You are on page 1of 70

Chapter 8

Appendices

Chapter Contents
8.1: Spontaneous Emission Calculations............................................254
8.2: Three Dimensional Analytic FEL Model Results ......................258
8.3: Bunching Monitor Calculations ...................................................264
8.4: Strong Focusing in Planar Undulators........................................269
8.5: Calculations for the Emittance Slits.............................................293
8.6: Beam Trajectory Calculation.........................................................296
8.7: IR Optics Calculator.......................................................................298
8.8: Black Body Background Estimation............................................300
8.9: Additional Support Systems.........................................................304

253

8.1: SPONTANEOUS EMISSION CALCULATIONS


A LabVIEW Virtual Instrument (VI) a type of program with a graphical
interface was written based on a MathCAD calculation by Pellegrini. Here we
present the upgraded MathCAD document which contains the angular and
frequency corrections discussed in Section 2.3. The LabVIEW VI, while faster for
calculation purposes, is not suited for printing and viewing on paper. All
calculations below are for a single electron; the final result must be multiplied by
the number of electron in the beam.
We begin by defining some constants and obtaining user inputs:
INPUTS
Nu 5

32.6
1. 10 2

Number of periods (in gain length or


undulator)
Beam energy in mc2

CONSTANTS
c 3. 108

m/sec

8.85. 10 1 2

Rho

q 1.6. 10 1 9

u 0.015

Period in meters

E 0 5.11. 105

au 1

Undulator Parameter

r e 2.81. 10 1 5 m

I 200

Beam current in Amps

r 23. 10 6

Wavelength in meters

r b 300. 10 6

Beam radius

Next, some simple relations are calculated:

254

Volts

Caluclated Terms
u

Lg

4.

N u. u

Ng

a u2

u
. 1
.2
2
2. . c
R

Nz
ku

Resonant energy

2
Resontant frequency

N u. u

Integration steps

N z = 50

Undulator wavenumber

Then, the electron trajectories are formulated:


Electron Trajectory

. sin( k . s )
u

1
s( s )

x( s )

14

R = 1.781 10

dz = 0.002

Number of steps

dz
2.

au

R = 1.059 10

L u = 0.075

Lu

x( s )

3. . . N u = 1.088

4.

N g = 1.088

Length of Undulator (or integration distance)

u
10

dz

L g = 0.069

Number of periods in one gain length

Lg

Lu

1D Gain length

3. .

a u2
2

.( 1

cos( 2. k u. s ) )
2. 2

a u. u
. cos( k . s )
u
2. .

Radiation terms are defined:

255

Radiation terms
1

1. . s. 1
c

a( , , , s )

+
v x( , , , s )
v s( , , , s )
v z( , , , s )

a u2
2
2. 2

cos( )

a u2
8. 2 . k

( x( s ) . sin( ) . cos( ) )

s( s ) . sin( ) . cos( ) . cos( )


s( s ) . sin( ) 2

. sin( 2. k . s ) ...
u
u

x( s ) . ( cos( ) 2

sin( ) 2 . sin( ) 2 )

x( s ) . sin( ) . cos( ) . cos( )

s( s ) . sin( ) . cos( ) . sin( )

x( s ) . sin( ) 2 . sin( ) . cos( )

These are followed by the radiation integrals. Here the integrals are performed
over slices along the undulator. The slices are used to facilitate the angular and
bandwidth corrections:
Radiation Integrals (of the nth slice in z)
n. dz
I 1( , , , n )
(n

1 ) . dz

v x( , , , s ) . exp( a( , , , s ) ) d s

n. dz
I 2( , , , n )
(n

1 ) . dz

v s( , , , s ) . exp( a( , , , s ) ) d s

n. dz
I 3( , , , n )
(n

1 ) . dz

v z( , , , s ) . exp( a( , , , s ) ) d s

The energy spectrum is given next (again for one electron):


U( , , , n )

r e. E 0. 2
4. 2 . c3

. I ( , , , n ) . I ( , , , n ) I ( , , , n ) . I ( , , , n ) ...
1
1
2
2
+ I 3( , , , n ) . I 3( , , , n )

The angular integral and corrections are defined next:

256

Angular Integrals
2.
E ( , , n )

U( , , , n ) d

Phi integral

( n )

1.
R
n

ca( n )

2.

a u2
2

. ( n )
R

Coherence angle

ca( N z ) = 0.011

Finally, the energy is obtained by summing over all the slices, and integrating
over the bandwidth and angle for each slice:
n

1 , 2 .. N z
R

1 .
Lu

( n )
2

( n )
2

ca( n )
E ( , , n ) d d
0

257

8.2: THREE DIMENSIONAL ANALYTIC FEL MODEL RESULTS

The following items are MathCAD documents with the Chin, Kim and Xie (CKX)
3D FEL model. The MathCAD document is taken from an earlier file of Pellegrinis.
The values used are similar to those given in Table 3.12 and the results are discussed
in Section 3.1.3.

First, some constants are defined:


FEL Calculator for Emittance, Current and Energy Spread
(using the K. -J. Kim formalism)
Constants
c 3. 108

speed of light (m/s)

r e 2.81. 10 1 5

electron radius (m)

e 1.6. 10 1 9

electron charge (c)

E 0 5.11. 105

electron rest energy (V)

K 0 93.5

undulator parameter scaling (e/mc)

Next, standard parameters are calculated from the user inputs:

258

Calculated parameters
B

K
K 0. u

Peak undulator field (T)

u
. 1
2. r

IA

e. c
re

JJ

Resonant energy (on-axis)

Alfven current (A)


K2

J0
4

K2
2

K2

J1

2. K2

2. K2

2. . u
2. . K

Bessel function factor

Natural focusing Betafunction

2. . c
r

Resonant frequency

One dimensional values and the D parameter are calculated below:


Formulas

( I , E , n )

1
64. 2

. K. JJ.

FEL parameter
( I 0 , 0 , n0 ) = 0.027

4. K2

D( I , E , n )

. 1
L1 G( I , E , n )

u 2 I
.
. 1

I A n.

1
3

. I . JJ
K2 I A
2
u

. 2.

1D Gain Length

3. ( I , E , n )

n
.
r

Z R( I , E , n )

4. .

p( I , E , n )

3
4

( . ( I , E , n ) ) 2 . . c. 2.

Radiation Rayleigh range

Plasma frequency

Scaled threedimensional parameters are calculated next:

259

u 1( I , E , n )

u
2. . . D( I , E , n )

f 4( I , E , n )

4. . n
r.

Emittance parameter

f 4( I , E , n )

u 2( I , E , n )

( I , E , n )

Focusing parameter

u 1( I , E , n )
ln( u 2( I , E , n ) )
0.238. ( I , E , n ) 0.0139. ( I , E , n ) 2
( f 4( I , E , n ) . u 1( I , E , n ) ) 2
0.149 0.0268. ln( u 1( I , E , n ) )

f 1( I , E , n )

0.759

f 2( I , E , n )

g 1( I , E , n )

44.03

g 2( I , E , n )
g 3( I , E , n )

3.32. ( I , E , n )

D( I , E , n )
f 1( I , E , n ) .

5.45. ( I , E , n ) 2

0.713.

D( I , E , n )

f 2( I , E , n )

68.65.

D( I , E , n )

g 1( I , E , n ) . g 2( I , E , n )

Then, the three dimensional gain length, saturated power and saturation length
are calculated:
3D Gain Length (m)
L G( I , E , n )

u
. exp( g ( I , , ) )
E n
3
4. . D( I , E , n )

Saturation Power (W)


P( I , E , n )

L1 G( I , E , n ) 2
( I , E , n ) . . E 0. I.
L G( I , E , n )

260

Saturation Length
L sat( I , E , n )

L G( I , E , n ) . ln

2. . P( I , E , n )
( I , E , n ) 2 . . E 0. . e

User inputs are entered at the end (for convenience) and calculated parameter
values are returned:
Inputs

Calculated Parameters

u 0.015

Undulator period (m)

= 22.1163

Resonant energy

K 1

Undulator parameter

0 = 0.075

Natural focusing

r 23. 10 6

Resonant wavelength (m)

B = 0.713

Peak field

0.075

Betafunction (m)

= 8.195 10

Resonant frequency

JJ = 0.91

Bessel factor

13

Baseline values
I 0 200

Current (I)

E0 0. 10 4

Energy spread

n0 5. 10 6

Normalize emittance

E1 1. 10 3
E2 5. 10 3

The basic outputs are presented:

261

Outputs
( functions of current, energy spread and emittance)
L G( I 0 , E0 , n0 ) = 0.05
L1 G( I 0 , E0 , n0 ) = 0.052
7

P( I 0 , E0 , n0 ) = 6.58 10
P 1( I 0 , E0 , n0 ) = 0.017

L sat( I 0 , E0 , n0 ) = 1.094
11

p( I 0 , 0 , n0 ) = 2.269 10

Finally, some graphs are generated (many others can also be produced):
I

50 , 60 .. 300

Gain Length vs. Current for Various Energy Spreads


0.15

L G I , E0 , n 0

0.1

L G I , E1 , n 0
L G I , E2 , n 0 0.05

0
50

100

150

200
I

262

250

300

Number of sample points

N 25
i

1 .. N

1 .. N

I min 50

min 2.5. 10 6

I max 300

max 10. 10 6

Ii

I min

( I max

I min ) .

i
N

Li , j

1
1

min

( max

j
min ) .
N

L G( Ii , 0 , j )
Gain Length
0.116

C
u
r
r
e
n
t

L
Emittance

263

1
1

8.3: BUNCHING MONITOR CALCULATIONS

Using the relations derived in Section 2.5, we can numerically evaluate a CTR
bunching foil for use on PBPL. In addition, we can compare the analytic results
with numerical solutions.

First, user inputs are obtained (some of these are redundant):


Coherent Transition Radiation Bunching Monitor
Work by J. Rosenzweig, A. Tremaine and G. Travish
MathCAD by Gil Travish
Modified: 4/10/96
Inputs

25

Beam energy [cm^2]

u 0.015

Undulator period [m]

z 7. 10 4

Bunch length [m]

r 3. 10 4

Bunch radius [m]

au 1

Undulator parameter

N b 6. 109

Number of electron in the bunch

b 1 0.0018

Bunching parameter (first harmonic)

Some constants are defined:

264

Constants
c 3. 108

Speed of light [m/s]

5.11. 105

E0

Electron rest energy [eV]

1.6. 10 1 9

Electron charge [C]

1
137

Fine structure constant

Some variables are defined:


Initial Calcs
r

kr

2.
r

a u2

u
. 1
.2
2

Resonant wavenumber [1/m]

1
2

k r = 3.491 10

Relativistic beta

r = 1.8 10

= 0.999

The exponentialintegral sum is defined:


n

1 .. 20

Ei( x )

(n

exp( x ) .

1 )! .
n

( 1 )n

Exponential Integral function

The angular cutoff used for the approximation is also computed:


a

2
k r. r

Resonant wavelength [m]

Approximate angular cut off for CTR

The full differential spectrum is calculated next:

265

Full equations
DIfferential spectrum (dk, dq)
N f( k , )

2.
2
.
1. N b b 1 .
sin( ) 3
. exp
4
2. . k
( 1 . cos( ) ) 2

( k. r. sin( ) ) 2

. exp

(k

The spectrum is integrated numerically over frequencies (wavenumbers):


Anglur Spectrum: Spectrum numerically integrated over1% frequency band
1.005. k r
N i( )
N f( k , ) d k
.k
0.995 r
7

N i( a ) = 2.976 10

and angles:
2. a
N i( ) d

N ii
0

N ii = 1.194 10

A comparison to the approximate analytic relations is made:


Approximate Relations
Angular Spectrum

N a( )

2.
2
.
1. N b b 1 .
sin( ) 3
. exp
4
. . k . ( 1 . cos( ) ) 2
8
r z
7

N a( a ) = 2.297 10

Error in approximate relation


N i( a )

N a( a )

N i( a )

= 0.228

266

( k r. r. sin( ) ) 2

k r )2

Further small angle approximation

N aa( )

. N b2 . b 12
3
.
. exp
. . k .
1 2
8
r z 2
2

( k r. r. ) 2

N aa( a ) = 2.297 10

Integrate above to yield the exponential-integral function


k r. r 2

N a1

1
2. 4

x = 17.546
. N b2 . b 12
.
.

( exp( x ) . Ei( x ) . ( 1

x)

1)

( 1 )n . ( 1

x)

. k r. z
5

N a1 = 9.225 10
Insert expression for Ei

N a2

. N b2 . b 12
.
4. 2. . k r. z
N a2 =

(n

1 )! .
n

9.225 10

267

Approximate sum by using first and second term


. N b2 . b 12
.
. 2. . k .
4
r z

N a3

1
x

1
x2

.( 1

x)

Incorrect

N a3 = 1.129 10
This simplifies to

N a4

. N b2 . b 12
.
. 2. . k .
4
r z

1 2
x

N a4 = 1.129 10

268

8.4: STRONG FOCUSING IN PLANAR UNDULATORS


The need to maintain a well focused beam is critical to successful operation
future short wavelength, highgain FELs. This Appendix examines alternating
gradient (AG) sextupole focusing in planar undulators and compares it to
quadrupole undulator focusing. The equations of motion for a single electron in
an undulator field with a strong sextupole component are examined. Analytic
and smooth approximation solutions are provided for AG sextupole focusing. It
is shown that the mean electron longitudinal velocity can be kept constant through
each focusing and defocusing section, but that the velocity differs between these
sections. The effects of this stepwise velocity modulation, as well as the beam
size variation, are explored computationally and compared to theory. Examples
using the proposed SLAC 4 nm FEL, the UCLA 10.6 m FEL as well as a Paladin
based device are also given.

8.4.1: Focusing Schemes


Quadrupole focusing has been used on a number of undulators. Typically,
external quadrupole magnets in a FODO lattice are superimposed over the
undulator field (see Figure 8.1). Use of external magnets requires that no permeable
materials be used in the undulator: hybrid undulators are not compatible with

269

external magnets. Alternatively, the quadrupoles can be interspersed with


undulator sections. Regardless of the arrangement, quadrupole focusing differs
from natural focusing in that it causes electron velocity modulation during betatron
oscillations. The effect is to modulate the phase of the electrons relative to the
undulator and optical beam. These oscillations can degrade FEL performance by
effectively debunching or detrapping the electrons (see Section 8.4.2). The reduction
in gain may be undesirable in single pass amplifiers.

Figure 8.1: A cartoon of an undulator magnet with external quadrupole focusing.

Canting the undulator poles has been used to achieve quadrupole like
focusing without external magnets (see Figure 8.2). By introducing a slight tilt to
each undulator pole, a focusing field is introduced near the axis; however, this

270

method can suffer from the same problems as external quadrupole focusing. In
addition, the canting can require more complicated undulator mechanics.

Figure 8.2: An undulator magnet with canted poles. The canting angle as well as
the perspective are exaggerated for clarity.

Solenoidal confinement has been considered for high current (~ kA) low
energy (~ MeV) beams. Unfortunately, offaxis electrons drift transversely.
Rotational stabilization has been suggested as a way of countering this problem.
The requirement that the solenoid field be higher than the undulator field would
make this technique impractical for many systems. Also, the need for an external
magnet extending the undulator length is undesirable (costly). Thus, solenoidal
confinement is not desirable or effective for highenergy beams or in undulators
with high peak fields.
Ion focusing is a promising concept for undulator beam transport (see
Figure 8.3). The idea of introducing a plasma into the beamline has been criticized
in the literature; however, the potential benefits are great. An ion channel can
offer strong focusing without phase modulation. Unfortunately, beam erosion,

271

ion column collapse and ionhose instabilities (as well as others) need to be
avoided.
Undulator

Plasma

Beam

Ion
Channel

Beam

Expelled
Electrons
Figure 8.3: Ion focusing for an FEL is depicted in this diagram.

An elegant focusing scheme based on the converse to optical guiding has


been theoretically derived. Given a sufficiently strong optical field, the electron
beam can be guided. The focusing is similar to natural focusing in its effect. The
mutual focusing effect would be most pronounced in a high power system. Thus,
an inverse FEL would be a candidate to observe this phenomenon.
The following Sections of this Appendix review Scharlemanns original
scheme, and propose an extension to include alternating gradient or strong
focusing. The theory which is developed is then applied to examples.

272

Figure 8.4: A sketch of the lower half of an undulator employing Scharlemanns


shaped poles.

Sextupole focusing satisfies the phase preservation requirement of an FEL;


however, if it is constant gradient, it is weak focusing. The idea of shaping the
poles of a planar undulator to provide focusing was examined by Scharlemann.
Proper shaping of the pole faces can symmetrize the natural planar undulator
focusing (see Figure 8.4). This concept has proven itself in application. Sextupole
focusing avoids the problems and complexities associated with other focusing
schemes, and can produce focusing up to the strength of the natural vertical
focusing. This concept has proven itself in application. The major drawback of
this scheme is the weakness of focusing, as is discussed in Section 2.6.2.

8.4.2: Strong Sextupole Focusing


A calculation of the equations of motion for alternatinggradient focusing
is presented in this Appendix. The inspiration for this treatment comes from
Scharlemanns work on weak sextupole focusing and the calculation by Dattoli
and Renieri using a smallness parameter to describe the effective horizontal
defocusing sextupole strength.

273

Implementing strong focusing with sextupoles, which requires alternating


between focusing and defocusing sections, can overcome the natural focusing
strength limitation. A set of poles which focus in, say, x and defocus in y is
followed by a set which defocuses in x and focuses in y. This is analogous to
strong focusing with quadrupoles. Since sextupole fields are quadratic, the offaxis
orbit taken by the design electron moves through a region with a linear gradient,
giving rise to a quadrupolelike focusing (this is an example of the feed down
effect in beam dynamics). One constraint is that the design orbit (and velocity) of
the electrons must match closely in the two types of undulator sections. It has
already been verified that this is the case in a weak focus (Scharlemann) and a
weak defocus (Dattoli and Renieri). It is still necessary to show that the results
are in fact valid for strong focusing.
The geometry used in the calculations is as follows: The beam propagates
along the positive zaxis with the undulations occurring in the xz plane and the
undulator field along with natural focusing occurring in the yz plane.
We begin, as Scharlemann did, with a normalized undulator magnetic
field which satisfies the Maxwell equations, and has a symmetric dependence on
the transverse distance from the undulator axis:
bk

b = x 0 x sinh( kx x)sinh( ky y )cos( ku z)


ky

+ yb0 cosh( kx x)cosh( ky y )cos( ku z)

(8.1)

b0 ku
cosh( kx x)sinh( ky y )sin( ku z)
ky
where kx and ky are the focusing wavenumber in x and y, respectively,
e
2
b0 =
B =
au is the normalized undulator field and the remainder of the
2 u
u
mc
notation is as given in Table 1.5. The Maxwell equations additionally require that

274

2
2
2
ku = k x + k y .

(8.2)

When the focusing strength in one plane exceeds the natural focusing strength,
the strong focusing regime is entered, and k (x,y)>k u, k(y,x) becomes imaginary.
A derivation of the field to arbitrary order is also possible, but analysis
shows that the correction terms beyond second order are negligible (for known
FEL parameters). For small k xx and k yy the field may therefore be approximated
by
2
bx = b0 kx xy cos( ku z)
2 2
2
.
(8.3)

kx x 2 k y y
by = b0 1 +
+
cos( ku z)
2
2

The electron equations of motion are straightforward to derive:


c

x = zby ybz ,

(8.4)
y = ( xbz zbx ),

z = ybx xby .

Here the dot is used to indicate a derivative with respect to time. A natural scale

length of the problem is the undulator period; by separating the fast oscillations
(those occurring at the undulator frequency) from the secular behavior, it is
possible to simplify the equations. Following Scharlemann we define r(x,y,z)=r0+r1
where r0 is constant over the undulator period (the slow betatron oscillation) and
r1 varies within a period (the fast undulator oscillation). Then, the equations of
motion can be written, to leading order, as

275


cz0
by ,

cz

(8.5)
x1 = 0 by ,

cx
cz

y0 = 1 bz 0 bx ,

where the brackets ( ) indicate averaging over an undulator period. For planar

x0 =

undulators, the term 1 can be neglected at this order in the analysis. It is easiest
to integrate the expression for x1 while inserting the expression for the magnetic
field to yield

x1 =

2 2
c
k 2 x 2 k y y0
b0 1 + x 0 +
sin( ku z) .
ku
2
2

(8.6)

Averaging and simplification gives the desired solution for the equations
of motion for the three cases of weak focusing, strong focusing in x, and strong
focusing in y:
2

x0 + c 2 kx x0 0

for kx,y<ku,
2 2

y0 + c ky y0 0

x0 + c 2 kx x0 0

for ky >ku,
2 2

y0 c ky y0 0

2 2

x0 c kx x0 0

for kx >ku,
2 2

y0 + c ky y0 0

(8.7)
(8.8)
(8.9)

where the notation


k ( x , y ) =

b0
k( x , y )
2ku

(8.10)

has been introduced. The constraints on the transverse focusing betatron


wavenumbers for these three cases follow from Equation 2.174:
2
2
k x + k y =

e
b0 ,
2mc 2 2

(8.11)

276

2
2
k x k y =

2
2
k x + k y =

e
b0 ,
2mc 2 2

(8.12)

e
b0 .
2mc 2 2

(8.13)

Note that for the alternating gradient cases the focusing strengths relative
to the natural case are |kx |/k u and |ky |/k u for the x and y directions, respectively.
The above sets of equations can each be integrated by using the relation between
the derivatives with respect to time and distance (z). Scharlemann has shown
that the additional term coming from the longitudinal acceleration (velocity
modulation) does not contribute to the average focusing, and the relation
d/dt=vz(d/dz) is a good approximation. Now it remains to evaluate the average
transverse velocity,
2
=

1
2
2 2
x1 + x0 + y0 .
c2

(8.14)

The x1 term is averaged by noting that <sin2 (kuz)>=1/2. Then, eliminating


higher order terms leaves
2
x1 =

2
c 2b0
2 2
2 2
1 + k x x0 + k y y 0 .
2
2 2 ku

(8.15)

The above equation holds for all three cases considered. So, the average transverse
velocities for each case become:

(8.16)

(8.17)

(8.18)

2
=

2
b0
2 2
2 2
1 + kx x + ky y for kx,y<ku,
2
2 2 ku

2
=

2
b0
2 2
2 2
1 + kx x ky y for ky >ku,
2
2 2 ku

2
=

2
b0
2 2
2 2
1 kx x + ky y for kx >ku,
2
2 2 ku

277

where xb and yb are the amplitudes of the transverse betatron oscillation, i.e.,
x0=xb sin[kbxz+x] for a focusing section, or x0=xbsinh[k bxz+x] for a defocusing section.
It is now possible to see that each < 2> is constant. That is, an electrons velocity
averaged over an undulator period is constant through a betatron oscillation
within a particular case of focusing. This indicates that the (longitudinal) phase
of the electrons within a ponderomotive bucket is not modulated. Hence, one
would expect that sextupole focusing is not deleterious to electron bunching and
FEL gain. In fact, as was discussed in the Introduction, the gain is expected to be
higher since the beam density remains greater under focus.
The above statements hold true for weak (constant gradient) focusing. For
strong focusing they apply only within a particular focusing section. In the
transition from, for example, horizontal focusing (Equation 8.17) to defocusing
(Equation 8.18), the velocity is not, in general, constant. Since it can be shown
that the betatron amplitudes xb and yb are constant for each electron and remain
2
the same across a lens boundary, one can see that is, in general, different in

defocusing and focusing sections. It is not feasible to make the velocities equal in
the two types of sections for all electrons: any realistic beam will have a spread in
the betatron amplitudes. However, this shortcoming does not in itself necessarily
imply that strong sextupole focusing is problematic.
Indeed, phasespace mixing and possible detrapping are possible at the
boundaries between focusing and defocusing sections. This situation inspires an
analogy to tapered undulators. Theory indicates that the tapering should be
performed gradually, but practical considerations can necessitate stepped tapering.

278

Likewise, it is expected that if the focusing is not too strong the FEL gain will not
be adversely affected.
The maximum phase change, , experienced across a focusing/defocusing
section boundary is given by

kr
2
dz .
2

(8.19)

The integral is trivial since the velocity is constant (Equations 8.168.18), and we
are ignoring the effects of actual energy change induced by this phase change.
Ignoring motion in y and integrating over one focus (or defocus) section of length
Lq yields
2

k b2 k
= r 03 x n Lq ,
2 ku

(8.20)

where

2
x

(8.21)

is the average focusing betafunction and has units of inverse length. The Examples
section uses this result to evaluate the effectiveness of sextupole AG focusing.
In the smooth approximation, if the average betafunction is the order of
the gain length (~Lg ), then FEL operation (power output) is maximized.
Perturbations caused by the focusing on a scale longer than the gain length
should not be significant. The next section examines an approximate solution to
AG sextupole focusing and addresses the issue of focusing strength, to allow for
a quantitative analysis of this scheme.

279

8.4.3: Matrix Description of AG Focusing


It is useful and straightforward to solve for the focusing effects of the AG
sextupoles by using the transfer matrix description of the linear equations of
motion. It is possible to use this method by substituting the equivalent quadrupole
strength of the sextupole channel. This analysis will elucidate the effects of focusing
on the beam size variation.
The transfer matrices for half of the focus (F) and defocus (D) section in a
strong focusing lattice are defined by this prescription as follows
1

cos
sin

2
k
2
F=


k sin
cos

2
2

(8.22)
1

sinh
cosh 2
k
2
D=

k sinh
cosh

2
2

respectively, where =kb Lq and Lq is the effective quadrupole length of each section.
Then, the total transfer matrix for one cell (one period of the focusing channel) is
given by
M1 =

F F
D
2 2

D D

M2 = F
,

2 2

where a cell is started from the middle of a focus (defocus) section. Then,
1

cos cosh
(sin cosh + sinh )
,

k
M1 =

cos cosh

k (sin cosh sinh )


1

cos cosh
(sin + sinh cos )
.

k
M2 =

cos cosh

k (sin sinh cos )

280

(8.23)

(8.24)

The parameter , the phase advance per cell, is then defined by


2 cos = Tr( M ) = 2 cos cosh .. For small angles (</4) we may expand this
transcendental expression to yield, in the smooth approximation,

2
.
3

(8.25)

The average betafunction can be defined to be a geometric average of the


minimum and maximum betafunctions: max(min) sin = [ M1( 2 ) ]1, 2 . This produces
the relation

2 3
.
2
k Lq

(8.26)

We note at this point the strong dependence on kb, and that this betafunction
is

3 times larger than that for a thin lens FODO channel. Although this implies

a larger beam (and so less dense), the variation of the beam size is smaller than in
the thin lens case. This is advantageous in an FEL since large fluctuations in the
beam size may be deleterious to gain and optical beam quality. The above relation
can be used to show that in a given FEL the phase change of Equation 8.19 is
approximately constant for small . The resulting relation can be written as

2 3 n k r
.

(8.27)

According to the FEL resonance condition, kr scales as 2 implying that scales


with . This indicates that high energy, shortwavelength devices may be unable
to use AG sextupoles. Note that the scaling of this result is expected from Equation
2.128. In fact, the limit on the emittance differs only by a factor of

3 . Thus,

Equation 8.27 imposes a more stringent limit on the emittance than the 1D
nofocusing limit.

281

In order to make the strong focusing based on sextupole fields attractive,


it must be clearly superior to weak focusing. This requires that

< weak ,

(8.28)

where the quantity weak is defined as the betafunction obtained for a round beam
using Scharlemanns pole shaping scheme. Let the ratio between the strong and
weak betatron wavenumber be Rkstrong / kweak . Then kstrong = Rknatural / 2 and
Equation 8.28 becomes
2 3
R>

kweak Lq

1/ 2

4 3
=

b0 Lq

1/ 2

(8.29)

This requirement is sometimes an overestimate of the ratio R because of


the previous requirement that the smooth approximation be valid. To show what
happens when this requirement is lifted, consider the case of 90 phase advance
per cell (=/2). Then,
1
min =
k

max =

1 /2
e ,
k

(8.30)
1
2.2
/2
=

.
e
k
k
Notice that the betafunction in this case is independent of the quadrupole

length (assuming Lq>lu). In fact, the ratio R is independent of all parameters and it
is only required that
R>2.2.

(8.31)

The variation of the undulator magnetic field by a strong sextupole


component may be large enough to degrade the FEL synchronism condition, and
this effect must be examined. The fractional variation of the magnetic field over
the beam size can be expressed as

282

(k ) (Rku x ) ,
B
= x x =
B beam
2
4
2

(8.32)

where x is the transverse beam size. Similarly, the variation of the magnetic field
over the undulation orbit is
2

1 Rb
= 0 .
2 ku

B
B

(8.33)

undulator

Saturation Length [m]

80
76
72
68
64
60
0

20
40
60
80
Phase Advance per Cell []

100

Figure 8.5: Results of numerical simulations show the length of undulator required
to reach saturation as a function of the strong sextupole focusing phase advance
per cell. Large phase advances imply poor FEL performance.

As the examples in the last section will show, requiring that these variations
be small compared to unity is not unreasonable. However, a large phase advance

283

per cell can introduce problems. While a phase advance per cell of 90 degrees
minimizes the average beam envelope, it creates large fluctuations in the beam
size. Numerical simulations confirm that when the phase advance is large and
hence the beam is modulated a great deal, the FEL action will be degraded (see
Figure 8.5). This statement also holds for quadrupole focusing. Thus, in practice,
the phase advance per cell must be smaller than 90.

8.4.4: Implementing AG Sextupole Focusing

Figure 8.6: Sextupole AG focusing for planar undulators using pole shaping. A
set of poles which focus (F) is followed by a set of poles which defocus (D) in
order to form a FD lattice. This is repeated (FDFDFD) the length of the undulator.

Sextupole focusing can be implemented by machining a parabolic curve


into the permeable metal pole pieces of a hybrid undulator (see Figure 8.6). It
might also be possible to achieve an effective sextupole component by simpler
methods. One scheme recently discussed is the use of side arrays of permanent
magnets to shape the undulator field. Another idea under consideration is the
use of planar permanent magnets (see Figure 8.7).

284

Beam pipe

Figure 8.7: Permanent magnets configured in planar arrays to produce a sextupole


field onaxis.

8.4.5: Numerical Examples


Three examples of highgain FEL amplifiers are now discussed to show
when AG sextupole focusing in undulators is potentially useful. The proposed
SLACbased xray FEL poses a number of challenges, including the need to
propagate the electron beam along ~50 meters of undulator. Table 8.1 presents
the nominal beam and undulator parameters. The natural roundbeam focusing
betafunction is ~80 m, whereas the design gain length requires a ~10 m betafunction.
So, weak focusing is insufficient to maintain the desired beam size and attain the
design gain length.

285

Table 8.1: The nominal beam and undulator parameters for the proposed SLAC
xray FEL.

Parameter

Value

Electron Beam Energy

7 GeV

Beam Emittance (normalized, rms) 3 mm-mrad


Peak Current

2500 A

Pulse Length

160 fs

Undulator magnetic field

0.8 T

Undulator period

8.3 cm

Radiation wavelength

4 nm
1.7 x 10-3

FEL parameter
AG Phase variation

~2

We consider implementing strong focusing in this undulator. An average


betafunction of 7.8 meters can be obtained with a focusing section length of 6 m
and R=21.4 when a 90 degree phase advance per cell lattice is used. Note that this
section length is about equal to a gain length, and thus the deleterious effects of
the sudden change in zvelocity of the electrons at the transitions might be
mitigated. It should be noted that in this example, the fractional variation of the
magnetic field over the beam cross section is small ( ~ 2 10 4 ), and the fractional
variation of the undulator field over the undulating design orbit is even smaller
( ~ 4 10 5 ). However, the large beam size variation would still be harmful. Thus,
we examine focusing channels with smaller phase advances per cell. For ~10,
an average beta of ~5 meters can be achieved with 0.5 m sections and R=70. The
phase variation of Equation 8.27 is of the order of 2; strong sextupole focusing is
not ideal for these parameters. It is useful to compare the results of numerical

286

simulations with semianalytic theories. Figure 8.8 shows the results of such a
comparison. As expected, the sextupole focusing shows poorer performance than
the smooth approximation analytic theory. Both the theory and simulations attempt
to account for 3D effects, but to simplify comparison the energy spread of the
beam was taken to be zero.

Power Gain Length [m]

4.5
Numerical Gain Length
Theoretical Gain Length

4.0

3.5

3.0
0.0

2.0

4.0

6.0

avg [m]

8.0

10.
0

Figure 8.8: Sextupole focusing in the SLAC based xray FEL. Analytic results
using a smooth approximation are plotted for comparison.

We now examine an example with a much lower beam energy: the PBPL
FEL. While the initial design calls for a single undulator section 60 cm long,
future plans include adding a second section for a total length ~120-160 cm. The
need for focusing might then become significant.

287

The shortperiod modified hybrid undulator has flat poles (see Section
6.1.1). The natural vertical focusing has an equilibrium betafunction of 10.5 cm,
which if converted to the equivalent weak focusing round beam case would yield
a function of 14.8 cm. While the phase variation (Equation 8.27) is small (~0.3),
strong focusing would yield only a modest improvement in the average
betafunction because of the short undulator period. The ~9 cm betafunction
attainable for reasonable phase advance per cell would only yield a modest increase
in FEL performance.

Power Gain Length [m]

1.5

1.0

0.5
Numerical Gain Length
Theoretical Gain Length

0.0
0.0

1.0

2.0

3.0

avg [m]

4.0

Figure 8.9: Sextupole focusing performance calculated numerically for the Paladin
undulator using a SLAC beam is compared to the results from analytic theory
with a smooth approximation. The natural (weak) focusing case would be off the
vertical scale on this plot.

288

5.0

The third example is based on the use of the Paladin undulator at SLAC.
This example uses a similar beam but with lower energy than the first example
(see Table 8.2). The reduced beam energy decreases the phase variation across
the focusing/defocusing boundaries enabling the AG sextupole focusing to
approach the smooth approximation performance (see Figure 8.9).

Table 8.2: One set of parameters for use of the Paladin undulator at SLAC.

Parameter

Value

Electron Beam Energy

1 GeV

Beam Emittance (normalized, rms) 3 mm-mrad


Peak Current

2000 A

Pulse Length

~200 fs

Undulator magnetic field

0.38 T

Undulator period

8.0 cm

Radiation wavelength

52 nm

FEL parameter

5 x 10-3

AG Phase variation

~0.6

8.4.6: Comparison of Sextupole to Quadrupole Focusing


The code TDA3D was modified to allow for sextupole focusing. This code
solves the averaged FEL equations in 3D and takes into account known phenomena
for the regime studied here. The sextupole focusing is accounted for in the

289

simulation by modifying the vector potential of the undulator (au). Quadrupole


focusing is simulated by adding a term to the particle equations of motion.
The example parameter set discussed here is the SLAC based xray FEL.
The parameters are given in Table 8.2. It serves as a good test case due to the long
length of the undulator and low beam emittance. Notice that applying Equation
8.28 yields a betafunction of 5 meters for peak FEL performance.
In order to compare reliably sextupole and quadrupole focusing, identical
lattices were calculated (same period, betafunction and phase advance per cell).
Monoenergetic beams were used (no energy spread) in a focus/defocus (FD)
lattice (no drifts). A study of the effect of the phase advance per cell was first
done. Typically, a phase advance per cell of 90 degrees is used to minimize the
average beam envelope. However, this creates large fluctuations in the beam
size. As expected, simulations confirm that when the phase advance is large and
hence the beam is modulated a great deal, then the FEL action is degraded. To
avoid this added effect, subsequent comparisons were performed with a phase
advance per cell of about ten degrees.
Figure 8.10 shows the results of a series of simulations. Three sets of data
points are plotted: quadrupole lattice, sextupole lattice and 3D semianalytic
calculations. The quadrupole set is clearly the best. As expected, there is an
optimal focusing strength. Peak quadrupole performance occurs close to, but not
precisely at the theoretically predicted 5 meter betafunction. A figure of merit for
the effect of focusing on an FEL is given by the variation of the phase over a
betatron period. This is related to the extent of detrapping of electrons from the

290

ponderomotive well. For quadrupole cases, this effect is small. For the sextupoles
used in this example, detrapping becomes significant for a betafunction 5 meters.

4.4
Gain Length (Theory)

4.2

Power Gain Length [m]

Sextupole

4.0

Quadrupole

3.8
3.6
3.4
3.2
3.0
2.8
0

10

Focusing Beta [m]


Figure 8.10: A comparison of quadrupole and sextupole focusing in an FEL.
Analytic results are also plotted for comparison. The emittancelimited optimal
focusing is indicated by the vertical line (at 5 meters).

8.4.7: Strong Focusing Conclusions


Alternatinggradient sextupole undulator focusing has been examined for
use in free electron lasers. Sextupole fields may be an attractive option for strong
focusing provided that variation of the longitudinal velocity and transverse

291

positions of the beam particles are minimized. The work presented here has
arrived at three limits which an alternating gradient focusing scheme for an FEL
must satisfy:
1) The velocity (or, equivalently, phase) modulation between
focusing and defocusing sections must be small.
2) The beam size variation (or phase advance per cell) must be
small.
3) The fractional variation of the magnetic field across the beam
should be small.
Also, while not a fundamental limit, AG focusing should be stronger than weak
(or natural) focusing to be practical.

292

8.5: CALCULATIONS FOR THE EMITTANCE SLITS


The following are direct entries and results from a MathCAD worksheet. The
first section is composed of inputs; parameters chosen for the PBPL system:
Calculations needed for the design of the slit based emittance measurement system (MKS units).
Inputs
I 200

Beam current [Amps]

E 16

Beam energy [MeV]

n 5 . 10

Normalized emittance [m-rad]


(desired to measure)

3
1.5. 10

Beam initial size [m]

6
d 50. 10

The slit width [m]

750 . 10

L d 0.70

Slit separation, should be smaller than initial beam size,


much larger than slit width.
Actual drift distance [m]

The next section are numerical constants used by the worksheet:


Constants
4
1.7. 10

I0

Alfven current

Iron density

5 . 10 3

Depth of stopping iron.

Lr

1.4. 10 2

Radiation length of Fe.

Some useful parameters are calculated next:

293

Calculated parameters
1.5. . 100

dedx

E
.511

Lorentz factor

= 1.597 10 7

Angular divergence of beam at focus

= 31.311

Physical emittance

Minimum ionizing stopping power for iron

= 1.065 10 4

Finally, the figures of merit are calculated:


Relations and Checks
The ratio of the space charge to the emittance terms in the rms
envelope equation before the beam goes through the slits:
R0

2
2. I .

R 0 = 67.632

2
I 0 . . n

>>1 => SC dominated

This is the ratio of the space charge to the emittance terms in the rms
envelope equation for the beamlets after the beam goes through the slits.
R

R0 d 2
.
3

must <<1, or...


R = 0.043

R<1 = 1

Maximum tolerable slit width, shoud be many times actual slit width.

dm

n.

3 . . . I 0
2
I

d m = 3.8 10 4

d m> d = 1

Fraction of energy lost in stopper:

l
dedx.
E

= 0.422

294

> 0.2 = 1

RMS multiple scatter


21 .
E

E
.
. dedx
Lr

1
1

= 1.032

> 0.2 = 1

Ratio of beam rms angle compared to slit acceptance:


l .
d

a = 0.011

a< 1 = 1

Optimum drift length to phoshor.


L do

d. w
2 . 3 0.25.

L do = 0.691

Ratio of of image at phosphor to the slit width, greater than 1 for resolution.

L d. .
d

12 = 5.163

Overlap of beamlets at phosphor, should be much less than one.


L d.

. 2 = 0.199

Ratio of the maximum contribution to the rms


angular width from the space charge, to the emittance part.
Rb

2. I
2
I 0.

Ld d
.
n w

R b = 0.224

295

8.6: BEAM TRAJECTORY CALCULATION


A MathCAD worksheet was used to find the optimal betafunction for
maximizing the average beam current through the undulator:
Beam Trajectory Calculator:
calculates the average current density traversing an undulator

Inputs
I 200

Peak Current [Amps]

n 5. 10 6

Normalized Emittance [m-rad]

Beam Energy [mc2]

33

a u 1.05

Undulator Parameter [-]

u 0.015

Undulator Period [m]

L u 0.6

Undulator length [m]

0.01 , 0.02 .. 2

Beta-star (initial beta)

Alpha (initial divergence)

Calculations
. u
n
2. . a u
( s )

J( s )

Natural Betafunction [1/m]


s 2
0

0. 1
I
n
.

Betafunction along undulator

Peak current density [Amps/m^2]

n. ( s )

296

Lu
1 .
J( s ) d s
Lu 0

J avg( 0 , L u )
J avg( 0 )

I.

L u. n.

02

. ln L
u

L u2 .

ln( 0 ) .

8 10

J avg 0 , 0.3
J avg 0 , 0.6 4 1 09
J avg 0 , 1.2

0
0.5

1.5

f( 0 , L u )

I.

L u. n.

.
n

.
2

3
2

...
2

1 ln L u
+ .
2

Lu

Lu

02

Lu

L u2

root( f( x , 0.3 ) , x ) = 0.09037

r( y )

root( f( x , 0.6 ) , x ) = 0.181

root( f( x , y ) , x )

1 . ln( 0 )
2
0

0.3 .. 2

0.4

root( f( x , 1.2 ) , x ) = 0.361


r( y )0.2

r( L u )
Lu

= 0.301
0
0

0.5

1
y

297

1.5

8.7: IR OPTICS CALCULATOR


Inputs are taken; the optical beam radius is obtained from the TDA3D simulation.
Optics and Detector Calculator for UCLA IR FEL
Inputs
TDA gives HWHM
w0 =2*sigma

w 0 2.355. 380. 10 6

Optical beam radius [m]

L1 1

Distance of optic from exit of undulator [m]

10.6. 10 6

Radiation wavelength [m]

d 0.3. 10 3

Detector diameter [m]

Calcs
. w 02

Zr

Rayleigh range

Z r = 0.237

.w

Optical beam divergence [rad]


0

b = 3.77 10
c

2. w 0. 2. b
c = 1.35 10

c< = 0
w lens

L1
Zr

Coherence length
5

Check that we are in coherent (diffraction) limit


.w

Size of beam at lens

w lens = 3.77 10

298

f1
M

d
.w
2

First lens focal length

L1

f2

Required magnification
ignoring diffraction

M. f 1

Second lens focal length

299

f 2 = 0.168

8.8: BLACK BODY BACKGROUND ESTIMATION


User inputs are taken, some of which are redundant.
Winston Cone Calculator

Inputs
r 20. 10 6

FEL radiation wavelength [m]

r d 1.75. 10 3

detector radius [m]

r c 9.5. 10 3

large radius of cone [m]

r w 0.5. 2.54. 10 2

radius of detector window [m]

l c 2.6. 2.54. 10 2

length of cone [m]

l dw 3. 2.54. 10 2

distance from detector to window [m]

d 5. 10 9

response time of the detector [s]


room temperature [K]

T 300
u 30. 10 6

cutoff wavelength of detector [m]

l 2. 10 6

minimum sensative wavelength of detector [m]

Constants
h 1.05459. 10 2 7
c 2.997. 108

h-bar in erg-sec

k 1.38066. 10 1 6

Boltzman's constant in erg/K

speed of light in m/sec

Calculated Quantities
r

2. . c
r

FEL radiation frequency

The angular distributions are calculated.

300

Field of view quantities


Cone opening angle
c( l cone )

atan

2. r w
l dw

c( l c ) = 1.19

l cone

180
c( l c ) .
= 68.199

Field of View solid angle


c( l c )
2. .

FOV

sin( ) d

FOV = 3.95

0
Distance in Space for integration cone
R
c. d
R = 1.498
Integration Volumedefined by cone-window solid angle and detector response time
R
r2 d r. FOV

Volume

Volume = 4.43

0
Black body radiation
Energy per unit volume [ergs/m^3]
l
U( T )

2 . c3
u

3
h.
exp .
k T

d
1
U( T ) = 54.595

Energy [ergs] in all of cone (neglecting isoltropic nature of blackbody


G( T )

U( T ) . Volume

G( T ) = 241.858

Energy [Joules]
E( T )

G( T ) . 10 7

E( T ) = 2.419 10

Equivelent Numer of Photons at FEL wavelength


N( T )

G( T )
h. r

15

N( T ) = 2.436 10

The volume integral for the Winston cone is calculated:

301

New fangled volume integral- this assumes that of all the black body photons radiated in the
cylinder defined by the FOV, only the ones emitted in a solid angle subtended by the cone
aperture get collected.
R
c( l cone )
r cone
atan
r
sin( c ) d c
0
NFVI( l cone , r cone )
2. .
r2 .
sin( ) .
d dr
2
0
0
NFVI( l c , r c ) = 1.32 10

Energy of collected photons [ergs]


U( T ) . NFVI( l c , r c )

H( T )

H( T ) = 0.007

Equivelent number of photons at FEL radiation wavelength


N 2( T )

H( T )
h. r

10

N 2( T ) = 7.259 10

SHOT NOISE FLUCTUATIONS OF DC BACKGROUND:

N 2( T ) = 2.694 10

And, the volume integral for the aperture (cold stop) is calculated:
Comparison with COLD STOP
You only wish your volume integral was this cool-the above integral modified for a cold stop as
opposed to a Winston cone. As black body radiators approach the FOV cone edge, the angle
subtended by the detector goes to zero.
R
atan
Cool

2.

2. .

rc
r
sin( ) . 1

cos

atan

0
0

Cool = 1.716 10

302

rc
r

d dr

Energy of photons collected assuming a cold stop instead of a cone


J( T )

U( T ) . Cool

J( T ) = 9.366 10

Equivelent number of photons


N 3( T )

J( T )
h. r

N 3( T ) = 9.433 10

SHOT NOISE
3

N 3( T ) = 9.712 10

303

8.9: ADDITIONAL SUPPORT SYSTEMS

8.9.1: Radiation Shielding Bunker


The PBPL bunker serves to protect the human occupants of the laboratory
as well as those in adjacent rooms. The primary design criterion used was to
allow only nearbackground levels of radiation outside the bunker walls while
the accelerator is producing the highest levels of radiation inside [247]. The final
specifications were arrived at using established empirical guidelines as well as
conservative estimates of the machine performance. Standarddensity
steelreinforced concrete was chosen for the bunker construction. Concrete is a
common choice for neutron shielding bunkers. Concrete offers both protection
from thermal neutrons as well as xrays. It was preferred over other materials
such as water, BoratedPolyethylene and higher density concrete. This decision
was based on factors such as cost, availability of qualified construction crews,
and earthquake safety.
The physical space of the bunker was chosen to be as large as possible
while still allowing for a control room and RF equipment within the rooms
allocated to PBPL. The final inner dimensions of approximately 9 m long, 3 m

304

wide and 2.5 m high ( 30' 10' 8.5' ) allows sufficient space for the full beamline,
as well as walkways for human access.
The shielding door (see Figure 8.11) was built to accommodate large pieces
of equipment being transported into and out of the bunker while still providing
the same shielding as the fixed concrete bunker walls. A sandwich of steel and
BoratedPolyethylene was chosen to match the gammaray and neutron
shieldingcapacity of concrete. Additionally, the door had to be structurally sound
in order to withstand the stress caused by repeated openings and closings.

Polyethylene
Steel

Figure 8.11: A crosssection of the shielding door showing the layers of material.

The bunker was also designed to accommodate the signal and control
cables used for the accelerator. Other penetrations for the waveguide, laser,
experiment signal and air conditioning were also incorporated, and are illustrated
in Figure 8.12.

305

31'-2"
Air in
4'-5"

Wire penetrations

Air out

Door
6'-6"
3'-7"

Figure 8.12: A side diagram of the shielding bunker showing some of the
penetrations.

8.9.2: Facilities Support


The needs of the accelerator and related experiments were taken into account
during the initial construction of PBPL. These needs lead to direct requirements
on the facilities. While the details of the design process are not appropriate for
inclusion here, the final specifications are recorded for reference.
The laboratory is temperature controlled using an air cooling system
independent of the building and based on a chilled water system provided
throughout UCLA [248]. One three ton cooler is used for the laser room in
order to achieve short term (hour) temperature stability of 1 centigrade. A
second unit of larger capacity (four tons) is used for the laboratory and bunker
mainly to remove heat produced by equipment. The result of the independent
cooling system is that the laboratory is insensitive to building temperature swings
caused by occupancy changes. Further, the ambient pressure of the laboratory is
slightly higher than the remainder of the building, which along with micron air

306

filters serves as an effective dust barrier an important consideration for stable


laser operation. The effects of the pressure difference between the laser room and
bunker on the laser transport are unknown.
In an effort to isolate the laboratory from outside electrical noise (and
viceaversa), a transformer was installed in the laboratory to generate all electrical
needs [249]. The building mains provide 200 A of threephase 480 VAC.
Threephase 208 VAC and 120 VAC are produced in the laboratory and distributed
to various subpanels (see Figure 8.13). All outlets are isolatedground type with
independent ground lines running to their respective subpanel to minimize ground
loops. Each electrical outlet was placed on an independent circuit (fuse) to prevent
the failure of one piece of equipment from disrupting the operation of others.
Finally, an uninterruptible power supply powers and isolates the computers.
sp

raceway

sp

sp

sp

DOOR

sp

Figure 8.13: The layout of the PBPL electrical system. Sub panels are indicated by
an sp.

Cooling water is provided through the building deionized closed circuit


system. Large (2" ID) pipes provide cooling water to the bunker. Plumbing boards
with filters, regulators, control valves and flow monitors provide hookups for the

307

accelerator as well as the experiments. Temperature control for the gun and linac
are provided by independent closed circuit heat baths.
The environmental stability, such as temperature, has played a significant
role in PBPL operations by determining such factors as long term stability and
component reliability.

8.9.3: Alignment System


During assembly the beamline components need to be aligned to each
other and to a reference line. The precision of alignment depends on the given
component and on the beam parameters. Typically, beampipe alignment is
noncritical (ignoring wakefields and resistive wall instabilities), while diagnostics,
accelerating structures and optics have stringent alignment tolerances. A standard
alignment system involves beamline components mounted on translation stages
or other movable platforms, with optical methods used to survey the objects.
Recent systems have implemented electronic sensors, interferometers and in situ
alignment using the beam itself as a diagnostic.
The PBPL system relies on conventional machining to achieve the desired
alignment. The system has only one degree of freedom along the beamline axis.
The transverse position and rotation are fixed by solid supports placed on optical
tables and butted against linear bearings (see Figure 8.14). This system has the
distinct advantage of not requiring expensive optical alignment tools. It also
dispenses with tedious alignment procedures and costly fiducial marks having to

308

be placed on all objects. This fixed alignment technique, of course, suffers from
not being correctable in situ (i.e., objects and brackets must be removed from the
beamline to be reworked). It also requires careful machining, and is limited by
the precision of the machining, optical tables and linear bearings. Thus, this
method is best suited to short systems with tolerances no better than standard
machining capabilities ( 0.001" = 10 3 in = 2.54 10 5 m = 25m).
Beamline Center

Beamline Center

Typical Mount
10.000
5.875

1.250

1.750

Optical Table

Figure 8.14: A cartoon of the fixed alignment system employing optical tables, an
alignment rod and machined brackets and supports.

The overall tolerance of the PBPL beamline can be estimated from the
minimum spot size:
=

(8.34)

309

where the notation is indicated in Table 1.5. Here we assume a normalized emittance
of 5x10-6 mrad and beam energy of ~17 MeV ( 33) and the focusing betafunction
equivalent to the undulators 0.11 m. Then, the beam size (one standard deviation)
is given by ~130 m. The alignment tolerance of the beamline can then be
crudely estimated by requiring it to be smaller than the minimum beam size.
Hence, beamline components should be aligned to within better than 100 m of
the ideal axis.
A more careful estimate of the alignment tolerance required can be
performed using beam propagation codes (matrix solvers) such as TRACE3D
and TRANSPORT. Having established the mechanical requirements of the
beamline, we turn to the requirements inside the beamline.

8.9.4: Vacuum System


Producing vacuum levels of 10-6 to 10 -9 Torr in beamlines several meters
long is common and relies on standard commercially available technology. Vacuum
systems still present technical and practical hurdles, however. Maintaining a
reliable vacuum system with thousands of seals, dozens of valves, feedthroughs
and windows is not without challenge.
The PBPL vacuum system was built to be as flexible as possible to allow
for rapid weektoweek changes and upgrades. With this need in mind, most
components used were offtheshelf commercial items. The main features of
the vacuum system are listed below:

310

All nonmagnetic stainlesssteel components


Knifeedge (Conflat) seals with copper gaskets
Majority of beamline >1" ID (2.75" flanges)
Ion pumps
Isolation Gate valves
Dry Nitrogen backfill valves
Pump down ports throughout beamline
Ionization gauges for monitoring gun, linac and waveguide
Portable turbomolecular pumpdown station
Portable helium leak checker
The system typically maintains vacuum levels in the mid 10-9 torr near the
gun, low 10-8 torr between the gun and linac and inside the linac, and mid 10 -7 to
high 10-7 torr in the remainder of the beamline. The waveguide vacuum system is
separated from the beamline using RF windows, but uses similar ion pumps and
maintains levels of <10-8 Torr near the pumps.
In order to achieve the UHV levels required in the gun, linac and interceding
section, bakeouts were performed using resistive heater tapes. This method
proved to be an endless source of problems. The heater tapes were not controlled
by thermocouples, causing occasional overheating and required constant vigilance
on the part of the operator. The fiberglass insulating material is a skin irritant to
the installers and a dust problem for the laser optics. Experience and frustration
eventually lead to using Nomex strips for insulation. This material is
nonirritating, nontoxic and nonflammable. However, it still produces dust and
cannot withstand as high temperatures as can Fiberglass. Further, experience has

311

shown that moderate bakeouts at < 150 C are sufficient to produce UHV. Due
to the proximity of magnets and diagnostics in the remainder of the beamline it
was deemed unwise and unnecessary to bake the system beyond the linac.

8.9.5: Control and Data Acquisition


To CAMAC
Patch Panels
ADC

Binary In
DAC

Binary Out

Monitoring

Control
Power
Supplies

BPM

Actuators

Interlock/
Monitor
Magnets

RF

Vacuum

Figure 8.15: The PBPL control system

The control area for the PBPL accelerator is designed for human operators
[250]. This should be contrasted with the remainder of the lab, which is designed

312

for the specific systems to operate optimally. The control system is based on
computers, electronics and human operators (see Figure 8.15. A central console
houses the main control computer and video systems (see Figure 8.16). Additional
racks house power supplies, CAMAC crates [251], RF controls, safety systems
and experimental hardware.
Control Console Front

CAMAC Rack
Solenoid Control

Store 1

Store 2

Slit motion control

D1

Magnet Supply Rack

D3

D2

D4

Degaussing Switcher

Switcher 1

Scope

Quad Supplies

Patch Panel

Gate valve / counter


mic / camera
Vacuum

Patch Panel

Master Camera

Patch Panel
Dipole Supplies

Computer
Monitor

Timing
CAMAC Crate
QDC, Delay, Motion

Switcher 2 decoder
TBC

Cables + Future
Crate Area

Computer

Steering Magnet Supplies

Current Supply

Switcher 2

Degaussing Supply (Bipolar)

VCR
Video
Printer

Drawer

Figure 8.16: The control room main console, CAMAC rack and power supplies.

8.9.5.1: Computer System


PBPL employs three Macintosh type computers for data acquisition, control
and general computational needs. Two of the computers are in the control area
and are connected through GPIB (IEEE 488) interfaces to the CAMAC crates and
scopes. The third computer is dedicated to laser diagnostics. Each of the computers

313

is equipped with a video digitizing card with 256 gray levels (8 bits). Two of the
computers are also equipped with general purpose analog and digital input/output
boards. The computers run LabVIEW software for the data acquisition and control.
The three computers are directly connected to a buildingwide ethernetbased
local area network (LAN). Printing and communications are handled though the
LAN (see Figure 8.17).

Ethernet
Computers

LabVIEW
Software

Video
subsystem

Printers,
Peripherals

GPIB
Oscillioscopes,
etc.

Trigger

CAMAC Crate
Figure 8.17: The control computer topology.

The main computer is a Quadra 950 with 24 MB of RAM, 1 GB drive. The


memory is required for the extensive video analysis performed on this machine.
The computational power of the Quadra was generally found to be sufficient for
the laboratorys needs.

314

8.9.5.2: Timing
The timing and triggering of the accelerator is difficult to think of as a
system, since it is spread across the laboratory and is comprised of several seemingly
disjoint components. Nevertheless, these components serve the same goal: to
provide pulses and gates to the various timecritical devices. The basic design
relies on a master oscillator and precision delay units to provide synchronization
and critical timing, while a series of trigger generators and delay boxes provide
less stringent timing pulses.

Thyratron Driver

Kilowatt Amp
Master Trigger
Generator

Safety Interlock

Vacuum
Interlock

Trigger
Distribution Box
Video System

Diagnostic Gate

Figure 8.18: The PBPL trigger system.

A block diagram of the triggering system is given in Figure 8.18. Following


the diagram from the top, the master oscillator (MO) operates at ~ 38.08 MHz
and is provided by the laser modelocker (see Section 4.6.1). The MO provides
signal for the laser system and, after seventyfive times frequency multiplication,
to the RF system. By using the MO to modelock the laser as well as provide a low

315

level RF source, the two systems are assured of being synchronized. In fact,
timing jitter needs to be reduced further and this is accomplished using a feedback
system. The tolerable timing jitter can be estimated by requiring that the electron
beam energy jitter (shot to shot) be small compared to the beam energy or
comparable to the energy spread. A good rule of thumb is that the timing (or
phase) jitter be less than one degree of RF phase. In our case (2.856 GHz) this
works out to be approximately one picosecond.
In addition to the RF synchronization, fast (low jitter) timing signals are
often required for diagnostics such as streak cameras. These are provided by a
digital delay unit (Stanford Research Systems Model 535). The unit provided
pulses of various widths. It is triggered by the laser regenerative amplifier at a
repetition rate of 5 Hz. The regenerative amplifier, in turn, is locked to the 60 Hz
power line zero crossing.

Master

Video

Diagnostics

KW Amp

Thyratron
Time
Figure 8.19: Timing sequence for a few signals on the RF and control systems.

316

The remaining signals are considered slow, and are only critical on the
microsecond timescale. These are provided through a set of delay boxes and gate
generators. A timing scale for some of the signals is shown in Figure 8.19.
It should be noted that since the regenerative amplifier is designed and
tuned to operate at the fixed repetition rate of 5 Hz, the laser triggering rate is not
adjusted. Rather, the RF triggering rate is controlled by using delays to skip
trigger signals produced from the laser.

8.9.5.3: Magnet System


The component that required the most effort and computer control hardware
was the magnet system. This is rather ironic considering that the magnets have
the least requirements in terms of data rate and signal bandwidth. Nevertheless,
the need to precisely control over thirty magnets of three different types presented
a challenge. This system does have some stringent requirements:
Current control to better than 0.005 A.
Ability to degauss each magnet
Knowledge of precise magnetic field or current
Need to slowly ramp supply current to magnets
Ability to save and restore settings
Manual override for direct user control
Meeting these requirements at available budget levels required a
combination of systems. Power supplies were both commercially obtained as

317

well as fabricated inhouse. CAMAC DACs were used to control the power
supplies, while ADCs were used to read back the current (via a shunt resistor) or
the manual control setting. For the dipole spectrometer, a Gaussmeter (Hall probe)
was used to accurately read back the field.

UniPolar 10A

For Quads
dI/I<10^-3

Computer
CAMAC
DAC
Patch Panel

UniPolar 15A

For
Dipoles

UniPolar 10A

Degaussing
Control Box

For Quads
dI/I<10^-3

UniPolar 15A

For Dipoles

Bipolar Current Controllers


Quads

Steering Magnets

Figure 8.20: The PBPL magnet system.

318

BiPolar Supply

For
degaussing

Dipoles

In order to degauss the magnets, a separate bipolar supply and switching


box was utilized. During degaussing, a magnet is switched from its normal supply
to the degaussing supply, the degaussing routine is run (under computer control),
and finally the magnet is switched back to its normal supply. This system saved
the cost of having to buy all bipolar supplies, while still allowing for quality
degaussing. Figure 8.20 is a schematic of the overall magnet system.
8.9.5.4: Video System
Video has proven to be a valuable data standard for laboratory use. High
bandwidth, large dynamic ranges and low costs have driven many laboratories
to rely on video data acquisition systems. PBPL has made extensive use of video:
over 15 cameras and 20 channels of video are implemented.
Daystrum
Computer

TBC

VCR

Decoder

Displays 4

Printer
MemScan
Switcher 1

Store 1

DA 1

Displays 1

Comp. Cntrl.
Switcher 2

Store 2

DA 2

Displays 2

Bunker
Cameras
Laser Room

TheQ
Computer

Displays 3

Computer control and analysis section

Figure 8.21: The PBPL video system.

The overall video system layout is shown in Figure 8.21. All video signals
are routed through the control console (via a patch panel) and the inputs are sent

319

into two switchers. Output from switcher 1 (computer controllable) is sent through
a distribution amplifier (DA) to various display monitors and the computer
digitizer. Switcher 2 is used to display a second source, select video to be recorded
by a VCR, or digitized on the auxiliary control computer. The video signals can
also be routed through a freeze frame unit which is synchronized with the system
trigger.
Synchronization of the video system to the RF/Laser timing system is
done indirectly. A master camera is used to generate a genlock (composite video)
signal which is distributed to the remaining cameras. The master camera is line
locked (to the 60 Hz). Since the RF/Laser timing system is also line locked, the
video system is synchronous. This arrangement was chosen for convenience and
cost. However, it suffers from not being directly triggered by the RF/Laser (master
oscillator) timing system. Thus, when a change is made to the system timing, the
video system must be rephased.
The Q
Computer
Saturnus2
Computer
TBC
Master Trigger IN
(Line Locked)

Standard Pulse
Gate/Delay Generator
Store 1

Store 2

Decoder
Figure 8.22: The video trigger distribution system.

320

The video system also requires a number of triggers for the various
components. These triggers are generated from the master trigger box and routed
through a gate generator and pulse distribution amplifier (see Figure 8.22).
8.9.5.5: Control and Acquisition Capabilities
The previous sections provided a brief description of some of the important
subsystems. We have covered how the systems work, now we discuss what the
system can do. While it is not appropriate to describe the entire control system
here, it is useful to list the main capabilities.
Controls:
Diagnostic actuators (software interlocked)
Video channel selection (based on diagnostic number)
Remote camera lens aperture
Magnet power supplies [252]
Magnet degaussing (software sequence through all magnets)
RFlaser phase (electronic and manual)
RF power and phase to gun and linac
RF trigger rate
Stepping motors (slits, optics, etc.)
Safety interlock (doors, radiation, etc.) [253]
Acquisition:
Magnet currents (settings readback)
Dipole field
RF power levels

321

Charge readings: ICT, Faraday cups


Video digitization and analysis

XYZZY

322

You might also like