Higgs discovery prospects p and statistics at the LHC (ATLAS) Eilam Gross Weizmann institute of Science/ATLAS

Transcription

1 Higgs discovery prospects p and statistics at the LHC (ATLAS) Eilam Gross Weizmann institute of Science/ATLAS 1

2 2 ATLAS with a Perspective

3 3 A Legend Comes True

4 4 A Legend Comes True

5 5 The team Israel and Pakistan working for the Atlas collaboration together for a family picture.

6 Higgs Generates All Current Masses L = g Φff = g ( H + v) f Hff Hff ff m g = f Hff g = m / Hff f H g Hff f f v v The heavier the particle, its coupling to Higgs is bigger 6

7 Constraints from EW precision measurements 114 < m < 166 H 7

8 The Technical Challenge ~10 9 proton-proton collisions/sec ~1/ /10 12 collisions produce a Higgs Boson One can only accommodate ~200 collisions/sec Rejection rate> % How do you make sure not to loose a Higgs Boson? Make sure we push the button (trigger) only when interesting things happen This is a tremendous technical challenge! Which require state of the art fast systems. σ tot ~ 100 mb, σ bb ~ 1 mb, σ jet > 1 nb, σ H > 1~pb The small S/B requires control samples to understand the background 8

9 9 Higgs LHC

10 Some Schematic Backgrounds QCD DrellYan Z + jets qg Zq Z / γ Z / γ gg Zqq Z / γ 10

11 σ * BR The Analysis Challenge The Higgs decay to bb is still dominant up to ~150 GeV, the Weak Bosons then enter the game However, H 4 leptons has a clear signature as well as H γγ with BRs of O(10-3 ). This is a challenge in analysis which makes the need for measuring background from data mandatory! The decay to ττ, though only a few % is still very appealing for the medium light SM Higgs For MSSM Higgs bb and ττ are dominant 11

12 LHC Brief Status As reported by Sergio Bertolucci, LP07 Engineering run originally foreseen at end 2007 now precluded by delays in installation and equipment commissioning. Beam commissioning starts May 2008 First collisions at 14 TeV c.m. July 2008 First month devoted to collecting millions of minimum bias events and di-jets events for first alignment, calibrations and performance studies of the detector ) 12

13 LHC Brief Status As reported by Sergio Bertolucci, LP07 Pilot run pushed to reach cm -2 s -1 by end 2008 This is equivalent to O(100 pb -1 )/experiment Use Z ee,μμ to calibrate, align and understand the trackers, the EM CAL, and the Muon system. Use W jj to calibrate jets and study b-tag performance (tt blν bjj) Rediscover SM Physics at s=14 TeV, this is your new background After all, the LHC is designed to go beyond the known SM Physics, first and foremost to discover the Higgs Boson. 13

14 Cosmic Ray Tests Tracks in the Muon chambers and in the Transition-Radiation Tracker TRT 14

15 And now. The main course HIGGS SEARCH TECHNIQUES AND PROSPECTS 15

16 16 The Official References

17 17 The Official References

18 m H ~120 GeV H γγ, classical bump hunting, BR ~ Irreducible background (now at NLO compared with TEVATRON) From the production point of view: The reference inclusive production: gg H In the last few years the VBF is stealing the show.. 100fb -1 Trigger on 2 isolated Photons with pt>20 or one very energetic photon 18

19 m H ~120 GeV 2 E 2 Ψ γγ γ m ~ 2 E (1 cos Ψ ) To get a 1% resolution in mass needs uniform resolution of EM calorimeter in η and a good angular resolution For high luminosity vertex Resolution in φ is obtained position must be taken from from EM calorimeter calorimetry ~16 mm Resolution in η is achieved with a fit constraining the photons to emerge from the primary vertex (zvertex known within a few 10s of mm), this will not work for high luminosity (10 34 ) 19

20 m H ~120 GeV Since σ γj+jj ~ 10 6 σ γγ one needs a rejection of at least 1000 against fake photons (jets) to maintain the level of the fake photons BG below the irreducible BG Current studies indicate σ γj+jj ~ 20% σ γγ maintaining an 80% q γ photon detection efficiency. γ g γ Fake photons are produced by leading q pions from fragmenting jets π 0 Need high performance of EM calorimeter to tell photons from Pions (e.g. use shower shapes based on detector t granularity) 20

21 Photon Conversions Photon conversions before the calorimeter ~30% conversions in tracker Single and Double conversions due to a lot of material in the Atlas Silicon trackers Performance of channel depends a lot on the percentage of restored conversions (construction of converted photons) 21

22 m H ~120 GeV 100fb -1 S/B~1:20. requires a full control of the background So how do we do it? 22

23 Side Bands Background Analysis Toy model with SIGNAL 23

24 Side Bands Background Analysis 24 Toy model with SIGNAL Only one way around. FITS! Fit the data Signal and BG altogether This is the only way to reduce the systematics ti Hard to believe but if we assume a 120 GeV Higgs, this toy MC delivers a median significance of 3.9σ (with no systematics but BG fitted with data, and taking shapes into account with the Profile likelihood method) Look elsewhere effect is crucial here

25 m H ~120 GeV Sensitivity can be increased by inclusive searches separating to H+0j H+1j H+2j (dominated by VBF, qq qqh) Resolution on mass ~1.6 GeV (@ m H =120) Excellent prospects, especially for L>10fb -1 25

26 The Golden Channel H ZZ* 4 leptons Very low BR of H ZZ* for low-medium m low-medium m H Yet, very clean and therefore appealing especially H ZZ* 4μ but also the 4e Trigger by requiring two high pt electrons or muons (~10,15) or one high pt electron or muon (~20,25) 26 Success of channel depends d on good e/μ identification, efficiency for isolated leptons and energy resolution For muons, eff close to 100%

27 Reducible Background Some collisions produce similar topologies to the signal Irreducible ZZ is an irreducible ibl background dominant after selection. XSC known to NLO. 27 Zbb or tt are reducible backgrounds (mainly by requiring isolated leptons [tracker and calorimeter] with a pair emerging from an on shell Z boson and anti-b IP based cuts)

28 Irreducible Background ZZ is an irreducible background dominant after selection. XSC known to NLO. Background shapes and normalization from DATA sidebands (reduce PDF and luminosity uncertainties) Clean channel (but low statistics) With 30 fb -1 a 150 GeV Higgs Boson can be observed with a significance of 7σ and a resolution of 1.8 GeV (Very Preliminary, just to give an idea) 28

29 The Golden Channel H 4 leptons Sensitivity best for ~160 GeV and heavier Higgs Boson 29

30 A Lesson in Systematic In absence of systematics significance can be approximated to be However if there is systematics, say, Δb the significance is reduced to s s s = 2 b(1 +Δ b) Δ b ( b ) + ( Δ b ) 2 2 s b For 5σ one needs s 5 b > Δ For 10% systematics this implies s 0.5 b > 30

31 A Lesson in Systematics, tth b b tth, H bb A multi-jets final state Looked promising until fast simulation was replaced by full simulation and systematics killed it (FOR THE TIME BEING) due to uncertainties in the shapes of the combinatorial 4j BG 31

32 An Essential Tool: Transverse Mass In hadron colliders the interaction is between 2 quarks somewhere along the beam.. A typical event will have a large imbalance in the Z direction but in principle should be balanced and therefore could be reconstructed (up to traceless particles like neutrinos) in the transverse plane, e.g. qq H+1jet H H j p Z miss j Transverse Plane 32

33 An Essential Tool: Transverse Mass This is why in many cases people use quantities which are well defined in the Transverse plane, first and foremost the transverse missing energy (missing E T ) or M T 2 W lν mt = 2 p E (1 cos ) T T Δϕ What is the source of the Transverse mass? In this example the source of the missing TRANSVERSE energy is (only) a neutrino with a zero mass z There are two unknown variables, pmiss, Emiss with one 2 constraint p = 0 miss 2 2 The mass M = ( p cannot be determined, but lep + pmiss ) using Lagrange multipliers we can find an extremum for the mass as a function of the measured known quantities ( m ) = ( P + P ) ( p + p ) 2 T T 2 T T 2 T miss lep lep miss m = 2 p E (1 cos Δϕ) 2 T T T

34 An Essential Tool: Transverse Mass Example (parton level) tt bqqb ν 34

35 m H ~160, Gauge Bosons Playground H WW νν H WW lνlν with l-l l spin correlation Dominant for m H ~2m W Signature of 2 isolated leptons + missing E T No mass peak, use transverse mass need to understand missing E T 35

36 m H ~160, Gauge Bosons Playground H WW νν H WW lνlν with l-l l spin correlation Dominant for m H ~2m W Signature of 2 isolated leptons + missing E T No mass peak, use transverse mass need to understand missing E T 36 ν ν W + e + W - e - Higgs Spin 0

37 m H ~160, Gauge Bosons Playground H WW νν Main background from Wt(b) and tt, rejected via jet veto, yet systematics is still >10% Signature of 2 isolated leptons with small acoplanarity+ missing E T VETO on hard jets 37

38 m H ~160, Gauge Bosons Playground H WW lνlν ATLAS M=160GeV 30fb Main interest is near M H ~160 GeV (BR H WW 95%) Sensitivity in the lower mass region can be extended looking at VBF production

39 VBF Central Jets Veto VBF qq qqh Han, Valencia, Willenbrock (1992); Figy, Oleari, Zeppenfeld (2003,2004) Forward jets Z / W H φ Z / W Higgs Decay η 39 W b t t b W

40 The Breakthrough: VBF qqh, H WW,ττ Signature: central jet veto (only signal decay products in the central region) twohighp T jets with large Δη separation Success of channel relies on understanding of the Missing Energy Note also the massive use of central jet veto necessiates its understanding! 40 φ Higgs Decay Forward jets η Han, Valencia, Willenbrock (1992); Figy, Oleari, Zeppenfeld (2003,2004) Jet Jet Full line Parton level, data is the reconstructed

41 VBF H WW Backgrounds qq qqww νν + 2 hard jets tt Central jet veto Anti b-tag Try to estimate background from data 41 QCD WW Use reconstructed missing ET Use di-lpton mass as a discriminator

42 An Essential Tool: Collinear Approximation The Taus are boosted Assume the leptons and neutrinos are collinear with the Taus P± = xp ± ± l τ Solution is valid if the scale 0<x<1 Define α ± 1 x E P ± νν = = = x E P Given the missing momentum in the transverse plane P/ x, y one can solve for α ± P / x P / y ± Px P l y l α ± = P P 42 P xl xl yl + + P yl ± l νν ± ± l One finds 2 2( + 1)( 1) and the resolution M = α + α + P+ P H l l ΔM H 1 1 = T T M P H x P l y P P l l l 2 P P xl sinφ i.e. No collinear mass for back to back leptons yl Another view: The collinear mass is the mass of the Higgs Boson needed in order to boost the system to the Higgs rest frame

43 An Essential Tool: Collinear Approximation One finds ( )( ) M = 2 α + 1 α + 1 P+ P 2 H + l l The resolution ΔM 1 H MH M H MH M H Δ/ Px Py + P l x + P l +Δ/ y Py + + P l x l M H 2det α+ + 1 α + 1 α+ + 1 α + 1 det P P = P x + y + l l T T P P l l x P l y l + sinφ depends on the acoplanarity angle and the missing E T resolution τ φ τ 2ν+l 2ν+l 43

44 2ν+l τ τ φ 2ν+l 44

45 H H ττ + 1j ττ + 0 j 45

46 46

47 H ττ qqh, H ττ Channels lepton-lepton lepton-hadron hadron-hadron The ττ decay mode extends the sensitivity to lower Higgs masses where there is a BR(H ττ) Mass can be reconstructed t for ττ using the collinear approximation (for lepton- lepton and lepton-hadron 47

48 H ττ Dominant Backgrounds Reducible Irreducible Drell Yan Z ll+jets Drell Yan Z ττ ll+jets tt+jets where the b-jets mimic forward jets (tt has a huge cross section) QCD & EW WW 48 ττ e μ + E T 30 fb 1

49 Use of Control Data Samples 49

50 The Breakthrough: VBF Han, Valencia, Willenbrock (1992); Figy, Oleari, Zeppenfeld (2003,2004) qqh, H WW,ττ The ττ decay mode extends the sensitivity to lower Higgs masses where there is a BR(H ττ) Mass can be reconstructed for ττ using the collinear approximation ττ eμ + E T 30 fb 1 50

51 Very Preliminary Combination Based on ATLAS TDR and ATLAS VBF Scientific Note Done with the Profile Likelihood method assuming the asymptotic chi squared behavior 51

52 The Profile Likelihood for Significance Calculation 2log λμ ( ˆ ± N σ ˆ μ ) = N = 2log λμ ( ) N 2 In particular if we generate background only experiments, λ(μ=0) is distributed as χ 2 with 1 d.o.f Discovery has to do with a low probability of the background only experiment to fluctuate and give us a signal like result. To estimate a discovery sensitivity we simulate a data compatible with a signal (s+b) and evaluate for this data λ(μ=0). For this data, the MLE of μ is 1 52 Comb Stat Forum Sep07 - Eilam Gross and Ofer Vitells

53 53 ILLUSTRATIVE E.G. & o. Vitells

54 Exclusion with Profile Likelihood Exclusion is related to the probability of the would be signal to fluctuate down to the background only region (i.e. the p-value of the s+b observation ) Here we suppose the data is the background only and the exclusion sensitivity is given by N = 2 λμ ( = 1) Exclusion at the 95% C.L. means N=2 54

55 ATLAS Exclusion Sensitivity ILLUSTRATIVE E.G. & o. Vitells 55

56 Cases for ILC OK, so LHC discovered a Higgs Boson, now what? What happens if we observe one Higgs and nothing more? Is it a SM Higgs Boson? What is its width? What is its spin? What are its couplings? Even though some properties can be probed with the LHC, if the LHC is a Higgs hunter, the ILC would be a Higgs Probe.. 56

57 57 From LHC to ILC

58 From LHC to ILC K. Desch, LCWS 07 Also workshop for the LHC early phase for the ILC, Fermilab, 07 58

59 The Case of Observing One Scalar Even if the Higgs is supersymmetric there is a large region in the parameters space where the LHC can observe only one of the Higgs Bosons An accurate measurements of the couplings & BRs of the observed Higgs can reveal its nature 59

60 Some Things ILC Does Better Moriond QCD Helenka Przysiezniak Here m h =120 LHC: for 300 fb -1 and BR( H bb) / BR( H WW ) R = 110<mH<190 GeV LC % [ BR( H bb) / BR( H WW )] SM LHC 20% Δg 2 /g 2 ~ 10%-45% (except for b) ΔΓ H /Γ H ~ 10%-50% hep-ph/ Dührssen Δm A = 30% for m A = 800 GeV also in parameter regions where LHC is blind 60 LHC/LC interplay in the MSSM Higgs sector Georg Weiglein and Sven Heinemeyer,Lidia Zivkovic,E.G.,Klaus Desch JHEP09(2004) (September 2004) J. High Energy Phys

61 The Ultimate Goal: Probing the SSB Sector Derive 95%CL bounds from χ 2 fit to m vis shape Moriond QCD Helenka Przysiezniak SM assumed to be valid except for self coupling. Assume m H precisely known, and BR(H WW) known to 10% or better. gg HH (W + W - )(W + W - ) (jjl ± ν) (jjl ± ν) (l = e,μ) for m H >150 GeV/c2. The self coupling λ is determined to % within 1σ for 150<m H <200 GeV This will have to wait for a future collider 61

62 The Ultimate Goal: Probing the SSB Sector Self Higgs coupling is not a piece if cake for a LC as well! Needs >1 TeV with a polarized electron beam and 1 ab -1 to probe it to the level of ~10% This by itself might justify a LC (If we observe a scalar) K. Desch, LCWS 07 Also workshop for the LHC early phase for the ILC, Fermilab, 07 62

63 Some Things LHC Can Also Do Combining CMS and ATLAS with high luminosity (yet accessible within a few years) with channels like H γγ, H ZZ(*) 4l, WBF H ττ l+hadr an accuracy of 0.1% might be achieved for the mass with m H ~ GeV, that is not so different from LC Moriond QCD Helenka Przysiezniak This is under the condition that the systematics on the energy scale is below the 0.1% for photons/electrons and below 1% for jets. 63

64 Conclusions LHC is a discovery machine LHC is scheduled to deliver data starting end of 2008 The optimist expects to see a hint of a Higgs Boson by the end of 2009 The pessimist will wait another year SUSY might come earlier 64