Impact Parameter Resolution Measurements

1
Impact Parameter Resolution Measurements from
900 GeV LHC DATA
Boris Mangano Ryan Kelley (UCSD)
2
Outline

• Samples and Event selection.
• Description of the method.
• Validation of the method on MC samples.
• Measurement of Transverse and Longitudinal
  Impact Parameter resolutions on 900 GeV DATA.

3
Samples and Event Selection
DATA /MinimumBias/BeamCommissioning09-BSCNOBEAMHA
LO-Dec19thSkim_336p3_v1/RAW-RECO MC /MinBias/S
ummer09-STARTUP3X_V8K_900GeV-v1/GEN-SIM-RECO

• Event Selection (see backup slides)
• BSC trigger and BH Veto (already applied at the
  level of the central skim)
• Tech.Trigger bit 0
• PhysicsDeclared bit
• NoScrapingEvent filter
• Only B3.8 T runs
• Only E900 GeV runs

4
Preliminary Remarks Different Components
affecting the IP measurement
Case 1) Exactly known trajectories and perfectly
known collision point (SimTracks and SimVertex)
Case 2) Exactly known trajectories, but measured
collision point (SimTracks and reco PV)
Both Tk1 and Tk2 have d0 different from their
true values because of the smearing due to the
vertex position resolution d0 d0true ?
vertex smearing
5
Preliminary Remarks Different Components
affecting the IP measurement
Case 3) Both trajectories and collision point are
not exactly know because they are both
measured Real World
reco Tk1
Tk1
reco Tk2
Tk2
reco PV
Both Tk1 and Tk2 have d0 different from their
true values because of the smearing due to the
vertex position resolution and the smearing due
to the finite resolution on the track parameters
d0meas d0true ? vertex smearing ? track
impact parameter resolution
6
Preliminary Remarks Different Components
affecting the IP measurement
d0meas d0true ? vertex smearing ? track
impact parameter resolution
The agreement between the resolutions of the
primary vertex position measurements in DATA and
MC has already been proved by Y.Gao, F.Fiori et
al. https//twiki.cern.ch/twiki/bin/view/CMS/Prim
aryVertexResolution In the rest of the talk,
focusing on the effect of vertex resolution on
the IP measurement vertex smearing
Final observable measured by this analysis
7
Description of the method
1) Select only events with the primary vertex
fitted from 10-16 tracks (see next). 2) For a
given track with measurements on 8 Tracker layer
(2 pixels), refit the primary vertex using all
the other tracks in the event. 3) For the same
track, evaluate d0 and dz w.r.t. the refitted
vertex position. 4) Repeat 2-3 for all the other
tracks in the event that pass the selection. 5)
Save the previously evaluated d0 and dz values in
bins of tracks pT and eta. 7) Repeat 2-5 for all
the selected events.
8
Description of the method (II)
8) Fit each pT and eta bin using the function
F VertexSmearingFunction ? IP
ResolutionFunction
Measured d0 wrt recoPV ?m
Vertex smearing pdf ?m
Impact parameter resolution ?m
Measured from data
Estimated from MC
Extracted from the Fit of the convolution
function F
9
Evaluation of the VertexSmearing function
The d0 and dz of simulated SimTracks were
evaluated with respect to the reconstructed PV
position on the MC sample.
The d0 and dz calculated in this way were put in
separate histograms for distinct SimTracks pT
and eta values.
Each bin is fitted by a 2-gaussian sum function.
10
Evaluation of the VertexSmearing function
Vertex smearing for d0 as a function of the
tracks eta
Vertex smearing for d0 as a function of the
tracks pT
Vertex smearing pdf ? ?m
Tracks pT GeV/c
Tracks eta
Vertex smearing for dz as a function of the
tracks eta
Vertex smearing for dz as a function of the
tracks pT
Vertex smearing pdf ? ?m
Tracks pT GeV/c
Tracks eta
11
Evaluation of the Impact Parameter Resolution
F VertexSmearingFunction ? IP
ResolutionFunction
Data are fit using the convolution function F,
where only the parameters of the IP
ResolutionFunction (a single gaussian) are
unconstraint.
Vertex smearing pdf ?m

• The use of a single gaussian for modeling the
  resolution function is appropriate if
• The pT and eta of the tracks contributing to the
  same histogram are sufficiently close.
• Only the core of the distribution is considered
  (2sigma fit)

12
Validation of the method on MC Transverse Impact
Parameter Resolution
Additional Track cut eta lt 0.4
Good agreement between the measured resolutions
(red) and the mc-truth-determined resolutions
(black).
13
Validation of the method on MC Transverse Impact
Parameter Resolution
Track selection 0.75 lt pT lt 0.85 (GeV)
Track selection 1.0 lt pT lt 1.4 (GeV)
14
Validation of the method on MC Longitudinal
Impact Parameter Resolution
Track selection 0.75 lt pT lt 0.85 (GeV)
Track selection 1.0 lt pT lt 1.4 (GeV)
15
Results from 900 GeV DATA
Track cut eta lt 0.4
16
Results from 900 GeV DATA
Track selection 0.75 lt pT lt 0.85 (GeV)
Track selection 1.0 lt pT lt 1.4 (GeV)
17
Results from 900 GeV DATA
Track cut eta lt 0.4
18
Results from 900 GeV DATA
Track selection 0.75 lt pT lt 0.85 (GeV)
Track selection 1.0 lt pT lt 1.4 (GeV)
19
Conclusions
Method to estimate the Track Impact Parameter
Resolution from DATA has been implemented and
described in this talk. The Method has been
successfully validated on MC and used on 900 GeV
DATA from December 2009 runs. Results based on
900 GeV DATA are very close to values expected
from STARTUP MC. Impact Parameter resolutions are
just slightly worse. Overall, the analysis will
benefit from higher LHC energy events for which
there are more high pT tracks from the collision
and the uncertainty on the primary vertex
position is smaller. TWIKI page and Analysis
Note in preparation https//twiki.cern.ch/twiki/b
in/viewauth/CMS/ImpactParameterResolutionDec2009
20
BACKUP SLIDES
21
Event Selection for DATA (and that you dont
usually apply on a MC sample)

• Require Tech.Trigger bit 0 (i.e. BPTX
  coincidence) it selects events in beam-beam
  bunch crossings. Discards single-beam BXs and
  empty BXs.
• Require Tech.Trigger bit 40 or 41 (i.e. BSC
  triggers) selects events with activity
  compatible with a collision
• Veto on Tech.Trigger bit 36-39 (BeamHalo
  triggers)

22
Event Selection for DATA (II) (and that you dont
usually apply on a MC sample)

• Discard events for which B!3.8, e.g. events
  collected during the magnet ramping to its
  nominal value.
• 4) For this study only, discard events from 2.36
  TeV runs
• Discarded runs 123967-123997, 124119,124120,12427
  5
• 5) Require PhysicsDeclared bit this bit is set
  to TRUE when all the CMS components are declared
  OK for data taking e.g. all the HighVoltages of
  the Tracker sensors are ON.

23
Event Selection for DATA (III) (and that you
dont usually apply on a MC sample)
6) Reject beam-background events (AKA pixel
monster events, AKA scraping event). The
NoScraping event filter has been used it
requires gt25 of tracks in the event to be
highPurity tracks.
24
Validation of the method on MC Longitudinal
Impact Parameter Resolution
Track selection eta lt 0.4
Some bins looks problematic
25
Validation of the method on MC Longitudinal
Impact Parameter Resolution
After smoothing of the response function, the
final fit is more stable
There are less problematic bins
26
Effect of the resolution on the Primary Vertex
position
1) Are the reconstructed primary vertices of the
MC (used to estimate the response function)
consistent with DATA?
From Y.Gao https//twiki.cern.ch/twiki/bin/view/C
MS/PrimaryVertexResolution
2) Anyway, how much the Impact Parameter
Resolution measurement is sensitive to the vertex
smearing? See next slides..