Secondary Vertex reconstruction for the D

1
Secondary Vertex reconstruction for the D

• Elena Bruna, Massimo Masera, Francesco Prino
• University of Torino and INFN

ALICE Physics Week Erice, Dec. 6th 2005
2
Outline

• Motivations for a good secondary vertex
  reconstruction capability
• Three different methods to find the secondary
  vertex for D ? K-pp
• Comparison between the methods ? find the
  candidate for the D analysis
• Future plans

3
Why D ? K-pp ?
drawbacks
Advantages

• Combinatorial background for this 3-body channel
  is larger than for D0 ? K-p.
• The average PT of the decay product is softer (
  0.7 GeV/c compared to 1 GeV/c for the D0)

• D has a long mean life (311mm compared to
  123 mm of the D0)
• D ? K-pp can be fully reconstructed in the
  detector
• D ? K-pp has a relatively large branching
  ratio (BR9.2 compared to 3.8 for D0 ? K-p).

• The signal selection strategy is based on
• Good secondary vertex reconstruction capability
  (c? (D) 300mm ? resolution of 200mm would be
  bad, 50mm would be a dream)
• Efficient system of cuts to discriminate the
  signal from the background
• On the single track (PT,d0,)
• On the signal candidates (invariant mass,
  distance between primary and secondary vertices,)

4
Simulation strategy
Our purpose exclusive reconstruction of D in
the ALICE barrel (Inner Tracking System employed
in the search for secondary vertices)
Too large statistics (108 events) would be
required to study the signal!!
Central Pb-Pb event (blt3.5 fm, dN/dy 6000,
vs5.5 TeV)
9 D/D- in ylt1
Signal and background events separately generated
with the Italian GRID

• 5000 signal events with only D decaying in Kpp
  (using PYTHIA)
• Check the kinematics and the reconstruction
• Optimize the vertexing algorithm
• 20000 background events (central Pb-Pb events
  using HIJING)
• cc pairs merged in addition in order to reproduce
  the charm yield predicted by NLO pQCD
  calculations ( 118 per event)
• Tune the cuts (impact parameter cut,) on the
  tracks to be analyzed by the vertexing algorithm
• Evaluate the combinatorial background

5
Straight Line Vertex finder

• Originally developed to find the primary vertex
  in p-p
• Based on the Straight Line Approximation of a
  track (helix)

• Main steps
• The method receives N (N3 in our case) tracks as
  input
• Each track is approximated by a straight line in
  the vicinity of the primary vertex
• An estimation of the secondary vertex from each
  pair of tracks is obtained evaluating the
  crossing point between the 2 straight lines
• The coordinates of secondary vertex are
  determined averaging among all the track
• pairs

track1
track3
track2
6
Results with the Straight Line Vertex Finder
Straight Line
Improvements
RMS179 µm
X coord

• Add a cut (DCAcut) on the distance of closest
  approach (DCA) between the two straight lines
• negligible effect on RMS,
• cuts only 1 of good vertices
• to be tested on background events
• Use a weighted mean of the 2 points defining the
  DCA (take into account the errors on the tracks
  parameters)
• improves Z resolution by 6mm

Finder- MC (mm)
Y coord
RMS183 µm
Finder- MC (mm)
RMS166 µm
Z coord
Finder- MC (mm)
7
From Straight Line to Helix
Straight Line
Helix
RMS169 µm
RMS179 µm
Further development use of the track as helix,
without the straight line approximation (no
longer analytic, minimization is required)
X coord
Finder- MC (mm)
Finder- MC (mm)
Y coord
RMS171 µm
RMS183 µm
Result very small improvement (10mm) WHY?
Finder- MC (mm)
Finder- MC (mm)
RMS166 µm
Z coord
RMS162 µm
Finder- MC (mm)
Finder- MC (mm)
8
Straight Line Approximation
Straight Line Approximation
Decay dist 300 mm
track
d (µm)
Secondary vertex
d (µm) is the distance between the secondary
vertex and the tangent line
Primary vertex
PT (GeV/c)

• d is of the order of tens of nm
• the small differences between Helix and Straight
  Line Finder seem not to be due to the linear
  approximation, but rather to the fact that in the
  Helix method the errors on the track parameters
  are considered in the minimization

PT 0.5 GeV/c
d (µm)
Decay dist (µm)
9
New secondary vertex finder

• Straight Line Approximation used ? analytic
  method
• The 3 tracks are taken at the same time, no
  pairs of tracks

Vertex coordinates (x0,y0,z0) from minimization
of
Where d1,d2,d3 are the distances (weighted with
the errors on the tracks) of the vertex from the
3 tracks
P1 (x1,y1,z1)
SecondaryVertex (x0,y0,z0)
sx sy
d1
10
Resolution of the vertex finder
RMS x
RMS y
At high Pt of D (Ptgt5-6 GeV/c), the RMS in the
bending plane increases, instead of going down to
15µm (spatial pixel resolution) as expected.
RMS z
New method improves RMS of 40µm for D Pt
2GeV/c for x, y and z with respect to previous
Helix vertex finder based on DCA of pairs of
tracks.
11
Resolution at high Pt
In the signal events, as the Pt of the D
increases, the daughters become more and more
co-linear, resulting in a worse resolution along
the D direction.
p
p
K-
bending plane
D
12
Resolution in the rotated frame
Along the Pt of the D (x coord.)
Orthogonal to the Pt of the D (y coord.)
? Along the Pt of the D as Pt increases (for
Ptgt5-6 GeV/c) the angles between the decay
tracks become smaller in this coordinate the RMS
increases ? Orthogonal to the Pt of the D the
RMS decreases as expected

• Checks with events only made of pions show that
  the RMS on the bending plane
• Decreases down to 50 µm if the 3 tracks have Pt
  2 GeV/c
• Reaches a value of 20 µm (in agreement with
  spatial pixel resolution) if the 3 tracks have Pt
  100 GeV/c

13
Track dispersion cut on fSigma
(for X coordinate)
Accepted Vertices / Tot Vertices (True vertices)
Fake vertices (tracks coming from 3
different D vertices)

• fSigma lt 0.7 cm cuts 1 of the events and gives
  a RMS of 115 µm

• fSigma lt 0.4 cm cuts 30 of the events and
  gives a RMS of 100 µm

14
Conclusions on the finders

• The Straight Line vertex finder
• DCA cut negligible effect on the RMS of the
  residual distributions, slightly reduced number
  of good vertices (1)
• The use of a weighted mean improves Z resolution
  by 6 mm
• Cutting on the dispersion fSigma improves the
  resolution (by 30mm)

• The Helix vertex finder
• Has better resolution w.r.t. Straight Line
  finder (by approximately 10 mm)
• DRAWBACK the DCA between helices is obtained by
  minimization
• Weighted mean improves Z resolution by 8 mm
• fSigma cut improves the resolution (by 30 mm)

• The Minimum Distance vertex finder
• Has better resolution w.r.t. Helix finder (by
  approximately 40 mm)
• Has less overflows and underflows w.r.t.
  previous finders
• Is an analytic method
• Weighted mean and fSigma cut improve the
  resolution (by 20mm)
• Is presently THE candidate for first D analysis

A cut on fSigma has to be tuned (it can be done
at analysis level)
15
Future Plans Tuning the cuts
GOAL tune the cuts on both signal and background
events and find the cuts giving the best S/B.
(S/B 11 was found for the D0?K-p)

• CUT TIPOLOGIES
• On the single tracks used to feed the vertexer
  (Particle Identification, pT, track impact
  parameter)
• ? reduce the number af all the possible
  combinations of track-triplets in a central Pb-Pb
  collision ( 1010 without any initial cut!!). It
  MUST be cut by 4-5 orders of magnitude before
  using the more time-consuming vertexer.
• In progress.
• Once the triplets are combined, additional cuts
  (invariant mass, Dalitz plots and, possibly,
  impact parameter based) are mandatory before
  using the vertexer. These cuts are done on the
  triplets.
• To be done.
• The third kind of cuts is applied on the quality
  of the secondary vertices found (vertex
  dispersion-fSigma, distance between primary and
  secondary vertices, pointing angle,)
• To be done.

16
Single track cuts
GOAL find a compromise between the number of
background triplets and the number of signals we
want to take
HOW for each triplet (both signal and bkg) a
loop on all the possible cuts (d0,Pt p,Pt K) is
done
Cut on the track impact parameter (d0)
Particle Id. given by the generation initial
approach
The number of BKG triplets is reduced by a factor
of 100 when doing the cut on the Invariant Mass
within 3s
17
Backup slides
18
Track dispersions/1
?R dist (Vertex FOUND Vertex MC)
?R lt 1000 µm
1000lt?R lt3000 µm
3000lt?R lt5000 µm
?R gt 5000 µm
fSigma bigger for not well found vertices
19
Resolution at high Pt /1

• Checks with events only made of pions show that
  the RMS on the bending plane
• Decreases down to 50 µm if the 3 tracks have Pt
  2 GeV/c
• Reaches a value of 20 µm (in agreement with
  spatial pixel resolution) if the 3 tracks have Pt
  100 GeV/c

3 pion vertex RMS in the bending plane vs. Pt
20
Before tuning the cuts /1
Reconstructed signal events D invariant mass
Mean
Integrated over PT
MEAN 1.867 GeV/c2 RMS 0.019 GeV/c2
this is not a complete
reconstruction of the signal tracks are grouped
by means of info. stored at generation time.
MINV Resolution (SIGMA of the gaussian fit)
Knowledge of MINV resolution vs PT is important
when selecting the signal candidates
21
Single track cuts /2
The number of BKG triplets is reduced by a factor
of 100 when doing the cut on the Invariant Mass
within 3s (see slide 37)
BkgTriplets
No cut on the track impact parameter (d0)
Cut on d0 ? lower cuts on Pt (useful up to Bkg
105)
Particle Id. given by the generation initial
approach
22
Tuning the single track cuts
When tuning a cut, one has to keep in mind how
the Pt distribution of the D is modified
Pt reconstructed D Mean2.5 GeV/c
Pt reconstructed D Pt cut (p) 0.75 GeV/c Pt
cut (K) 0.6 GeV/c Mean4.7 GeV/c
Ratio With cut / Wo cut