Failure modes of resistive plate chambers
Outline
• Resistive Plate Chamber (RPC) operation
• Mechanical tolerances
• Failures due to resistivity changes – eg Oil
bridges
• Aging in new production BaBar RPCs
• Malter effect
• Water in glass RPCs
• Conclusions
LC Santa Cruz
1
David Strom – UO
RPC operation
• Number of electrons at the
head of shower is given by
Muon
pads
ne = eα`
- HV
Bakelite
where α is the Townsend coefficient (depends on gas and E)
and ` is the shower length
Ar
-
Ar
α` ' 20
Ar
-
⇒ ne = 5 × 108
-
-
- -
40,000 V/cm
-
Ar
Ar
-
Ar
-
-
Bakelite
• Streamer is limited in part by
the high resistivity of the bakelite
LC Santa Cruz
-
-
• Streamer mode (space charge
dominated discharge) occurs
when
E=
Ar
pads
2
David Strom – UO
Typical gas mixture
• Argon to provide for efficient gas amplification
• Isobutane (or another hydrocarbon) to absorb UV photon
• Freon ( e.g. 134a , C4H2F4 ) ”quench gas”, controls charge and
physical size of streamers
• The detectors will operate over a very wide range of these gases.
• The Isobutane fraction can be as low as 4%
Caution: flammable mixtures easily produced, especially at low 134a
fractions!
• Streamer production relatively tolerant to N2, O2 and H2 O contamination
LC Santa Cruz
3
David Strom – UO
• The ratio of Ar/134a can vary
from 10 to 0.25
• Streamer charge and size
(area is in mm2) increase with
Ar fraction.
2001 NSS, Onodera, et al.
•
Charge
distributions
of
streamers is relatively narrow
All Strips
6000
4000
2000
0
• Fraction of double streamers
small
LC Santa Cruz
100
150
of
in
200
300
400
Strip 10
50
200
200
400
600
800
Triplet charge (pC)
Strip 20
0
700
800
900
Triplet charge (pC)
Strip 15
200
1000
400
600
800
Triplet charge (pC)
Strip 25
750
600
400
500
200
250
0
600
600
400
800
4
500
100
0
•
Charge
distributions
avalanches
exponential
parallel plate geometry
0
200
400
600
800
Triplet charge (pC)
0
200
400
600
800
Triplet charge (pC)
David Strom – UO
Bakelite (or glass) resistivity controls time needed (typically milliseconds) to rebuild field after a streamer occurs
In BaBar bakelite was required to have
ρ = 28 − 120 × 1010Ωcm
at 20◦ C. Resistivity of bakelite varies substantially with both humidity and temperature. Higher resistivities can be used for cosmic ray
detectors.
The temperature effect is large:
∆ρ/ρ ∼ −10%/ ◦C
It is speculated that at high temperature streamers lower values of ρ
can lead to large discharges and significant aging of the detectors.
LC Santa Cruz
5
David Strom – UO
Mechanical Tolerances
•
Townsend
coefficients
rapidly increase with electric
field (from Imonte simulation)
• If gap width increased,
Townsend coefficient decreases faster than streamer
length ` increases
• Chamber becomes inefficient when α` < 20
• This analysis courtesy of
C. Lu, Princeton
LC Santa Cruz
6
David Strom – UO
Basic result:
dV
' 2300V/mm
d gap
In Babar a few ”popped buttons” (unglued spacers) can easily lead to
a 3mm gap width rather than the nominal 2 mm width.
• To avoid excess aging chambers should be kept no more than 500 V
above streamer threshold
⇒ mechanical tolerance of only 200 µm
LC Santa Cruz
7
David Strom – UO
Problems associated with linseed oil coating
• Linseed oil coatings of inner surface lower the current drawn through
the gas and singles of rates of the detectors by a factor of 5 to 10.
• The linseed oil is thought to provide two functions:
• It makes a smooth inner surfaces leading
to a more uniform electric field
• It can absorb UV photons produced in the
avalanche
• Main advantage of glass RPCs is that they avoid this coating
LC Santa Cruz
8
David Strom – UO
Efficiency History
Babar problems
Possibly due to linseed oil
bridges
Barrel
1
Average RPC Efficiency
0.8
0.6
RPCs with eff ≥ 10%
All RPCs
0.4
• Temperature rose to 36◦ C in
the experimental hall
• Currents increased
⇒ Many chambers temporarily
disconnected
• Efficiency can be increased by
lowering amount of Freon
⇒ See 200 and 420 days
• But efficiency still declines continuously
LC Santa Cruz
9
0.2
0
100
June
1999
200
300
Jan.
2000
400
500
July
700
Jan.
2001
Forward Endcap
1
600
800
July
0.8
0.6
0.4
0.2
0
100
June
1999
200
300
Jan.
2000
400
500
July
600
700
Jan.
2001
800
July
Backward Endcap
1
0.8
0.6
0.4
0.2
0
June
1999
100
200
Jan.
2000
300
400
July
500
600
Jan.
2001
700
800
July
Henry Band
David Strom – UO
Efficiency Maps
• Inefficiency appears to be mainly
concentrated around edges of
the chambers
• There is some evidence that the
efficiency also occurs near the
rows of spacers
• High voltage plateau’s become
very broad
LC Santa Cruz
10
David Strom – UO
Efficiency Plateaus
During original testing
LC Santa Cruz
After operation in BaBar
11
David Strom – UO
Test Stand Studies
• Can we reproduce the problems
in the lab?
• SLAC test stand shows that
trigger chambers made prior to
the BaBar production are sensitive to heat.
• Other tests (e.g. at Oregon)
show that damage can be done
to chambers at temperatures of
only 28◦ C
⇒ Problems could occur even
at moderate temperatures!
LC Santa Cruz
12
David Strom – UO
Materials Studies and Models
• Effects of linseed oil columns will
depend on the resistivity of the linseed
oil:
HV
Bakelite
Bakelite
⇒ The resistivity of linseed oil
depends on how it has cured and
if contaminants are present
R bak
Sample
R gap
resistivity
[109 Ωcm]
Vgap
R bak
• A model of high and low resistivity
linseed oil columns:
polymerized US linseed oil
(skin/oil mix)
145.9
US linseed oil
(cured in air for 30 days)
42.3
US linseed oil
(cured in air for 3 days)
27.9
uncured US linseed oil
14.4
uncured linseed
(production oil)
7.7
uncured oil
(removed from bad RPC )
measurements from SLAC
0.21
and
Princeton
Unclear why oil removed from bad
RPC has so low resistivity
LC Santa Cruz
13
David Strom – UO
Experience with prototypes for endcap replacement
Efficiency (no beam) for layer 18
• 24 endcap modules (12 chambers) were
prototypes
• The prototype chambers have a single
coat of 30% linseed oil, 70% pentane.
• Inner surface of opened chambers
Efficiency
Efficiency
replaced with prototypes 12/00
1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
smooth
• Some damage seen in one of two chambers heated in test stand
⇒ Thinner linseed oil surface more sensitive to dust, contamination
• Modules in the shallow layers of the
0
detector have stable, good efficiency
100
200
Days of operation
0
100
200
Days of operation
• Modules in the deepest layer of the calorimeter show significant
damage after ∼ 120 days of operation.
LC Santa Cruz
14
David Strom – UO
• The layer 18 prototypes were exposed
to high levels of background from beam
Current (microamps)
processes.
• Since detailed monitoring began, the
charge through the gas has grown linearly
with time.
•
The
350
Measured current
300
Predicted current
250
decline
in
efficiency
started
200
at about 120 days corresponding to
150
∼500
C/m2
(∼
108
streamers/cm2 )
in
100
the gas.
• A model which takes the temperature
of the leakage current into account and
which assumes that
50
0
140
Ileakage ∝ Qgas
160
180
200
220
240
260
280
Day of operation
Predicted and measured current
at injection.
describes the data well.
⇒ Can this model explain the
decline in efficiency?
LC Santa Cruz
15
David Strom – UO
Voltage 7250 V
current in chamber 6 (microamps)
• Water vapor (70% relative
humidity at 20◦) was added
to the gas of test stand
chamber 6 on day 528. Rate
was nominally 1 cm3/min,
but was much lower for
chamber 6 because it is
somewhat leaky.
100
80
60
40
20
high rate of water
0
Efficiency
• On day 529 a high rate
of gas was flown through
chamber 6
(flow rate off-scale on flow
meters, ∼ 15 cm3/min)
524
526
528
530
532
534
536
days
532
534
536
days (Oct 1 = 1)
1
water added to gas
high rate
Ch 4
0.8
Ch 5
Ch 6
0.6
• Current immediately decreased in 6
0.4
0.2
• Efficiency
immediately
improved in 6
LC Santa Cruz
0
524
16
526
528
530
David Strom – UO
Discussion
The observed behavior of chamber 6 is consistent with the Malter effect:
- HV
Graphite
Bakelite
Oil
Insulating
skin
e-
e+
Ar
+
+
Ar
Ar ions
Ar
E
Ar
-
-
• Chamber current locally depletes charge carriers in linseed oil skin
LC Santa Cruz
17
David Strom – UO
• Ions collect on the insulating linseed oil surface
• Accumulated ions will produce a large electric field across the linseed
oil surface
• Electrons can then be accelerated into the gas volume where avalanches
are produced (Malter Effect)
• The large current from Malter electrons keeps the gap voltage below
streamer threshold. A large current and inefficiency is observed
LC Santa Cruz
18
David Strom – UO
• Adding water vapor to the gas decreases the surface resistivity of
the linseed oil and prevents the accumulation of ions
• The Malter Effect also explains a common phenomena observed
with many chambers: when the chambers are first switched on their
efficiency decreases and the current increases
• The increased current occurs as the ions collect on islands of insulator on the linseed oil surface causing the Malter Effect
• As the chambers become drier, these islands become larger due to
the depletion of ion conductivity (see Jerry’s Notebook).
• On 2 of 3 chambers tested, the water had no effect
LC Santa Cruz
19
David Strom – UO
Glass RPCs
• ”Float glass” has resistivity of roughly 1012 Ω cm, comparable to the
higher resistivity bakelite
3
12
Volume resistivity (10 W cm)
10
10
2
10
1
10
-1
-10
0
10
20
30
40
50
60
o
Temperature ( C)
Hoshi, et al
LC Santa Cruz
20
David Strom – UO
• In the Belle experiment, it was found necessary to control any mosture
in the glass very tightly. Reportedly
- Water can combine with Fluorine which forms in the
streamers to produce HF
-HF can etche the glass allowing for the adsorption of
water onto the glass etch
-The water forms a conducting layer which ”shorts” the
surfaces to nearby spacers, reducing the gap voltage below streamer threshold.
LC Santa Cruz
21
David Strom – UO
Gas without freon 134a can be used (Hoshi et al.,) eg:
percent butane
4% isobutane, 10% O2, 10% Ar and 76% CO2 has 90% efficiency
instead of 95% for freon based mixtures.
Caution: a simple analysis based
on adiabatic flame temperatures
and complete combustion indicates that this mixture may still
be flammable.
10
CO2
9
8
7
0.76 CO2+ 0.10 Ar+ 0.10 O2
6
5
N2
4
3
2
1
0
LC Santa Cruz
22
0
10
20
30
40
50
60
70
80
fraction inert+02
David Strom – UO
Conclusion
• Results are mixed for large scale deployment of RPC
• Detectors are relatively inexpensive, but are not ”easy to build” –
careful QA/QC needed during production
• Double gap chambers are more robust against failure
• Must be able to replace faulty RPC chambers during the lifetime of
the experiment.
LC Santa Cruz
23
David Strom – UO