Focusing

The transfer matrix M has two eigenvectors. Any initial condition (x, x’) can be
expressed as a linear superposition of these two eigenvectors,
x
 x  AV1  BV2
  in
x
Particles been transferred after n repetition periods: M    A1nV1  Bn2V2
n

 x in
, where V1 and V2 are the eigenvectors of M, 1 and 2 the eigenvalues.

The requirement of beam stability is equivalent to the requirement that 1n and
2n do not grow as n.

For an oscillation to be stable after an arbitrary period N, the quantity

x
N
M  
 x 
, where M is the transfer matrix for one period, must remain finite.
32
Considering the case where A= (x, x’) is the eigenvector of the matrix M, then the
characteristic equation of transfer matrix is given by

MA  A  det(M  I )  0
2    Tr ( M )  det(M )  0 (70)

Tr( M )   i , det(M )   i (71)

i i

Since M has a determinant equal to one,

12  1
|  | 1
In general we can write these two eigenvalues as

1  eim , 2  eim (72)

From Eq.(71), the trace of M equals to

Tr(M )  1  2  2 cos m

33
Therefore, we obtain a condition for stability of betatron motion:
1
 1  Tr ( M )  1 (73)
2
The parameter m is the betatron phase advance per repetition period.

Recall the explicit form of transfer matrix M for one period given by Eq.(66),
2   (2 cos  p )  1  0

  cos  p  i sin  p  exp( i p )

1
cos  p  cos(2pQ / N )  Tr ( M ) (74)
2
, where N is the number of period.

For stability p must be real and this implies:

tr (M )  2 and   1 (75)

34
 Liouville’s Theorem and the Invariant of Beam Motion

Liouville’s theorem: under the influence of conservative forces, the density of

particles in some given point in phase space remains constant in time.
p
S
R
q(t2), p(t2) fi: a map which describes the
trajectory of particle motion in the
q2= f1(q1, p1) phase space, for i=1, 2
q(t1), p(t1)
p2= f2(q1, p1) Fig. 29
q
If r is the particle density in phase space, the number of particles within an
infinitesimal volume in 6-D space is r ( x, y, z, px , p y , pz )dxdydzdp x dp y dpz

The phase space current created by the motion of particles is

   
J  r where   dQ dt (76)

Q is a 6 - D position vector in 6 - D phase space

Q  ( x, y , z , p x , p y , p z ) (77)

J  ( rx, ry , rz, rp x , rp y , rp z ) (78)
35
From the continuity equation of phase space current, i.e. the number of particles is
conserved,  r
 J   0 , where the del  is a vector derivative in 6-D space.
t
If we assume that the particle position does not depend on its momentum and vice
versa,
r 
   J
t
 
  r ( rr )   p ( rp )
   
 r   r r  r r  r  p   p r  r   p  p (79)

, where r  ( x, y, z ,0,0,0)

p  (0,0,0, p x , p y , p z )
 r  ( x,  y,  z , 0,0,0)
 p  (0,0,0,  p x ,  p y ,  p z ) (80)

   
 p r p p
Knowing that r   2   (81)
m0 c m0c 2 m02c 4  p 2c 2
36
 not depend on the position
 
 2
c p
r  r  r  0 (82)
m02c 4  p 2c 2

  Because the electrical field is independent of the

p  p  p F particle momentum
  
  p  q( E  r  B) (83)

  
 p  p  q p  (r  B)
   
 qB  ( p  r )  r  ( p  B)
 
 qB  ( p  r )
 2
c p
 qB  ( p  ) (84)
m0 c  p c
2 4 2 2

     
Note that   (a  b )  b  (  a )  a  (  b ) (85)
p 2  p x2  p y2  p z2 (86) 37
Carry out the explicit derivatives as shown above, we obtain the following result

p  p  0 (87)

r  
Therefore, we get the relation   r  r r  p   p r
t
 
r dr dp
0  ( r r )   ( p r ) 
t dt dt
 
dr (r , p, t )
 (88)
dt
We have proved the invariance of phase space density of a particle beam under
the influence of electromagnetic fields! The motion of particle beam is described by
a Hamiltonian system.

Without damping nor excitation of particles, the Liouville’s theorem implies that the
area in phase space that is occupied by the particle beam stays constant. The area
of the beam distribution in phase space is preserved as the particle beam evolves
in time (traverses in an accelerator). The phase space area occupied by the
particle beam is an invariant of the beam motion in an accelerator. We readily
get the following relation,

S

dq2 dp2  J f ( x) dq1dp1
R
(89) 38
 , where Jf(x) is the Jacobian matrix of canonical transformation

J f ( x)  (f i x j )
 f1 f1 
  
 x1 xn 
(90)
  
 f m f m 
  
 x1 xn 

Since the area occupied by the particle beam in phase space is an invariant, the
determinant of the Jacobian matrix must be unity.
Using the transformation depicted in Fig.29 as an example,
(q1, p1 )  (q2, p2 ) If (q, p)= (x, x’)  x2 x2 
 
J f ( x)  (f i x j ) x x1 
J f ( x)   1
 x2 x2 
 q2 q2   x
   1 x1 
q p1 
 1  M 11 M 12 
 p2 p2 
   (92)
(91)
 q p1 
 1  M 21 M 22  39
 The transfer matrix M is the Jacobian matrix,
 x2   M 11 M 12  x1 
      (93)
 x2   M 21 M 22  x1 
The determinant of the transfer matrix (map) M has a value of unity!

From Hamiltonian dynamics, the Jacobian matrix of a canonical transformation is

symplectic [Ref.].
The transfer matrix M is symplectic and satisfies the symplectic condition,

~  0 1 ~
MSM  S , where S    , S S  I , S 2   I (94)
 1 0 
The consequence of symplectic condition is det(M)= 1. (95)
For the 3-D particle motion in an accelerator, the volume of the beam distribution in
the 6-D phase space is an invariant of the beam motion in the accelerator.
g g
*Another notation for Jacobian: J  g , h   u 
 
 u,  h h (96)
[Ref.] H. Goldstein, Classical Mechanics, 2nd ed. (1980) u  40
http://en.wikipedia.org/wiki/Symplectic_integrator
 Adiabatic Damping of Betatron Motion in a Linac

When a charged particle energy is slowly/adiabatically increased in an accelerator,

how will the betatron motion be affected by the acceleration process?
Argument 1:
The slope of transverse particle trajectory is:
dy
y 
ds
dy

vs dt where py is the transverse momentum
vy determined by the particle source
 (electron gun), and p is the longitudinal
vs momentum.
mvy py
 
mvs p (97)
Since the acceleration of particle takes place in the longitudinal direction, p is a
function of time, or longitudinal position in the case of a linac. The transverse
momentum py is only determined by the initial conditions when particles are created.
At any particle energy, the longitudinal momentum is expressed as p= m0cb. 41
 py 1  py 
y     (98)
m0cb b  m0c 

The item in the bracket is invariant during the acceleration process.

The emittance equals to the area enclosed by the y-y’ contour in the phase space
divided by p. Therefore, the emittance during the acceleration process is inversely
proportional to the product b,
0
 (99)
b
where 0 is the emittance at the beginning when the particle beam is created by
the source. The natural (geometric) emittance decreases as the particle energy is
increased. We define a normalized emittance N which is invariant during the
particle acceleration:
 N  b (100)

where b  vs / c, and   1 1 b 2 Something is wrong here!

Which part is incorrect?

Only the normalized emittance is invariant during the particle acceleration. 42

The geometric emittance is inversely proportional to the particle energy.
Argument 2:

The equation of betatron motion during the particle acceleration should be modified
as following:
d ds d d
   vs
dt dt ds ds
dp y d
 ( py)
dt dt
d
 vs ( py)
ds
 vs ( py  py)
  (101)
 qv  B

For a focusing quadrupole magnet, B= -Gy, then we have the following:

 
vs ( py  py)  qv  B  qvsGy
p qG p
y  ( ) y  y0 y  ( ) y  K ( s ) y  0 (102)
p p p
qG ( s)
K (s)  43
damping term p( s)
For simplicity, let’s use the piecewise constant magnetic field to look for solutions of
the betatron motion with damping term. When the particle energy is slowly increased,
we can write the solution with the following approximate form:

y(s)  A(s) cos( K s   ) (103)

, where the amplitude function A(s) is a slowly varying function of coordinate s.

Then, we just do the algebra:

y  A cos( K s   )  K A sin( K s   )
y  A cos( K s   )  2 K A sin( K s   )  KA cos( K s   )

p
y  ( ) y  K ( s ) y
p
 A cos()  2 K A sin()  KA cos() 
p
( )[ A cos()  K A sin()]  KA cos() (104)
p
0
44
 p
A cos()  2 K A sin()  ( )[ A cos()  K A sin()]  0 (105)
p
p p
[ A  ( ) A] cos( K s   )  K [2 A  ( ) A] sin( K s   )  0 (106)
p p

2nd order slowly varying coefficient •The coefficient in front of

small enough to be neglected sine function must be zero

p
2 A  ( )A  0
p
A  1 p

A 2 p
dA  1 dp

A 2 p
1
ln A  ln  const. (107)
p
a a
A  y ( s)  cos( K s   ) (108)
p( s) p( s)
45
, where a is a constant depending on the initial conditions.
The amplitude of betatron motion is inversely proportional to the momentum
when the particle energy is adiabatically increased !
The beam emittance is related to the rms amplitude of betatron motion by the
following,
2
xrms
1 1
   (109)
bx p m0 bc

The invariant emittance N is written in the normalized form:

 N  b (110)

Only the normalized emittance is invariant during the particle acceleration. The
geometric emittance is inversely proportional to the particle momentum.

46
 Transformation Matrix of Courant-Synder Parameters

From the Courant-Snyder invariant as shown in Eq.(69), we can contemplate the

existence of another transformation for the envelope function b (s),  (s), and  (s).

At any position s0 ,  (s0 ) y0 2  2 (s0 ) y0 y0  b (s0 ) y0 2   (111)

Knowing that the phase space coordinate (y, y’) at any subsequent position s can be
found by the transfer matrix M,

 y ( s)   m11( s | s0 ) m12 ( s | s0 )   y0 
 y( s)  m ( s | s ) m ( s | s )   y  (112)
   21 0 22 0   0

If we can use the transformation of (y0, y’0 ) to (y, y’) and substitute the relationships
to the Courant-Snyder invariant at position s0, we should be able to obtain an
equivalent expression in the form of:

 y 2  2 yy   y2   (113)

 (s)  (s) b (s)

47
y0 , y’0 , mij (s|s0 )
From the following transformation,
1
 y0   m11( s | s0 ) m12 ( s | s0 )   y ( s ) 
 y  m ( s | s ) m ( s | s )  y( s )
 0  21 0 22 0   
 m22 ( s | s0 )  m12 ( s | s0 )  y ( s) 
     (114)

 21m ( s | s 0 ) m11 ( s | s 0  
) y ( s ) 

After some calculation, we arrive at the following transformation matrix for the
Courant-Snyder parameters:

 b ( s)  m11  2m11m12   b ( s0 )

2 2
m12
 ( s)    m m m m  m m  m m

 ( s ) 
   11 21 11 22 12 21 12 22   0 

  ( s)   m21
2
 2m21m22 2
m22    ( s0 ) 

(115)

48
The position dependence of Courant-Synder parameters along a drift space can
be easily found as:
 b ( s) 1  2s s   b ( s0 )
2

 ( s)   0 1  s

 ( s ) 
    0  (116)
  ( s)  0 0 1    ( s0 ) 

For the special case when  (s0)= 0, e.g. in the middle of straight section,
we have the following relations:

 s2 
b 0  
 b ( s)  b0 
 ( s )     s 
   b  (117)
  ( s )   0

 1 
 b 0 

49
 A simple application of transfer matrix

Septum (a pulsed dipole magnet)

qk e- xs

Pulsed dipole (kicker) Fig. 37

 xs   m11 bi b f sin(y f y i )   0 
 x      (118)
 s   m21 m22  q k 

xs  q k bi b f sin(y f y i ) (119)

 Homework *) How to estimate the required strength of pulsed dipole to generate a

horizontal displacement xs downstream? (do it in the class)
50
Example: Beam extraction from a synchrotron

SE1
KE

Booster lattice
MSE

Fig. 38
‘B6P3.mad8’ lattice:
51
Qx= 16.406, Qy= 11.391, x0= 5.82 nm-rad, xh= 0.99, xv= 1.82
The required kick strength: 2 extraction kickers,
xSE1 each kicker (L= 0.6 m, q= 1.4 mrad)
q (120)
sin KS b KE b SE1 0.2 m
QF
SE1

Fig. 39
Lattice ‘BTS_B6P3-A1.mad8’
2 extraction kickers (in-vacuum)
52
From booster ext. kicker SE1 BTS lattice
SE2
2 kickers

Si1 Si2

Fig. 40

53
 Beta Functions in a Transfer Line

Can we still use the concept of beta function in a transfer line where no periodicity
exists in the magnet structure?
x(s)  K (s) x(s)  0
x(s)  w(s)eiy (s )
We still need to find the general solutions for the above equation, except there is no
periodicity in K(s) and the amplitude function w(s) for a transfer line.
Mathematically we can still guess the solution x(s) has the general form as shown
above. Because there is no periodicity in K(s) for this case, there is no periodicity in
b(s). As a result, we will not have the periodic motion in the normalized phase space,
which is a characteristic feature of magnet structures with periodicity.

 YES, we can still define the beta function b(s) as we have done for the solutions of
Hill’s equation. BUT the beta function does not have periodicity! The b(s) and (s)
functions depend on the incoming beam (the initial Courant-Synder parameters) and
the lattice design of transfer line as well. This can be readily understood from
Eq.(115).

54
 Circular accelerator
b(s) and (s) depend on the whole magnet structure. Any change in magnet at
one location will affect the b(s) and (s) everywhere around the circular
accelerator.

Transfer line
b(s) and (s) depend on the property of incoming beam and the magnet
structure of transfer line. Any change in magnet at one location will only affect
the b(s) and (s) downstream along the transfer line.

Up to now, all the discussions are about the motion of particles with the nominal
design energy. In reality, charged particles are generated with some energy
distribution among them. Most charged particles are not on the design energy.
We need to understand the dynamics of those off-energy particles in an
accelerator. Besides, there are always fabrication errors in the magnets

Transverse motion with driving terms (energy error)

55
 Transverse Motion with Error Terms (Driving Terms )

Consider an inhomogeneous second order linear differential equation,

d2y dy
2
 P ( x )  Q( x ) y ( x)  F ( x ) (121)
dx dx
if 1(x) and 2(x) are solutions of the homogeneous differential equation, i.e.

1 P( x)1  Q( x)1  0

2  P( x)2  Q( x)2  0
Using the procedure of variation of constants, we can get a general expression for
the particular solution in terms of the complementary solutions:
F (t )2 (t ) F (t )1 (t )
x x
y p ( x)  1 ( x)  dt  2 ( x)  dt
x0
W (1 (t ), 2 (t )) x0
W (1 (t ), 2 (t ))
(122)
F (t )2 (t ) F (t )1 (t )
x x
yp ( x)  1( x) 
x0
W (1 (t ),2 (t ))
dt  2 ( x)x
W (1 (t ),2 (t ))
dt
0

, where W(1, 2) is the Wronskian

56
[Ref.] E.A. Coddington, An Introduction to Ordinary Differential Equations (Prentice Hall, 1961)
 1 2
W (1 , 2 )  (123)
1 2

Recall the general properties of the homogeneous solutions of Hill’s equation as

given by Eqs.(39 & 40), the particular solution of an inhomogeneous Hill’s equation
is:
z z
x p ( z )  S ( z )  F (t )C (t )dt  C ( z )  F (t ) S (t )dt (124)
z0 z0

, where F(t) is the driving term of an inhomogeneous Hill’s equation.

How the hell do we get an inhomogeneous Hill’s equation? Who orders these
errors?
i) off-momentum particle (a spread in the particle momentum is
inevitable)
ii) errors in field strength of magnets (fabrication tolerance)
iii) errors in alignment (position and orientation) of magnets
iv) errors in the electric current that drives magnets (power supply, quality
of electricity delivered from the power company) 57
  Digression on mathematics:
Particular solution of 2nd order inhomogeneous differential equation

Consider the following differential equation,

d2y dy
2
 P ( x )  Q( x ) y  F ( x)
dx dx

1(x) and 2(x) are solutions of homogeneous differential equation, i.e.

 1 P( x)1  Q( x)1  0

2  P( x)2  Q( x)2  0
Using the procedure of variation of constants, we can write the particular solution
of inhomogeneous differential equation as:
y p ( x)  u( x)1 ( x)  v( x)2 ( x) (d1)

then, yp ( x)  u1  u1  v2  v2

 (u1  v2 )  (u1  v2 )
If we require the following condition,
u1  v2  0 (d2) 58
 then, yp ( x)  u1  v2 (d3)

yp ( x)  u1  u1 v2  v2 (d4)

We substitute Eqs.(d1, d3, d4) into the original 2nd order inhomogeneous
differential equation, we get:
(u1  u1 v2  v2)  P( x)(u1  v2 )  Q( x)(u1  v2 )  F ( x)
u[1 P( x)1  Q( x)1 ]  u1  v[2  P( x)2  Q( x)2 ]  v2  F ( x)

u1  v2  F ( x) (d5)

Equations (d2) & (d5) give the following set of relations,

 u1  v2  0 (d6)


u1  v2  F ( x)
(d7)

We substiute Eq.(d6) to Eq.(d7): replacing u’ by v’

2
(v )1  v2  F ( x)
1
v(12  12 )  F ( x)1 59
From the definition of Wronskian,
1 2
W (1 , 2 )   12  12
1 2
v W (1 , 2 )  F ( x)1 ( x)

F (t )1 (t )
Therefore, x
v( x)  x W (1 (t),2 (t ))dt (d8)
0

Next we replace v’ by u’:

1
u1  (u )2  F ( x)
2
u(21  12 )  F ( x)2
 u W (1 , 2 )  F ( x)2 ( x)
Therefore,
F (t )2 (t )
x
u ( x)    dt (d9)
x0
W (1 (t),2 (t )) 60
Finally we put Eqs.(d1),(d8), and (d9) together, and we have a general
expression for the particular solution in terms of the complementary solutions:

F (t )2 (t ) F (t )1 (t )
x x
y p ( x)  1 ( x)  dt  2 ( x)  dt
x0
W (1 (t ), 2 (t )) x0
W (1 (t ), 2 (t ))
F (t ) 2 (t ) F (t )1 (t )
x x
y p ( x)  1( x)x
W (1 (t ), 2 (t )) 
dt   2 ( x)
x
W (1 (t ), 2 (t ))
dt
0 0

End of digression

61
 Dispersion Function

Consider the horizontal motion with a driving term due to off-momentum error,
1 dp dp
Eq.(25) x( s)  K x ( s) x( s)  ( )  o( x ) (125)
F(t)
r p0 p0
z z
C (t ) dp S (t ) dp
x p ( z)  S ( z)  ( )dt  C ( z )  ( )dt (126)
z0
r (t ) p0 z0
r (t ) p0
z z
C (t ) dp S (t ) dp
xp ( z )  S ( z )  ( )dt  C ( z )  ( )dt (127)
z0
r (t ) p0 z0
r (t ) p0

The closed orbit of an off-momentum particle can be expressed in the form:

dp dp
x p  D( s ) xp  D( s) (128)
p0 p0
, and we look for the closed solution. The solution must satisfy the closed orbit
condition:
D(s  L)  D(s) , D(s  L)  D(s) (129)
, where L is the circumference of the circular accelerator. This function D(s) is 62
referred to as the dispersion function.
Using the piecewise method, we can find the particular solution of inhomogeneous
Hill’s equation for D(s), 1
D( s)  K x ( s) D( s)  (130)
r
• Drift space and quadrupole magnet are straight sections
1
0 Same as the eq. of horizontal betatron motion
r
• Sector dipole:
D  r (1  cos q )
D  sin q (131)

The 2×2 transfer matrix in Eq.(39) can be expanded into a 3×3 matrix:

   
 x ( s )   C ( s , s 0 ) S ( s , s0 ) D   x ( s 0 ) 
   
x ( s)   x( s)  C ( s, s0 ) S ( s, s0 ) D  x( s0 )  M ( s, s0 ) ~
~ x ( s0 ) (132)
 dp   0 0 1   dp 
 p0   p 
 0 
Particular solutions of inhomogeneous Hill’s eq. for off-momentum 63
If we set dp/p0=1 in Eq.(132), we get the 3×3 transfer matrix for the dispersion function,

 D( s)   m11 m12 m13   D( s0 ) 

D( s)   D( s)  m21 m22 m23   D( s0 )  M ( s, s0 ) D( s0 )
~ ~
(133)
 1   0 0 1   1 

•For circular accelerators the periodicity imposes a closed orbit for the dispersion
function D(s), so that D(s0  L)  D(s0 ) , D(s0  L)  D(s0 )
, where s0 is any point around the accelerator, and L is the period length of the
accelerator lattice.

 D( s0 )   m11 m12 m13   D( s0 ) 

 D( s )  m m23   D( s0 )
 0   21 m22
 1   0 0 1   1 

(1  m22 )m13  m12m23 (1  m11)m23  m21m13

D( s0 )  D( s0 )  (134)
2  m11  m22 2  m11  m22
64
mij is the element of one-period transfer matrix.
In comparison to Eqs.(29-32), we have the following results:
i) Drift space
1  0
M  0 1 0
(135)
0 0 1

ii) Sector dipole (hard edge)

 cos q r sin q r (1  cos q )
 1 
M x   sin q cos q sin q 
r
(136)

 0 0 1 

1  0
M y  0 1 0
0 0 1

65
iii) Quadrupole
 1 
 cos  |K|
sin  0
K> 0  
(focusing): M   | K | sin  cos  0
(137)
 0 0 1

 

K< 0  1 
(defocusing):  cosh  |K|
sinh  0
 
M   | K | sinh  cosh  0
 0 0 1
(138)

 

66
 iv) Rectangular dipole

 1 0 0  cos q r sin q r (1  cos q )  1 0 0

 tan(q 2)   1   tan(q 2) 
Mx   1 0  sin q cos q sin q  1 0
 r r  r  (139)
 0 0 1  0 0 1   0 0 1

Momentum Compaction Factor (Dilation Factor)

dp (140)
x x ( s )  xb ( s )  D ( s )
d  ds(1  ) p0
r x
DL   ds
r r
A global property of
DL dp
x c  ( ) ( ) accelerator lattice
L p0
1 D( s)
Fig. 41  
L r (s)
ds (141)
67
ideal orbit
 The Complete Solution of an Inhomogenous Hill’s Equation

The betatron motion with driving terms which are independent of the transverse
displacement is given by the following form,
y( z)  K ( z) y( z)  F ( z) , K ( z)  K ( z  L)

The solutions of homogenous Hill’s equation in a circular accelerator satisfy the

following properties,
 y ( z )   C ( z ) S ( z )   y0   y0 

 y( z ) C ( z ) S ( z )  y   M   y 
     0  0
, where det(M)=1, i.e. CS’-C’S=1, C(z): cosine-like function, and S(z): sine-like
function.
The complementary solutions of homogenous Hill’s equation can be expressed in
the following notation,
1 ( z )  y0C ( z )
 (142)
2 ( z )  y0 S ( z )
Then, we substitute Eq.(142) into the general expression of particular solution as
given by Eq.(122).
[Ref.] “linearized error analysis for an accelerator…”, Nucl. Instrum. and Methods, A346 (1994)1, R.K. Koul, F.E.
68
Mills.
We get the following result for the particular solution of an inhomogenous Hill’s equation:
z z
y p ( z )  S ( z )  F (t )C (t )dt  C ( z )  F (t ) S (t )dt
z0 z0
z
  [ S ( z )C (t )  C ( z ) S (t )]F (t )dt
z0
z
  G ( z , t ) F (t )dt (143)
z0

, where G(z,t)= S(z)C(t)-C(z)S(t). G(z,t) is the Green’s function.

The derivative of yp(z) is: z z

yp ( z )  S ( z )  F (t )C (t )dt  C ( z )  F (t ) S (t )dt (144)
z0 z0

The complete solution of an inhomogeneous Hill’s equation is:

y ( z )  y0C ( z )  y0 S ( z )  y p ( z )
(145)
y( z )  y0C ( z )  y0 S ( z )  yp ( z )

, where y0  y( z0 ) , y0  y( z0 ) 69

Therefore, we have the following results:
C ( z0 )  1 , S ( z0 )  0 , C ( z0 )  0 , S ( z0 )  1 (146)
y p ( z0 )  0 , yp ( z0 )  0

The complete solution must satisfy the periodic condition which indicates the
following:
 y ( z0 )  y ( z0  L)

 y( z0 )  y( z0  L)
(147)

, where L is the circumference of circular accelerator.

The periodic condition (closure condition) as shown in Eq.(147) gives the following
results:
y0  y0C ( z0  L)  y0 S ( z0  L)  y p ( z0  L) (148)

y0  y0C ( z0  L)  y0 S ( z0  L)  yp ( z0  L) (149)

Let’s define the following notations:

C1  C ( z0  L) , S1  S ( z0  L) , y p ( z0  L)  u1 , yp ( z0  L)  u1
C1  C ( z0  L) , S1  S ( z0  L) (150) 70
Then, we can rewrite Eqs.(148, 149) as
y0  y0C1  y0 S1  u1 (151)
y0  y0C1  y0 S1  u1 (152)

Since z0 is an arbitrary longitudinal coordinate, we can obtain the explicit form of y(z)
if we can solve y(z0) !

Using the explicit expression of one-turn transfer matrix in Eq.(66), we arrive at the
following result for y0,
z0  L
1
y0  y ( z0 ) 
2sin p Q 
z0
F (t ) b 0 b (t ) cos[y (t )  y ( z0 )  p Q]  dt (153)

Hence, we have the complete solution of an inhomogenous Hill’s equation:

sL
1
y(s) 
2sin p Q  F (t )
s
b ( s ) b (t ) cos[y (t ) y ( s)  p Q]  dt (154)

To evaluate the above integral, we need to define the origin of longitudinal coordinate
first. Then, we can determine the explicit value of betatron phase (s) and proceed
with the integral. This dependence on the choice of coordinate origin is not
71
convenient in practice.
The integral in Eq.(154) can be written as two separate terms
L sL
1
y(s) 
2sin p Q s 
{ F (t ) b ( s ) b (t ) cos[y (t )  y ( s)  p Q]dt   F (t )
L
b ( s) b (t ) cos[y (t ) y ( s)  p Q]dt}

Using the periodicity of circular accelerators, we know that

b (t )  b (t  L)
F (t )  F (t  L)
 (t )   (t  L)   ( L)

Therefore, Eq.(154) can be rewritten as 2pQ

L sL
1
y(s)  {
2sin p Q s  cos[y (t )  y ( s)  p Q]dt   F (t  L )
L
b ( s) b (t  L) cos[y (t  L)  y ( L)  y ( s)  p Q]dt}

By changing the variable of the second integral from t to t’= t-L, we arrive at
L s
1
y(s)  {
2sin p Q s  
cos[y (t )  y ( s)  p Q]dt  F (t ) b ( s) b (t ) cos[y (t )  y ( s)  p Q]dt}
0

In the first integral y (t ) y ( s)  0 , and the second integral  y (t ) y ( s)  0

L s
1
y(s)  {
2sin p Q s  
cos(|y (t )  y ( s) | p Q) dt  F (t ) b ( s) b (t ) cos(  | y (t )  y ( s) | p Q) dt}
0
L
1

2sin p Q 0 
F (t ) b ( s ) b (t ) cos(| y (t )  y ( s) | p Q) dt 72
The complete solution of Hill’s equation with an error term independent of
displacement is expressed as the integral of error distribution with Green’s function
L
1
y(s) 
2sin p Q 0 
F (t ) b ( s ) b (t ) cos(| y (t )  y ( s) | p Q) dt

L
(155)

 F (t )G ( s, t )dt
0

, where the Green’s function is

1
G ( s, t )  b ( s ) b (t ) cos(| y (t ) y ( s) | p Q)
2sin p Q

In real accelerators the error distribution is discrete. If we use the thin-lens

approximation for the error source, we can approximately treat b, , and F as
constant over the length  of the error source. Then we can approximate the integral
by a series of summation over N error terms.

b (s) N
y(s)  
2sin p Q n 1
 n b n cos(|y n y ( s) | p Q)

where  n  n Fn
73
If we perform the Floquet transformation to the normalized coordinate as defined in
Eqs.(53) and (54), the complete solution of Hill’s equation with an error term F(s) in
the normalized coordinate system (u(), Pu()) is given by
ds
d 
Qb ( s)
L
1
y(s) 
2sin p Q 0 
F (t ) b ( s ) b (t ) cos(| y (t )  y ( s ) | p Q)dt

2p
y(s) 1
u ( )  
b ( s ) 2sin p Q 0
Q b 3/2 ( ) F ( ) cos(|  (t )   ( s) | p Q) d

2p
1
=
2sin p Q  f ( ) cos(|  (t )   (s) | p Q)d
0

f ( )  Q b 3/2 F ( )

The error term f() is a periodic function of 2p. We can express the error term
by Fourier series.



1
f ( )  Q b ( ) F ( )  f n e jk , f k   Q b 3/2 F ( )e  jk d
3/2

k  2p



Q fk jk
u ( )  e
k  Q  k
2 2

74
 Equation of Betatron Motion with Dipole Errors

error term due to

Recall the transverse equation of motion for an
momentum deviation:
off-momentum particle as given by Eq.(125), the
corresponding driving term F(s) is:
1 dp
F ( s)  This is equivalent to a dipole error
r p0
Let’s plug the driving term (error term) caused by a momentum deviation into Eq.(155).
The explicit expression of closed orbit deviation for an off-momentum particle is:
dp b ( s ) b (t )
L
1
xD ( s )  ( )
2sin p Q 0 p0 r
cos(| y (t )  y ( s) | p Q)  dt (156)

Therefore, the horizontal dispersion function D(s) is given by:

b ( s ) b (t )
L
xD ( s ) 1
D( s)  
(dp p0 ) 2sin p Q 
0
r
cos(|y (t )  y ( s) | p Q)  dt (157)

Integer tune should

be avoided! 75
 Other notation for the dispersion function,
D(s)   (s)
The equations of betatron motion with field errors are given as followings:

 p0 1 dp  1 dp DBy
x( s)   K x ( s)( )  2 ( ) x( s)  ( )  0 (158)
 p r p0  r p0 By r
p0 DB
y( s)  K y ( s)( ) y( s)  0 x (159)
p By r field errors
(terms other than nominal dipole & quadrupole)
p0
, where By0  By ( s,0,0) ,  qB0 qBy(0)> 0 must be satisfied
r
Note that the driving terms of differential equations have opposite sign for the
horizontal and the vertical direction respectively.

error term due to dipole DB

F ( s) 
field of bending magnet:
By0 r
Let’s plug F(s) into the expression of complete solution as given in Eq.(155) and
assume the dipole field error only occurs at location s1. The closed orbit distortion
is given by the following result: 76
 b ( s ) b ( s1 ) DB 
xc ( s )  cos(| y ( s1 )  y ( s) | p Q) (160)
2sin p Q B0 r

, where  is the length of error source (a dipole magnet for this case). The change of
circumference due to a single dipole error ∆𝐶 = (𝑥𝑐𝑜 𝜌)𝑑𝑠 = 𝐷(𝑠0 )𝜃𝑑 (𝑠0 ) .

The closed orbit distortion xc(s) due to an uniform dipole error can be expressed in
terms of the deflection angle and the Green’s function by Eqs.(12) & (155) respectively,
xc ( s )  G ( s, s1 )q
DB  b ( s ) b ( s1 )
q ; G ( s, s1 )  cos(| y ( s1 )  y ( s) | p Q) (161)
By r
0
2sin p Q

An integer tune should not be used in order to avoid the orbital

resonance driven by errors of dipole fields.
i.e. Q≠0, 1, 2, 3,…, N

Considering the dipole-type errors, Eq.(140) is expanded as

dp dp
x ( s )  xb ( s )  D ( s ) x ( s )  xb ( s )  D ( s )  xc.o ( s )
p0 p0
77
Example: A misaligned quadrupole magnet with a transverse offset dx from the design orbit.
f
By
Dx DB   Dx
x
e-
s xc ( s)  G ( s, s1 )q x
q x  K x   Dx (162)
Fig. 42

Amplification factor— the ratio of rms closed orbit distortion to the rms
misalignment of the quadrupole magnet.
xCOD
Ax 
Dx
  | G ( s, s ) |
i
i
2
K i2  2i (163)

 If the error kick occurs at a location where the beta function is larger, then the
resulting distortion of closed orbit is also larger.

[Refs.] P.M. Gygi-Hanney, J.M. Jowett, E. Keil, CERN/LEP/TH/88-2 (CERN internal report, 1987)
A. Streun, SLS-TME-TA-1999-0014 (Swiss Light Source internal report, 1999) 78
Example: Closed orbit distortion in TPS booster vs. leakage of DC extraction septum

Total estimated integrated leak field~ 6 T-mm

Accelerator parameters at 150 MeV (no error)

E [GeV] 0.15
0.000

Qx 14.380 -0.002

bx(SE) [m] 6.286 -0.004

-0.006

bx(Q1) [m]

By (T)
15.533 -0.008

|(SE)- (Q1)| [rad] (14.348-13.780)*2p -0.010

-0.012

leakage - Iy
-0.014

-0.016
-1000 -500 0 500 1000
Z (mm)

79
 SE1
File= b6p4dleak.mad8
Q1

• The leak field of DC septum is modeled at the middle of septum as an angular kick
80
 Angular kick (integrated leak field)= 1000 Gauss-cm
 MAD8 simulation: Xco(max)= 10.54 mm, Xco(rms)= 5.38 mm
closed orbit distortion due to a dipole kick
15

10

5
X(co) [mm]

-5

-10
error kick

-15
0 50 100 150 200 250 300 350 400 450 500
s [m]
Analytic formula MAD8 simulation

X(Q1)= -7.65 mm X(Q1)= -7.48 mm 81

 Floquet transformation of betatron motion: 𝑢 𝑄𝑥 = 𝑥(𝑠)/ 𝛽(𝑠)

closed orbit distortion due to a dipole kick

(after Floquet transformation)
0.1

0.08

0.06
X_Floquet [sqrt(mm)]

0.04

0.02

-0.02

-0.04

-0.06

-0.08
error kick
-0.1
0 3 6 9 12 15
Qx

82
The change in circumference due to the dipole field errors
The dipole field error will cause the closed orbit distortion, which results in the
change of path length. The circumference of particle orbit is given by
x( s )
L  (1  x r ) 2  x2  y 2 ds  L0   r
ds  o( x2 )

Using the expression of Eq.(155)

1
x( s ) 
2sin p Q 0 
F (t ) b ( s ) b (t ) cos(| y (t )  y ( s) | p Q) dt

We plug the above equation to the expression of change in the accelerator

circumference,
x( s )
DL   r
ds
Eq.(160)

1 DB b ( s ) b (t )
  ( 
2sin p Q 0 B0 r
)
r
cos( y (t )  y ( s)  p Q) dt  ds

DB  1 b ( s ) b ( s0 )
(
B0 r
) 
2sin p Q r
cos( y ( s)  y ( s0 )  p Q) ds

From the expression of dispersion function as given in Eq.(157), we can

rewrite the change in the path length due to a dipole field error as
DB  1 b ( s ) b ( s0 )
DL  (
B0 r
) 
2sin p Q r
cos( y ( s )  y ( s0 )  p Q) ds
83
 DB 
DL =( ) D( s0 )
B0 r
=q 0 D( s0 )
where q0 is the angular kick due to a single dipole field error at the
longitudinal position s0. When there are several dipole field errors
distributed in an accelerator, the change in circumference is given by
DB ( s )
DL   D( s)
B0 r
ds

Since the revolution frequency of particle beam must stay in synchronism

with the RF frequency (harmonic number must be an integer), a change
in the circumference must be offset by a change in the beam energy.
From the definition of the momentum compact factor, we get the
corresponding momentum error as
q 0 D( s0 )

c L
The dispersive closed orbit distortion caused by this energy offset is
q 0 D( s0 ) D( s )
xD ( s )   D ( s ) 
c L

Therefore, the resultant closed orbit distortion due to a dipole field error
is b (s)b ( s )
q0 q 0 D( s0 ) D( s )
xc.o ( s )  x x 0
cos( y ( s )  y ( s0 )  p Q) 
2sin p Q r c L
84
Application of Floquet transformation– error finding

1) Moving averaging the turn-by-turn BPM data to get DC term,

𝑥 = 𝑥𝑐𝑜𝑑 + 𝐷𝑥 𝛿 + 𝑥𝛽 , DC = 𝑥𝑐𝑜𝑑 + 𝐷𝑥 𝛿, AC = 𝑥𝛽 = 𝜖𝑥 𝛽𝑥 𝑐𝑜𝑠𝜑

𝑥−𝐷𝐶
2) Perform the Floquet transformation, plot 𝑣𝑠. 𝑝ℎ𝑎𝑠𝑒 𝜑
𝛽𝑥

3) Use the piecewise Sine function and adjust the phase to fit the transformed
TbT BPM data with moving average. 𝐴𝑥 sin 𝜑 + 𝜑0 → 𝑎𝑑𝑗𝑢𝑠𝑡 𝐴𝑥 𝑎𝑛𝑑 𝜑0 to fit the
transformed TbT BPM data with moving average.

4) The boundary between different piecewise Sine functions indicates the

location of error source

[Ref.] J-P. Koutchouk, “Trajectory and closed orbit correction”, in Frontiers of Particle Beams;
Observation, Diagnosis and Correction, H. Araki et al., Eds, Springer-Verlag, 1989.
 Error finding by Floquet transformation

s [m]

error error
Ax sin(  0 )

x  DC
bx

 Courtesy of M.S. Chiu

Correction of closed orbit distortion in a circular accelerator —
For a given accelerator lattice model with m beam position monitors (BPMs)
and n correctors (short steering dipoles), the response of accelerator lattice to
each individual corrector can be expressed in terms of matrix form:
∆𝑐 = 𝑇𝜃
, where ∆𝑐 is a m-dimensional column vector and 𝜃 is a n-dimensional column
vector respectively. ∆𝑐 is the collection of measured signals at m BPMs and 𝜃 is
the collection of angular deflections exerted by n correctors. The 𝑚 × 𝑛 matrix T
represents the response of lattice model. Suppose the orbit deviation measured
before correction is ∆0 , the measured orbit distortion after applying the
correction, i.e. turning on correctors, is ∆ . We have the following equation,

∆= ∆0 + ∆𝑐
The square sum of each individual element of orbit vector after the correction,
∆ , is written as 𝑚

𝑆= ∆0𝑖 + ∆𝑐𝑖 2

𝑖=1
To reduce the closed orbit distortion in the accelerator, we should try to minimize
the value of S. The minimum of S is attained when the first derivative of S is
zero: 87
 S D
 0  (D 0i  D ci ) ci  0
q j q j
(D 0i  D ci )Tij  0
, where i = 1, …, m and j = 1, …, n. The dimension of ∆0 + ∆𝑐 is m. Therefore,
the desired settings for correctors to reach the minimum orbit distortion is given
by ~  
T (D 0  D c )  0
~ ~
TD0  TDc
~ ~ 
 T D 0  T Tq
0
 ~ 1 ~
q  (T T ) T D 0
Discussions on various correction strategies can be found in following
references.
[Refs.]
1. W. Herr, CERN-SL-95-07(AP), “Algorithms and procedures used in the orbit correction package COCU”
2. A.S. King, M.J. Lee, P.L. Morton, IEEE Trans. NS-20(1973)898 or SLAC-PUB-1203
3. B. Autin & Y. Marti, CERN-ISR-MA-73-17, “Closed orbit correction of A.G. machines using a small
number of magnets” 88
 The lattice response matrix T can be rewritten as 𝑇 = 𝑈𝑊𝑉 by the analysis of
singular value decomposition (SVD). U and V are 𝑚 × 𝑚 and 𝑛 × 𝑛 orthogonal
matrices, and W is a 𝑚 × 𝑛 diagonal matrix with positive or zero elements. The
solution of corrector vector 𝜃 can be further reduced by the SVD of response
matrix as below
q  (TT ) 1T D 0
 (VWUUWV ) 1  VWU  (D 0 )
 [diag (1 w2j )]  VWU  (D 0 )
 V [diag (1 w j )]U (D 0 )

This is the same result as solving a set of equations 𝑇𝜃 = −∆0 using the SVD
method [ref]. Therefore, the residual closed orbit distortion after the correction is
minimum in the sense of least square. The strength (length of vector 𝜃 ) of
correctors is the smallest to satisfy the constraints imposed by the correction.

[Ref.] W.H. Press et al., “Solution of linear algebraic equations”, in Numerical Recipes (Cambridge)

89
Alternative approach to find the solution of closed orbit in the presence of a dipole error
dp
x ( s )  xb ( s )  D ( s )  xCOD ( s)
p0  x1   x1 
 x  q   x 
 1   1
x x1’-q x1’

s
q Location of
steering error,
thin-lens approximation, 0 magnitude q

Question to be answered: how to predict

the particle trajectory at any longitudinal
position s?

A naïve solution: if we know the explicit

expression of 𝑥1 , 𝑥 ′1 , we can determine the
particle trajectory by multiplying the general
transfer matrix as given in Eq.(65).
90
Step 1) we can find the explicit expression for ( x1, x1’ ) by solving the following set of
equations
 x1  cos 2pQ   sin 2pQ b sin 2pQ   x1 

 x  q  
 1     sin 2pQ cos 2pQ   sin 2pQ   x1 
 qb1
x 
 1 2 tan pQ
How to use this results to store the injected
 q 1 beam during accelerator commissioning?
 x1  (1  )
 2 tan pQ

Step 2) consider the case that the observation point is downstream of the dipole error
and not crossing the origin of longitudinal coordinate
 b2 
 (cos   1 sin  ) b1b 2 sin  
 2 
x b   x1 
 x     (1    )
1

 2 (1   2 ) b1   x1 

1 2
sin   cos  (cos    2 sin  )
 b1b 2 b1b 2 b2 
0
1 dipole error   y ( s2 ) y ( s1 )    0
2 q b1b 2
x ( s2 )  cos[y ( s2 ) y ( s1 )  pQ] 91
2 sin pQ
Step 3) consider the case that the observation point is upstream of the dipole error
and crossing the origin of longitudinal coordinate

 x1   x2 
 x   M ( s1  s2 )  x 
0
2

1 dipole error
 1  2
 x2  1  x1 
 x   M ( s1  s2 )  x 
 2  1
q b1b 2
x ( s2 )  cos[pQ y ( s2 ) y ( s1 )]
2 sin pQ
y ( s2 ) y ( s1 )  0

Combining the results in preceding steps and the property of cosine as an even
function, we get the general solution of closed orbit at any location s when there is a
dipole error at s1:
q b ( s1 ) b ( s )
x( s )  cos(|y ( s1 ) y ( s ) | pQ)
2 sin pQ

For N discrete dipole errors, the closed orbit is given by the summation of all terms
b (s) N
y(s)  
2sin p Q n 1
q n b n cos(|y n y ( s) | p Q)
92
where q n  n DBn B0 r
 Betatron Motion with Gradient Errors

The Hill’s equation for unperturbed betatron motion is

y(s)  K0 (s) y(s)  0

When we include a gradient error to the quadrupole strength, the Hill’s equation of
perturbed betatron motion is,
y(s)  [ K0 (s)  DK (s)] y(s)  0

y(s)  K (s) y(s)  0

where K(s)= K0 (s)+DK(s), and DK(s) is a small perturbation.

The perturbed quadrupole strength K(s) satisfies a weaker periodic condition:

K (s)  K (s  L)
, where L is the circumference of circular accelerator.

93
 m(s1 ) Betatron tune shift due to gradient errors
＋ －
The transfer matrix for an infinitesimal localized
s1 gradient error is:
 1 0
m( s1 )    (164)
s2   DK ( s1 ) ds1 1 

Assuming the gradient perturbation with an infinitesimal

M0(s1+L| s1 ) length ds1 at location s1.
Fig. 43
The perturbed one-turn transfer matrix M(s1+L | s1) is given by the following relation:
M (s1  L | s1 )  M 0 (s1  L | s1 )  m(s1 ) (165)
Recall the following relation:
K(s)  b(s)  betatron phase advance y(s)  betatron tune Q

Applying the explicit form of one-turn transfer matrix as given by Eq.(66) and the
property shown in Eq.(74), we get the expression of betatron tune shift DQ caused by
a small gradient error :
1
DQ  b ( s1 )DK ( s1 )ds1 (166) 94
4p
 Betatron function beating due to gradient errors

m(s1 ) error
＋ －
M0(s2+L| s1 ) s2 s1 s2+L
Fig. 45
s1
In principle we know that the betatron
s2 function, i.e. the amplitude function, will be
perturbed by a gradient error according to the
betatron envelope equation as given by
Eq.(48).
How to solve for the perturbed betatron function?
M0(s1| s2 )
We can still make good use of the one-turn
Fig. 44 transfer matrix.

The perturbed one-turn transfer matrix for any downstream location s2 of the gradient
error at s1 can be expressed by the unperturbed transfer matrices as,
M (s2  L | s2 )  M 0 (s2  L | s1 )  m(s1 )  M 0 (s1 | s2 ) (167)

A naïve solution: knowing the tune shift due to gradient error, we can learn about the
perturbed betatron function from the m12 element of one-turn transfer matrix. 95
From Eq.(65)

 b1 
 (cos y   sin y ) b b sin y 
b
2 1 2

M 0 ( s1 | s2 )   2 
  (1  1 2 ) ( 2  1 ) b2 
 siny  cosy (cosy  1 siny )
 b1b 2 b1b 2 b 1 
y   ( s1 )   ( s2 ) the betatron phase advance from s2 to s1.

 b2 
 (cos   1 sin  ) b1b 2 sin  
b1
M 0 ( s2  L | s1 )   
  (1  1 2 ) (   2 ) b1 
 sin   1 cos  (cos    2 sin  )
 b b
1 2 b b
1 2
b2 
  2pQ0 y

Considering a gradient error in a focusing quadrupole as given by Eq.(164), the matrix

element M12 in Eq.(167) is expressed explicitly as:

96
 ~
b 2 sin(2pQ)  b 2 sin(2pQ0 )  DKds1b1b 2 siny sin  (168)

M12 on the L.H.S. of Eq.(167) M12 on the R.H.S. of Eq.(167)

~
b Is the perturbed betatron amplitude function, and Q is the perturbed betatron tune.

We use the betatron tune shift as given by Eq.(166) and keep the first order terms in
the detailed calculation. Our goal is to obtain the explicit expression for the deviation
of b(s).
The deviation of betatron amplitude function at any location s downstream of the error
source (s1) is:
Db ( s)  DKds1
 b ( s1 ) cos[2( ( s1 )   ( s)  pQ0 )] (169)
b ( s) 2 sin(2pQ0 )

, where (s) is the betatron phase at location s.

Equation (169) indicates that there will be an amplitude modulation at two times
the betatron frequency due to gradient errors. This phenomenon is also called
‘beta beating’. It also shows that a half-integer tune will lead to orbital resonance
in the presence of gradient errors.
A half-integer tune should be avoided, i.e. Q≠N+1/2 . 97
Example: beta beating
TLS low emittance lattice@ 1.5 GeV, ex= 13.8 nm-rad, Qx= 7.460, Qy= 4.398

Q2

Only one Q2 has 1% error in the quadrupole strength K

98
 Beta Beating
120
Fig. 46
Q2[K]= 2.65377842*1.01 70

Dbeta/beta0[%]
(Qx, Qy)= (7.481, 4.394)

20

0 20 40 60 80 100 120
-30

-80
s [m]

Beta Betaing

120

70
Dbeta/beta0[%]

20

0 2 4 6 8
-30

-80
99
Qx
Example: TPS storage ring with bare lattice (one period)

OPTICAL FUNCTIONS TPS 79H2 (26.18, 13.28)

30

b emttance = 1.6 nm-rad

x
25
Optical functions (m)

20
b
y

15

10

5
10*  x

0
0 10 20 30 40 50 60 70 80
S(m)

100
 Girder Support and COD Correction Scheme

CV CH CV CH CV CH CV

Precision ~ 15 mm

7 BPM each cell

3 HC(+1) and 4 VC(+1) each cell for SVD
but all sextupoles are with HC and VC.
101
Betatron phase y(s) vs. longitudinal coordinate

102
Example: A rotated (misaligned) quadrupole w.r.t. the vertical axis in TPS storage ring
yaw & displacement

x  Beta beating (max.) :

0.23 mm
16.4% in bx
y
s 5.3% in by.
 Xmax(cod) = 2.055 mm
Xrms(cod) = 0.961 mm

Feed-down effects of misaligned magnets:

Misaligned sextupole magnets quadrupole errors and dipole error
Misaligned quadrupole magnets dipole errors

103
Bare lattice with one quadrupole misaligned-- yaw (sextupole on)

Amplitude function Closed orbit

 Beta beating is observed.  Closed orbit distortion in the x

direction
 Xmax(cod) = 2.055 mm
Xrms(cod) = 0.961 mm
104
Beta beating

 Maximum beta beating is 16.4% in bx  Maximum beta beating is 0.12% in

and 5.3% in by. bx and 0.06% in by.

∆𝛽 𝛽𝑒𝑟𝑟𝑜𝑟 −𝛽0
(%)= x100%
𝛽 𝛽0
Feed-down effects of sextupole magnets

105
Example: Troubleshooting performance degradation after Tai Power annual maintenance

Beta beating

Measured on 2016/12/20 EPU48@ double mini-betY

106
Deviations of quadrupole magnets setting

Absolute deviation of set value Relative deviation of set value

QL1
At DMB Section (cell20~21)

DMB

Measured on 2016-12-20, compared with lattice data on 2016-11-2

107
Energy error due to deviation of driving current of bending magnet in TPS storage ring. The
deviation of dipole driving current DI~ 0.41 A which resulting in a significant drop of beam
lifetime (12 hr less than 4 hr), orbit deviation for beamlines, and tune shift.

0.41 A (DI/I= 6.7e-4)

108
Although we performed lattice correction with LOCO when the output of dipole
current was still drooping and brought the working tune back to the original values,
the beam lifetime was greatly reduced from 12 hr to less than 4 hr. Users
complained the photon beam disappeared. Why?

109
Example: Adjustments of quadrupole strength required by LOCO correction, performed
on 2018-8-4

This result seems to suggest the beam energy is deviated from the nominal value
𝜕𝐵
𝐾= (𝐵0 𝜌), when 𝐵0 𝜌 is increasing K is decreasing, vice versa
𝜕𝑥

110
 Observation of strong correlation between betatron tune and dipole current, 2018-8-8

I (mA)

Tbeam (min)
Δ𝐼
= 5.72 × 10−5
𝐼
Dipole

0.153
nx
0.150 p-p: 3x10-3
0.244
ny
0.242

RF
Simulations using the model lattice. All quadrupole strengths varied by a factor 1+/- 3x10-5

Quadrupole strength Betatron tune

B0 (dipole) decreased K0*(1+3x10-5) DQx= 2.06x10-3, DQy= 7.6x10-4
B0 (dipole) increased K0*(1-3x10-5) DQx= -2.07x10-3, DQy= -5.6x10-4

Simulations confirm the observed fluctuation of betatron tune is mainly

caused by the slow variation of dipole current

112
• Earthquake 2019/04/18 13:01:43

13:01:43 非注射期間
發生的順序如下：
 Earthquake
 SRF trip
 Beam trip

1. 發生震度三級地震

Courtesy of C.H. Huang (I&C group)

 2019/04/18 13:01:43

3. Beam trip

2. SRF trip

Courtesy of C.H. Huang (I&C group)

 2019/04/18 13:01:43

2. Beam trip

Beam trip

1. 發生震度三級地震
Earthquake
x (mm)

y (mm)

Courtesy of C.H. Huang (I&C group)

Example: Spectacular and unusual
An earthquake in New Zealand (M 7.8) at 11:02 UTC on November 13th 2016 (12:02 local LHC
time) is visible together with the earth tides on the mean LHC orbit (here expressed in relative
energy change) during a very long fill with proton and lead ion beams.

https://lhc-beam-operation-committee.web.cern.ch/lhc-beam-operation-committee/LBOC-Documents.html

New Zealand magnitude-7.8 earthquake:

http://www.bbc.com/news/world-asia-37968715

116
 Tune diagram

Qy

Qx

transverse resonance: mQx+nQy=p 117

Horizontal Phase Space with 5th order resonance

118
Example: Multipolar errors of magnets nonlinear resonance, diffusion of particles,
reduction of momentum aperture, reduction of dynamic aperture, shorter beam
lifetime

x’(mrad)

y’(mrad)
x(mm) y(mm)

Tune diagram Dynamic aperture

Y(m)

Qy

119
Qx X(m)
 Chromaticity (gradient error due to momentum deviation)

From the expression of focusing strength as given by Eqs.(21, 22 and 24), the
effective focusing strength as experienced by an off-momentum particle is different
from the on-momentum particle. Therefore, the betatron tune of an off-momentum
particle is different from the nominal betatron tune of on-momentum particle.
F
on-momentum dp
 DK ( s)  DQ
p0

off-momentum

Fig. 47
Let’s recall the Hill equation for off-momentum particle as shown in Eqs.(25, 26), the
corresponding error in the effective quadrupole strength due to the momentum
deviation of particles is given by:
p0 1 gradient error linearly proportional to dp
K ( s )( )  K ( s)
p 1  (dp p0 )
dp dp (170)
 K ( s)(1  )  K ( s)  K ( s)( ) 120
p0 p0
Therefore, the total betatron tune shift for off-momentum particles traversing in a
circular accelerator is given by summation over all error sources in Eq.(166):
1
DQ 
4p  b ( s)DK ( s)ds
1 dp

4p  b ( s ) K ( s )(
p0
)ds (171)

Note the following facts:

b(s) is larger at K(s)> 0 (focusing quadrupole) bF
b(s) is smaller at K(s)< 0 (defocusing quadrupole) bD
For a circular accelerator designed with alternating gradient focusing, the number
of focusing quadrupoles is about the same as the defocusing quadrupoles.
Therefore, we get the following relation:
 1 dp
DQ  ( )  [ b F ( s)  b D ( s)]  K ( s) ds (172)
4p p0
The chromaticity of an accelerator is defined as:
DQ 1
x 
(dp / p0 ) 4p  b ( s) K ( s)ds (173) 121
Typically the magnitude of bF is several times larger than that of bD at locations of
quadrupole magnets,

[b F (s)  b D (s)]  K (s) ds  0 (174)
Therefore, the natural chromaticity x of a circular accelerator with alternating
gradient focusing is negative!
The natural chromaticity is a property of accelerators which arises solely from
quadrupole magnets. This is an effect caused by systematic errors.
A negative chromaticity in a circular accelerator will induce the transverse head-tail
instability driven by the resistive wall impedance. The chromaticity must be corrected
to a small positive value in order to avoid the head-tail instability. In addition to the
above consideration, a larger chromaticity will also result in a larger betatron tune
shift which in turn will increase the danger of resonance crossing.

How to counteract the gradient error due to the momentum deviation?

We need a device which behaves like a quadrupole magnet and its strength
is linearly proportional to the momentum deviation dp.

122
 Transverse collective instability

•Coupled bunch instability •Head-tail instability

snapshot of coherent oscillation
m=0, x=0 m=0, x0
1

0.5

0
m=1 m=2

-0.5

-1

-1 -0.5 0 0.5 1 [References]: Frontiers of Particle Beams: Intensity Limitation, 123

M. Dienes et al (eds.), Springer-Verlag LNP 400 (1992)
 Sextupole Magnets and Chromaticity Correction
y

S S
By
Bx
x

N
N

Fig. 48 Fig. 49

From Eq.(170) we would like to have a magnetic device which can provide a focusing
force with the following form to compensate the gradient error caused by the
momentum deviation: A quadrupole strength linearly
dp
F SF x proportional to dp/p0 124
p0 (175)
We are looking for a magnetic field with the following form,
 dp
x ( s )  xb ( s ) 
dp D  xb
D( s ) B  x2 p0
p0

The focusing force Fx(s) should have the following form,

By
Fx ( s)   SF x (176)
x
SF 2
By ( s) ~x  f ( y) (177)
2
 By Bx
From   B  0   Suppose there is no longitudinal
x y dependence and ignore the edge effect.

Bx By
  SF x Bx (s) ~ S F xy  g ( x) (178)
y x
  B By
From   B  0  x
 0
x y
g ( x) f ( y) SF 2
SF y    0  f ( y)   y , g  const. (179) 125
x y 2
Finally we find the functional form for the required magnetic field,
 SF 2
B( s)  yˆ ( x  y 2 )  xˆS F xy (180)
2
Recall the 2-D Taylor’s expansion for a function f(x,y) It’s a sextupole magnet !

f f 1 2 f
f ( x, y )  f (a, b)  [( x  a)  ( y  b) ]  ( x  a ) 2

x y 2 x 2

2 f 1 2 f
( x  a)( y  b)  ( y  b ) 2
 ....
xy 2 y 2

n
1 f (a, b) f (a, b) k
  [( x  a)  ( y  b) ]  o( x n 1 ) (181)
k  0 k! x y

We expand the magnetic fields Bx(x,y) and By(x,y) as 2-D Taylor’s series:
Bx (0) Bx (0) 1  2 Bx 2  2 Bx 1  2 Bx 2
Bx ( x, y)  Bx (0,0)  x y x  xy  y   (182)
x y 2 x 2 xy 2 y 2

By (0) By (0) 1  By 2  By

2 2
1  By 2
2

By ( x, y)  By (0,0)  x y x  xy  y   (183)
x y 2 x 2
xy 2 y 2
126
We look for quadratic terms for sextupole magnet only,

1  2 Bx 2  2 Bx 1  2 Bx 2
Bx ( x, y)  x  xy  y (184)
2 x 2
xy 2 y 2

1  By 2  By 1  By 2
2 2 2

By ( x, y )  x  xy  y (185)
2 x 2
xy 2 y 2

Maxwell’s equations and Eq.(180) give us the following relations:

  2 By  2 Bx
  SF
 B  0 x 2 xy
(186)

 2 By  2 Bx
 (187)
xy y 2
  2 By  2 By
  (  B)  0   SF (188)
x 2
y 2

 2 Bx  2 Bx
 2 (189)
x 2
y
127
Consider the radial displacement for an off-momentum particle
dp
x( s)  xb ( s)  D where   (190)
p0

We plug Eq.(190) into sextupole field as shown in Eq.(180). We get the followings,
SF
By  [( xb  D ) 2  y 2 ]
2
S S S
 F xb2  S F Dxb  F D 2 2  F y 2 (191)
nonlinear effect 2 2 2
Bx  S F xb y  S F Dy (192)
x-y coupling
The fractional momentum error  is typically on the order of 10-2. If we only keep the
first order terms in xb and ignore the x-y coupling for a moment, then the field errors
DB are approximated respectively as:
DBy  S F Dxb (193)

DBx  S F Dy (194)

Recall the equations of betatron motion with field errors as given by Eqs.(158 &159).
If we substitute Eqs.(193 &194) to Eqs.(158 & 159) respectively, we will end up with
the following equations when a sextupole magnet is present: 128
  SF D 


xb ( s)   K x ( s)  K x ( s)    x( s )  0 (195)
 B0 r 

 SF D 
y( s)   K y ( s)  K y ( s)    y(s)  0 (196)
 B0 r 
gradient errors

We substitute the gradient errors as highlighted by the dashed lines in Eqs.(195 &
196) into Eq.(171). The betatron tune shifts for the horizontal plane and the vertical
plane are given respectively by
1 dp
DQx 
4p  b x ( s )[  K x ( s )  m x D ( s )]ds  (
p0
) (197)

1 dp
DQy 
4p  b y ( s )[  K y ( s )  m x D ( s )]ds  (
p0
) (198)

1  By
2
SF
, where mx  0  0 is the sextupole strength in the horizontal plane.
By r By r x 2 129
Therefore, the chromaticity after the addition of a sextupole magnet is given by:

~ 1
xx 
4p  b x ( s)[ K x ( s)  mx Dx ( s)]ds (199)

~ 1
xy 
4p  b y ( s)[ K y ( s)  mx Dx ( s)]ds (200)

If the polarity of sextupole is chosen such that SF >0 (or mx >0), we call this as a
focusing sextupole for the horizontal motion.
Add a horizontal focusing sextupole mx  0 : Dx x  0 and Dx y  0

We need to add another sextupole with opposite polarity, say SD <0, to compensate
the vertical chromaticity.
1  By
2
SD
my  0  0 (201)
By r By r x 2

Add a horizontal defocusing sextupole my  0 : Dx x  0 and Dx y  0

130
Taking into account the contributions of two sextupole families, Eqs.(199 & 200)
become:
~ 1
xx 
4p  b x ( s)[ K x ( s)  mx Dx (s)  my Dx ( s)]ds (202)

~ 1
xy 
4p  b y ( s)[ K y (s)  mx Dx ( s)  my Dx ( s)]ds (203)

We need sextupole magnets to correct the natural chromaticity in circular

accelerators. But sextupole magnets also introduce nonlinear effects which should
be carefully mitigated.

General guidelines for chromaticity correction:

1) Sextupole magnets should be placed as close as possible to quadrupoles

where beta functions are large
2) Sextupole magnets should be placed in locations where dispersion
functions are large in order to reduced the strengths of sextupoles.
3) At least two families of sextupoles are needed in order to correct the
horizontal and vertical chromaticity to small positive values in both planes.
 achieving the same level of correction with smaller sextupole strengths
131
Example: TPS Booster Machine Study on 2015/03/20

Tune Control during ramping (correction of Eddy current effects)

Horizontal

Vertical

132
When we increase the beam C2
energy, i.e, the bending field also
increased, there is an induced emf.

dB
0
E
dt
DB
y

E
x dB
s  0 (energy ramping up)
dt

L
B g

According to the Lenz’ Law, the induced magnetic field DB is in the

direction against the change of magnetic flux dB/dt.
133
From Faraday’s Law, there is an emf induced by the time variation of magnetic flux in
the loop C2.
    
 E  d    t  B  dA
The direction of induced E-field is shown as blue arrows in the above picture.

Considering the loop C2 inside the vacuum chamber wall, then

there will be an induced eddy current J flowing in the direction
of blue arrows.
C2

J J
n̂

-x x

h
2a

134
The general expression of magnetic field B(x, t) at a horizontal position x is:

m0 NI 0 (t ) m0ha 2  m0hx 2 
B ( x, t )   B(t )  B(t ) Perturbation technique
g g g
m0ha 2  m0hx 2  )
 B0 (t )  B0 (t )  B0 (t )  o( B
g g
m0 NI 0 (t )
, where B0 (t ) 
g

Effective dipole field: Sextupole field:

m0ha B 0 (t )2 m0hx 2 B 0 (t ) 1 d 2B 2
B0 (t )   2
x
g g 2! dx

Closed orbit deviation Extra focusing sextupole, increase of

during the energy ramp horizontal chromaticity during the ramp
 decrease of vertical chromaticity 
RW head-tail instability is possible !
135
For a resonant ramping scheme,
assuming the resistivity of stainless steel= 90 × 10-8 ohm-m, rep. rate= 3 Hz

dipole field depression due to eddy current

0

-0.2
thickness= 0.7 mm
-0.4
delta_B/B [1e-3]

-0.6

-0.8

-1

-1.2

-1.4
0 20 40 60 80 100 120 140 160 180
time [ms] extraction

136
 Chromaticity during the Energy Ramp, ramp rate= 2 Hz
1.5
Ramp rate= 2 Hz
1
horizontal
0.5

chromaticity
-0.5

-1

-1.5

-2 Cx

-2.5

-3
vertical Cy

-3.5
0 50 100 150 200 250
Time [ms]

Chromaticity during the Energy Ramp, ramp rate= 2.5 Hz

Ramp rate= 2.5 Hz 2

1 horizontal
chromaticity

-1

-2
Cx

-3 Cy

vertical
-4
0 20 40 60 80 100 120 140 160 180 200
137
Time[ms]
Tune shift during ramping

After tune correction

Horizontal

*Time 0 ms = Injection Point.

Vertical

138
Tune shift during ramping

*Time 0 ms = Injection Point.

139
 Root-mean-square Beam Size

This is the effective beam size when there is nonzero dispersion function.

  x
2 2  y2  y 2
x

dp 2  [ yb ( s )]2
 [ xb ( s )  Dx ( s )( )] (204)
p0   yb y
  x b x  [ Dx ( s) p ]2   x0
y   x 0,  x 
1  1 
x’

(b)1/2

Dxp Fig. 50 140