Post on 25-Apr-2018
Final Combination of CDF’s Searches for the Higgs Boson
Tom Junk Fermilab
On behalf of the CDF Collabora5on
1/18/13
Joint Experimental-‐Theore5cal Physics Seminar January 18, 2013
in the Standard Model and Extensions
T. Junk CDF's Higgs Searches 1
1/18/13
Tevatron Performance Proton-‐an5proton collisions with Run I: 1988 – 1996 : CDF Collects over 100 pb-‐1 of data at ECM = 1.8 TeV Run II: 2001 – 2011 : CDF Collects 10.0 W-‐1 of data at ECM = 1.96 TeV -‐-‐ 12 W-‐1 delivered. Many thanks to the Beams Division for spectacular performance of the Tevatron
T. Junk CDF's Higgs Searches 2
1/18/13
The CDF II Detector First CDF pp event: 1985 End of opera5ons: 2011
T. Junk CDF's Higgs Searches 3
1/18/13
1
10
10 2
10 3
100 125 150 175 200 225 250 275 300mH [GeV]
(pp
H+X
) [fb]
Tevatrons=1.96 TeV
pp–H (NNLO+NNLL QCD + NLO EW)
pp–
WH (NNLO QCD + NLO EW)
pp–ZH (NNLO QCD + NLO EW)
pp– qqH (NNLO QCD + NLO EW)pp–tt–H (NLO QCD)
[GeV]HM100 200 300 400 500 1000
Higg
s BR
+ T
otal
Unc
ert
-310
-210
-110
1
LHC
HIG
GS
XS W
G 2
011
bb
cc
ttgg
Z
WW
ZZ
SM Higgs Boson Production and Decay Rates
Primary analysis modes: mH<130 GeV: VHàVbb mH>130 GeV: ggàHàWW
Addi5onal modes: Hàττ, Hàγγ, HàZZà4l cHàcbb
Gluon Fusion
Associated Produc5on
Vector Boson Fusion
t
t cH
T. Junk CDF's Higgs Searches 4
1/18/13
Talk Outline
• Updates to CDF’s METbb Search • CDF’s Hàbb combina5on • Combined Results of all CDF SM Higgs boson searches • Fourth-‐Genera5on and Fermiophobic Results • Constraints on non-‐SM Couplings
T. Junk CDF's Higgs Searches 5
1/18/13
Associated production VH
H
b
b!
l-
l
Jet
Jet
ν
l
ν
ν!
1)
2)
3)
V
Friday, November 9, 12
“lvbb”
“METbb” or “vvbb”
“llbb”
Drawing Credit: CMS Higgs TWiki. CDF also seeks qqHàqqbb and cHàcbb
The Main Associated Production Search Channels
T. Junk CDF's Higgs Searches 6
1/18/13
The Status as of Summer 2012 CDF, D0, and Tevatron Publica5ons: CDF lvbb: Phys. Rev. Lec. 109, 111804 (2012) Includes new HOBIT b-‐tagger CDF llbb: Phys. Rev. Lec. 109, 111803 (2012) Includes new HOBIT b-‐tagger CDF METbb: Phys. Rev. Lec. 109, 111805 (2012) Uses old SECVTX+JetProb b-‐taggers CDF Combined Hàbb: Phys. Rev. Lec. 109, 111802 (2012) D0 lvbb: Phys. Rev. Lec. 109, 121804 (2012) D0 llbb: Phys. Rev. Lec. 109, 121803 (2012) D0 METbb: Phys. Lec. B 716, 285 (2012) D0 Combined Hàbb: Phys. Rev. Lec. 109, 121802 (2012) Tevatron Combined Hàbb: Phys. Rev. Lec. 109, 071804 (2012)
T. Junk CDF's Higgs Searches 7
1/18/13
10-4
10-3
10-2
10-1
1
90 100 110 120 130 140 15010-4
10-3
10-2
10-1
1
mH (GeV/c2)
Back
grou
nd p
-val
ue ObservedExpected if SM Higgs signal
at each mH separately
Expected ±1 s.d.Expected ±2 s.d.
1
2
3
0
100
200
300
400
500
600
700
90 100 110 120 130 140 1500
100
200
300
400
500
600
700
mH (GeV/c2)
(W
H+
ZH) x
Br(H
bb– ) (fb
)
Measured±1 s.d.
±2 s.d.Predicted
At mH=125 GeV/c2, the measured rate is
! (WH + ZH )!Br(H" bb ) = 291#113+118 fb
CDF’s Hàbb Results, Summer 2012
The SM predic5on is
! (WH + ZH )!Br(H" bb ) =120±8 fb
Cross Sec5ons: Baglio, Djouadi; Harlander, Mantler, Marzani, Ozeren Branching Ra5os: LHC Cross Sec5on Working Group Denner et al., arXiv:1101.0593, arXiv:1201.3084
p-‐value using the cross sec5on as test sta5s5c: 2.7σ
T. Junk CDF's Higgs Searches 8
Phys. Rev. Lec. 109, 111802 (2012)
1/18/13
Tevatron Hàbb Results, Summer 2012
Maximum local significance: 3.3σ at mH=135 GeV/c2 Global significance: 3.1σ Local significane at mH=125 GeV/c2 is 2.8σ
Cross sec5on fit at mH=125 GeV/c2
(!WH +! ZH )!B(H" bb ) = 230#0.080+0.090 (stat+syst) fb
SM Predic5on: 0.12 ± 0.01 pb
Phys. Rev. Lec. 109, 071804 (2012)
T. Junk CDF's Higgs Searches 9
1/18/13
Current Status from the LHC
A Higgs-‐like par3cle is firmly established, its mass looks to be between 125 and 126 GeV, and its proper3es are consistent with the SM Higgs boson
Many more superb results are available at hcps://twiki.cern.ch/twiki/bin/view/AtlasPublic/HiggsPublicResults hcps://twiki.cern.ch/twiki/bin/view/CMSPublic/PhysicsResultsHIG
T. Junk CDF's Higgs Searches 10
1/18/13
CDF’s HOBIT b-tagger – Improved Sensitivity With Respect to the SECVTX, Roma, and BNess taggers
HOBIT Tight and Loose opera5ng points chosen to have similar mistag rates as older SECVTX opera5ng points. More signal More b background Similar LF background
A neural-‐network b-‐tagger using inputs from other b-‐taggers, as well as lepton ID to include semileptonically decaying B hadrons.
J. Freeman et al., Nucl. Instrum. Methods A 697, 64 (2012). see also M. Stancari’s JETP Seminar, Mar. 7 2012
Opera3ng Point
Mistag rate
SECVTX efficiency
HOBIT efficiency
Tight 1.4% 39% 54%
Loose 2.9% 47% 59%
per-‐jet efficiencies shown
T. Junk CDF's Higgs Searches 11
1/18/13
VHàMETbb Analysis
Select Events with 2 or 3 jets: 25 < ET,J1 < 200 GeV 20 < ET,J2 < 120 GeV If a third jet is present, 15 < ET,J3 < 100 GeV |ηjet| < 2 with at least one jet with |η|<0.9 No iden5fied lepton MET > 35 GeV ΔϕMET,J2>0.4; ΔϕMET,J3>0.4 if present
B-‐tagging: Double-‐5ght HOBIT tag (TT) Tight tag + Loose Tag (TL) One 5ght tag (1T)
Event Preselection
Inversions of the event selec5on requirements define control samples
T. Junk CDF's Higgs Searches 12
1/18/13
Signal (and b-tagged background) gain by using HOBIT compared with SECVTX and JetProb
Categories are non-‐overlapping. The 1T category does not include TL or TT events
T. Junk CDF's Higgs Searches 13
1/18/13
Data-Based Tagged QCD Multijet Prediction Control Region: Same cuts as Preselec5on but:
• no b-‐tag required • 35 GeV < MET < 70 GeV • Δφ(MET,j2) < 0.4
Nearly all events are QCD mul5jet events in this sample. Measure tag frac5ons for TT, TL, 1T in this sample as func5ons of
• HT • missing pT • the frac5on of charged energy in the cones of jets 1 and 2 • the number of reconstructed primary ver5ces • The momentum-‐weighted sum of the sines of the angles between reconstructed muon candidates and jet axes
Result is three Tag Rate Matrices. Apply these to preselected data minus W+jets, Z+jets, cbar, single top, and dibosons to model the tagged QCD mul5jet background
New for the HOBIT analysis
T. Junk CDF's Higgs Searches 14
1/18/13
)T
p, T
E(0 1 2 3
Even
ts /
0.10
10
210
310
410
)2) (GeV/c2
,j1
Invariant Mass(j0 100 200 300 400
2Ev
ents
/ 8
GeV
/c
1
10
210
310
410 Data QCD Multijet TopV + HF EWK Mistags VV
Sphericity0 0.2 0.4 0.6 0.8 1
Even
ts /
0.03
3
10
210
310
410
Validation of Key Variables in the Tagged Preselection Sample
Variables also checked in • QCD-‐enriched region: low MET, low Δφ(MET,jet) • Electroweak control region Requires the presence of a high-‐pT isolated lepton (e or μ)
T. Junk CDF's Higgs Searches 15
1/18/13
Signal Region: NNQCD>0.6. Retains 87% of signal while rejec5ng 90% of QCD mul5jet backgrounds
TT+TL+1T together
QCD-rejection Neural Network
Signal Region
QCDNN0 0.2 0.4 0.6 0.8 1
Even
ts /
0.03
310
210
310
410
510Data QCD Multijet TopV + HF EWK Mistags VV
10)×VH (Used to Normalize QCD
0.1 < NNQCD < 0.6 events used to normalize the QCD rate Extrapola5on uncertain5es assessed in predic5ng rates for NNQCD > 0.6. Larger uncertain5es than Moriond 2012 analysis
Signal Region
T. Junk CDF's Higgs Searches 16
1/18/13
Predicted Yields in the Three Tagged Signal Samples
Process 1T TL TT
QCD Mul5jet 5941 ± 178 637 ± 25 222 ± 16
Top 1174 ± 158 302 ± 40 271 ± 34
V+heavy flavor jets 3124 ± 718 286 ± 83 211 ± 65
Electroweak mistags 1070 ± 386 55 ± 21 13 ± 6
Diboson 305 ± 46 48 ± 6 41 ± 5
Total expected background 11612 ± 949 1329 ± 112 759 ± 86
Observed Data 11955 1443 692
ZHàννbb, llbb (mH=125 GeV) 9.7 ± 1.0 5.4 ± 0.5 5.4 ± 0.5
WHàlνbb 9.8 ± 1.0 5.3 ± 0.5 5.3 ± 0.5
T. Junk CDF's Higgs Searches 17
1/18/13
)2 = 125 GeV/cH
(mSIGNN0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1T e
vent
s / 0
.04
0
200
400
600
800
1000
1200Data QCD Multijet TopV + HF EWK Mistags VV
10)×VH (
0.8 0.85 0.9 0.95 1
1
10
210
)2 = 125 GeV/cH
(mSIGNN0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TL e
vent
s / 0
.04
0
20
40
60
80
100
120
140
0.8 0.85 0.9 0.95 1
1
10
)2 = 125 GeV/cH
(mSIGNN0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TT e
vent
s / 0
.04
0
10
20
30
40
50
60
70
0.8 0.85 0.9 0.95 1
1
10
Signal-Separation Neural Network Output Distributions for the Three b-tagged Signal Samples
Double Tight
Single Tight Tag
Tight+Loose
T. Junk CDF's Higgs Searches 18
1/18/13
)2 (GeV/cHm90 100 110 120 130 140 150
95%
CL
Uppe
r Lim
it/SM
1
10
ObservedExpected
1 s.d. expected±
2 s.d. expected±
)2Higgs Mass (GeV/c90 100 110 120 130 140 150
95%
C.L
. lim
it / S
M
1
10
68% Confidence interval95% Confidence intervalExpected 95% C.L. limitObserved 95% C.L. limit
Observed and Expected Limits
This Result: HOBIT TT+TL+1T
SECVTX+JP SS+SJ+1S
Phys. Rev. Lec. 109, 111805 (2012)
At mH=125 GeV, obs = 3.06xSM exp = 3.33xSM
At mH=125 GeV, obs = 6.7xSM exp = 3.6xSM
8% sensi5vity improvement at mH=125 GeV/c2. Average expected improvement over whole mass range: 14%. Big change in observed result! New limit is a drop of 55% rela5ve to the published one at mH=125 GeV/c2
T. Junk CDF's Higgs Searches 19
1/18/13 T. Junk CDF's Higgs Searches 20
Discussion of METbb Results
Many cross-‐checks run to assess compa5bility of the new HOBIT result and the published SECVTX+JetProb result: 1. Reanalysis of 1S, SJ, and SS tag categories using new framework, selec5on, and systema5cs 2. Valida5on of b-‐jet tagger modeling 3. Effects of Data Migra5on 4. P-‐value for the difference in observed limits using pseudoexpriments 5. Background modeling
1/18/13
Analysis Updates Besides Changing b-Taggers • B-‐tag scale factor uncertainty handling improved – proper an5correla5ons between exclusive samples evaluated. • QCD Mul5jet uncertain5es are larger, and improved methods for extrapola5ng from the control regions are used. • Events with with jets with ET1>200 GeV and ET2>120 GeV now rejected • Addi5onal MET cut at 35 GeV in early stage of analysis to get away from trigger turn-‐on now applied
• V+heavy flavor jets backgrounds now itemized separately with independent sources of uncertainty. The charm tag rate and gluon spli�ng rates are different depending on the sample. • Z+jets with Zàbb background model included. Re-‐Doing the SECVTX+JetProb analysis with these changes measures the impact of the changes and provides a cross check of the Winter/Summer 2012 METbb analysis
T. Junk CDF's Higgs Searches 21
1/18/13
)2 (GeV/cHm90 100 110 120 130 140 150
95%
CL
Uppe
r Lim
it/SM
1
10
ObservedExpected
1 s.d. expected±
2 s.d. expected±
HOBIT
)2 (GeV/cHm90 100 110 120 130 140 150
95%
CL
Uppe
r Lim
it/SM
1
10
Observed [5]Expected [5]
1 s.d. expected± 2 s.d. expected±
Observed (Cross-check)Expected (Cross-check)
SECVTX+JP Published analysis + Redo Crosscheck
Results of SECVTX+JetProb Redo Cross Check
At mH=125 GeV/c2, the observed HOBIT limit is 47% lower than the Re-‐Done SECVTX+JP observed limit. (Was 55% lower than published).
T. Junk CDF's Higgs Searches 22
1/18/13
Combine the Redone SECVTX+JetProb METbb with llbb+lvbb
1
10
90 100 110 120 130 140 150
1
10
mH (GeV/c2)
95%
CL
Lim
it/SM
CDF Run II Preliminary H bb–, L = 9.45 fb-1
Observed (w/ published METbb)Expected w/o HiggsExpected ±1 s.d.Expected ±2 s.d.
Observed (w/ Re-DoneSECVTX+JetProb METbb)
SM=1
Black Curve: Same limit curve as in Phys. Rev. Lec. 109, 111802 (2012) Red Curve: Published llbb, lvbb, and the Redone SECVTX+JetProb crosscheck P-‐value remains unchanged. at 2.7σ at mH=125 GeV/c2
T. Junk CDF's Higgs Searches 23
CDF finds no significant issues with the previous version of the METbb analysis and stands firmly behind last summer’s published results.
CDF Combined VHàVbb limits/SM
1/18/13
B-Tag Modeling Validation: Electroweak Control Region
)2 = 125 GeV/cH
(mSIGNN0 0.2 0.4 0.6 0.8 1
TT E
vent
s / 0
.04
1
10
210 Data QCD Multijet TopV + HF EWK Mistags VV
Double-‐Tight HOBIT TT
)2 = 125 GeV/cH
(mSIGNN0 0.2 0.4 0.6 0.8 1
SS E
vent
s / 0
.04
1
10
210
Double-‐Tight SECVTX SS
Preselec5on requirements, but require a high-‐pT isolated lepton (e or μ) Look for mismodeled correla5ons between NNSIG and b-‐tagging
T. Junk CDF's Higgs Searches 24
No modeling issues seen with either b-‐tag algorithm as a func5on of NNSIG
1/18/13
1T TL TT 1S 55% 35% 15% SJ 4% 20% 30% SS 1% 14% 51%
Data Event Overlap Fractions: NNSIG>0.8
T. Junk CDF's Higgs Searches 25
Denominator: HOBIT analysis events
Studying the Effects of Event Migration
HOBIT Tagger is more efficient: Expect promo5on of signal events to higher tag categories.
We do not see problems in modeling: Suspect sta5s5cal fluctua5ons.
1/18/13
Event Rate Summaries – HOBIT analysis vs. SECVTX+JP
Now: 6.5 events deficit at high score. Was: spot on
Tag category Background (fit) DataTT 39.5± 4.6 33SS 37.6± 4.6 37TL 67.4± 6.8 80SJ 45.6± 5.1 62
T. Junk CDF's Higgs Searches 26
High Score: NNSIG > 0.8 Observed Event Counts
Excesses seen in both old and new 5ght-‐loose categories
1/18/13 T. Junk CDF's Higgs Searches 27
Events Gained and Lost: TL and SJ
Top three NNSIG score events in the SJ analysis are no longer selected. Two are now LL and one is 1L. Bin has high s/b and a large weight in the result. A test: We added these three events by hand to TL. SECVTX+JP(redo) vs. HOBIT limit discrepancy changes from 47% to 31% with the events added back. The NNSIG func5on is unchanged since
Moriond 2012. Selected, tagged events receive the same NNSIG scores they always did.
1/18/13
Events Gained and Lost: TT and SS
T. Junk CDF's Higgs Searches 28
Lost More Events than Gained in the high-‐NNSIG region. Studied adding 5 extra candidates to the TT data by hand, following the background shape (not just the last bin). Change in limit discrepancy: from 47% to 33% Combined effect of TT and TL candidates: reduces discrepancy from 47% to 19% Improvement in sensi5vity is 8% at mH=125 GeV
1/18/13
How Likely is it? Compare CDF METbb HOBIT analysis with the re-‐done SECVTX+JP analysis Paired pseudoexperiments: • HOBIT channels TT, TL, 1T • SECVTX channels SS, SJ, 1S • Use expected overlaps in the NNSIG < 0.8 and NNSIG > 0.8 regions to generate independent samples of events in 15 categories: TTSS, TTSJ, TT1S, TTnone TLSS, TLSJ, TL1S, TLnone 1TSS, 1TSJ, 1T1S, 1Tnone noneSS, noneSJ, none1S Calculate limit for each pair of pseudoexperiments & compare
P-‐value for |Limit difference| to be as big as observed: 7% Highly correlated over mH range: P-‐value for all limits to be low: 3-‐5%
T. Junk CDF's Higgs Searches 29
1/18/13 T. Junk CDF's Higgs Searches 30
)2 (GeV/cHm90 100 110 120 130 140 150
Ove
rlap
Frac
tion
0
0.2
0.4
0.6
0.8
1
> 0.8SIG
Overlap Fraction of Data Events with NN
= 100Hm = 125Hm = 150Hm
How Correlated are the Searches at Each mH? Limited resolu5on on input variables to NNSIG correlates results of searches at nearby masses. Between 2 and 3 independent search results over the mass range 90 < mH < 150 GeV/c2
1/18/13 T. Junk CDF's Higgs Searches 31
Tag category Background (fit) DataTT 264.8± 25.1 265SS 228.8± 21.0 217TL 506.1± 38.8 506SJ 312.5± 22.6 291
Studying Backgrounds in the Intermediate NNSIG Score Region
Intermediate Score: 0.48 < NNSIG < 0.8 Background fits dominated by events at even lower NNSIG score Limits on Higgs boson signal fairly insensi5ve to these bins.
)2 = 125 GeV/cH
(mSIGNN0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TT e
vent
s / 0
.04
0
10
20
30
40
50
60
70
0.8 0.85 0.9 0.95 1
1
10
No evidence for mismodeling in this region
1/18/13
Final CDF Combined VHàVbb Cross Section Measurements
The SM predic5on is
! (WH + ZH )!Br(H" bb ) =120±8 fb
0
100
200
300
400
500
600
700
90 100 110 120 130 140 1500
100
200
300
400
500
600
700
mH (GeV/c2)
(W
H+
ZH) x
Br(H
bb– ) (fb
)
Measured±1 s.d.
±2 s.d.Predicted
At mH=125 GeV/c2, the final measurement is
! (WH + ZH )!Br(H" bb ) = 206#104+110 fb
Final
! (WH + ZH )!Br(H" bb ) = 291#113+118 fb
0
100
200
300
400
500
600
700
90 100 110 120 130 140 1500
100
200
300
400
500
600
700
mH (GeV/c2)
(W
H+
ZH) x
Br(H
bb– ) (fb
)
Measured±1 s.d.
±2 s.d.Predicted
Summer 2012
At mH=125 GeV/c2 the previous measuement is
T. Junk CDF's Higgs Searches 32
1/18/13
Summary of the Hàbb Updates and Investigations
• Previous METbb (SECVTX+JetProb) analysis is s5ll valid • Analysis redone with new framework, cuts, and systema5c uncertainty handling • Previous METbb result reproduced with small shi�s in limits • No change in combined limits or p-‐values
• Switching to a new b-‐tagger reclassified events – only 50% overlap with events in the old analysis in the highest-‐weight region. • New best-‐fit cross sec5ons are lower. • We must choose the more sensi5ve analysis: HOBIT METbb is 8% more sensi5ve than the previous version at mH=125 GeV/c2, and 14% stronger on average in 90 < mH < 150 GeV
T. Junk CDF's Higgs Searches 33
1/18/13 T. Junk CDF's Higgs Searches 34
Predicted Event Overlap Fractions, signal MC
Roman font – normalized to HOBIT yields. Italics: Normalized to SECVTX+JP yields
Combining the SECVTX+JetProb Analysis and HOBIT Analysis? Reasons to do it: • More sta5s5cal power • Observed results may be intermediate between published and new results
Reasons not to do it: • Expected sensi5vity gain is very small – signal events are expected to remain tagged, merely shuffled from one category to another • More exclusive tag categories (15 instead of 3). Smaller data control samples mean higher systema5c uncertain5es • HOBIT uses SECVTX variables as inputs, among others. It’s already a combina5on. • Combining HOBIT and SECVTX+JetProb was not on our original menu of analysis improvements, for the above reasons. We made the decision to switch all analyses from SECVTX+JetProb to HOBIT without knowledge of the data outcome. Switching a posteriori would bias the final result.
1/18/13
Updates Since the last full CDF SM Higgs Combination
• Revert to the published 6.0 W-‐1 Hàττ Search. 8.3 W-‐1 preliminary result not published.
• Update the cH search: train MVA’s at each mH. • Improve correla5ons/an5correla5ons in b-‐tag uncertain5es in llbb / lvbb analyses • Upgraded the Central-‐Central Hàγγ search to use an MVA instead of mγγ. Searches including plug photons and conversions s5ll use mγγ. CDF’s SM combina5on last done for Winter 2012 conferences.
T. Junk CDF's Higgs Searches 35
1/18/13
All CDF SM Search Channels
T. Junk CDF's Higgs Searches 36
1/18/13
1
10
0 2 4 6 8 10 12 14Integrated Luminosity (fb-1)
Expe
cted
Lim
it/SM Summer 2005
Summer 2006Summer 2007Winter 2008Fall 2008
Fall 2009Summer 2010Summer 2011Winter 2012
mH=115 GeV/c2
SM=1
1
10
0 2 4 6 8 10 12 14Integrated Luminosity (fb-1)
Expe
cted
Lim
it/SM Summer 2004
Summer 2005Summer 2007Winter 2008Fall 2008
Spring 2009Fall 2009Summer 2010Summer 2011Winter 2012
mH=160 GeV/c2
SM=1
Sensitivity Evolution over Time
We collected data, and we also learned how to get more out of the data. • Becer MVA’s • Becer Event Selec5on • Becer lepton ID • Becer jet energy resolu5on • More triggers • More analysis categories • Sharing improvements between analyses
T. Junk CDF's Higgs Searches 37
1/18/13
10-4
10-3
10-2
10-1
1
10
10 2
10 3
10 4
10 5
10 6
-4 -3 -2 -1 0log10(s/b)
Even
ts CDF dataBackground fitSM Higgs signal
mH=125 GeV/c2
10-4
10-3
10-2
10-1
1
10
10 2
10 3
10 4
-4 -3 -2 -1 0log10(s/b)
Even
ts CDF dataBackground fitSM Higgs signal
mH=165 GeV/c2
SM Combined Data Summaries at mH=125 and 165 GeV/c2
Background Hypothesis fit to data Bins of similar s/b added together
T. Junk CDF's Higgs Searches 38
1/18/13
-150
-100
-50
0
50
100
150
-3 -2.5 -2 -1.5 -1 -0.5 0log10(s/b)
Even
ts/0
.25
Data-background
SM Higgs signal
±1 s.d. on background
mH=125 GeV/c2log10(s/b)
Even
ts/0
.25
-10
-5
0
5
10
-1 -0.75 -0.5 -0.25 0-10
-5
0
5
10
-150
-100
-50
0
50
100
150
-3 -2.5 -2 -1.5 -1 -0.5 0log10(s/b)
Even
ts/0
.25
Data-background
SM Higgs signal
±1 s.d. on background
mH=165 GeV/c2log10(s/b)
Even
ts/0
.25
-10
-5
0
5
10
-1 -0.75 -0.5 -0.25 0-10
-5
0
5
10
SM Combined Data Summaries at mH=125 and 165 GeV/c2
Same as before, but ficed background subtracted from the data. Data error bars are sqrt(s+bfit)
T. Junk CDF's Higgs Searches 39
1/18/13 T. Junk CDF's Higgs Searches 40
Sensitivity of the Main Search Channels
1
10
10 2
100 110 120 130 140 150 160 170 180 190 200
1
10
10 2
mH (GeV/c2)
R95m
ed
H + -
Htt–H tt–tb–
VH Vbb–
H ZZ 4lH W+W-
Combined
SM=1
1/18/13
1
10
100 120 140 160 180 200
1
10
mH (GeV/c2)
R95 Expected if no Higgs
ObservedExpected ±1 s.d.Expected ±2 s.d.
Expected if mH=125 GeV/c2
SM=1
CDF’s Final SM Higgs Rate Limits
Excluded regions: 90 < mH < 102 GeV/c2 and 149 < mH < 172 GeV/c2 Expect to exclude if no Higgs: 90 < mH < 94 GeV/c2, 96 < mH < 106 GeV/c2, and 153 < mH < 175 GeV/c2
T. Junk CDF's Higgs Searches 41
1/18/13
SM/Best Fit 0 1 2 3 4 5 6 7 8 9 10 11
b VbVH
-+ H
-W+ WH
Hbbt tHtt
2 = 125 GeV/cHmCombined (68%) Single Channel
0
0.5
1
1.5
2
2.5
3
3.5
4
100 120 140 160 180 2000
0.5
1
1.5
2
2.5
3
3.5
4
mH (GeV/c2)
Rfit
Observed±1 s.d.±2 s.d.
Expected ifmH=125 GeV/c2
Best-Fit Cross Sections at mH=125 GeV/c2
ttH! ttbb : 9.49"6.28+6.60 #SM
H! !! : 7.81"4.42+4.61 #SM
H!W +W " : 0.00"0.00+1.78 #SM
H! ! +! " : 0.00"0.00+8.44 #SM
VH!Vbb : 1.72"0.87+0.92 #SM
Combined : 1.54!0.73+0.77 "SM
T. Junk CDF's Higgs Searches 42
SM Combina5on
1/18/13
10-4
10-3
10-2
10-1
1
10
100 120 140 160 180 20010-4
10-3
10-2
10-1
1
10
mH (GeV/c2)
Back
grou
nd p
-val
ue ObservedExpected if SM Higgs signal
at each mH separately
Expected if mH=125 GeV/c2
Expected ±1 s.d.Expected ±2 s.d.
1
2
3
CDF’s Combined SM Higgs Boson Search p-value Observed local significance at mH=125 GeV/c2 is 2.0σ Expected significance at mH=125 GeV/c2 is 1.6σ assuming a signal is present.
2.5σ local significance at mH=120 GeV/c2
Look-‐Elsewhere Effect: • mH~125 GeV has been firmly established by the LHC • CDF’s mass resolu5on is not very sharp -‐-‐ ~2 independent search results in Hàbb, ~2 in HàWW • Technical challenge – MVA’s trained at each mH separately. Histograms of predic5ons exchanged. Would need to exchange correlated pseudoexperiments to compute LEE exactly.
T. Junk CDF's Higgs Searches 43
1/18/13
Extensions to the SM: Fourth-Generation Models (SM4)
A heavy fourth genera5on of quarks would scale the gg→H produc5on rate at colliders by a factor of ~9. But watch out for H→ν4ν4 decays. E. Arik et al., Acta Phys. Polon. B 37, 2839 (hep-‐ph/0502050)
Kribs, Spannowsky, Plehn, Tait, Phys. Rev. D 76, 075016 (2007) T. Junk CDF's Higgs Searches 44
1/18/13
10-1
1
10
120 140 160 180 200 220 240 260 280 300
10-1
1
10
mH (GeV/c2)
95%
C.L
. Lim
it/SM
4(lo
w m
ass)
pre
dict
ion
SM4(low mass)=1
ExpectedObservedExpected ±1 s.d.Expected ±2 s.d.SM4(high mass)
10-1
1
10
120 140 160 180 200 220 240 260 280 30010-1
1
10
mH (GeV/c2)
(gg
H)×
Br(H
W+ W
- ) (pb
)
Exp. 95% C.L. limitObs. 95% C.L. limit±1 s.d. exp. limit±2 s.d. exp. limitSM4 (low mass)SM4 (high mass)
Searches for ggàHàWW – Model Independent and SM4 interpretation
SM4 Cross Sec5ons computed at NNLO in QCD by Anastasiou, Boughezal, and Furlan
Searches op5mized specifically for ggàH (NN’s not trained with WH, ZH, or VBF). Limit on cross sec5on 5mes b.r. shown along with SM4 model predic5ons.
Low-‐Mass scenario: ml4=100 GeV, mv4=80 GeV High-‐Mass scenario: ml4=mv4=1 TeV Both Scenarios: md4=400 GeV, mu4=450 GeV
Low-‐mass scenario exclusions: Expected exclusion: 123 < mH < 231 GeV/c2 Observed exclusion: 124 < mH < 203 GeV/c2
T. Junk CDF's Higgs Searches 45
1/18/13
1
10
100 110 120 130 140 150 160 170 180 190 200
1
10
mH (GeV/c2)
R95
f
Expected if no Higgs
Observed
Expected ±1 s.d.
Expected ±2 s.d.
FHM=1
A Test of the Fermiophobic Higgs Model Model tested: Assume SM-‐like behavior for the Higgs boson, except switch off all couplings to fermions. Decays for Hàbb, Hàττ, and Hàgg are highly suppressed. ggàH produc5on is neglibly small. WH, ZH, VBF produc5on cross sec5ons are as predicted by the SM HàWW, ZZ par5al widths are as predicted by the SM. Hàγγ is modified – loss of the fermion loop increases the decay rate. Hàγγ search is re-‐op5mized for this search because the pT spectrum of the H is harder in WH and ZH than for ggàH Branching ra5os recomputed using the modified decay widths.
Included channels: Hàγγ, HàWW, HàZZà4l
Observed exclusion: 100 < mH< 113 GeV/c2 Expected exclsuion: 100 < mH< 122 GeV/c2
T. Junk CDF's Higgs Searches 46
1/18/13
Constraining the Couplings of the Higgs Boson to Fermions and Gauge Bosons
We follow the procedures and nota5on of the LHC Higgs Cross Sec5on WG A. David et al., arXiv:1209.0040
The model: SM-‐like, but Hff couplings are scaled together by κf HWW coupling is scaled by κW HZZ coupling is scaled by κZ For some studies, we scale the HWW and HZZ couplings by κW=κZ=κV Standard Model is recovered if κf = κW = κZ = 1
T. Junk CDF's Higgs Searches 47
1/18/13
Step 1: Scale cross sec3ons for each process according to couplings -‐-‐ Precy much by defini5on! Unless NNLO VBF includes the EW ggH piece. From Bolzoni, Moch, Maltoni, and Zaro’s papers, it seems as if the EW ggH piece is not in the NNLO VBF calcula5on.
LO rela5ons, but mostly true at higher order (most QCD affects the colored ini5al-‐state par5cles. There is ggàWH at higher order, however.)
Constraining Couplings
! (gg!H ) =! SM (gg!H )(0.95" f2 + 0.05" f"V )
! (WH ) =! SM (WH )"V2
! (ZH ) =! SM (ZH )"V2
! (VBF) =! SM (VBF)"V2
A. David et al., LHCHXSWG-‐2012-‐001. arXiv:1209.0040
T. Junk CDF's Higgs Searches 48
1/18/13
Aglie�, Bonciani, Degrassi and Vicini, arXiv:hep-‐ph/0610033 Anastasiou, Boughezal, Petriello, JHEP 0904 003 (2009) Ac5s, Passarino, Sturm, Uccira5 PLB 670, 12 (2008) Grazzini and de Florian PLB 674, 291 (2009)
Two-Loop Electroweak Contributions to the ggH Coupling
• Approximately a 5% contribu5on to the ggàH produc5on cross sec5on at mH=125 GeV/c2. Main contribu5on is from interference with the LO process. • Contribu5on not included in VBF calcula5on (HAWK: Denner, Dicmaier, Mück)
T. Junk CDF's Higgs Searches 49
1/18/13
Step 2: Recompute all Higgs boson decay branching ra3os from scaled par3al widths α and β come from Spira et al., arXiv:hep-‐ph/9504378 α=1.28 β = -‐0.21
Constraining Couplings
!(H" gg) = !SM (H" gg)(0.95! f2 + 0.05! f!V )
!(H"W +W # ) = !SM (H"W +W # )!V2
!(H" bb ) = !SM (H" bb )! f2
!(H" " +" # ) = !SM (H" " +" # )! f2
!(H" cc ) = !SM (H" cc )! f2
!(H" ZZ ) = !SM (H" ZZ )!V2
!(H" ## ) = !SM (H" ## )$!V +%! f2
Br(H! XX) = "(H! XX)"i
i#
Other modes, like Hàμ+μ-‐ and HàZγ have very small widths
T. Junk CDF's Higgs Searches 50
1/18/13
Some work from theorists Espinosa, Grojean, Mühlleitner, and Troc, “First Glimpse at Higgs’ Face”, arXiv:1207.1717v2 (updated Aug. 21 with post-‐ICHEP data) JHEP 1212, 045 (2012).
a=κV, c=κF T. Junk CDF's Higgs Searches 51
1/18/13 T. Junk CDF's Higgs Searches 52
More Complete Treatment of Signal Scalings Example: HWW Search. All Signals have HàWW, but different mixtures of produc5on mechanisms in each search channel.
Pie charts show rela5ve contribu5ons.
ggH WH ZH VBF
ggH WH ZH VBF
ggH
WH
ZH
VBF
Zero Jets
One Jet
Two or More Jets
Same Sign di-‐ leptons
Tri-‐Leptons Z-‐rejected
ggH WH ZH VBF
ggH WH ZH VBF ggH WH ZH VBF
Tri-‐Leptons Z-‐selected
Each signal contribu5on should be scaled by the appropriate factor
1/18/13
-3 -2 -1 0 1 2 3
-3-2
-10
12
3
12345678
Vf
WH(H bb)
-3 -2 -1 0 1 2 3
-3-2
-10
12
3012345678
Vf
ggH(H WW)
Signal Rate Enhancement Factors for ggàHàWW and WH->lvbb
T. Junk CDF's Higgs Searches 53
1/18/13
Signal Rate Enhancement Factors for ttHàttbb and WH->WWW
-3 -2 -1 0 1 2 3
-3-2
-10
12
3
5101520
2530
Vf
WH(H WW)
-3 -2 -1 0 1 2 3
-3-2
-10
12
3
2
4
6
8
10
Vf
ttH(H bb)
T. Junk CDF's Higgs Searches 54
W+
W-‐
t
t
H
b
b
1/18/13
f-5 0 5
Post
erio
r den
sity
0
0.1
0.2
0.3 Best fit68% C.L.95% C.L.
Posterior Constraint on the Hff Coupling Factor κf
• Uniform prior assumed • κW = κZ = 1 assumed
Excess in the Hàγγ searches drives the asymmetry from posi5ve and nega5ve coupling scale factors
T. Junk CDF's Higgs Searches 55
1/18/13 T. Junk CDF's Higgs Searches 56
W-2 -1 0 1 2
Post
erio
r den
sity
0
0.2
0.4
0.6Best fit68% C.L.95% C.L.
Posterior Constraint on the HWW Coupling Factor κW
• Uniform prior assumed • κf = κZ = 1 assumed
Excess in the Hàγγ searches drives the asymmetry from posi5ve and nega5ve coupling scale factors
1/18/13 T. Junk CDF's Higgs Searches 57
Z-2 0 2
Post
erio
r den
sity
0
0.1
0.2
0.3Best fit
68% C.L.
95% C.L.
Posterior Constraint on the HWW Coupling Factor κZ
• Uniform prior assumed • κf = κW = 1 assumed
1/18/13
V
0 0.5 1 1.5 2
f
-4
-2
0
2
4
6Local maxima SM68% C.L. 95% C.L.
=1WZ
Two-Dimensional Constraints: Bosons vs. Fermions
T. Junk CDF's Higgs Searches 58
Uniform prior assumed Large κV with small κf constrained by trileptons, same-‐sign dileptons Large κf and small κV constrained by cHàcbb
1/18/13 T. Junk CDF's Higgs Searches 59
W
0 0.5 1 1.5 2 2.5
Z
-2
-1
0
1
2
3Local maxima SM68% C.L. 95% C.L.
floatingf
Two-Dimensional Constraints: W vs. Z Couplings
Uniform prior assumed κf integrated over (“marginalized”) Less constraint on HZZ than on HWW All couplings consistent with SM predic5ons
1/18/13
Summary (1) • CDF’s Searches for the Higgs boson in the SM, SM4, and Fermiophobic models are now finalized • Publica5ons on the final METbb (HOBIT) search plus the combina5on are being submiced. • Previous METbb (SECVTX+JP) analysis is s5ll valid – no mistakes affec5ng the Summer 2012 result were found. • Switching to a new b-‐tagger reclassified events – only 50% overlap with events in the old METbb analysis in the highest-‐NNSIG region. • New best-‐fit cross sec5ons are somewhat lower. • We must choose the more sensi5ve analysis: HOBIT METbb is 8% more sensi5ve than the previous version at mH=125 GeV/c2
T. Junk CDF's Higgs Searches 60
1/18/13
Excess of events persists in the SM Higgs search near mH=125 GeV/c2. Higgs boson does not look like those of the FP model, or SM4. Couplings to W, Z, and fermions are consistent with SM predic5ons Extrac5ng coupling informa5on from the data requires full predic5ons of signal rates and shapes in all channels. Channels that contribute licle to the total SM sensi5vity can have outsized impacts on exo5c coupling scenario tests.
Summary (2)
T. Junk CDF's Higgs Searches 61