Content uploaded by Darren Kelly
Author content
All content in this area was uploaded by Darren Kelly on Jun 10, 2016
Content may be subject to copyright.
DUST IN ACCELERATOR VACUUM SYSTEMS
Darren R.C. Kelly, DESY, Notkestr.85, 22603 Hamburg, Germany
Abstract
Many accelerators of electron beams are susceptible to per-
sistent beam lifetime disruptions, with correspondingly re-
duced performance. One distinguishes between three pre-
vailing explanations of these disruptions: (1) trapping of
positive atomic ions in the negatively charged beam; (2)
trapping of small highly positively ionised micro-objects
(“dust”) in the negatively charged beam; (3) disruptions
due to stray magnetic objects trapped in the magnetic field
of undulators.
The lifetime disruption of certain electron storage rings
that employ ion-getter pump systems are attributed by most
researchers to explanation (2), the trapped dust hypothe-
sis. Systematic experimental studies of HERA, PETRA
and DORIS reinforce the suspicion that specifically this
type of pump system is the culprit.
Examples of beam lifetime disruptions are presented, to-
gether with a summary of observations and experiments
performed at various afflicted storage rings to investigate
dust trapping and the connection between ion getter pumps
and dust particle release. Observations of the disrupted
beam are found to agree with the dust trapping hypothesis.
1
2
3
4
5
6
7
8
9
10
5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 6
lifetime [h] & BLM rate/100
time [h]
HERA-e lifetime disruption on 12dec1995
lifetime [h]
count/100 loss monitor 159 (NL191)
Figure 1: Example of the reaction of a beam loss monitor
during a lifetime disruption in HERA-e at injection energy
12 GeV on 12 Dec 1995. The beam loss monitor count Ri,j
divided by 100 for monitor NL191 (number j=159) and the
lifetime τin hours are shown against time ti. Reductions
in the lifetime coincide with local loss rate increases.
1 INTRODUCTION
Various electron storage rings are afflicted by a spurious re-
duction of the beam lifetime – apparently caused by the de-
flection of beam electrons by positively charged dust parti-
cles of size order 1µm trapped in the electric field potential
of the negativelycharged beam: Super-ACO [1], TRISTAN
AR [2], CESR [3], HERA-e, PETRAII, DORISIII [4, 5],
PF [6]. The complementary experience at these machines
has built a consistent picture of the symptoms of macropar-
ticle lifetime disruptions, and this picture agrees quantita-
tively with the dust trapping model detailed in [3, 4].
This electron beam lifetime problem is beam charge de-
pendent, i.e. it only occurs with electron beams, as evi-
denced by the problem-free switch to positron operation in
HERA and DORIS.
Observations confirm that the ion getter pumps of all
above-mentioned machines are implicated in casting dust
particles into the beam pipe. The lifetime disruption can be
provoked by switching an ion getter pump on and off, or
by abruptly increasing the pump voltage, when a discharge
within the pump can be sometimes be observed.
Such lifetime disruptions are not observed in the non-
evaporative getter (NEG) pump based storage ring LEP
or at ESRF (with NEG pumps and auxilliary lumped ion
pumps), with the exception of the possible capture of mag-
netic dust in undulators [7]. Trials over a limited region of
HERA with NEG pumps indicate a strong if not complete
reduction of the release of disrupting particles within this
NEG pump region.
It will be illustrated in this report that the extensive ob-
servations of beam lifetime disruptions in a number of ac-
celerators are well described by the trapped dust hypothe-
sis. However, the formation or liberation of dust particles
within the complicated environmentof the ion-getter pump
chamber is poorly understood. Relatively few direct ob-
servations of the processes within these pumps systems are
available, whereas many observations of pump configura-
tions and operating conditions conducive to the onset of
beam lifetime disruptions have been performed, in particu-
lar with HERA, PETRA and DORIS. Such observations are
however not sufficient to form a causal model of dust par-
ticle release into the beam pipe; our knowledge of the pro-
cesses on the pump side of the vacuum chamber gas con-
duction slits is as poor as the available diagnostics. A num-
ber of interesting observations and experiments of dust-like
disruptions at other afflicted storage rings will also be dis-
cussed.
2 AN EXAMPLE OF LIFETIME DISRUPTIONS IN
HERA-E
Of all available diagnostics of the disrupted beam, the 214
beam loss monitors of HERA [8] are perhaps the most
revealing. In Fig.1 we see an example of a typical life-
35470-7803-4376-X/98/$10.00 1998 IEEE
Figure 2: The beam loss monitor count ratio Ni,j =Ri,j /Ri−1,j reveals the influence of dust particles moving longitudi-
nally around the HERA electron ring. The grayscale corresponds to a range 0.5...3 inNi,j at 12 GeV, where 1 represents
background (no change). The ratios are plotted over the (i, j)-plane of BLM monitor numbers j=1...214 for the entire
machine against time ti. Current in mA, energy 12 GeV and lifetime in h are shown onan extended scale
time disruption in HERA-e at injection energy 12 GeV1.In
this particular HERA-e run a number of abrupt beam life-
time reductions seemed to correlate with increases in the
rate Ri,j of beam loss monitor number j=159 at position
NL191 over times ti.
More global insight into the dust trappings is obtained
when the time development of the beam loss monitor count
ratio Ni,j =Ri,j /Ri−1,j of all monitors j=1...214
around the machine is displayed simultaneously in one di-
agram for a range of times ti. Thus changes in the reaction
of each monitor are reflected by departure from Ni,j =1,
which value then represents predominantly the synchrotron
radiation background and beam electron losses due to de-
flection from residual gas molecules in the vacuum cham-
1In HERA-e at 12 GeV the beam loss monitor reaction to scattered
electrons dominates over the loss monitor’s reaction to synchrotron radi-
ation, permitting easy recognition of events, whereas at e-p luminosity
energy 27.5GeV transient reactions can be identified but at 3% of the
count are difficult to resolve against the dominant synchrotron radiation
background, so examples in this paper are restricted to injection energy
12 GeV .
ber. In Fig.2 the count ratio is associated with grey shades
for a range 0.5 – 3 in Ni,j for the same HERA-e run on
12 Dec 1995 at 12 GeV . The reader will be rewarded by
taking a few minutes to peruse and understand this some-
what overwhelming diagram, which illustrates most of the
important aspects of the dust trapping problem. A number
of features of the disrupting particles can be identified in
this diagram:
•The longitudinal flight of many particles can be easily
discerned as bright flight trails, and their velocities can
be measured to be around 10 to 100 m/s (in HERA-e
at beam energy 12 GeV and current 30-40 mA).
•Both transient and instransient disruptions of the life-
time can be seen to correspondwith particles entering
the beam. The lifetime is plotted against time on the
extended axis.
•It can be seen that there are hundreds to thousands
of particles passing through the beam per hour, only
a handful of which are permanently trapped. Many
3548
particles do not survive their flights along an arc, per-
haps due to thermal, structural or dynamic instability.
There is clearly a narrow stability window.
•The reaction of monitor number 159 (position NL191)
shown individually in Fig.1 can be identified at the
instant of lifetime reduction (extended axis).
3 THE TRAPPED DUST HYPOTHESIS
Most if not all symptoms of the electron beam lifetime dis-
ruption can be quantitatively explained by the prevailing
dust trapping model as detailed in [3, 4], which I sum-
marise briefly. Macroparticles, perhaps of SiO2or metallic
oxides from the beam pipe and ion pump surfaces, are cast
into the beam pipe by numerous ion pumps at frequent in-
tervals, where they are rapidly ionised and drawn into the
electron beam by the beam’s strong electric field. The par-
ticles are transversely trapped and rapidly reach an equilib-
rium charge determined by competition between ionisation
by beam electrons and deionisation through field evapora-
tion and photoelectron capture. The equilibrium charge ob-
tained by SiO2particles of sizes 0.1−1µm – as computed
by integration of a trapped particle’s equation of motion
[12] with charge development after [4] – is listed in Ta-
ble1 for current 20 mA and energy 27.5 GeV in HERA-e.
The particles oscillate transversely at frequencies of a few
kHz. Particles of low mass-to-charge ratio, i.e. of radius
0.1µm are transversely unstable.
The macroparticles are driven downstream by Mœller
scattering at about 12 ms−2[4] until they are possibly
trapped in horizontally defocussing quadrupoles by restor-
ing kicks due to the longitudinal asymmetry of the beam
bunches from the β-function gradient there [3]. In HERA
particles of radius <1µm achieve an equilibrium charge
meeting this longitudinal trapping criterion.
Table 1: Charge number Q, mass number to charge number
ratio Q/A, and and transverse oscillation frequencies ob-
tained by particles of different radii Rtrapped in the core
of the HERA electron beam at 20 mA current at 27.5 GeV.
RQ A/Q f
xfz
[µm] [Hz] [Hz]
1.0 5.81E+06 1.04E+06 1510.88 3152.41
0.5 4.75E+06 1.58E+05 3863.99 8062.09
0.3 4.04E+06 4.02E+04 7668.93 16001
0.1 2.72E+06 2.23E+03 32576.5 67969.8
Particles of size 0.1 – 1 µm should be thermally stable in
HERA-e at typical current 40 mA and energy 27.5GeV ,
although [4] has predicted that these particle might be ther-
mally unstable at very high electron currents 40 mA.
In view of this model the myriad of activity in the beam
loss monitor diagram Fig.2 becomes understandable. A
large number of dust particles with a distribution of sizes
are cast into the beam pipe at many locations in the ma-
chine. Only the handfull of particles per hour meeting
the transverse, longitudinal and thermal stability conditions
survive to cause intransient lifetime disruptions,the others
leaving merely brief trails of losses as they are swept down-
stream before melting, falling out of the beam, or becoming
structurally unstable.
4 EXPERIMENTAL INVESTIGATIONS
A number of chance observations and machine studies at
various institutes have added greatly to our knowledge of
the lifetime disruption problem, some of which are now
briefly described.
Since the inception of CESR sudden lifetime dropshave
been observed and by comparison of the magnitude of
the observed lifetime drops with a model of a trapped
macroparticle’s equilibrium charge as a function of particle
mass the typical particle size was estimated to be of order
1µm[3].
At TRISTAN AR dust particles were dropped into the
beam pipe via a 1 mm nozzle to see whether macroparticles
could indeed be trapped in an electron beam [14]. Metal
particles consisting of Cu, Al, Ti (sizes 0.1-8 µm) were not
trapped, whereas metallic oxides such as Cu0 (0.35 µm),
TiO2(0.3 µm), and the NEG compound Zr-V-Fe (∼1µm)
were trapped for many minutes at currents order 10 mA and
energy 6.5 GeV. Particles of diameter 0.5 µm consisting of
C (diamond), SiC and Al2O3were trapped even at very low
currents (less than 0.1 mA).
Surprisingly,the TRISTAN AR beam lifetime was found
to be poor after dumping and refilling, and bremsstrahlung
signals and electron losses typical of trapped particles were
still observed. This “hysteresis” effect has also been ob-
served at HERA, and is notyet understood. Simple consid-
erations predict that the image charge force on a dust parti-
cle exceeds the beam electric field force within a few mm
of the vacuum chamber wall for both HERA and TRISTAN
AR [3, 15].
During dust trapping investigations at TRISTAN AR
bremsstrahlung observations with γ-ray detectors [15] sug-
gested both longitudinal motion and transverse oscillations
of dust particles.
At CERN a troublesome ion pump installed above an
electrostatic separator was reported [7]. The pump was
found to arc frequently. When the pump was moved below
the separator and around a 90oelbow the problem disap-
peared.
At Super-ACO a CCD camera 10 m from an ion pump
in a beam-line from Super-ACO was found to be “sand-
blasted” with titanium and stainless steel after an accidental
gas inlet [7].
The degree and frequency of beam lifetime disruptions
seems to increase with higher beam energy and beam
current. This correlation has been systematically studied
in HERA-e [9]. Analysis of the correlation reveals that
dozens of particles of radius 0.3 µm are typically trapped
3549
in HERA at current 20 mA and energy 27.5GeV resulting
in a total lifetime reduction from 10 h to about 3 h.
It has been suggested [16] that the likelihood of spon-
taneous ion pump discharge and consequent macroparticle
release is related to the number of photoelectrons released
per metre from the vacuum chamber surface, a quantity ap-
proximately proportional to the product E×Iwhere Eand
Iare the beam energy and current respectively. This de-
fines a locus for onset of the disruptions, although clearly
one is dealing with a stochastic phenomenon not a sharp
threshold. There were no indications of lifetime disruptions
due to macroparticles at the NEG-pump based LEP storage
ring with beam current 8 mA at 45 GeV[19], whereas life-
time disruptions are prevalent at HERA at current 30 mA
at 12 GeV and at current 13 mA at27.5 GeV.
Investigations in PETRA showed that a range of adapted
ion pumps were capable of causing lifetime disruptions
[17], either spontaneously or by abrupt switching of the
pump high-voltage. These pump variations included: (1)
a regular PETRA ion pump where many cylindrical Pen-
ning cells share a common cathode; (2) a regular HERA
ion pump where perforated anode foils offer an open dis-
charge region to the side; (3) a HERA ion pump with a
baffle between the gas conduction slits and the pump an-
ode, blocking the direct route to the beam pipe; (4) an “in-
verted” pump with high negative voltage attached to the
cathode instead of the anode. A “dummy” pump with a
closed, inactive rectangular tube replacing the cylindrical
Penning cells did not give rise to lifetime disruptions.
A possible mechanism for dust particle release from ion
pumps is the liberation of pump material such as titanium
by strong discharges known to occur spontaneouslywithin
the pump cells, and provokable by abrupt switching of the
pump high voltage. Resistors of strength 100 MΩwere in-
stalled before integrated dipolepumps in HERA in the hope
of dissipating the energy of the discharge sufficiently to
prevent liberation of particles [18]. However beam lifetime
disruptions where found nevertheless to occur frequently
and analysis of beam loss monitor responses did not show a
reduction in the frequency of particle release into the beam
pipe.
It has been shown that both strong repeated beam kicks
and a carefully tuned beam excitation sweep with a feed-
back kicker can restore the disrupted beam lifetime of PE-
TRA and HERA [17, 11, 12]. In HERA, where the lifetime
is “multiply-disrupted”, presumably by dozens of disrupt-
ing particles, the lifetime and the rates in particular beam
loss monitors and experiment detectors could be seen to
improve in discrete steps as the parameters of the beam ex-
citation were scanned, providing strong support for the dust
trapping model. Bremsstrahlung detector observations at
PETRA likewise indicated discrete beam disruption.
A clearing field constructed using beam position mon-
itors button electrodes immediately downstream from
HERA’s horizontally defocussing quadrupoles was shown
to deflect the longitudinal flights of particles travelling suf-
ficiently slowly past the clearing field [13]. The location
of the BPMs is not suitable for construction of a clearing
field capable of improving the beam lifetime in HERA by
removing particles trapped within horizontally defocussing
quadrupoles, but the experiment provides further support
for the dust trapping model. The installation of general
clearing-field electrodes to improve the HERA-e beam life-
time is a technically cumbersome option.
5 CONCLUSION
A wealth of experimental and observational information
has been gathered by researchers at many electron storage
rings afflicted by spontaneous beam lifetime disruptions,
most of which is consistent with the model of macroparti-
cles (dust) of size order 1 µm being trapped in the electron
beam.
Yet despite this consistent picture of the symptoms, the
cause is not fully understood. As researchers of viral infec-
tions know, the tiniest creatures can be responsible for the
greatest grievances and warrant the most extensive combat.
An elegant, inexpensivesolution to the electron beam life-
time problem compatible with general machine operation
in HERA is not yet available. Trials indicate that replacing
the ion getter pump system with NEG pumps will likely
restore good electron beam lifetime. But a deeper under-
standing of dust generation/liberation in accelerator pump
systems would be welcome now that so much is known
about dust in electron accelerator beam pipes.
6 ACKNOWLEDGEMENTS
Thanks to DESY colleagues and other electron storagering
colleagues too numerous to list for discussions and commu-
nications.
7 REFERENCES
[1] P. Marin. ‘Microlosses of beam current on Super-ACO oper-
ated with electrons’, LURE, ORSAY RT/90-01 (1990)
[2] H. Saeki, T. Momose, and H. Ishimaru. ‘Observations of dust
trapping phenomena in the TRISTAN accumulation ring and
study of dust removal in a beam chamber’, Rev. Sci. Instrum.
62, 874 (1991)
[3] D. Sagan. ‘Mass and charge measurement of trapped dust in
the CESR storage ring’, Nucl. Instrum. Methods. A330 371
(1993)
[4] F. Zimmerman. ‘Trapped dust in HERA and prospects for
PEP-II’, Technical Report PEP-II AP Note No.: 8-94, (Stan-
ford Linear Accelerator Center, 1994).
[5] D. R. C. Kelly. ‘Characterisation of lepton beam lifetime be-
haviour in HERA’, Technical Report DESY HERA 95-01’,
(Deutsches Elektronen Synchrotron, 1995).
[6] K. Fuke ‘Photon Factory activity report #6’,R-1, 1988
[7] P. Marin. ‘Problems encountered with stray micro-objects in
various ultra-high vacuum systems’, LURE, ORSAY RT/93-
04 (1993)
3550
[8] Bialowons, W., Ridoutt, F., and Wittenburg, K. (1994), ‘Elec-
tron beam loss monitors for HERA’, in Fourth European
Particle Accelerator Conference (EPAC94), Vol. 2, p. 1628,
World Scientific
[9] D. R. C. Kelly. ‘Many-event lifetime disruption in HERA and
DORIS’, Technical Report DESY HERA 95-02, (Deutsches
Elektronen Synchrotron, 1995).
[10] D. Sagan. ‘Some aspects of longitudinal motion of ions in
electron storage rings’, Nucl.Instrum.Meth.A307 171 (1991)
[11] D. R. C. Kelly. ‘The effect of beam excitation on the HERA
electron beam lifetime disruption’. In Fifth European Particle
Accelerator Conference (EPAC96).
[12] D. R. C. Kelly. ‘The effect of transverse beam excitation and
kicking on the HERA electron beam lifetime disruption’. To
be submitted to Nucl. Instrum. Methods.A (1997)
[13] D. R. C. Kelly. ‘The effect of a static clearing field on
trapped dust particles in the HERA electron ring’, To be sub-
mitted to Nucl. Instrum. Methods.A (1997)
[14] K. Kanazawa, (TRISTAN)pers.communication, (1995)
[15] H. Saeki, T. Momose, and H. Ishimaru. ‘Motions of trapped
dust particles around the electron beam in the TRISTAN ac-
cumulation ring’, Rev. Sci. Instrum. 62, 11 (1991)
[16] J. Kouptsidis,‘Lebensdauerprobleme in Elektronenspeicher-
ringen’, in HERA Seminar, Bad Lauterberg DESY (1995)
[17] H. Ehrlichmann,‘Untersuchung zur Elektronenlebensdauer
in PETRA’, in HERA Seminar, St.Englmar DESY (1996)
[18] W. Bialowons,‘HERA Betrieb mit Elektronen’, in HERA
Seminar, Hamburg DESY (1997)
[19] CERN, LEP machine study 10 June 1995: S. Myers,
F. Willeke, J. Kouptsidis, W. Bialowons, H. Ehrlichmann,
K. Wittenburg, D. Kelly.
3551