1. ttH can be feasible using shape ….? Samir FERRAG University of Glasgow Les Houches, 23/06/2007 Not Official ATLAS results

2. H  bb, via ttH with mH< 130 GeV Complex and challenging channel (H  bb):Complex and challenging channel (H  bb): –One lepton (trigger) –4 b-jets + 2 jets  Need excellent b-tagging: signal efficiency ~O(  4 b ) –Fully reconstructed t –High jet multiplicity CombinatoricsCombinatorics Difficult to simulateDifficult to simulate Background:Background: –Reducible: tt+jj, W+jets, WWbbjj Extracted from dataExtracted from data Reduced by imposing reconstructed tt pairReduced by imposing reconstructed tt pair –Irreducible: ttbb, combinatorics Slight difference in kinematics with the signal (Likelihood)Slight difference in kinematics with the signal (Likelihood)

3. Evolution of results After an optimistic evaluation on the TDR: –Updating pdf: 20% smaller cross sections –Matrix element based Monte Carlo for ttbb background  higher jet multiplicity Realistic environment –Significance down to < 2 

4. Counting experiment: Constraints on the ttH analysis Higgs mass=120GeV Higgs mass=130GeV 10 fb -1 100 fb -1      ATLFAST analysis ATLFAST analysis  The constraints on ttH are challenging

5. CE: Constraint on ttH for 10fb -1 55 44 33 22 6% of PDF uncertainties (on ttbb+ttjj) S/B systematic uncertainty on B (%) Requested S/B For 3  significane For 6% PDF uncertainty Need 1 to 2 % of accuracy On background to reach 2  significance Higgs mass=120GeV S/B=0.14 From Markus analysis (ATLFAST)

6. Signal Reconstruction The signal is plagued by wrong combinatorial pairs (>70%) which makes it looking- like the background --> - lower the S/B power Signal Background Wright combinations

7. Likelihood2 for the Signal/Background Robustness of distributions validated by a comparison between MC@NLO and ACERMC

8. Likelihood1 to reduce combinatorics

9. CE: Sensitivity results (without systematics) Current Analyses: A1- Cuts Based Analysis: C. Collins-Tooth  s/  b=1.9 A2- Selected Likelihood: G. Aad  s/  b=3.1 90 -150 GeV s/  b is computed in this Interval (Signal Interval) 5% –For 100fb -1 (in 2011) and Background accuracy of 5% : S/B=30% for a 5  discovery (m Higgs =120 GeV) S/B=16% for a 3  evidence (m Higgs =120 GeV)

10. Shape Analysis: Getting and using  2 H0H0 H 0 ---> t 0 ( where t 0 is a template for the null hypothesis, H 0. ) H 1 ---> t 1 ( where t 1 is a template for the signal plus background hypothesis, H 1. ) ‘Data’ ---> gives us one  0 2 (H 0 ) and  1 2 (H 1 ). Generate ‘Data’ 10000 times (in pseudo experiments). 2lnQ =-  2 H1H1  2 = 9.876 e.g. 1 - CL b = Probability that the background will fluctuate to look like the signal Equivalent in counting experiment: …….5   1-CL b = 2x10 -7. CL s+b = Confidence Level for excluding H 0. Use:- CL s+b 1 - CL b

11. Improvement from using shape Shape Analysis  2 (CE) One Bin  2 H1H1 H0H0

12. Effect of shapes Probability of measuring with 30fb -1 33 55 Counting Experiment19%<1% Shape38%3% Shape 15% on Signal X- sec 30%1% Probability of measuring with 30fb -1 22 33 Counting Experiment30%6% Shape34%8% Cut based analysis Likelihood analysis The hape of distributions helps for both likelihood and shape analysis but is it robust?

13. Backup slides

14. CE: Constraint on ttH for 100fb -1 6% of PDF uncertainties S/B systematic uncertainty on B (%) Requested S/B For 3  significane For 6% PDF uncertainty 55 44 33 22 Need 4% of accuracy On background to reach 3  significance S/B=0.14 From Markus analysis (ATLFAST) Higgs mass=120GeV