RESEARCH ARTICLE

A New Boson with a Mass of 125 GeVObserved with the CMS Experimentat the Large Hadron ColliderThe CMS Collaboration*

The Higgs boson was postulated nearly five decades ago within the framework of the standardmodel of particle physics and has been the subject of numerous searches at accelerators around theworld. Its discovery would verify the existence of a complex scalar field thought to give mass tothree of the carriers of the electroweak forcethe W+, W, and Z0 bosonsas well as to thefundamental quarks and leptons. The CMS Collaboration has observed, with a statisticalsignificance of five standard deviations, a new particle produced in proton-proton collisions at theLarge Hadron Collider at CERN. The evidence is strongest in the diphoton and four-lepton(electrons and/or muons) final states, which provide the best mass resolution in the CMS detector.The probability of the observed signal being due to a random fluctuation of the background isabout 1 in 3 106. The new particle is a boson with spin not equal to 1 and has a mass of about125 gigaelectron volts. Although its measured properties are, within the uncertainties of thepresent data, consistent with those expected of the Higgs boson, more data are needed to elucidatethe precise nature of the new particle.

The standard model (SM) of particle phys-ics (13) describes the fundamental par-ticles, quarks and leptons, and the forcesthat govern their interactions. Within the SM, thephoton is massless, whereas the masses of theother carriers of the electroweak force, the W

and Z0 gauge bosons, are generated through asymmetry-breakingmechanism proposed by threegroups of physicists (Englert and Brout; Higgs;and Guralnik, Hagen, and Kibble) (49). Thismechanism introduces a complex scalar field,leading to the prediction of a scalar particle: theSM Higgs boson. In contrast, all known elemen-tary bosons are vector particles with spin 1. Inthe SM, the scalar field also gives mass to thefundamental fermions through a Yukawa inter-action (13). The Higgs boson is predicted todecay almost instantly to lighter particles.

The theory does not predict a specific massfor the Higgs boson. Moreover, the propertiesof the Higgs boson depend strongly on its mass.General arguments indicate that its mass shouldbe less than about 1 TeV (1013), although searchesfor the SM Higgs boson conducted before thoseat the Large Hadron Collider (LHC) have ex-cluded the mass region below 114.4 GeV (14).Searches at the Tevatron have excluded a narrowmass region near 160 GeV (15) and recently re-ported an excess of events in the range from 120to 135 GeV (1618).

The LHC is installed in a circular tunnel 27 kmin circumference and 100 m underground, strad-

dling the border between France and Switzer-land, near Geneva (19). The LHC acceleratesclockwise and counterclockwise beams of pro-tons before colliding them head on. These col-lisions were at a total center-of-mass energy of7 TeV in 2011 and 8 TeV in 2012, the highestenergies reached to date in a particle acceler-ator. These high-energy collisions enable theproduction of new, and sometimes very heavy,particles by converting energy into mass in ac-cordance with Einsteins well-known formulaE = mc2. The LHC can produce all known par-ticles, including the top quark, which, with a massof about 173 GeV, is the heaviest known ele-mentary particle. It was predicted that the SMHiggs boson could also be produced at the LHCif it has a mass less than about 1 TeV.

The SM predicts the cross section for theproduction of Higgs bosons in proton-protoncollisions as a function of its mass. The crosssection increases with the center-of-mass energyof the collision and decreases with increasingHiggs mass. Despite the high collision energy,the predicted probability of Higgs boson produc-tion is extremely small, about 1010 per collision.Thus, to detect a significant number of Higgsbosons a huge number of collisions must be an-alyzed, which requires very high luminosity. Themaximum instantaneous luminosity achieved sofar is 7.6 1033 cm2 s1, close to the LHC peakdesign value that was not expected to be attaineduntil 2015. This was achieved by having 1368bunches of protons in each beam, spaced 50 nsapart (corresponding to a separation of about16 m), with each bunch containing about 1.5 1011 protons squeezed to a transverse size of about

20 mm at the interaction point. Each bunch cross-ing yields more than 20 proton-proton collisions onaverage. The multiple collisions per bunch cross-ing, known as pileup, are initially registered as asingle collision event by the detectors. Resolv-ing the individual collisions within these eventsis an important challenge for the detectors atthe LHC.

The Compact Muon Solenoid (CMS) detec-tor surrounds one of the LHCs interaction points.Heavy particles, such as SMHiggs bosons, createdin LHC collisions will typically be unstable andthus rapidly decay into lighter, more stable par-ticles, such as electrons, muons, photons, andhadronic jets (clusters of hadrons travelling in asimilar direction). These long-lived particles arewhat CMS detects and identifies, measuring theirenergies and momenta with high precision inorder to infer the presence of the heavy particlesproduced in the collisions. Because the CMS de-tector is nearly hermetic, it also allows for thereconstruction of momentum imbalance in theplane transverse to the beams, which is an im-portant signature for the presence of a neutrino(or a new, electrically neutral, weakly interactingparticle) in the collision.

We report the observation of a new particlethat has properties consistent with those of theSM Higgs boson. This paper provides an over-view of the experiment and results that aredescribed in greater detail in (20). The study ex-amines five SM Higgs boson decay modes.Three modes result in pairs of bosons (gg, ZZ, orW+W), and two modes yield pairs of fermions(bb or t+t), where g denotes a photon, Z and Wdenote the force carriers of the weak interaction,b denotes a bottom quark (and b its antiquark),and t denotes a tau lepton. In the following, weomit the particle charges and use b to refer to boththe quark and antiquark. The unstable W, Z, b,and t particles decay to final states containing elec-trons, muons, neutrinos, and hadronic jets, all ofwhich can be detected (directly or, in the case ofneutrinos, indirectly) and measured with the CMSdetector. An independent observation was madeby the ATLAS collaboration (21, 22), which fur-ther strengthens our interpretation.

Overview of the CMS detector. The CMSdetector measures particles produced in high-energy proton-proton and heavy-ion collisions(23). The central feature of the detector is a su-perconducting solenoid 13 m long, with an in-ternal diameter of 6 m. Within its volume itgenerates a uniform 3.8-T magnetic field alongthe axis of the LHC beams. Within the field vol-ume are a silicon pixel and strip tracker, a leadtungstate (PbWO4) scintillating crystal electro-magnetic calorimeter, and a brass/scintillator had-ron calorimeter (HCAL). Muons are identifiedand measured in gas-ionization detectors em-bedded in the outer steel magnetic-flux-returnyoke. The detector is subdivided into a cylindricalbarrel part and endcap disks on each side of the

*To whom correspondence should be addressed. E-mail:cms-spokesperson@cern.chThe complete list of authors and affiliations is availableas supplementary material on Science Online.

ARTICLE

www.sciencemag.org SCIENCE VOL 338 21 DECEMBER 2012 1569

ARTICLES

on

Dece

mbe

r 8, 2

014

www.

scien

cem

ag.o

rgDo

wnloa

ded

from

o

n De

cem

ber 8

, 201

4ww

w.sc

ience

mag

.org

Down

loade

d fro

m

on

Dece

mbe

r 8, 2

014

www.

scien

cem

ag.o

rgDo

wnloa

ded

from

o

n De

cem

ber 8

, 201

4ww

w.sc

ience

mag

.org

Down

loade

d fro

m

on

Dece

mbe

r 8, 2

014

www.

scien

cem

ag.o

rgDo

wnloa

ded

from

o

n De

cem

ber 8

, 201

4ww

w.sc

ience

mag

.org

Down

loade

d fro

m

on

Dece

mbe

r 8, 2

014

www.

scien

cem

ag.o

rgDo

wnloa

ded

from

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushikairtight,waterproof

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

interaction point. Forward calorimeters comple-ment the coverage provided by the barrel andendcap detectors. The CMS detector has a largeangular acceptance, detecting particles over thefull azimuthal range and with q larger than 0.8,where q is the polar angle relative to the beamaxis. Figure 1 shows the CMS detector and itsmain components.

The 66 million silicon pixels and 9.3 millionsilicon strips forming the tracker are used todetermine the trajectories of charged particles.The multilayer silicon detectors provide accu-rate tracking of charged particles with excellentefficiency, which is especially important for thehigh-pileup conditions at the LHC. The magneticfield curves the trajectories of charged particles,

allowing the measurement of their momenta. Thetrack-finding efficiency is more than 99%, andthe uncertainty in the measurement of transversemomentum, pT (projection of the momentumvector onto the plane perpendicular to the beamaxis), is between 1.5 and 3% for charged tracksof pT ~ 100 GeV. By extrapolating tracks backtoward their origins, the precise proton-protoninteraction points, or collision vertices, can bedetermined. Decay vertices of long-lived parti-cles containing heavy-quark flavors, such as Bmesons, can similarly be identified and recon-structed. Such b-tagging is particularly usefulin searches for previously unobserved particles,such as the Higgs boson.

The electromagnetic calorimeter (ECAL)absorbs photons and electrons. These produceshowers of particles in the dense crystal material,which yield scintillation light detected by photo-detectors glued to the rear faces of the 75,848crystals. The amount of light detected is propor-tional to the energy of the incoming electron orphoton, allowing their energies to be determinedwith a precision of about 1% in the region ofinterest for the analyses reported here. Becauseelectrons are charged particles, they can be dis-criminated from photons by matching the ECALsignal with a track reconstructed in the tracker.

Hadrons can also initiate showers in theECAL, but they generally penetrate further into

SUPERCONDUCTING SOLENOIDNiobium titanium coil carrying ~18,000A

PRESHOWERSilicon strips ~16m2 ~137,000 channels

SILICON TRACKERSPixel (100x150 m) ~16m2 ~66M channelsMicrostrips (80x180 m) ~200m2 ~9.6M channels

MUON CHAMBERSBarrel: 250 Drift Tube, 480 Resistive Plate ChambersEndcaps: 468 Cathode Strip, 432 Resistive Plate Chambers

FORWARD CALORIMETERSteel + Quartz fibers ~2,000 Channels

STEEL RETURN YOKE12,500 tonnes

HADRON CALORIMETER (HCAL)Brass + Plastic scintillator ~7,000 channels

CRYSTAL ELECTROMAGNETICCALORIMETER (ECAL)~76,000 scintillating PbWO4 crystals

Total weightOverall diameterOverall lengthMagnetic field

: 14,000 tonnes: 15.0 m: 28.7 m: 3.8 T

CMS DETECTOR

Fig. 1. Schematic view of the CMS detector showing its main components.

Fig. 2. Diphoton (gg) invariantmass distribution for the 7- and 8-TeV data collected by CMS in2011and2012, respectively (blackpoints with error bars). The dataare weighted by the ratio of thesignal to signal plus backgroundfor each event class. The solid redline shows the fit result for signal-plus-background; the dashed redline with color bands shows onlythe background with its uncertain-ties at 1s (yellow) or 2s (cyan).(Inset) The central part of the un-weighted invariant mass distri-bution. Integrated luminosity was5.1 fb1 in 2011 and 5.3 fb1 in2012.

110 120 130 140 150

S/ (S

+ B)

Weig

hted E

vents

/ 1.5

GeV

0

500

1000

1500

DataSignal + Background FitBackground Fit Component12

(GeV)m120 130

Even

ts /

1.5

GeV

1000

1500

Unweighted

(GeV)m

21 DECEMBER 2012 VOL 338 SCIENCE www.sciencemag.org1570

The Higgs Boson

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

the detector, reaching the HCAL surroundingthe ECAL. The measurements of particle en-ergies in the HCAL are not as precise as thoseof the ECAL but are well adapted to the needsof the CMS physics program.

The solenoid is surrounded by a large detectorsystem that identifies and measures momenta ofmuons. It comprises three different types of gas-ionization detectors that enable muon momentato be measured with a precision of less than 1%in the region of interest relevant for the searchpresented here.

The combination of information from alldetectors is used to reconstruct the particle con-tent in a collision event through an algorithmknown as particle flow. The quarks and gluons,created in a hard collision of the constituents ofthe protons, combine and form jets of collimatedhadrons in the detector. Once reconstructed fromdata, the jet energy is calibrated to provide anaccurate measurement of the energy of the un-derlying quark or gluon. A vector sum of themomenta of all visible particles is computed,and the missing transverse momentum deducedfrom momentum conservation leads to the in-ference of the presence of undetected particles,such as neutrinos.

Although the LHC typically produces closeto half a billion collisions in roughly 20 million

bunch crossings per second, only a tiny fractionof these contain potentially interesting new phe-nomena, so it is neither necessary nor feasibleto record all of the data from every single colli-sion. CMS uses a two-level online trigger sys-tem to reduce the event rate from about 20 MHzto about 500 Hz, keeping only those eventsthat are worthy of further investigation. The firstlevel uses custom electronics close to the detec-tor to analyze coarse information from the cal-orimeters and muon detectors to reduce the rateto 100 kHz or less. The second level uses a com-puting farm of 13,000 processor cores to ana-lyze the full information from all subdetectorsin order to make the final decision on whetherto record an event. CMS has thus far selectedseveral billion events, corresponding to more than4 petabytes of stored event data. The recordedevents are sent to computing centers at CERNand around the world to fully reconstruct theparticles produced in each collision and allowsubsequent analyses.

Searching for the SM Higgs boson. At theLHC, the SM Higgs boson should be producedmost efficiently through gluon-gluon fusion:Gluons from each of the colliding protons fusetogether to form a Higgs boson. Two additionalimportant production processes are vector bosonfusion (VBF), where quarks inside the protons

emit W or Z bosons that fuse to form the Higgsboson, and associated production (VH), wherea vector boson V (either a W or Z) is producedtogether with the Higgs boson. The interactingquarks in the VBF events also give rise to high-energy jets produced at small angles that can bedetected and used to help identify this event type.Both VBF and VH events have better signal-to-background ratios relative to gluon fusion butoccur far less frequently (2428).

For every inverse femtobarn (fb1 = 1039

cm2) of integrated luminosity at the LHC, about20,000 SM Higgs bosons are expected to beproduced if the Higgs mass is close to 125 GeV.The majority of these decay to final states thathave large backgrounds, making identificationdifficult or impossible. Dedicated methods havebeen developed to exploit channels with lowerdecay fractions by selecting certain kinematicalregions of the decay products where the signal-to-background ratio is sufficiently large to makethe observation of SM Higgs bosons possible.Extensive use is made of particle-isolation crite-ria to reject the high-rate jet background, because,in general, the particles from Higgs decays ap-pear relatively isolated from each other and otherparticles in the detector.

Billions of detailed simulated events havebeen generated to develop and refine the analysis

Fig. 3. Event recorded with the CMS detector in 2012 at a proton-proton center-of-mass energy of 8 TeV. The event shows characteristics expected from the decayof the SMHiggs boson to a pair of Z bosons, one of which subsequently decays to a

pair of electrons (green lines in the tracker matched to green towers in the ECALin the central region of the detector) and the other decays to a pair of muons(red lines). The event could also be due to known SM background processes.

www.sciencemag.org SCIENCE VOL 338 21 DECEMBER 2012 1571

ARTICLES

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

techniques needed to estimate the SM Higgsboson signals and backgrounds (2932). Sam-ples of simulated events reconstructed with thesame software as used for the LHC data allow,for example, the estimation of background yieldsor the prediction of the expected significancefor the observation of new particles. However,in the presented analyses the background estima-tions are derived mostly from the control samplesin data.

We studied five SM Higgs decay modes:Hgg, HZZ, HWW, Htt, and Hbb.The gg, ZZ, and WW channels are of compa-rable sensitivity in the search for a Higgs bosonwith a mass around 125 GeV and are more sen-sitive than the bb and tt channels. Both the ggand ZZ channels provide precise mass measure-ments of the parent particle. The presence of anSM Higgs boson decaying to these final stateswould appear as relatively narrow peaks in theinvariant mass spectra of gg and ZZ.

An integrated luminosity of 5.1 fb1 was col-lected by CMS in 2011 at 7 TeV, allowing a firstthorough investigation into the existence, or non-existence, of the SM Higgs boson over a widemass range. This led to CMSs first significantexclusion of the SMHiggs boson in the medium-and high-mass region between 127 and 600 GeV(3338). Data from the ATLAS experiment ex-cluded a similar region (39). This left a smallwindowwhere a low-mass SMHiggs boson couldstill exist.

In the low-mass region below 127 GeV, the2011 data analyses also showed an excess overthe background-only expectation in the vicinityof 124 GeV. ATLAS observed a similar excessat around the same mass value (39). The ob-served excess in CMS was inconclusive, beingaround three standard deviations (3s) above thebackground-only expectation. After taking intoaccount the possibility that a signal-like excesscould appear randomly in the data between 110and 145 GeV (the look-elsewhere effect), thissignificance was reduced to about 2s. Therefore,there was still a nonnegligible chance that thisexcess could be due to a random upward fluc-tuation in the background, making it look likea signal. More data were needed to establishwhether this excess was genuine or not. It waspredicted that, in the case of a Higgs boson sig-nal, around 10 fb1 more data would be requiredto reach a statistical significance of around 5s.However, the LHC operation at 8 TeV in 2012(giving a 20% increase in Higgs boson produc-tion cross section compared to 7 TeV), coupledwith improved analyses with 20 to 30% highersensitivity, reduced this additional required lumi-nosity to around 5 fb1. By the summer of 2012,CMS had collected an additional 5.3 fb1 ofcollision data at this new energy.

Because the 2011 analysis (3338) showedan excess of events at about 125 GeV and toavoid a potential bias in the choice of selection

criteria for the 2012 data that might artificiallyenhance this excess, we performed the analysis ofthe 2012 data blind: The region where the sig-nal may be present was not examined until afterall the analysis criteria had been fully scrutinizedand agreed upon within the collaboration.

Search for the SM Higgs boson decay intotwo photons. The predicted probability for a125-GeV SM Higgs boson to decay into twophotons is about 0.3%. Yet this decay mode isone of the most important, because both photonscan be measured very accurately in the CMSECAL and the backgrounds can be precisely es-timated. The presence of a signal would manifestitself as a narrow peak above a smoothly fallingbackground in the invariant mass distributionof the two photons.

The energy resolution and precise knowledgeof the absolute energy scale of the ECAL are keyelements of this analysis. These were achievedby calibrating each channel of the ECAL in situ,using diphoton decays of p0 and h0, for example.The stability of the ECAL response was ensuredby the use of a sophisticated real-timemonitoringprocedure that corrects any deviations with aprecision of a few per mill. Decays of Z bosonsinto electron pairs were then used to determinethe energy resolution and energy scale, takingadvantage of the precise knowledge of the Zmass and width.

An additional challenge is to determine fromwhich of the many collision vertices in the eventthe two photons originate, which affects the pre-cision of the mass measurement of the parentparticle. The collisions occurring in a single LHCbunch crossing, as many as 40 in 2012, are spreadover a distance of about 10 cm along the beamaxis at the center of CMS. Because photons donot leave tracks in the detector, there can be am-biguity as to which collision vertex they belongto. A variety of techniques were used to deter-

mine the diphoton vertex, including use of eventkinematics and an understanding of photon con-versions into electron pairs.

Multivariate analyses (40,41) based on boosteddecision trees were used to identify the photonsand to extract their energies and uncertaintieson a photon-by-photon basis. To optimize sensi-tivity, we categorized diphoton events into fourclasses with decreasing restrictions on the qual-ity according to many variables, including the un-certainty on the diphoton mass measurement,the kinematics of the photons, and whether thephotons convert into electron pairs in the materialbefore reaching the ECAL. For example, eventswhere both photons are in the central region ofCMS and do not convert into electron pairs in thetracker were given the highest classification be-cause they are the most precisely measured. Wealso included additional classes for diphotonevents that have two additional jets with proper-ties consistent with those expected for the VBFproduction process.

Photons of high quality (determined from thespatial distribution of electromagnetic showersand isolation criteria) were selected with en-ergies above 30 to 40 GeV, depending on theevent class. Figure 2 shows the diphoton in-variant mass spectrum from all the data collectedby CMS from 2011 to mid-2012, after selectionsas defined in (20). The spectrum was built upfrom the event classes, with each class weightedby the ratio of the signal to signal-plus-backgroundestimated from simulation.An excesswas observedat 125 GeVon an otherwise smoothly falling back-ground spectrum. The background consists mostlyof collisions where two photons are produced inSM processes and a smaller fraction from eventswhere at least one of the photon signals is notgenuine but originates from the debris of jets.

The observed excess is consistent in shapeand size with that expected for diphoton decays

Fig. 4. Distribution of the four-lepton reconstructed invariant massfor the sum of the 4e, 4m, and 2e2mchannels. Points represent the data,shaded histograms represent thebackground, and the empty histo-gram the signal expectations. Thedistributions are presented asstacked histograms. The measure-ments are presented for the sumof the data collected at center-of-mass energies of 7 and 8 TeV dur-ing 2011 and 2012, respectively.Error bars represent standard de-viations. (Inset) The four-lepton in-variantmass distribution after selectionof events with signal-like kinemat-ics, as described in the text.

80 100 120 140 160 180

Even

ts /

3 G

eV

0

2

4

6

8

10

12

14

16 DataZ+X

*, ZZZ=125 GeVHm

(GeV)m120 140 160

Even

ts /

3 G

eV

0123456

(GeV)m

21 DECEMBER 2012 VOL 338 SCIENCE www.sciencemag.org1572

The Higgs Boson

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

Vishnu Kaushik

of SM Higgs bosons. To evaluate the signifi-cance of the signal, we fitted the backgroundspectrum over the whole mass range with a fifth-order polynomial function (or a third-order poly-nomial for the VBF categories) and measuredthe magnitude of the excess above the back-ground. The diphoton decay mode has a signalsignificance of 4.1s relative to the background-only hypothesis. This excess is present in both2011 and 2012 data and is consistent between thetwo data sets.

The observation of the diphoton final statealso implies that the new particle is a boson andhas an integer spin different from unity (42, 43).

Search for the SM Higgs boson decay intotwo Z bosons. If the SM Higgs boson has a massof 125 GeV, about 2.6% of them are predictedto decay into two Z bosons. At least one of theZ bosons is necessarily virtual, that is, it has adifferent mass than the 91 GeV Z mass. The Zbosons each decay into pairs of leptons or quarks.We concentrate on the Z decays into leptons, par-ticularly electrons (e) and/or muons (m), becausethese have the smallest SMbackgrounds. In CMS,we analyzed separately the three different finalstates in this channel, namely 2e2m, 4e, and 4m,and then combined the results.

The invariant mass of the ZZ system can bereconstructed and measured with good accuracyin CMS from the four-lepton momenta. Hence,the presence of a Higgs boson in the data shouldmanifest itself as a peak in the ZZ invariant massspectrum in the presence of a small continuumbackground.

There are numerous SM processes (not in-cluding Higgs boson decays) that can lead to thesame final states. They include direct ZZ pro-duction from quark-antiquark annihilation andgluon-gluon fusion, as well as processes involvinga single Z boson produced with associated heavy-quark jets and top-antitop pair-production. Apart

from the rate of direct ZZ production, which wecan determine accurately from simulation, the ratesof other backgrounds were extracted from data.

The leptons fromZ decays are, in general, wellisolated in the detector; that is, their trajectoriesare far from the debris of jets or other particlesproduced in the collision. Despite the large par-ticle multiplicity per event from pileup interac-tions, the overall efficiency for selecting isolatedleptons remains very high.

We selected collisions with four isolated lep-tons originating from the same vertex, for whichthe transverse momentum of each muon is atleast 5 GeVand of each electron is at least 7 GeV.(These criteria were determined by using a largesample of single-Z events collected in the past2 years.) Both Z boson candidates are requiredto decay to two same-flavor leptons of oppositecharge, and the invariant mass of the dileptonsproduced in the Z boson decays must be in therange from 40 to 120 GeV for the heavier of thetwo and 12 to 120 GeV for the lighter one.

Figure 3 shows a typical event containing tworeconstructed Z bosons, with a ZZ invariant massaround 125 GeV. The ZZ invariant mass spec-trum for selected events is shown in Fig. 4. Be-cause leptons (especially electrons) can oftenradiate an energetic photon at an early stage oftheir trajectory through the detector, the energyor momentum of such leptons can be consider-ably underestimated. We therefore searched forenergetic photons close to these leptons andadded their energies when appropriate.

The invariant mass spectrum in Fig. 4 showsa Z peak at 91 GeV resulting from decays ofZ bosons into two leptons and an energetic vir-tual photon that materializes through a seconddilepton pair. There is also a statistically signif-icant peak near 125 GeV. This completely inde-pendent analysis indicates the presence of a signalin the same region as that found in the diphoton

decay mode. This is to be expected if indeed thesignals correspond to the same parent particle.

The signal-to-background separation im-proves further by exploiting the decay kine-matics expected for signal events, especially thedecay angles and invariant masses of the twopairs of leptons (44). Analyzing events in thepeak at 125 GeV confirmed that many of theseevents have the requisite characteristics; this re-inforced our interpretation that the signal is gen-uine. The statistical significance of the excessobserved by combining data from 2011 and2012, accounting also for the decay-angle charac-teristics, is 3.2s relative to the background-onlyhypothesis. The maximum significance occurs ata mass of 125.6 GeV.

Search for Higgs boson decays in other chan-nels. Apart from the gg and ZZ channels dis-cussed above, CMS also searched for decays ofSM Higgs bosons to two W bosons, two t lep-tons, or two b quarks.

For the WW decay mode, the final statesmust contain two opposite-sign leptons (eitherelectrons or muons) and significant missing trans-verse momentum, resulting from the undetectedneutrinos from W decays. In contrast to the ggand ZZ modes, the invariant mass of the twoWbosons cannot be precisely reconstructed. Thepotential excess in the data over the backgroundexpectation provides only a continuum insteadof a sharp resonance peak. We used multivariateanalysis techniques to optimize the sensitivityto a possible signal present in data. We classifiedthe events into a number of exclusive categories,for example, according to lepton flavor contentand whether there are jets present (to enhanceVBF production relative to gluon fusion).

These different event classes were subject todifferent backgrounds and have different sen-sitivities. The challenge for this analysis was toestimate the backgrounds from the SM, whichwas generally achieved through techniques basedon control regions in the data and complementedthrough simulation. Special attention was paid todetermining the missing transverse momentum,particularly in the presence of large pileup (as inthe 2012 data sample).

We selected events in which the pT of themost energetic lepton is greater than 20 GeVandthat of the second-most-energetic lepton is above10 GeV and that have missing transverse mo-mentum typically above 20 GeV. The results ofthis analysis, combining all the classes across the2011 and 2012 data, show (Fig. 5) a broad excessof events over the expected background, con-sistent with the presence of a new particle at amass near 125 GeV. The statistical significance isabout 1.5 to 2.0s relative to the background-onlyhypothesis.

We also explored whether this new particledecays into fermion pairs, as it would be ex-pected to if the associated field gives mass to thefermions in addition to the W and Z bosons, by

Fig. 5. Distribution of the invariantmass of lepton pairs for the zero-jetem category in the search at 8 TeV forthe SM Higgs boson decay to a pair ofW bosons. The signal expected fromthe SM Higgs boson with a massmH =125 GeV is shown added to the back-ground. Error bars indicate standarddeviations.

(GeV)m0 50 100 150 200

Entri

es/ 5

GeV

0

10

20

30

40 data

= 125 GeVH m WW VV top Z+jets W+jets

www.sciencemag.org SCIENCE VOL 338 21 DECEMBER 2012 1573

ARTICLES

looking for instances where the particle decaysinto heavy fermions. The heaviest fermions intowhich a 125-GeVSMHiggs boson can decay arethe t leptons and b quarks.

The detection of t leptons is challenging be-cause they are unstable and decay less than 1012

s after production, either into a lighter chargedlepton (electron or muon) and neutrinos or intoa neutrino and either one or three charged pionspossibly accompanied by neutral pions. CMShas tools to detect and reconstruct such decaysand separate these from backgrounds. Severaldecay channels were explored, including thecombination of one t lepton decaying exclu-sively into leptons and the other into hadrons.A natural process to calibrate the analysis isthrough Z boson production, where 3.4% of theZ decays are into t+t pairs, and CMS has suc-cessfully used this decay channel (45).

The main challenge for this search is to assessthe backgrounds, most of which are extractedusing control samples in data. Different t decaychannels are analyzed separately and classifiedaccordingly, including classes with accompa-nying jets. The results of these individual an-alyses were then combined for a final result.

As in the case of the WW decay mode, thepresence of neutrinos in the decay products ofthe t leptons prevents a full event reconstruc-tion, and, instead of a resonance peak, a broadenhancement over background is expected. Wehave not yet found such an enhancement, butthe current sensitivity to this channel does notexclude the presence of the SM Higgs boson.With the LHC on course to triple the integratedluminosity by the end of 2012, studies of the ttchannel will become more sensitive.

Lastly, CMS conducted a search for SMHiggs bosons decaying into two b quarks. Eachquark gives rise to a jet that is recognized in theanalysis (tagged) as originating from b quarks.For the tagging of b quarks, we searched forsecondary vertices in the jets, caused by decaysof B hadrons that travel a few millimeters beforedecaying. The energy of the original b quark isestimated from the energies of all the particles inits jet and has a large uncertainty. The reconstructedmasses of objects obtained from these jets aretherefore expected to be distributed over a regionof about 20 GeV in the mass range of interest.

At low mass (below about 135 GeV), the SMHiggs boson decay into b quarks has the largestrate of the five search modes we report in thispaper, and we therefore expect a large numberof such decays in the data. This signal is, how-ever, overwhelmed by a large background fromSM b quark production, making the search lesssensitive. To have a more favorable signal-to-background ratio, we searched for this signal inthe (rarer) associated production process involv-ing a Wor Z boson, which can be detected fromtheir leptonic decays. We required these bosonsto have transverse momenta above 50 GeV.

Tominimize the background in the bb channel,we again used severalmutually exclusive classes ofevents, which were analyzed separately. Theseclasses are based on the transversemomentum ofthe jet pair and the nature and decay of the asso-ciated boson. For the final result, we combined allthese channels and used all of the available datafrom 2011 and 2012. The result shows a smallexcess above the background-only expectationover a large mass range, including the regionnear 125 GeV. The sensitivity of this analysis isabout 1.5 times lower than required for conclud-ing whether a signal is present (as expectedfrom SM prediction) or if the coupling to b quarksis different from what we would expect. Again,tripling the amount of collision data should bedecisive.

In conclusion, neither fermion decay modeshows, at present, a statistically significant en-hancement over the background-only expectation.

Nevertheless, at the present level of sensitivitythe results in these channels are consistent withthe production of the SM Higgs boson, in agree-ment with observations in the other three (diboson)decay modes.

Observation of a new particle.The final resultcombines all the information collected through aglobal fit (46) to the five different search channels.The result reflects the probability for the back-ground to deviate from the expectation by at leastthe observed amount, assuming the absence of theSM Higgs boson in this mass range. This proba-bility, known as the local P value, is evaluatedby using sets of simulated data that incorporateall experimental uncertainties and correlationsamong analyses. The result is shown (Fig. 6) foreach of the five search channels individually, aswell as for the combination of all five channels.For the combination the minimum P value at125 GeV is of the order of one in three million.

(GeV)Hm110 115 120 125 130 135 140 145

Loca

l p va

lue

-1210

-1010

-810

-610

-410

-210

112

3

4

5

6

7

Combined obs.Exp. for SM HHHZZHWWHHbb

Fig. 6. The observed probability (local P value) that the background-only hypothesis would yield asmany (or more) events as seen in the CMS data, as a function of the mass of the SM Higgs bosonfor each of the five search channels individually and the combination of all five channels. Thedashed line shows the median expected local P values for the SM Higgs boson with a mass mH.

(GeV)Xm123 124 125 126 127 128 129

SM /

0

1

2

3

4

5

6 Combined (untagged)H (VBF tag)H

ZZH

ZZ + H H

SMBest fit /-1 0 1 2 3

bbH

H

WWH

ZZH

H

= 125.5 GeVH m

Fig. 7. (Left) 68% CL contours for the relative signal strength s/sSMH versus the Higgs boson massfor the high-precision decay modes (gg and ZZ). (Right) The relative signal strength s/sSMH for thefive Higgs boson decay modes examined by CMS. The symbol s/sSM denotes the production crosssection multiplied by the relevant branching fraction, relative to the SM expectation.

21 DECEMBER 2012 VOL 338 SCIENCE www.sciencemag.org1574

The Higgs Boson

This probability corresponds to a local signifi-cance of 5s. The probability of observing thislarge a fluctuation anywhere in the mass rangeof 114 to 130 GeV, where the Higgs boson hadnot been excluded by previous data, is small andresults in a global significance of 4.6s. The glob-al significance is smaller than the local value be-cause of the look-elsewhere effect. Both measuresconvincingly show that this is not a backgroundfluctuation, but rather the observation of a newparticle. The expected sensitivity with the presentdata for a 125 GeV SM Higgs boson amounts toa local significance of 5.8 1.0s, consistent withthe signal observed at 5s.

In addition to being able to say with highconfidence that a new particle has been ob-served, and that it is a boson with spin not equalto one, we were also able to derive some of itsproperties, such as its mass. And, as mentionedabove, once the mass is known the SM allowsus to calculate many other properties, such asthe fractions of Higgs bosons decaying in differ-ent ways, and compare these expectations withour measurements. This is expressed as the sig-nal strength, that is, the measured productionrate of the signal, which can be determined foreach decay mode individually and for the over-all combination of all channels, normalized tothe predicted Higgs boson production rate. Thesignal strength was defined to be equal to onefor the SM Higgs boson. The measured signalstrength was highest in the diphoton channel,namely 1.6 0.4, whereas that in the ZZ chan-nel was 0:70:40:3. By using the high-resolution di-photon and ZZ channels discussed above, whichshow a resonance peak, we obtained the 68%confidence level (CL) contours for the signalstrength versus the boson mass (Fig. 7 left). Wealso show the combination of the diphoton andZZ decaymodes,where the relative signal strengthsof these two modes are constrained by the ex-pectations for the SM Higgs boson. To extractthe value of the mass in a model-independentway, we allowed the signal yields of the com-bined channels to vary independently. The com-bined best-fit mass is 125.3 0.4 (statistical) 0.5 (systematic) GeV.

The signal strengths for all five channels aredepicted in Fig. 7 (right). The overall combinedsignal strength, including all channels, is 0.87 0.23. Hence, these results are consistent, withinrelatively large statistical and systematic uncer-tainties, with the expectations for the SM Higgsboson.

The CMS data also rule out the existence of theSMHiggs boson in the ranges of 114.4 to 121.5GeVand 128 to 600GeVat 95%CL (20). Lowermasseswere already excluded by CERNs Large Elec-tron Positron collider at the same CL (14).

More data are needed to establish whetherthis new particle has all the properties of theSM Higgs boson or whether some do not match.The latter may imply new physics beyond the

SM. This particle has the potential to be a portalto a new landscape of physical phenomena thatis still hidden from us. The CMS experiment isin an excellent position to undertake this researchin the years to come.

References and Notes1. S. L. Glashow, Nucl. Phys. 22, 579 (1961).2. S. Weinberg, Phys. Rev. Lett. 19, 1264 (1967).3. A. Salam, in Elementary Particle Physics: Relativistic Groups

and Analyticity, N. Svartholm, Ed. (Nobel Symposium 8,Almqvist and Wiskell, Stockholm, 1968), pp. 367377.

4. F. Englert, R. Brout, Phys. Rev. Lett. 13, 321 (1964).5. P. W. Higgs, Phys. Lett. 12, 132 (1964).6. P. W. Higgs, Phys. Rev. Lett. 13, 508 (1964).7. G. S. Guralnik, C. R. Hagen, T. W. B. Kibble, Phys. Rev.

Lett. 13, 585 (1964).8. P. W. Higgs, Phys. Rev. 145, 1156 (1966).9. T. W. B. Kibble, Phys. Rev. 155, 1554 (1967).

10. J. M. Cornwall, D. N. Levin, G. Tiktopoulos, Phys. Rev.Lett. 30, 1268 (1973).

11. J. M. Cornwall, D. N. Levin, G. Tiktopoulos, Phys. Rev. D10, 1145 (1974).

12. C. H. Llewellyn Smith, Phys. Lett. B 46, 233 (1973).13. B. W. Lee, C. Quigg, H. B. Thacker, Phys. Rev. D Part.

Fields 16, 1519 (1977).14. ALEPH, DELPHI, L3, OPAL Collaborations, and LEP

Working Group for Higgs Boson Searches, Phys. Lett. B565, 61 (2003).

15. T. Aaltonen et al.; CDF Collaboration; D0 Collaboration,Phys. Rev. Lett. 104, 061802 (2010).

16. CDF Collaboration, Combined search for the standardmodel Higgs boson decaying to a bb pair using the full CDFdata set (2012), http://arxiv.org/abs/1207.1707.

17. T. Aaltonen et al.; CDF Collaboration; D0 Collaboration,Phys. Rev. Lett. 109, 071804 (2012).

18. D0 Collaboration, Combined search for the standard modelHiggs boson decaying to bb using the D0 Run II data set(2012), http://arxiv.org/abs/1207.6631.

19. L. Evans, P. Bryant, J. Instrum. 3, S08001 (2008).20. CMS Collaboration, Phys. Lett. B 716, 30 (2012).21. ATLAS Collaboration, Phys. Lett. B 716, 1 (2012).22. ATLAS Collaboration, Science 338, 1576 (2012).23. CMS Collaboration et al., J. Instrum. 3, S08004 (2008).24. J. R. Ellis, M. K. Gaillard, D. V. Nanopoulos, Nucl. Phys. B

106, 292 (1976).25. H. M. Georgi, S. L. Glashow, M. E. Machacek,

D. V. Nanopoulos, Phys. Rev. Lett. 40, 692 (1978).26. S. L. Glashow, D. V. Nanopoulos, A. Yildiz, Phys. Rev. D

18, 1724 (1978).27. S. Alioli, P. Nason, C. Oleari, E. Re, J. High Energy Phys.

0904, 002 (2009).28. P. Nason, C. Oleari, J. High Energy Phys. 1002, 37 (2010).29. T. Sjstrand, S. Mrenna, P. Z. Skands, J. High Energy Phys.

0605, 026 (2006).30. S. Gieseke et al., Herwig++ 2.0 Release Note (2006),

http://arxiv.org/abs/hep-ph/0609306.31. J. Alwall et al., J. High Energy Phys. 0709, 028 (2007).32. S. Agostinelli et al., Nucl. Instrum. Meth. A 506, 250 (2003).33. CMS Collaboration, Phys. Lett. B 710, 26 (2012).34. CMS Collaboration, Phys. Lett. B 710, 403 (2012).35. CMS Collaboration, Phys. Rev. Lett. 108, 111804 (2012).36. CMS Collaboration, Phys. Lett. B 710, 91 (2012).37. CMS Collaboration, Phys. Lett. B 713, 68 (2012).38. CMS Collaboration, Phys. Lett. B 710, 284 (2012).39. ATLAS Collaboration, Phys. Rev. D 86, 032003 (2012).40. H. B. Prosper, paper presented at XII International

Workshop on Advanced Computing and AnalysisTechniques in Physics Research (ACAT08), 3 to7 November 2008, Erice, Italy, no. PoS(ACAT08)010.

41. P. C. Bhat, Annu. Rev. Nucl. Part. Sci. 61, 281 (2011).42. L. D. Landau, Dokl. Akad. Nauk 60, 207 (1948).43. C. N. Yang, Phys. Rev. 77, 242 (1950).44. CMS Collaboration, J. High Energy Phys. 4, 36 (2012).45. CMS Collaboration, J. High Energy Phys. 08, 117 (2011).46. ATLAS and CMS Collaborations, technical report

ATL-PHYS-PUB 2011-11, CMS NOTE 2011/005 (2011),http://cdsweb.cern.ch/record/1379837.

Acknowledgments: We congratulate our colleagues in theCERN accelerator departments for the excellent performanceof the LHC machine. We thank the computing centers inthe Worldwide LHC Computing Grid for the provisioning andexcellent performance of computing infrastructure essential toour analyses and the administrative staff at CERN and the otherCMS institutes. We gratefully acknowledge the contributions ofthe technical staff at CERN and other CMS institutes and thesupport from all the funding agencies that contributed to theconstruction and the operation of the CMS detector: the AustrianFederal Ministry of Science and Research; the Belgian Fondsde la Recherche Scientifique, and Fonds voor WetenschappelijkOnderzoek; the Brazilian funding agencies [Conselho Nacional deDesenvolvimento Cientifico e Technologico(CNPq), Coordenaode Aperfeioamento de Pessoal de Nvel Superior (CAPES),Fundao de Amparo Pesquisa do Estado do Rio de Janeiro(FAPERJ), and Fundao de Amparo Pesquisa do Estado doSao Paulo (FAPESP)]; the Bulgarian Ministry of Education, Youth,and Science; CERN; the Chinese Academy of Sciences, Ministryof Science and Technology, and National Natural ScienceFoundation of China; the Colombian funding agency(COLCIENCIAS, Departamento Administrativo de Ciencia,Tecnologa, e Innovacin); the Croatian Ministry of Science,Education, and Sport; the Research Promotion Foundation, Cyprus;the Ministry of Education and Research, Recurrent financingcontract SF0690030s09 and European Regional DevelopmentFund, Estonia; the Academy of Finland, Finnish Ministry ofEducation and Culture, and Helsinki Institute of Physics; theInstitut National de Physique Nuclaire et de Physique desParticulesCNRS, and Commissariat lnergie Atomique et auxnergies AlternativesCEA, France; the Bundesministerium frBildung und Forschung, Deutsche Forschungsgemeinschaft,and Helmholtz-Gemeinschaft Deutscher Forschungszentren,Germany; the General Secretariat for Research and Technology,Greece; the National Scientific Research Foundation, andNational Office for Research and Technology, Hungary; theDepartment of Atomic Energy and the Department of Scienceand Technology, India; the Institute for Studies in TheoreticalPhysics and Mathematics, Iran; the Science Foundation, Ireland;the Istituto Nazionale di Fisica Nucleare, Italy; the KoreanMinistry of Education, Science and Technology and the WorldClass University program of NRF, Republic of Korea; theLithuanian Academy of Sciences; the Mexican funding agencies[Centro de Investigacin y Estudios Avanzados, (CINVESTAV),Consejo Nacional de Ciencia y Technologa (CONACYT),Secretara de Educacin Pblica (SEP), and UniversidadAutnoma de San Luis Potos Fondo de Apoyo a la Investigacin(UASLP-FAI)]; the Ministry of Science and Innovation,New Zealand; the Pakistan Atomic Energy Commission; theMinistry of Science and Higher Education and the NationalScience Centre, Poland; the Fundao para a Cincia e aTecnologia, Portugal; JINR ( Joint Institute for Nuclear Research)(Armenia, Belarus, Georgia, Ukraine, Uzbekistan); the Ministryof Education and Science of the Russian Federation, the FederalAgency of Atomic Energy of the Russian Federation, RussianAcademy of Sciences, and the Russian Foundation forBasic Research; the Ministry of Science and TechnologicalDevelopment of Serbia; the Secretara de Estado de Investigacin,Desarrollo, e Innovacin and Programa Consolider-Ingenio2010, Spain; the Swiss funding agencies [EidgenssischeTechnische Hochschule (ETH) Board, ETH Zrich, Paul ScherrerInstitut (PSI), Swiss National Science Foundation, UniversittZrich, Canton Zrich, and State Secretariat for Education andResearch (SER)]; the National Science Council, Taipei; theThailand Center of Excellence in Physics, the Institute for thePromotion of Teaching Science and Technology of Thailandand the National Science and Technology DevelopmentAgency of Thailand; the Scientific and Technical ResearchCouncil of Turkey, and Turkish Atomic Energy Authority; theScience and Technology Facilities Council, UK; U.S.Department of Energy, and NSF.

Supplementary Materialswww.sciencemag.org/cgi/content/full/338/6114/1569/DC1Complete Author List

10.1126/science.1230816

www.sciencemag.org SCIENCE VOL 338 21 DECEMBER 2012 1575

ARTICLES