Nov 6-7/25 PionLT/KaonLT Analysis Meeting Notes ----------------------------------------------- (Notes by GH) Today: PionLT will be discussed first Please remember to post your slides at: https://redmine.jlab.org/projects/kltexp/wiki/Kaon_LT_Meetings Thursday: Present ----------------- Regina - Garth Huber, Nathan Heinrich, Muhammad Junaid, Alicia Postuma, Nermin Sadoun FIU - Pete Markowitz Virginia - Richard Trotta CUA - Tanja Horn, Chi Kin Tam JLab - Dave Gaskell CSULA - Konrad Aniol Junaid ------ PionLT Q2=3.85, W=2.62 LT-sep - changes made to LT, TT functional forms - main difference, removed p/|t| type term and changed exp(p1/t)*p2/|t|^2 to p1/|t|^p3 form - still see a phi-oscillation at low epsilon, but R=Data/MC ratios better at high epsilon: fairly flat with R~1 - Pete: please recheck your error calculations on the ratios, they seem over-estimated in comparison to the fit, the chi-square is too good - W,Q2 distribution comparisons between Data and MC - differences between Data & MC got better with more iterations up to a point, then stabilize. Are the comparisons good enough now? - shifts in the W,Q2 plots are 10-20 MeV - the applied MM offset is ~10 MeV, W,Q2 are not recalculated with the offset, so it's not a surprise that the shift is of this magnitude - seems unlikely that it would get much better without improving the experimental offsets - some discussion about implementing the wandering beam energy in the analysis, as suggested by Gabriel last week - Dave: hcana does not use HallC:p to correct the beam energy, only the value in standard.kinematics is used - to implement it, one would need to know not only the value of HallC:p during the run, but also its value at the time the arc energy measurement was done. Then one could correct for beam variations from the ratio - in his investigations of SIDIS data, the effect was not large enough to be worthwhile to correct - *NB* Garth suggests to first make plots of HallC:p to see if it is a worry or not. Particularly for the KaonLT 10.6 GeV data, the accelerator was in very rough shape, and he thought the beam energy wandering was larger than usual, so it might make a difference there - Richard will do some investigations and report back Next steps: - putting together a report on the LT-separations for this setting - setting up analysis of Q2=3.85, W=2.02 setting Nathan ------ PionLT CoinLumi analysis - completed a first draft of technical report and received comments back from Garth that he's implementing - will send a copy of the draft to Sameer, any questions he has will help to further improve the report - expects to circulate report to everyone else early next week Chi Kin ------- KaonLT Q2=3.0, W=3.14 analysis - compared Counts/mC/DeltaMM with Richard - his counts were before pion subtraction, while Richards include subtraction, as expected CK results are slightly higher - then apply pion subtraction, agreement with RT is better but some outliers - CK applying the full pi+ subtraction as determined from neutron peak fit - RT only applies ~70% of the pi+ subtraction and does a polynomial fit for the rest of the background - *NB* MM plots after pi+ subtraction show some evidence of over-subtraction at low t, higher -t looks better. Needs to make refinements - also using different t-binning than Richard, maybe the lowest -t bin needs some adjustment to get more counts - *NB* a table comparing the cut limits and #counts between CK and RT would be helpful for our review - SIMC work - tries Richard's old functional form in SIMC, then tests code with a slightly different set of parameters to see how much things change - wants to see if his code gives similar sig_uns to Richard - respects positivity of sigL,T with constraint sigLT=rhoLT*sqrt(sigL*sigT) sigTT=rhoTT*sigT - Rosenbluth fit doesn't always give sigL,T>>sigLT,TT - *NB* Garth: maybe try setting LT=TT=0 and concentrate on sigT - next steps for pi+ subtraction - will modify Sigma0 peak fit and subtract it only at higher -t, since it does not contribute at low -t (due to higher -t_min value) - Garth: you might need a different background fit strategy at low -t (no Sigma0 peak) and high -t (both Lambda and Sigma0 peaks) - question on whether need to throw away same bin at high epsilon if stats poor in low epsilon bin - Answer: No, the SHMS settings coverage is different at high and low epsilon, their phi coverages are not the same. - what is important is to throw away the same bin for Data and MC, but to keep as many good bins as possible at high epsilon Friday: Present --------------- Regina - Garth Huber, Nathan Heinrich, Muhammad Junaid, Vijay Kumar, Nermin Sadoun, Nacer Hamdi, Alicia Postuma JLab - Dave Gaskell York - Stephen Kay FIU - Pete Markowitz Ohio - Julie Roche Glasgow - Kathleen Ramage, Rachel Montgomery CUA - Chi Kin Tam, Tanja Horn, Sameer Jain Virginia - Richard Trotta JMU - Ioana Niculescu, Gabriel Niculescu Alicia ------ KaonLT u-channel low epsilon first look - Q2=3.0, W=3.14, SHMS center, low epsilon - preliminary PID cuts only - will raise the Aero NPE cut - RF time available for PID, need to adjust the cut - pi+ leak-through greatly suppressed by the PID cuts - see small eta, eta' peaks - pronounced pi0/DVCS blob at low MM - Tanja: how do you plan to disentangle pi0 from DVCS with no dedicated detector? - no guarantees we can do this, will look at MM^2, and compare to SIMC for both processes - even a conservative upper limit would be useful, and could strengthen the case for the dedicated u-channel pi0 experiment - MM shape study - nice agreement between omega Data and MC shapes, MM resolution is pretty good - need to run more statistics for Pythia simulation - Q2=2.1, W=2.95, pythia dist does a poor job at higher MM - Q2=4.4, W=2.74 - no MM offset was needed for low epsilon data, only high epsilon - Richard agrees that the MM offsets were smaller for low epsilon - Nacer: curious if there is a small peak at MM~1.15 GeV? - so far, thinks it's just a statistical fluctuation, will need to see what happens with more analysis - Garth: the closest PDG candidate would be the h1(1170), with width=360 MeV, so not likely - Q2=3.0, W=3.14, SHMS center, low epsilon - shows MM plots for 8 phi-bins - recognizable omega peaks in all phi-bins for SHMS center - made a table of omega counts/bin for all settings - Q2=5.5 setting has very poor low epsilon statistics - Bill had a cutoff of 70 counts per t-phi bin for a valid MM fit - probably will need a ~100 count threshold here, since the MM range is wider Richard ------- KaonLT high Q2 LT-sep - booked at beam energy wandering correction mentioned by Gabriel last week, will show something in a future meeting - Q2=4.4, W=2.74 - SIMC now using proper vertex variables as discussed at a prior meeting - functional forms - now using sin^2(theta) for both LT,TT - the sin^2(theta) instead of sin(theta) for LT seemed to fix some issues - also Wfac now using Wsetting instead of event-by-event value - Data/MC ratios now 0.7-1.5, centered at 1, reasonably flat - Nacer: are your fit parameters determined in a constrained fit? - Richard has an algorithm that hunts over a wide parameter range and checks that it's not a false local minimum - Gabriel: has looked at Richard's code. His understanding is that there is no fitting theory that works when the dependence on fitting parameters is highly nonlinear. Richard's algorithm tries to address this - sig_uns look a bit better now, high epsilon marginally higher than low epsilon, reasonably flat phi-dependence for IT=3 - however, things change x2 from IT=3 to IT=9, and the phi-dependence becomes much more pronounced - some outliers at edge of distributions, phi~0, 360deg, need investigation - also some huge changes in error bars with iteration that seems strange. Richard thinks it's caused when hitting the positivity constraint sigLT=rhoLT*sqrt(sigL*sigT) sigTT=rhoTT*sigT - will try more iterations, but believe the functional forms are OK, since the kinematic comparison plots between Data and MC look pretty good Nacer ----- KaonLT Q2=0.5 model optimization - using sig2007 inspired functions, but with non-zero LT,TT of the form p1*exp(-p2*(t-tmin))*sin(theta)/(Q2+p3)^2 Wfac=pK/(sqrt(W2)*(W2-mp*2)) where pK is the K+ momentum - after one iteration, low and high epsilon Data/MC ratios reasonably flat R~0.7 - sigL slightly negative, about 1sigma below zero - then tried changing sigT to p2/(Q2+t^p3)^2 form - Data/MC ratios similar, fairly flat ~0.8 - after more iterations, gets single bump in ratio around phi~180deg in - highest -t bin - Junaid: a single bump means LT is off - will try adjusting the functional form - Garth: suggests to fix LT,TT parameters to values for an iteration where ratio was fairly flat and concentrate just on L,T - some discussion on when to consider changing the parameterization - checks the change in parameters and yields after 4 iterations, if reasonably stable then it's time to change the functional form - this seems a sensible approach - Sigma0 LT-sep - using same functional form as Lambda - 1st iteration gives R=Data/MC ratio near 1, except at edges of distribution (phi~0, 360deg) at higher -t, where R~4 for low epsilon - high epsilon ratios look much nicer, R~1 - then tried fixing LT, TT parameters, set L=0, fit only T - ratios generally near 1 at low epsilon, R~0.8 at high epsilon - then allowed both LT, TT to vary, while L=0 - big bump in ratios near phi~180deg - *NB* Garth: TT cross section looks much larger than LT, so to reduce the number of free parameters in the fit, keep LT fixed and only vary T, TT Vijay ----- PionLT Q2=0.375, 0.425 CoinTime blocking correction - found some errors in last week's calculation, which removed some outliers - Q2=0.375 now has correction 0.9 to 1.0 - results now also for Q2=0.425, looks similar to other Q2 vs CoinRate - a few outliers at 1kHz correspond to Left SHMS setting, will look a bit more into this - calculating uncertainties from variation in CoinTime window Sameer ------ KaonLT CoinTime blocking correction - changed TDC time window to get rid of weird peaks at -300 (ROC1) old window: 4200-6200 new window: 5000-6000 - this did not get rid of the weird peaks - also changed windows for pTRIG1,4 (SHMS,HMS) but didn't have an effect, which seems a bit strange as this worked for Nathan - ROC2 does not have these peaks - *NB* Nathan: ROC1,2 are equivalent, it's equally valid to use either one, so you should use the one that's cleaner - using ROC1, gets corrections >0.9 - *NB* 3 low outliers from valid physics runs at Q2=2.1 that needs more investigation - Nathan: suggest to calculate also the error bars, first the statistical uncertainty, then the cut-dependent systematic uncertainty - see his draft report for more info Next Week Meetings ------------------ - Thurs: Nov 13 @ 16:00 Eastern/15:00 Regina - KaonLT will go first **NB* Richard will take notes, GH will be at Ali's Convocation - Fri: Nov 14 @ 11:00 Eastern/10:00 Regina - we will continue where we left off, GH will take notes