SPALLATION NEUTRON SOURCE AND OTHER HIGH INTENSITY

PROTON SOURCES*

WEIREN CHOU
Fermi National Accelerator Laboratory
P.O. Box 500
Batavia, IL 60510, USA
E-mail: chou@fnal.gov

This lecture is an introduction to the design of a spallation neutron source and

other high intensity proton sources. It discusses two different approaches:
linac-based and synchrotron-based. The requirements and design concepts of
each approach are presented. The advantages and disadvantages are compared.
A brief review of existing machines and those under construction and proposed
is also given. An R&D program is included in an appendix.

1. Introduction

1.1. What is a Spallation Neutron Source?

A spallation neutron source is an accelerator-based facility that produces pulsed
neutron beams by bombarding a target with intense proton beams.
Intense neutrons can also be obtained from nuclear reactors. However, the
international nuclear non-proliferation treaty prohibits civilian use of highly
enriched uranium U235. It is a showstopper of any high efficiency reactor-based
new neutron sources, which would require the use of 93% U235. (This explains
why the original proposal of a reactor-based Advanced Neutron Source at the
Oak Ridge National Laboratory in the U.S. was rejected. It was replaced by the
accelerator-based Spallation Neutron Source, or SNS, project.)
A reactor-based neutron source produces steady higher flux neutron beams,
whereas an accelerator-based one produces pulsed lower flux neutron beams. So
the trade-off is high flux vs. time structure of the neutron beams. This course
will teach accelerator-based neutron sources.
An accelerator-based neutron source consists of five parts:
1) Accelerators
2) Targets
3) Beam lines

*
This work is supported by the Universities Research Association, Inc., under contract
No. DE-AC02-76CH03000 with the U.S. Department of Energy.

1
2

4) Detectors
5) Civil construction
A project proposal includes 1) through 5), plus a cost estimate, a schedule
and environment, safety and health (ES&H) considerations. This course will
teach part 1) only, although part 2) is closely related to 1) and a critical item in
the design of a spallation neutron source.

1.2. Parameter Choice of a Spallation Neutron Source

The requirements of neutron beams for neutron scattering experiments are as
follows:
• Neutron energy: low, about a few milli electron volts.
• Neutron pulse: sharp, about 1 µs.
• Pulse repetition rate: 10-60 Hz.
When an intense proton beam strikes on a target (made of carbon or heavy
metal), neutrons are produced via spallation. The production rate is roughly
proportional to the power deposited on the target.
Proton energies between 1 and 5 GeV prove optimal for neutron
production. At 1 GeV, each incident proton generates 20-30 neutrons.
The beam power P is the product of beam energy E, beam intensity N
(number of protons per pulse) and repetition rate f:

P (MW) = 1.6 × 10-16 × E (GeV) × N × f (Hz) (1)

Typical parameters of a modern high power spallation neutron source are:

• P ~ 1 MW
• E ~ 1 GeV
• N ~ 1 × 1014
• f ~ 10-60 Hz

1.3. Linac-based vs. Synchrotron-based Spallation Neutron Source

There are two approaches to an accelerator-based spallation neutron source:
linac-based and synchrotron-based.
A linac-based spallation neutron source has a full-energy linac and an
accumulator ring. It works as follows:
• A heavy-duty ion source generates high intensity H- beams.
• A linac accelerates H- pulses of ~1 ms length to ~1 GeV.
• These H- particles are injected into an accumulator via a charge
exchange process, in which the electrons are stripped by a foil and
dumped, and the H+ (proton) particles stay in the ring.
 3

• This injection process takes many (several hundreds to a few

thousands) turns.
• The accumulated protons are then extracted from the ring in a
single turn onto a target. The pulse length is about 1 µs.
• This process repeats 10-60 times every second.
A synchrotron-based spallation neutron source has a lower energy linac and
a rapid cycling synchrotron. It works differently.
• A heavy-duty ion source generates high intensity H- beams.
• A linac accelerates H- pulses of ~1 ms length to a fraction of a
GeV.
• These H- particles are injected into a synchrotron via the same
charge exchange process.
• This injection process takes many (several hundreds) turns.
• The H+ (proton) beam is accelerated in the synchrotron to 1 GeV
or higher and then extracted in a single turn onto a target. The
pulse length is about 1 µs.
• This process repeats 10-60 times every second.
Compared with a linac-based spallation neutron source, a synchrotron-
based one has the following advantages:
• For the same beam power, it would cost less, because proton
synchrotrons are usually less expensive than proton linacs.
• For the same beam power, it would have lower beam intensity,
because the beam energy could be higher.
• Because the injected linac beam has lower power, the stripping foil
is easier. Also, larger beam loss at injection could be tolerated.
• A major problem in a high intensity accumulator ring (a DC
machine) is the e-p instability. However, this has never been
observed in any synchrotron (an AC machine) during ramp.
The disadvantages of a synchrotron-based spallation neutron source
include:
• AC machines (synchrotrons) are more difficult to build than DC
machines (accumulators). In particular, the hardware is
challenging, e.g., large aperture AC magnets, rapid cycling power
supplies, field tracking during the cycle, eddy current effects in the
coil and beam pipe, high power tunable RF system, etc.
• AC machines are also more difficult to operate than DC machines.
Therefore, the reliability is lower.
One needs to consider all these factors when deciding which approach to
take for a spallation neutron source.
4

1.4. Spallation Neutron Source vs. Other High Intensity Proton Sources
Spallation neutron sources are an important type of high intensity proton
sources. However, a high intensity proton source may find many other
applications. For example:
• To generate high intensity secondary particles for high-energy physics
experiments, e.g., antiprotons (Tevatron p-pbar collider), muons (AGS,
JHF), neutrinos (K2K, MiniBooNE, NuMI, JHF), kaons (CKM,
KAMI, JHF), ions (ISOLDE), etc.
• To generate neutrino superbeams as the first stage to a neutrino factory
and a muon collider. (Such a high intensity proton source is called a
Proton Driver.)
• Nuclear waste transmutation (JHF, CONCERT).
• Energy amplifier (CERN).
• Proton radiography (AHF).
The design concept learned from this course can readily be applied to the
design of other high intensity proton sources.

2. High Intensity Proton Sources: Existing, Under Construction, and

Proposed

There are a number of high intensity proton sources operating at various

laboratories over the world. There are presently two large construction projects:
the SNS at the Oak Ridge National Laboratory in the U.S., and the JHF at the
KEK/JAERI in Japan. Each has a construction budget of about 1.3 billion US
dollars and is scheduled to start operation around 2005-2006. There are also
numerous proposals for Proton Drivers and other high intensity proton sources.
Table 1 is a summary based on a survey conducted during the Snowmass 2001
Workshop. 1
Among the existing machines, the highest beam power from a synchrotron
is 160 kW at the ISIS at Rutherford Appleton Laboratory in England. The
highest beam power from an accumulator is 64 kW at the PSR at Los Alamos
National Laboratory in the U.S.
The SNS is a linac-based spallation neutron source. The design beam
energy is 1 GeV, beam power 1.4 MW. The JHF is a synchrotron-based facility.
It has a 400 MeV linac, a 3 GeV rapid cycling synchrotron with a beam power
of 1 MW, and a 50 GeV slow ramp synchrotron with a beam power of 0.75
MW.
Several proposals of proton drivers have been documented and can be
found in Ref. [2]-[4].
 5

Table 1. High intensity proton sources: existing, under construction, and proposed
(Snowmass 2001 survey)

Machine Flux Rep Rate Flux† Energy Power

(1013 /pulse) (Hz) (1020 /year) (GeV) (MW)
Existing:
RAL ISIS 2.5 50 125 0.8 0.16
BNL AGS 7 0.5 3.5 24 0.13
LANL PSR 2.5 20 50 0.8 0.064
ANL IPNS 0.3 30 9 0.45 0.0065
Fermilab Booster (*) 0.5 7.5 3.8 8 0.05
Fermilab MI 3 0.54 1.6 120 0.3
CERN SPS 4.8 0.17 0.8 400 0.5
Under Construction:
ORNL SNS 14 60 840 1 1.4
JHF 50 GeV 32 0.3 10 50 0.75
JHF 3 GeV 8 25 200 3 1
Proton Driver
Proposals:
Fermilab 8 GeV 2.5 15 38 8 0.5
Fermilab 16 GeV 10 15 150 16 4
Fermilab MI Upgrade 15 0.65 9.8 120 1.9
BNL Phase I 10 2.5 25 24 1
BNL Phase II 20 5 100 24 4
CERN SPL 23 50 1100 2.2 4
RAL 15 GeV (**) 6.6 25 165 15 4
RAL 5 GeV (**) 10 50 500 5 4
Other Proposals:
Europe ESS (**) 46.8 50 2340 1.334 5
Europe CONCERT 234 50 12000 1.334 25
LANL AAA - CW 62500 1 100
LANL AHF 3 0.04 0.03 50 0.003
KOMAC - CW 12500 1 20
CSNS/Beijing 1.56 25 39 1.6 0.1
†
1 year = 1 × 107 seconds.
(*) Including planned improvements.
(**) Based on 2-ring design.

3. Design Concept of a Linac-based Spallation Neutron Source

A linac-based spallation neutron source has three major accelerator components:

• Linac front end
• Linac (full energy)
• Accumulator
We will discuss the design concept of each component in the following sections.
6

3.1. Linac Front End

The linac front end consists of an ion (H-) source, a pre-accelerator (Cockcroft-
Walton or RFQ), a low energy beam transport (LEBT), and a chopper.

3.1.1. H- source
H- ions have almost been universally adopted for multi-turn injection from a
linac to an accumulator ring. These ions are generated in an H- source. There are
several different types: surface-plasma source (magnetron), semi-planatron,
surface-plasma source with Penning discharge (Dudnikov-type source), RF
volume source, etc. This is a highly specialized field. There are regular
conferences and workshops devoted to this topic. The main challenges are to
provide ion beams with high brightness (i.e., high intensity and low emittance)
and to operate at high duty factor with a reasonable lifetime.

3.1.2. Cockcroft-Walton and RFQ

The kinetic energy of the H- particles from an ion source is about a few tens of
keV. These particles are accelerated by a pre-accelerator, which can be either a
Cockcroft-Walton or RFQ. The former has been in use for many years and has
a maximum energy of about 750 keV. In a number of laboratories it has been
replaced by the latter, which is a common choice of new accelerators. This is
because an RFQ has higher energy (several MeV) and a much smaller physical
size. Its beam has higher brightness and is bunched. (The Cockcroft-Walton
needs a buncher.) The design issues of an RFQ include high beam current, high
efficiency, small emittance dilution, and higher order mode (HOM) suppression.

3.1.3. LEBT
When an RFQ is used, one needs a low energy beam transport (LEBT) as a
matching section between the ion source and the RFQ. It consists of lenses that
focus the beam from the ion source, which is relatively large in radius and
divergence. The LEBT also usually contains source diagnostics and provides the
differential vacuum pumping between the source and the RFQ.

3.1.4. Chopper
The purpose of a chopper is to chop the beam so that it can properly fit into the
RF bucket structure in an accumulator. This would greatly reduce the injection
loss caused by RF capture. The requirements of a chopper are: short rise- and
fall-time (10-20 ns), short physical length (to reduce space charge effects), and a
 7

flat top and a flat bottom in the field waveform (to reduce energy spread in the
beam).
There are several different types of choppers:
• Transverse deflector: This is a traveling wave structure. It has
short rise- and fall-time. The shortcoming is its physical size
(about 1-meter long). It is used at the Los Alamos National
Laboratory and the Brookhaven National Laboratory.
• Electric deflector: This is a split-electrode structure for deflecting
the beam right after the LEBT. It was built at the Lawrence
Berkeley National Laboratory and will be installed in the linac
front end of the SNS project. 5
• Beam transformer (energy chopper): This is a new type of chopper
and is based on the fact that an RFQ has a rather small energy
window. A pulsed beam transformer that provides 10% energy
modulation to the beam in front of an RFQ can effectively chop
the beam. It has short rise- and fall-time and a short physical
length (about 10 cm). A prototype has been built by a KEK-
Fermilab team and is installed at the HIMAC in Japan for beam
testing. 6

3.2. Linac
The linac is the main accelerator. Its function is to accelerate the H- beam to full
energy (~ 1 GeV) before injection into the accumulator. Because the particle
velocity changes over a wide range during the acceleration (β = 0.046 at 1 MeV,
β = 0.875 at 1 GeV), the linac is partitioned to several parts. Each part uses a
different design to best match the corresponding β values.

3.2.1. Low energy part (below 100 MeV, β < 0.4)

Drift tube linac (DTL) is a common choice of this part. It is a matured
technology and has been used in every proton linac over the world. A potential
concern is that some vacuum tubes used to drive the RF cavities could have
supply problem because the vendors may terminate their production.
There is also an effort to develop superconducting RF cavities (the so-
called spoke cavity) for low β acceleration.

3.2.2. Medium energy part (100 MeV - 1 GeV, 0.4 < β < 0.9)
This is the bulk part of the linac. There are two design choices. One is room
temperature coupled-cell linac (CCL), another superconducting (SC) linac.
8

The CCL is a matured technology and has been used in all existing linacs
(e.g., Fermilab, Los Alamos National Laboratory, Brookhaven National
Laboratory, etc.). The highest energy using this technology reaches 800 MeV
(LANSCE at Los Alamos). However, the new project SNS has decided to use
an SC linac for good reason.
SC linacs have been proved reliable and efficient in electron machines (e.g.,
LEP and CEBAF). But still, it is a challenge for employment in proton machines
when operating in short pulse mode and accelerating particles with different β
values. In the past decade, SC linac technology has been making good and
steady progress. 7 Compared with a room temperature linac, an SC linac has the
following advantages:
• Higher accelerating gradient.
• Larger aperture (which is particularly important for high intensity
beams).
• Lower operation cost.
• Lower capital cost if higher energy is required. (There is an energy
threshold above which an SC linac becomes more economical.)
In addition to the SNS, CERN and KEK are also planning to use an SC
linac in their future machines (SPL and JHF Stage 2 linac, respectively).
Among various challenges to an SC linac, a crucial one is the RF control.
The allowable phase error (< 0.5°) and amplitude error (< 0.5%) are demanding.
One needs to investigate the choice of RF source (number of cavities per
klystron), redundancy (off-normal operation with missing cavities), feedback
and feedforward technique.

3.2.3. High energy part (above 1 GeV, β > 0.9)

In this range, particles travel at a velocity near that of the light and behave
similar to electrons. An SC linac is an obvious and probably also the only choice
from economical considerations. Several new high-energy proton linac
proposals (2.2 GeV at CERN, 3 GeV at Los Alamos, and 8 GeV at Fermilab)
have all picked this design.

3.3. Accumulator
As the name indicates, an accumulator is a ring that accumulates many turns of
injected particles and ejects them in a single turn. The purpose is to convert long
beam pulses (~ 1 ms) to short beam pulses (~ 1 µs) for experiments. It is a DC
machine. Its hardware is more or less straightforward (a main advantage of the
 9

accumulator approach). But this by no means implies an "easy" machine. On the

contrary, there are a number of challenges due to high beam power.

3.3.1. Beam loss control

This is the most challenging problem. For a 1 MW beam power, 1% beam loss
would give 10 kW, which already exceeds the full beam power on the targets
for most of the existing physics experiments. Therefore, allowable beam loss
must be much lower than 1%.
There are two types of beam loss: controllable and uncontrollable. One uses
specially designed collimators and dumps to collect the former so that the loss
can be localized. The uncontrollable beam loss would spread over the entire
machine and must be kept very low. The rule of thumb is that it must be below 1
W/m in order to make hands-on maintenance possible. For a 100-meter
machine, 1 W/m gives the total uncontrollable beam loss of 100 W, which is
0.01% of the total beam power. This is a goal not impossible but very difficult.
In the PSR at Los Alamos, which is a 64 kW accumulator, the total beam
loss is a fraction of a percent. Most of them are unstripped H0 and H- particles,
which are collected by special beam dumps.

3.3.2. Collimators and remote handling

Collimators are a critical part of an accumulator. They are used to localize the
beam loss and leave a majority part of the machine "clean."
Modern collimators use a 2-stage design. The primary collimator scatters
the halo particles; the secondary collimator (which can be more than one)
collects them. There is one set of collimators in each transverse plane.
Longitudinal collimators are also used, which are placed in high dispersion
regions. The design efficiency of collimators is 95% or higher.
The area near the collimators is very "hot" (highly radioactive). One must
use remote handling for maintenance in this area. Robot arms and cranes are
often employed. This should be an integral part in the machine design.
Invaluable experiences can be learned from LANSCE (Los Alamos, U.S.) and
PSI (Switzerland). These machines have been handling MW beams for years
and have designed several remote-handling systems that work reliably. 8

3.3.3. H- injection
This is another difficult part of the design and has many technical issues
involved.
10

• The stripping foil (usually made of carbon) must stand for high
temperature and large shock waves. It must also have high stripping
efficiency and a reasonable lifetime.
• The stripped electrons and unstripped H0 and H- particles must be
collected.
• During the many-turn (hundreds to thousands) injection, the orbit
bump needs to "paint" the particles in the phase space so that a uniform
distribution can be obtained. This would reduce the space charge
effect. The orbit bump also needs to minimize the average number of
hits per particle on the foil.
• The beam emittance dilution due to Coulomb scattering from the foil
should be kept under control.
• There are proposals from the KEK and Los Alamos for laser stripping.
The R&D is being pursued.

3.3.4. Lattice
The main requirement is multiple long straight sections, which are used for,
respectively, injection, extraction, RF and collimation.

3.3.5. e-p instability

This is a main beam dynamics problem in an accumulator. In the PSR at Los
Alamos, e-p instability is the bottleneck limiting the beam power. When the
beam intensity reaches a threshold, rapid beam oscillations (usually in one
transverse plane) occur that leads to fast beam loss. This instability is believed
to be caused by electrons trapped in the proton bunch gap. These electrons come
mainly from secondary yield. When a primary electron hits the wall, secondary
electrons are generated, which are accelerated by the proton beam and hit the
opposite side of the wall, generating more electrons, so on and so forth, causing
an avalanche.
This so-called electron cloud effect (ECE) has also been seen in electron
storage rings, in particular in the two B-factories: PEP-II and KEK-B. These
two machines effectively used solenoids to suppress this effect. However,
solenoids or clearing electrodes appear to be less useful in the PSR. (This is a
puzzle to be resolved.) Instead, the following measures have been found
effective in raising the instability threshold in the PSR: beam scrubbing (which
conditions the wall), inductive inserts (which make the proton bunch gap
cleaner), and sextupoles (which couple the motion in the two transverse planes
and stabilize the oscillation in one plane).
 11

This is an active research field. 9

3.3.6. Hardware
The challenge is the magnet, which must have large aperture in order to
accommodate large beam size and beam halo. Good field quality is necessary to
ensure large dynamic aperture.
Other technical systems, including power supplies, RF, vacuum and
diagnostics, are relatively straightforward.

4. Design Concept of a Synchrotron-based Spallation Neutron Source

A synchrotron-based spallation neutron source also has three major accelerator

components:
• Linac front end
• Linac (lower energy)
• Synchrotron
The designs of the linac front end and the linac are similar to that of a linac-
based spallation neutron source. However, a synchrotron design is very different
from an accumulator. Therefore, we will focus on the synchrotron in this
section.

4.1. Lattice
There are two basic requirements on the design: a transition-free lattice, and
several dispersion-free straight sections. For high intensity operation in proton
synchrotrons, transition crossing is often a major cause of beam loss and
emittance blowup. One should avoid it in the first place. Dispersion in the RF,
which is placed in one or more straight sections, may lead to synchro-betatron
coupling resonance and should also be avoided.
For a medium energy synchrotron (above ~ 6 GeV), regular FODO lattices
(in which γt ∝ √R, γt the lattice transition γ, R the machine radius) are ruled out
because they would use too many cells to achieve a high γt. Otherwise a
transition crossing is inevitable when the γ of the beam approaches γt during
ramp. There are several lattices that can give either a high or an imaginary γt so
that a transition crossing would not occur. For example, (a) a flexible
momentum compaction (FMC) lattice, which has a singlet 3-cell modular
structure with a missing or short dipole in the mid-cell; (b) a doublet 3-cell
modular structure with a missing or short dipole in the mid-cell. Figure 1 is an
example of (b), which is designed for a new 8 GeV synchrotron at Fermilab.
12

The choice of phase advance per module is of critical importance in this type
of lattice. There are two reasons. (i) The chromaticity sextupoles are placed in
the mid-cell, where the beta-function peaks and available space exists. In order
to cancel the higher order effects of these sextupoles, they need to be paired
properly. (ii) The phase advance per arc in the horizontal plane must be multiple
of 2π in order to get zero dispersion in the straights without using dispersion
suppressors (which are space consuming). Other requirements in the lattice
design include: ample space for correctors (steering magnets, trim quadrupoles,
chromaticity and harmonic sextupoles, etc.), ample space for diagnostics, low
beta and dispersion functions (to make the beam size small), large dynamic
aperture (to accommodate beam halo), and large momentum acceptance (to
allow for bunch compression when necessary).

ReesGarren Racetrack Rbend Lattice, No Trims

SUN SunOS 5.X version 8.21/0 03/05/02 16.53.47
25.0 3.0
β (m)

Dx (m)
βx βy Dx
22.5

2.5
20.0

17.5
2.0

15.0

12.5 1.5

10.0

1.0
7.5

5.0
0.5

2.5

0.0 0.0
0.0 5. 10. 15. 20. 25. 30. 35. 40.
s (m)
δ E/p 0 c = 0 .
Table name = TWISS

Figure 1. Lattice module of the Fermilab new 8 GeV synchrotron design. Each
module has three doublet cells. The dipole in the mid-cell is short. The phase
advance per module is 0.8 and 0.6 in the h- and v-plane, respectively. There are
five modules in each arc.

4.2. Space Charge

Amongst various beam physics issues, the space charge is a major concern. It is
often the bottleneck limiting the beam intensity in an intense proton source, in
particular, in a synchrotron, because the injection energy is low.
A useful scaling factor is the Laslett tune shift

(2)
 13

∆ν = - (3r/2) × (N/ εN) × (1/βγ2) × Bf

in which r is the classical proton radius (1.535 × 10-18 m), N the total number of
protons, εN the normalized 95% transverse emittance, β and γ the relativistic
factors, and Bf the bunching factor (ratio between peak and average beam
current). It shows the space charge effect is most severe at injection because βγ2
takes the minimum value. The situation becomes worse for high-intensity
machines not only because the intensity is high but also because the injection
time is long. Numerical simulation is the main tool to study this effect. A
number of 1-D, 2-D and 3-D codes have been or are being written at many
institutions. An example is shown in Figure 2. These codes are particularly
useful to the design of the injection orbit bump current waveform for achieving
uniform particle distribution in the beam, reducing emittance dilution and
minimizing average number of hits per particle on the stripping foil during the
phase space painting process. Several other measures, e.g., tune ramp, inductive
inserts, quadrupole mode damper and electron beam compensation are under
investigation for possible cures of the space charge effects.

Figure 2. Space charge simulation using Track2D (by C. Prior). It shows the
particle distribution after 45 turns injection in the Fermilab new 8 GeV
synchrotron with (left) and without (right) the space charge effect.

4.3. Other Beam Dynamics Issues

In addition to the space charge, there are several other beam dynamics issues
that need to be studied concerning an intense proton source.
• Electron cloud effect (ECE). This has been discussed in Section 3.3.5.
It is interesting to note that, by far all reported ECE is either in DC
machines (accumulators and storage rings) or AC machines in DC
14

operation (i.e., on flat top or flat bottom). No ECE has been seen in AC
machines during ramp. Does this imply that AC machines are immune
to ECE? If true, this would be an important advantage of the
synchrotron approach. However, this is solely an empirical
observation. Lack of a reliable theory for understanding and analyzing
the ECE is a loophole that urgently needs to be filled.
• Microwave instability of bunched beam below transition. Because the
machine will always operate below transition, the negative mass
instability due to space charge would not occur. Would then this
machine be immune to the microwave instability?
• Tune split. A split between the horizontal and vertical tunes is required
in order to avoid the strong resonance 2νx - 2νy = 0 that could be
excited by the space charge. However, it is not clear how big the split
needs to be. Does it have to be an integer? Or would a half-integer
suffice?

4.4. Beam Loss, Collimation and Remote Handling

This part is similar to that for accumulators as discussed in Sections 3.3.1 and
3.3.2. There is, however, an important difference. Because the injected linac
beam power is low (e.g., 6% of the full beam power if the acceleration range of
the synchrotron is 16), higher beam loss at injection (which usually accounts for
most of the total loss) can be tolerated. This is an advantage of the synchrotron
approach.

4.5. Slow Extraction

In addition to one-turn extraction, synchrotrons are also used for experiments
that require slow extraction (many turns). A critical issue is the efficiency.
Although the efficiency of one-turn fast extraction can exceed 99%, it is much
lower for multi-turn slow extractions. At high-intensity operation, the beam loss
in existing machines during slow spill is usually around 4-5%. This is not
acceptable for the next generation of high-intensity machines, in which the
beam power will be 1 MW or higher and one percent loss would mean 10 kW or
higher. This is a serious problem in the case of KAMI and CKM at the Fermilab
Main Injector, and kaon and nuclear physics programs at the JHF. A recent
ICFA mini-workshop was devoted to this topic. 10
 15

4.6. Hardware

4.6.1. Magnets
Magnets are one of the most expensive technical systems of a synchrotron. A
critical parameter in the magnet design is the vertical aperture of the main
bending magnets. The magnet cost is essentially proportional to the aperture. It
should be large enough to accommodate a full size beam including its halo. The
following criterion can be used in design:

A = {3 εN × βmax /βγ}1/2 + Dmax × ∆p/p + c.o.d. (3)

in which A is the half aperture, εN the normalized 95% beam emittance, βmax the
maximum beta-function, Dmax the maximum dispersion, ∆p/p the relative
momentum spread, c.o.d. the closed orbit distortion. The parameter 3 is the
estimated size of the beam halo relative to the beam size.
Because this is an AC machine, field tracking between the dipoles and
quadrupoles at high field is an important issue. Trim quads or trim coils are
needed. The peak dipole field should not exceed 1.5 Tesla. The peak quadrupole
gradient is limited by the saturation at the pole root (not pole tip).
The choice of the coil turn number per pole is a tradeoff between the coil
AC loss and voltage-to-ground. The former requires the use of many small size
coils, whereas the latter requires the opposite, namely, small number of turns.
There are two ways to compromise. One is to employ stranded conductor coils,
as shown in Figure 3, which was adopted in the JHF 3 GeV ring design.
Another is to connect several coils in parallel at the magnet ends, as done in the
ISIS. The ratio of the AC vs. DC coil loss should be kept around 2-3. The
voltage-to-ground should not exceed a few KV.
The aperture and good field region should include a rectangular area
(instead of an elliptical area). This is because there will be a significant number
of particles residing in the corners of the rectangle.
16

Figure 3. Stranded conductor coil for reducing coil AC loss.

4.6.2. Power supplies

This is another expensive technical system. There are several choices for the
power supplies in a rapid cycling machine. (1) A single harmonic resonant
system, e.g., the Fermilab Booster which resonates at 15 Hz. (2) A dual-
harmonic resonant system, e.g., the Fermilab new 8 GeV synchrotron which
uses a 15 Hz component plus a 12.5% 30 Hz component as shown below: 11

I(t) = I0 - I cos(2πft) + 0.125 I sin(4πft) (4)

in which f = 15 Hz, I0 and I are two constants determined by the injection and
peak current. The advantage of this system is that the peak value of dB/dt is
decreased by 25%, which leads to a saving of the peak RF power by the same
amount. (3) A programmable ramp system, e.g., the AGS Booster and AGS.
Although this is a most versatile system (e.g., allowing for a front porch and a
flat top), it is also most expensive.

4.6.3. RF
The RF system is demanding, because it must deliver a large amount of power
to the beam in a short period. In addition, it must be tunable, because the
particle revolution frequency increases during acceleration. Cavities with ferrite
tuners have been in use for decades. Recently the development of the Finemet
cavities at the KEK has aroused strong interest at many laboratories. Thanks to a
US-Japan collaboration, Fermilab has built a 7.5 MHz, 15 kV Finemet cavity
and installed it in the Main Injector for bunch coalescing. 12 The main
 17

advantages of the Finemet cores are high accelerating gradient and wide
bandwidth. The former is especially important for high intensity small size
rings, in which space is precious. The main concern, however, is its high power
consumption. For example, the Fermilab Finemet cavity needs a 200 kW power
amplifier to drive it. New types of magnetic alloys are under investigation for
performance improvement.

4.6.4. Vacuum
Vacuum pipe for a rapid cycling machine is probably one of the most
challenging items. Ceramic pipe with a metallic cage inside has been
successfully employed at the ISIS. However, this is a costly solution, because it
occupies a significant portion of the magnet aperture. Assuming the ceramic
wall and the cage need a 1-in vertical space, a 4-in aperture magnet would have
to increase its vertical gap by 25% to 5-in in order to accommodate this pipe.
This would directly be translated to a 25% increase in the magnet and power
supply costs, equivalent to tens of millions dollars.
Thin metallic pipe is an alternative. However, it must be very thin (several
mils) in order to minimize the eddy current effects (pipe heating and induced
magnetic field). Such a thin pipe is mechanically unstable under vacuum.
Several designs have been tried to enhance its stability, including ceramic
shields, metallic ribs and spiral lining. Prototyping of the first two designs did
not work well. The third one looks promising and is currently under
investigation. 13
Another alternative is that the magnets employ external vacuum skins like
those in the Fermilab Booster. Perforated metallic liners are used in the magnet
gap to provide a low-impedance environment for the beam.

4.6.5. Diagnostics
In addition to the conventional diagnostics for measuring beam position, tune,
profile, intensity and loss, intense proton sources have several specific
requirements. A system that can diagnose beam parameters during multi-turn
injection is highly desirable. The method for fast, accurate non-invasive tune
measurement is being developed. A circulating beam profile monitor covering a
large dynamic range with turn-by-turn speed will be crucial for studying beam
halo. (A similar instrument has been developed for the linac beam halo
experiment at Los Alamos. 7) There was also an ICFA mini-workshop devoted
to this topic. 14
18

4.7. New Ideas

In the past several years, there are a number of new or revitalized ideas
proposed to the high intensity proton source study. Here are a few examples:

4.7.1. Inductive inserts

They are made of ferrite rings and also can have bias current for impedance
tuning. Their inductive impedance would fully or partially compensate the space
charge impedance, which is capacitive. The first successful experiment was at
the PSR. 15 Two ferrite modules made by Fermilab have been installed in the
ring. They help increase the e-p instability threshold, which is a major
bottleneck of that machine. Another experiment is going on at the Fermilab
Booster.

4.7.2. Induction synchrotron

This is a longitudinally separated function machine. In other words, the
longitudinal focusing and acceleration are carried out by two separate RF
systems. The former uses barrier RF buckets, the latter a constant RF voltage
curve. One useful feature of this type of machine is tunable bunch lengths. So
the so-called superbunch acceleration could be possible. Because a superbunch
is similar to a debunched beam, the peak beam current is low. Thus, the space
charge effect can be reduced and beam intensity increased.

4.7.3. Barrier RF stacking

The application of Finemet and other magnetic alloys makes it possible to build
broadband barrier RF cavities with high voltage (~10 kV or higher). They can
be used to stack beams in the longitudinal phase space. This is particularly
useful when the beam intensity of a synchrotron is limited by its injector (e.g.,
the intensity of the Fermilab Main Injector is limited by the Booster).
Compared to the slip stacking, an advantage of barrier RF stacking is the greatly
reduced beam loading effects due to a lower peak beam current. 16,17

4.7.4. Fixed field alternating gradient (FFAG) accelerator

Although MURA first proposed this idea about 40 years ago, it was almost
forgotten. Only the recent activities at the KEK brought it back to the world's
attention. KEK has successfully built a 1 MeV Proof-of-Principle (PoP) proton
FFAG and is building a 150 MeV one. 18 FFAG is an ideal machine for high
intensity beams. Its repetition rate can be much higher than a rapid cycling
synchrotron (in the range of kHz). One problem of the FFAG, however, is that it
 19

is difficult (if not impossible) to fit it into an existing accelerator complex,

which usually consists of a linac and a cascade of synchrotrons.

4.7.5. Repetition rate increase in existing synchrotrons

This is a brute force approach but can be appealing because it is straightforward.
For example, the Brookhaven National Laboratory has a proposal for increasing
the AGS repetition rate from 0.5 Hz to 2.5 Hz. 19 The Fermilab Main Injector
upgrade also includes a rep rate increase (from 0.53 Hz to 0.65 Hz). 20

5. Design Concept of a Proton Driver

5.1. Differences between a Proton Driver and a Spallation Neutron

Source
A proton driver is a high intensity proton source. It can be used as a spallation
neutron source. But it can do more. It can generate neutrino superbeams and
other high intensity secondary particles (muons, kaons, pions, antiprotons, etc.)
for high-energy physics experiments. It can also be used as the first stage of a
neutrino factory and a muon collider.
There are two main differences between a proton driver and a spallation
neutron source.
• The beam energy of a proton driver is higher. A commonly used
production target is carbon. For a carbon target, the π- cross-
section is much lower than π+ when the proton beam energy is
below 4 GeV. Therefore, for polarized muon experiments, a proton
driver must be 4 GeV or higher. Furthermore, for neutrino
oscillation experiments, a proton source with tunable energy in the
range of several GeV up to about 100 GeV is preferred.
• The bunch length of a proton driver is shorter. The pion yield (i.e.,
number of pions per unit proton beam power) has a strong
dependence on the proton bunch length. This is the only parameter
that we have control to minimize the 6-D phase space volume of
the pions. Moreover, to obtain highly polarized pion beams also
requires short proton bunch length. The typical bunch length in a
proton driver is a few ns (instead of µs as in a spallation neutron
source).
20

5.2. How to Achieve Higher Energies

In a synchrotron-based design, this is not difficult. The energy covers a wide
range: from as low as 3 GeV (JHF, 1 MW) to as high as 120 GeV (Fermilab
Main Injector upgrade, 2 MW).
In a linac-based design, however, this is severely limited by the cost. The
existing highest energy proton linac is the LANSCE (0.8 GeV) at Los Alamos.
The SNS linac under construction at Oak Ridge is 1 GeV. There are proposals
for 2.2 GeV (CERN), 3 GeV (Los Alamos) and 8 GeV (Fermilab) proton linacs.
But none of these has become a construction project.

5.3. How to Obtain Short Bunch Lengths

In a synchrotron-based design, the bunch length is determined by the RF bucket
length, i.e., by the RF frequency. A short bunch length implies the use of a high
frequency RF system. However, sometimes there are good reasons to use low
frequency RF (e.g., to limit the number of bunches). In this case, a bunch
rotation technique can be used for compressing the bunch length.
It should be pointed out that there is a new beam dynamics problem
associated with bunch rotation, namely, the path length dependence on
momentum spread ∆p/p and space charge tune shift ∆ν. This is especially
important for proton drivers, in which due to large momentum spread (a few
percent) and large tune shift (a few tenth), the dependence of the path length ∆L
on ∆p/p and ∆ν can no longer be ignored. In other words, the momentum
compaction factor α = (∆L/L) / (∆p/p) cannot be treated as a constant during
bunch rotation. It is dependent upon the momentum and amplitude of each
particle. This will result in a longer bunch after rotation. Simulation study must
take this effect into account.
In a linac-based design, a compressor ring (separate from an accumulator
ring) will be needed in order to provide the required bunch length and bunch
structure.

6. Summary

Two recent major spin-offs from high-energy accelerators are synchrotron light
sources and high intensity proton sources. Both have found wide-range
applications in the field of basic sciences (e.g., material science, molecular
biology, chemistry, etc.) as well as in industrial research and development (e.g.,
chip technology, nano technology, medical and pharmaceutical research, etc.).
Spallation neutron source is an important type of the latter.
There are two approaches to a spallation neutron source. One is linac-based,
another synchrotron-based. Each approach has its pros and cons. The PSR at the
 21

Los Alamos National Laboratory and the SNS project at the Oak Ridge National
Laboratory belong to the former, while the ISIS at the Rutherford Appleton
Laboratory and the JHF project at the KEK/JAERI represent the latter. (Note
that the JHF is a multi-purpose facility unlike the SNS, which serves solely as a
neutron source.)
There are close connections between the design of a spallation neutron
source and a proton driver. The latter is a strong contender for a near term
construction project in the high-energy physics field in the U.S., Europe and
Japan. The studies of the two types of machines benefit each other.
The work on high intensity proton sources has been a dynamic field in the
accelerator world. There are numerous challenging problems as well as great
expectations. Out of the world's three large accelerator projects currently under
construction - LHC, SNS and JHF - two are high intensity proton sources.
Several more have appeared on the horizon. We'd like to encourage young
people to join this field and bring with them their energy, enthusiasm and fresh
ideas.

Acknowledgements

The author would like to express his gratitude to the 3rd OCPA International
Accelerator School for the invitation to give a lecture and for the hospitality
during his stay in Singapore.

Appendix A

There have been numerous conferences and workshops on high intensity proton
sources sponsored by the Beam Dynamics Panel of the International Committee
for Future Accelerators (ICFA). For example, there are a series of ICFA mini-
workshops on various specific topics, including transition crossing, particle
losses, RF, beam loading, transverse and longitudinal emittance measurement
and preservation, injection and extraction, beam halo and scraping, two-stream
instability, diagnostics and space charge simulations. These mini-workshops can
be found on the web http://www-bd.fnal.gov/icfa/workshops/workshops.html.
Paper proceedings are also available from the workshop organizers.
There was an ICFA-HB2002 workshop in April 2002 at Fermilab, which
covered almost all the aspects concerning high intensity proton sources. The
web address is http://www-bd.fnal.gov/HB2002/.
22

There was an ECLOUD'02 workshop also in April 2002 at CERN for the
study of electron cloud effect. The proceedings and presentations are posted at
http://wwwslap.cern.ch/collective/ecloud02/.
An international workshop on induction accelerators took place in October
2002 at the KEK. The web address is http://conference.kek.jp/RPIA2002/.

Appendix B

In July 2001, about 1,200 physicists over the world gathered at Snowmass,
Colorado, USA, for three weeks to discuss the future of high-energy physics.
One specific topic was high intensity proton sources. A detailed 26-point R&D
program was crafted. This program is directly related to the spallation neutron
source work. The Executive Summary is attached, which can be used as
guidance for planning future R&D. The full context can be found on the web:
http://www-bd.fnal.gov/icfa/snowmass/.

Executive Summary of Snowmass2001 on High Intensity Proton Sources

The US high-energy physics program needs an intense proton source (a 1-4

MW Proton Driver) by the end of this decade. This machine will serve multiple
purposes: (i) a stand-alone facility that will provide neutrino superbeams and
other high intensity secondary beams such as kaons, muons, neutrons, and anti-
protons (cf. E1 and E5 group reports); (ii) the first stage of a neutrino factory
(cf. M1 group report); (iii) a high brightness source for a VLHC (cf. M4 group
report).
Based on present accelerator technology and project construction
experience, it is both feasible and cost-effective to construct a 1-4 MW Proton
Driver. There are two PD design studies, one at FNAL and the other at the BNL.
Both are designed for 1 MW proton beams at a cost of about US$200M
(excluding contingency and overhead) and upgradeable to 4 MW. An
international collaboration between FNAL, BNL and KEK on high intensity
proton facilities addresses a number of key design issues. The sc cavity,
cryogenics, and RF controls developed for the SNS can be directly adopted to
save R&D efforts, cost, and schedule. PD studies are also actively pursued at
Europe and Japan.
There are no showstoppers towards the construction of such a high intensity
facility. Key research and development items are listed below ({} indicates
present status). Category A indicates items that are not only needed for future
machines but also useful for the improvement of existing machine performance;
 23

category B indicates items crucial for future machines and/or currently

underway.

1) H- source: Development goals - current 60–70 mA {35 mA}, duty cycle 6–

12% {6%}, emittance 0.2 π mm-mrad rms normalized, lifetime > 2 months
{20 days}. (A)
2) LEBT chopper: To achieve rise time < 10 ns {50 ns}. (B)
3) Study of 4-rod RFQ at 400 MHz, 100 mA, 99% efficiency, HOM
suppressed. (B)
4) MEBT chopper: To achieve rise time < 2 ns {10 ns}. (B)
5) Chopped beam dump: To perform material study & engineering design for
dumped beam power > 10 kW. (A)
6) Funneling: To perform (i) one-leg experiment at the RAL by 2006 with
goal one-leg current 57 mA; (ii) deflector cavity design for CONCERT. (all
B)
7) Linac RF control: To develop (i) high performance HV modulator for long
pulsed (>1ms) and CW operation; (ii) high efficiency RF sources (IOT,
multi-beam klystron). (all A)
8) Linac sc RF control: Goal - to achieve control of RF phase error < 0.5° and
amplitude error <0.5% {presently 1°, 1% for warm linac}. (i) To investigate
the choice of RF source (number of cavity per RF source, use of high-
power source); (A) (ii) to perform redundancy study for high reliability; (B)
(iii) to develop high performance RF control (feedback and feedforward)
during normal operation, tuning phases and off-normal operation (missing
cavity), including piezo-electric fast feedforward. (A)
9) Space charge: (i) Comparison of simulation code ORBIT with machine data
at FNAL Booster and BNL Booster; (ii) to perform 3D ring code bench
marking including machine errors, impedance, and space charge (ORNL,
BNL, SciDAC, PPPL). (all A)
10) Linac diagnostics: To develop (i) non-invasive (laser wire, ionization,
fluorescent-based) beam profile measurement for H-;(ii) on-line
measurement of beam energy and energy spread using time-of-flight
method; (iii) halo monitor especially in sc environment; (iv) longitudinal
bunch shape monitor. (all A)
11) SC RF linac: (i) High gradients for intermediate beta (0.5 – 0.8) cavity; (A)
(ii) Spoke cavity for low beta (0.17 – 0.34). (B)
12) Transport lines: To develop (i) high efficiency collimation systems; (A) (ii)
profile monitor and halo measurement; (A) (iii) energy stabilization by
HEBT RF cavity using feedforward to compensate phase-jitter. (B)
13) Halo: (i) To continue LEDA experiment on linac halo and comparison with
simulation; (ii) to start halo measurement in rings and comparison with
simulation. (all B)
24

14) Ring lattice: To study higher order dependence of transition energy on

momentum spread and tune spread, including space charge effects. (B)
15) Injection and extraction: (i) Development of improved foil (lifetime,
efficiency, support); (A) (ii) experiment on the dependence of H0 excited
states lifetime on magnetic field and beam energy; (B) (iii) efficiency of
slow extraction systems. (A)
16) Electron cloud: (i) Measurements and simulations of the electron cloud
generation (comparison of the measurements at CERN and SLAC on the
interaction of few eV electrons with accelerator surfaces, investigation of
angular dependence of SEY, machine and beam parameter dependence);
(A) (ii) determination of electron density in the beam by measuring the tune
shift along the bunch train; (A) (iii) theory for bunched beam instability that
reliably predicts instability thresholds and growth rates; (A) (iv)
investigation of surface treatment and conditioning; (A) (v) study of fast,
wide-band, active damping system at the frequency range of 50–800 MHz.
(B)
17) Ring beam loss, collimation, protection: (i) Code benchmarking &
validation (STRUCT, K2, ORBIT); (A) (ii) engineering design of
collimator and beam dump; (A) (iii) experimental study of the efficiency of
beam-in-gap cleaning; (A) (iv) bent crystal collimator experiment in the
RHIC; (B) (v) collimation with resonance extraction. (B)
18) Ring diagnostics: (i) Whole area of diagnosing beam parameters during
multi-turn injection; (ii) circulating beam profile monitor over large
dynamic range with turn-by-turn speed; (iii) fast, accurate non-invasive
tune measurement. (all A)
19) Ring RF: To develop (i) low frequency (~5 MHz), high gradient (~1
MV/m) burst mode RF systems; (B) (ii) high gradient (50-100 kV/m), low
frequency (several MHz) RF system with 50-60% duty cycle; (B) (iii) high-
voltage (>100 kV) barrier bucket system; (B) (iv) transient beam loading
compensation systems (e.g. for low-Q MA cavity). (A)
20) Ring magnets: (i) To develop stranded conductor coil; (ii) to study voltage-
to-ground electrical insulation; (iii) to study dipole/quadrupole tracking
error correction. (all B)
21) Ring power supplies: To develop (i) dual-harmonic resonant power
supplies; (ii) cost effective programmable power supplies. (all B)
22) Kicker: (i) Development of stacked MOSFET modulator for DARHT and
AHF to achieve rise/fall time <10-20 ns; (B) (ii) impedance reduction of
lumped ferrite kicker for SNS. (A)
23) Instability & impedance: (i) To establish approaches for improved estimates
of thresholds of fast instabilities, both transverse and longitudinal
(including space charge and electron cloud effects); (ii) to place currently-
used models such as the broadband resonator and distributed impedance on
a firmer theoretical basis; (iii) impedance measurement based on coherent
tune shifts vs. beam intensity, and instability growth rate vs. chromaticity,
 25

including that for flat vacuum chambers; (iv) to develop new technology in
feedback implementation. (all B)
24) FFAG: (i) 3-D modeling of magnetic fields and optimization of magnet
profiles; (ii) wide-band RF systems; (iii) transient phase shift in high
frequency RF structures; (iv) application of sc magnets. (all B)
25) Inductive inserts: (i) Experiments at the FNAL Booster & JHF3; (A) (ii)
programmable inductive inserts; (B) (iii) development of inductive inserts
which have large inductive impedance and very small resistive impedance;
(B) (iv) theoretical analysis. (B)
26) Induction synchrotron: (i) Study of beam stability; (ii) development of high
impedance, low loss magnetic cores. (all B)

References

1. The Snowmass 2001 Workshop web site: http://www.snowmass2001.org/.

2. W. Chou, C. Ankenbrandt, and E. Malamud, editors, "The Proton Driver
Design Study," FERMILAB-TM-2136 (December 2000).
3. G.W. Foster, W. Chou, and E. Malamud, editors, "Proton Driver Study II,"
FERMILAB-TM-2169 (May 2002).
4. M. Vretenar, editor, "Conceptual Design of the SPL, a High-Power
Superconducting H- Linac at CERN," CERN 2000-012 (December 2000).
5. J. Staples et al., Proc. 1999 PAC (New York, USA), p. 1961.
6. W. Chou et al., “Design and Measurements of a Pulsed Beam Transformer
as a Chopper,” KEK Report 98-10 (September 1998); W. Chou et al., Proc.
1999 PAC (New York, USA), p. 565; Y. Shirakabe et al., Proc. 2000 EPAC
(Vienna, Austria), p. 2468.
7. T.P. Wangler, "Linac Based Proton Drivers," AIP Conference Proc. 642, p.
43 (2002).
8. E. Wagner, "Remote Handling and Shielding at PSI," AIP Conference
Proc. 642, p. 265 (2002).
9. For references see http://wwwslap.cern.ch/collective/ecloud02/.
10. The 10th ICFA Mini-Workshop on Slow Extraction, October 15-17, 2002,
Brookhaven National Laboratory, USA. The proceedings can be found on
the web: http://www.agsrhichome.bnl.gov/ICFA2002/.
11. Chapter 5 of Ref. 3.
12. D. Wildman et al., Proc. 2001 PAC (Chicago, USA), p. 882.
13. Z. Tang, "A New Kind of Vacuum Tube for Proton Driver," FERMILAB-
TM-2188 (September 2002).
14. The 11th ICFA Mini-Workshop on Diagnostics, October 21-23, 2002, Oak
Ridge National Laboratory, USA. The proceedings can be found on the
web: http://www.sns.gov/icfa/.
15. M.A. Plum et al., Phys. Rev. ST-AB 2, 064210 (June 1999).
26

16. K. Koba and J. Steimel, "Slip Stacking," AIP Conference Proc. 642, p. 223
(2002).
17. K.Y. Ng, "Doubling Main Injector Beam Intensity using RF Barriers," AIP
Conference Proc. 642, p. 226 (2002).
18. Y. Mori, "Progress on FFAG Accelerators at KEK," ICFA Beam Dynamics
Newsletter, No. 29, p. 20 (December 2002).
19. W.T. Weng and T. Roser, "The AGS High Power Upgrade Plan," AIP
Conference Proc. 642, p. 56 (2002).
20. Chapter 13 of Ref. 3.