Outline: • DIRC Concept and Design • Detector Performance – Operations – Photon Yield

Jochen Schwiening, SLACRICH2004, Playa del Carmen, Nov 30-Dec 6, 2004 1/24

Outline:

• DIRC Concept and Design

• Detector Performance – Operations – Photon Yield – Detector Resolution

• Physics Application Examples

Jochen SchwieningJochen SchwieningJochen SchwieningJochen Schwiening

PPERFORMANCE OF THE ERFORMANCE OF THE BBAABBAR AR DDIRC IRC

PPERFORMANCE OF THE ERFORMANCE OF THE BBAABBAR AR DDIRC IRC


BBAABBARAR DIRC NIM P DIRC NIM PAPERAPER

New NIM paper describing design, construction, performance of the DIRC:

The DIRC Particle Identification System for the BABAR Experiment.

SLAC-PUB-10516 (Jun 2004), 83pp, to be published in Nucl. Instr. Methods A.

New NIM paper describing design, construction, performance of the DIRC:

The DIRC Particle Identification System for the BABAR Experiment.

SLAC-PUB-10516 (Jun 2004), 83pp, to be published in Nucl. Instr. Methods A.

→ download from SPIRES www.slac.stanford.edu/pubs/slacpubs/10000/slac-pub-10516.html

(I also brought a few copies.)

→ download from SPIRES www.slac.stanford.edu/pubs/slacpubs/10000/slac-pub-10516.html

(I also brought a few copies.)


a Stanford Linear Accelerator Centerb CEA-Saclay, c LPNHE des Universités Paris 6 et Paris 7d LAL, Universite Paris Sude Ecole Polytechnique, LPNHEf Lawrence Berkeley National Laboratoryg University of California, Santa Barbarah Colorado State Universityi University of Cincinnati

Novel RICH detector used for the first time in BABAR.

DIRC combines with dE/dx from drift chamber and vertex detector

(mostly in the 1/2 region) ashadronic particle identification

system for BABAR.

Novel RICH detector used for the first time in BABAR.

DIRC combines with dE/dx from drift chamber and vertex detector

(mostly in the 1/2 region) ashadronic particle identification

system for BABAR.

BBAABBARAR DIRC C DIRC COLLABORATIONOLLABORATION

The BABAR-DIRC Collaboration

R. Aleksan,b D. Aston,a N. van Bakel,a D. Bernard,e G. Bonneaud,e F. Brochard,e

Ph. Bourgeois,b D.N. Brown,f J. Chauveau,c M. Convery,a S. Emery,b A. Gaidot,b X. Giroux,b Ph. Grenier,e G. Grosdidier,d T. Hadig,a G. Hamel de Monchenault,b

B. Hartfiel,i A. Höcker,d M. John,c R.W. Kadel,f M. Krishnamurthy,j M. Legendre,b J. Libby,a A.-M. Lutz,d J. Malcles,c G. Mancinelli,h B.T. Meadows,h Ll.M. Mir,f

D. Muller,a J. Ocariz,c T. Petersen,d M. Pripstein,f B.N. Ratcliff,a L. Roos,c S. Schrenk,e M.-H. Schune,d K. Suzuki,a J. Schwiening,a V. Shelkov,f M.D. Sokoloff,h S. Spanier,j A.V. Telnov,f G. Therin,c Ch. Thiebaux,e

G. Vasileiadis,e G. Vasseur,b J. Va'vra,a M. Verderi,e R.J. Wilson,g G. Wormser,d A. Yarritu,a Ch. Yéche,b Q. Zeng,g M. Zitob


6.5mrad@4GeV/c

• 1.7 < |p| 4.2 GeV/c Pion/Kaon separation in rare charmless decays, e.g., B + – / B K

(time dependent decay asymmetry measures sin2α • |p| < 2 GeV/c

B/B flavor tagging with Kaons via b c s cascade (sin2β measurement)

PID using dE/dx of the BABAR Vertex Detector and Drift Chamber is only effective for |p| <0.7 GeV/c

~<+–

Design Constraints: • CsI Calorimeter needs to detect photons down to 20 MeV,

thus small radiation length (<20%) and small radial size required. • Radiation robustness (expect 10 krad within 10 year lifetime).

• /K separation at 4 GeV/c: 6.5 mrad 3 separation requires 2.2 mrad resolution

Covering all B Decays at BABAR requires Particle Identification (PID) up to 4.2 GeV/c momentum.

PPARTICLEARTICLE I IDENTIFICATION AT THE DENTIFICATION AT THE B-FB-FACTORYACTORY


• A charged particle traversing a radiator with refractive index n with v/c> 1/n emitsCherenkov photons on cone with half opening angle cos c = 1/n.

• If n>2 some photons are always totally internally reflected for 1 tracks.

• Radiator and light guide: Long, rectangular Synthetic Fused Silica (“Quartz”) bars (Spectrosil: average <n()> 1.473, radiation hard, homogenous, low chromatic dispersion)

• Photons exit via wedge into expansion region (filled with 6m3 pure, de-ionized water).

• Pinhole imaging on PMT array (bar dimension small compared to standoff distance).(10,752 traditional PMTs ETL 9125, immersed in water, surrounded by hexagonal “light-catcher”,transit time spread ~1.5nsec, ~30mm diameter)

• DIRC is a 3-D device, measuring: x, y and time of Cherenkov photons,

defining cc tpropagation of photon.

DIRC PDIRC PRINCIPLERINCIPLE


3.1 GeV positrons on 9.0 GeV electrons

center of mass energy M(4S) =10.58 GeV/c2

= 0.56

TTHEHE SLAC PEP-II B- SLAC PEP-II B-FACTORYFACTORY


BABAR layoutBABAR layoutInstrumented Flux Return19 layers of RPCs, (being upgraded to LSTs) 1.5 T Solenoid

DIRCDIRCDrift Chamber40 layers (24 stereo)

ElectromagneticCalorimeter6 580 CsI crystals

Silicon Vertex Detector5 layers of double sided silicon strips

e– (9.0 GeV)

e+ (3.1 GeV)

TTHEHE B BAABBARAR D DETECTORETECTOR


DIRC thickness: 8 cm radial incl. supports19% radiation length

at normal incidenceDIRC radiators cover:

94% azimuth, 83% c.m. polar angle

BABAR-DIRC Timeline:

November 1998: installed SOB and one bar box, PMTs in water;

April 1999: BABAR moves into beam line, added 4 more bar boxes;

November 1999: all 12 bar boxes installed, start of first physics run.

Summer 2004: PEP-II peak luminosity 9.2×1033 cm−2sec−1 BABAR total recorded ~245M BB pairs.

TTHEHE DIRC DIRC ININ B BAABBARAR


• Calibration constants stable to typically rms < 0.1ns per year.

• No problems with water or gas systems.

The two most significant operational issues:

• Sensitivity of the DIRC to machine background interacting in the water of the SOB (primarily DAQ issue)

• Concerns about PMT longevity due to PMT window degradation

DIRC is Stable and RobustDIRC is Stable and Robust

DIRC ODIRC OPERATIONALPERATIONAL E EXPERIENCEXPERIENCE

DIRC has run in factory mode for 5 years, PMTs have been immersed in ultra-pure water for 6 years


PMT front window corrosion• Discovered after ~ 1 year immersion Oct. 99.

• Status Oct. 99: ~ 50 frosty tubes and ~ 2/3 visibly milky.

• Only front glass affected, side glass fine.

Studies

- Strongly corroded (frosty) tubes are a bad batch of PMT glass (no zinc).

- Milkiness results from sodium depletion in near surface.

No obvious immediate effect (water provides good coupling) but ...

Might lose PMT efficiency with time.

Might lose vacuum in some of the ~ 50 frosty tubes on 10 year time scale (front window thickness: 1mm).

milky

frosty

electron microscope image

DIRC PDIRC PMT MT LLONGEVITYONGEVITY

Frosty PMT


DIRC standoff region filled with ~6000 l ultra-pure water(recirculated ~2.5 times per day, typical resistivity ≈ 8-10MΩcm, pH ≈6.5)

DIRC WDIRC WATERATER S SYSTEMYSTEM

Analysis of samples of DIRC water system:

• transparency (laser, 3 wavelengths)

→ stable at >98%/m for 325nm

• chemical composition (spectroscopy, outside company)

from sodium content comparison in supply vs. return water calculate PMT front glass loss rate

→ ~2-4μm/PMT/year, not a problem


DIRC PDIRC PHOTON HOTON YYIELDIELD

Example: number of “dead PMTs”

• record in each run number of PMTs with

rate <<10% of expected hit rate

• loss rate vs. time is remarkably god fit to a simple line – we lose ~2.4 PMTs per month

(not understood in detail, green ranges excluded from fit)

Detailed monitoring of photon yield using:

• LED pulser calibration;

• PMT aging tests;• comparison of signal photon yield

in real Bhabha and di-muon events; • observed rates of signal and

background in all PMTs;

Detailed monitoring of photon yield using:

• LED pulser calibration;

• PMT aging tests;• comparison of signal photon yield

in real Bhabha and di-muon events; • observed rates of signal and

background in all PMTs;


Consistent result from all studies:photon yield loss few % per year.

very minor impact on PID performance over 10 year lifetime of DIRC.

DIRC PDIRC PHOTON HOTON YYIELDIELD

time

Num

ber

of s

igna

l pho

tons

per

trac

k Example:

di-muon events (e+e–→μ+μ–)

study number of signal photons from fit to “ring” as function of time

• loss rate few %/year

• interesting variations with time (not fully understood)

run 1 run 2 run 3 run 4

e+e–→μ+μ–

raised PMT HV to compensate for gain loss


Luminosity (1033/cm2s)

max

imum

sca

ler

rate

(kH

z)

Succession of lead shielding installed in 2000 and 2001.

Current TDC and shielding configuration “background safe” at 1034 /cm2·s.

Thanks to lead shielding, PMT rates acceptable even at 3 times design lumi.

PMT Rate vs. Luminosity shows that lead shielding essential in protecting DIRC from

few MeV photon accelerator induced background (radiative Bhabhas etc).

DIRC TDC1: ~5% inefficiency at 250 kHzSpring 2000

Fall 2000

Spring 2001

BBACKGROUNDSACKGROUNDS IN THEIN THE DIRC DIRC

New TDC chips were installed during shutdown Fall 2002.

DIRC TDC2: <5% deadtime at 2.5MHz


DIRC “Ring” images:

• limited acceptance for total internal reflection,

• reflection ambiguities (initial reflection up/down, left/right, reflection off mirror, wedge

up to 16 (cc) ambiguities per PMT hit),

• toroidal detection surface,

Cherenkov ring images are distorted:

complex, disjoint images

Low energy photons from accelerator hit Standoff Box.

At current luminosity that causes rates of 80-200 kHz/tube.

80-200 kHz 10,752 PMTs 300 nsec trigger window

500-1300 background hits (~10% occupancy)

compared to

50-300 Cherenkov photons

DIRC RDIRC RECONSTRUCTIONECONSTRUCTION


300 nsec trigger window 8 nsec t window (~500-1300 background hits/event) (1-2 background hits/sector/event)

Calculate expected arrival time of Cherenkov photon based on

• track TOF • photon propagation in radiator bar and in water

t: difference between measured and expected arrival time

(t) = 1.7 nsec

t(nsec)

Time information provides powerful tool to reject accelerator and event related background.



Example: Comparison of real event to simulated response of DIRC to e//K/p.

Calculate unbiased likelihood for observed PMT signals to originate from e///K/p track or from background.

( Likelihood: Pdf(c) Pdf(t) Pdf(N) )

Two complementary reconstruction algorithms:

• iterative process to maximize event likelihood, full correlation of all tracks;

• individual track fit provides c, c, number of signal/background photons.

Reflection ambiguities: t cut reduces these from up to 16 to typically 2-3

Particle ID is based on log likelihood differences

of the five hypotheses.



Expectation: ~9.5 mrad

dominated by:7mrad from PMT/bar size, 5.4mrad from chromatic term,2-3mrad from bar imperfections.

Single Photon Cherenkov angle resolution:c,: difference measured c, per photon solution and expected track c (di-muons)

~10% Background under c, peak:combinatoric background, track overlap, accelerator background,

electrons in radiator bar, reflections at fused silica/glue interface, ...

(c,) = 9.6 mrad

DIRC PDIRC PERFORMANCEERFORMANCE


Number of Cherenkov photonsper track (di-muons) vs. polar angle:

Resolution of Cherenkov angle fitper track (di-muons):

(c) = 2.4 mrad

Track Cherenkov angle resolution is

within ~10% of design.

Should improve with advances in track- and DIRC-internal alignment.

Very useful feature in BABAR environment:higher momentum correlated with larger polar angle values

more signal photons, better resolution (~ 1/N )

Between 20 and 60 signal photons per track.



• Select D0 candidate control sample with mass cut (0.5 MeV/c2)

and K are kinematically identified

• calculate selection efficiency and mis-id

• Correct for combinatorial background (avg. 6%) with sideband method.

kinematically identified and K

example:2.5<|p|3GeV/c

K

D – D0 –

K– +



D– D0 –

K– +

Kaon selection efficiency for

L K > L

Kaon selection efficiency for

L K > L

mis-id as K

(6% comb. background)

(track in DIRC fiducial, comb. background corrected)

Kaon selection efficiency

typically above 95%with mis-ID of 2-10% between 0.8-3GeV/c.



/K separation power:Measure Cherenkov angle resolution as function of track momentum

for pions and kaons, kinematically identified in D* decays.

about 4.3 separation at 3GeV/c, close to 3 separation at 4GeV/c



Kaon Tagging for sin(2) measurement:

• highest efficiency • low mis-tag fraction w• dominant contribution to measurement

(sin2 1/Q.NCP

Q: effective efficiency

28.1 ± 0.7 %28.1 ± 0.7 %65.6 ± 0.5 %65.6 ± 0.5 %TotalTotal

2.7±0.3 %2.7±0.3 %31.5±0.9 %31.5±0.9 %20.0±0.3 %20.0±0.3 %Incl.Incl.

6.7±0.4 %6.7±0.4 %20.9±0.8 %20.9±0.8 %19.8±0.3 %19.8±0.3 %Kaon IIKaon II

10.7±0.4 %10.7±0.4 %10.0±0.7 %10.0±0.7 %16.7±0.2 %16.7±0.2 %Kaon IKaon I

7.9±0.3 %7.9±0.3 %3.3±0.6 %3.3±0.6 %9.1±0.2 %9.1±0.2 %LeptonLepton

Q=Q=(1-2(1-2ww))22wwEfficiencyEfficiencyFlavor tagFlavor tag


0( ) (CP odd) modesScc K

BBAABBARARBBAABBARAR2004


The DIRC is a novel type of particle identification system, well matched to asymmetric B-factory environment, capable of -K separation for momenta up to ~ 4 GeV/c.

Five years of experience in PEP-II/BABAR B-factory mode: DIRC very reliable, robust, easy to operate, 98.5% of channels fully functional.

Machine backgrounds up to 200 kHz/PMT at 9·1033/cm2·s no problem for reconstruction;

new TDC chips, installed in 2002, keep DIRC safe for 1034/cm2·s luminosity.

Single photon time and Cherenkov angle resolution and photon yield close to nominal.

Track Cherenkov angle resolution within 10% of design.

DIRC plays significant role in almost all BABAR physics analyses published to date.

R&D program under way to prepare DIRC for 1034 /cm2·s and beyond.(see posters by J. Va’vra et al.)

Read more details in www.slac.stanford.edu/pubs/slacpubs/10000/slac-pub-10516.html

CCONCLUSIONONCLUSION


BBACKUPACKUP S SLIDESLIDES


Sodium and Boron levels (return-supply) in SOB water = f(time)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

4/12/99 10/29/99 5/16/00 12/2/00 6/20/01 1/6/02 7/25/02

Date

Sod

ium

and

Bor

on

leve

ls in

wat

er [

ppb]

Sodium level

Boron level

PMT front surface removal rate = f(time)

y = -0.0013x + 50.536

y = -0.0019x + 70.187

0

1

2

3

4

5

6

4/12/99 10/29/99 5/16/00 12/2/00 6/20/01 1/6/02 7/25/02

Date

Rat

e of

rem

oval

[m

icro

ns/p

mt/

year

]

Calculation using Sodium

Calculation using Boron

Linear (Calculation using Sodium)

Linear (Calculation using Boron)

Calculation using Boron

Chemical analysis of return water

from standoff box:

• measure Sodium and Boron levels;• measurements since 1999; • levels indicate loss of 2-3m per year

from PMT windows (1mm thick).

Chemical analysis of return water

from standoff box:

• measure Sodium and Boron levels;• measurements since 1999; • levels indicate loss of 2-3m per year

from PMT windows (1mm thick).



Concern with mechanical stability of PMTs when front glass gets thinner (Super-K...)

• breaking test of PMT bundle at 4m and 8m water column (DIRC max. < 4m);• used sharp plunger to break PMT front glass; • no breaking of neighboring PMTs observed at either depth.

Concern with mechanical stability of PMTs when front glass gets thinner (Super-K...)

• breaking test of PMT bundle at 4m and 8m water column (DIRC max. < 4m);• used sharp plunger to break PMT front glass; • no breaking of neighboring PMTs observed at either depth.

plunger

PMT bundlewith light catchers

broken PMT



Most spectacular failure mode of PMTs:

“Christmas Tree”

• loss of vacuum in PMT at base;• discharge in PMT creates many photons,

emitted via front or clear side glass;• extra photons are detected by neighbors

(sometimes scatter through entire SOB);• rates in affected PMT in MHz range,

several 100kHz in neighbors;• PMT dies after few hours – days;• HV of affected PMT immediately

(automatically) lowered to preserve data quality;

• observed rate: 5-6 per year.

Most spectacular failure mode of PMTs:

“Christmas Tree”

• loss of vacuum in PMT at base;• discharge in PMT creates many photons,

emitted via front or clear side glass;• extra photons are detected by neighbors

(sometimes scatter through entire SOB);• rates in affected PMT in MHz range,

several 100kHz in neighbors;• PMT dies after few hours – days;• HV of affected PMT immediately

(automatically) lowered to preserve data quality;

• observed rate: 5-6 per year.

Christmas Tree PMT

Rate in one sector

Rate in all sectors



After over five years, ~160 PMTs out of 10,752 are dead or inefficient 98.5% fully functional

After over five years, ~160 PMTs out of 10,752 are dead or inefficient 98.5% fully functional



DIRC is Stable and Robust

• Calibration constants stable: typical rms of T0 per channel ~ 0.1ns

(light pulser and data stream).

• Monitor humidity of nitrogen return line frombar box: dew points constant at -45... -55C,no leaks after installation.

• Water purification system keeps resistivityat 18.5Mcm (input) and 9.5Mcm (return).

• Water transmission in SOB remains stable at 98%/m (442nm, 325nm), 95%/m (266nm).

occu

panc

ym

ean

T0

(ns)

sigm

a T

0 (n

s)

elapsed time (days)

2ns

colors indicate sectors

DIRC CDIRC CALIBRATIONALIBRATION S STABILITYTABILITY


• Charged Hadron Spectra (, K, p/p) analysis.

• Cuts optimized to keep Mis-ID < 1-2% everywhere.

• In return, must accept somewhat lower ID efficiency especially at high momenta.

Example for combination of DIRC likelihoodswith drift chamber and vertex detector likelihoods


Documents

Outline: • DIRC Concept and Design • Detector Performance – Operations – Photon Yield