Gregorio Bernardi / DØ-France/ October 14-2008
D0-France, October 14th 2008 Gregorio Bernardi, LPNHE-Paris for the CDF and DØ Collaboration
WH, High Mass, Combined Higgs Searches and Prospects at the Tevatron
• Introduction
• Standard Model Higgs Searches
• Combination techniques
• Combination results:
• Prospects
2Gregorio Bernardi / DØ-France/ October 14-2008
SM Higgs boson production
• gg fusion– Dominates at hadron machines– Usefulness depends on the Higgs decay
channel
• WH, ZH associated production– Important at hadron colliders since can
trigger on 1 or 2 high-pT leptons, Jets and MET
• t t H (and bbH) associated production
– High-pT lepton, top reconstruction, b-tag
– Low rate at the Tevatron
• Vector Boson Fusion
– Two high-pT forward jets help to “tag” event
– Important at LHC, first results at Tev!
H
3Gregorio Bernardi / DØ-France/ October 14-2008
SM Higgs Production and Decays
Production cross section (mH 115-180) in the 0.8-0.2 pb range for gg H in the 0.2-0.03 pb range for WH associated vector boson production
Dominant Decays bb for MH < 135 GeV WW* for MH > 135 GeV
Production Decays
Gregorio Bernardi / DØ-France/ October 14-2008
SM Higgs Low mass searches: datasets/methods
Missing ET 2 b-jets
Lepton Missing ET 2 b-jets
CDF DØ CDF DØ CDF DØ
Prel: 2.7 fb-1 1.7 fb-1 2.7 fb-1 2.1 fb-1 2.4 fb-1 2.3 fb-1
Pub: 1.0 fb-1 1.1 fb-1 1.0 fb-1 0.9 fb-1 1.0 fb-1 0.4 fb-1
Common to all analyses: b-tagging, Jet calibration & resolution, lepton-identification, Background cross-section
Differences: instrumental bckd, multivariate techniques
2 Leptons 2 b-jets
WH ZHbb ZHl l bb
5Gregorio Bernardi / DØ-France/ October 14-2008
Signal
u
dW*+
W+
H0
b
b
l+
l b
b
u
d
W+ l+
lLargebackground(also Wcc)
u
d
W+ l+
l
W+2 light jetwith one/two false b-tag
u
d
W*+
t
b
b
l+
l
W+
Single Top
b
b
u
d
W+ l+
l
Z0
“Diboson”
u
d
t
t
“ttbar”:Jets+leptonsfrom W decay
u
d
q
q
“Non-W from QCD”
and “Non-W from EW”,e.g. all Zbb processesIn which one lepton is lost
SM Higgs Searches @ Tevatron: WHlbb
Gregorio Bernardi / DØ-France/ October 14-2008
Selecting WlbbSelect events using W decay signatures
Require one high-pT leptons: pT > 15 GeV)
Neutrinos missing transverse energy
WHlbb: MET > 20 GeVReconstruct vector boson mass
Use “OR'ing” of muon triggers: 100% efficiency +15% in sensitivity.
Single EM or EM+jets for e-channel (95% eff)
Select >=2 high-pT, central jets as a first step towards a Higgs signaturepT>20 GeV,|eta|<2.5
Reweight W+jets ALPGEN to data in eta, phi, delta-eta and delta-phi of jets, before b-tagging
7Gregorio Bernardi / DØ-France/ October 14-2008
Neural Network b-tagger
Secondary Vertex Tagger
Jet LIfetime Probability
Counting Signed Impact Parameter
NNbtag
LOOSE(eff=70%,
fake=4.5%)
All Higgs analyses uses Neural Network b-tagging algorithm,
Combining best aspects of secondary vertex algorithm
statistical impact parameters algo
Asymmetric tagging: Tight tagging for Single TagLoose tagging for Double Tag Large improvement compared
to the individual taggers:Loose 72% b-tagging eff.
6% mistagTight 50% b-tagging eff. 0.3% mistag
B
Impact Parameter (dca)
Vertex Tagging(transverse plane)
Decay Lengh (Lxy)
Hard Scatter
(Signed) Track
Gain > 30% compared to our individual taggers (secondary vertex or impact parameter)
8Gregorio Bernardi / DØ-France/ October 14-2008
Dijet mass, GeV
WHl bb (l=e,): effect of b-tagging
Pre-tag Exclusive single tag (Tight)
Starting from a W+ 2 jet selection, apply NN_btagging:
Dijet mass, GeV
9Gregorio Bernardi / DØ-France/ October 14-2008
Dijet mass, GeV
WHl bb (l=e,): after b-tagging
Dijet mass, GeV
Exclusive single tag Double Loose tag
Higgs x10Backgrounds are measured one after the other (Wbb, Single
Top), WZ with Zbb remains the golden benchmark on which we can tune our analysis tools.
Starting from a W+ 2 jet selection, apply NN_btagging orthogonal samples
10
Gregorio Bernardi / DØ-France/ October 14-2008
WHl bb (l=e,): Neural Net and Limits
Use neural network to separate signal from backgroundFit the NN output
pT(j1)
pT(j2)
R(jj)
(jj)
pT(jj)
M(jj)
pT(l,MET)
NNwh
Future improvements (short term): include forward electrons and 3 jets sample. Improve NN with more backgd rejection and use Matrix Element approach
Higgs x10
11
Gregorio Bernardi / DØ-France/ October 14-2008
Summary of WH published Results
PRL’05:S. Beauceron’s Ph.D thesis
PLB’07:L. Sonnenschein’s Habilitation
PRL’08:J. Lellouch’s Ph.D thesis
In the making:4 fb-1 result / N. Huske’s future thesis (’10)7 fb-1 result/ J. Brown’s future thesis (’11)
12
Gregorio Bernardi / DØ-France/ October 14-2008
ICHEP Preliminary WHlbb
• WHlbb - signature: high pT lepton, MET and b jets– Backgrounds: W+bb, W+qq(mistagged), single top, Non
W(QCD)– Key issue: estimating W+bb background
• Shape from MC with normalization from data control regions
– Innovations: CDF: 20% acceptance from isolated tracks, NN jet corrections
TWO approaches Neural Net and Matrix Element+Boosted Decision Trees
Working on combining both for Tevatron combination.
Analysis Lum (fb-1)
Higgs Events
Exp. Limit
Obs. Limit
CDF NN 2.7 8.3 5.8 5.0
CDF ME+BDT 2.7 7.8 5.6 5.7
DØ NN 1.7 7.5 8.5 9.3
Results at mH = 115GeV: 95%CL Limits/SM
13
Gregorio Bernardi / DØ-France/ October 14-2008
Other SM Higgs Searches
• CDF and DØ are performing searches in every viable
mode:
- CDF: VHqqbb: 4 Jet mode.
– CDF: H with 2jets• Simultaneous search for Higgs in
VH, VBF and ggH production modes
• Interesting benchmark for LHC
– DØ: H • Also model independent and
fermiophobic search
– DØ: WHbb, new mode• Dedicated search with hadronic
decays
– DØ: ttH, new mode
Analysis: Limits at 160 and
115GeV
Exp. Limi
t
Obs. Limi
t
DØ WHWWW 20 26
CDF VHqqbb 37 37
CDF H 25 31
DØ WHbb 42 35
DØ H 23 31
DØ ttH 45 64
14
Gregorio Bernardi / DØ-France/ October 14-2008
Low mass (mH <~135 GeV): dominant decay: bbWHqq
bbZHqq
bbZHqq Use associated production modes to get better signal/background
High mass (mH>~135 GeV): dominant decay:
(*)WWH WWHgg
(*)WWWWHqq Intermediate mass:
bbH
VBF also contributes ~ 6%
SM Higgs Searches at Higg Mass at the Tevatron
15
Gregorio Bernardi / DØ-France/ October 14-2008
Search for WH WWW* in di-lepton mode
Limits ~20 times higher than SM at 160 GeV useful in the intermediate region 125-140 GeV .. and also for fermiophobic Higgs searches
Data (Signal@ 160 GeV) 19 (0.10) 15 (0.21) 5 (0.11)
Topological Likelihood discriminant, built on 3 variables per channel (ET
miss, ’s
Same charge di-lepton from W’s (one from HWW, the other from prompt W)
Gregorio Bernardi / DØ-France/ October 14-2008
High Mass: gg ➙H ➙WW*
➙ll(l=e,mu)
17
Gregorio Bernardi / DØ-France/ October 14-2008
WW production W + jet/ production:
H WW* ll
Major backgrounds
Selection Strategy:• Presection: lepton ID, isolation,trigger, opposite charge leptons • Remove QCD and Zl+l-: ET > 20 GeV
• Higgs Mass Dependent Cuts: Invariant Mass (Ml+l-); Min. Transverse Mass Sum of lepton pT
l and ET (pT
l + ET)
• Anti tt(bar) cut: HT = PTjet < 100 GeV, or less than
2 jets• Spin correlation in WW pair: (l,l) < 2.0
•Then apply advanced analysis technique, Neural Net+Matrix element
Now measured at the Tevatron by both expts. in agreement with NLO calculation: ~13.5 pb
Gregorio Bernardi / DØ-France/ October 14-2008
H WW* ll : Selection
Signal (x10)MT
min/GeV
Signal
19
Gregorio Bernardi / DØ-France/ October 14-2008
5 Matrix Elements discrim.
H
Neural Network Input Variables /Matrix Elements
Gregorio Bernardi / DØ-France/ October 14-2008
Matrix elements discriminant
Matrix Elements + Neural Net Final Discriminant
21
Gregorio Bernardi / DØ-France/ October 14-2008
SM Higgs: HWW
• Most sensitive Higgs search channel at the Tevatron
Analysis Lumi (fb-1)
Higgs Events
Exp. Limit
Obs. Limit
CDF ME+NN 3.0 17.2 1.6 1.6
DØ NN 3.0 15.6 1.9 2.0
Results at mH = 165GeV : 95%CL Limits/SMBoth experiments
Approaching
SM sensitivity!
22
Gregorio Bernardi / DØ-France/ October 14-2008
Systematics, including correlations, are taken into account when combining results.
Correlations of uncertainties of analyses inside one experiment (e.g. Jet calibration), or across experiments (e.g. PDF, theoretical cross-sections)
Main systematics (depending on channel):
- luminosity and normalisation - QCD background estimates - input background cross-sections - jet energy scale and b-tagging - lepton identification - K-factors on W/Z+ Heavy Flavor
Combining the results
23
Gregorio Bernardi / DØ-France/ October 14-2008
Limit SettingLEP: low background, small systematics Tevatron: high background, large systematics (at low mass)
But SMALL signals in both cases Background only (b) and signal plus background (s+b)hypotheses are compared to data using Poisson likelihoods.
For the searches and to set limits, Tevatron experiments use generalized CLs method (modified frequentist, DØ) and Bayesian methods (CDF), and cross-check each other.
Systematic uncertainties are included in the likelihood, via Gaussian smearing of the expectation (‘profile likelihood’).
New compared to LEP: Background is constrained by maximising profile likelihood (‘sideband fitting’), useful in particular at low mass.
Limit setting approaches agree to better than 10%
24
Gregorio Bernardi / DØ-France/ October 14-2008
Constraining Systematic Uncertainties with Data
“Profiling”
AKA side band fitting Nuisance parameters introduced in the chi2 of the fit allow shifting of central value of the background estimation
Systematic uncertainty width gets also constrained
Shape of the systematic is also taken into account
Background prediction + uncertainty
25
Gregorio Bernardi / DØ-France/ October 14-2008
Best fit for nuisance parameters
No smoking gun after all the checks, proceed to derive combined limit…
Centered close to good
sh
ift
in u
nit
s o
f syste
mati
c
un
cert
ain
tyDØ Higgs SM combination
Fitted Nuisance parameters for each systematic uncertainty
26
Gregorio Bernardi / DØ-France/ October 14-2008
Combining the Results
WHl bb 2.7 + 1.7 fb—1 / NN+NN e/, 1b/2b
ZHll bb 1.0 + 1.1 fb—1 / NN+NN e/, 1b/2b
ZH bb 2.7 + 2.1 fb—1 / NN+DT Z, Wł (2b)
H (gg,VBF,WH,ZH) 2.1 + __ fb-1 / NN . + 2 jets
H __ + 2.3 fb-1 / Di- mass
HWW* 2.4 + 2.3 fb—1 / NN&ME ee, e,
WHWWW* 1.7 + 1.1 fb—1 2D LHood
ee, e,
Total of 28 CDF + DØ channels combined
Channel Lumi /Technique Final state
27
Gregorio Bernardi / DØ-France/ October 14-2008
Post-Moriond 2008, with up to 2.4 fb-1
Observed limit at mH= 160 Gev: 1.1 x SM (3.6 @ 115 GeV)Very close to excluding a 160 GeV SM Higgs. @ ICHEP: ~ 3 fb-1
3.3 SM at mH=115 GeV 1.6 SM at mH=160 GeV
arXiv:0804.3423
28
Gregorio Bernardi / DØ-France/ October 14-2008
SM Higgs Limits with 3 fb-1
• Limits calculating and combination– Using Bayesian and CLs methodologies.– Incorporate systematic uncertainties using pseudo-
experiments (shape and rate included) (correlations taken into account between experiments)
– Backgrounds can be constrained in the fit
• Low mass combination difficult due to ~70 channels, not updated yet
but Expected sensitivity of CDF/DØ combined: <3.0xSM @ 115GeV
Combination of CDF and D0 done at high Mass > 150 GeV
29
Gregorio Bernardi / DØ-France/ October 14-2008
SM Higgs Combination at High Mass
High mass only
Expected: 1.2 @ 165, 1.4 @ 170 GeV
Observed: 1.0 @ 170 GeV
30
Gregorio Bernardi / DØ-France/ October 14-2008
SM Higgs Combination at High Mass
• Result verified using two independent methods(Bayesian/CLs)
M Higgs(GeV) 160 165 170 175
Method 1: Exp 1.3 1.2 1.4 1.7
Method 1: Obs
1.4 1.2 1.0 1.3
Method 2: Exp 1.2 1.1 1.3 1.7
Method 2: Obs
1.3 1.1 0.95 1.2
95%CL Limits/SM
SM Higgs Excluded: mH = 170 GeV
We exclude at 95% C.L. the production of a SM Higgs boson of 170 GeV
31
Gregorio Bernardi / DØ-France/ October 14-2008
• Since 2005, our high Higgs mass experimental sensitivity has improved by a factor of 1.7 (i.e. taking out gain due to luminosity)– NN discriminants– Lepton acceptance
• For 2010, we estimate an additional improvement in analysis sensitivity by a factor of 1.4– increased lepton efficiency (10% per lepton)– multivariate analyses (~30% in sensitivity)
• Potential improvements not included in estimate– add channels– …
Projection assumptions: High mass Higgs (Dec. 07)
32
Gregorio Bernardi / DØ-France/ October 14-2008
• Since 2005, our analysis sensitivity has improved by a factor of 1.7 well beyond improvement expected from sqrt(luminosity)– Acceptance/kin. phase space/Trigger efficiency– Asymmetric tagging for double b-tags – b-tagging improvements (NN b-tagging)– improved statistical techniques/event NN discriminant for channel with largest effort applied (WH) factor was 2.1
• For 2010, we estimate that we will gain an additional factor of 2.0 beyond improvement expected from sqrt(luminosity)– b-tagging improvements
• DØ: Layer 0 (~8% per tag efficiency increase)• add semileptonic b-tags (~5% per tag efficiency
increase)– Di-jet mass resolution (18% to 15% in σ(m)/m)– increased lepton efficiency (10% per lepton)– improved/additional multivariate techniques (~20% in
sensitivity)
Sensitivity and Projections – MH = 115 GeV
33
Gregorio Bernardi / DØ-France/ October 14-2008
Expected Higgs sensitivity in 2009/2010
Assumes two experiments
2010 2009
Projection: DØ X 2
By the time LHC produces Physics (2009), precision EW meas. + Tevatron might allow SM Higgs only with mass between 118 and 145 GeV, definitely only a light Higgs boson, which will take some time to be found at LHC (> 1 fb-1) LHC/Tevatron complementarity H vs H bb
2009
34
Gregorio Bernardi / DØ-France/ October 14-2008
Higgs sensitivity, 3-σ evidence
Assumes two
experiments2010 2009
Projection: DØ X 2
• With data accumulated by the end of 2010, we will be able to explore much of the SM Higgs mass region allowed by the constraints from precision measurements and LEP direct exclusion– Expected 95% CL exclusion over whole allowed range, (except
possibly around 130 GeV) - assuming the Higgs does not exist at these masses
– Three-sigma evidence for a Higgs possible over almost entire range, and probable for the low end and high end.
35
Gregorio Bernardi / DØ-France/ October 14-2008
ConclusionNew Higgs analyses available, large common effort by both CDF and
Dzero collaborations. Sensitivity constantly progressing ~linearly with Luminosity.
The Tevatron experiments have achieved sensitivity to the SM Higgs boson production cross section at High Mass, still working on low Mass
• 2009 will also teach us how well we perform at low mass, with the golden WZ/ZZ (Zbb) benchmark
• The Higgs boson search is entering its most exciting era, since it is for the first time within reach in the very near future (< 3 years).
• We exclude at 95% C.L. the production of a SM Higgs boson of 170 GeV– We expect large exclusion, or
evidence, with full Tevatron data set and improvements
Report for conseil scientifique In2p3, july 2000
36
Gregorio Bernardi / DØ-France/ October 14-2008
37
Gregorio Bernardi / DØ-France/ October 14-2008
Expected
38
Gregorio Bernardi / DØ-France/ October 14-2008
Systematics: ZH-llbb / CDF vs DØ