Kaon_LT Experiment

                Nov 20-21/25 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

Thursday: Present

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

Regina - Garth Huber, Nathan Heinrich, Muhammad Junaid, Alicia Postuma,

   Nermin Sadoun

Virginia - Richard Trotta

CUA - Chi Kin Tam, Tanja Horn

JLab - Dave Gaskell

Ohio - Julie Roche

Junaid

------

PionLT pion absorption correction via Geant4

- NGC on, Aerogel n=1.011 simulation for P_shms=5.127 GeV/c

- added an uncertainty calculation

- correction is 0.9654 +/- 0.00005

RF cut PID efficiency

- added HGC to PID selection cuts to both numerator (Did) and denominator

  (Should)

- get 0.99802 +/- 0.00007

- will apply both corrections to the normalized physics yields

  - *NB* Garth: it would be great if you could put together a table of

    systematic uncertainties similar to what is in the Blok paper

    - please include Nathan's new systematic errors there as well

Next steps

- setting up for next Q2 LT-sep, checking RF cut offsets for these data

Nathan

------

PionLT CoinLumi analysis

- sent everyone a copy of his Lumi report for comment

  - no comments received yet from Sameer

- question for Dave on where to find the documentation for the hodoscope gate

  width

  - after some hunting, Dave finds:

    Found an elog from 2019 that says discriminator widths were 50 ns

    https://logbooks.jlab.org/entry/3747761

  - Dave also asks Bill Henry to recheck the gate widths manually, and gets

    back an answer surprisingly quickly

    https://logbooks.jlab.org/files/2025/07/4423016/IMG_6761.jpeg

    which confirms that the widths are still 50ns

  - Nathan confirms that 50ns is the value he used in his ELLT calculation, so

    everything should be good now

Alicia

------

pi+n BSA paper has been resubmitted to PLB

- awaiting on final communication from journal, hopefully about page proofs

- working on arXiv submission, filling in author list metadata

u-channel replay of Q2=3.0, W=2.32 data

- the intent was to see if new HMS matrix elements for P_hms=6.59 GeV/c made

  any improvement to the MM reconstruction resolution

  - differences in MM are quite small, we will have to live with the worse

    resolutionfor this setting

- pi+n contamination in omega region is worse than for Q2=3.0, W=3.14 setting

  - the two MM peaks are not fully resolved, comparison to clean pion data

    indicates omega is 1-1.5sigma to left of pi+n, so very close to the nominal

    1.5sigma optical criterion for resolving two close peaks

  - unfortunately, no RF cut for this setting

  - also, Python generator behaves poorly for this setting, as reported earlier

    (extraneous MM shape)

    - Garth: suggests to consult with Henry Klest on Pythia settings

- retained EM^2 PM^2 so MM^2 can be calculated

  - SIMC predicts DVCS and pi- to be offset from each other by ~1 sigma, but

    MM^2 data is shifted by ~1 sigma to right of both peaks

    - doesn't trust data peak positioning so far away from omega, pi+n region

    - applying a shift of 4 MeV^2 to data, to overlap data with pi0 simulation

      shows reasonable agreement in peak shape, lends credence that this region

      is mostly pi0, as expected

    - low epsilon data doesn't need any MM offset, as peaks already in correct

      position, and also RF cut available

Next steps:

- will look in more detail at low epsilon MM^2 data

- will contact Henry Klest re. Pythia

Richard

-------

KaonLT Q2=3.0, W=2.32 LT-sep

- looking at t-phi bins with anomalously low cross sections

  - the issue is that these are bins with no pi+n leakthrough to anchor the

    background fit underneath Lambda peak

  - Chebyshev polynomial fits for background are poorly constrained, and

    clearly oversubtracting, giving outlier yields near zero

  - now looking at 2 different types of background fits

    - quadratic w/far edge fixed to MM=1.15 GeV data

    - Chebyshev poly fitting the rest

    - still adjusting fit parameters, but things looking better

  - *NB* Garth: the error bars due to this background fitting uncertainty are

    significantly underestimated, the way you'r calculating these errors is

    fine only for those settings where pi+n leakthrough gives a constraint, but

    you need to take into account the uncertainty in the background estimation

    where it is poorly constrained

- using Q2=4.4 parameterization

  - Goal: all ratios within 3sigma of mean, and mean within 1.5sigma of unity

  - then will apply these fit functions also to Q2=5.5 data

Chi Kin

-------

KaonLT Q2=3.0, W=3.14 LT-sep

- last week, theta_pq (CM frame) was not calculated correctly

  - wrong reference frame used, giving negative weights in re-weighting script

  - this is fixed now, things looking more consistent

- iterations looking a bit better now

  - did 8 iterations, fits stabilized but still not good enough

  - looked at t-binning, with goal to improve the yields in poorly-populated

    bins

    - narrowed the t-bin ranges:

      new bin limits: 0.19,0.26,0.31,0.38,0.49

      - previously had 0.17 to 0.6

      - Richard was using only data up to 0.4

Friday: Present

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

Regina - Garth Huber, Nathan Heinrich, Muhammad Junaid, Nacer Hamdi,

   Vijay Kumar, Nermin Sadoun

York - Stephen Kay

CUA - Sameer Jain, Tanja Horn, Chi Kin Tam

JMU - Ioana Niculescu, Gabriel Niculescu

Ohio - Julie Roche

JLab - Dave Gaskell

CSULA - Konrad Aniol

Nacer

-----

KaonLT Q2=0.5 LT-sep

- still looking for right parameterization

  - fitting only sigT=p1/Q2+|t|^p2/(Q^2+p3)^2, L=LT=TT=0

  - switched to simpler Wfac=1/(W^2-mp^2)^2

  - did 4 iterations

    - Data/MC ratios looking quite a bit better, R near 1 and flatter than

      before

      - high epsilon has a bit bigger oscillations than low epsilon

    - sigT has non-monotonic t-dependence

- would like to try adding back sigL now

  - *NB* Garth: suggests a polynomial for sigT, the fit |t|^p2 form is

    monotonic and cannot reproduce the observed t-dependence

    - TT could be a simple straight line fit TT=a+b*t, no Q2-dependence

    - L could be some simple monotonic, like L=p1+|t|^p2

    - keep LT=0

- kinematic and focal plane plots for Data and MC

  - overall the plots look good, but there are some mismatches between Data and

    MC that generated some discussion

    - SHMS_xpfp is a bit narrower than data

    - HMS xptar has a shift between data and MC for high epsilon, right SHMS

  - *NB* Gabriel: puzzled that the HMS_xptar shift is as large as it is

    - Dave: this is a known issue with HMS xptar reconstruction

      - the data should be corrected by a couple of mr, could have an impact on

        phi-distribution, i.e. at edges of acceptance Data and MC will

        mismatch, giving rise to oscillations in the Data/MC ratio

      - Junaid did not see this effect

      - *NB* Nacer and Junaid will compare 0th order matrix offsets used in

        replays

Vijay

-----

PionLT Low Q2 LT-sep

- systematic checks for Coin Blocking correction

  - varied timing cuts +/-4ns

    - resulting uncertainties are correlated with epsilon for both Q^2:

      - low epsilon: +/-0.5%, mid and high epsilon: +/-0.3%

  - how should we apply these uncertainties to data?

    - Nathan: what about calculating it run-by-run and adding it in quadrature

      with the statistical errors?

    - Garth: these are not random errors, should not be added in quadrature

      with statistical errors.  The issue is that random errors are magnified

      by 1/Delta-epsilon in the L/T-separation, non-random errors are not

      magnified

      - Nacer: wouldn't it be more conservative to take the biggest error and

        apply it with the statistical?

	- Dave: the issue is that it's too conservative, under-reports the

  	  quality of the data, and results in lower quality L/T-separation

    - *NB* Dave: need to break this uncertainty into parts (see systematics

      table in Blok paper).  Part of the uncertainty is global, and part of it

      epsilon-correlated)

      - Garth: take the smallest (high epsilon) value ~0.3% as a scale

        systematic uncertainty, and the difference between the low and mid

        epsilon values and this smallest value as the epsilon-correlated part

	- Dave: agrees with this suggestion

    - *NB* Garth needs to have a discussion with students on how systematic

      uncertainties were treated in Fpi-2 analysis (Blok paper)

- shows Data MC overlay for kinematic and focal plane plots for 2 Q2=0.425

  settings:

  - high epsilon, Right-2

  - low epsilon, Center

  - agreement looks fairly good, HMS yptar shift seems smaller than Nacer's

Sameer

------

KaonLT Coin Blocking correction

- error bars are statistical and systematic added in quadrature, following our

  discussion he will quote them separately next time

  - Nathan also needs to separate them in his analysis

- systematic uncertainty is bigger than statistical, which indicates a problem

  in the analysis

  - *NB* Nathan: the right cut is too tight into the distribution, it needs to

    shift by ~20ns

- Nacer: what boiling factor should we use for KaonLT data?

  - *NB* Nathan: Richard's boiling factor is ~2x Nathan's

    - would like someone to redo this study using the same methodology as in

      Nathan's report

      - perhaps Richard used what we now understand to be the wrong LiveTime,

        given the newer studies?

      - hopefully get a boiling number similar to Nathan, in which case use his

        number as it's higher statistics

	- if it's significantly different, then need to investigate in more

	  detail

  - Garth: agrees with this suggestion, it would be good for Nacer to add

    this to his list.  Nathan thinks reproducing his study should not take long

    to do

Gabriel

-------

Looking in more detail at KaonLT offsets

- ideally want a common set of offsets for all settings

- using physics data to investigate the variation in the offsets with setting,

  which will also be helpful in understanding the uncertainties in the offsets

  -  uses Lambda, Sigma, pi+n MM peaks as constraint

    - Lambda, Sigma: compare data to SIMC MM values

    - pi+n: recalculate MM using pion mass and compare to SIMC

- follows a method similar to Richard:

  - generates many offsets, evaluates MM for each

    - finds best set of offsets, then creates a new generation of offsets near

      these values

    - process is time consuming, only 1 generation done so far, expect 2-3

      generations necessary to converge

    - gets a nice set of histos of offsets satisfying some criteria for each

      generation

- not using hcana MM values, calculating everything on own from spectrometer

  vectors

  - using HallC:p value of the beam energy, not the value from

    standard.kinematics

  - *NB* Garth: we had a discussion about this at a meeting that Gabriel

    missed.  The issue is that HallC:p is not corrected for the Arc Energy

    Measurement.  Need to find the value of HallC:p at the time of the Arc

    Energy Measurement (AEM), and then correct all other values by the ratio

    Beam=(HallC:p_now)/(HallC:p_AEM)*(AEM-GeV)

    - also note that for 10.6 GeV beam energy the bremsstrahlung of the beam in

      the Hall C Arc is too large to ignore.  The Arc Energy Measurements are

      corrected for this loss (via a calculation), while HallC:p is not

      corrected for it.

  Next Week Meetings

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

- Thurs: Nov 27: no meeting due to USA Thanksgiving

- Fri: Nov 28 @ 11:00 Eastern/10:00 Regina

  - Canadians and UK collaborators are invited

  - Richard thought he would be available to attend