Kaon_LT Experiment

                 July 31/25 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

Present

-------

Regina - Garth Huber, Alicia Postuma, Nathan Heinrich, Ivan Zhenchuk,

   Nacer Hamdi

Virginia - Richard Trotta

JMU - Gabriel Niculescu

JLab - Dave Gaskell

CUA - Tanja Horn

FIU - Pete Markowitz

Nathan

------

PionLT Lumi studies

- now has a set of timing cuts that work well

  - no weird shifts in CT across HMS or SHMS focal planes

  - 2ns beam structure resolution in CT plot is noticeably improved

  - working on implementing these cuts in rest of PionLT replay, and will

    notify Junaid of changes

- high rate and low rate versions of CTime.CoinTime_RAW_ROC2 and

  CTime.ePiCoinTime_ROC2

  - weird parts of distribution at far left and right are now gone.  Most

    likely this is due to correct channels for timing cuts being used now

  - the plots now have shoulders extending from -200 to +300, noticeably

    narrower than before

  - runs at 3 rates show the height of these shoulders is rate dependent, and

    the shoulder width is constant

  - running jobs for all Lumi settings, hopefully the Lumi studies will now

    have a reduced rate dependence with proper timing cuts being used

  - then can use these plots to determine a coincidence blocking correction

    that will hopefully resolve any remaining rate dependence

Richard

-------

KaonLT LT-sep

- still looking at HGC cuts and background fits

- expanded range of MM plots

  - it is now clear that there is significant proton leakthrough for left SHMS

    settings at both high and low epsilon

  - *NB* Dave: Peter Bosted is using an aerogel NPE>4 cut for pion ID with

    n=1.015 due to the large proton background

    - Richard is using NPE>1.6 cut, suggest to investigate a higher cut

  - the proton leakthrough explains why the pi+ leakthrough subtraction does

    not properly get rid of the background underneath the Lambda peak

  - some discussion on background removal strategies

    - Tanja and Dave suggest fitting a polynomial

    - longer term, Alicia's Pythia studies suggest an alternative way of

      fitting the background

Alicia

------

u-channel: feeding Pythia events into SIMC

- need to specify electron @ HMS central kinematics, protons are then generated

  over a wide range

  - while Pythia is slow, the good news is that get all 3 SHMS settings from a

    single Python file

- ypfp was not calculated properly last week due to a variable typo, now the

  variables are correctly centered on the spectrometer acceptance

  - now getting reasonable Pythia results, but the efficiency of passing events

    through SIMC is low

- found a workaround for screen terminal issue last week

  - now using tmux, ran 1B Pythia -> 1.2M pass energy cut -> 80k SIMC evts

    takes 1-2 days to run this #events

- comparison of Pythia result to data MMp

  - Q2=3.0, W=3.14, center SHMS

  - Pythia has a very good resemblance of the higher MM region to right of

    omega peak

  - Pythia seems to reproduce the changing features of the data with L,R,C settings

  - there is some discrepancy in the MM region, where Pythia drops off but the

    data extend a bit further

- change to MC fitting procedure to data

  - choose MM regions for different MC fits, sum data and MCs for each region,

    and do a simultaneous fit of all regions

  - this provides scale factors for each MC to data

  - some evidence of over subtraction to right of omega, MM~1.0

- plots of: Bkd subtracted data comparison to omega MC

  - sum of MCs vs data

  - difference of all MC subtractions, giving residuals vs MM

- plots for SHMS center for 8 phi bins with new algorithm

- running Pythia for other settings now

Gabriel: this all looks very good, but we're putting a lot of faith in Pythia

  getting the correct bkd shape

  - Pythia documentation claims it's less reliable for W<10 GeV

  - while most of the MM shape seems to be phase space, different cross section

    models within Pythia could give a slightly different shape

    - *NB* does Pythia have any knobs, which allow us to vary the shape

  - how do we evaluate systematics for this?

- Alicia plans to look at the Data-MC difference plots for different scaling

  factors and use this as a guide to the sysematics studies

- Dave: maybe compare to polynomial fit we discussed for Richard to evaluate

  systematics

Junaid

------

PionLT LT-sep

- gave slides for Garth to show, as recovering from Owl Shift now

- met with Dave and discovered an error in the weight recalculation script

  - was using some reconstructed variables to calculate weights instead of

    vertex values

  - modified Ntupleinit.f and results_write.f in SIMC to add epsiloni, phicm

    and thetacm to ROOT tree

  - comparison of weight script and SIMC shows they're getting the same weights

    - *NB* Garth will want to see some more information on the weight script

      checks

- working on first iterations for Q2=3.85, W=2.62

Nacer

-----

KaonLT Q2=0.5 LT-sep

- incorporated weight script and SIMC changes from Junaid

- using Tanja's Fpi-2 functional form and initial parameters

  - Lambda: Data/MC ratios 0.5-0.2, flat vs phi at low -t, some phi-dep in

    ratios at higher -t

  - did simultaneous Rosenbluth fit of L/T/LT/TT

  - then fit function params to L/T/LT/TT vs t in unconstrained fit

  - Dave: the sigL fit looks odd, clearly the functional form is too

    complicated for these data

    - Nacer will investigate this after successfully doing a few iterations

  - weight script check: compare SIMC output w/ initial params vs weight

    script, they appear to be consistent

- then feed the new parameters into weight script for first iteration

  - weight script crashes, gets weights up to 10^(36)

  - tried fixing params 5,6 in sigL to initial values but keep other params to

    new values and refit

  - no longer get a crash, but now get ratios up to 30, similar issue to what

    Junaid had last week

    - Dave and Tanja strongly suspect a common software bug between Nacer and Junaid

  - the fit of new sigL parameters did not converge, fit failed so the

    parameters that are being used for sigL are meaningless

  - *NB* it's extremely important to be sure fits succeed.  You should continue

    the next step in the iteration with an unconverged parameter, instead, stop

    and fix the issue before continuing to avoid meaningless results!

  - *NB* Gabriel: for debugging purposes, suggest to change a few parameters by

    hand by a small amount and confirm that the results only change by a small

    amount.  If the change is larger than expected, it indicates where the bug

    might be

  - *NB* Garth: this is a good suggestion.  In addition, you could try

    estimating the effect of each parameter by hand (using central kinematic

    values) fixing to small values the ones that are responsible for the

    10^(36) weight, and allowing the others to float.

Ivan

----

KaonLT BSA

- just finished setting up for CoinTime and RF cut systematic studies

  - has a software framework for them, will use scripts to investigate

    variations in BSA

  - working on Co-Op report outline

Next Meeting

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

- Thur Aug 7 @ 15:30 Eastern/13:30 Regina

  - KaonLT will go first

  - Garth cannot attend (LRPC meeting at TRIUMF), Tanja will take notes