May 14-15/26 PionLT/KaonLT Analysis Meeting Notes ------------------------------------------------- (Notes by GH) Today: PionLT will be discussed first Thursday: Present ----------------- Regina - Garth Huber, Nathan Heinrich, Alicia Postuma, Dex Yadlowski, Muhammad Junaid, Nacer Hamdi, Nermin Sadoun JLab - Dave Gaskell FIU - Pete Markowitz Virginia - Richard Trotta CUA - Chi Kin Tam, Sameer Jain, Tanja Horn Glasgow - Kathleen Ramage, Rachel Montgomery Ohio - Julie Roche Nathan ------ PionLT Q2=5.0 LT-sep - tracking down a bug in t-binning scrhipt - mid epsilon data doesn't show up in plots - an issue with too many dummy target events Nermin ------ PionLT LD2 Luminosity analysis - investigating current cuts, to ensure a stable rate sub-sample for Lumi study - there was a run w/ many beam fluctuations, after applying the cuts, the trend of the rate data is more stable - now looking again at the other runs - has one run @ 80uA, with high DAQ rate - Nathan: run with DAQ rate of 2.5-3 kHz should give believable results, even though the livetime might be lower than at lower rate Nacer ----- KaonLT Q2=0.5 LT-sep - presented "final" cross sections - includes all normalization factors, including CoinTime Blocking - improved functional forms - KLambda: - L = p0*|t|*exp(-p1*|t|)/(|t|+mK**2)**2 T = p2*exp(-|p3*t|) LT= p4*exp(-p5*|t|)*sin(theta) TT= p6*exp(-p7*|t|)*sin(theta)**2 Wfac=1/(W**2-mp**2)**2 - Rosenbluth "money" plots have high eps slightly above low eps data - L shows a slow increase w/t, from 0.15 -> 0.30 T has a monotonic decrease w/t, from 0.9 -> 0.6 LT is negative, near zero at lowest -t and monotonically increasing - Data/MC ratios look pretty flat low eps: particularly flat at low -t, a bit worse at higher -t high eps: ratios really look good, flat and near 1 - Kinematic distribution plots SIMC reproduces data pretty well, except for SHMS_xpfp, where SIMC is narrower than data - KSigma0: - use same L/T/LT/TT functional forms - Rosenbluth "money" plots: high and low epsilon data nearly equal - L ~0 T flat w/t, ~0.25 LT flat w/t, -0.04 TT is negative, trending away from 0 at higher -t - Data/MC ratios at both high, low epsilon ~1 with larger errors than Lambda and random fluctuations - Kinematic distribution plots: SIMC reproduces data shape well, except for SHMS_xpfp - Dave: regarding the kinematic distribution plots for Sigma0, for making nice plots to show at conferences suggests to use coarser binning to remove some of the fluctuations Next steps - will look next at Sigma0/Lambda ratios for overlapping t-bins - will start systematic uncertainty studies Chi Kin ------- KaonLT high Q2 LT-sep - looked again at the Out-of-plane offssets from Heep data - followed same method as Garth - shows plots of Pmy/Pp vs Pe'/Pp for all data from 3.8 to 10.6 GeV - error bars look quite large compared to fluctuations of data - was using the width of the Pmy distribution, rather than the fit uncertainty in the mean - if using the mean fit uncertainty, the error bars will look too small - alternate plot (SIMC_pmy-Data_pmy)/Pp vs gives slightly diffferent results - Dave: if using scipy fit package, need to set absoute_sigma=TRUE to get accurate errors - CKT checks script, was using iMinuit, so that should not be the problem - some questions about what offsets to use in SIMC input file - Dave: Please see https://hallcweb.jlab.org/hclog/0307_archive/030713122338.html - Henk Blok did a calc of what happens at FP if there is a vertical beam offset and how it affects the reconstruction. - calculation uses the full forward matrix, not the sequential matrix - for particle with initial X0 that is offset vertically, the X' reconstructed is offset by X' = -1.14*X0 + 1.00*X'_0 - for SHMS, we only have the sequential forward ME, not the full forward matrix, so Henk's calculation cannot be applied directly *NB* Dave will see if he can find something that can be used for SHMS - central_theta_offset in SIMC needs to have opposite sign, you have to think carefully of how the offset is defined before applying it Richard ------- KaonLT high Q2 LT-sep - now using Kin's diamond cuts, as reported last week, there was an issue w/ left setting diamond being too small - this issue has been fixed, will end up with better statistics when data replayed - now splitting background fits into 2 MM regions (example numbers only, exact numbers vary with setting): - Region 1: 1.05-1.15 - Region 2: 1.10-1.20 - First fit Region 1, and subtract it from data - then fit subtracted data for Region 2 - then will compare resulting radiative tail to SIMC (stil in progress) - example shown for Q2=4.4, W=2.74 - high epsilon right SHMS setting has events to left of Lambda peak, other settings (center, left, low epsilon) look much cleaner - Kinematic distributions at low epsilon comparo between Data and SIMC look okay - playing around with functional form for LT LT= p9*exp(-p10*|t|)*sin(theta)**2 - sin(theta) instead of sin(theta)**2 results in a bit worse Data/MC ratios - once background fit procedures are finalized, will run on other settings Friday: Present --------------- Regina - Garth Huber, Nathan Heinrich, Alicia Postuma, Dex Yadlowski, Nacer Hamdi, Nermin Sadoun, Vijay Kumar York - Stephen Kay JLab - Dave Gaskell JMU - Gabriel Niculescu, Ioana Niculescu CUA - Sameer Jain, Chi Kin Tam Glasgow - Kathleen Ramage, Rachel Montgomery Kathleen -------- PionLT LD2 Luminosity analysis - read over Nathan's luminosity report, still making sure the analysis scripts work - looking at 6.4GeV (singles), 7.9GeV (coin) - shows one run for debugging purposes: Run 16715 LD2 target - checked PID cuts - next step is to run script for all runs - Nathan: please first check at least one Carbon setting, to make sure you get the same result as Nermin - if you do, the code works and you can go on to LD2 Sameer ------ PionLT Q2=1.6 LT-sep - shows diamond cut plots for low epsilon Left, Center - Nathan: the diamond cuts look too small, please make bigger at top and bottom of diamond - upper left corner needs to be adjusted inward a bit - t-resolution study: high and low epsilon - the t-range of this setting is very different than Junaid's, need to make a lot of modification - Garth: judging by eye, you can try a range like 0.03<-t<0.25 - 16 phi bins Vijay ----- PionLT Low Q2 LT-sep - working on radiative corrections systematic - following same method outlined in Blok et al. paper - this involves studying the Data/MC ratio as a function of MM cut at high and low epsilon, if SIMC correctly models the MM tail of the data, the ratio should be insensitive to the exact placing of the upper MM cut - scanned upper MM cuts from 0.97 to 0.99GeV - variation of Data/MC ratio vs cut is different for low, med, high epsilon - outer range of variation is about +/-1.5% - *NB* will look at the variation separately for each t-bin (integrating over phi). Alicia notes that in BSA analysis (different kinematics) the Radiative Tail description was t-dependent - changed functional form used for sigL - now using exactly the same form as Born Term Model equation, including substitution of t-dependence in g_piNN form factor L ~ (-t)/(t-mpi**2)/(Lgpinn**2-t**2) where Lgpinn=0.44 - used equation has 2 free parameters T has a linear mx+b form - the Q2=0.375 results with this form have good Data/MC ratios - L has well-defined turn-over in lowest t-bins T has monotonic decrease w/ -t Alicia ------ KaonLT u-channel analysis - Aerogel proton-knockon correction - following method described in Bill's thesis - combined KaonLT Heep, PionLT Heep and KaonLT physics settings to get a wider range in momentum - data from 3 different aerogel trays are used - assumes the interaction probability scales with linear density aerogel density is proportional to (n-1) - applying this correction to rescale all data to same n gives a more linear trend vs SHMS_momentum than w/o correction - error analysis for the proton-knockon correction - use Binomial Errors since Ndid and Nshould are highly correlated - however, Binomial Errors assume no error on Nshould, so instead take delta-Nshould=sqrt(Nshould), gives a slightly larger error than Binomial, but still smaller than purely uncorrelated errors - resulting chi-square of data is still too large, implying the errors are underestimated - added an ad-hoc correction to give chi-square=1 NewErr =sqrt(Err_binomial**2 + adhoc**2) - the adhoc correction is presumably due to a systematic in the data, but not sure what this might be due to - NewErrs for all data are much more similar now, while before some points had much smaller errors than others - resulting correction: Q2=3.0, 3.14: knockon_corr=83.6% +/- 1.6% 3.0, 2.32: 94.9% +/- 1.3% - Nathan: suggests to compare "Garth method": cyclicly remove point-i and check the variation in the slope (i.e. knockon correction), to see if it gives similar results to your error-rescaling method - Dave: a likely source of this systematic uncertainty comes from the fact that the HeepCoin data does not fully fill the SHMS focal plane. There are aerogel density variations from spot-to-spot across the focal plane, and this should give some additional variation beyond the statistical uncertainties *NB* suggests to compare X,Y_Aerogel for the different HeepCoin settings to see how much the illuminated focal position varies by Next Meetings ------------- - Thurs: May 21 @ 13:00 Eastern/11:00 Regina - KaonLT will go first - Fri: May 22 @ 12:00 Eastern/10:00 Regina - we will continue where we left off PS: Gabriel and Ioana would like some time next Friday for their presentation and discussion, so people are requested to have their slides ready on Thursday if at all possible