Reliability issues for linear colliders

Abstract

To deliver a high integrated luminosity over several years of operation, a linear collider must not only meet its energy and luminosity performance goals, but also have a very high hardware availability and operating efficiency. The first challenge is the size and complexity of the facility. If the typical reliability of existing High Energy Physics accelerators is simply scaled to the size of a 500 GeV linear collider, the overall system availability will be too low. The final design must incorporate a more rigorous failure analysis as well as built-in overheads and redundancy. An additional challenge is the complexity of the tuning procedures required to preserve a very small beam emittance. These include beam-based alignment of magnets and rf structures, automated trajectory correction, feedback, emittance and luminosity optimization, and more. Another issue is the inherently large power densities in the beams, which can damage any beamline components they intercept. An extensive machine protection system is necessary to inhibit beam in case of a fault and automatically execute a recovery sequence. This paper will present the important issues in the context of the proposed linear collider designs.

Full text

Page 1

SLAC–PUB–9969
June 2003
Reliability Issues for Linear Colliders*
N. Phinney, C. Adolphsen, M.C. Ross
Stanford Linear Accelerator Center, Stanford University, Stanford, California 94309
Abstract
To deliver a high integrated luminosity over several years of operation, a linear
collider must not only meet its energy and luminosity performance goals, but also
have a very high hardware availability and operating efficiency. The first challenge
is the size and complexity of the facility. If the typical reliability of existing High
Energy Physics accelerators is simply scaled to the size of a 500 GeV linear collider,
the overall system availability will be too low. The final design must incorporate a
more rigorous failure analysis as well as built-in overheads and redundancy. An
additional challenge is the complexity of the tuning procedures required to preserve
a very small beam emittance. These include beam-based alignment of magnets and rf
structures, automated trajectory correction, feedback, emittance and luminosity
optimization, and more. Another issue is the inherently large power densities in the
beams, which can damage any beamline components they intercept. An extensive
machine protection system is necessary to inhibit beam in case of a fault and
automatically execute a recovery sequence. This paper will present the important
issues in the context of the proposed linear collider designs.
Invited talk presented at 2003 Particle Accelerator Conference
Portland, OR, USA
May 12 – May 16, 2003

* Work supported by Department of Energy contract DE–AC03–76SF00515.

Page 2

RELIABILITY ISSUES FOR LINEAR COLLIDERS*
N. Phinney#, C. Adolphsen, M.C. Ross, SLAC, Stanford, CA 94309, USA
Abstract
To deliver a high integrated luminosity over several
years of operation, a linear collider must not only meet its
energy and luminosity performance goals, but also have a
very high hardware availability and operating efficiency.
The first challenge is the size and complexity of the
facility. If the typical reliability of existing High Energy
Physics accelerators is simply scaled to the size of a 500
GeV linear collider, the overall system availability will be
too low. The final design must incorporate a more
rigorous failure analysis as well as built-in overheads and
redundancy. An additional challenge is the complexity of
the tuning procedures required to preserve a very small
beam emittance. These include beam-based alignment of
magnets and rf structures, automated trajectory correction,
feedback, emittance and luminosity optimization, and
more. Another issue is the inherently large power
densities in the beams, which can damage any beamline
components they intercept. An extensive machine
protection system is necessary to inhibit beam in case of a
fault and automatically execute a recovery sequence. This
paper will present the important issues in the context of
the proposed linear collider designs.
HISTORICAL PERSPECTIVE
The early generations of accelerators were high energy
physics machines which were technically innovative.
Their primary emphasis was on achieving breakthroughs
in energy and luminosity, usually under tight cost
constraints. Given the overhead of fills and ramping for
storage rings, the luminosity uptimes achieved were in the
range of 50%.
The relative importance placed on reliability has
evolved with the advent of accelerator user facilities such
as the synchrotron light sources, and with the new
generation of high energy physics ‘factories’. The large
energy-frontier colliders such as the Tevatron at FNAL,
HERA at DESY, LEP at CERN, and SLC at SLAC have
achieved hardware availabilities in the range of 70-90%.
In contrast, the B-factories at SLAC and KEK have closer
to 95% availability for the colliders themselves.
Synchrotron light or spallation sources have invested
significant effort into improving reliability and now reach
98-99.5% [1].
While it is true that these facilities are often smaller
than the energy-frontier machines, and in some respects
less demanding as to performance, the reliability achieved
does not appear to scale with the size of the complex.
Rather, it appears that the user facilities and factories have
higher standards for acceptable availability and therefore
allocate the necessary resources to reach that target level.
AVAILABILITY GOALS
A reasonable goal for a future linear collider would be
to have a hardware availability of 80-85%. Hardware
downtime should include unscheduled repairs (something
critical breaks), scheduled repairs (either at regular
intervals or when enough problems have accumulated),
and all associated cooldown, warmup and recovery times.
Typically in the past, only the light sources have included
maintenance periods in their downtime accounting, but
this is really appropriate for all facilities. Modern
accelerators do not require routine ‘preventative’
maintenance and interventions are only ‘scheduled’ when
there is broken hardware. Hence, they take away from the
overall beam time that might otherwise be delivered. Note
that each maintenance intervention takes on the order of 3
shifts, including edge effects and recovery. A ‘day’ every
3 weeks represents already a 5% hit.
The overall operating efficiency or beam availability is
typically significantly smaller than the hardware uptime.
The integrated luminosity delivered is closer to half of
what might be expected from the peak rate, even for the
high performance ‘factories’. Beam inefficiencies include
Machine Development (time spent studying and
improving the accelerator), the impact of tuning
procedures, injection and the luminosity decay during a
store (for storage rings), Machine Protection trips and
recovery (for linacs), and last but not least, the simple fact
that accelerators do not manage to deliver the same
luminosity every pulse or every store. A reasonable goal
for a linear collider would be a beam efficiency of 75-
80%, which would produce a delivered luminosity equal
to ~65% of peak performance.
Achieving this availability goal will be a challenging
task for a facility the size of a linear collider, but it is
necessary in order to integrate significant luminosity.
Experience with the SLC and more recently with
recommissioning the upgraded Tevatron and HERA has
shown that poor reliability can impact the peak luminosity
achievable as well as the integrated performance. If the
hardware interruptions are too frequent, the machine is
not up long enough to effectively make progress on the
luminosity issues. It was only after the SLC achieved
reasonable reliability that the many beam tuning
challenges for a linear collider could be addressed. The
more complex next generation of colliders must be
designed from the start for high availability so that the
inevitable new problems can be overcome rapidly and
effectively.
SIGNIFICANT ISSUES
Several aspects of a linear collider make achieving high
reliability particularly challenging. First is the sheer size
of the facility and the number of components which must
____________________________________________
*Work supported by the U.S. Dept. of Energy under contract DE-
AC03-76SF00515.
#nan@slac.stanford.edu

Page 3

be functional if the machine is to operate. If the typical
reliability of existing HEP accelerators is simply scaled to
the size of a 500 GeV linear collider, then the resulting
uptime will be unacceptably low. Fortunately, this is
amenable to engineering solutions. Reliability with large
numbers of components was studied extensively for the
Superconducting Super Collider project in Texas and
more recently for projects such as the Large Hadron
Collider in Switzerland and the Spallation Neutron Source
in Tennessee. Reliability must be addressed up front by
failure analysis, and appropriate remedies must be
implemented. Adequate engineering margins are essential
if components are to perform reliably in the long term.
The key issue is whether sufficient engineering and
financial resources are actually allocated during
development and construction to produce a reliable
system.
Linac rf systems
The main linac rf system demands particular care
because of the large number of components with
relatively short lifetimes. Table 1 lists the component
counts as given in the 2003 Technical Review Committee
Report [2]. The klystrons (and some modulator components)
must be replaced frequently and are considered a
consumable expense. In addition, the modulators,
klystrons, distribution system, and structures or cavities
will experience brief faults or breakdown events where
the hardware can be reset and continue operation after an
appropriate timeout. Because each unit contributes only a
small fraction of the total energy, a fault will typically not
interrupt operation, but simply cause that pulse to be
slightly low in energy. All linear collider designs plan to
include spare rf units which can be switched in when a
unit faults or needs repair. Critical issues are the
frequency and impact of faults, the adequacy of the spares
overhead, and the accessibility and duration of repairs.

TESLA JLC-C JLC-X/NLC
Modulators 572 2138 508
Klystrons 572 4276 4064
Rf Distribution n/a 2138 2032
Structures/
Cavities
20592 6784 12192
Spare rf units 2% 5% 5%
Table 1: Main linac rf components required for 500 GeV
center-of-mass (both linacs)
In the linear collider designs based on warm rf
technology, the klystrons and modulators are installed in a
separate support housing where they are accessible for
repair while the collider is delivering luminosity. Since
they can be replaced more or less continuously, the
number of spares required is determined by estimated
fluctuations in the failure rate. In the present JLC-X/NLC
designs, 5% overhead has been allocated to cover both
faults and failures. The design based on superconducting
rf technology described in the TESLA Design Report [3]
has a single tunnel. The modulators are installed in
support housings but the klystrons, transformers, and
high-power pulsed cables are in the tunnel with the
accelerator and can only be repaired during a shutdown.
The stated goal is to have a maintenance intervention no
more often than every three weeks. This would be
difficult to achieve without substantially more overhead
than the allocated 2% spares.
Tuning procedures
Another aspect which makes a linear collider
particularly challenging is in the complexity of the tuning
procedures required to preserve a very small beam
emittance. In all areas of the collider from the damping
rings to the interaction point, the component alignment
tolerances are extremely tight (micron-scale) and cannot
be achieved by traditional survey techniques. All of the
designs foresee extensive use of beam-based alignment.
In addition, the tight tolerances make the machines very
sensitive to vibration (nanometer-scale) and to slow drifts
due to temperature and ground motion effects. As a result,
beam-based feedback systems are mandatory, and both
invasive and non-invasive retuning will be required at
intervals.
Regardless of the main linac rf technology, no linear
collider can be considered a static machine and tuning is
required on a variety of timescales. Feedback is essential
to keep the beams in collision. Without it, they would drift
apart between pulses of the machine by as much as tens of
nanometers at a noisy site, such as Hamburg, to a fraction
of a nanometer at a quiet site, such as the LEP tunnel.
TESLA plans to bring the beams into collision and
optimize the positions within a single long bunch train.
NLC/JLC-X use pulse-to-pulse feedback at 100-120 Hz to
damp motion at frequencies below about 10 Hz.
Trajectory feedback is required to keep the beams
centered in the strong final focus sextupoles or the
luminosity degrades within minutes. Trajectory feedback
is required elsewhere to damp transients and correct slow
drifts. Energy feedback must compensate for fluctuations
in the total linac energy due to rf faults as well as to a
variety of rf phase or amplitude errors. Re-steering of the
main linacs and damping rings will be needed on the time
scale of hours and dispersion correction of the rings on
the time scale of days.
Alignment tolerances
The alignment tolerances differ for the two techno-
logies, as do the methods forseen to correct errors. The
quadrupole and cavity tolerances are 10 and 100 times
looser for the superconducting main linacs, but the X-
band linac will have high precision position monitors on
both structures and magnets, and movable stages on each
magnet or girder to effect the required alignment. In the
damping rings, the situation is reversed. The X-band
damping rings are similar to third generation light sources
and have tolerances which are no more than a factor of 3

Page 4

tighter than what has already been achieved [4]. The
damping rings for TESLA will have tolerances which are
another 3-10 times tighter, and these must be maintained
over a 17 km circumference. Meeting these tolerances
may require movable stages for precision alignment
and/or more rigid supports than presently foreseen. Table
2 compares the tolerances for the TESLA and NLC
damping ring designs with those for the recently
completed Swiss Light Source (SLS) as given in Ref. 4.
In the final focus, the alignment tolerances are similar for
both designs but the superconducting collider is more
sensitive to vibration because the low repetition rate limits
the frequency to which feedback can be effective.

SLS NLC TESLA
Energy [GeV] 2.4 2 5
Circumference [m] 288 300 17,000
Sext vert align [µm]
71 31 11
Quad roll align [µrad]
374 322 38
Quad vert jitter [nm] 230 75 76
Table 2: Alignment tolerances for the TESLA and NLC
damping rings compared with tolerances at the Swiss
Light Source (SLS)
MACHINE PROTECTION
The small, very intense, beams in a linear collider
require a new approach to machine protection (MPS)
untested at any existing or soon to be completed machine.
The pulsed time structure of the beam, as opposed to the
CW nature of storage rings like the Tevatron or LHC is an
additional difficulty. A single, nominal intensity, bunch
will damage almost any accelerator hardware it happens
to strike downstream of the damping rings. Since it is not
possible to stop a given beam bunch once extracted from
the damping ring, there is little fundamental difference in
the MPS exposure or design strategy for the different
machines. The long inter-bunch interval in TESLA allows
the beam to be switched off somewhat more quickly than
in JLC/NLC. The minimum time required to turn off the
beam is one full interpulse period for the JLC/NLC short
train and about 1/10 of the train length (~100 µs or 300
bunches) for TESLA.
Protection system schemes have been proposed for both
TESLA and JLC/NLC which appear feasible [5]. They
must automatically control changes in beam power, both
by halting operation when a fault is detected and by
restoring operation when the fault is cleared. They rely
heavily on the use of a pilot bunch and a fast permit
system. The permit signal is derived from beam data
taken on the previous pulse and from a system that
monitors fast devices. Before operation can be resumed
after a fault, the MPS must provide for the production of a
sequence of pilot and low power pulses that prove the
fitness of the downstream systems for high power
operation.
TUNNEL CONFIGURATION
The TESLA Design Report [3] proposed a collider built
within a single tunnel. This tunnel would contain the
beam lines for the superconducting main linacs, damping
rings, injectors, injector linacs, positron production, and
beam delivery systems. The linac klystrons with pulse
transformers, rf controls and high power pulsed cables, as
well as many power supplies and electronics, would also
be installed in the same tunnel. In contrast, the X-band rf
machine would have separate tunnels for the injectors and
damping rings and separate accessible support housings
for klystrons, power supplies, electronics, etc. to facilitate
repair during operation.
A single tunnel would require interrupting operation at
frequent intervals to access the tunnel to replace failed
klystrons and repair other components. Great care would
be needed to ensure that all in-tunnel components had
extremely high reliability. Because a single tunnel would
house almost all beamlines, linac access would also
impact the rings and injectors. The single tunnel also
limits flexibility in initial commissioning. All of these
issues would need to be carefully assessed with regard to
reliability and efficiency. The single tunnel choice was
driven by cost considerations and constraints of the DESY
site, but could well be reconsidered for a superconducting
linear collider built elsewhere.
CONCLUSIONS
To deliver the integrated luminosity demanded by the
physics goals, a linear collider will need to be designed
for very high hardware availability and beam efficiency.
Nominal goals of 80-85% for hardware availability and
75-80% for beam efficiency will not be achieved without
considerably more effort than has often been devoted in
the past. A robust design requires rigorous failure
analysis, generous built-in overheads and redundancy for
critical components. Complex tuning procedures will
demand an unprecendented level of automation. Overall
these goals should be achievable, but only if sufficient
attention and resources are allocated from the earliest
design stage through commissioning and operation.
REFERENCES
[1] L. Hardy, “Accelerator Reliability - Availability”,
Proc. of EPAC02 (Paris, France, June 3-7 2002)
[2] International Linear Collider Technical Review
Committee Second Report 2003, SLAC-R-606,
January 2003.
[3] R. Brinkmann, et al., (eds.), TESLA Technical
Design Report Part-II, DESY-2001-011, March 2001.
[4] T. Raubenheimer, A. Wolski, "Comparison of Align-
ment Tolerances in the Linear Collider Damping
Rings with those in Operating Rings," LCC-112,
January 2003.
[5] C. Adolphsen, et al., “The Next Linear Collider
Machine Protection System,” SLAC-PUB-8130,
Proc. of PAC99, new York, NY, 29 Mar-2 Apr 1999