W ANTED :
T HE H IGGS B OSON
D EAD OR A LIVE ?
Kai-Feng Chen
National Taiwan University
Joint Physics Colloquium on January 3rd, 2012
The God Particle?
The Goddamn Particle? or
The Higgs boson is often referred to as
"the God particle" by the media, after the title of Leon Lederman's book.
Lederman initially wanted to call it the
"goddamn particle," but his editor would not let him...
Higgs?
ORIGIN OF MASS?
u d
c s
t b e
ν e
μ ν μ
τ ν τ Q U A R KS LE PT O N S
γ
FO R CE CA R R IE R S
g
W
Z
The Higgs Boson is an elementary particle predicted to exist by the Standard Model (SM). It is the last SM particle that has not yet been confirmed by the experiments.
The Standard Model describes:
-
How the particles interact;-
How different particles behave;-
How the force between particles are manifested.-
and, maybe explain the origin of mass?The Higgs mechanism was proposed in 1964 independently by three groups of physicists: by Peter Higgs; F. Englert and R. Brout; and by
G. Guralnik, C. R. Hagen, and T. Kibble. They were awarded the 2010 J. J. Sakurai Prize for this work.
The 1964 PRL papers by Higgs et al. displayed the field that would become the well known Higgs boson eventually.
After 47 years, it is still the major objective at the Large Hadron Collider!
THE HIGGS MECHANISM
AN ANALOGY
Image a fairly crowed airport terminal, people are scattering around normally...
THE HIGGS MECHANISM
AN ANALOGY
If a super star just arrives the terminal...
THE HIGGS MECHANISM
AN ANALOGY
Now Faye Wang acts like a massive object,
due to the fact that followers are strongly interacting with her....
A “massive”
particle
“Higgs bosons”
THE HIGGS MECHANISM
The Higgs mechanism operates in a way similar to this analogy.
Particles that have mass (e.g. weak force carriers and fermions) move through the Higgs field, interacting with the Higgs bosons.
Heavier particles interact more with the Higgs field taking on more mass, while massless particles (e.g. photons) have no direct
interactions with the Higgs boson.
b τ γ
Z W g
t
Massive particles =strong direct connections with Higgs
Massless particles =
connection with 2nd order loops
THE HIGGS MECHANISM
The Standard Model, which is based on the Lagrangian, must be symmetric under gauge transformations.
However, explicit mass terms for the gauge bosons are forbidden by gauge invariance. But the W/Z bosons are known to be massive!
The way out is provided by Spontaneous Symmetry Breaking (SSB).
The Lagrangian is still invariant but the gauge symmetry is broken by the vacuum.
In the simplest way, the SSB can be achieved by introducing one complex scalar doublet. This gives 4 degrees of freedom:
➡
3 give the masses to W+, W–, Z0 bosons.➡
1 left for the Higgs boson.In some of SM extensions may contain more Higgs doublets.
(= more Higgs bosons!)
HOW TO FIND THE HIGGS BOSON?
First, allocate 10 billion euros to build the Large Hadron Collider.
Second – persuade thousands of physicists and engineers to live in Geneva and work day and night.
Finally – be patient and wait for the conclusion.
Since the Standard Model is so successful, most physicists believe the (SM-) Higgs boson can be found by the experiments at LHC. But don’t be too surprise if people cannot find it.
If the Higgs is not found, it means that the current model (which is the simplest solution) is not working. But in this case, discovery of something else is almost
guaranteed.
LET’S HUNT FOR HIGGS!
How to look for Higgs through its decays?
How the Higgs bosons are produced?
What are the predictions and experimental bounds?
How to tell its (non-)existence in terms of probability?
What are the newest results from the LHC?
BUMP HUNTING IN A NUTSHELL
H
γ
γ
The Higgs boson
should be short lived, quickly decay into some other particles.
(e.g. photons)
# of event
M(H) The detector can measure
the decay products.
The Higgs mass can be
“reconstructed” using
the measured energy and momentum of the particles.
Collecting the measured mass from many events, the
Higgs mass bump should be visible.
BUMP HUNTING IN A NUTSHELL
γ
γ
# of event
M(H)
# of event
M(H) The Higgs boson
should produce a peak on the mass spectrum
Background (e.g. two random photons)
should generate a
“flatter”distribution.
H
γ
γ
If Higgs exists
background only
14
DECAY OF HIGGS
Comment #1
Decay to heaviest particles, if they are kinematically
allowed.
Comment #2
Decay to massless particles through loops.
Comment #3
“Best channel” is actually driven by background level.
BR(H) bb_
!+!"
cc_ gg
WW ZZ
tt-
## Z#
MH !GeV"
50 100 200 500 1000
10 -3 10 -2 10 -1
1
σ(gg → H → b¯b) ∼ 20 pb
σ(b¯b) ∼ 500 µb
H → γγ (BR ∼ 10
−3)
140 ∼< MH
∼< 180 GeV
H → W W
∗→ lνlν
MH∼< 140 GeV
the “best channel”
bb γγWW ZZ
BEFORE THE ERA OF LHC
80.3 80.4 80.5
155 175 195
mH [GeV]
114 300 1000
m [GeV]
m W [GeV]
68% CL
!"
LEP1 and SLD LEP2 and Tevatron
July 2011
top mass measurement
W mass
measurement
Dependency on Higgs mass
Since Higgs tends to interact strongly with heavy stuff – it is obvious that the
measurements of top quark and W/Z bosons will give us some hints of Higgs.
= “still” allowed region from direct searches (without LHC)
very limited space allowed for M(W), M(Top) & M(Higgs)!
BEFORE THE ERA OF LHC
0 1 2 3 4 5 6
100
30 300
mH [GeV]
!"2
Excluded
!#had =
!#(5)
0.02750±0.00033 0.02749±0.00010 incl. low Q2 data
Theory uncertainty
July 2011 mLimit = 161 GeV
= regions excluded by LEP and Tevatron
(to be discussed in next slides!)
M(Higgs) = 92 GeV, or < 161 GeV
+34–26Higgs mass is NOT predicted in the SM.
But if we assume the SM is 100%
correct and no other contributions,
the Higgs mass can be constrained by existing precision measurements:
In any case, we still need to observe the Higgs boson directly...
LEP DIRECT SEARCHES
LEP = Large Electron-Position Collider
was the collider right before LHC at CERN.
Four experiments on the ring (ALEPH, DELPHI, OPAL, L3).
Search for Higgs using the so-called
“Higgs-strahlung” process:
(Maximum Higgs mass reach)
√s − M(Z) = 206.7 − 91.2 = 115.5 GeV
LEP DIRECT SEARCHES
18
Mass plot(s).
0 10 20 30
Events / 3 GeV/c2
!s– = 200-210 GeV
LEP loose background hZ Signal (mh=115 GeV)
all cnd= 119 bgd= 116.51 sgl= 10.02
> 109 GeV 1715.76 7.1
0 5 10
Events / 3 GeV/c2
!s– = 200-210 GeV
LEP medium background hZ Signal (mh=115 GeV)
all cnd= 34 bgd= 35.69 sgl= 5.3
> 109 GeV 53.93 3.88
0 2 4 6
0 20 40 60 80 100 120
Reconstructed Mass mH [GeV/c2]
Events / 3 GeV/c2
!s– = 200-210 GeV
LEP tight background hZ Signal (mh=115 GeV)
cnd= 18all bgd= 13.97 sgl= 2.9
> 109 GeV 4 1.21 2.21
From a high efficiency ...
• The reconstructed mass is ONE of the discriminating variable.
(others include L or ANN output, b-tagging variable, ...)
• That’s the reason why we use them for “illustrative purposes only”
... to a high purity selection.
Pierre Lutz /SACLAY LEP Jamboree (page 5) 07/22/2002
Higgs Mass Lower Bound
LEP excludes a 114.4 GeV Higgs boson @ 95% CL.
(expected 115.3 GeV)
Exp. Obs.
ALEPH 113.5 111.4 DELPHI 113.3 114.1 L3 112.4 112.0 OPAL 112.7 112.7
10 -6 10 -5 10 -4 10 -3 10 -2 10 -1
1
100 102 104 106 108 110 112 114 116 118 120
mH(GeV/c2)
CL s
114.4 115.3
LEP
Observed Expected for background
Pierre Lutz /SACLAY LEP Jamboree (page 18) 07/22/2002
4 experiments combined
A mild hint found ~118 GeV, but it is hard to concluded ⇓
mostly H→bb
LEP combined
Observed limit >114.4 GeV
(Expected limit >115.3 GeV)
at 95% confidence level
= ±2σ (~95%) from background hypothesis
= ±1σ (~68%) from background hypothesis So called “CLS” method, to be addressed.
TEVATRON SEARCHES
Tevatron is the hadron collider at Fermilab with a CM beam energy of 1.96 TeV (the highest CM energy before the LHC!).
Two general purpose experiments: CDF and D∅.
Higgs are mostly produced through:
gluon fusion Higgs-strahlung
Although the searching mass region can be much higher than LEP
(which is totally limited by its CM energy), but the background level is also higher.
TEVATRON SEARCHES
The production rate for Higgs is almost 10 orders of magnitude lower than the QCD (=jets)
processes.
The search for Higgs is highly dependent on the background level. (e.g. a direct search for H→bb will not work, but need to tag the associated W/Z boson.)
Adopt multivariate analysis tools, such as neural network, matrix element, boost decision tree, etc.
Need to combine many analyses.
Higgs signal x 50
TEVATRON SEARCHES
Tevatron combined exclusion 156 < M(H) < 177 GeV/c2
at 95% confidence level
A global combination of 12 search channels
It’s already very close to the best sensitivity of Tevatron
(closed in 2011).
HOW TO READ THE LIMIT PLOT?
σ/σSM
M(H) 1
σ/σSM
M(H) 1
σ/σSM
M(H) 1
= +
A typical limit plot
The “expected”
limit curve and its uncertainties (±1σ,±2σ bands)
The “observed”
limit curve σ/σSM vs. M(H) =
limit on relative cross sections to the SM versus
the given Higgs mass
However, there are still some difficult questions:
Q1: What’s exactly the meaning of exclusion limit?
Q2: Why there are “expected” and “observed” curves?
Q3: How can we tell the strength of an excess?
M(H) 1
σ/σSM
m1 m2
Comment #1
Any region above the “observed limit” curve is excluded. The “σ/σSM = 1” is excluded
between m1 and m2, indicates SM Higgs with M(H) ∈ [m1,m2] is excluded.
M(H) σ/σSM
excluded
excess
deficit
Comment #2
If the “observed limit” is above the “expected limit”, one can interpret such behavior as an
“excess”. But one cannot read the significance (# of σ) from such an exclusion plot.
CONSTRUCTION OF LIMIT
AN EXAMPLE
Suppose you have a magic Swiss-franc coin.
You want to toss it (do experiments!) to know if it has equal probabilities for
head and tail.
However, since it’s a MAGIC coin,
it will cost you $100,000 per tossup
(ouch!)...
So you prepared a proposal, explained the importance of this experiment, and fired it to the funding agency. Fortunately you
received $1000,000 to do this experiment. Congratulations!
So you get the right to toss it for 10 time...
CONSTRUCTION OF LIMIT
AN EXAMPLE
Fair Model Cheat Model
Alien Model
You-Must-Be-Very-Lucky Model
50% 50% 0
40% 60% 0.2 20% 80% 0.6 1% 99% 0.98 First, do a survey on arXiv and check
several possible predictions:
Before doing the experiment, we shall use Monte Carlo (pseudo
experiments) to examine what is the EXPECTED lowest “unfairness”
can be excluded with only 10 tossups, at the 95% confidence level.
Then we can just toss the coin 10 times and obtain the OBSERVED limit on the “unfairness”, also at the 95% confidence level.
define the
“UNFAIRNESS”
Generate Monte Carlo (10 tossups x 10000 trials x 2 hypothesis) according to
The Null Model
(“unfairness” = 0)
(4,6) (1,9) (7,3) (5,5) (6,4) ...
The Alternative Model
(“unfairness” = 0.6, as an example)
(1,9) (0,10) (3,7) (6,4) (2,8) ...
Then collect the trials and for each hypothesis and for data (the real 10 tossups):
For each trial (=10 tossups), the relative likelihood (or Δχ2) can be calculated:
-2lnQ Null Model
x 10000 trials
Alternative Model
x 10000 trials
# of trials
The Data
(the real 10 tossups) x only 1 trial
−2 ln Q = −2 ln(Lalternative/Lnull)
One can use these two distributions to calculate
“
CL
b” and “CL
s” for the statistics analysis.CL
b = 1 – “p-value”CL
s+bCL
s= CL
s+b/CL
bFor a given “unfairness”, one can obtain a CLs value:
Tuning the given “unfairness”, until the CLs value is equal to 0.05 (=1 – 95%):
Higher “unfairness”
Lower “unfairness”
At this moment, the given
“unfairness” is the
OBSERVED limit at 95%
CLs confidence level.
Replace the data point (↓) with the average (and its width) of the null model to
determine the EXPECTED limit and the associated uncertainty band with the same
“unfairness” tuning.
average
±1σ,±2σ
uncertainty bands
HERE WE GO...
Limit on the “unfairness”
Y-model (0.98)
0.59(expected) 0.72(+1σ) 0.89(+2σ)
A-model (0.6) C-model (0.2)
0.43(–1σ) 0.29(–2σ)
F-model (0)
Tail Head Observed Limit
10 0 0.26
9 1 0.29
8 2 0.35
7 3 0.42
6 4 0.50
5 5 0.59
4 6 0.72
3 7 0.81
2 8 0.91
1 9 0.98
0 10 N/A
The EXPECTED limit is 0.59 for 10 tossups =
We expected to exclude any model that gives an “unfairness” > 0.59 at 95% C.L.
For the OBSERVED limit, we have to do the experiment now!
Actually we need at least 100 tossups to exclude the C-model (40%/60%)!
EXCESS?
Tail Head Scanned p-value for A-model
10 0 1.0
9 1 0.999
8 2 0.99
7 3 0.95
6 4 0.83
5 5 0.62
4 6 0.38 (0.9σ)
3 7 0.17 (1.4σ)
2 8 0.055 (1.9σ)
1 9 0.011 (2.5σ)
0 10 0.00098 (3.3σ)
The strength of an excess is given by the “p-value” (=1–CLb), defined by the likelihood that the observed data is actually a fluctuation from null hypothesis.
(lower p-value = stronger excess; higher p-value = weaker excess.)
1σ
0 2σ 3σ 4σ 5σ
p=0.317 (=1–68.3%)
p=0.00270
⇓ p=0.000000573 ⇓
Fluctuation
Hin t
Evidence
Discovery
BEFORE MOVING FORWARD...
The constraints from EWK precision data prefer a light Higgs boson:
if the Standard Model is correct.
The direct searches from LEP exclude the SM Higgs boson below 114.4 GeV/c2. The analyses at Tevatron exclude 156~177 GeV/c 2.
Exclusion of Higgs should be
calculated by the “CLs” statistics method introduced above.
LHC’s major objectives – find or fully exclude the SM Higgs, and look for any possible new physics scenarios.
M(H) = 92 GeV/c
+34–26 2It’s Showtime!
THE LARGE HADRON COLLIDER
Lake Geneva
LHCb
ATLAS ALICE
CMS
CERN main campus Mt. Jura
27 k m
Geneva airport
THE LHC AT CERN
2 b-tagged jets
The LHC is the proton–proton collider at CERN, primary physics targets are:
-
The origin of mass, the Higgs boson.-
What is the dark matter!? Supersymmetry particles?-
Matter versus antimatter: the CP violation.-
Understanding of the space and time.-
and many others...7 experiments:
-
General purpose: ALTAS and CMS.-
B-physics: LHCb.-
Heavy ion: ALICE.-
Forward physics: TOTEM and LHCf.-
Monopole search: MoEDAL.Start its 7 TeV (3.5 on 3.5) run since March 2010.
CERN prepared a nice X’mas gift for the particle physicists all over the world.
Atlas and CMS experiments reveal their newest results on Standard Model Higgs searches.
THE DECEMBER 13 EVENT
THE ATLAS EXPERIMENT
3000+ scientists and engineers (including ~1000 students)
from 174 institutes in 38 countries.
From Taiwan:
Academia Sinica
The colorful
Atlas control building
THE CMS
EXPERIMENT
3000+ scientists and engineers (including ~840 students)
from 173 institutes in 40 countries.
From Taiwan:
NTU and NCU
w/o track
Track ECAL Shower
HCAL Shower Less calorimeter
energy
+ +
HCAL Shower
ECAL Shower
Track
Track
w/o track
Muon Chamber
+ + + +
Lots of tracks ECAL & HCAL
Shower
+
MUON ELECTRON
CHARGED HADRON (e.g. proton, pion)
NEUTRAL HADRON (e.g. neutron)
PHOTON
(e.g. light quark, gluon)
JET
MISSING ENERGY
(e.g. neutrino)
invisible, look for detector energy imbalance
PARTICLE DETECTION
photon
(isolated EM shower)
photon
HCAL showers (produced by hadrons)
tracks
(mostly charged pions)
a H→γγ candidate from CMS
HIGGS PRODUCTIONS AT LHC
38
H t, b
g g
H
q q
W, Z W, Z
H g
g
Q
Q¯
¯ H q q
W, Z
W, Z
Q ¯Q
W, Z
SM Higgs production at LHC
Gluon fusion (gg! H) is the dominant production mechanism at LHC.
Irreducible backgrounds in H ! WW, ZZ, !! are from qq annihilation. Signal to Noise better than at Tevatron except in VH. VBF and VH also very useful at LHC
gluon fusion
vector-boson fusion (VBF) Higgs-strahlung (VH)
Production rate of Higgs at LHC is roughly 10x to the Tevatron. Overall
S/N is better in principle.
Still dominated by gluon fusion, while the vector- boson fusion and Higgs- strahlung channels are also very useful.
THE CHALLENGE
dijet 108 pb
Huge background from non-Higgs processes Events produced in 2011 per Experiment
Higgs 5,000~50,000
(~100 visible)
ZZ ~35,000 WW ~200,000 Z ~150,000,000 W ~500,000,000
dijet ~500,000,000,000
The actual Higgs signal is highly dependent on its mass.
THE CHALLENGE
An Atlas event with 20 reconstructed vertices
A Z→μμ candidate
= 20x actual resolution (just for visibility)
The detectors record more than one interaction in a single snapshot. Number of interactions per crossing is high for higher luminosity. One has to pick up the right event from
the right interaction. This is price to pay.
41
HIGGS DECAY CHANNELS
SM Higgs Decay Modes Vs Mass
Mode! Mass Range! Data Used (fb-1)! CMS Document!
H ! ""! 110-150! 4.7! HIG-11-030!
H ! bb ! 110-135! 4.7! HIG-11-031 !
H ! ##! 110-145! 4.6! HIG-11-029!
H !WW !2l 2$! 110-600! 4.6! HIG-11-024!
H ! ZZ !4l! 110-600! 4.7! HIG-11-025!
H ! ZZ !2l2#! 190-600! 4.7! HIG-11-028 !
H ! ZZ !2l2j! 130-165/200-600! 4.6! HIG-11-027!
H ! ZZ !2l2$! 250-600! 4.6! HIG-11-026!
Channels with higher sensitivities
H→γγ for M(H)<130 GeV/c2
-
good mass resolution-
acceptable S/N (comparing to bb)H→ZZ(*)→4l for 125<M(H)<300 GeV/c2
-
good mass resolution-
best S/NH→WW(*)→2l2ν for 125<M(H)<180 GeV/c2
-
larger production rate-
good S/NH→ZZ→2l2ν for M(H)>300 GeV/c2
-
distinct signature (Z + missing energy)-
good S/NATLAS ANALYSES
channel mass range
(GeV/c2)
Luminosity
(fb–1) Number of signals S/N
H→γγ 110~150 4.9 ~70 ~0.02
H→ττ→2l+ν 110~140 1.1 ~0.8 ~0.02
H→ττ→l+had. 100~150 1.1 ~10 ~0.005
HW/Z→bbl(l) 110~130 1.1 ~6 ~0.005
H→WW(*)→lνlν 110~300 2.1 ~20 (130 GeV) ~0.3 H→ZZ (*) →4l 110~600 4.8 ~2.5 (130 GeV) ~1.5
H→ZZ→2l2ν 200~600 2.1 ~20 (400 GeV) ~0.3
H→ZZ→2l2q 200~600 2.1 2~20 (400 GeV) 0.05~0.5
H→WW→lν2q 240~600 1.1 ~45 (400 GeV) 0.001
Three channels (which have highest sensitivities in the low mass region) were updated on Dec/13, the rest analyses were shown at the summer conferences already.
CMS ANALYSES
channel mass range (GeV/c2)
sub- channels
mass resolution
Luminosity (fb–1)
Document ID
H→γγ 110~150 4 1–3% 4.7 HIG-11-030
H→ττ 110~145 9 20% 4.6 HIG-11-031
H→bb 110~135 5 10% 4.7 HIG-11-029
H→WW→lνlν 110~600 5 20% 4.6 HIG-11-024
H→ZZ→4l 110~600 3 1–2% 4.7 HIG-11-025
H→ZZ→2l2τ 190~600 8 10–15% 4.7 HIG-11-028
H→ZZ→2l2ν 250~600 2 7% 4.6 HIG-11-027
H→ZZ→2l2q 130~164/200~600 6 3% 4.6 HIG-11-026
Channels with best mass resolution: H→γγ and H→ZZ(*)→4l
All eight analyses were shown on Dec/13; CMS combination documented in HIG-11-032.
ALL
Ready for tons of LHC results?
HIGH MASS: H→ZZ→2l2ν
Find a clean Z candidate + extra missing energy.
Major background sources: Z+jets, top pair, WW.
Reconstruct the transverse mass, MT.
MT shape analysis improves the sensitivity by 10%.
270–400 GeV/c2 excluded
H(400)x1
HIGH+LOW MASS:
H→ZZ→2l2q
Find a clean Z→ll candidate + Z→2jets candidate, no missing energy.
Categorized by presence of 0,1,2 b-jets Major background sources: Z+jets.
Use scaler Higgs assumption in an angular likelihood discriminant.
Background normalized to data sideband.
M(lljj)=580 GeV/c2
H(400)x100
2 electrons
2 jets
Low Mass High Mass H(150)x5
H(400)x2
HIGH MASS: H→ZZ→2l2τ
10 observed events, 10.2 expected background.
Background shapes are taken from MC simulation and normalized to the values obtained using
data-driven techniques.
H(400)x1 H(200)x1
H→WW→2l2ν
Only two oppositely charged, isolated leptons and some missing energy.
With additional 0,1,2 jets from vector- boson-fusion process.
Scalar Higgs ⇒ small opening angle (ΔΦ) between two charged leptons.
Two neutrinos in the event, cannot form a mass peak ⇒ a counting experiment.
Challenge to remove the large backgrounds.
Large background
from ttbar/WW/Z+jets.
Collinear leptons
electron (pt=34 GeV) muon
(pt=32 GeV)
ΔΦ
missing energy (47 GeV)
⇓H(150)x1
ATLAS H→WW→2l2ν
Events with full selection criteria Events with full selection criteria
Observed data 94 events
(10 ee/42 eμ/42 μμ) Expected background 76 ± 11
Expected signal, M(H)=130 19 ± 4
145 < M(H) < 206 GeV/c2 excluded at 95% C.L.
where selection changes
transverse mass after full selection
H(130)x1
H(150)x1
~1.9σ for M(H)=130 GeV/c2
CMS H→WW→2l2ν
Two independent studies: cut-based versus BDT(boost decision tree, shown here).
Categorizing the events with same flavor/
opposite flavor/0,1 jet.
H(130)x1
H(130)x1
129 < M(H) < 270 GeV/c2 excluded at 95% C.L.
opposite favor, 1 jet
opposite favor, 0 jet
THEGOLDEN CHANNEL H→ZZ→4l
Mass can be fully reconstructed;
no missing particles.
Best width and mass resolution.
Extremely clean: S/N ~ 1 The best single channel
for Higgs discovery!
CMS 4μ candidate
Atlas 4μ candidate
ATLAS H→ZZ→4l
Expected background: 62 ± 9 Observed events: 71
Counting in M(4l)<180 GeV/c2 Expected background: 9.3 ± 1.5
Observed events: 8
Zoom in
Full region
180
3 events in
ATLAS H→ZZ→4l
135 < M(H) < 156 GeV/c2 181 < M(H) < 234 GeV/c2 255 < M(H) < 415 GeV/c2
High mass Low mass
Excluded at 95% C.L.
The 3 events produce a local p-value ~2.1σ
CMS H→ZZ→4l
Expected signal
distributions Full region
For M(4l)<160 GeV/c2
Expected background: 67.1 ± 6.0 Observed events: 72
Expected background: 9.5 ± 1.3 Observed events: 13
Zoom in
3 events in
CMS H→ZZ→4l
Full region Low mass
134 < M(H) < 158 GeV/c2 180 < M(H) < 305 GeV/c2 340 < M(H) < 460 GeV/c2
Excluded at 95% C.L.
The 3 events also produce a local p-value ~2.5σ around 119 GeV, but the atlas excess is around 125 GeV...
High mass regions were more or less excluded:
With H→WW→2l2ν along, no Higgs in 129~270 GeV/c2 With H→ZZ→2l2ν along, no Higgs in 270~400 GeV/c2
With H→ZZ→4l along, pushes to limit to 460 GeV/c2
Have to go to low mass, which is very difficult for hadron colliders...
LOW MASS SEARCHES:
H→bb
At low mass, H→bb is the dominant
channel, but overwhelmed by enormous QCD dijet background (S/N<1/1M..).
The best option: qq→VH→Vbb.
Tag another vector boson + strong b- tagging + BDT analysis.
Reconstruct Z(→ll,νν)H and W(→lν)H. b
b
W(→eν)H W(→μν)H
displaced vertices from b-jet
Tag a clean/boosted W/Z
from Higgs-strahlung production
59
LOW MASS SEARCHES:
H→ττ
SM Higgs production at LHC
Gluon fusion (gg! H) is the dominant production mechanism at LHC.
Irreducible backgrounds in H ! WW, ZZ, !! are from qq annihilation. Signal to Noise
Gluon fusion,
with 0 or 1 jet in addition
Boosted mode,
with 1 high pT (>150 GeV/c) jet
Vector-boson fusion with 2 additional jets (best channel)
H(120)x5
H(120)x5
H(120)x5
LOW MASS SEARCHES:
H→bb & H→ττ
H→bb H→ττ
Search for Higgs in low mass region
is very difficult (background level is very high).
The best channel is still H→γγ (low branching fraction, but cleaner).
LOW MASS SEARCHES:
H→γγ
Very simple final state: two isolated photons.
Smaller effective cross-section: σ~40 fb
Smooth irreducible background (S/N~0.02)
Experimental aspects:
-
Reject non-photon background (fakes from jets, π0, etc.)-
M(γγ) mass resolution and calibration-
Vertex finding-
Optimizing the sensitivities in different event categories.pT = 86 GeV
pT = 56 GeV no hard tracks,
just two ECAL clusters
LOW MASS SEARCHES:
H→γγ
σ~1.9 GeV/c2 σ~1.4~2.3 GeV/c2
Simulated
H→γγ peak ⇒
⇐ The calibration of calorimeter has been confirmed by the Z→ee data.
The resolution of peak position is much better then the width.
Width of the peak is
around 2 GeV.
LOW MASS SEARCHES:
H→γγ
⇐ The Atlas calorimeter has
fine η segmentations (4mm strips) can well separate
π0 and photon.
beam axis
Deduce Z of primary vertex
Z(γ1) – Z(γ2) ⇒ Calorimeter pointing capability reduces vertex uncertainty to 1.5 cm
ATLAS H→γγ
The events around 126 GeV has a local p-value ~2.8σ Global p-value is around 1.5σ only.
(to be discussed later)
CMS H→γγ
Background is normalized to data.
(conservation of excess & deficit...)
The events around 124 GeV has a local p-value ~2.5σ away from Atlas excess by 2 GeV
(already differ by a full width ~ unlikely to be a fluctuation...) H(120)x5
If we can combine everything together, shall we become stronger?
CMS COMBINED
Combining all analyses into single limit plot
Excluding 127~600 GeV/c2 at 95% C.L.
(expected to exclude 117~543 GeV/c2)
ZOOM IN THE LOW MASS
We cannot exclude the presence of the SM Higgs boson below 127 GeV/c2 because of a modest excess of events in the region
between 115 and 127 GeV/c2.
all combined bb+ττ+WW
(low resolution channels)
γγ+ZZ(4l)
(high resolution channels)
ATLAS COMBINED
Excluding 112.7~115.5 GeV/c2 131~237 GeV/c2
251~453 GeV/c2 at 95% C.L.
(expected to exclude 124.6~520 GeV/c2)
Remark: some of the channels were not yet updated from the summer analysis (1 fb–1).
low mass region
THE EXCESS(ES) IN LOW MASS
Maximum local significance ~3.6σ
(2.8σ H→γγ + 2.1σ H→4l + 1.4σ H→2l2ν )
at 126 GeV /c2 Maximum local significance ~2.6σ
at 119 GeV /c2 and 124 GeV/c2
(H→4l) (H→γγ)
Getting excited about the high significance? Wait a minute!
We still have to consider the
LEE
(Look-Elsewhere Effect)...THE LOOK-ELSEWHERE EFFECT
The look-elsewhere effect is a phenomenon, where an apparently statistically significant observation may have actually arisen by chance because of the size of the parameter space to be searched.
– from Wikipedia
⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑ ⇑...
EXPERIMENTS
Actually, this is not a single experiment.
When we look for different Higgs masses, these are equivalently MANY experiments.
The significance is overestimated if we take the maximum local value.
Maximum local significance: 2.6σ
LEE corrected in full range (110~600): 0.6σ
LEE corrected in low mass (110~145): 1.9σ
Maximum local significance: 3.6σ
LEE corrected in full range (110~600): 2.2σ
LEE corrected in low mass (110~146): 2.5σ
The excess observed in the low mass region has a modest statistical significance and could be still reasonably a fluctuation of the background.
THE LOOK-ELSEWHERE EFFECT
Analogy #1
In our magic coin example, if you have examined a bag of coins, the chance to find an unfair coin is definitely higher than just one coin.
Analogy #2
Surely you can find many peaks on a random noise distribution. It is not too difficult to find a single peak with 3σ as well.
This is the same as the Higgs hunting, scanning over a large mass region.
THIS IS EXPECTED...
Don’t be too disappointed: actually both experiments do not expect to see a >3σ effect in low mass region with the current data sets:
Both experiments only expect to see a 2~3σ excess in 115~127 GeV/c2 even if SM Higgs is there....
The expected p-value if SM Higgs exists
CMS
VERSUS ATLAS
The excesses from two experiments differ by 2 GeV. This is hard to be explained by the peak resolution (since the Z mass
peak is already well calibrated).
The Atlas excess at 126 GeV/c2 is
supported by both H→γγ and H→γγ, but it’s too close to CMS lower bound.
The CMS excesses are at different places
(H→γγ is at 124 GeV/c2, H→γγ is at 119 GeV/c2) CMS exclusion
Atlas
exclusion
BEST FITTED CROSS
SECTIONS
H→γγ
H→4l
All combined
BEST FITTED CROSS SECTIONS
M(H)=119.5 GeV/c2 M(H)=124 GeV/c2
CLOSING REMARKS
“The main conclusion is that the Standard Model Higgs boson, if it exists, is most likely to have a mass constrained to the range 116-130 GeV by the ATLAS experiment, and 115-127 GeV by CMS. Tantalising hints have been seen by both experiments in this mass region, but these are not yet strong enough to claim a discovery.”
OFFICIAL CERN STATEMENT
This is a true, accurate statement.
The Standard Model Higgs boson is not the only target of LHC experiments.
We are looking for almost everything we can think of, but nothing found (yet).
(ie. no SUSY, no extra dimensions, no new fermions, no new vector bosons, no black holes, etc....)
CLOSING REMARKS
CMS and Atlas have published 200 papers
(with another 200 papers in the pipeline), but none of them shown a significant deviation from the Standard Model.
We already know that the Standard Model is imperfect, even with a not-yet-concluded light Higgs boson.
(e.g. non-zero neutrino mass, hierarchy problem, etc.)
It is very strange that we observed nothing other than the Standard Model particles.
Maybe we just need to wait a little bit more time.
In year 2012, LHC will deliver 20 fb–1 integrated luminosity, probably at 8 TeV.
We shall be able to draw a final conclusion for the Higgs boson,
and maybe, even better, we start to see some other new physics signals.
CLOSING REMARKS
If you believe in Higgs, you may safely interpret the excess as the Higgs boson, and continue your reach life.
If you don’t believe in Higgs, you can still claim this is just a statistical fluctuation of background.
Then just wait for one more year for the final judgement.
BACKUP SLIDES
83
H t, b
g g
H
q q
W, Z W, Z
H g
g
Q
Q¯
¯ H q q
W, Z
W, Z
Q ¯Q
W, Z
BR(H) bb_
!+!"
cc_ gg
WW ZZ
tt-
## Z#
MH !GeV"
50 100 200 500 1000
10 -3 10 -2 10 -1
1
σ(gg → H → b¯b) ∼ 20 pb
σ(b¯b) ∼ 500 µb
H → γγ (BR ∼ 10−3)
140< M < 180 GeV H → W W∗ MH
∼< 140 GeV Higgs width
Higgs production at LHC
Higgs decay
EPS ver. EPS ver.
FROM
PRESHOWER TO
HIGGS
Higgs searches are very difficult, but it is not due to the rarity...
We all want to work on Higgs!
Unless we are well prepared, it’s better not to join the battle too early...
OUR COMMITMENT:
THE PRESHOWER DETECTOR
Endcap ECAL
Preshower
A THIN detector, only 19.52cm!
Endcap HCAL
OUR COMMITMENT:
THE PRESHOWER DETECTOR
NTU+NCU team dominate the Preshower group!
WHY PRESHOWER?
Higgs
γ γ
γ γ
Single incident (isolated) photon
Two closely-spaced incident photons
γ
γ γ
Silicon sensors had chosen for improving
spatial resolution for endcap ECAL.
WHY PRESHOWER?
We built the Preshower detector
We maintain the detector
We provide the calibration and alignment constants.
We develop the algorithm to suppress π0 background
Then, we can finally join the work on H→γγ!
It’s a long track, but if we can maintain all the required work very well, we can eventually be part of the Higgs analysis!
Q&A
Q: Why does the current search stop at 600 GeV?
A: The SM Higgs boson of high mass becomes very wide, which leads to large theoretical uncertainties in predictions of its production and mass line shape m(H*). For a SM Higgs boson with mass of 600 GeV, the production uncertainties are about 30% and rapidly grow for higher masses. Advanced theoretical calculations for the SM Higgs boson with a mass greater than 600 GeV are expected next year, according to the LHC Higgs cross section group.
Q: You say the mass range [127, 600] GeV is excluded at 95% CL. Is it sufficient level of confidence? Is the presence of a SM Higgs boson there now truly excluded or…?
A: No, it is never excluded at 100%. A 95% exclusion is a common practice in HEP. One can read exclusion limits at any desired confidence level form the CLs plot we provide as a part of our results. For example, at 99% CL, today's results exclude the mass range of [128,525] GeV.
Q: Last summer you had an excess at ~140GeV, which was similar in shape and significance to the one you have now at ~124GeV. Did the 140 move down to 124? Or is the 140 excess still there? If it’s the latter, can the 140GeV excess be an indication for a BSM Higgs with a x-section significantly lower than the SM one for that mass?
A: No, the modest excess we had at 140GeV did not move down to 124GeV. Due to the excellent momentum resolution of our detector, these two modest excesses are seen in the data independently. In fact, the modest excess around 140GeV is still there, but its significance is now considerably smaller as the new data we collected since the EPS conference did not bring as many new events in that mass range. This is why we make it very clear that one needs be prudent and not get overexcited about modest excesses of events.
Q&A
Q: Is the observed excess an indication for the Higgs boson ?
A: With the current amount of data, the excess is not unlikely to be a plain statistical fluctuation. On the other hand, it is not inconsistent with the Standard Model Higgs boson either. It also may turn out to be due to some other unaccounted backgrounds. Much more data coming in 2012 will allow us to pin down the true nature of the observed excess.
Q: At the low end you had some excess at 140GeV in the summer, which went away and now there is a new excess at 124? Why did this excess move? How likely are these bumps to move around? How stable (reliable?) is this analysis?
A: No, the modest excess we had last summer at 140GeV did not move down to 124GeV. Due to the excellent momentum resolution of our detector, these two modest excesses are seen in the data independently. They do not move around. In fact, the modest excess at 140GeV is still there, but the number of events observed is already below that expected from SM backgrounds, hence we can exclude that this excess is due to SM Higgs. Should we mention potential BSM Higgs with production x-section significantly smaller than SM Higgs?
Q: What happens to the two dips in the p-value distribution if one were to eliminate one event from each one of them? Is any one of them become significantly more likely to be consistent with background-only interpretation and hence less signal-like?
A: If one were to eliminate one event from the 119GeV excess, its significance would have been reduced from 2.8 sigma down to 1.9 sigma, which makes it a lot less impressive. Eliminating one event from the 124GeV excess, its significance would have been reduced from 2.5 sigma down to 2.4 sigma, basically remaining intact.
Q&A
Q: Could the two bumps (119 & 124GeV) come from one source? If it's probable, that is an interesting piece of information. If on the other hand it's not, then we know that one has to deal with each one separately and maybe one of them is "worth" more than the other (e.g. see one candidate test for that in the previous question).
A: In principal no, due to the excellent momentum resolution of our detector. However, other physics effect, which are yet to be studied carefully may move a few events around.
Q: What's the right/relevant LEE for today’s analysis? Is it the full available mass range of (115,1000) or the not-yet-excluded range (<130 today or <140 last summer) or (take a deep breath!) since ATLAS has already showed its excess at 125GeV there is no need for the LEE here at all.
A: We are quoting two LEEs, one for the full mass range of (100,600) and one for the allowed mass range from the LEP direct searches and precision EWK fits of (110,145). They give you an idea for the sensitivity of this search and the significance of the current result.
Q: What's the likelihood of the composition of excess of events that CMS has in the different channels when comparing with SM expectations? This may sound trivial, yet it carried non-negligible weight since this excess may represent something (or nothing) very different (unlikely) from SM Higgs. The quantitative answer to this question may go a long way to substantiate a statement that if it looks like a duck and it walks like a duck, it is more likely to be (though we certainly cannot say it yet) a duck.
A: As we have shown in the X plot, we have excess of events in all decay channels that we studied so far.
They are compatible with the SM Higgs, but also with statistical fluctuations. The likelihood of the data from