1. Sinéad Farrington 8th December 2014
Higgs to tt at ATLAS
Sinéad Farrington
8th December 2014

2. The Higgs Boson Professor Peter Higgs
Emeritus Professor at Edinburgh
Also Brout, Englert, Kibble, Guralnik,
Hagen
Devised a mechanism to account
for the generation of mass
Predicts one new particle, the Higgs boson
Specifically to give mass to W/Z bosons
Yukawa couplings allow the same
particle to give mass to up- and down-type
fermions
Unseen until 2012

3. How to look for Higgs at the LHC?
We didn’t know the Higgs Boson’s mass (not predicted directly by the theory)
Very different composition of PRODUCTION and DECAY mechanisms depending on mass

4. Were there any clues? Most likely Higgs mass: 95+30-24 GeV
Yes!
Most likely Higgs mass:
GeV
(from indirect evidence)
Mass > 115 GeV
(direct evidence until 2012)

5. Many ways to search for the Higgs
PRODUCTION
DECAY
Most likely mass ranges

6. Standard Model Higgs Production

7. New Boson: Status until Nov 2013
Observed by its decay to ZZ*, gg, WW* bosons (CMS and ATLAS)
Combined mass from ZZ, gg: 125.5± GeV
Spin/CP measurements agree with SM expectation of JP=0+

8. New Boson: Status until Nov 2013
Evidence for Vector Boson Fusion and gg fusion production
Signal strength m=s/sSM
All consistent with 1
CMS data gives the same picture
Properties are compatible with SM Higgs Boson

9. New Boson

10. Nobel Prize
Prize motivation: "for the theoretical discovery of a mechanism that contributes to our understanding of the origin of mass of subatomic particles, and which recently was confirmed through the discovery of the predicted fundamental particle, by the ATLAS and CMS experiments at CERN's Large Hadron Collider”
Today’s seminar is about the search for the Higgs boson decaying to tau lepton pairs
Another step in this “confirmation”

11. New Boson: Decay to Fermions
Status until Nov 2013 (evaluated at 125 GeV)
Tevatron H to bb: 2.8 s
CMS H to bb: 2.1 s
CMS H to tt: 2.85 s
ATLAS H to tt: 1.1 s
Search for Higgs to fermions decay important part of knowing whether we have seen the SM Higgs
Does the New Boson couple to fermions?
Indirect evidence from gg fusion through top loop
Furthermore: Couple to leptons?
If yes, are we sure the same particle is responsible for boson and all fermion decays?
Yukawa
PRL 109, (2012)
CMS-HIG
CMS-HIG

12. Standard Model Higgs (at 125 GeV)

13. H tt ne/m nt nt h+ (K/p) h- h+ Lepton-lepton 12.4%
Lepton-hadron 45.6%
Hadron-hadron 42.0% H t+ t-
e-/m-
ne/m nt nt
h- (K/p) h- h+
Perform the search in all combinations of decays
Involves all lepton identification methods
Additionally for the Vector Boson Fusion mechanism, require jets
Neutrinos lead to missing energy (MET)
Complex signatures!

14. ATLAS dataset High pile-up conditions, challenging environment
Analysed
20.3 fb-1 of 8 TeV data
4.3 fb-1 of 7 TeV data

15. Stability of electron ID
Efficiency of electron identification quite stable versus number of primary vertices

16. Stability of Hadronic tau ID
Hadronic tau’s identified by multivariate method (boosted decision tree)
Shower shape, decay length, etc
Tau Energy Scale (TES) derived from Z tt mass distribution

17. Missing Transverse Energy
Multiple neutrinos in ditau decays
MET resolution is an important aspect of mass reconstruction

18. ATLAS H to tt Analysis
Does the same boson observed to decay to WW*, ZZ*, gg, couple to t leptons?
Try to answer this with a multivariate analysis (BDT)
Data blinded
BDT trained to distinguish SM Higgs signal samples from backgrounds

19. Triggers and preselection
Lepton-lepton
Single and di-lepton triggers
N(lepton)=2, N(jet pt>40GeV)≥1
Mll and MET cuts to suppress Drell-Yan and multijet
Lepton-hadron
Single lepton triggers
N(lepton)=1, N(tau)=1
MT<70 GeV cut to suppress W+jets
Hadron-hadron
Di-tau triggers
N(tau)=2
MET>20GeV, DR(tt) and Dh(tt) cuts suppress multijets
Apply preselection
Train BDT on remaining events
Validate background modelling on these events

20. Analysis Categories Vector Boson Fusion (54-63% of signal, rest is gg)
Two forward jets with leading pt>40-50 (30-35) GeV, Dh(jj)> 2
Boosted (gg fusion is ~ 71-74% of the signal, rest is gg,VH)
Pt(H)>100 GeV
Veto events with b-tags in lep-lep and lep-had
Suppress top background
In had-had use “rest” of events to constrain backgrounds

21. Backgrounds
Backgrounds estimated using data directly or MC normalised to control regions
Z tt: dominant background, modelled by data
Others: MC for Dibosons, H WW
Data normalisation for Z ee/mm and top
Fake e/m/t: W+jets, top, QCD multijet modelled by data

22. Z tt Background Embedding method Advantages
Harvest Z mm events from data
Replace the muons with simulated taus
Gives a hybrid Z tt event
Advantages
Take from data: MET resolution, pile-up, jets, Z kinematics, VBF W/Z backgrounds modelled in data

23. Backgrounds from “fakes”
Estimated from data
e or m fakes estimated from sample of anti-isolated leptons
Hadronic tau fakes estimated
In lep-had channel from sample with hadronic tau failing ID
In had-had channel from events which do not have opposite sign t’s

24. Top Background
Shape from MC; normalisation from b-tagged control region
Normalisation performed separately for boosted/VBF categories
Validation regions defined to check shapes
Mll>100 GeV (lep-lep)
HT>350 GeV (lep-had)

25. BDT Input variables

26. Pre-fit steps Check modelling of all input variables
And the modelling of the correlations among them
Control regions are fitted simultaneously with signal regions to constrain
Z ee/mm + jets in lep-lep, lep-had
Top in lep-lep, lep-had
W+jets in lep-had
Fakes in lep-lep
QCD(multijet) in had-had
Fit performed in and 140+ GeV sidebands
Provides check of background model, especially Z tt

27. Di-tau mass
Mass reconstruction not straightforward, owing to neutrinos in the final state
Use likelihood method (Missing Mass Calculator, MMC) using all measured kinematics and their resolutions and tau mass constraint
This variable is included in the BDT, mass resolution:

28. Control regions

29. The Fit

30. Post-fit distributions
VBF
lep-lep
lep-had
had-had
BOOSTED

31. Systematic Uncertainties
Signal strength m=s/sSM
Dominant theory uncertainty: matching, t and b quark treatment
Dominant expt uncertainty: background normalisations

32. Results
ATLAS observes significant excess of data events in high S/B region
Expected significance at 125 GeV is 3.5 s
Observed significance at 125 GeV is 4.5 s

33. Results
ATLAS observes significant excess of data events in high S/B region
Expected significance at 125 GeV is 3.5 s
Observed significance at 125 GeV is 4.5 s

34. Compatibility with 125 GeV
Weight each event by ln(1+S/B) for corresponding bin in BDT score
Excess is consistent with SM Higgs at 125 GeV
Signals at 110, 125, 150 are shown for the best fit m at 125 GeV

35. Crosscheck Analysis
Cut based analysis performed as a crosscheck (8 TeV data only)
Expected significance: 2.5σ
Observed significance 3.2σ

36. Couplings
Signal seen in all channels and both production mechanisms

37. Couplings
Consistent with SM within one sigma

38. ATLAS Channels Combine this picture with the ATLAS H mm result
Expected limit 8.2xSM
Observed 9.8xSM
If the Higgs coupled universally to leptons, we would have already observed H mm !
So we know that Higgs couples to fermions, but not universally

39. Summary
ATLAS has observed evidence for decay of a particle consistent with the SM Higgs boson
4.5 standard deviation significance
CMS also produced evidence at a similar time (3.4 s)

40. Outlook Run 2 will yield Higher luminosity and energy
Higgs cross section increases:
But Z cross section only increases by ~1.8x
Challenges for triggering
Spin/CP measurements in fermions
H bb observation?
H mm observation?
Production mode
ggF
VBF VH
ttH
s(14TeV)/
s(8TeV)
2.6
2.1
4.7

41. VBF Higgs to tt?

42. H to tt H to tt is the newest of the evidence modes at ATLAS and CMS
Projections have been made by both experiments extrapolating analyses to the future
CMS evaluate two scenarios:
1: leave systematic uncertainties the same
2: Halve theory uncertainty; scale others by luminosity
Luminosity (fb-1)
Dm (%) [scenario 1,2]
300
[8,14]
3000
[5,8]

43. H tt ATLAS recent result uses Boosted Decision Tree
ATLAS-CONF
H tt
ATLAS recent result uses Boosted Decision Tree
Perform projections from a
simple cut based analysis
Assume no improvement in
theory uncertainty(!)
Assume experimental challenges
(pile-up, trigger) compensate for
increased signal:background cross section
Pessimistic?
300fb-1
3000fb-1
Uncertainty
All
No theory
Dm/m
0.22
0.16
0.19
0.12

44. Higgs seen at CERN

45. Many ways to search for the Higgs
PRODUCTION
DECAY
Most likely mass ranges

46. H tt nt ne/m h+ t+ nt h- h+ (K/p) nt t- ne/m h+ nt h- h- (K/p) e+/m+

47. H tt Experimental signature Experimental challenges (significant)
Electron or muon with neutrinos (missing energy)
Electron or muon identified fairly cleanly
Hadrons
Large rate for tau leptons to
decay this way
Experimental challenges (significant)
Difficult to differentiate these signatures from backgrounds
Production of generic jets of hadrons
Z+jet production, W+jet production, pairs of top quarks

48. H tt challenges
Background sources calibrated with several control regions

49. Future
Key properties of this new boson will take some time to ascertain
This was always anticipated
In fact we are fortuitous in nature’s choice for the Higgs mass – all decay modes are accessible at this point
Key to characterising this particle are
Production and decay rates
Spin: first measurements made public last week!
Mass (to greater precision)
Switch from search mode to precision physics

50. What does a Higgs event look like?ET
t 
e+/m+
ne/m nt nt
h+ (K/p) h- h+ H t+ t-
Distinctive signature
Reconstruct each element
e-/m-
ne/m nt nt
h- (K/p) h- h+

51. Jets
t 
t 
Jets are an important part of VBF signature