Presentation is loading. Please wait.

Presentation is loading. Please wait.

Referee Report on Open charm production results for summer conferences, 2010 Peter Clarke Marcel Merk “Observations” and “Comments” The referees thank.

Similar presentations


Presentation on theme: "Referee Report on Open charm production results for summer conferences, 2010 Peter Clarke Marcel Merk “Observations” and “Comments” The referees thank."— Presentation transcript:

1 Referee Report on Open charm production results for summer conferences, 2010 Peter Clarke Marcel Merk “Observations” and “Comments” The referees thank the Open Charm task force members for their willingness to answer many questions over the past two weeks and are impressed by the turnaround speed by which these answers were provided.

2 Four decay measurements in one note D 0 production via D 0 ->K π + cc D* 0 production via D*->D 0 π slow + cc D + production via D  Kππ and D  KKπ as a cross check D s + production via Ds  φ(KK)π + cc Goal: open charm cross section in 8 bins of p t and 5 bins y For D 0 analyses two independent analyses available – Selection A (“conventional cuts” + PID) Optimized for low background – Selection B (“kinematic & geometric variable” + no PID) Optimized for S/sqrt(S+B) : higher efficiency & higher background Many discussions with referees over last two weeks: – Established procedure: Selection A = default; Selection B = systematic check

3 Internal and Public Notes Detailed Internal Note soon available ~ 100 pages – Impressive document with a lot of analysis details and systematic checks Presentation of individual analyses rather non- uniform (language and contents) – Should improve presentation in the public note

4 Selections Dataset used: TCK 51710 for total efficiency calculations – Larger dataset for distributions (TCK 51710 & TCK 81710) – Consistent datasets between individual analyses The D 0,D*,D +,D s + selections use different methods – Optimizations: 1) by-eye, 2) S/sqrt(S+B), 3) sigma(S)/S Data-MC comparisons only possible for distributions after selection – Some data-MC distributions differ (eg D 0 IPχ 2 ), but effects taken into account Pre-selections are often as tight as final selection – Cut variation analyses not always possible D 0 selection in D* analysis is very similar to D 0 analysis A

5 Efficiencies Factorization of efficiencies consistently done in all analyses Differences seen between magnet up and magnet down data: – Different in different analyses – No systematic error applied D0 selection: – Analysis A (use PID): lower Background – Analysis B (no use PID): higher efficiency, few extra bins  Difficult to test consistency, requires more time

6 Yield Extraction Different background subtraction methods used in different selections – D 0 analysis: Sideband background fitting (analysis A) vs sideband subtraction (analysis B) – D s analysis: sWeight method used in optimization Primary charm extraction based on IP shape fitting, without use of per-event errors and χ 2. – Less statistical sensitivity compared to using errors – IP shapes differ bin-by-bin in (pt,y), while χ 2 distributions are uniform Secondary charm component shows difference between data and MC – Typically few % Internal combinatoric background (multiple charm candidates per event)

7 Cross Sections Extracted D 0 cross sections “analysis A” and “analysis B” differ significantly in overall scale factor (~30%) – Reason not understood; most likely not due to PID cuts Ratio of D* cross section to D 0 cross section seems anomalous Cross check cross section D +  Φ(KK) π disagrees overall with D +  Kππ (~30%) Magnetic field up/down differences observed Although heroic effort was done, there was not sufficient time to understand the anomalies in scale factors between cross checks Detailed checks were done of many possible systematic effects.  All distributions shapes agree within errors

8 Referee Consensus – Hesitant to assign additional ad-hoc systematic error for “ununderstood” effects – Instead suggest to show the cross section p t and y distributions with arbitrary normalization Allows shape comparison with theory for ICHEP

9 The Status Today – For CKM The D 0 analyses A and B have both identified a bug and have implemented a few improvements The D* analysis has done many cross checks but remains unchanged The D + analysis has implemented some improvements and has done cross checks The D S + analysis remains unchanged but is cross checked

10 The D0 analysis Analysis “A”: – A problem with double counting in analysis “A” was identified and fixed – Background subtraction improved Likelihood vs Chi2 fits Compare BG fit to sideband subtraction Analysis “B”: – A problem in the efficiency calculation was found and fixed – Systematic error due to binning identified and included

11 Even yields and Cross Section Comparison Analysis A:Analysis B: Cross Check: Compare the cross section estimate of “A” and “B” in each bin of P T and Rapidity and plot the “Pull”. Stat error: assume A is a subset of B Syst error: include independent errors  “A” and “B” agree

12 D 0 Cross Section “A” and “B”

13 The D* analysis Nothing has changed in the D* analysis between ICHEP and now D* analysis: Combine D 0 with slow pion Implemented independent D 0 selection and cross section calculation: Analysis “C”

14 Comparison between “A” and “C” Agreement between D 0 analyses gives confidence in the D* analysis Selections are very similar Algorithms are the same (with some corrections related to slow pion π s ) Yield extraction methods almost equivalent D* efficiencies look reasonable when compared to D 0 Error calculation in the comparison: Statistical errors subtracted (conservative) Different systematic error  Consistent

15 D + and D S + Analysis The D + analysis had some modifications: – Track quality cuts taken uniform with other analyses – Tighten PID cuts and revisited systematic error – More robust secondary charm component determination (not critical) – Floating the widths of mass peak fits Yields increased by 10% Good agreement mass fitting with sideband subtraction  Cross section increased by ~10% The D S analysis has remained unchanged

16 The D + and D S + Cross Checks Compare the D + cross section extraction using the decay modes: – D +  Kππ and D +  KKπ (low statistics mode) Use the D + analysis to reconstruct D s +  KKπ and compare to the alternative D s +  φπ analysis

17 Comparison D +  Kππ to D +  KKπ Plot the ratio and fit a horizontal like: Fitted constant 0.947 +- 0.091  Two methods are in agreement

18 Comparison D S  KKπ and D S  φπ Plot the ratio and fit a horizontal line: Chi2/NDF = 3.49/3 Fitted constant 1.21 +- 0.17  Two methods are in agreement

19 D 0 cross section vs theory

20 D* Cross Section vs Theory

21 D + Cross section vs Theory

22 D S + Cross Section vs Theory

23 The Recommendation (this afternoon) Peter (Clarke) and Marcel are now comfortable with the progress of the charm group, and in particular the comprehensive level of cross checks that exist for all channels. In several cases there are truly independent checks, and in others indirect checks. We are satisfied that enough thought and checking has gone into this that we can now proceed to publish cross sections top CKM and this will be the recommendation.

24 The D 0 cross section approval timeline


Download ppt "Referee Report on Open charm production results for summer conferences, 2010 Peter Clarke Marcel Merk “Observations” and “Comments” The referees thank."

Similar presentations


Ads by Google