c. h. shepherd-themistocleous rutherford appleton laboratory, uk

Charged Higgs – Uppsala 2006 C.H. Shepherd-Themistocleous- RAL 1

C. H. Shepherd-Themistocleous Rutherford Appleton Laboratory, UK

Identification of tau particles in the CMS detector

Rutherford Appleton Laboratory

cH±arged 2006, Uppsala University, Sweden, 13-16 September 2006


Outline

Properties of particles

CMS detector

identification techniques in hadronic decays at CMS

• Isolation• Decay length• Impact parameter• Invariant mass

N.B. HLT performance later this afternoon


Characteristics of Tau decays

Lifetime c 87 m (0.29 ps) , m1.78 GeV/c2

Decays: 65% hadronic , 35% leptonic• Hadronic

– 1 prong 50 % : n

– 3 prong 15 % : 3n

tau jets at LHC:• Very collimated

• Low multiplicity

– One, three prongs

• Hadronic, EM energy deposition

– Charged pions

– Photons from 0


tau tagging

Properties used for tagging at CMS– Narrow jets

• ECAL isolation

• Tracker isolation

– Significant lifetime

• Impact parameter

• Decay length

– Invariant Mass

Backgrounds – QCD jets

– Electrons


Compact Muon Solenoid

Tracker

4T solenoid

Muon chamber

s

HCAL

Iron yoke

Total weight: 12,500 tOverall diameter: 15 mOverall length: 21.6 mMagnetic field: 4 T

Si microstripsPixels

BarrelDrift tubes (DT)Resistive platechambers (RPC)

EndcapsCathode Strip Chambers (CSC)Resistive platechambers (RPC)

Plastic scintilator/brass sandwich

Scintillating PbWO4

crystals

ECAL


The CMS Tracker

5.4 m

Endcap Strips

Outer Barrel Strips

2.4

m

Inner Barrel Strips

PixelsThe World’s largest

Silicon Tracker = 250 m2

!

10 layers of Silicon Strip Sensors surrounding 2-3 layers of Silicon

Pixel Sensors.

15000 silicon modulescontaining 76000000 pixels +

strips !

TEC

TOB

TID


ECAL

High resolution electromagneticcalorimetry is central to the CMS design

m / m = 0.5 [E1/ E1 E2

/ E2 / tan( / 2 )]

Where: E / E = a / E b c/ E

Aim: Barrel End cap Stochastic term: a = 2.7% 5.7% (p.e. stat, shower fluct, photo-detector, lateral leakage)Constant term: b = 0.55% 0.55% (non-uniformities, inter-calibration, longitudinal leakage)Noise: Low L c = 155 MeV 770 MeV High L 210 MeV 915 MeV(dq relies on interaction vertex measurement)


Events used

tau sample – taus only i.e. no pile up, no underlying event

– pT jet > 30 GeV, uniform in |

QCD sample

– di-jets events in Pythia ET 30-150 GeV, Rsep > 1.5

– True energy is that found when using cone size 0.5.

Matching: R(Calorimeter jet axis – MC jet axis) < 0.2

Efficiency for QCD events to pass preselection and matching ~12%


ECAL isolation

candidates Pisol < cut value

Efficiency of rejection of QCD jets increases with ET. — Low pT tracks (<2 GeV/c) bent out of cone

Achieve 80% efficiency with bkg rejection of factor of 5 for QCD jets with pT > 80 GeV/c(wrt presel. and matching)


Tracker Isolation

axis of calo jet

Jets reconstructed with iterative cone algorithm

Look for tracks inside jet-track matching cone Rm (0.1) with pt > 6 GeV

Form signal cone around track with highest pT.

Tracks inside Rs with z d0 within z (2mm) of leading track deemed to be from tau

Tracks reconstructed within Ri.

Require pT > pTi (1 GeV) and within

z (2mm) of leading track.

Tracks 8 Si hits with at least 2 pixel hits.

Isolation requires no non-tau tracks within Ri.


Tracker Isolation Performance

Single Tau (30<ET<150 GeV)

QCD jets50<ET<170 GeV

Ri Ri

Bins 130-150, 80-110, 50-70, 30-50 GeV

ET

inc

Single tau simulated events QCD events generated in bins of pT.

Efficiency wrt preselected and matched events


Impact parameter tag1 prong 3 prong

QCD tail due to fake tracks. Hits onreconstructed track from various true tracks.Majority at large Extrapolation distances larger

Little discriminationin 3-prong events

IP for highest pT track


d0 performance

• Efficiency of transverse d0 significance cut. d0 < 300 m

• Mean error ~ 15 m (1-) 16.7 m (3-) : QCD 17.9 (1-) 22.2(3-) m

Tracker isolation required Significance cut


Electron Rejection

Electrons can fake 1-prong taus.

Events selected requiring ECAL and Tracker isolation

Background suppressed using HCAL information

Minimum requirement on energy of most energetic HCAL tower in the jet.

HCAL cut electron jet

ET 40-60 GeV

jet

ET 100-140 GeV

> 1 GeV 0.08 0.936 0.977

> 2 GeV 0.03 0.854 0.942

Performance of HCAL cut for leading track pT > 10 GeV

All distributions normalised to 1.

Tail due to gaps in ECAL


Decay length tag

Very collimated jets lead to shared hits in pixel layers.Decay length < 35 mm required.

Events required to pass tracker isolation and have 3 tracks in the signal cone.

Probability ~ 63% for 3-prong decays

b and c quark jets not a major problem(~12% c 3% b)

-jetsQCD jets


SV - decay length

Secondary vertex resolution in jet events

Resolution transverse to jet axis Resolution parallel to jet axis

• Analysis used the Kalman vertex fitter (KVF)


Decay length performance

Performance as a function ofsigned transverse decay lengthsignificance

Error in decay length dominated by secondary vertex. - Primary vertex:

QCD events use pixel vertex finder events smear z by 60m

Rejection factor of 5 possible for a signal efficiency of 70-80%(efficiency calculated wrt MC preselection and matching and tracker isolation)


Mass tag

Mass reconstruction uses track momenta and energy of ECAL clusters.

Clusters matched to tracks are removed to avoid double counting. — Clusters only used if track - cluster R > 0.08 — Use clusters within cone of 0.4

Due to 1-prong decays

ET 30-50 GeV ET 130-150 GeV


Mass tag performance

— Tracker isolation required

— Mass cut < 2.5 GeV/c2


tag efficiency determination

Method: Use in single muon triggers


efficiency II

Principal backgrounds: t t , W + jet, QCD

Error on tag tag efficiency using 30fb-1 of data.

This method allows verification of MC at Z energies. The efficiency is a function of jet energy. – Greater collimation

• lower probability of signal tracks outside signal cone• greater probability of tracks sharing hits


ID performance

From R. Kinnunen’s talk on Wednesday

For this channel: Rm = 0.1, Rs = 0.07, Ri = 0.4 with veto on tracks with pT > 1 GeV in the isolation coneECAL isolation for jet: ET

cell (0.13<R<0.4) < 5.6 GeV, R defined around the jet direction Electron contamination suppressed with a cut on maximal HCAL cell (ET > 2 GeV) inside the jet cone pleading track / E > 0.8, to exploit the opposite helicity correlations in in the H± -> and W± -> decays, leading to harder pions from H± ->

Efficiency: signal 11-15%, tt->WWbb -> bbl 5%, W+3jets 1%, tt->WWbb -> bbl’’l 2%

Signal process defined as gg -> tt -> W±H±bb->lbb, l = e or mH+ = 140 GeV


ID performance II

From R. Kinunnen’s talk on Wednesday

Associated production gg -> tbH± , H± -> Level 1 jet trigger ET > 93 GeV

-selection efficiencies including pre-selection, trigger and off-line mmH± (GeV) (GeV) 200 400 tt Wt W+3j QCD

Total efficiency 3.8% 9.0% 5.0x10-4 6.7x10-4 1.9x10-3 9.2x109.2x10-5-5

Offline Offline identification:- jet reconstruction in the direction of the triggeredjet, ET > 100 GeV- leading track within R < 0.1 around the jet direction- small signal cone around the leading track, r=0.04- one or three tracks in the signal cone- isolation of the signal cone in 0.04<R<0.4- addional quality cuts for the leading track: transverse impact parameter < 0.3 mm and at least 10 hits in the tracker (signal efficiencies ~95%)- ET of maximal HCAL cell in the jet > 2 GeV to remove electron contamination- pleading track/E jet > 0.8, exploits the helicity correlations


Summary

Reconstruction utilizes characteristics of tau jets

Principal methods are:– Isolation in ECAL & tracker– Decay length– Impact parameter – Invariant mass

A method for determining efficiency from data has

been studied.


Backup slides


Choice of jet cone size

Cone size of 0.4 chosen. Contains 98% of jet energy and good energy resolution


Energy scale corrections

Tau jets need softer corrections to their

energy, wrt QCD jets.– For the same transverse energy,

pions in Tau jets have harder transverse momentum than pions in QCD jets

– In Tau jets there is a larger amount of electromagnetic energy (due to the presence of p0)

– Corrections parametrized as function of ET and

The jets corrections optimised for true hadronic taus significantly underestimate energy scale for QCD jets


Performance


trigger

Level-1 (~µs) 40 MHz High-Level ( ms-sec) 100 kHzEvent Size ~ 106 Bytes

Level-1 (~µs) 40 MHz High-Level ( ms-sec) 100 kHzEvent Size ~ 106 Bytes

40 MHzClock drivenCustom processors

100 kHzEvent drivenPC networkTotally software

100 HzTo mass storage

two trigger levelstwo trigger levels


L1 trigger

QCD ET samples in the range 50-170 GeV were used for the HLT studies. They represent more than the 90% of the total L1 Rate

A factor ~103 of QCD background rejection is required at HLT

— Reduce rate from ~kHz -> ~Hz

Active towers patternsallowed for tau jetscandidates


HLT

Two trigger algorithms “Calo+Pxl trigger ” and “Trk trigger”

– “Calo+Pxl trigger”: only pixel hits and calorimeter isolation used• Fast – limited track reconstruction• Good performance for isolation• preferred for decays with two taus in the final state (like A/H-

>tautau)

– “Trk trigger”: (some) hits of the microstrip inner tracker used, no calorimeter isolation

• slower than “Calo+Pxl”• much better resolution for track momenta • useful in channels like charged Higgs boson decay ( plus missing energy selection)

– tight cut on the pT of the leading track


ECAL isolation

Efficiency evaluated for bbH bb+ samplew.r.t. L1 trigger.

QCD di-jet events in range pThat:50-170

GeV used to evaluate background suppression.

Rejection factor 3 is given by Pisol < 5.0 GeV

Used with Pixel isolation to form trigger

Isolation is applied to the most energetic) HLT calorimeter jet.


Calo + Pixel HTL

Track isolation alg. similar to offline. Tracks constructed from 3 pixel hits only.

Isolation cone varied from 0.2 to 0.6 (step 0.05). Signal cone 0.07, matching cone 0.1, leading track pT > 3 GeV/c.

Single tag Double tag

Efficiency calculated w.r.t L1 trigger


Charged Higgs Trigger

Channel considered

gg->tbH+, gb->tH , H+ -> (tau hadronic decay)

L1 output rate: ~ 3kHz

HLT selection

ETmiss>65 GeV: output rate ~30 Hz

After applying Tracker isolation + momentum cut (PT

LT>20GeV):— output rate: ~ 1Hz


performance in example channel


HCAL


H0/A0 decaysProvides best reach large tan : + , + had, had + had had+had final state:

Backgrounds: QCD ( muli-jet fake ) ; Z/* tt ; W+jet, W . Requires hadronic trigger Large associated production allows good rejection with b tag. “jet” (1-, 3- prong) tagging, lifetime

Potential SUSY background. - decays. negligible

b tagging

QCD ~ 106 Mass resolutionrejection ~ 15%

~

ExploitbbH0/A0

production


Provides clear signature for BSM physics.

Production: – mH

± < mt: tt , t H± b– mH

± > mt: gb t H± ,gg tbH ±, qq’ H±

Backgrounds: tt ; Wtb, W

Signal : Look for lepton from top + had

Spin correlations from H harder than from W . Require 80% jet energy carried by +

Plot transverse mass. (missing >1 )– Signal endpoint ~ mH

– Background endpoint mw

H± decays


add properties of sub detectors

resolutions

c. h. shepherd-themistocleous rutherford appleton laboratory, uk

Documents

nontau tracks

low pt tracks

fake tracks

p npo tau jets

various true tracks

prong taus

bins of pt

pt pti