Kaon_LT Experiment

                 Jan 25/24 PionLT/KaonLT Analysis Meeting Notes

                 ----------------------------------------------

                          (Notes by GH & Richard Trotta)

                     Today: PionLT will be discussed first

Please remember to post your slides at:

https://redmine.jlab.org/projects/kltexp/wiki/Kaon_LT_Meetings

Present

-------

Regina - Ali Usman, Muhammad Junaid, Vijay Kumar, Nacer Hamdi, Garth Huber

JLab - Dave Gaskell

CSULA - Konrad Aniol

FIU - Pete Markowitz

CUA - Tanja Horn, Richard Trotta

Junaid

------

Efficiency studies

- CSV files of eff values created for all data

- working on scripts for lumi, heep, and finalizing plotting scripts

- Plotting scripts for efficiencies read in csv of efficiencies and plot from

  there

- updating REPORT FILES but Heep efficiencies are calculating incorrectly.  The

  issue is the error bar of efficiencies, not the efficiencies themselves.

  - Ali: there might be a bug in the TEMPLATE file, such as the error formula

    somehow dividing by zero.  If efficiencies themselves are okay, then DEF

    files ok.

Richard

-------

Issues replaying Dummy runs

- dtype error

  - python cut script gave error for some runs and not processed

  - fixed code to catch error when there is an empty ROOT tree.  Code now says

    which trees are empty

- CT cut issue

  - met w/Stephen on Jan 24

  - CT offset was not applied to Dummy runs.  SK missed a few Dummy runs in

    standard.kinematics

  - Vijay found a similar issue for some runs

Weird features in some t-distributions

- some (e,e'K+) runs had a spiky t-distrubtion

  - getting events lower than the -t_min for K+ in that setting, which is a

    sudden jump in distribution from the allowed region

  - most likely due to insufficient PID cut, both pi+ and proton leakthrough,

    some evidence of both pi+n and p+omega peaks

  - after tightening HGC cut, the t-distribution looks better

  - seems to have evidence of Lambda(1405) and other resonances in Q2=2.115

    center, low epsilon data

    - this is pleasant news, in the online analysis only the high epsilon data

      (w/o diamond cut) showed strong evidence of the higher hyperon states

- GH: suggests to generate *clean* pi+ and proton spectra, normalize them to the

  pi+n and p+omega peaks and *subtract* them to get an even cleaner K+

  distribution

  - absolutely must be finalized before starting iterations!

  - also critical to determine the correct PID efficiencies to apply to data

Fixed binning issue

- long-standing binning issue, some bins were always empty when binning over

  Q2-bar, W-bar, t, epsilon

- rewrote both binning scripts for average kinematics and yields

- now looks much better

Separated cross-section script

- scripts now working!

- now can debug things properly with average kinematics and yields

  - some issues were encountered in not treating t-bins in a consistent

    sequence

- now can generate Iteration-0 plots, then can try to do the next iteration

  - process in general works, but need to resolve PID issues first

- DG: your high epsilon data show a weird left-right phi asymmetry that is

  unphysical

  - need to look more closely at what is causing this

SIMC MM weirdness

- plot shown for Q2=3.0, W=3.14, low-eps, center

  - K+L MM distribution has a bizarrely long tail that is much bigger than the

    data, exactly the opposite situation than was encountered by Alicia, where

    the pi+n MM tail is under-predicted by SIMC

- GH: maybe it's a radiative tail flag that is wrong??

SIMC negative counts for some kinematics at highest Q2

- for Q2=5.5, the Iteration-0 script gives negative weight events

  - corresponds to sigT<0 for some kinematics

  - the initial fit parameters in the SIMC model are completely inappropriate

    for this Q2, and need adjustment

  - will set to zero initially

  - GH: if this doesn't work, then find by hand some reasonable fit parameter

    values to try initially

Ali

---

- N-Delta Transition GPDs white paper contribution for ECT* draft was sent to

  collaborators and when feedback is returned he will send to everyone

- Optimizing BSA code, looking at left and right settings

  - Q2=2.115, not every bin has enough statistics

  - If Events<Threshold cut applied, sets these bins to zero

    - the issue is that the SIDIS+rho fits failed if not enough events

    - GH: Bill had to do something similar in the omega analysis, as the

      rho+PhaseSpace fits would fail if not enough statistics, or if radiation

      corrections were too high.  This is all explained in his thesis if you

      want

  - weighted average of everything should cover empty bins for off-center

    settings

  - Made changes to weighted average scripts, should have plots for all

    settings combined soon

Dave

----

hsxpfp vs delta correction for high HMS momenta

- looked at a subset of runs provided by Ali

- applied correction: delta=delta+C0*hsxpfp

  - correction factor C0<0

- plot of C0 vs P_HMS shows a monotonic decrease to about 5 GeV/c, then

  is reduced in magnitude from 5 to 6.5 GeV/c

  - likely the reversal in trend is due to partial HMS saturation correction

    applied to Matrix Elements

- 6.59 GeV/c correlation is not totally linear, did a best fit linear

  correction, but some slight curvature remains

- Will provide table of numerical C0 values

- some discussion on how to implement this in hcana

  - if recalculate hcana values on the fly, then just recalc delta with above

    correction, and not make any modifications to hcana

  - Richard will implement into ltsep python framework to recalculate delta

  - fortunately, this does not affect the PionLT data, as the effect

    disappeared when the power supply was replaced

- Final step will to be to optimize SIMC resolution for pi+n to get better

  agreement between data and MC

Vijay

-----

Busy with making talk for conference in India

- EIC workshop in Bangalore

- Will send around draft slides in next few days

- Leaving for India on Feb 2.  Talk is Feb 9

Nacer

-----

SHMS DC calibration

- resolved all bugs for KaonLT calib w/Junaid

- new t0 offsets needed

- Need to compare parameterization every few runs

  - systematically checking groups of runs.  6638-6691 shown

- Pre-calibration X-T and new T0 (red) is usually better than post-calibration X-T

  - this is because it over-calibrates (i.e. over-fits the wiggles in the data)

    in particular when there are low stats

- Basically only need to calibrate t0 and that's it. This is just tedious

  because need to check how it varies 

  - set up to check once per 8hr shift of data

  - ~115 runs still to go, should be done around mid to late February

  - batch jobs failing and such take a while

Next Meeting

------------

- Thur Feb 1 @ 15:00 Eastern/14:00 Regina

  - KaonLT will go first