Kaon_LT Experiment

                 Dec 4-5/25 PionLT/KaonLT Analysis Meeting Notes

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

                                (Notes by GH)

                    Today: KaonLT 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, Alicia Postuma, Muhammad Junaid, Nathan Heinrich,

   Nermin Sadoun

Virginia - Richard Trotta

CUA - Tanja Horn, Chi Kin Tam

JLab - Dave Gaskell

Alicia

------

pi+n BSA paper

- PLB asked for a 3rd round of edits and sent to reviewer, even though

  the reviewer said further review was not required

  - the referee responded quickly, seemed slightly annoyed to have been

    asked again

- PLB has now formally accepted the paper, trying to deal with

  copyright forms, running into yet more problems with their website

- paper is also now on the arXiv:2512.01928

u-channel pi0 region investigations

- Q2=3.0, W=3.14, low epsilon, SHMS center setting

- Bill requested that Alicia try fitting both pi0 and DVCS simulations

  to the MM^2 data, rather than a fixed 90%:10% ratio

  - the fit looks surprisingly good:

      ~25% DVCS:75% pi0

      ~100 counts DVCS vs 300 counts pi0

- project could be assigned to a future summer student to complete, now

  moving back to omega region analysis

proton PID studies

- tried a "geometric cut" on the AeroNPE vs RFtime distribution

  - for Q2=3.0, W=3.14 high epsilon, RF is available for Center and

    Right SHMS, but not Left

  - looked at MM, get a few more omega events but unfortunately also a

    few more pi+ events, the cut needs some fine tuning

- the Left SHMS setting has no RF cut available, so more pi+

  leakthrough

  - would probably need to apply a tighter Aero cut and a "geometric

    cut" on HGC vs Aero

Junaid

------

Q2=3.85, W=2.02 LT-sep preparations

- will show main plots tomorrow, today just a quick update

- finished running SIMC for this setting

- this setting has both Right-1 and Right-2 settings, which are

  different by about 1degree, apparently we did an HMS saturation test

  as part of run plan, they have to be analyzed separately

- high and low epsilon have different Aerogel index

  - was using same Aerogel cuts, will have to investigate to see if any

    changes to the cuts are required

Nathan

------

PionLT CoinLumi systematic uncertainties

- reading Blok paper, looking at how to divide syst unc into scale,

  point-to-point and partly-correlated categories

  - Dave: overall systematic uncertainty scale comes from HeepCoin

    analysis, i.e. how well we are able to reproduce the known elastic

    cross sections

- EDTM syst unc is evaluated by increasing errors to give ChiSquare=1

  - Garth: can look at how the ChiSquare varies by rate

- Tracking syst unc: Ali did not evaluate systematic uncertainties in

  his tracking study, will look into this

  - Dave: suggest to look at Abishek's thesis

    - SIDIS data are flatter, allow some systematic studies not easily

      possible with more rapidly varying exclusive data.  Can take his

      values (with citation) where appropriate

- also looking at Junaid's LT-sep scripts, will try to reproduce his

  Q2=3.85, W=2.62 results

Friday: Present

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

Regina - Garth Huber, Nathan Heinrich, Alicia Postuma, Nermin Sadoun,

   Muhammad Junaid, Nacer Hamdi

Virginia - Richard Trotta

CUA - Chi Kin Tam, Tanja Horn

Ohio - Julie Roche

JMU - Ioana Niculescu, Gabriel Niculescu

FIU - Pete Markowitz

York - Stephen Kay

JLab - Dave Gaskell

Richard

-------

KaonLT Q2=3.0, W=2.32, low epsilon, Center SHMS

- checking empirical fits of background near Lambda peak

  - current method:

    - first subtract pi+ leakthrough, this still leaves substantial background

      in Lambda region

    - then fit a quadratic to shoulder around MM=1.2

       p0*(x-1.12)+p1*(x-1.12)^2

    - finally fit a 2nd order Chebyshev to region under Lambda

  - this leaves an unfit background at MM<1.05, Richard asks if he should apply

    another polynomial to subtract this region?

    - *NB* Gabriel: you should try to use the Lambda peak shape from SIMC to

      guide you on how much background you need to subtract

      - you need to match the data to MC Lambda peak shape over a subset of the

        MM range (about 1.08-1.15)

      - only need to estimate background in that region

      - region lower than MM cut won't be relevant

      - estimate the background 2 ways: take the difference between them as a

        systematic

   - Nacer: your background fit is over a narrow MM range, maybe fitting over a

     broader MM range will give a more stable result

     - another possible method is SideBand Subtraction, avoids issues with

       challenging fit for each t-phi bin

- empirical fit error calculation method

  - define normalized background yield in analysis window

  - propagate fit uncertainty to integrated background using the full

    covariance matrix

  - if covariance matrix is invalid, use diagonal approximation

  - evaluate using central differences

  - convert background-integral uncertainty into a fractional yield uncertainty

  - Chi Kin: try to avoid the more complicated formula using full covariance

    matrix, try using just the diagonal approximation

- question for us: currently using a 25 count/t-phi-bin threshold in the

  analysis, if the threshold is raised to 100 counts, end up removing many

  t-phi bins

  - *NB* Garth: can you instead use your background error estimate to guide

    whether to keep the t-phi bin or not?

    - a clean MM histo with few counts might be easier to analyze than a high

      count MM histo with very uncertain background

Chi Kin

-------

KaonLT follow up to Richard's background investigations

- method used:

  - first subtract Sigma0 MC to give a flatter MM dist to right of Lambda peak

  - then define 2 SideBands left and right of Lambda peak and fit a Chebyshev

    polynomial underneath the Lambda

  - issue: the Lambda radiative tail is oversubtracted, agreement between data

    and Lambda MC is good in main peak region, but poor in tail region

  - it would be better to fit sidebands and Lambda together

    - tried approximating Lambda MC peak with CrystalBall distribution and fit

      this together with the Chebyshev

      - this didn't always work, too many free parameters to fit

    - *NB* Gabriel: avoid the CrystalBall distribution, just fit SIMC and

      background directly to the data.  This will have fewer free parameters

      - for the SIMC, just a scale and small MM offset to fit

- Q2=3.0, W=3.14 LT-separations

  - changed from Richard's more complicated W-factor to Wfac=1/(W^2-mp^2)^2,

    i.e. same as Nacer

    - high epsilon Data/MC ratios slightly improved

    - low epsilon ratios improved too

  - lowest t-bin: high epsilon sig_uns ~25% larger than low epsilon

  - next t-bin: sig_uns fairly flat with phi, high epsilon ~10% larger than low

    epsilon

  - other 2 t-bins: high epsilon = low epsilon within errors

  - sigL has a significant t-slope

  - results look fairly encouraging

Next week: plan to show first plots for Q2=4.4, W=2.74

Nacer

-----

KaonLT Q2=0.50 LT-sep

- found a good model for sigT, still working on sigL

  - the exponential function does not fit well, perhaps will move to flat or a

    polynomial function

- HMS xptar Data vs MC comparison

  - Dave: looks like there's a consistent shift between them, the entire

    distribution needs to be shifted by 2.5mrad

    - the HMS matrix elements have an issue, likely the 0th order offsets were

      not included correctly when doing the fitting

    - this effect is seen also in other data sets, affects the phi

      reconstruction

   - Nacer uncommented the hphi_offset=-4.9E-3 line in his file and replayed

     data

     - the data shifts, but in the wrong direction, needs a positive offset

     - *NB* Dave: not sure where that number came from

       - please try hphi_offset=+2.8E-3 and htheta_offset=0

         determined by DG using 2022 inclusive data

     - *NB* Nacer and Richard/ChiKin need to replay their data with this offset

   - Junaid is using Christine's HMS matrix elements and 0th order offsets

     determined with NPS data, sees no HMS xptar shift

     hphi_offset=+5.8E-4

     - *NB* Dave: you need to check PionLT data for P_HMS<5 GeV/c, where older

       ME are being used

   - other distribution showing a shift is SHMS-xp_fp'

     - Garth: possibly this is caused by the HMS offset, since exclusive data

       are highly correlated between the 2 spectrometers

     - Dave: see what it looks like after modifying the HMS offset and we can

       discuss again

Junaid

------

Q2=3.85, W=2.02 LT-sep preparations (continued)

- Aerogel cut: 1.5npe

  - *NB* Garth: need to evaluate the pi+ cut efficiency for this cut

  - *NB* Dave: you are only showing the 1D Aero-NPE plot, it's impossible to

    tell whether this cut is what you need.  Not saying that this cut is

    inappropriate, but rather that it is impossible to evaluate

    - please look at other variables with this cut, such as CoinTime

    - Garth: can also look at MM to see background near pi+n region

- MM offsets: from fit of Data to MC

  - obtain offsets of 1-5 MeV

- Diamond cuts

  - low epsilon diamond needs to be a bit tighter, to remove unpopulated

    regions in other SHMS settings and improve comparison with SIMC

  - high epsilon diamond seems too high in W compared to low epsilon, for both

    data and MC

    - *NB* Garth: please double check that standard.kinematics is set correctly

    - if it remains, then need to remove non-overlap region from low epsilon

      diamond

Next steps:

- will complete these studies, then move to t-binning

Next Week Meetings

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

- Thurs: Dec 11 @ 16:00 Eastern/15:00 Regina

  - PionLT will go first

- Fri: Dec 12 @ 11:00 Eastern/10:00 Regina

  - we will continue where we left off