Lecture 5

RF Acceleration in Linacs
Part 2 Stuart Henderson, Jeff Holmes, Yan Zhang USPAS January 2009

Outline
Traveling-wave linear accelerators Longitudinal beam dynamics Material M t i lf from Wangler, W l Chapters Ch t 3 3, 4 4, 6

Guided Electromagnetic Waves in a Cylindrical Waveguide

We can accomplish each of these by transporting EM waves in a waveguide Take a cylindrical geometry. The wave equation in cylindrical coordinates for the z field component is
2 E z 1 E z 1 2 E z 1 2 E z =0 + + r z 2 r r r r 2 2 c 2 t 2

Assume the EM wave propagates in the Z direction. Lets look for a solution that has a finite electric field in that same direction:
E z = E z (r , , z, t ) = E0 (r , ) cos(k z z t ) The azimuthal dependence must be repetitive in :

E z = R(r ) cos(n ) cos(k z z t ) The wave equation yields:

2 R (r ) 1 R (r ) 2 n2 2 + 2 k z 2 R(r ) = 0 + 2 1 r r r c4 r 2 4 3 2 k c

Cylindrical Waveguides
Which results in the following differential equation for R(r) (with x=kcr) 2
d R 1 dR 2 2 + + ( 1 n / x )R = 0 2 dx x dx

The solutions to this equation are Bessel functions of order n, Jn(kcr), which look like this:

Cylindrical Waveguides
The solution is:
E z = J n (kc r ) cos(n ) cos(k z z t )

The boundary conditions require that Ez (r = a) = 0 Which requires that

J n (kc a ) = 0 for all n

Label the n-th th zero of Jm: For m=0, x01 = 2.405

2
2

J m ( xmn ) = 0

2.405 2 2 2 = k + k = + k c z z c2 a

Cutoff Frequency and Dispersion Curve

The cylindrically symmetric waveguide has

k02 = kc2 + k z2
A plot of vs. k is a hyperbola, p Curve called the Dispersion Two cases: > c: kz is a real number and the wave propagates < c: kz is an imaginary number and the wave decays exponentially with distance Only O l EM waves with ith f frequency above cutoff are transported!

2 = c2 + (k z c) 2

Phase Velocity and Group Velocity

The propagating wave solution has
E z = E0 (r , z ) cos( )
vp =

A point of constant propagates with ith a velocity, elocit called the phase velocity,
kz

= k z z t

The electromagnetic Th l t ti wave i in cylindrical li d i l waveguide id h has phase h velocity that is faster than the speed of light:
vp = c 1 /
2 c 2

>c

This wont work to accelerate particles. We need to modify the phase velocity to something smaller than the speed of light to accelerate particles The group velocity is the velocity of energy flow:

PRF = v gU
A di And is given i b by:
vg = d dk

Traveling Wave Structures

Recall that in the cylindrical waveguide, the electromagnetic wave has phase velocity that is faster than the speed of light:
vp = c 1 /
2 c 2

>c

This wont work to accelerate particles. We need to modify the phase velocity to the speed of light (or slower) to accelerate particles in a traveling wave Imagine a situation where the EM wave phase velocity equals the particle velocity v PARTICLE Then the particle rides rides the wave wave
vP

A disk-loaded waveguide can be made to have a phase velocity equal to the speed of light. These structures are often used to accelerate electrons The best and largest example of such an accelerator is the SLAC two-mile long linac

DiskDisk -loaded waveguide structure

KEK

Energy Gain in a DiskDisk-Loaded Waveguide

Define D fi Ea: longitudinal accelerating field amplitude U: stored energy per unit length Pw: traveling wave power dPw/dz: power dissipation per unit length 2 rL = Ea /( dPw / dz ) Shunt impedance per unit length We have Q = U /(dP / dz )
w

Pw = v gU
2 Ea = rL Pw / Qvg

dPw = Pw = 2 0 Pw dz Qvg
We have ha e two t o choices for the accelerating str structure, ct re considered now in turn

Constant Impedance Traveling Wave Structure

Consider a disk-loaded waveguide with uniform cell geometry along the length, then Q, vg, rL, 0 are independent of z: Power decays exponentially along the length of the structure The Electric field amplitude is

Pw ( z ) = e 2 0 z

dEa / dz = 0 Ea

E a ( z ) = E 0 e 0 z
At the end of a waveguide g of length g L Ea ( L) = E0 e 0 Pw ( L) = P0 e 2 0
0 = 0L = L
2Qvg
L

The energy gain is

W = q cos Ea ( z )dz = qE0 L

0

1 e 0

cos

W = q 2rL P0 L

1 e 0

cos

Constant Impedance Structure Parameters

Constant Gradient Traveling Wave Structure

A more common design keeps the gradient constant over the length length, which requires that the attenuation 0 depend on z dPw = 2 0 ( z ) Pw dz Which can be integrated to yield The attenuation factor is The energy gy g gain is
z Pw ( z ) = P0 1 (1 e 2 0 ) L

1 1 e 2 0 0 ( z) = 2 L 1 ( z / L)(1 e 2 0 )
L

W = q cos Ea ( z )dz = qE0 L cos

0

W = q rL P0 L(1 e 2 0 ) cos

To achieve a constant gradient, the SLAC linac structure tapers from a radius of 4.2 4 2 to 4 4.1 1 cm cm, and the iris radii taper from 1.3 to 1.0 cm over 3 meters

Constant Gradient Traveling Wave Structure

The group velocity is

L 1 ( z / L)(1 e 2 ) = vg ( z ) = 2Q 0 ( z ) Q 1 e 2
0 0

The Th filli filling time ti is i

tF =
L 0 L dz Q dz 2Q 2 0 = (1 e ) = 0 0 1 ( z / L )(1 e 2 0 ) v g ( z ) L

For typical parameters parameters, the filling time is ~1 sec, and the beam pulse is 1-2 sec

Constant Gradient Structure Parameters

SLAC Linac
Largest in the world. Reached energies of 50 GeV

Synchronicity condition in multicell RF structures

TM010 Cavities Drift spaces

1 l1

2 l2

3 l3

4 l4

5 l5

Suppose we want a particle to arrive at the center of each gap at =0. Then we would have to space the cavities so that the RF phase advanced by
2 if the coupled cavity array was driven in zero-mode Or by if the coupled cavity array was driven in pi-mode

Synchronicity Condition
Zero-mode:
= t =
2c

ln = n

t=

2c ln = 2 c n

RF gaps (cells) are spaced by , which increases as the particle velocity increases Pi-mode Pi mode:
= t =
2c

t=

2c ln = c n

ln = n / 2
RF gaps (cells) are spaced by /2, which increases as the particle velocity increases

Longitudinal Dynamics
The drift space length between gaps was calculated for a particular particle with a very specific energy. This is the reference particle, or the synchronous particle. What happens to particles slightly faster or slower than the synchronous particle that the linac was designed to accelerate? Linacs are operated to provide longitudinal focusing to properly accelerate particle over a range in energies or arrival time Slower particles arrive at the next gap later than the synchronous particle
They experience a larger accelerating field

Faster particles arrive at the next gap earlier than the synchronous particle
They experience a smaller accelerating field

Equations of Motion I

Consider an array of accelerating cells with drift tubes and accelerating gaps The array is designed at the n-th cell for a particle with synchronous phase, kinetic energy and velocity sn, Wsn, sn. Note that the synchronous phase is not zero! We express the phase phase, energy and velocity for an arbitrary particle in the n-th n th cell as n, Wn, n Assume that the particles receive a longitudinal kick at the geometric center of the cell, and drift freely to the center of the next cell The half half-cell cell length is

ln 1 =

N s ,n 1 2

Where N=1/2 for Pi-mode and 1 for zero-mode The cell length (center of one drift tube to center of next) is therefore

Ln = N ( s ,n 1 + s ,n ) / 2

Equations of Motion II
The RF phase changes as the particle advances from one gap to the next according to
n = n 1 +
2ln 1 + n 1c 0

mode 0 mode

The phase change during the time an arbitrary particle travels from gap n-1 to gap n, relative to the synchronous particle is

( s ) n = n s ,n
Where we have used
1 1 =

1 n 1 1 = 2N s ,n 1 2N s ,n 1 2 s , n 1 s , n 1 n 1
1 1 2 , for << 1 s + s s

Using g We get

W mc 2 s3 s

( s )n = 2N

(W

mc

n1 2 3

Ws,n1 )
2 s,n1 s,n1

Equations of Motion III

Next, derive the difference in kinetic energies of the arbitrary particle and the synchronous particle

(W Ws )n = qE0TLn (cos n cos s ,n )

To figure out the dynamics, we could track particles through gaps on a computer using these difference equations To get a feel for the dynamics on paper, we can convert these difference equations to differential equations by replacing the discrete action of the fields with a continuous field So we replace
d ( s ) ( s ) dn
d (W Ws ) (W Ws ) dn

n=

s N s

giving

d ( s ) W Ws = 2 ds mc 2
3 s 3 s

d (W Ws ) = qE0T (cos cos s ) ds

Equations of Motion IV
Assume acceleration rate is small small, and that E0T, T s and s are constant We arrive at the equations of motion:
w = dw d = B(cos cos s ) = = Aw and ds ds qE 0T W Ws 2 and A B with w = = = mc 2 s3 s3 mc 2

Finally

d 2 = AB(cos cos s ) 2 ds
Aw2 + B (sin cos s ) = H 2

1 Aw2 + V = H 2
Where V is the potential energy term, and H (the Hamiltonian) is total energy

Stable RF Bucket
There is a potential well when - <s< 0 There is acceleration for -/2 <s< /2 The stable region for phase motion is 2 < < -s Th separatrix The t i defines d fi th the area within which the trajectories are stable. The stable area is called the bucket Stable motion means that particles follow a trajectory p j y about the stable phase, with constant amplitude given by H

Hamiltonian and Separatrix Parameters

We can calculate the Hamiltonian to complete the discussion At the potential maximum where, = -s , =0 and w=0

H = B (sin( ( i (s ) (s cos s ))
The points on the separatrix must therefore satisfy

Aw 2 + B (sin cos s ) = B (sin s s cos s ) 2

We can calculate the size of the separatrix. We will do the energy width. Th maximum The i energy width idth corresponds d t to = s
2 Awmax + B(sin s s cos s ) = B(sin s s cos s ) 2

Giving for the energy half-width of the separatrix. The energy acceptance is twice this value:

wmax

Wmax 2qE0T s3 s3 = = (s cos s sin s ) 2 2 mc mc

Phase Width
The maximum phase width is determined from the two solutions for w=0. One solution is 1= -s. The other solution 2 is given by g y
sin 2 2 cos s = s cos s sin s

The total phase width is = s 2 The phase width is zero at s=0 and maximum i at t s=-/2, /2 giving i i =2 2 (see ( Wangler figure 6.4)

Small Amplitude Oscillations

Look at small amplitude oscillations. Letting -s be small,

+ AB sin( ( s )( s ) = 0

This is an equation for simple harmonic motion with an angular frequency given by

2 qE0T sin(s ) = 2mc 2 s3 s

2 l

Note that as the beam becomes relativistic, relativistic the frequency goes to zero From the equation of motion we can calculate the trajectory of a particle: W w 2 ( s ) 2 3 3 2 2 w = = qE T sin( ) / 2 mc = 1 + s s s 0 0 0 2 mc 2 w0 (0 ) 2 This is the equation of an ellipse in w, -s phase space Particles on a particular ellipse circulate indefinitely on that trajectory

Longitudinal Phase Space Motion

We studied the approximation of small acceleration rate, and constant velocity, synchronous phase, etc. In a real linac, the velocity increases, and the phase space motion and separatrix becomes more complicated. The acceptance takes a shape called the golf-club

= const

const

Longitudinal Dynamics: Real data from SNS Drift Tube Linac

Simulated DTL1 Acceptance

Longitudinal Acceptance Scan

Data

FWHM=24 deg

Measurement of SNS SC Linac Acceptance (Y. Zhang)

Measurement Simulation

Adiabatic Phase Damping

Louivilles theorem: The density in phase space of non-interacting particles in a conservative or Hamiltonian system y measured along g the trajectory j y of a p particle is invariant. Or, if you prefer: phase space area is conserved Area of ellipse: Which gives

A Area = 0 W0
0 = const ( s s ) 3 / 4

W0 = const ( s s ) 3 / 4

Since area is conserved an initial distribution with phase width ()i acquired a smaller phase width after acceleration:

(0 ) f

( ) i = (0 ) i ( ) f 3 / 4
3/ 4

The End

That concludes our whirlwind tour of Linear Accelerators Now, on to Rings.