Searches for Higgs Bosons at CDF
Thomas WrightUniversity of Michigan
SLAC Experimental SeminarFebruary 13, 2007
2
Run 2 at the Tevatron
• World’s highest-energy collider
• Record luminosity 278E30 cm-
2s-1
• Will have 2 fb-1 on tape in ~1 week!
• Integrating 40-45 pb-1/week
• Results shown here use data samples of ~1 fb-1
• Look for updates on expanded samples this summer
3
The CDF II Detector
4
The CDF II Detector
• Azimuthally symmetric “barrel” geometry
• Central detectors cover ||<1
• Plug calorimeter extends to ||<3.6
• Silicon extends to z = 50 cm (luminous region z ~ 25 cm)
• Tracking out to ||<2
5
Particle Identification
• Charged leptons identified by characteristic energy deposition patterns
• Presence of neutrinos is inferred from energy imbalance – “missing energy”
• Because net pz of the scattering partons is not known, mostly work in the transverse plane (i.e pT, ET, missing-ET)
• B-jet identification uses the silicon tracker– 8 layers, 704 ladders,
722432 channels– Total sensor area = 6 m2
– SVX II – 5 double-sided layers (r + rz)
– L00 r only – mounted directly to beampipe (R = 1.4 cm)
6
The CDF II Trigger System
• Interaction rate very high, but most not “interesting”
• Limited bandwidth to mass storage – must be choosy
• Level1 system– Synchronous – no deadtime– Single CAL towers (photons
and jets), COT tracks (with pair correlations), track-tower matches (electrons and taus), muons, missing energy
• Level2 system– Asynchronous - ~5%
deadtime– All Level1 objects, plus CAL
clusters (jets) and silicon tracking
• Level2 accept triggers full detector readout (few % deadtime)
• Level3 runs a version of the offline reconstruction – final rate reduction before writing to tape
• Always tuning the system to accommodate higher luminosity
2.5 MHz crossing rate (396 ns)
Output 20-30 kHz
Synchronous
Latency 25-30 s
Output ~700 Hz
Output 70-90 Hz
Readout latency ~650 s
7
The Higgs Boson of the Standard Model
• Electroweak symmetry can be broken using the “Higgs mechanism”– 4 new scalar fields– Three give mass to the W’s
and Z– Other manifested as a single
scalar – the “Higgs boson”
• If there is such a particle, precision electroweak measurements favor a low mass– LEP2 searches exclude
mH < 114.4 GeV/c2 @ 95% CL
– SM fit requires mH < 166 GeV/c2 @ 95% CL (199 if including LEP2 direct searches) as of last year
• New CDF W mass (most precise single measurement) moves best fit to 80+36-26 GeV/c2 (was 85+39-28)
8
Higgs Production and Decay
Ideally, use gg H bb, WW
But, QCD bb background too high
For low mH, use WH+ZH, H bb (associated production)
At high mH the WW decay mode opens up – can use gg H production
(pb-1)
9
The H WW* l l Channel
• Largest BR for mH > 135 GeV/c2
• Uses gg H production– Now including VBF
(6-10% extra cross section)
• Event selection– Two isolated leptons with
pT > 20 and 10 GeV/c– Opposite charge– Missing-ET > mH/4– If missing-ET aligned with
lepton, > 50 GeV– mll > (mH/2)-5 GeV/c2
– pT,1+pT,2+missing-ET < mH
– Jet veto
• Including W BR’s, acceptance is 0.3-0.7% depending on mH
W l BR not included
10
H WW* Backgrounds
• Predominantly WW• Also Drell-Yan and other
diboson channels, and from fake leptons
• Not possible to reconstruct Higgs mass due to multiple neutrinos
• Can exploit scalar nature of Higgs– Leptons from H WW*
are closer in
• Treat each bin of as a separate counting experiment
W-
W+
e+
e-
11
H WW* Cross Section Limits
• Sensitivity about 6x SM level at 160 GeV
• For 6 fb-1 and combined with DØ, within factor 2 of SM
• Get more events
• WZ search almost doubled acceptance by adding new lepton types
• Get more out of events
• More than just
• Multivariate (NN)
• Matrix elements
exc
lud
ed
at
95%
C.L
.
12
Higgs Production and Decay
Ideally, use gg H bb, WW
But, QCD bb background too high
For low mH, use WH+ZH, H bb (associated production)
At high mH the WW decay mode opens up – can use gg H production
(pb-1)
13
B-Jet Identification (b-tagging)
• B-hadrons are long-lived – search for displaced vertices
• Construct event-by-event primary within beamspot (10-32 m)
• Fit displaced tracks and cut on Lxy significance ( ~ 200 m)
• Calibrate performance from data (low-pT lepton samples)
Tag this jet
b-fraction ~80%
measure tag efficiency in data and MC
14
Fake B-Tags (mistags)
• Fake tags are (almost) symmetric in displacement Lxy
• Rate of tags with Lxy<0 is a good estimate for the mistag rate
• Parametrize mistag rate which can be applied to any sample
• ~30% correction for tags from /KS and interactions with detector material
Lxy > 0
Lxy < 0
15
The WH lbb Channel
• Event selection– Isolated e or with
pT>20 GeV/c– Missing-ET > 20 GeV– Exactly two jets with
ET>15 GeV– At least one b-tagged jet
• Acceptance is 1.8-2.1%
• Backgrounds include– Non-W events (fake lepton,
fake missing-ET, b decays)– W + mistagged jets– W + heavy flavor jets– Diboson production
(WW, WZ, ZZ)– Z – Top quark production
(including single top)
• This channel uses a neural net filter on the b-tags to reject half of the background (~10% signal loss)
16
WH Backgrounds
Use W+1-jet bin to test W+HF bkgd
Top pair cross sectionMeasured from theW+3,4-jets events
SM Higgs would be about two events
17
WH Dijet Mass
At least one jet b-tagged with NN Both jets b-tagged
18
WH Cross Section Limits
exc
lud
ed
at
95%
C.L
.
19
The ZH bb Channel
• Distinctive final state of b-jets recoiling against missing-ET
• Event selection– Missing-ET > 75 GeV– Lepton veto– Exactly two jets with ET >
60 and 20 GeV– Missing-ET not aligned with
either jet• Acceptance is 0.5-0.8% (for ZH)
– About half the events would be WH with lost lepton
• Backgrounds include– QCD with fake missing-ET
– QCD bb production– W/Z + jets– Top production– Diboson production
Missing ETb-jet
b-jet
y
x
2nd jet
Fake Missing ET
1st jet
180o
A di-jet QCD event:
20
ZH Dijet Mass
21
ZH Cross Section Limits
22
The ZH llbb Channel
• Look for ZH also in Z decays to charged leptons (e or )
• Lose in BR, gain in background
• Z+jets and top pairs
• Instead of dijet mass, enhance S/B using neural networks
23
ZH Cross Section Limits
24
Combined Limits (Relative to SM)
• Not including new WW*
• Factor of ~8 away from SM prediction at 115 GeV
• Expect factor 2-3 from more luminosity
• Another factor of √2 from combination with DØ
• As with WW*, need to improve the analyses in order to reach SM sensitivity
25
Multivariate B-Tagger
• Multivariate b-tag algorithm in the pipeline
• Much like the one used by SLD
• Plots are for simulation only, performance characterization on data is in progress
• Up to 30% higher b-tag efficiency compared to the NN tagger already used in the WH search channel
• More double-tagged events
• Better S/B
• Better dijet mass resolution
26
Double-Tagging in WH
Similar limits for exclusive 1-tag and 2-tag samples
20% improvement over inclusive ≥1-tag result
27
The W/Z+H qqbb Channel
• Was competitive with the leptonic W channel in Run I (but a little lucky)
• 70% of W/Z decays are into hadrons
• Base selection is four jets with two b-tagged– Search in tagged dijet mass– Can use the four-jet trigger
designed for all-hadronic top pairs – also working on a new dedicated trigger
• Background dominated by QCD multijet production (with real tags)– Data-driven estimates in
progress
28
Higgs in the MSSM
• Minimum of five scalars: h, H, A, H+, H-
– Separate couplings for up-type and down-type fermions
• Properties of the Higgs sector largely determined by mZ and by two other parameters:
– mA : mass of pseudoscalar– tan : ratio of down-type
to up-type couplings
• If tan is small, then h looks a lot like a standard model Higgs
• If tan is large– Production via b quarks
can be greatly enhanced (factor ~tan2)
– Decays to bb (~90%) and (~10%) dominate
• LEP-II searches have excluded mA<93 GeV/c2 and tan < 5-10
b
0
b
b
0
29
MSSM Higgs Masses
• At high tan, the A becomes nearly degenerate with h or H– Can search for single resonance, double the cross section
• The other neutral Higgs is like SM Higgs (no production enhancement)
30
Scenario Dependence
• Higgs properties are largely, but not completely, determined by mA and tan
• Loop corrections introduce dependence on other SUSY parameters– M. Carena et al., Eur.Phys.J. C45
(2006) 797-814 (hep-ph/0511023)
• b is a function of the other SUSY parameters and depends on the “benchmark” scenario
• b ~ tan (sign of critical)
• For tan = 50, = -200– mh
max: b = -0.21
– no-mixing: b = -0.11
• Need both channels to get the full story
31
The gg+bb Channel
• High cross section and unique final state (not QCD)
• Best signature is one decay into e or and the other hadronically (46% BR)– Now also e+ channel (+6%)
• Event selection– One e or with pT > 10
GeV/c– One hadronic with pT > 15
GeV/c, mass < 1.8 GeV/c2 (including 0’s)
– Or, e+ with pT > 6 GeV/c– Opposite charge– Missing-ET not recoiling
against leptons (rejects W l)
• Acceptance is 1.2-2.1%
• Backgrounds include– Z – W l +jet fake had– QCD multijet (both fake)
32
Higgs Discriminant
• Because of multiple neutrinos cannot reconstruct invariant mass• In cases where taus are not back-to-back can use missing-ET projections
– Low efficiency• Instead, use “visible mass”
– Mass of the visible parts of the two taus and the missing-ET
33
Sample Fit
• Signal is normalized to 95% CL exclusion limit
34
MSSM Di-Tau Limits
• Background-only pseudoexperiments indicate <2 significance when considering the entire mvis range
35
MSSM Interpretation
36
The gg bb bbbb Channel
• bb final state not unique enough• Require one of the additional b’s
to have high pT and b-tag it• Results are for a different cross section
than the case (factor ~4)
• Basic event selection is three b-tagged jets• Search in dijet mass of two leading jets
• Background expected to be ~100% QCD multijet production
• Almost all bbq, bbc, bbb• Data-driven estimates exist
• At high tan the Higgs can developsignificant width – fit templates arefunction of cross section
• We expect CDF result on 1 fb-1 outwithin a few weeks
*editorial comment belongs to me, not DØ
Dawson, Jackson, Reina, Wackeroth hep-ph/0603112
?*
37
SM Higgs at the LHC
• Discovery prospects are excellent
• Strategies for low-mass region are different – more focused on backgrounds than cross section
• Experience gained at the Tevatron will be very useful– tt event reconstruction– dijet mass reconstruction– W/Z + jets background
estimation techniques– b-tagging and ID at
hadron colliders
• Starting from a bump at the Tevatron gets us there that much faster!
38
Tevatron Prospects
39
Summary
• CDF is searching for the Higgs in a variety of production and decay scenarios– Tools are in place to combine results from different channels– Lots of effort going into adding new channels and improving
the existing ones
• MSSM Higgs searches looking quite interesting
40
Backup Material
42
The Standard Model
• Matter is made out of fermions: – quarks and leptons– 3 generations
• Forces are carried by Bosons:– Electroweak: ,W,Z– Strong: gluons
• Higgs boson:– Gives mass to particles
• Much is still unknown– Is the EW symmetry really
broken by a Higgs? What kind of Higgs(es)?
– Are there any other particles? New gauge bosons (Z’, W’)? Extra generations of quarks? Extra dimensions? Superpartners?
• Let’s have a look!
HH
43
What about Neutrinos?
• Not nearly enough material to stop them in the detector
• Instead, infer their presence by energy imbalance
• i.e. add up everything you see, then reverse it
• Commonly called “missing-ET”
• Only get net neutrino momentum if >1
• Only works in transverse plane
44
PDFs
45
46
47
Non-W Background to WH Channel
• Use missing-ET and isolation ratio (assumed uncorrelated) in sidebands to extrapolate into signal region
• Isolation ratio = (lepton pT)/(non-lepton energy in cone with - radius 0.4 around the lepton)
Signal region D predicted by
Model event kinematics from sideband
48
B-Tag Efficiency Measurement
• Large b-hadron mass gives a wide pT,rel distribution relative to non-b contributions
• Fit untagged and tagged jets with b and one of four non-b templates to get b-tag efficiency
• Spread of results using the four non-b used as a systematic error
49
Dijet Mass Resolution
• Raw: what we use now
• H1: track + CAL energy flow
• MTL: correct for soft leptons
• Hyperball: multivariate nearest-neighbor algorithm, pick the most likely “true” dijet mass
50
W + jets Simulation
• Lots of activity in recent years• We use the ALPGEN generator
– Tree-level W + N partons– Also W+c+Np, W+cc+Np,
W+bb+Np• HERWIG parton shower adds
soft gluon radiation• Monte Carlo prediction
normalized to observed number of W+jets
• Fraction of events containing heavy quarks calibrated from data– b-tag rates in data and
ALPGEN multijet samples– Scale ALPGEN prediction by
1.5 0.4
51
Untagged “Control Sample”
52
ZH bb Backgrounds
• QCD bb background normalization fixed in Control Region 1 – extrapolate into others
• Other backgrounds checked in Control Region 2
• Now search in the signal region
53
The CDF II Detector
SOLENOID
= 2.0
END WALL HAD CAL.
CENTRAL OUTER TRACKER
CLC
CENTRAL HAD CALORIMETER
PLUG HAD CAL.
= 1.0
= 3.0
PLU
G E
M C
AL.
MUON CHAMBERS
Silicon Vertex Detecto
r
CENTRAL EM CALORIMETER
54
Electrons and Photons
Electron ID :
•Covers ||<3.6
•||<2 (w/ trk)
•ID eff. ~ 80-90%
Photon ID :
•Coverage : ||<2.8
•ID eff. ~ 80%
55
Muons
Muon ID :
•Coverage : ||<1
•ID eff. ~ 90-100%
56
Taus
h ID
cone
isolation
Tau ID :•Narrow iso. cluster
•Low # tracks
• 0 identification
•Coverage : ||<1
•ID eff. ~ 46%
Neutrinos inferred from energy imbalance
57
Jets and b-Tagging
Jet ID :
•Cluster of CAL towers
•Coverage : ||<3.6
Heavy Flavor Jet Tagging :
• Id HF jets via finding displaced vertex
•Coverage : ||<1.5
•Efficiency : 40-45%
x
y
z
Lxy
bdo