1. Wouter Verkerke, NIKHEF Generating W+jets background with AlpGen 2.06 – the end Wouter

2. Wouter Verkerke, NIKHEF Why would you care about W + jets background? Important for everybody – Top / SUSY / Higgs If ttbar is your signal, or if it is a major background you should worry about W+jets as well –Same final state (SL ttbar final state= W + 4 jets) –Except that ttbar has 2 b-jets and generic W+jets doesn’t (too much) Big question: how much W+jets background is there? Turns out to be a difficult question on the generator level, since a jet is reconstruction level object

3. Wouter Verkerke, NIKHEF First order approach: W + 4 partons To first order W + 4 jets background can be estimated by the W + 4 partons ME process There exists a dedicated LO generator for this process: AlpGen –Can generate many more processes actually. W  l

4. Wouter Verkerke, NIKHEF The ‘Rome’ ttbar commisioning analysis For the original ‘Rome’ ttbar commisioning analysis we used sample ‘A7’ as background A7 sample is W+4 partons calculated by Alpgen processed through Jimmy (=Herwig + Underlying Event simulation) for fragmentation, hadronization, parton showers etc… A7

5. Wouter Verkerke, NIKHEF So why worry? Main concern is that A7 sample (W + 4 partons) does produce a lot of W + 4 jets background, it may not be the only source Why? 1) Because Alpgen does not calculate the complete W + 4 parton cross section. –Analytical calculation of gluon radiation has divergences (soft collinear limit) that make it impossible to calculate the cross section for the full W + 4 parton phase space –We around this through the concept of parton showers: (Soft) gluon radiation is attached a posteriori to output of ME element calculation. –Happens in HERWIG phase of processing: means that W+4 parton ME may ultimately have >4 partons. –Also means that W + 3 parton ME followed by ‘hard’ parton shower can generate W + 4 jet event 2) Because a W + 2/3/5 parton event may be (mis)- reconstructed as W + 4 jets

6. Wouter Verkerke, NIKHEF Background from W + n parton visualized We know that we underestimate the level of background –Only generating W + 4 partons now, but W + 3,5 partons may also result in W + 4 jet final state due to splitting/merging W  l W + 4 partons (97 pb * ) W + 3 partons (315 pb * ) W + 5 partons (29 pb * ) parton is reconstructed as 2 jets 2 parton reconstructed as single jets * These are the cross sections with the analysis cuts on lepton and jet pT applied at the truth level or parton added by PS technique

7. Wouter Verkerke, NIKHEF How to calculate a better W+4jets estimate Need to take contributions from W + 0,1,2,3/5 parton ME into account as well Double counting problem: W + 4 partons can be –W + 3 ME partons + 1 Parton Shower –W + 2 ME partons + 2 Parton Showers –W + 1 ME partons + 3 Parton Showers –W + 0 ME partons + 4 Parton Showers Even worse when considering jets: W + 4 jets can be –W + m partons + n Parton Shower (~30 combinations possible in total) How to avoid double counting: Introduce a convention that decides which part of the 4-jet phase space is generated by ME (=AlpGen) and which part of phase space if generated by PS(=Herwig)

8. Wouter Verkerke, NIKHEF A prescription to eliminate of double counting Which techniques works best where? –PS better at low pT & collinear jets –ME better at high pT and non-collinear jets Make division in 4-jet phase space based on 2 jet parameters: –If pT(jet) pT_CUT  Jet should be generated with ME technique –If R(j-j) dR_CUT  Jet should be generated with ME technique Easy idea, but cannot ‘configure’ PS technique to do this, so we proceed as follows: 1)Generate W + n Partons ME, add parton showers, reconstruct jets on truth level 2)Determine a posteriori if configuration is ‘legal’, i.e. if PS resulted in a jet that should have been generated by W + (n+1) ME process (i.e. if pt(jet) to large) 3)Veto all ‘illegal’ configurations a posteriori Technique called MLM matching –NB: There are important details concerning scales that I didn’t cover

9. Wouter Verkerke, NIKHEF Does MLM matching work ? Look at P T distribution of W-boson at Tevatron –Region of high W-boson transverse momentum described by matrix element computation –Sum of MLM-matched W + n ME parton samples describes CDF data well W+1jetsW+0jets W+2/3/4jets (Plot taken from presentation by M. Mangano) W P T W-boson = net P T radiation

10. Wouter Verkerke, NIKHEF Applying MLM to estimate W + 4 reco jet background Generate samples of W + n ME partons + PS sample (n=0,1,2,3,4,5) Look at contribution of each sample to W + 4 reco jets final state Sample (# of ME partons)#Reco jetsSample (# of ME partons) #Events # Alpgen ME partons versus # reconstructed jets Constribution of ME parton samples in selected events (4 reconstr. jets)

11. Wouter Verkerke, NIKHEF So it works! Or not… Main point of following MLM procedure is that it gives a more reliable estimate of W+4jets background Key concern: Is the estimate from MLM matching of W+(0,1,2,3,4,5) jet samples stable under a variation of the matching parameters –In other words: If I would put the transition point between PS and ME at pT(Jet)=20 GeV instead of 40 GeV, would I get the same answer? Solution: Check this explicitly by rerunning several configurations –Lots of work… –Q: What is ‘reasonable’ range to vary pT/R parameters? PS are known to describe data poorly at very high pT. ME are known to describe data poorly at very low pT? Is there are region where both descriptions are good?

12. Wouter Verkerke, NIKHEF Recent results: Comparing pT20 and pt40 samples Now working with completely new & automated setup to generate W+njet samples –All generation/ reconstruction steps performed on grid –Fully automated job submission and bookkeeping –Upgraded to Alpgen2.06 –Updated p.d.f to CTEQ6 following Atlas convention –Now generating all leptonic W decays (e,,), was only e before (Very preliminary) results on visible cross section after analysis cuts Matching at pT=20Matching at pT=40 W+0jets0.000.02 W+1jets0.01 W+2jets0.020.08 W+3jets0.040.44 W+4jets0.172.52 W+5jets6.651.30 Total6.904.37 Hmmm!

13. Wouter Verkerke, NIKHEF New results on W+jets background from Alpgen Estimate of W+jet background surviving ttbar commissioning analysis cuts –Numbers are cross section  efficiency (multiply by L to obtain #events) after standard analysis cuts 1 good electron pT>20 GeV, 4 good jets pT>40 GeV, Missing ET > 20 GeV JetPt>40pT match =10pT match =20pT match =40pT match =80 R match =0.314.338.815.543.72 R match =0.713.227.516.205.73 More PS More ME Amount of background predicted by Alpgen stable within ~25% in centrial region (pT match =20,40 / R match =0.3/0.7)

14. Wouter Verkerke, NIKHEF Relative contribution of samples What is the relative constribution of the W+0,1,2,3,4,5 subsamples to the total for the various matching choices? –For pT=10 only W+5 samples contributes, not sure if matching really works here… pT match =10pT match =20pT match =40pT match =80 R match =0.3 R match =0.7

15. Wouter Verkerke, NIKHEF Selected plots from MLM matched W+jets background simulations Distribution of jet pTs with breakdown by sample –Histograms show cumulative distributions of W+0,1,2,3,4,5 parton samples pT(jet1)pT(jet2)pT(jet3)pT(jet4) pT(match)=40 pT(match)=20

16. Wouter Verkerke, NIKHEF Selected plots from MLM matched W+jets background simulations Comparison of cumulative distribution of jet pTs with matching scale at 20,40,80 GeV –All histograms to unity – just comparing the shapes pT(jet1)pT(jet2) pT(jet3) pT(jet4) Note that pt match =80 sample falls short in pT<80 GeV region (= region generated by PS)

17. Wouter Verkerke, NIKHEF Properties of the W: electron pT and ETmiss Comparison of cumulative distribution of pT(elec) and ET(miss) with matching scale at 20,40,80 GeV –All histograms to unity – just comparing the shapes –Good agreement now (also in pT of electron) pT(elec)ETmiss

18. Wouter Verkerke, NIKHEF A closer look at some basic distributions: m(jjj) Distribution of 3-jet invariant mass with breakdown by sample (‘top background’) –Histograms show cumulative distributions of W+0,1,2,3,4,5 parton samples m(jjj) pT(match)=20 GeV m(jjj) pT(match)=40 GeV m(jjj) pT(match)=80 GeV

19. Wouter Verkerke, NIKHEF A closer look at some basic distributions: m(jjj) Comparison of cumulative distribution of 3-jet invariant mass (‘top background’) with matching scale at 10,20,40,80 GeV –All histograms to unity – just comparing the shapes

20. Wouter Verkerke, NIKHEF Summary on W+jets background Studied W+4 jets background from MLM matched Alpgen samples with W+0,1,2,3,4,5 ME partons at varying MLM matching thresholds –pT(match)=10,20,40,80 GeV / Rjj(match)=0.3,0.7 Background to ttbar analysis stable to ~25% under reasonable variation in matching parameters Stable region of MLM matching predict background that is ~30% lower than original Rome ‘A7’ estimate –Appears to be related to choice of R=0.4 for generation of ‘A7’ sample W+4jets background appears under control. For final studies CSC samples 8220-8237 might be suitable (pTmatch=20, soft lepton cut), but need to look into filtering criteria –Samples used to be ‘Higgs / top common’, but has morphed into Higgs only due to application of VBF event filter. –This means there are currently no CSC samples that we can use to estimate W+jets background in ttbar for CSC commisioning notes. –We will need our own sample. Can reuse ME files from 8220-8237 but need to apply different filter (which is mostly a logistical / coordination exercise).

21. Wouter Verkerke, NIKHEF New W+jets background in ttbar analysis Estimate for 200 pb-1 of data (e+)