Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Stanford Linear Accelerator, shown in an aerial digital image. The two roads seen near the accelerator are California Interstate 280 (to the East) and Sand
Hill Road (along the Northwest). Image data acquired 2004-02-27 by the United States Geological Survey
R. M. Jones
The University of Manchester, UK
PHYS4722, Physics of Particle Accelerators
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 1
Linac -References
Linear Accelerators, North Holland Publishing Co. John Wiley Interscience Div. (1970)
eds. Pierre Lapostolle and Albert Septier.
The Stanford Two-Mile Accelerator, W.A. Benjamin (1968) ed. Richard Neal.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 2
RF Linear Accelerators
RFQ Operating
at 425 MHz
DTL tank #2 in the
Fermilab Linac operating
at 425 MHz R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 3
Chronological Development
Cyclotron
Van Der Graff (1930) Lawrence (1930)
Synchrotron
Oliphant (1943)
Synchrocyclotron and Betatron
Electrostatic Field Based
McMillan and Veksler (1944)
Alvarez Linac
McMillan (1946)
Strong Focusing
Courant and Snyder (1952)
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 4
Livingston Curve
Computer scientists
see the imminent end
of Moores law
around 2020, the
exponential growth
in accelerator
technology, described
by the so-called
Livingston curve, is
perhaps also
reaching its technical
limit?
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 6
Mechanical Analogue to RF Linac
Bowling alley model of particle
accelerator: gravity accelerates balls
motion on the sloping part of the
track. The height of each horizontal
section corresponds to a voltage source
in the particle accelerator. The amount
of acceleration is limited by the
difference in levels between the top
and foot of the entire track.
1929 Widere
The ion velocity fits the criterion that the distance between the midpoint
of each gap is given by /c c/f .
The Widere design is a standing wave cavity in which the accelerating
electric field maxima and nodes remain fixed in space. At a moment when
the rf voltages are maximum on each electrode, acceleration takes place in
every other gap.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 8
Birth of Linacs Contin.
After crossing these gaps, the particles (ions) enter the metallic
cylindrical electrodes which act as a shield to the electric fields. These
particles then exit the electrode after the field in the next gap has
changed polarity.
The length of these cylindrical electrodes must become longer as the
particles are accelerated to make the beam arrive at the next gap
when the rf voltages are appropriate for accelerating.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 9
Birth of Linacs -Contin
The history of particle accelerators for ion beams is often described in
association with the development of cyclotrons, primarily due to their wide-
spread use in the medical field.
However, what is often not acknowledged is that ion linear accelerators
("linacs") were developed in parallel with the cyclotron and other circular
accelerators.
While Lawrence and Livingston designed the first small cyclotron in 1930,
R. Widere had already published a paper in 1928 on his results from an rf
powered linear accelerator for ions.
This device, followed a proposal in 1925 by G. Ising. It consisted of a series
of cylindrical tubes, placed along the longitudinal axis of an evacuated glass
cylinder. Alternate tubes were connected to opposite terminals of an rf
generator.
By selecting the frequency and applied rf voltage, a variety of heavy ions
could be accelerated across the gaps and bunched simultaneously.
Sloan and Lawrence used such an array of 30 electrodes excited with a
voltage of 42 kV at an rf frequency of 10 MHz in 1931 to accelerate Hg+ ions
to 1.26 MeV.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 10
Drift Tube Linacs (DTL)
Luis Alvarez ,
Inventor of the DTL.
In the late 1940's, after the second World War, a program was initiated
by E.O. Lawrence at the University of California Radiation Laboratory
(now known as the Lawrence Berkeley National Laboratory) with the US
Atomic Energy Commission to investigate electronuclear breeding of
Pu239, U232 and tritium by bombarding depleted uranium with
accelerator-produced neutrons.
A series of high power rf linacs for protons and deuterons was built and
tested starting in 1950 at the site which is now the Lawrence Livermore
National Laboratory.
When the first linac was built, the only other proton linac which had
been operated was the 32 MeV linac built by L. Alvarez, although a 68
MeV p+ linac was under construction at the University of Minnesota.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 11
Drift Tube Linacs (DTL)
The large linacs in the Material Testing Accelerator program (12-48 MHz)
were disassembled in the mid 1950's, but the scientists who worked on these
systems went on to build many successful proton linacs at research facilities
such as Brookhaven National Laboratory, Argonne National Laboratory,
Fermi National Accelerator Laboratory and Los Alamos National Laboratory.
Most of these modern linacs, which are used for physics research, were
based on the original 200 MHz design of L. Alvarez.
Typical linear gradients achieved in these Alvarez linac structures are 1-2.5
MV/m, with the gradients in the gaps ranging from 6-10 MV/m. Using the
development of strong focusing magnets (quadrupole magnets) that occurred
in 1938, these structures also employed quadrupole magnets within the
cylindrical drift tubes.
These magnets are needed to keep the ion beam focused, because it becomes
slightly defocused by the radial electric field components present in the
accelerating gap. The need to have magnets in the drift tubes, and several
other problems such as the defocusing in the gap, require that the relative ion
velocity, = /c, must be greater than 0.04 for effective use of such a "drift
tube linac". This is particularly true for high beam currents of protons, which
will diverge rapidly due to the repulsive charge of the other ions in the beam
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 12
Drift Tube Linacs (DTL)
Fundamental Principles of DTL
Operation:
5. The drift tube length and spacing increases to keep pace with the increasing particle velocity as they gain energy.
6. The bunches are timed to arrive in the centre of the gap, as the field is increasing, so that those arriving early gain
less energy, and those arriving late gain more energy.
7. The beam is focused in the transverse direction by strong permanent magnet quadrupoles inside each drift tube.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 13
Drift Tube Linacs (DTL)
Thus,the origin lies at the geometrical center of the gap only if the field is symmetrical.
The energy gain is given by:
2z
E=q m sin( + )dz
L L
2z 2z
= qcos m sin( )dz + qsin m cos( )dz.
L L L L
The first term vanishes as a consequence of the definition of the electrical center:
2z
L m
(z)cos(
L
)dz
E=qsin L mdz here quantity in parenthasis is the transit factor.
L
m (z)dz
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 16
Thus,we obtain:
E=qFsin
(z)dz dz.
m
L
dz
L
L
(z)cos(2z / L)dz
The transit factor, F=
L
m
is evaluated with a suitable
(z)dz
L m
computer code, such as 'superfish'. The remaining term in parenthasis is the mean
value of the field, averaged over a cell and this finally gives an expression for the energy
gain:
E=qFm,av Lsin .
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 17
However, the increments of momentum are constant and we show this
below.
Tn =E0 1+n2 1
Tn =E0 1+2C2 1
However, recall: Tn = E0 2 2 1/ 2 1 and thus the final kinetic energy depends
( Ln )
on E0 and the mechanical parameters of the accelerator. Thus, relating this to the above
equation, leads us to the conclusion that C is similarly determined once the machine is
fabricated.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 19
Radio Frequency Quadrupole (RFQ)
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 22
Radio Frequency Quadrupole (RFQ)
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 23
Radio Frequency Quadrupole (RFQ)
Non-
Equally Equally
Spaced Spaced
Electrodes Electrodes
As in any RF linear accelerator, the reason for the time dependent field is
clear without time dependence the overall energy gain would be zero as the
particles would receive equal amounts of acceleration as well as
deceleration.
The spatial period of the electrode displacements must match the axial
distance traversed by a synchronous particle during one RF period.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 26
High Energy Linacs -Introduction
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 27
RF Linear Accelerators
In general, the aim is to transfer energy from the RF wave to electron beam
consisting of bunches of charged particles
If we inject the RF into a waveguide then on average, a electron beam
traversing the waveguide gains no energy from the e.m. field.
Why? Because the phase velocity of the RF wave is larger than the velocity
of light and so it runs ahead of the electron beam the electron beam sees both
the accelerating and decelerating part of the RF field and this averages to
zero.
In order to exchange energy, the phase velocity of the RF has to be matched
with that of the electron beam. This is achieved by slowing the RF field down
in an aptly named slow wave structure which consists of the original
waveguide loaded down, periodically, with irises. This is also known, for
obvious reasons, as a disk loaded structure and is illustrated below
Disk loaded Accelerating
Structure Suitable for
Electron Acceleration. The
characteristic parameters are
also indicated -a and b refer
to the iris radius and cavity
radius, respectively (very
common nomenclature)
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 28
RF Linear Accelerators
The influence of the irises on
slowing down the wave is readily
seen by referring to the dispersion
diagram.
The original smooth waveguide is
also illustrated.
The irises form a periodic
structure within the cavity, reflecting
the wave as it passes through and
causing interference.
This process is similar to the interference of light in a diffraction grating
Loss-free propagation in the grating occurs at z = pd, with p=1, 2, 3..
Thus it is clear that 2/p = 2d/ z with p=1, 2 , 3
Applying this to the disk-loaded linac, we find only certain wavelengths
propagate characterised by mode number p.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 29
RF Linear Accelerators
One can imagine using any value
of p at all.
In practise there are a limited
number of modal configurations
used and these are illustrated
adjacent.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 33
E(r,,z)=e- z E1 (r, ,z)
H(r,,z)=e- z H1 (r, ,z)
The field must also satisfy the usual transverse boundary conditions.
For the lowest order, monopole mode in a disk-loaded waveguide the field is:
Ez = E0n J0 (kc,nr)e j z n
n
Er = j E0n J1 (kc,nr)e j z n
kc,n
j k0
H =
Z0
kc,n
E0n J1 (kc,nr)e j z ,where :
n
n2 = k02 kc,n
2
, with k 0 = / c
The propagation constant of the fundamental differs from that
of the unloaded guide. In fact, one finds that 0 decreases as
the perturbation due to the irises increases. The phase velocity of the
space harmonics is given by:
vpn =
2n
0 +
d
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 34
and the group velocity is the same for all space harmonics:
1 1
d dn d
vgn = = = d = vg
dn d
One can see this by referring to the dispersion diagram.
Thus, for a given frequency, an infinite series of space harmonics
are excited. The fundamental space harmonic has the largest Fourier
amplitude. Also, for a structure designed to be synchronous
with the fundamental space harmonic the integrated effect on the beam
is such that all non-synchronous space harmonics average to zero over
the length of the cavity
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 35
Thus, one designs the disk-loaded structure to interact with the
fudamental space harmonic with a disk to disk period of d=()/(/c) with
=1 for the mode and 2/3 for the 2/3 mode, etc.
Also, at synchronism the light line intersects with the dispersion curves
and thus 0 = /vb (vb is the velocity of the beam).
The details of the dispersion curves are obtained with finite element or
finite difference computer codes such as Superfish, HFSS, MAFIA,
Microwave Studio, GdfidL.
However, given a limited number of points on the curves (zero and pi)
one can use a circuit model to within remarkably good precision to map
out the remaining part of the dispersion curve.
The method outlined for disk-loaded structures applies equally well to
more elaborate cavities such as the superconducting Niobium cells in the
ILC which consist of elliptical irises and cavities.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 36
The gradient is predicted to be ~ 19.9 MV/m. In operation 15 20 MV/m
have been obtained.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 37
RF Linear Accelerators
Single cavity structures are
also used.
In damping rings and FFAGS Single Cell Cavity
for example. Resonating at 500 MHz
Developed for Storage
Clearly the cost becomes Ring DORIS
prohibitive when several
thousand or more are required.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 38
It is advantageous to add as many cells as possible per cavity in order to
reduce the number of couplers required.
There are, of course, practical limitations on the number of cells
allowed per cavity:
1. as the number of cells increase the spacing between neighbouring modes
decreases
2. The power density requirement increases with the number of cells and
hence one may be forced to reduce the number of cells to avoid the
surface of the cavity suffering from electrical breakdown. Nonetheless,
the ILC had 55 cells with a gradient of 65MV/m in 100ns. The ILC
superconducting cavities have adopted a conservative approach as only
9 cells are contained in each cavity and 9 cavities per module in the
present RDR design.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 39
Fundamental RF Linac Parameters
The limiting quantity in linacs is usually the RF power, either peak power or
average power.
Since the power dissipated per unit length of the structure Pd is proportional to
the square of the RF field, a useful parameter is the shunt impedance per unit
length :
R = E2z/Pd.
(Sometimes, R' is defined in terms of r.m.s. rather than peak values, making it a
factor two less!)
The shunt impedance is also defined in terms of the stored energy U as:
2
j z
E (z)e
L
z
c
dz
R=
4U
R' is typically given in M /m. In proton linacs operating at 200 MHz values
of 35 M/m are reached, whereas electron linacs at 3 GHz have values around
100 M /m.
The Q-value is a factor of merit of an RF cavity as a resonator. It is defined as
the ratio of the stored energy to the energy dissipated per radian of the RF cycle.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 40
The group velocity is important for the following three reasons:
1. The filling time -the time to fill a cavity or structure of length l
with energy: Tf = l / vg
2
2. As P=W'vg and W' ~ Eacc , it is preferable to have a low group
velocity in order to maximise the energy density and the accelerating field.
3. R', Q and R'/Q all depend on vg . As a rule, decreasing vg increases R' and
decreases Q and hence R'/Q is increased.
A wave traveling down the structure is attenuated due to wall losses. The
rate of attenuation is obtained from the continuity of power flow:
W' P
+ + Pd' = 0
t t
P P
+ vg + P=0
t t Q0
and in the steady state, the time derivative is zero:
P=P0e2z , =
2Q0 vg
This can be interpreted in terms of the time required
for a field in a cavity to die down to 1/e of its initial value.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 41
The accelerating field is proportional to the tangential magnetic field at the wall
(although with some reshaping of cavities matters are changed a little):
Eacc ~ Htw ~ i w / beff
The stored energy and energy transport both scale as:
W' ,P ~ Eacc
2 2
beff
Thus, bearing mind beff ~ 1 and putting all components together:
2
R'~1/(beff R'w )~1/2
Q ~ / R'w ~ 1/ 2
R'/ Q ~
vg = P / W' ~ 0
For a constant group velocity structure the filling time is also a constant
for a given structure length. But to optimise the field, the structure length
changes and:
Tf ~ 3/ 2
Thus, from the point of view of RF power, the frequency should be as
large as possible.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 42
However, we have ignored two further issues:
1. High frequency implies smaller structures and this can lead to RF electrical breakdown
2. The beam excites higher order modes, which constitute a wakefield and these lead to
a beam break up instability, at worst, or at the very least a dilution in the beam emittance.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 43
Travelling Wave Linear Accelerators
SC TESLA 9-Cell
Cavity operating at
1.3 GHz. For the ILC
the baseline design
gradient is 35 MV/m.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 45
High Energy Linacs
Finally we note that instead of operating a SW-cavity in the -mode one
can operate in the 0-mode with good efficiency.
Again, forward and backward space harmonics add up in phase in every
cell and both contribute to the acceleration. Since the electromagnetic fields
in adjacent cells are in phase the seperating walls can be left out without
affecting the field distribution.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 46
Monopole Mode Dispersion Relations
Utilize an L-C circuit model.
This represent an infinitely repeating periodic C C C
structure. L L L L L L
For our purposes the Ls and Cs are the same for In-1 In In+1
47
Dividing by 2jL gives:
r2 1
1- 2
i n + (i n +1 + i n 1 ) = 0, where r = .
2 2 LC
For an infinitely periodic structure the phase changes across one cell by .
i n = i 0 e jn
r2 jn j j jn
1- 2 e + (e + e )e = 0
2
r2
1 - 2 + Cos = 0
/ 2 CAD Model of 5-Cell
=
1 + Cos Positron Capture Cavity
(designed and fabricated at SLAC)
where the /2 is recognised as the cell resonant frequency
(verify by letting =/2).
In general <<1
=> ~ / 2 1 Cos
2
48
0 ~ / 2 ( 1 / 2 ) and ~ / 2 ( 1 + / 2 ) .
Thus the bandwidth of the cell defines
the coupling coefficient:
=
( 0 )
.
/ 2
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 51
2 r2 r
Using = 1 2 j then:
Q
i1 + i 2 = 0
2
in + ( in+1 + in1 ) = 0
i N +1 + i N = 0
2
Multiplying and subtracting we obtain an upper triangular matrix:
1 1
+ =
i1 i2 0, ( - + =
)i2 i3 0, ( - )i3 + i4 , ....
2 / 2 - (1/ / 2)
= cos
2
1 cos2
1 =
2 cos
1 cos cos3
= 2cos + =
1 cos2 cos2
2
1
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 52
cos(N + 1) + cos(N 1) sin sinN
=
2cosN cosN
The determinant of a triangular matrix is equal to the product of
the diagonal elements for a homogeneous system we set it equal to zero:
det=(-)N sin sinN = 0
Eigenvalues are given by n = n / N (n=0,1,2..N)
2 r2 r
and n = 1 2 j = 2cos n
n nQ
For Q>>1 we obtain:
1
n =r / 1 + cos(n / N) ~ / 2 1 cos(n / N)
2
For a general value of Q solving for n :
1
j 1
n = / 2 + 2 + 1 + cos(n / N)
2Q 4Q
The real component of frequency is increased due to the presence of losses:
Re{n }~ / 2 [1 + cos(n / N) + 1/(8Q2 )]
2
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 53
Similarly, for the eigenvectors:
in = cos(p 1)n = cos (p 1)n / N
Clearly, n is the phase shift per cell. The bandwidth defines the
separation between the 0 and mode:
=N 0 ~ / 2
Thus, the coupling factor, = / / 2 (=M/L) is the normalised bandwidth.
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 54
Now, the mode separation should be larger than that defined by the
Q losses bandwidth (=r /Q0 ):
N < Q0 and N / 2 < Q0
2 4
Monopole 3/6
Curve
2/6
0 /6 Light line
R.M. Jones, Physics of Particle Accelerators, PHYS 4722, Linacs Lecture, 2011. 56