BID Status Report part I

1
BID Status Report part I
S?bastien Greder, Lorenzo Feligioni
On behalf of bid group
Taggers Description -CSIP, JLIP, SVX, SLT
- definition positive/negative tag
Taggability V0 removal Mistag Rate -
neg.tag rate in Data and Monte-Carlo corrections
B-tagging efficiency - Methods - B-tagging
efficiency in Data - B-tagging efficiency in
MC - Data/MC Scale factor Available Tools
2
Tagger Description Introduction
- The b quark has a long lifetime 1.5 ps -
Can fly over 1-2 mm. Lifetime based taggers
CSIP, JLIP, SVX based either on tracks
impact parameter or vertex decay length -10
b decay produce a muon Muon based tagger
SLT
m
3
Counting Signed Impact Parameter CSIP (2)
L. Chabalina, R.Demina, A. Khanov,
F. Rizatdinova
Based on impact parameter significance S(IP)
(sometimes named distance of closest approach
(dca) significance) S(IP) IP/s(IP)
Note IP is a signed quantity w.r.t to jet axis
(so is S(IP)) - positive if q lt p/2
- negative if q gt p/2
Jet axis
Track
q
I.P
Primary vertex
4
Counting Signed Impact Parameter CSIP (2)
Requirements to tag a jet positively - at
least 2 tracks with S(IP) gt 3 - or at least 3
tracks with S(IP) gt 2 Requirements to tag a jet
negatively - at least 2 tracks with S(IP) lt
-3 - or at least 3 tracks with S(IP) lt -2
5
Jet Lifetime Impact Parameter JLIP (1)
D. Bloch, B. Clement, D. Gele, S.Greder, I.
Ripp-Baudot
Fit a resolution function (R) on the negative
S(IP) distribution. Assume these tracks to
originate from primary vertex. Use then R to
define a probability for a track to originate
from the primary vertex
6
Jet Lifetime Impact Parameter JLIP (2)
Combine then each track's probability to compute
the probability for N tracks to originate from
primary vertex Pjet the jet lifetime
probability (tracks are required to be in dR
0.5 cone around jet axis)

Pjet(-) p(-) 5 ?j1Ntrk (-logp)(j-1)/(j-1)!
With p(-) Õj1Ntrk Ptrk(siglt0gt0)
Requirements to tag a jet positively - Pjet
() lt cut Requirements to tag a jet negatively
- Pjet (-) lt cut
7
Secondary vertex tagger SVX (1)
L. Feligioni, M. Narain, A. Schwartzman, P.
Schieferdecker
-Build up track-based jets and fit their
tracks to a secondary vertex -Select tracks with
high IP to build secondary vertices -Tracks
with high IP will form vertices with high
decay length significance S(Lxy)
Lxy/s(Lxy). (Like S(IP), S(Lxy) is a signed
quantity)
Track-based jet
Z axis
8
Secondary vertex tagger SVX (2)
Jet are then tagged by requiring a dR lt 0.5
matching between the jet axis and the secondary
vertex.
Requirements to tag a jet positively - match
a secondary vertex with S(Lxy) gt cut
Requirements to tag a jet negatively -
match a secondary vertex with S(Lxy) lt -cut
9
Soft lepton tagging SLT
K. Hanagaki, J. Kasper, J. Butler
- Defined "a la RunI" require a muon to be
matched within dR lt 0.5 to the jet axis no
Ptrel cut is required.
At this step try to mesure muon reconstruction
and track-match efficiencies by looking into
J/Psi-gt mumu signal in CSG sample. Muon are
required to have Pt gt 4GeV, h lt 2.

10
Taggability
A jet is required to be taggable before tagging
it ensures a "minimum" quality requirement
before tagging procedure
Definition - build-up track-based jets, dR
0.5, track Pt gt 0.5 GeV - require tracks to
have at least 1 smt hit (ladderFdisk) - at
least 1 track with Pt gt1GeV - track
dca(x-y) lt 0.2 cm - track dca(z) lt 0.4 cm
A calorimeter jet is taggable if if it
matched to a track-jet within a dRlt 0.5 cone.
Efficiencies, mistags in data and Monte-Carlo
are calculated w.r.t taggable jets.
11
V0 Removal procedure
Decays in flight like Kos , L, g conversions do
contribute into signal for impact parameter based
taggers
V0 removal algorithm is included into CSIP
package It makes use of all the tracks in
the event and flag them No-V0 flagged tracks are
then filtered out to feed the taggers
12
V0 Removal procedure
Fractions of V0 candidates without V0 removal in
jet trigger data normalized to taggable jet
(left) and tagged jets (right)

• All
• Kos
• g conversions
• L

13
Mistags Definition and origin in MC
Mistag is defined as the light quark tagging
efficiency
The origin of tracks in light jets has been
scrutinized (Sasha Khanov) in
Wlight jets events
The main contributions come from tracks
originating from the primary vertex and fakes
(defined as reconstructed track not matched with
a mc track within 10s of all tracks' parameters)
Negative tags
Positive tags
14
Mistags how to get them in data ?
Heavy flavour is most likely to be positively
tagged forget about using positive tags !
On the other hand light flavour has a nearly
symmetric distribution
That's nice, take negative tag rate in data as
mistags !
BUT unfortunately reality is far from being as
simple as Monte-Carlo because
if(jet-gtGetFlavour() "light") doesn't work in
data ! WHY - Heavy flavour
contaminates also negative tags. -
Long lived particles contributes in positive tags
and are thus missing in negative
tags
15
Mistags corrections factors
Need to correct data negative tag rates to get
the estimated light tagging efficiency in data
Scaling factor to correct for the heavy flavour
contamination in negative tags
Missing contribution from long lived particles
Both scaling factors are evaluated in Monte-Carlo
16
Mistags examples plots
SVX
SFhf
SFll
JLIP
17
Mistags estimated light tagging efficiency
Shown here for CSIP as predicted from EMQcd
sample for 4 working points
This prediction parametrizations have to be used
to reweight Monte-Carlo light jets