1 Bunch length measurement with the luminous region Z distribution : evolution since 03/04 B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Origin of the discrepancies between on- and off-line : Origin of the discrepancies between on- and off-line : Reminder Reminder Comparison using 2 samples from the same runs Comparison using 2 samples from the same runs Comparison using 2 samples containing the same events Comparison using 2 samples containing the same events Differences in the selection between on- and offline Differences in the selection between on- and offline

2 Reminder : different results between on- and off-line B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Measurement with 2 samples at RF voltage = 3.2 and 3.8 MV Data type  LER  HER  z 2  2 #events (3.2) #events (3.8) online M 1.50M online   M 1.50M offline M 3.30M offline   M 3.30M  Important variation between on- and offline. Why ?

3 Comparison of 2 samples made of the same runs B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Z [mm] = 1.10  0.1 mm RMS = 7.7  0.1 mm = 1.27  0.2 mm RMS = 7.8  0.2 mm online offline Data type  z 2  2  z 2 β * (y)  2 online 314 ~5  online 314  ~5  offline ~5  offline 322  1 ~5  Hardly consistent Cuts on the vertex χ 2 and on tan(λ 1 ) are applied + same frame

4 Comparing 2 samples containing exactly the same events B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady The timestamp is available both in online and offline samples: used it to select a sample containing exactly the same events online and used it to select a sample containing exactly the same events online and offline offline compared the vertex coordinates: z differs by ~5010 μm, Δz/z <6% compared the vertex coordinates: z differs by ~50  10 μm, Δz/z <6% => only a part of the discrepancy => only a part of the discrepancy

5 Differences in the selection… B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady offline online Z [mm] sqrt(x 2 +y 2 ) [mm] Z [mm] Cuts on the vertex χ 2 and on tan(λ 1 ) applied

6 Differences in the selection… B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady online Z [mm] cos(trk1-trk2) in rest frame Z [mm] Cuts on the vertex χ 2 and on tan(λ 1 ) applied offline cos(trk1-trk2) in rest frame

7 Differences in the selection: add cuts on tan(λ 2 ), Differences in the selection: add cuts on tan(λ 2 ), cos(trk1-trk2), sqrt(x 2 +y 2 ) B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Z [mm] online offline Data type  z 2  2  z 2 β * (y)  2 online 314 ~5  online 314  ~5  offline 314 ~4 273  offline 314  ~4 273  Z [mm] = 1.10  0.1 mm RMS = 7.71  0.1 mm = 1.22  0.2 mm RMS = 7.73  0.2 mm

8 Discrepancies between on- and off-line B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Not yet completely understood Not yet completely understood A small part is due to differences in the reconstruction A small part is due to differences in the reconstruction Samples built with the same runs lead to consistent results Samples built with the same runs lead to consistent results after a few extra cuts, but: after a few extra cuts, but: need more statistics to concludeneed more statistics to conclude -> get some other samples built with the same runs -> get some other samples built with the same runs need to reproduce, as much as possible, the same cuts in both samplesneed to reproduce, as much as possible, the same cuts in both samples

9 On the way B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Codes to subtract the slow and bunch number dependant z variationsCodes to subtract the slow and bunch number dependant z variations Codes to fit using unbinned likelihoodsCodes to fit using unbinned likelihoods Fit in slices of zFit in slices of z

10 Measurement with 2 samples taken at RF voltage = 3.2 and 3.8 MV B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Data type  LER  HER  z 2  2 #events (3.2) #events (3.8) online M 1.50M online   M 1.50M offline M 3.30M offline   M 3.30M if we subtract the bunch number dependent Z variation if we subtract the bunch number dependent Z variation offline offline   => Important variation between on- and offline. Why ? Large correlation between  LER and  HER ( > 99%)Large correlation between  LER and  HER ( > 99%) too large to find precisely the individual values ? too large to find precisely the individual values ? MC-TOYs have the same correlation and work correctly. MC-TOYs have the same correlation and work correctly. effect of fitting a PDF which doesn’t describe the data properly ? effect of fitting a PDF which doesn’t describe the data properly ? need more MC-TOY tests to check that. need more MC-TOY tests to check that. several discrepancies observed between on- and offline : several discrepancies observed between on- and offline : RMS of both RF distributions 0.1 mm larger in offline data RMS of both RF distributions 0.1 mm larger in offline data An offset of ~1mm in Z An offset of ~1mm in Z => Origin ? Different frames ? Something in the slow Z movement subtraction ? => Origin ? Different frames ? Something in the slow Z movement subtraction ? Cuts ? => We’ll try the offline analysis with exactly the cut than online. Cuts ? => We’ll try the offline analysis with exactly the cut than online.

11 Measurement with long coast data B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Data type  LER  HER  z 2  2 #events online k online 5.6   k offline k offline 6.4  14.6  k  Not enough stat. + correlations ? + correlations ?

12 Z variation as a function of the bunch number Z variation as a function of the bunch number B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady [mm] Bunch number Slow Z movement subtracted Slow Z movement not subtracted Mini-trains ?

13 Z variation as a function of the bunch number high vs. low I B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Bunch number [mm] High I Low I

14 Z-RMS variation as a function of the bunch number B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady 0 Bunch number 3492 Z-RMS [mm]

15 Systematic uncertainties B. VIAUD, C. O’Grady B. VIAUD, C. O’Grady Varying the parameters fixed in the fit within their Varying the parameters fixed in the fit within their known errors and re-compute the results. known errors and re-compute the results. How to evaluate the uncertainty due to the fact the PDF How to evaluate the uncertainty due to the fact the PDF used in the fit doesn’t describe properly the data ? used in the fit doesn’t describe properly the data ? try several PDFs (asymmetric bunches) ? try several PDFs (asymmetric bunches) ? let Beta * _y float ? let Beta * _y float ? use TOYs to produce distorded distributions compared to use TOYs to produce distorded distributions compared to the nominal PDF ? the nominal PDF ? ? ?

16 New Since Last Collab Mtg

17 New Offline-Style Analysis Necessary for analyzing MC/data with same code. New (simple) cuts: ntracks==2ntracks==2 Chi2(vertex)<3Chi2(vertex)<3 Mass(2track)>9.5GeVMass(2track)>9.5GeV E(charged showers)<3GeVE(charged showers)<3GeV 0.7<tan(lambda1)<2.50.7<tan(lambda1)<2.5 Note that all our units are mm (like PEP). Also, subtract Z motion of beamspot more trivially now (new value every 10 minutes).

18 New Offline Analysis Code Z [mm]  Check we see the same effect in data (from late July 2005) => Similar effect. Similar values of the fitted parameters   ~10   ~3

19 Monte Carlo with a gaussian Z distribution Z-distribution is generated in the mu-pair MC as a gaussian with Z-distribution is generated in the mu-pair MC as a gaussian with =0 mm and  = 8.5 mm =0 mm and  = 8.5 mm => No obvious effect due to the => No obvious effect due to the reconstruction / selection reconstruction / selection   ~1.1 Z [mm]

20 Z vertex resolution from MC 30um resolution is very small on the scale we are looking, so feels difficult for it to be a resolution effect. Z Reconstructed – Z True (mm)

21 Z-distribution/bunch length measurement as a function of bunch current Data/theory discrepancy could be due to Beam-Beam effect proportional to the bunch current Data/theory discrepancy could be due to Beam-Beam effect proportional to the bunch current => Compare Z-distribution at high and low current => Compare Z-distribution at high and low current Used data taken on July the 31 st and July the 9 th Used data taken on July the 31 st and July the 9 th LER: 2.4 A -> ~ 0.7 A LER: 2.4 A -> ~ 0.7 A HER: 1.5 A -> ~1050 A HER: 1.5 A -> ~1050 A Selected each time the first and last runs of the period Selected each time the first and last runs of the period during which the currents drop. during which the currents drop.

22 31 st of July Standard fit (waists Z-position or  * (y) not allowed to float ) Standard fit (waists Z-position or  * (y) not allowed to float ) No obvious difference at this statistics. When waists Z-position or  * (y) are allowed to float : Chi2 ~ 1, fitted values of Zwaist and  * (y) similar to those obtained with the usual sample. Z [mm]   ~1.4   ~1.3 High current Low current RMS=7.0 mm Z [mm] RMS=7.14 mm

23 9 th of July Standard fit (waists Z-position or  * (y) not allowed to float ) Standard fit (waists Z-position or  * (y) not allowed to float ) No obvious difference at this statistics. When waists Z-position or  * (y) are allowed to float : Chi2 reduced, fitted values of Zwaist and  * (y) ~ consistent with those we usually see.   ~1.4   ~1.3 High current Low current RMS=7.14 mm RMS=7.0 mm High current Low current   ~1.6   ~0.8 Z [mm] RMS=7.2 mm RMS=7.01 mm

24 Conclusions No obvious z-distribution distortion observed when analysis run on monte- carlo With available statistics, no obvious beam- beam effects in high/low beam-current runs.

25 How do we proceed? Analyze monte-carlo with correct hourglass shape (tried once, but hourglass in monte-carlo was not correct we believe). Unlikely cause, IMHO. Backgrounds (tau, 2-photon)? Unlikely cause, IMHO. Effect of parasitic crossings (now have bunch number in ntuples … so should be easy). Unlikely cause, IWHO. Think about asymmetric bunches more Perhaps help Ilya/Witold study at simulation? Some machine studies?

26 Reminder I Reminder I Fit the following PDF on the luminous region Z distribution: Fit the following PDF on the luminous region Z distribution: Number of particles per bunch, Z c : Z where the bunchs meet Allowed to float

27 Reminder II The theoretical distribution cannot describe the shape of the data.The theoretical distribution cannot describe the shape of the data.  Trying to understand this before proceeding with bunch length measurement!!  ~ 7.25 mm   ~13 Z [mm]

28 Reminder III Better data/theory agreement if the waists Z-position or  * (y) are allowed to float in the fit Better data/theory agreement if the waists Z-position or  * (y) are allowed to float in the fit Waists Z positions /  * (y) values seem unlikely ! Are they real ? Which other effect could simulate this lack of focalisation ??   ~2.2 Z [mm]   ~2.4