Development of m icromegas for the ATLAS Muon System Upgrade

Development of micromegas for the ATLAS Muon System Upgrade

Jörg Wotschack (CERN)

MAMMA CollaborationArizona, Athens (U, NTU, Demokritos), Brandeis, Brookhaven, CERN, Carleton, Istanbul (Bogaziçi, Doğuş), JINR Dubna, MEPHI Moscow, LMU Munich, Naples,

CEA Saclay, USTC Hefei, South Carolina, Thessaloniki

Work performed in close collaboration with CERN/TE-MPE PCB workshop (Rui de Oliveira)Lots of synergy from the RD51 Collaboration

2

Outline The ATLAS Muon System upgrade The micromegas technology Making micromegas spark-resistant Performance & ageing studies Two-dimensional readout structure Towards large-area micromegas chambers First data from ATLAS Next steps

PH Det. seminar, 18 Nov 2011 Joerg Wotschack (CERN)

Not covered: electronics

The MAMMA*) R&D activity Using the micromegas technology to build muon chambers for the ATLAS

upgrade was first suggested in an ATLAS Muon Collaboration brainstorming meeting back in 2007 by I. Giomataris (CEA Saclay) By this time MMs had been successfully used in several experiments at very high

rates (COMPASS, NA48) but the largest chambers did not exceed 0.4 x 0.4 m2

The idea looked intriguing to some of us Profiting from the know-how of the CERN PCB workshop, the first prototype

chamber, 0.4 x 0.5 m2 in size, was built still in 2007; it worked very nicely By 2009 the excellent performance of MMs and their potential for large-area

muon detectors was demonstrated 2010 was dedicated to making micromegas spark resistant 2011 the first resistive large-area chambers successfully built and tested

PH Det. seminar, 18 Nov 2011 Joerg Wotschack (CERN) 3

*) Muon ATLAS MicroMegas Activity

Joerg Wotschack (CERN) 4

The ATLAS Muon System Upgrade

PH Det. seminar, 18 Nov 2011


Count rates*) in the ATLAS Muon System at √s = 14 TeV for L = 1034 cm-2s-1


Small WheelRates in Hz/cm2

Rates at inner rim 1–2 kHz/cm2

*) ATLAS Detector paper, 2008 JINST 3 S08003


The ATLAS detector


Small Wheels


Rates: measurement vs simulation


Measured and expected count rates and in the Small Wheel detectors Data correspond to L = 0.9 x 1033 cm-2s-1 at √s = 7 TeV

Rate in Hz/cm2 as a function of radius in the large sectors

Ratio between MDT data and FLUGG*)

simulation (CSC region not shown)*) FLUGG simulation gives rates about factor 1.5 – 2 higher than old simulation


Three reasons for new Small Wheels

Small Wheel muon chambers were designed for a luminosity of L = 1 x 1034 cm-2 s-1

The rates measured today are 2–3 x higher than estimated; all detectors in the SW will be at their rate limit at 5 x 1034 cm-2 s-1

Eliminate fakes in high-pT (> 20 GeV) triggers

At higher luminosity pT thresholds of 20-25 GeV are a MUST Currently over 95 % of forward high pT triggers are fake

Improve pT resolution to sharpen thresholdsNeeds ≤1 mrad pointing resolution



The problem with the fake tracks


Current LVL1 end-cap trigger Only the vector BC at the Big Wheels is

measured Momentum defined by assumption that track

originated at IP Random background tracks can easily fake this Currently 96% of forward high-pT triggers (at

LVL1) have no track associated with them

Proposed LVL1 trigger Add vector A at Small Wheel Powerful constraint for real tracks A pointing resolution of 1 mrad will also

improve pT resolution

Small Wheel performance requirements

Rate capability 15 kHz/cm2 (L ≈ 5 x 1034 cm-2s-1) Efficiency > 98% Spatial resolution ≈100 m (Θtrack< 30°) Good double track resolution Trigger capability (BCID, time resolution ≤ 5–10 ns)

Radiation resistance Good ageing properties

Joerg Wotschack (CERN) 10PH Det. seminar, 18 Nov 2011

All requirements can be met with micromegas

11

2.4 m

ATLAS Small Wheel upgrade proposal*)


CSC chambers

Today:MDT chambers (drift tubes) +TGCs for 2nd coordinate (not visible)

Replace the muon chambers of the Small Wheels with 128 micromegas chambers (0.5–2.5 m2) Combine precision and 2nd

coord. measurement as well as trigger functionality in a single device

Each chamber comprises eight active layers, arranged in two multilayers a total of about 1200 m2 of

detection layers 2M readout channels

*) other candidates: combined systems sMDT+TGC and sMDT+RPC


The micromegas technology


-800 V

-550 V

Micromegas operating principle Micromegas (I. Giomataris et al.,

NIM A 376 (1996) 29) are parallel-plate chambers where the amplification takes place in a thin gap, separated from the conversion region by a fine metallic mesh

The thin amplification gap (short drift times and fast absorption of the positive ions) makes it particularly suited for high-rate applications


The principle of operationof a micromegas chamber


Conversion & drift space

MeshAmplificationGap 128 µm

(few mm)


Micromegas characteristics Drift region of a few mm with

moderate electric field of 100–1000 V/cm, function of the gas

Narrow amplification gap with high electrical field (40–50 kV); typically a factor 70–100 is required to make the mesh fully transparent for electrons

Charged particles ionize the gas in the drift region, the electrons move with typically 5 cm/µs (or 20 ns/mm) to and through the mesh and are amplified in the amplification gap

By measuring the arrival time of the signals a MM functions like a TPC



An established technology Micromegas were (and are) used successfully in the

COMPASS experiment as vertex chambers in a high-rate muon beam 12 planes with dimensions 400 x 400 mm2

Strip pitch: 360 µm; 1024 channels/plane Superb performance in high rates (>100 kHz/cm2) Spatial resolution ≈70 µm; time resolution 9 ns

T2K near-detector TPC readout 72 chambers of 350 x 360 mm2

‘Mass production’ using bulk micromegas technique



Pillars ( ≈ 300 µm)

The bulk-micromegas* technique

PCB

Photoresist (64 µm)r/o strips

Mesh

*) I. Giomataris et al., NIM A 560 (2006) 405

The bulk-micromegas technique, developed at CERN, opens the door to industrial fabrication



Bulk-micromegas structure


Standard configuration Pillars every 2.5 – 10 mm Pillar diameter ≈300 µm Dead area ≈1–2% Amplification gap 128 µm Mesh: 350 wires/inch Wire diameter 20 µm

Pillars (here: distance = 2.5 mm)


Early performance studies All initial performance studies were done

with ‘standard’ micromegas chambers We used different gas mixtures, initially

Ar:CF4:iC4H10 (88:10:2, 95:3:2), moved later to Ar:CO2 (80:20, 85:15, 93:7)

The chambers were read out with the ALTRO chip (up to a maximum of 64 channels) and the ALICE Date readout system (Thanks to H. Muller and ALICE!)

End 2010 we switched to a new readout electronics (APV25, 128 ch/chip) and a new ‘Scalable Readout System’ (SRS) developed by H. Muller et al. in the context of RD51


P1 (40 X 50 cm2), H6 Nov 2007

2008: Demonstrated performance


Standard micromegas (P1) Safe operating point with

efficiency ≥99% Gas gain: 3–5 x 103

Very good spatial resolution

250 µm strip pitch

σMM = 36 ± 7 µm

Ar:CF4:iC4H10 (88:10:2)

(MM + Si telescope)

X (mm)

y (m

m)

Inefficient areas

Conclusions by end of 2009 Micromegas (standard) work

Clean signals Stable operation for detector gains of 3–5 x 103

Efficiency of 99%, limited by the dead area from pillars Required spatial resolution can easily be achieved with strip

pitches between 0.5 and 1 mm Timing performance could not be measured with the ALTRO

electronics However: sparks are a problem for LHC operation

Sparks lead to a partial discharge of the amplification mesh => HV drop & inefficiency during charge-up

The good news: no damage on chambers, despite many sparks



2010: Making MMs spark resistant Tested several protection/suppression schemes

A large variety of resistive coatings of anode Double/triple amplification stages to disperse charge, as

used in GEMs (MM+MM, GEM+MM) Settled on a protection scheme with resistive strips Tested the concept successfully in the lab (55Fe source,

Cu X-ray gun, cosmics), H6 pion & muon beam, and with 2.3 MeV and 5.5 MeV neutrons



The resistive-strip protection concept



Large number of resistive-strip detectors tested

Small chambers with 9 x 9 cm2 active area

Large range of resistance values Number of different designs Gas mixtures

Ar:CO2 (85:15 and 93:7)

Gas gains 2–3 x 104

104 for stable operation


R16

R16, first chamber with 2D readout

24

Small resistive-strip detectorsChamber RGND

(MΩ)Rstrip

(MΩ/cm)Readout coord.

(NR:Nro)Strip pitch

(µm)R11 15 2 x (1:1) 250

R12 45 5 x (1:1) 250

R13 20 0.5 x (1:1) 250

R14 100 10 x (1:1,2,3,4,72) 250

R15 250 50 x (1:1,2,3,4,72) 250

R16 55 35 x-y 250

R17a,b 100 45 x-y 250 Used for ageing tests

R18 200 100 x-y 250

R19 50 50 xuv 350/900/900 Mesh & GEM

R20 80 25 x 250 +HV on strips

R21 250 150 x-y 500/1000 +HV on strips

MBT0 100 100 x-u (x-v) 500/1500 2 gaps/+HV on strips



Detector response


Sparks in resistive chambers Spark signals (currents) for resistive chambers are about a factor 1000 lower than for

standard micromegas (spark pulse in non-resistive MMs: few 100 V) Spark signals fast (<100 ns), recovery time a few µs, slightly shorter for R12 with strips with

higher resistance Frequently multiple sparks



Performance in neutron beam

Standard MM could not be operated in neutron beam

HV break-down and currents exceeding several µA already for gains of order 1000–2000

MM with resistive strips operated perfectly well,

No HV drops, small spark currents up to gas gains of 2 x 104


Standard MM Resistive MM


Spark rates in neutron beam (R11)

Typically a few sparks/s for gain 104

About 4 x more sparks with 80:20 than with 93:7 Ar:CO2 mixture

Neutron interaction rate independent of gas

Spark rate/n is a few 10-8 for gain 104

Larger spark rate in 80:20 gas mixture is explained by smaller electron diffusion, i.e. higher charge concentration



Sparks in 120 GeV pion & muon beams

Pions, no beam, muons Chamber inefficient for O(0.1s)

when sparks occur

Stable, no HV drops, low currents for resistive MM

Same behaviour up to gas gains of > 104


Gain ≈ 104Gain ≈ 4000 8000


Spatial resolution & efficiency in hadron and muon beamsS3 and R12 (250 µm strips)

Analysis of data taken in July 2010 (by M. Villa)


Spatial resolution as function of gain for S3 (non resistive) and R12 with resistive-strip protection

Efficiency measured in H6 pion and muon beam (120 GeV/c); S3 is a non-resistive MM, R12 has resistive-strip protection

Resistive strip chambers fully efficient in a wide gain windowResolution improves with resistive strips


R11 rate studies


Clean signals up to >1 MHz/cm2,but some loss of gain

Gain ≈ 5000


Test beamNov 2010


Active area9 x 9 cm2

Four chambers with resistive strips aligned along the beam

NEW: Scaleable Readout System (SRS) – RD51 (H. Muller at al.)

APV25 hybrid cards


APV25 data



R11

R12

R13

R15

Tim

e bi

ns (2

5 ns

)

Char

ge (2

00 e

- )

Strips (250 µm pitch) Strips (250 µm pitch)

ve ≈ 4.7 cm/µs


R11

R12

R13

R15

Delta ray

Tim

e bi

ns (2

5 ns

)

Char

ge (2

00 e

- )


Inclined tracks (40°) – µTPC


R12

R11

Tim

e bi

ns (2

5 ns

)

Char

ge (2

00 e

- )

ve ≈ 2 cm/µs


… and a two-track event


R11

R12 Tim

e bi

ns (2

5 ns

)

Char

ge (2

00 e

- ) ve ≈ 2 cm/µs


Ageing



Ageing studies

Micromegas have shown excellent ageing properties in the past, no long-term effects were observed

However, no results are reported for MMs with resistive coating

We have started an ageing test campaign at CEA Saclay that involves high-rate exposure of resistive chambers to few MeV X-rays and thermal neutrons



Long-time X-ray exposure 1 A resistive-strip MM (R17a) has

been exposed at CEA Saclay to 5.28 keV X-rays for ≈12 days

Accumulated charge: 765 mC/4 cm2

Expected accumulated charge at the smallest radius in the ATLAS Small Wheel: 30 mC/cm2 over 5 years at sLHC

No degradation of detector response in the irradiated area (nor elsewhere) observed; rather the contrary (still to be understood)



Long-time X-ray exposure 2 In a second measurement the same chamber (R17a) was exposed again to the X-ray

source, irradiating a non-irradiated area of the chamber. In parallel an ‘identical’ chamber (R17b) was measured without being irradiated continuously. Exposure time: 21.3 days

Accumulated charge: 918 mC/4 cm2



Exposure to thermal neutrons R17a was then moved to the Orphee reactor at Saclay and has been under radiation from 7

– 17 Nov 2011 Neutron flux is ≈0.8 x 109 n/cm2/s with energies of 5–10 x 10-3 eV Total exposure on-time: 40 hrs equivalent to ≈20 years LHC at 5 x 1034 cm-2s-1 (including a

safety factor of 3) Exposure finished yesterday evening Detector response perfectly stable over full duration of irradiation



Two-dimensional readout &

other structural issues



2D readout (R16 & R19) Readout structure that gives two readout coordinates from the same gas gap;

crossed strips (R16) or xuv with three strip layers (R19) Several chambers successfully tested


x strips: 250/150 µm r/o and resistive strips

y: 250/80 µm only r/o strips

PCB

Mesh

Resistivity valuesRG ≈ 55 MΩRstrip ≈ 35 MΩ/cm

Resistive strips

x strips

y stripsR16


R16 x-y event display (55Fe γ)


R16 x

R16 y

Char

ge (2

00 e

- )

Tim

e bi

ns (2

5 ns

)


R19 with xuv readout strips

x strips parallel to R strips u,v strips ±60 degree


R strips v stripsu stripsx strips

Mesh

R19 R v u xDepth (µm) 0 -50 -100 -150

Strip width (mm) 0.25 0.1 0.1 0.25

Strip pitch (mm) 0.35 0.9 0.9 0.35

Q collected (rel.) 0.84 0.3 1

Tested two chambers with same readout structure (R19M and R19G) in a pion beam (H6) in July

Clean signals from all three readout coordinates, no cross-talk

Strips of v and x layers well matched, u strips low signal, too narrow

Excellent spatial resolution, even with v and u strips

σ = 94/√2 µm

RD51 Coll. Mtg, Joerg Wotschack (CERN) 47

Simplify construction of micromegas

Mesh Mesh stretching over large-area is possible but complicates

the production process Mesh HV insulation requires special attention, work

intensive Mesh segmentation is difficult to achieve, dead areas,

labour intensive Two variants tried

Replace the mesh with a GEM foil Connect the mesh to ground and the R strips to HV

Readout structure

11/08/2011

MPGD2011, Joerg Wotschack (CERN) 48

Replacing the mesh by a GEM foilGEM is simply placed on 128 µm high pillars on the resistive strips

11/08/2011

1. Stripped GEM foil (lower Cu layer removed) R19G, successfully used

in test beam in July Works well, no striking

difference to 19M (with micromesh)

High rate looks OK

2. Standard GEM foil Works similarly as R19G Did not find with this

configuration a HV setting to profit from gain in GEM

Fragile, sparks in GEM destroyed GEM

3. Stripped GEM foil (upper Cu layer removed) Did not find good

operating point Works as MM but

problems with GEM transparency

Effort not pursued any further


Inverting the HV connection scheme Idea by A. Ochi (Kobe) ‘Why

not connect the R strips to HV and the mesh to ground ?’

R20 was built along these lines

Simplifies considerably the detector production

Clean signals, good energy resolution High gain Little charge-up Good high-rate performance

11/08/2011

PCB

Resistive Strips (+HV)

Copper readout stripInsulator

Mesh

Drift electrode (-HV)

Works very well, all recent detectors were built along these lines

R20


Performance of R20 (inverted HV)

11/08/2011

G≈10000

G≈5000

55Fe γ (5.9 keV)Ar:CO2 93:7

Signal drop: 5 % (compared to 10–15% for R11, R12,…)

8 keV Cu X-rays

Signal drop compatible with HV drop over resistive strips (0.5–1 GΩ)


Simplifying the readout board Readout strips on two sides of PCB

carrying the MM structure PCB thickness 0.8 – 1 mm Signals capacitively coupled across

the PCB No through holes Connectors on both sides of the

PCB Avoids multi-layer PCBs Leads to major simplification of

production and cost reduction Concept has been successfully

tested in R21

ATLAS Upgrade Plenary, 18 Nov 2011


Towards large-area MM chambers



Assembly of 1st large resistive MM in March 2011

Size: 1.2 x 0.6 m2

2048 circular strips Strip pitch: 0.5 mm 8 connectors with 256

contacts each Mesh: 400 lines/inch 5 mm high frame

defines drift space O-ring for gas seal Closed by a 10 mm

foam sandwich panel serving at the same time as drift electrode


Dummy PCB


Cover and drift electrode



Drift electrode HV connection


HV connection spring for drift electrode

O-ring seal

Al spacer frame


Chamber closed Assembly simple,

takes a few minutes Signals routed out

without soldered connectors



Chamber in H6 test beam (July 2011)


Large resistive MM

R19 with xuv readout(seen from the back)


Experience with large (1.2 x 0.6 m2) MM The large MM with resistive strips and

0.5 mm pitch has been successfully tested in July and November in the H6 beam


Event display showing a track traversing the CR2 chamber under 20 degree

Hit map showing the beam profile (top) and charge spectrum (bottom)


The 2nd large resistive chamber

Chamber was completed yesterday

Electrical tests are OK Employs the new HV scheme

with mesh on ground potential and resistive strips on +HV



Micromegas in ATLAS cavern



Four resistive MMs installed behind last muon station


R11 R12

R13x R16xy

Trigger (strips)

DCSmmDAQ

Laptopin USA15

≈120 mm

R16


Collision data and cavern background

The chambers operated from April to November 2011 Most of the time collision data were recorded with a stand-

alone trigger In a few fills we recorded several 10 M events with a random

trigger For each trigger the detector activity was measured for 28

time bins of 25 ns, i.e. 700 ns (accepted 500 ns) over a total area of 2 x 81 cm2

Measured background rate: 2.7–3 Hz/cm2 at 1034 cm-2s-1

About a factor 3 lower than the background rate measured in the nearby EOL MDT



ATLAS cavern


Two types of background events

Photon ? Neutron ? induced nuclear break-up

Total charge: 1700 ADC counts

Total charge: >10000 ADC counts



Background rate ≈ 2.7±0.2 Hz/cm2 at L=1034 cm-2s-

1


(Measured rate in close-by EOL2A07 MDT ≈ 8 Hz/cm2)

LHC luminosity (x1033 cm-2s-1)


Conclusions


What have we learned so far ? Micromegas fulfil all of the Small Wheel requirements

Excellent rate capability, spatial resolution, and efficiency Potential to deliver track vectors in a single plane for track

reconstruction and LV1 trigger We found an efficient spark-protection system that is

easy to implement; sparks are no longer a show-stopper MMs are very robust and (relatively) easy to construct

(once one knows how to do it) Large-area resistive-strip chambers can be built … and

they work very well

67PH Det. seminar, 18 Nov 2011 Joerg Wotschack (CERN)

68

What next? We are working on a detector of 1.2 x 0.5 m2 with four active

planes to be installed in the winter shutdown in ATLAS on the Small Wheel on one of the CSCs

Install a small detector (MBT0) in front of the end-cap calorimeter as prototype for a replacement of the Minimum Bias Scintillator trigger detector

Build a 1 m wide chamber (possible after the completion of the upgrade of the CERN PCB workshop, early 2012 ?)

Move to industrial processes for Resistive strip deposition Mesh placement

… and build a full-size module0 for the Small Wheel in 2012



Thank you !for your attention

… and thanks to the many colleagues who in one or the other way helped with our micromegas R&D project



Backup slides



A tentative Layout of the New Small Wheels and a sketch of an 8-layer chamber built of two multilayers, of four active layers each, separated by an instrumented Al spacer for monitoring the internal chamber deformations


Homogeneity and Charge-up

No strong dependence of effective gain on resistance values (within measured range)

Systematic gain drop of 10–15% for resistive & standard chambers; stabilizes after a few minutes

Charge-up of insulator b/w strips ?


R ≈ 45 MΩ R ≈ 85 MΩ


Trigger & readout New BNL chip: 64 channels; on-chip zero suppression,

amplitude and peak time finding Trigger out: address of first-in-time channel with signal above

threshold within BX Data out: digital output of charge & time for channels above

threshold + neighbour channels Trigger signals and data are driven out through one (same)

GBTx link/layer (one board/layer) Trigger: track-finding algorithm in Content-Addressable Memory (or

in FPGA) in USA15; latency estimated 25–32 BXs (of 42 allowed) Small data volumes thanks to on-chip zero-suppression and

digitization



Trigger/DAQ Block Diagram


GBTx Gigabit TranceiverChipset being developed at

CERN, will combineData, TTC, DCS on a single fiber

BNL chip specifications (prelim.)64 channels/chip (preamplifier, shaper, peak amplitude detector, ADC) Real time peak amplitude and time detection with on-chip zero

suppression Simultaneous read/write with built-in derandomizing Buffers Peaking time 20–100 ns; dynamic range: 200 fC Fast trigger signal of all and/or groups of channels Rate: 100 kHz SEU tolerant logicA similar older BNL chip (with longer integration time and smaller

rate capability) has been tested with our MMs and is working well


Documents

Development of m icromegas for the ATLAS Muon System Upgrade