Mar 26-27/26 PionLT/KaonLT Analysis Meeting Notes ------------------------------------------------- (Notes by GH) Today: KaonLT will be discussed first Thursday: Present ----------------- Regina - Garth Huber, Nathan Heinrich, Nermin Sadoun, Nacer Hamdi, Alicia Postuma, Muhammad Junaid CUA - Sammer Jain, Chi Kin Tam, Tanja Horn Virginia - Richard Trotta Ohio - Julie Roche Richard ------- KaonLT high Q2 LT-sep - investigating small |t| thetaCM behavior - the problem shown last week was that the #events for thetaCM~0 gets large in data but not MC - implemented Gabriel's suggestion, which was to replace thetaCM with sin(thetaCM) in the analysis, but it made no difference - followed the crucial clue, which was that the weird behavior is only in data, not MC. Maybe it's caused by pion leakthrough background events? - modified the background subtraction algorithm for variables other than MM - previously, applied to these variables a background scale factor determined from the integral of the MM background normalization - that method ignores the kinematic correlations between MM and other variables - new method: on an event-by-event basis, determine a background subtraction scale factor to apply to pion leakthrough subtraction events determined from where they are in the MM distribution, rather than an average normalization factor - then fill the other variables (e.g. Q2, W) with this MM-weighted background scale factor - the subtraction then takes into account the correlations between MM and other kinematic variables for background events - fixes the spike at small thetaCM - data vs MC comparisons are also now better for variables other than MM, Q2-dist is better, SHMS/HMS-delta dists have noticeably better comparisons between data and MC - *NB* issue yet to be solved: lower end of W-distibution is weirdly cut-off, looks artificial Chi Kin ------- KaonLT high Q2 syst unc studies - varied acceptance cuts - changes cut regions in low-epsilon distributions - of course, the high-epsilon data have a diamond cut, so changing the acceptance cut has no effect - SIMC yield, Data yield, average kinematics, will all change - still need to calculate new cross sections - Tanja: as long as the MC correctly describes the data, the cross sections should remain stable - varying diamond cuts - looking to decrease polygon size by 2.5%-15% - Garth: this seems too big of a change, the issue primarily is small variations on the edges of the diamond where the #events starts dropping and there is a small ambiguity on where to place the boundary - large diamond changes will affect first and last t-bins much more than the others, given where in the diamond these events come from - *NB* comparing RT and CKT's diamonds is likely to be more instructive - rad corr - last week: showed that Rad-On/Off gave a huge variation - Dave suggested to compare Rad-On/Off for pion case, and use that for the systematic - CKT not sure how to properly propage this to K+ t-phi bins, GH suggests an overall scale factor instead of t-phi dependent - Tanja: suggests to check how this was handled in Carmignotto, Mohrning, Niculescu theses - *NB* Garth: suspect last week's result indicates the MM cut is too narrow, see if you can widen the MM cut to reduce the sensitivity - Garth: for pi+ analysis, there is very little background underneath the neutron peak, so one can unambiguously investigate the rad corr systematic - for K+ analysis, there is some ambiguity in the Lambda tail, as it depends on how much pion background is subtracted. Possibly it makes more sense to quote a combined background-subtraction/RadCorr systematic, rather than separately - K decay correction in SIMC - changed the K+->mu+nu_mu branching ratio in SIMC, made no difference Sameer ------ PionLT Q2=1.6 W=3.08 yield analysis - replayed data for both high and low epsilon - initial PID checks - CoinTime shows pi+ and K/p bands vs MM - *NB* Junaid apply cuts in this specific order: 1) acceptance cut 2) then CoinTime 3) then look at other Detectors in various combinations - *NB* Garth: this is briefly described in the notes following Kathleen's presentation of March 13, suggest to look them up - Tanja: also look at recent Hall C theses to learn what was done there - after cuts, will check CoinTime and RF time offsets Friday: Present --------------- Regina - Garth Huber, Nathan Heinrich, Alicia Postuma, Muhammad Junaid, Nermin Sadoun, Nacer Hamdi CUA - Chi Kin Tam, Tanja Horn Virginia - Richard Trotta JMU - Gabriel Niculescu Glasgow - Kathleen Ramage, Rachel Montgomery FIU - Pete Markowitz Nacer ----- KaonLT Q2=0.5 LT-sep - implementing t-shift based on MM-shift, followed by 1 iteration - sigL: low -t peak is gone, t-dependence is much flatter now, the issue is that the first t-bin is negative by ~1.5sigma - no longer an exponential, the fit parameter in the exponent prefers zero - sigT: rises to compensate - interference terms: basically the same - iterate 2 more times - Data/MC ratios are better than before, generally more flat vs phi - Richard: they haven't shown their results yet, but after t-shift the cross-sections stayed similar to before the shift - Later note by GH: the reduced sensitivity could likely be due to the higher -t values of the high Q2 data and the correspondingly wider t-bin - Nacer will change the lower limit of the 1st t-bin a bit, to avoid the unphysical region near -tmin - *NB* Richard: suggests to plot W,Q2 for each t-bin (summing Left,Right,Center SHMS) and compare to the ave-Q2,ave-W values the model is calculated at - this is how he found the thetaCM problem, the ave-Q2,ave-W values were outside the range of the data distribution, now they agree well - shows plots of -t vs -tmin - generally good agreement between edge of distribution and -t=-tmin line - shows small #events -t<-tmin - Garth: have pion leakthrough been subtracted? It seems likely these are due to mis-identified pions - Nacer: indeed no pion leakthrough is subtracted on these plots - has CoinTime blocking info from Sameer, preparing to replay data to determine the correction factors Alicia ------ u-channel analysis - met w/ Bill on Tuesday for 2hr on shape study - he suggested to not worry on exact values of u-bins right now, concentrate more on the shape study now and fine-tune the u-bins after that is done - seems good advice, concentrating now on getting the MC fits to data to work properly Nathan ------ PionLT LT-sep framework checks - Junaid and Nathan now have exactly matching data yields, moving on to SIMC yield comparisons - SIMC yields agree to better than 1e-6 for one setting - now running other SHMS settings for checks - after that is done, will compare cross sections - identified an issue during Data yield comparisons - Junaid had defined ROC1 as ROC2 by mistake, due to a copy/paste error in python environment setup - once this was fixed, ROC2 needs to be defined as ROC2, things agreed - Junaid has pushed the fixed version to UTIL_PION - *NB* Kathleen, Nermin, Sameer should double check their python environments to make sure they did not inherit Junaid's bug - setting up replay scripts for other settings in x=0.39 scaling scan Junaid ------ PionLT Q2=3.85 W=2.62 LT-sep - applying t-shift based on MM-shift - MM-shifts were in different directions for lo, hi epsilon, t-shifts similarly will be in opposite directions - also some of the MM-shifts were fairly large, ~10 MeV - this is the "worst case" scenario in terms of sensitivity to t-shift - shows plots of -t vs -tmin after t-shift applied - before t-shift, had events -t<-tmin - hi eps: after t-shift, events move away from -t=-tmin line, now events have -t>-tmin - lo eps: before t-shift, there was a small gap between data and -t=-tmin line, after t-shift the gap is gone, events come right to the line - this confirms the respective t-shifts are applied with correct sign - LT-sep after t-shift - very large effect, sigT now has a steep rise in cross section at low -t, while sigL is fairly flat - more distressing: sigL is now negative by ~1.5sigma at low -t - plots of unsep cross sections vs phi also indicate this, the gap between high and low epsilon is much smaller after t-shift - Data/MC ratios do not look good, particularly at low -t, where they are 2 and higher, this at least tells us that the data are not consistent with this new behavior - for a test, applies t-shifts with same sign for hi,lo epsilon, rather than opposite sign. sigL is much better behaved - *NB* Garth: are we sure the MM offsets are correct? We should take a closer look at them given this info, maybe we can reduce some of the MM-shifts - *NB* Nacer: suggests to change the low -t bin lower limit, to exclude events very close to -tmin Kathleen -------- LD+ yield for Q2=1.6 W=3.08 6.4GeV (2022) data - last week: looked at detector cuts in more detail - Random Subtraction: selecting 3 peaks on either side of prompt, with a gap of 2 away from prompt peak (i.e. peaks 3-5 on either side) - Garth: some evidence that the 5th peak is a bit smaller than peaks 2,3,4 on both sides. Possibly they are getting close to the edge of the coincidence timing window. Suggest to use peaks 2-4 rather than 3-5, as their heights are more uniform - Dummy Target Subtraction: - calculated target thickness factor f=4.009 from target group report - normalized by total beam charge as a quick test - *NB* Nacer: need to form Qeff from the detector efficiencies and livetimes - the relevant information should be in the report files - MM plot after Random and Dummy subtractions - Nacer: to minimize sensitivity to the MM cut, put the cut in the plateau between the neutron peak and Delta-rise - *NB* Alicia: your next step should be to compare the MM peak to SIMC, to see how well they agree, that will also guide where to place the MM cut Next steps: - calculate Qeff from detector efficiencies, livetimes - will clean up macros used to make plots - will present a poster at IOP conference, will distribute a draft for comments Nermin ------ PionLT 9.117GeV Q2=1.6 W=3.08 SHMS=6deg detector efficiencies - SHMS Aerogel: - Last week: LD+ aerogel eff was 97% - added a HGC cut to the aerogel efficiency calculation to reduce K/p contamination, now the efficiency is 99% - new LD- detector efficiencies - using cuts on SHMS calorimeter, acceptance, CoinTime, obtained aerogel efficiency of 98% - after adding HGC, aerogel eff improved to 99% - HMS Cherenkov: - coin efficiency is lower for LD- than LD+ - this means HMS calorimeter is not removing all pi- from the electron sample used to calculate the efficiency - tightened the calorimeter cut, efficiency improved - HMS Calorimeter: - applied acceptance, CoinTime, Cherenkov cuts, got eff=0.978 +/- 1e-4 - tightened the Cherenkov cut, nothing improved - Garth: the issue when calculating the HMS calorimeter efficiency is that it is very difficult to remove all of the pi- contamination in the electron sample - Ali ended up using dedicated Heep runs with minimal pi- contamination in his HMS calorimeter efficiency study, and then apply this efficiency to the physics data - *NB* Junaid: he has done the HMS calorimeter efficiency study for HMS momenta from 1-4 GeV/c. Please check the HMS momentum for your setting, and if it is within this range then you can use his efficiencies (and uncertainties) directly - Junaid will send his slides on this Next steps: - Lumi study for LD2 cryotarget Next Meetings ------------------ - Thurs: Apr 2 @ 16:00 Eastern/14:00 Regina - PionLT will go first - Fri: Apr 3 @ 11:00 Eastern/9:00 Regina - NO MEETING DUE TO EASTER (GOOD FRIDAY) HOLIDAY