1. Measuring the W+charm Cross Section in CMS J. Alcaraz, I. Josa, J. Santaolalla (CIEMAT, Madrid) CMS EWK Working Meeting 27 May 2011 Measuring the W+charm Cross Section in CMS J. Alcaraz, I. Josa, J. Santaolalla (CIEMAT, Madrid) CMS EWK Working Meeting 27 May 2011 J. Alcaraz (CIEMAT) 7 March 2011 J. Alcaraz (CIEMAT) 7 March 2011

2. Wc analysis, EWK Working Meeting, 27 May 2011 2 Why is W+c interesting ● In “W+c”, the W production proceeds predominantly via “gluon + s-quark”: g + s -> cc + s -> c + W - ”. This means that this channel gives direct access to the s-quark PDFs:

3. Wc analysis, EWK Working Meeting, 27 May 2011 3 Why is W+c interesting ● This is also a non-negligible source uncertainty for the W mass measurement (at least a priori) (from M.W. Krasny et al, arXiv:1004.2597)

4. Wc analysis, EWK Working Meeting, 27 May 2011 4 Non-strange contributions to W+c ● Valence quark contribution for W - : g + d -> cc + d -> W - + c. This is strongly Cabibbo suppressed (|V cd | 2 / |V cs | 2 ~ 0.05), but it is partially compensated by the fact that a “d” is a valence quark -> its contribution is: ~ 15 %. ● Valence quark contribution for W + : g + d -> cc + d -> W + + c, but an “anti-d” is not a valence quark: it is much more suppressed in the W+ case (i.e. there may be small differences depending on the charge of the W). Contribution: ~ 5%.

5. Wc analysis, EWK Working Meeting, 27 May 2011 5 Non-strange contributions to W+c ● Gluon splitting of the type: u + d -> W + + g -> W + + cc. In this case, there are two c-quarks in the final state, but they are confused with our signal. These contributions are small, but not fully negligible. At the end of the day, s (W + + c)/ s (W - + c) ~ 1.0-1.1 according to our MCs (POWHEG- MADGRAPH). More gluon splitting pushes the ratio slightly up, more g+d → W - + c pushes the ratio slightly down.

6. Wc analysis, EWK Working Meeting, 27 May 2011 6 LHC versus Tevatron ● Charm fraction is ~ 5 times larger at the LHC (~40%, to be compared with ~8% at Tevatron for pT(jet)>20 GeV) ● So light-quarks dominate much more at Tevatron and, on top of that, the genuine charm fraction from strange quarks in the initial state is comparable to the total fraction from ccbar and bbar from gluon splitting -> complicated mixture ● And I guess that we also have better lifetime tagging... :) ● CDF has published a total charm cross section measurement (ptjet>20 GeV, |eta|<1.5) with 33% uncertainty (with 1.8 fb -1 ) ● D0 has a preliminary result for conferences on the charm ratio for ptjet>20 GeV and 1 fb -1 (7.1+-1.7%, i.e. a 24% uncertainty). ● Both measurements use semi-leptonic charm decays and test that their charge is opposite to the W charge.

7. Wc analysis, EWK Working Meeting, 27 May 2011 7 ● Simple: use the standard VBTF W selection and apply b-tagging criteria to the observed jets in the event. ● This will work for W production because it is almost impossible to produce W+b in the final state. For instance, g + u -> bb + u -> W + + b is very strongly suppressed (~|V ub | 2 ~ 10 -5 ), as well as g + c -> bb + c -> W + + b (strong charm PDF suppression and |V cb | 2 ~ 2*10 -3 ). ● Main backgrounds in practice will be ttbar and single top (giving W + b quarks in the final state). ● There is still gluon splitting of the type: u + d -> W + + g -> W + + bb. But this contribution is at the 1-2% level and not visible in the final distributions compared with top backgrounds. Strategy to measure W+c+X

8. Wc analysis, EWK Working Meeting, 27 May 2011 8 We have analyzed 2010 data... ● We did that for several reasons: ● We considered essential to bring the analysis from the beginning to the end, including systematics assignment, to be sure that there are no serious bottlenecks. ● At the same time we will get a feeling of the sensitivity and accuracy that we can get on this measurement ● 2011 data and MC are not fully ready yet (reprocessing, pending calibrations,...) ● It does not mean that we do not plan to analyze 2011 data, obviously ● We are only using the W->muon channel. Electrons can be used, but with at least one caveat: ● We can not get rid of QCD so easily (and the experience from the W polarization analysis is not very encouraging)

9. Wc analysis, EWK Working Meeting, 27 May 2011 9 Analysis of W+c in the muon channel ● Current set of VBTF cuts to select W->mu nu, 38X processing (Nov 4th):  Single-mu triggered (HLT_Mu15_v1 at the end of 2010 data-taking),  One muon with P T > 25 GeV, | h |<2.1,  VBTF tracker+muon quality cuts (|d xy |<2 mm, minimal number of hits, at least two segments, c 2 cut),  Z-> mm veto (two global muons with ptmax>20 GeV, ptmin>10 GeV)  ISO variable <0.1  M T > 50 GeV  I.e. no fit to the MT distribution to extract the cross section (unnecessary complication)  p T (hadron jet) > 20 GeV, no more than 3 jets above 40 GeV ● We use particle-flow jets, L2+L3 corrected according to official calibrations We finally plot the b discriminator of the most significant jet

10. Wc analysis, EWK Working Meeting, 27 May 2011 10 Reference MC for this study ● We use the POWHEG MC WITH PILEUP for W production. This should provide a reliable prediction for W + 1 hard jet + soft/collinear jets. POWHEG has some advantages: ● Straightforward access to 'single-charm' productionl: “W+c” or “W+nonc” information is directly accessible in the generator information with status=3 ● More direct comparison with theory calculations (pp -> W+c (+1 jet)). ● PDFs are already NLO (a sensible NLO comparison can not be done with MCs like Alpgen, MadGraph or Sherpa). ● And one disadvantage: ● W + ≥ 2 hard jets are not so reliably predicted by POWHEG. ● But we cross-check with W+jets MadGraph MC samples too We finally plot the b discriminator of the most significant jet (no implicit cut on jet E T for the moment (effective cut is ~ 20 GeV) MC PLOTS ARE NORMALIZED TO LUMI * XSECTION ((N)NLO), UNLESS 'FITTED'

11. Wc analysis, EWK Working Meeting, 27 May 2011 11 SSVHE as our default Simple secondary vertices (SSV, discriminator = log(1+decayLengthSignificance)) should be less sensitive to pileup (thinking on 2011 data). Good agreement with POWHEG out-of-the- box. Use negative vertices to control the light-quark contribution The W+c signal (red dashed histogram) is clearly visible As well at the ttbar and single-top backgrounds. QCD is negligible W+W+W+W+ W-W-W-W-

12. Wc analysis, EWK Working Meeting, 27 May 2011 12 We'll use TCHE as cross-check “TCHE”: significance of the second most significant IP in the jet. Good agreement too. Sensitivity for charm with small light-quark backgrounds for disc>3 As well at the ttbar and single-top backgrounds. QCD is negligible W+W+W+W+ W-W-W-W-

13. Wc analysis, EWK Working Meeting, 27 May 2011 13 Fit procedure  Use templates for signal, top, light-quark and “other” contributions. Fit the charm yield for W + and W - separately. Plots above are “after fit”  Negative vertices help to constrain the light-quark contribution below the charm signal peak (but note that positive and negative contributions are not symmetric: there are also K 0 and L contributions to positive vertices in light- quark jets)  Data-driven top templates W+W+W+W+ W-W-W-W-

14. Wc analysis, EWK Working Meeting, 27 May 2011 14 Top templates  Single-top is dominant over ttbar in the current selection. But both single-top and ttbar have similar templates, 'harder' than W+charm  We build a data-driven template from the control sample with 3 or more jets  Our data-driven template looks softer than what MC predicts, we we have not found hints of significant background in the control sample (looking at the invariant mass at secondary vertex) -> anyway this is taken into account as systematic uncertainty

15. Wc analysis, EWK Working Meeting, 27 May 2011 15 Results  The measurements are in the expected range (~1 for the charge ratio, ~40% for the charm fraction over the total)  Only statistical uncertainties shown. Systematics is discussed in the next slides

16. Wc analysis, EWK Working Meeting, 27 May 2011 16 Some additional distributions  Good agreement with MC, but no sensitivity to improve the analysis (except to reject a few top events in the tail) W + : Invariant mass at vertex (GeV) W - : Invariant mass at vertex (GeV)

17. Wc analysis, EWK Working Meeting, 27 May 2011 17 Some additional distributions  Excellent agreement with MC!!  We are thinking on cutting in this distribution (a significant fraction of light- quark decays corresponds to large decay length uncertainties) W + : decay length uncertainty (cm) W - : decay length uncertainty (cm)

18. Wc analysis, EWK Working Meeting, 27 May 2011 18 Some additional distributions  Good agreement with expectations  We will use this distribution to assign tracking systematics:  Determine a probability to lose a track that leads to a bad chi2 in the data-MC comparison (we use chi2=12/5 -> 3.5% probability) W + : number of tracks at vertex W - : number of tracks at vertex

19. Wc analysis, EWK Working Meeting, 27 May 2011 19 Systematic sources considered  Difference in efficiency between positive and negative muons: taken from Z- >mumu inclusive studies (~1%)  Muon resolution: taken from Z->mumu studies too (SIDRA/MUSCLE)  Pileup (+ missing E T ) uncertainties: half of the difference with respect to the result obtained with the W MC with 'no pile-up'  Jet scale/resolution: change the jet scale by 3% and repeat the analysis (this should also cover discrepancies in resolution: a few/ten percent of 15% is smaller than 3%).  Jet multiplicity: reweight the MC POWHEG multiplicity to exactly agree with the one seen in data  Tracking resolution: determine a probability to lose a track in vertex reconstruction that leads to incompatibility with data (chi2 probability < 5%). Also check the agreement data-MC in the decay length uncertainty distribution.  Top background: repeat the analysis with a data-driven template  Light-quark systematics: assume that all negative vertices are light-quarks (this will change the light-quark fraction by at least ~20% on the positive vertex side of the fit)

20. Wc analysis, EWK Working Meeting, 27 May 2011 20 Systematic sources considered  Other backgrounds: change DY and Wtaunu by 10%, QCD by 50%, bbbar by 50%  PDF uncertainties: usual uncertainty recipe for the CT10 set, but at 90% CL  Average energy fraction in charm fragmentation: change the average fraction for D* (or D if no D* is produced) by 2%, which covers safely the ranges of variation of LEP measurements (ALEPH, OPAL,...)  Change in charm fragmentation fractions c → D 0, D ±, D s, L c : change them according to uncertainties from arXiv:hep-ph/0603175 (a bit larger than the last one used by ATLAS). We have checked that the fractions present in MC are within one stnadard deviation from what is given in this reference.  PDF uncertainties: usual uncertainty recipe for the CT10 set, but at 90% CL  Average energy fraction in charm fragmentation: change the average fraction for D* (or D if no D* is produced) by 2%, which covers safely the ranges of variation of LEP measurements (ALEPH, OPAL,...)  Change in charm fragmentation fractions c → D 0, D ±, D s, L c : change them according to uncertainties from arXiv:hep-ph/0603175 (a bit larger than the last one used by ATLAS). We have checked that the fractions present in MC are within one stnadard deviation from what is given in this reference.

21. Wc analysis, EWK Working Meeting, 27 May 2011 21 Systematics for charm charge ratio  Details described in CMS-AN-11-156 (being finalized now)  Result: R c ± = 0.97 ± 0.21 (stat.) ± 0.06 (syst.)

22. Wc analysis, EWK Working Meeting, 27 May 2011 22 Systematics for charm ratio  Details described in CMS-AN-11-156 (being finalized now)  Result: R c = 0.443 ± 0.097 (stat.) ± 0.109 (syst.)

23. Wc analysis, EWK Working Meeting, 27 May 2011 23 Cross checks  Redo the analysis with TCHE tagging in the 3 < DISCR < 20 region W+W+W+W+ W-W-W-W-  Consistent with the SSVHE result within systematics

24. Wc analysis, EWK Working Meeting, 27 May 2011 24 Cross checks  Redo the analysis with a MadGraph MC instead of POWHEG W+W+W+W+ W-W-W-W-  R c ± with MadGraph is consistent within statistical uncertainties, but differs from the POWHEG result beyond systematic uncertainties -> studies going on to understand why (light-quarks, obsolete LO PDFs,...)

25. Wc analysis, EWK Working Meeting, 27 May 2011 25 ● Understand why MadGraph provides different results for the “charge ratio” (first indications point to a significantly different light-quark component, also different for positive and negative vertices with respect to POWHEG, but we have to understand the origin of the effect in more depth) ● Try some variations: ● Do not use vertices with too high decay length uncertainties ● Cut explicitly on eta of jets ● Cut even looser on number of jets (top background is 'fitable' anyway) ● Do comparisons with predictions from different PDFs ● And of course start analyzing 2011 data (things must be more under control now and reprocessing is almost finished) ● For the future: do the analysis as a fucntion of different (ptjet, eta) bins TO DO LIST

26. Wc analysis, EWK Working Meeting, 27 May 2011 26 ● Do we have enough MC for 2011 for a sensible analysis? We are afraid that the statistics in MC will be lower than in data (both for POWHEG and MadGraph) ● We already have a competitive result with 2010 data. We do not know how long will it take to get it done with 2011 data, and 200 pb -1 or so will already lead to measurements that are dominated by systematics ● Which means that we have to work hard too on reducing some systematic sources CAVEATS APPLICATIONS ● If we assume the current PDF knowledge, this analysis can be used to estimate the charm-tag efficiency with uncertainties of order ≤ 20% ● Measuring a single-top cross section is trivial (even with reasonable uncertainties). Why people still think it is difficult? There are several level arms to disentangle it from the ttbar background (number of jets, charge)