Kaon_LT Experiment

                Feb 26-27/26 PionLT/KaonLT Analysis Meeting Notes

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

                                (Notes by GH)

                    Today: KaonLT will be discussed first

Thursday: Present

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

Regina - Garth Huber, Alicia Postuma, Nacer Hamdi, Muhammad Junaid,

   Nermin Sadoun

FIU - Pete Markowitz

Ohio - Julie Roche

Virginia - Richard Trotta

JLab - Dave Gaskell

CUA - Sameer Jain, Chi Kin Tam, Tanja Horn

CSULA - Konrad Aniol

Nacer

-----

KaonLT Q2=0.5 LT-sep

- Lambda: changed t-binning to fill in gap between lowest 2 t-bins

  - divided 1st t-bin in half, now 9 bins:

    -t=0.075, 0.082, 0.087, 0.093, 0.099, 0.105, 0.112, 0.122, 0.140

  - Data/MC ratios look good after 7 iterations

  - 2nd t-bin sigL slightly higher than 1st bin by ~1 sigma, first t-bin went

    down a bit

  - Richard: notes that sigT has pole term formula in it, while sigL does not

    - Nacer will try with the pole term in sigL

- Sigma0: results not shown last week, Data/MC ratios look good

  - kinematic plot comparisons look good too

  - sigL essentially zero over whole t-range

- will also rerun w/o MM shift to see how big of a difference it makes

  - Garth will look into kinematics equation, to find implied t-shift when MM

    is shifted

Junaid

------

PionLT Q2=3.85, W=2.62

- also added another low t-bin, divided first 2 bins into 3 bins

  - the new point is mid-way between bins 1,3

  - Garth: the gap between the 2 highest t-bins is quite large, suggests to

    also divide last 2 t-bins into 3 bins, so all bins have roughly equal

    spacing

  - Data/MC ratios look better for high epsilon than low epsilon

- planning to start systematic studies next week

Richard

-------

KaonLT high Q2 yield comparisons with Chi Kin

- CKT and RT met at JLab and figured out the SIMC discrepancy, yields now match

  - 2 issues were found:

    - cut mismatch, CKT hadn't implemented an Aerogel tray geometry cut

    - miscommunication on which files were used

- effect of various cuts on SIMC yields:

  - MM + t: ~20% decrease from no cut

    + acceptance: ~50%

    + HGC hole cut: ~65%

    + diamond cut (low epsilon): ~80%

- plans for full replay of KaonLT data

  - will start w/ best setting (Q2=3.0, W=3.14) and work down

  - adding a new branch w/ only CoinTime cuts (no PID) for detector syst

    studies

  - will explicitly add PID cuts in analysis and vary them for studies

- plans for KaonLT systematic studies

  - also will start with Q2=3.0, W=3.14 and work down

  - Phase 1: check now data yields change

    - compare data yield results due to different bkd subtraction procedures

     (RT vs CKT)

   - vary MM, acceptance, HGC hole cuts, and check variation across t-bins,

     SHMS-settings, and epsilon-settings

  - Phase 2: Radiative corrections ON/OFF, check how SIMC yields change

Chi Kin

-------

KaonLT drafting paper: intro, iterations procedure, text not yet on OverLeaf

- systematics study: Q2=3.0 W=3.14

  - Dave: should not widen acceptance cuts into regions of poorly understood

    ME, only narrow them

  - Garth: similarly don't make diamond wider, as then parts of the low epsilon

    diamond will be empty, only narrow

    - also does not see the point of varying t-binning, this will change the

      t-central of the bin and of course the cross section will change

Alicia

------

Slides for Phi-007 Expt Readiness Review

 - did MM resolution studies, demonstrating we understand the expected Phi-007

   MM resolution, based on KaonLT p(e,e'p)omega data

 - proton arm momentum much lower in Phi-007 than KaonLT, TOF separation will

   be beter, so K+ leakthrough will be less

 - Phi-007 MM resolution expected to be 1.8x worse than KaonLT, based on

   magnetic optics properties

omega LT-sep

- ssypfp big spike at zero resolved

  - issue was a few runs when reprocessing data

  - did not have to replay data, only redo python processing after replay

    - the issue was that not all variables were enabled for some runs

- the omega MC of ssyptar dist is somewhat narrower than Bkd subtracted data

  - Dave: a possible issue is that SIMC does not incorporate collimator

    punchthrough for protons, only for pions

    - will try to locate Mark Jones' old proton punchthrough code for us

- in general, the Sum-MC kinematic distributions are in reasonable agreement

  with data

  - need 2-4x more Pythia MC statistics

Friday: Present

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

Regina - Garth Huber, Nathan Heinrich, Vijay Kumar, Muhammad Junaid,

   Nacer Hamdi, Nermin Sadoun, Alicia Postuma

FIU - Pete Markowitz

JMU - Ioana Niculescu, Gabriel Niculescu

Glasgow - Kathleen Ramage, Rachel Montgomery

Virginia - Richard Trotta

CUA - Tanja Horn

Nermin

------

PionLT LD2- PID study

- last week showed that NGC had very low NPE, looked into recalibrating

  - went through Cameron's NGC calib script

- Run 12035

  - placed Pcal cut to select electron sample

  - cuts placed on approximate boundaries of 4 NGC mirrors

  - apply SHMS cut: -10% < delta < +22%

  - optained new NGC gain factors via Poisson fit of fADC spectra

- statistics looked a bit low, so tried script on Runs 12026, 29, 35

  Gain Parameters:

  PMT1: 3.38   PMT2: 1.735   PMT3: 6.37   PMT4: 4.91

  - Nathan: these are much smaller gain values than he had (about 10) so it

    explains the observed low NPE

  - Garth: surprised how poorly these PMTs were gain matched

  - Nacer comments about PCAL spectrum:

    - 2 peaks around 1, 1.2

      - Nathan: these are most likely an electron and a pi- in the calorimeter

        at the same time.  The other option could be a miscalibration for some

	data

    - peak around 2

      - Nathan: this is 2 electrons in the calorimeter at same time

- after calibration, now gets a more reasonable NGC distribution from 2-6 NPE,

  more similar to what Das had in his setting (~4 NPE)

Next steps:

- NS will try to get a NGC calibration using even more runs to get better

  statistics for the Poisson fit

- will also do a NGC efficiency calculation for electrons

Kathleen

--------

PionLT LD+ PID study

- looking at Q2=1.6, W=3.08, Ebeam=6 GeV, theta_SHMS=5.50deg

- CoinTime study

  - with no PID cuts, hard to see prompt CoinTime peak, due to the very high

    rate with SHMS at small angle

  - then applied PID cuts, prompt peak is clear, shifted prompt peak to zero

  - did random subtraction using 3 peaks to left and right of prompt peak

- RFtime study

  - selected main peak for pi+nn events

Next step:

- refining HMS and HMS cuts

Vijay

-----

PionLT Low Q2 systematics studies

- completed model systematics for Q2=0.375

  - this week, varied sigLT and sigTT models by 1 error bar and looked at

    variation in sigL,T

    - 1st t-bin is more sensitive than the others, results only in a drop in

      the cross section

    - for the other t-bins, sigL,T goes up or down, depending on the model

      change

  - this means a 1-sided systematic error (from model dep) will be applied

    for the 1st bin

Next step:

- compute the combined uncertainties to apply to the data

- it would be good if the model syst could be studied also for Q2=0.425

Nathan

------

PionLT Q2=3.85, W=2.62 cross checks with Junaid's analysis

- showed plots from the analysis scripts

- next step is to compare yield totals, and kinematic plots with Junaid

- Nathan and Sameer had a 3 hour meeting on Wednesday on Lumi analysis

Gabriel

-------

Alternative PID and channel ID for KaonLT

- leaky pi+ has a different distribution than Lambda,Sigma0, so a subtraction

  pi+n data might not be ideal

- first showed a fit of 2 skewed Gaussians to MM data

  - did 10,000 fits and found the Lambda,Sigma0 yield for each fit

  - made a frequency plot of yields obtained from the 10k fits, found this led

    to a 10-15% variation in yield

  - Nacer: many of the skewed Gaussian fits produce a small peak near MM=1.11

    that seems unphysical

- then explained Quality-factor fitting (Q-fit) based on recent paper by

  Williams, Bellis and Meyer: arXiv:0809.2548v3 [nucl-ex]

  - it is a generalization of the SideBand subtraction method

  - compute the "Distance" between 1 event and another, using 7 variables:

    Ebeam, (delta, xptar, yptar)_SHMS, (delta, xptar, yptar)_HMS

  - select 1000 nearest neighbors to each event, do a MM fit of Gaussian peak

    and linear background for Lambda and Sigma0 regions, and find fitted signal

    fractions of Signal and Background

  - use the signal fraction to calculate the Quality-factor for the event

  - instead of incrementing data histogram by 1, increment it by the

    Quality-factor

  - the idea is to get a MM plot of just the Lambda, Sigma0 peaks with no

    background underneath

  - compared Q-fit Lambda, Sigma0 yields to standard fit of MM histogram, they

    agree nicely, much smaller errors than 2 skewed Gaussian fit

Next steps:

- plans to adapt Q-fit method to excited hyperon identification, and write a

  paper on the method

- also preparing the analysis framework for general release  

Next Meetings

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

- Thurs: Mar 5 @ 16:00 Eastern/15:00 Regina

  - PionLT will go first

- Fri: Mar 6 @ 11:00 Eastern/10:00 Regina

  - we will continue where we left off

- NOTE: this will be the last week at these times.  After this, we move to the

  summer schedule

  Thurs: 16:00 Eastern/14:00 Regina

  Fri:   11:00 Eastern/ 9:00 Regina