March 17, 2005A. Boyarski, SLAC Seminar1 New Insights to Wire Chamber Breakdown. By Adam Boyarski...

March 17, 2005 A. Boyarski, SLAC Seminar 1

New Insights to Wire Chamber Breakdown.

ByAdam Boyarski

Stanford Linear Accelerator Center


Motivation For This WorkDuring BaBar commissioning

Background rates were high. Saw high current spikes and steadily increasing currents in

some wires in the drift chamber. These increased currents could trip the HV supplies. This is an old problem, seen in other large detectors. Known solutions for Argon-based gas mixtures -

Add alcohol (Charpak et. al. 1972, Atac 1977), or Add water (Argus expt ~1985).

BaBar added 3500 ppm water vapor – no more spikes to date.

But BaBar uses a Helium-based gas (less known case), Will water solve the problem for the life of the experiment? Will it work at future increased luminosities at PEPII?

1) Need to study long-term and high-rate aging in BaBar gas.


Another Motivation - Some experiments report less dark current or

breakdown problems when using CO2 in the gas mixture. This poses a question, Could dissociated oxygen from CO2 be burning off

deposits? If so, could oxygen alone do an even better job? An oxygen additive in plasma experiments is known to

reduce deposits on cathodes.

2) Study an oxygen additive in a test chamber.


Test Chamber Requirements

Use a cell design similar to that in BaBar. Hexagonal cell, ~1cm wire spacing.

Use BaBar gas - helium(80%) isobutane(20%) but without additives to allow aging effects to proceed.

Use a 55Fe source for ionizing the gas. Pico-ammeter to measure chamber current. Multi channel analyzer to measure the pulse

spectrum shape. Ability to add water, alcohol, oxygen or CO2.


Test Chamber Box 10x10x30(L) cm. One hexagonal cell. 1 cm wire spacing. 1-Sense wire, 20 um,

gold coated, 2050 V. 6-Field wires, 120 um,

gold coated, 0 V. 6-Bias wires, 120 um,

gold coated, 1300 V. Same fields as BaBar.

55Fe source, mylar window, atten. foils.


First Observations As the chamber was run at various source

strengths, the spectrum shape changed with source strength. The 55Fe peak broadens at high rates.

Due to pulse pileup, base line shifts. (OK – can live with it).

The ratio (#small-pulses/55Fe-conversion) varies:

In a new chamber – ratio is constant, but In an aged chamber – ratio changes with ionization

rate. Next plot.


Spectra at Low & High Ionizations.

0

500

1000

MCA Channel0 2000 4000 6000 8000

Cou

nts

0

500

1000

Low Ionization

High Ionization

The number of small pulses per 55Fe conversion is larger at higher ionization levels in an aged chamber.


Understanding small pulses from 55Fe Small pulses may be a useful tool.

But first need to understand the small pulse spectrum. Small pulses are probably due to conversion at the cell edge.

Develop a Blob model.1. Gamma conversion gives ~170 ion pairs in a small (Blob)

volume around conversion point.

2. When a blob straddles a boundary of a cell, electrons in the blob are shared by the cell and its neighbor.

3. Assume blob is a sphere of radius R. 1. Cut sphere into 170 equal-probability slices. 2. Each slice has 1 electron on average. 3. Orient slices parallel to a local cell boundary. See next plot.


r

Cell Boundary

V=1

1-ion2-ions

Conversion sitesproviding

3-ions

Ion drift

Conversion site

Ion volume

Blob Model For Gamma Conversions

2

3

(N ions)


Blob Description Move sphere across a cell boundary, from outside to

inside of cell, marking the locations of the sphere center whenever a slice crosses over the boundary.

The total number of slices that are within the cell give the collected electron count versus conversion location.

The frequency of seeing N electrons is proportional to

the area of conversion sites, equal to the perimeter of the cell times the width of the N’th slice.

Since the first slice is the thickest, the 1-e frequency is the largest, then 2-e, then 3-e etc., as observed.

An ion density distribution should also be considered when calculating the slice widths of the sphere.


Small Pulse 55Fe SpectrumAnd fits to a Blob model

Best fit is with a quadratic density, (max at center to zero at radius R).

R=0.79 mm (1.6 mm diam.)

Consistent with the 2.5 mm range of a 5.9 KeV electron in the (80:20) gas.

Constant density, R=0.86 mm, is too flat.

Gaussian density, sigma=0.19 mm, gives too much of a peak.

Fe55 Peak is at Ch 5500

0

200

400

600

Cou

nts

0 200 400 600 800

Channel Number

Blob Density & R (mm)

DataQuadratic (0.79)Constant (0.86)Gaussian (0.19)


Single Photoelectrons & 1-e Cuts Establish channel

cuts for single electrons.

Shine light through a window to measure a photo-electron spectrum. A Polya function fits

the photo-electron spectrum well.

Channels 30-120 see approximately 50% of the 1-e avalanches.

Channel0 50 100 150 200 250 300

Cou

nts

0

500

1000

1500

2000

2500

3000

35001-Electron Polya Fit

ADC Cutoff

Data

1-e Acceptance Channels


The Spectrum and N1/NFe Ratio Channel cuts for 55Fe

pulses Amplifier gain adjusted for

peak at Ch ~5500. Used channels 3280-7590 to count 55Fe.

The ratio of single-electrons per 55Fe in a new chamber is

0.022 at low rates. 0.033 at high rates (larger

due to pileup, base line shifts).

Any higher values observed must be from aging effects.

Quiz – What’s at ch 2000?

Channel0 2000 4000 6000 8000

Cou

nts

0

50

100

150

55Fe

1-e


Data Acquisition Data was logged using Ortec Maestro

software on a PC running Windows/XP. The software can be configured to loop and

take spectra continually with specified time intervals.

Intervals from 2s – 10min were used. Data was appended to a log-file after each

interval, containing Counts N1, NFe, etc. during the interval, Environmental variables (time, chamber pressure,

temperature). Log-file is saved for each “run”.


Measurements in an Aged Cell. Transient single electron rates were measured in

an aged cell. Starting with a resting chamber (source off, HV on), Source was suddenly opened. Data was recorded until a steady single electron rate

was reached or a breakdown occurred. Repeated for various source strengths.

In each case, the N1/NFe ratio starts at the base-line value, then increases with time – very rapidly at the higher source strengths. Plot.


Single Electron Rates in Aged Cell. At various ionization levels (in nA/cm on SW). “Star” indicates high-current jump occurs, eventually.

Base Ratio

1.3

0.77 nA/cm

0.54

0.35

0.25

He:Isobutane 80:20 Gas

0 100 200 300 400 500 600

Time (Sec)

0.00

0.05

0.10

0.15

0.20

Num

ber

of S

mal

l Pul

ses/

Fe55


Transient Spikes For ionization levels below the breakdown level, when the

source is opened the single electron ratio increases rapidly and then decreases slowly - a transient spike is seen.

0 100 200 300 400 500 600 700 800

Time (Sec)

0.00

0.05

0.10

0.15

# S

mal

l Pul

ses/

Fe55

He:Isobutane 80:20 Gas


Spike Height The spike height was found to depend on the length of quiescent time

(settling time) prior to source turn on, shown in the table.

For HV on & source off, the E field clears out ions in the film in ~0.5 h.

For HV off, only the self field from stored charge pushes charges outward. Charge pushed to the metal side are neutralized there. Charge at the gas side of the film is neutralized by ions in the gas. Settling time depends on the ionization level in the chamber. With source off, the low cosmic ionization gives a long 70 hr settling time.

The charged film behaves like a storage capacitor.

High Voltage

Source Settling Time (h)

ON OFF 0.5

OFF ON 2

OFF OFF 70


Begin Additive Study. Start with an already aged chamber,

Chamber breaks down at low currents when no additive is present.

Add additive (alcohol or water), Measure transient spikes while running at high

ionization levels – but not too high, Restricted to less than 10nA/cm of current on SW to

allow reliable detection of single electrons. BaBar runs at approximately 0.3nA/cm so the test

chamber is at 30 times the level of BaBar.


Additive: 2-Propanol

Concentrations above 0.5% propanol show no small pulse activity.

At 0.25%, a spike is seen at the highest chamber current, 12.5nA/cm.

2-Propanol at 1% is a good additive.

1% Propanol

0.5% Propanol

0.25% Propanol

Small Pulse Response to Step Current 80:20 Gas + 2-Propanol

0.00

0.01

0.02

0.03

0.04

0.05

NS

mal

l / N

Pea

k

0.00

0.01

0.02

0.03

0.04

0.05

NS

mal

l / N

Pea

k

4.4 nA/cm

3.8 nA/cm

0.00

0.01

0.02

0.03

0.04

0.05

NS

mal

l / N

Pea

k

0 200 400 600 800 1000 1200 1400

Time (Sec)

12.5 nA/cm

12

10

7.7


Additive: Water

No spikes seen at highest currents for concentrations of 0.35% (the BaBar value) or half that.

Water at >0.2% is a good additive.

3500 ppm Water

1800 ppm Water

Small Pulse Response to Step Current 80:20 Gas + Water

0.00

0.01

0.02

0.03

0.04

0.05

NS

mal

l / N

Pea

k

I = 11.1 nA/cm

1.4 nA/cm

8.5

0.00

0.01

0.02

0.03

0.04

0.05

NS

mal

l / N

Pea

k

0 200 400 600 800 1000 1200 1400

Time (Sec)


Additive: Methylal

Some spiking seen at the 4% concentration.

Huge spikes and chamber breakdown at the 2% level.

Methylal is not good even at 4%.

2% Methylal

4% Methylal

Small Pulse Response to Step Current 80:20 Gas + Methylal

0.00

0.01

0.02

0.03

0.04

0.05

NS

mal

l / N

Pea

k

8.2 nA/cm

1.5

8.1

1.7

4.0 nA/cm

0 100 200 300 400 500 600 700

Time (Sec)

0.0

0.1

0.2

0.3

0.4

NS

mal

l / N

Pea

k

0.5


Alcohol and Water Summary These additives immediately improve a

damaged chamber.

When the additives are removed, then the small pulses return.

These additives prevent high dark currents and breakdown, but they do not clean or cure the chamber.


Long Term Running With Water The test cell was run with 0.35% water

Use a very high current (40nA/cm on SW, 130×BaBar) For a month duration. Charge collected = 80mC/cm of SW,

about that expected for BaBar’s lifetime. Changguo Lu at Princeton did a similar run to

230mC/cm.

No small pulse activities were measured during these runs.

Conclusion - the water additive should keep BaBar safe over its lifetime. No aging seen at 100× its present current.


Additive: Oxygen Different behavior than water

or alcohol. Training needed. Could only operate at low

currents at first without breakdown, but the current level could slowly be increased to the maximum source level in ~2 hours.

But when oxygen was removed, the chamber could still run at max level. The chamber was cleaned! (Although not to the level of a new chamber).

First such observation of cleansing with O2. Etching was previously seen with a CF4 additive.

Chamber Curing With 500 ppm Oxygen

0 20 40 60 80 100 120

Time (Min)

0

100

200

300

400

500

Ano

de C

urre

nt (

nA)


Table 1. Summary of measurements, showing the maximum stable current Imax(nA/cm) in a damaged

chamber with Helium:Isobutane (80:20) gas before the additive, then with the additive shown, and then after the additive is removed. Cases that did not reach break down at the maximum attempted current are marked with a “>” sign. T is the training time to reach Imax. Some additives cure a damaged chamber,

as indicated in the last column.

Additive (%)Before With Additive After

Cure?Imax T(hr) Imax Imax

Methylal 4 2

0.3 ~0~0

>83

0.4No

Propanol

1.00.5

0.25

0.7 ~0~0~0

>12>10>13

0.2

No

H2O 0.35 0.18

0.4 ~0~0

>27>9

0.5

No

O2 0.10 0.5 1.5 >32 >40 Yes

0.05 0.4 2 >29 >16 Yes

0.02 0.9 10 >35 >14 Yes

CO2 5 0.4 35 >40 >27 Yes

O2 + H2O(0.05%+0.35%)

0.4 40 10 3 Partly


Done. But A New Motivation. Done with the original mission of studying

additives for BaBar. But what is the mechanism for those large

single electron rates? Motivation 3), understand those single

electron rates and the breakdown mechanism.

Need a model. Could wire chamber breakdown be related to

breakdown in accelerator structures?


Modeling

Start with two known phenomena.1. Malter (1936) showed that a thin insulating film on a

cathode, charged positively, produces a high electron current from the cathode due to a high E-field in film.

2. Fowler-Nordheim (1928) used quantum mechanics to derive the field emission current from a metal surface in the presence of a high external E-field.

Link the two together to try to explain the observed single electron rates. (Malter didn’t do this).

Assume +ions neutralize and accumulate on cathode and grow a thin polymer (insulating) film over a period of time.

New +ions continually charge up the thin film and establish a high E-field before the charge dissipates.


Thin Film Field Emission Cell field E0 in gas.

Positive ion current I (A/m2)

+ions collect on film, -ve induced charge opposite.

High E field in thin film.

Triangular potential barrier for Fermi electrons in cathode (work function W).

Fermi electrons tunnel through barrier (field emission), through film to gas and avalanche at SW, producing 1-e avalanches.

Potential

-W

0

Insulating

Thin Film

E0E

_

__

Gas

x

e

Conducting Cathode

+

++

++

__


Field Emission Current JFE

Fowler-Nordheim derived equation for JFE(E). Good-Muller modified the calculation for a rounded tip potential. The latter is used here.

JFE depends on work function W (4.3eV for gold). Modify E → βE (β>1) to account for the increases

in E due to localized surface bumps – as experienced in RF accelerating cavities.

JFE (in A/m2) is:

105.43 10

255.4 10 EFEJ E e


JFE Graph As Function of βE A very rapid rise in J with βE.

E (V/m)

J FE (

A/m

2 )

1010

108 10

9 10

11

10-5

1010

100

105

10-10

1015


Field On Cathode Wire

So if we know βE, can calculate J. But what is E? E depends the (unknown) discharge rate of ions in the

film as well as the known charge-up rate from ionization.

Assume a leaking-capacitor (resistive) model, Over a very small surface area on cathode wire, the film

behaves like a parallel plate capacitor, with Dielectric const ε (=kε0), resistivity ρ, thickness d, area A. Time constant is ρε (=ρd/A×εA/d = RC).

Also known, any charge Q on or in the film produces a field E=Q/ε at the wire surface (Gauss law).


E(t) For Resistive Model Can derive the field E(t) on the wire

surface for a step increase I in current density on the film at t = 0,

Note – no dependence on the film thickness d.

/0 1)( teIk

EtE 0t

,


More Parameters – Area, Feedback, Absorption Area – η, the surface area on the cathode producing field

emission.

Feedback – A field emitted electron that reaches the anode will avalanche and produce additional G~105 positive ions.

Absorption – σ, field emitted electrons passing through the film can recombine with positive ions on the film.

The feedback and absorption parameters make an analytical solution difficult. Use a numerical solution (computer program) instead.


Resistive Model Fitting Model uses small time steps (dt=0.2 s). Loop over t to span the measurement times,

At each step, increment surface Q by+I55 * dt (source ionization current density)

-Iρ * dt (resistive leakage current density)+GηJ * dt (feedback) -recombination.

Compute new E, J, Iρ Calculate R1 (= 0.022 + GηJ/I55), record R1 and t. Bump t and loop some more.

Do so for each I55 value taken. Fit all the data to one common set of

parameters ρ, ε, β, η, σ. Next plot.


Resistive Model Fit For A Very Aged Cell.

2.1

2.4

Isw=2.9 nA3.76.413.6

T (sec)0 50 100 150 200 250

N1

/ NFe

0.0

0.2

0.4

0.6

0.8

1.0DataFit


Fit Parameters (MKS units)

Fit results: ε = 1.3ε0 , ρ = 4×1012 , β = 136 , η = 4×10-25 σ = 1.5×10-36 The large β and very small (unphysical) η are

indicative of a very high field, very localized area of field emission.

ε and ρ values are reasonable for polymers.


Spike Modeling. Resistive model does not give a spike, i.e. rise and fall in E.

Need a method that holds a burst of charge and bleeds it away slowly,

Can do so by limiting discharge/passage of J current through film with second resistive parameter ρJ for these electrons.

As E increases and an E threshold is reached, there is a burst of electrons which then discharge slowly. This negative electron charge reduces the E field at the cathode surface, thus limiting further J production and settling to some equilibrium value eventually.


Spike Fit – Both ρI & ρJ Model

T (Sec)0 50 100 150 200

N1

/ NFe

0.0

0.1

0.2

0.3

0.4

0.5

DataFit


Problem With Resistive Model Although the resistive model fits transient data for an aged

chamber fairly well, it has a problem. Model is not a function of film thickness. It predicts that a new chamber (d≈0) should look like an aged

one, which is incorrect.

Need to add another discharge mechanism that removes charge only when the film is thin.

Mobility provides such a mechanism. For thin films the transport time through the film is shorter

than the resistive ρε time constant, so mobility dominates in the discharge process.

For thick films, the transport time is longer than the resistive time, so the resistive mechanism dominates.


Resistivity + Mobility (ρμ) Model

dx

I

kEEI

k

E

k

EE

/1ln 000

For field E(x) within the film material at depth x, and a current density i(x) flowing over time dt with velocity v=μE (mobility constant μ), two coupled differential equations,

The increase in E due to i is dE = idt/ε = -idx/(εμE). The resistive decrease in i is di = -dE/ρ = idx/(ρεμE). These can be solved as

I is the external ionization current density hitting the film. Setting x=0 gives the steady state flow when the mobility driven ions traverses the full film thickness d.

In this model E depends on d, (actually d/μ for steady state, can’t separate).


E(d) From ρμ Model

d (10-9 m)

0 20 40 60 80

E (

106 V

/m)

0

200

400

600

800

The parameter values (ρ,ε,μ,I) used in the plot are shown later (from a fit).

E reaches a plateau of 0.6×109V/m at large thickness values (βE is 1.2×109V/m, with β=2 in the fit)

Need more aging data of single electron rate versus d to test this model.


Aging A New Chamber To Death Use a newly wired chamber. Run with a high source level

and normal HV. Monitor the current (a) and

the single electrons (b). HV was bump in day 7. Pause periodically for checks. See 1-e spike on restart (c). See self sustaining current in

day 24 (QSW=70mC/cm). Single e- seen much earlier,

in day ~10 (QSW=20mC/cm) with rapid rise in day 24.

Aging Time (day)11.463 11.465 11.467

N1

/ NF

e

0.0

0.1

0.2

0.3

Cur

rent

(nA

)

0

200

400

600

800

Aging Time (day)0 5 10 15 20 25

N1

/ NF

e

0.0

0.2

0.4

0.6

(a)

(b)

(c)

SSFE


Other Fitting Parameters Need another parameter α relating d to QSW, the

accumulated charge on the SW. d = αQSW

Also add a feedback parameter f, for the fraction of fed back ions that return to the position of the original field emitted electron, since scattering and the avalanche process smear the trajectories.

The previous σ parameter was dropped (it didn’t help). Total parameters are ε, ρ, β, η, μ, α, and f. Too many to fit,

so fix some parameters. Set ε = 2.3ε0, the value for polyethylene. Set β = 2, and α=1.0×10-6 (1 micron for 1 C/cm-1 on SW). Do a dt stepping fit for remaining parameters. Next plot.


Fit of ρμ Model To Aging Data

Aging Time (day)0 5 10 15 20 25

N1

/ NFe

R

atio

0.0

0.2

0.4

0.6

0.8Parametersk = 2.3 = 2.0

= 8.5x10-4

m2

= 1.192x1011

-m

= 8.3x10-18

m2V

-1s

-1

=1.0x10-6

m-C-1

cm

f = 3.46x10-5


Comments on ρμ Fit. This model reproduces the data very well.

But good starting values are needed for the fit to converge. Small iteration changes to the parameters can easily lead to sudden large J-currents, and the fit wanders away without converging.

Solution shown here may not be unique because of the manual starting values.

But the model establishes the mechanism for chamber breakdown, even if the fitted parameters may be off.


Comments The steady state single electron rate in a

chamber provides an indicator for the wellness of a chamber. In this test chamber, the chamber ran 3.5 times longer

after single electrons were first observed. But it is necessary to have steady ionization currents to

make the measurement (hours). Transitory spikes can also be used.

e.g. Does water suppress the polymer build up? The chamber that was aged for 25-days without water

was compared with a another new chamber aged with 0.35% water for the same running time (and a 2-month dry out time). Next plot.


Aging With Water Spike height for “Aged

with H2O” is approx. the same as “Aged without H2O” at half the accumulated charge.

Suggests that H2O does not prevent film growth, but maybe slows the rate down.

(0.080 C/cm)

T (sec)

0 100 200 300 400 500 600N

1 /

NFe

0.0

0.1

0.2

0.3

0.4

0.5

0.6 AgedWithout H2O (C/cm)

0.0

0.038

0.050

0.070

AgedWithH2O


Conclusion The water additive can keep the BaBar chamber

alive for its lifetime and at much higher rates. An oxygen (and CO2) additive provides a cathode

cleaning effect, to some extent. Breakdown in an aged chamber comes from:

Ion collection/neutralization, which builds an insulating polymer film on the cathode,

New ions provide a transitory charge on/in the film, The resulting >109V/m E-field causes FN field emission, These electrons avalanche and feed back more ions, Leading to self-sustaining field emission (SSFE).


Breakdown can be modeled.

Need ρ, μ discharging in general case, Only ρ needed for very aged chambers.

This model applies to chambers with gases other than that used here, and also other gas-devices (MSGC, GEM).

Aged chambers should be turned on slowly to allow the transient field emission spike to decay (~minutes) to prevent SSFE from occurring.


More information

The data presented here is published in:

NIM A515 (2003) 190-195, for the additive study.

NIM A535 (2004) 632-643, for the modeling study.

March 17, 2005A. Boyarski, SLAC Seminar1 New Insights to Wire Chamber Breakdown. By Adam Boyarski...

Documents

Transcript of March 17, 2005A. Boyarski, SLAC Seminar1 New Insights to Wire Chamber Breakdown. By Adam Boyarski...