Transcript Top Mass Measurements at CDF

The Top Quark
The Top Quark Mass
An Important thing to
know.
7/16/2015
B. Todd Huffman - Oxford
1
The Top Quark
 The top quark was
discovered only 10 years ago
The Standard Model
 Existence is required by the
SM, but striking
characteristics: its mass is
surprisingly large
 Studied only at the Tevatron
Particle Masses
t
Z W
7/16/2015
b
c
s
d
u


B. Todd Huffman - Oxford
e
  e
2
Why measure the Top Quark Mass?
 Related to standard model
observables and parameters
through loop diagrams
Summer 2005
 Consistency checks of SM
parameters
 Precision measurements of
the Mtop (and MW) allow
prediction of the MHiggs
 Constraint on Higgs mass can
point to physics beyond the
standard model
7/16/2015
B. Todd Huffman - Oxford
3
Dilepton Channel
Final State from Leading Order Diagram





What we measure
Branching fraction: 5% (lepton = e or )
Final state: 2 leptons, 2 b quarks, 2 neutrinos
Combinatorial background: 2 combinations
2 neutrinos: under constrained, kinematically complicated to solve Mtop
S:B = 2:1 and 20:1 requiring 1 identified b tag
7/16/2015
B. Todd Huffman - Oxford
4
Lepton+Jets Channel
Final State from Leading Order Diagram





What we measure
Branching fraction: 30% (lepton = e or )
S:B = 1:4 to 11:1 depending on the b-tagging requirement
Combinatorial background: 12 (0 b tag), 6 (1 b tag), and 2 (2 b tags)
1 neutrino: over constrained
Most precise Mtop measurements
7/16/2015
B. Todd Huffman - Oxford
5
All Jets Channel
Final State from Leading Order Diagram




What we measure
Branching fraction: 44%
Huge amount of background S:B = 1:8 after requiring at least 1 b-tag jet
Combinatorial background: 90 combinations
Backgrounds mainly from multi-jet QCD production
7/16/2015
B. Todd Huffman - Oxford
6
Top Quark Mass at CDF
 Robust program of top quark mass measurements
 Many measurements in all the different channels ->
consistency
 Different methods of extraction with different
sensitivity -> confidence
 Combine all channels and all methods -> precision
7/16/2015
B. Todd Huffman - Oxford
7
Mtop
 Different statistical and systematical sensitivities in each channel
CDF Measurement
(~350pb-1)
Statistical
(GeV/c2)
Jet En.
Scale
(GeV/c2)
Other syst.
(GeV/c2)
Dilepton
6
3
2
Lepton+Jets
4
2
1
All Jets
5
4
2
 Other sources arise from the assumptions employed to infer Mtop:







Initial state and final state radiation
Parton distribution functions
b-jet energy scale
Generators
Background modeling and composition
b-tagging efficiency
MC statistics
 Systematics dominated by the uncertainty on parton energies (Jet
Energy Scale, JES)
7/16/2015
B. Todd Huffman - Oxford
8
Jet Energy Scale
 Jet energy scale
 Determine the
energy of the quarks
produced in the hard
scattered
 We use the Monte
Carlo and data to
derive the jet energy
scale
 Jet energy scale
uncertainties
 Differences between
data and Monte Carlo
from all these effects
7/16/2015
B. Todd Huffman - Oxford
9
In-situ Measurement of JES
 Additionally, we use Wjj mass resonance (Mjj) to measure the jet
energy scale (JES) uncertainty
Mjj
Constrain the
invariant mass
of the non-btagged jets to
be 80.4
GeV/c2
Measurement of JES scales directly with statistics!
7/16/2015
B. Todd Huffman - Oxford
10
Data and Monte Carlo
W-jet pT
b-jet pT
ttbar pT
7/16/2015
Mttbar
B. Todd Huffman - Oxford
11
In this Talk
 Lepton + jets: Template analysis
 Lepton + jets: Matrix Element
 Lepton + jets: Decay Length
 All Hadronic: Ideogram
 Missing Di-leptons…cannot go into
detail on everything!
7/16/2015
B. Todd Huffman - Oxford
12
Mtop Lepton+Jets Results
Detected Top Candidate
Silicon Detector
Results
7/16/2015
B. Todd Huffman - Oxford
14
Top Mass - Guessing Jets
7/16/2015
B. Todd Huffman - Oxford
15
Top Mass - Templates
Run Monte carlo with
various mass hypotheses.
These are used as
‘templates’ that can be
compared to data using the
c2 difference between data
and the template.
7/16/2015
B. Todd Huffman - Oxford
16
Top Mass – Background Templates
Use W+Jet background
with fake electron and
mistagged b to check that
jet shapes are OK in MC.
Then use MC to generate
a ‘background’ mass plot.
7/16/2015
B. Todd Huffman - Oxford
17
Template Analysis
 Reconstructed mtop and mjj from data are compared to templates
of various true Mtop and JES (jet energy uncertainty shift) using
an unbinned likelihood
 Uses all four samples to increase sensitivity
2 b tags
1 b tag (T)
1 b tag (L)
0 b tag
10:1
4:1
1:1
0.6:1
Expected Number of
Events ( tt = 6.1pb)
47
104
64
no a priori
estimate
Data (680 pb-1)
57
120
75
108
Expected S:B
7/16/2015
B. Todd Huffman - Oxford
18
Template Analysis Results
 Using 360 candidate events in 680 pb-1 we measure
Mtop  173.4  1.7 (stat)  1.8 (JES)  1.3(syst) GeV / c2
 Using in-situ JES calibration results in 40% improvement on JES

Better sensitivity than the previous world average!
7/16/2015
B. Todd Huffman - Oxford
19
Matrix Element Analysis Technique
 Optimizes the use of kinematic and dynamic information
 Build a probability for a signal and background hypothesis
 Likelihood simultaneously determines Mtop, JES, and signal
fraction, Cs:
L(Cs , Mtop , JES) 
7/16/2015
Nevents
 (Cs Psignal,i (Mtop , JES)  (1 Cs ) Pbackground,i (JES))
i 1
B. Todd Huffman - Oxford
20

Matrix Element Analysis Technique
 For a set of set measured variables x:
W(x,y) is the probability that a parton level
set of variables y will be measured as a set
of variables x (parton level corrections)
dn is the differential cross section: LO Matrix
element
P(x; M top , JES) 
1

 d (y; M
n
top
) dq1 dq2 f(q1 ) f(q2 ) W(JES, x, y)
f(q) is the probability distribution than a parton will have a momentum q
 JES sensitivity comes from W resonance –this too is in the fit.
 All permutations and neutrino solutions are taken into account
 Lepton momenta and all angles are considered well measured
W(JES, x, y)   (p
3
y
lepton
p
4
x
lepton
4
) Wjet (JES E , E )  2 (iy  ix )
x
j
j 1
y
j
i 1
 Background probability is similar, no dependence on Mtop
7/16/2015
B. Todd Huffman - Oxford
21
Cross-check Monte Carlo with Data
 Compare Data and Monte Carlo calculating the invariant mass of 2
and 3 jets
 Signal probability evaluated at Mtop=174.5 GeV/c2 and JES=1 and
using the most probable configuration
Excellent agreement found between data and Monte Carlo
7/16/2015
B. Todd Huffman - Oxford
22
Results
 Using the 118 candidates in 680 pb-1 our Mtop is:
Mtop  174.1 2.0 (stat)  1.5 (JES)  1.3 (syst) GeV / c2
with JES = 1.019  0.022 (stat)
Better sensitivity than the previous world average!
7/16/2015
B. Todd Huffman - Oxford
23
New technique – B Decay Length
 The method has been published by C. Hill et al. at PRD 71, 054029
 B hadron decay length  b-jet boost  Mtop
 Uses the average transverse decay length of the b-hadrons <Lxy>
Relies on tracking, no JES and uncorrelated with other measurements
7/16/2015
B. Todd Huffman - Oxford
24
Mtop All Jets Results
All Jets
 Main challenges in this channel:
 Small signal fraction S:B = 1:8 after requiring at least 1 identified b-jet
 Large combinatorial background: 90 combinations
 Selection and events




ET/  ( ET) < 3 (GeV)1/2
ET  280 GeV
nb-tag  1
Exactly 6 jets
Expected
Events (310 pb-1)
Multi-jets (light)
182
Multi-jets (heavy flavor)
68
Total background
240
tt (6.1 pb)
40
Data
290
 Ideogram method from the Delphi experiment for the W mass
measurement, used in a Run II preliminary D0 for top mass
measurement in lepton+jets channel
7/16/2015
B. Todd Huffman - Oxford
26
Ideogram Overview
 Define a 2D event likelihood as
90


L event (Mtop , Cs )   wi CsSignal  (1 Cs )Background
i 1
 Weight each combination with kinematical and b-tagging

information:
wi
 Extract from kinematical fit to mtop and manti-top  m1, m2, 1,2, c2
Signal(m1i , m2i , 1i , 2i , Mtop )  pmSm  (1 pm )Scomb

Sm calculated convoluting Briet-Wigners
and Gaussian resolution functions
Scomb combinatorial background from
Monte Carlo
7/16/2015
B. Todd Huffman - Oxford
27
Template Shapes
Template for combinatorial background
Template for background
Signal, correct combination
Using the two fitted masses
gives a good separation
between
signal and background
7/16/2015
B. Todd Huffman - Oxford
28
Results
 Using 310 pb-1 and 290 candidates we measure
Mtop  177.1 4.9 (stat)  4.3 (JES)  1.9 (syst) GeV / c2
 First Tevatron Run II all jets Mtop measurement
 Systematically limited!

Similar statistical sensitivity as the lepton + jets channel
7/16/2015
B. Todd Huffman - Oxford
29
Summary of Mtop Results
CDF Measurement
Extracted value
(GeV/c2)
Statistical
(GeV/c2)
JES
(GeV/c2)
Other syst
(GeV/c2)
Lepton+Jets: Template
(680pb-1)
173.4
1.7
1.8
1.3
Lepton+Jets: ME(680pb1)
174.1
2.0
1.5
1.3
Lepton+Jets: Decay
Length (750pb-1)
183.9
+15.7
-13.9
0.3
5.6
Dilepton: Matrix
Element (750pb-1)
164.5
4.5
2.6
1.7
All Jets:
Ideogram (310pb-1)
177.1
4.7
4.3
1.9
We compare (confidence and consistency) and combine (precision)
7/16/2015
B. Todd Huffman - Oxford
30
Combining Mtop Results
 Excellent results in each
channel
 Combine them to improve
precision
 Include Run-I results
 Account for correlations
 Use BLUE (NIM A270 110, A500
391)
 Matrix Element analysis in
lepton+jets not yet included.
Working to understand
statistical correlations with
Template analysis
CDF April’06
2.6
(750 pb-1)
Mtop (CDF)  172.4  2.6 GeV / c2
c 2 / dof  6.2 / 6
7/16/2015
B. Todd Huffman - Oxford
31
Future
 We surpassed our Run II
goal of measuring to 3
GeV/c2 precision
 Have made
extrapolations based on
present methods
 Upper limit: Only (stat)
improves with luminosity
 Lower limit: Everything
improves with luminosity
 Reality: likely somewhere in
between
With full Run-II dataset CDF should measure Mtop to < 1%
7/16/2015
B. Todd Huffman - Oxford
32
Conclusions
 What has been shown.
 First Run-II determination in all-jets channel
 Multiple methods and channels
 Observed consistency builds confidence

CDF combined
Mtop (CDF)  172.4  2.6 GeV/c2
 CDF should reach 1% precision with full Run-II data set


Tevatron combination better still
Mtop (Tevatron)  172.5  2.3 GeV/c2
7/16/2015

B. Todd Huffman - Oxford
33
Conclusions
 Present uncertainties on Mtop and MW help constrain MHiggs to
about 40% MHiggs/ MHiggs
 Best fit favors light MHiggs
 where CDF/D0 are sensitive
 where difficult for LHC
 Mtop will continue to shrink
 New CDF/D0 MW expected soon...
 MW will also shrink
We'll continue to squeeze SM, will it hold?
7/16/2015
B. Todd Huffman - Oxford
34