top quark mass measurements at hadron colliders
DESCRIPTION
Top Quark Mass Measurements at Hadron Colliders. G. Watts (UW/Seattle, CPPM). For the DZERO, CDF, CMS, and ATLAS collaborations. July 15, 2014. The Top Quark. Just like other Fermions. Except:. The next heaviest quark!. The Mass gives the top quark a special role in the Standard Model. - PowerPoint PPT PresentationTRANSCRIPT
Top Quark Mass Measurements at Hadron CollidersG. WATTS (UW/SEATTLE, CPPM)
For the DZERO, CDF, CMS, and ATLAS collaborations
July 15, 2014
G. Watts (UW
/Seattle) FFP 2014 - Marseille
2The Top QuarkJust like other Fermions
Except:𝑚𝑡 40×𝑚𝑏
The next heaviest quark!
The Mass gives the top quark a special role in the Standard Model
• Only fermion which has a significant coupling to the Higgs• Plays key roll in many important physics processes
• Flavor physics, Electro-weak processes• It plays a special roll in a number of Beyond the Standard
Model theories as well
G. Watts (UW
/Seattle) FFP 2014 - Marseille
3The Top MassWe have known almost since it was discovered.
By far the most precisely measured quark mass!
While it behaves like any other quark in the Standard Model, its mass gives it a unique role.
• Only version for which the coupling to the Higgs is important• Stability of the SM Higgs
potential at high scalesA consistency check for the
Standard Model!• Shows up in a number of production
loops• at the LHC contains a top loop• Heavy Flavor physics (e.g. )
production
G. Watts (UW
/Seattle) FFP 2014 - Marseille
4Top Mass Is A Precision
Measurement
Each measurement deserves at least a
seminar
I have chosen a few extra results
Current World Average: 173.3 GeV.Known to better than 0.5 %!!
Higgs mass is known to better than 0.3%
Top is easier to discover: at TeV at TeV
No clean easy to see peak l!All final states involve jets
Top is harder to reconstruct:
G. Watts (UW
/Seattle) FFP 2014 - Marseille
5Te
vatro
nLH
C
G. Watts (UW
/Seattle) FFP 2014 - Marseille
6Decays
Dilepton eventsClean, but low statistics~4%
Lepton + Jet eventsGood compromiseReasonable background~30%
All Hadronic eventsHuge multi-jet background~44%
Top mass has been measured in all decay
channels.
𝑡 𝑡→𝑊+¿𝑏𝑊 −𝑏¿
Classified by the Ws’ decay
G. Watts (UW
/Seattle) FFP 2014 - Marseille
7The Tevatron & The LHCThe Tevatron is coming out with its final results
• of data at TeV• Well understood detector• Sophisticated analysis techniques
The LHC is just coming online in the world• TeV results well developed• 8 TeV results just appearing• Statistics are much better due to the much higher
The much larger statistics will eventually open the door to new measurement techniques.
G. Watts (UW
/Seattle) FFP 2014 - Marseille
8Extracting from Data
• Does not always give you 4-vectors (neutrinos!)• Detector/Object resolutions (e.g. Jet Energy Scale)• Background contamination• Incorrect reconstruction (e.g. bad jet assignment)• Top mass width• Etc.
Two common methods to address this:
Matrix Element Uses all the informationComputationally very expensive
Template Method Flexible, subsets the information used“Fairly easy” to implement
Detector gives you 4-vectors. Use Griffiths!
What do we measure? The Pole mass? The MC mass?
G. Watts (UW
/Seattle) FFP 2014 - Marseille
9The Jet Energy ScaleCommon curse for all methods
• Experiments normally measure in independent control sample.
• Resolution not good enough for a state-of-the-art top mass measurement.
In situ Jet Energy Scale measurement
𝑊→𝑞𝑞 ′Two poorly measured
objectsOne very
well measured
object
Many techniques will constrain to be as part of the global fitting process.
Global fit over the full sample• Scale all jets by a constant
factor to achieve constraint
Lepton+Jets
Flavor Jet Energy Scale
G. Watts (UW
/Seattle) FFP 2014 - Marseille
10The Matrix Element
ApproachA reverse Monte Carlo
MC Generates
100K events
Distributions of kinematic
variables for all objects
“Map of kinematic
phase space”
Turn that aroundGiven a single event in data, how dense a
part of kinematic phase space is it in?
Repeat for all major backgrounds and signal:
𝑃
G. Watts (UW
/Seattle) FFP 2014 - Marseille
11ME – Multiple Steps
ALPGEN + Pythia
Detector Simulation
Reconstruction
4 vectors of reconstructed objects
𝑃 (𝑚𝑡𝑜𝑝 )= 1𝜎𝑜𝑏𝑠
𝑡 𝑡 (𝑚𝑡𝑜𝑝 )∑𝑖=1
24
𝑤 𝑖
Normalization
Sum over all possible jet assignments• Which jet is the first tops?• Which jets belong to the
W?
A weight reflecting the probability of those jet assignments• -tagging
probabilities
G. Watts (UW
/Seattle) FFP 2014 - Marseille
12ME – Multiple Steps
ALPGEN + Pythia
Detector Simulation
Reconstruction
4 vectors of reconstructed objects
𝑃 (𝑚𝑡𝑜𝑝 )= 1𝜎𝑜𝑏𝑠
𝑡 𝑡 (𝑚𝑡𝑜𝑝 )∑𝑖=1
24
𝑤 𝑖∫𝑑𝜌 𝑑𝑚12𝑑𝑀1
2 𝑑𝑚22𝑑𝑀 2
2𝑑𝜌 ℓ𝑑𝑞1𝑥𝑑𝑞1
𝑦 𝑑𝑞2𝑥𝑑𝑞2
𝑦
10 dimensional integral over phase space• Mass of the tops, W’s• Directions of the b-quarks• Lepton and neutrino direction
Note no mention of data 4-vectors yet!
G. Watts (UW
/Seattle) FFP 2014 - Marseille
13ME – Multiple Steps
ALPGEN + Pythia
Detector Simulation
Reconstruction
4 vectors of reconstructed objects
𝑃 (𝑚𝑡𝑜𝑝 )= 1𝜎𝑜𝑏𝑠
𝑡 𝑡 (𝑚𝑡𝑜𝑝 )∑𝑖=1
24
𝑤 𝑖∫𝑑𝜌 𝑑𝑚12𝑑𝑀1
2 𝑑𝑚22𝑑𝑀 2
2𝑑𝜌 ℓ𝑑𝑞1𝑥𝑑𝑞1
𝑦 𝑑𝑞2𝑥𝑑𝑞2
𝑦
∑𝑝 𝑎𝑟𝑡𝑜𝑛 𝑓𝑙𝑎𝑣𝑜𝑟𝑠 ,𝜈
❑
|𝔐𝑡𝑡|2
Sum over incoming parton flavorsAll neutrino solutions
The Leading Order Matrix Element• Given all the phase space
parameters• Weight for the kinematics
values• Uses all available
information• At leading order
G. Watts (UW
/Seattle) FFP 2014 - Marseille
14ME – Multiple Steps
ALPGEN + Pythia
Detector Simulation
Reconstruction
4 vectors of reconstructed objects
𝑃 (𝑚𝑡𝑜𝑝 )= 1𝜎𝑜𝑏𝑠
𝑡 𝑡 (𝑚𝑡𝑜𝑝 )∑𝑖=1
24
𝑤 𝑖∫𝑑𝜌 𝑑𝑚12𝑑𝑀1
2 𝑑𝑚22𝑑𝑀 2
2𝑑𝜌 ℓ𝑑𝑞1𝑥𝑑𝑞1
𝑦 𝑑𝑞2𝑥𝑑𝑞2
𝑦
∑𝑝 𝑎𝑟𝑡𝑜𝑛 𝑓𝑙𝑎𝑣𝑜𝑟𝑠 ,𝜈
❑
|𝔐𝑡 𝑡|2 𝑓 ′ (𝑞1 ) 𝑓 ′ (𝑞2 )
√ (𝜂𝛼𝛽𝑞1𝛼𝑞2𝛽 )2−𝑚𝑞12 𝑚𝑞2
2Φ6
PDF’s
Phase Space FactorTransverse
momenta of incoming partons
G. Watts (UW
/Seattle) FFP 2014 - Marseille
15ME – Multiple Steps
ALPGEN + Pythia
Detector Simulation
Reconstruction
4 vectors of reconstructed objects
𝑃 (𝑚𝑡𝑜𝑝 )= 1𝜎𝑜𝑏𝑠
𝑡 𝑡 (𝑚𝑡𝑜𝑝 )∑𝑖=1
24
𝑤 𝑖∫𝑑𝜌 𝑑𝑚12𝑑𝑀1
2 𝑑𝑚22𝑑𝑀 2
2𝑑𝜌 ℓ𝑑𝑞1𝑥𝑑𝑞1
𝑦 𝑑𝑞2𝑥𝑑𝑞2
𝑦
∑𝑝 𝑎𝑟𝑡𝑜𝑛 𝑓𝑙𝑎𝑣𝑜𝑟𝑠 ,𝜈
❑
|𝔐𝑡𝑡|2 𝑓 ′ (𝑞1 ) 𝑓 ′ (𝑞2 )
√ (𝜂𝛼𝛽𝑞1𝛼𝑞2𝛽 )2−𝑚𝑞 12 𝑚𝑞 2
2Φ6𝑊 (𝑥 , 𝑦 ;𝜌 , 𝜌 ℓ ,…)
Transfer Functions• Given a generated jet with what is the probability DZERO
will reconstruct values x and y?• Detector and reconstruction resolution
G. Watts (UW
/Seattle) FFP 2014 - Marseille
16DZERO using the ME
MethodIn used at DZERO since Run I
GeV3.6
GeV
Total error is equivalent to March world average!
3 years of work (old result):
• Use different top mass in the Matrix Elements
• Vary the Jet Energy Scale in the transfer functions
G. Watts (UW
/Seattle) FFP 2014 - Marseille
17What Did 3 years get?
• Speed (CPU) to allow better MC stats• X100 increase means MC stats error
drops from ~0.25 GeV to ~0.05 GeV.
• New Jet Energy Scale Calibrations• ISR modeling
• Constrain by studies in Drell-Yan data
• General modeling improvements
The variable is sensitive to Z boson recoil ().Gives an experimental bound to ISR mis-modeling
Systematic error on reduced from ~0.25 to 0.06 GeV
G. Watts (UW
/Seattle) FFP 2014 - Marseille
18Template Method
Using a distribution sensitive to :
Simulated sample at GeV
Simulated sample at GeV
Simulated sample at GeV
Use a likelihood to estimate template
compatibility
Make it for each sample
𝑚𝑡
Can do in two dimension• Jet energy scale• Top mass
G. Watts (UW
/Seattle) FFP 2014 - Marseille
19Top Mass In Dilepton
Events4% of all decays, split into , and .
Very little SM background!
CDF’s basic selection: Observe 520 events, expect 78% purityATLAS’ basic selection: Observe 2913, expect 96% purity
Really excellent top labExcept…
For 2 !!! There are no 4-vectors for the two!!
G. Watts (UW
/Seattle) FFP 2014 - Marseille
20Template Method
Need distributions that are strongly correlated
with the top massTemplate method to
figure out the top mass
ATLASThe average in the event
Two permutations (take smallest)Avoid the missing resolution
Good separatio
n power
G. Watts (UW
/Seattle) FFP 2014 - Marseille
21CDF Template Variables
Fully reconstruct the top massProblem: detector measures missing
There are not enough constraints to solve for solution!The weighting method
𝜙1
𝜙2
Grid in the azimuthal angles• Fit for the top mass at
each grid location.• Resulting is the template
variable.• Weight by fit .
The fit includes terms for:• All the measurements (2 leptons, two jets, missing )• Top mass and the (constrained) W mass
G. Watts (UW
/Seattle) FFP 2014 - Marseille
22Statistics Isn’t The
Problem… Broad peak, but decent separation power.
Leading systematic:Jet Energy Scale!
This measurement is statistics limited.Can something be done?
G. Watts (UW
/Seattle) FFP 2014 - Marseille
23Statistics Isn’t The
Problem… Broad peak, but decent separation power.
Leading systematic:Jet Energy Scale!
This measurement is statistics limited.Can something be done?
CDF creates a second template variable:
GeV • Depends on 4-vector of leptons• Direction of jets• No Jet Energy Scale, no Missing
And combines the two, optimizing for minimal error𝑀 𝑡
𝑓𝑖𝑡=𝑤 ∙𝑚𝑡𝑓𝑖𝑡+ (1−𝑤 ) ∙𝑚𝑡
𝑎𝑙𝑡
G. Watts (UW
/Seattle) FFP 2014 - Marseille
24Dilepton Top Mass
ResultsStandard Template MethodJet Energy Scale isn’t fit: not enough constraints
Statistics already making a big difference here
G. Watts (UW
/Seattle) FFP 2014 - Marseille
25Top Mass in All Hadronic
Decays (CDF & CMS)44% of all decays. Largest single decay class.
Overwhelmed by SM QCD background!
6 Jets
After CMS requires 6 jets4 jets with GeV5th with GeV6th with GeV
Estimated signal purity is 3%Signal Efficiency is 3.5%!
G. Watts (UW
/Seattle) FFP 2014 - Marseille
26Improving the Purity
Unique Handles:2 -jets
2 Look for -tagged jetsPerform kinematic fit:• Know • The two are the same2
(CDF)
Mass of the pairs of light quark jets
is the well measured value of 80.4 GeV
Mass of the pairs of light quark jets
free parameters
Every jet permutation is triedMinimum is kept
G. Watts (UW
/Seattle) FFP 2014 - Marseille
27Improving the Purity
1. Requiring the fit to converge2. Very basic cuts on the
Raise CMS’s purity to 39%
Additional kinematic selection
CMS: CDF: Neural Network
Raise CMS’s purity to 54%CDF has a purity of 57%
G. Watts (UW
/Seattle) FFP 2014 - Marseille
28Extracting the Mass
𝑚𝑡𝑟𝑒𝑐𝑜 𝑚𝑡
❑
The Template MethodFit for both Jet Energy Scale and
G. Watts (UW
/Seattle) FFP 2014 - Marseille
29Lepton + Jets From CMS
Full TeV result: Analysis is very similar to the All-Jets analysis from CMS
Initial selection is > 100K events and 94% pure QCD background is negligible!
A simple kinematic fit to clean up incorrect jet assignments
• Each possible jet assignment gives • Each is weighted by the fit probability
Largest systematic error is the flavor dependent Jet Energy Scale (0.41 GeV)
G. Watts (UW
/Seattle) FFP 2014 - Marseille
30Conclusions
Field is still rapidly evolving World average submitted in
March CDF dilptons and all-hadronic DZERO matrix element CMS all-hadronic and
lepton+jets What is next?
Tevatron will finish putting out “final” mass measurements
LHC’s statistics and purity mean it should quickly surpass the Tevatron.
LHC Run 2 projections Other measurements with the
quark mass Top and anti-top have
consistent masses measurements that can
clarify which mass we measure.
Becoming like the W mass…
If you believe BICEP2
!
G. Watts (UW
/Seattle) FFP 2014 - Marseille
31Awaiting the next world
Combination…
Current World Combination
Tevatron CombinationCMS Combination
±0.95 ±0.76 ±0.64
G. Watts (UW
/Seattle) FFP 2014 - Marseille
32
Systematic Errors
G. Watts (UW
/Seattle) FFP 2014 - Marseille
33ATLAS Lepton+Jets Template
G. Watts (UW
/Seattle) FFP 2014 - Marseille
34ATLAS dilepton 7 TeV CDF dilepton
G. Watts (UW
/Seattle) FFP 2014 - Marseille
35
CDF all jets CMS all jets
G. Watts (UW
/Seattle) FFP 2014 - Marseille
36CMS All Jets 7 TeV
G. Watts (UW
/Seattle) FFP 2014 - Marseille
37
Tevatron Combination
DZERO Lepton+Jets ME