Impact parameter studies

with early data from

ATLASAaron Bundock

University of Liverpool

Overview● Analysis of transverse Impact Parameter distributions with the aim

of characterising the resolution, misalignment and material

budget within the ATLAS Inner Detector

● IP's are used extensively in b-tagging – need to maximise IP resolution by alignment of Inner Detector

● b quarks are heavy flavour and can be a sign of interesting physics, e.g. low mass higgs, SUSY…

● b-tagging also needed for efficient background rejection (W + jets, tt jj)

Inner Detector● Surrounds beampipe up to radius of ~1m● Pixel barrel layers at 5 cm (b-layer), 9 cm, and 12 cm with 2 x 3

endcaps● 4 Silicon barrel layers between 30 cm and 51 cm with 2 x 9

endcaps● Both pixel and semiconductor tracker cover range |η| < 2.5● Transition radiation detector located at radius 56-108 cm

● Pixel layers provide 3 hits per track● SCT layers give 4 hits/track● TRT gives ~ 34 hits/track

Impact Parameter

● Distance between the point of closest approach of a track and

primary vertex● Transverse IP d0 is this distance in transverse plane x,y

● Longitudinal IP z0 is the z-coordinate of this point

Impact Parameter resolution● Divided into intrinsic detector resolution (including misalignment)

and multiple scattering terms:

σd0track = σintrinsic σ MS

● Multiple scattering depends on amount of material in detector and

momentum of particle:

σMS = b

(pT2 sin θ)1/2

● Full IP resolution:

σ2d0

track = σ2intrinsic + b2

pT2 sin θ

● Total uncertainty in IP also depends on resolution of primary vertex:

σd0 = σd0track σPV

● Only valid for cylindrically symmetric material

Selection cuts for 900 GeV● Select only well-defined tracks● Reject fake tracks, long-lived particle tracks, material interactions● Want to select a good primary vertex to reduce error in IP

● Datasets used: 900 GeV

minimum bias trigger data

and Monte Carlo AODs

● Good run lists applied

Cut parameter Cut value

pT

|η|

# Silicon hits

# Pixel hits

# b-layer hits

# tracks in PV

> 0.5 GeV/c

< 2.5

> 6

> 1

> 0

> 3

Resolution plots● Produced by fitting gaussian to a d0 distribution for each bin of

1/(p2 sin3θ)● Intercept of linear fit to resolution plot depends on alignment, and

gradient depends on amount of multiple scattering (material

budget)

Gaussian fit range ± 2σ

Resolution plots @ 900 GeV• 900 GeV data + nominal MC for central region |η| < 1.2

• Also shown is nominal + 10% and + 20% material

Resolution plots @ 900 GeV• Data + nominal MC

• Also MC with ‘day 1’ alignment – initial guess at module alignment

Resolution plots @ 900 GeV• Data + nominal MC: forward region of detector: 1.2 < |η| < 2.5

• Slightly poorer resolution for endcaps (more multiple scattering)

• Resolution parameterization still agrees well for endcaps

Selection cuts for 7 TeV● Similar to the cuts used for 900 GeV● Increase number of tracks in primary vertex due to higher track

multiplicities● Increases resolution of primary vertex● At higher energies, can be many PVs● Need to restrict number of PVs to 1

Cut parameter Cut value

pT

|η|

# Silicon hits

# Pixel hits

# b-layer hits

# tracks in PV

# PVs

> 0.5 GeV/c

< 2.5

> 6

> 1

> 0

> 9

= 1

Resolution plots @ 7 TeV• Data + nominal MC (number of bins increased to 42)

• Ran over 7 TeV minimum bias trigger AODs with good run selection

Resolution plots @ 7 TeV• Data + nominal MC, forward region 1.2 < |η| < 2.5

Resolution plots @ 7 TeV• Look at resolution in bins of eta to see detector behaviour in different regions

• p0 represents intrinsic resolution and misalignment

• p1 represents material budget

misalignment

•This dataset is known to have a misalignment in one of the endcaps

April reprocessing: z vertex

reweight• April reprocessing of data and nominal MC

• Distribution of z-coordinate of primary vertex was broader in Monte Carlo than in data for April reprocessing

• Performed reweight of z-coordinate:

Latest resolution plot @ 7

TeV•Primary vertex cut changed:

• 1 primary vertex with > 9 tracks

• any other PVs must have < 5 tracks

• Bunch crossing identification cut: BCID == 1

Summary• Impact parameter studies on early ATLAS data show an excellent agreement between data and simulation

• Inner detector central region is well aligned, and most of forward region

• Excellent modelling of ID material budget in simulation

• Can move onto b-tagging studies now we have good impact parameter resolution…

Future Work• Study jet properties such as multiplicities, truth flavour, jet pT, jet eta, spatial distributions of jets

• Look at weights of various b-tagging algorithms

• Look at efficiencies and systematics

•Track inefficiencies

•Rejection vs. efficiency plots

Future Work

Future Work