Kaon_LT Experiment

                Nov 28/25 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, Nacer Hamdi, Nermin Sadoun,

   Muhammad Junaid, Nathan Heinrich

Glasgow - Rachel Montgomery, Kathleen Ramage

York - Stephen Kay

CUA - Chi Kin Tam

Junaid

------

Setting up for PionLT Q2=3.85, W=2.02 analysis

- completed RF cut studies.  Using cut 1.2<P_RF_Dist<3.4

  RF histogram after PID cuts is remarkably clean

- evaluated RF cut efficiency: 99.8%

- evaluated pion absorption correction w/ NGC and n=1.011 aerogel, i.e. only

  the SHMS momentum is different from the earlier setting

  - comparing the two corrections:

    P=5.127 GeV/c   corr=3.6%   previous setting

    P=3.493 GeV/c   corr=4.5%   this setting

Next steps:

- now setting up SIMC for this setting

- then will investigate MM cut and MM offset

- after that would be the diamond cut selection

Q2=3.85, W=2.62 analysis

- added CoinTime Blocking, pion absorption and RF cut efficiency corrections

  and did 2 more iterations

Nathan

------

PionLT wrapping up CoinLumi systematics

- as reported earlier, applying a small "ad-hoc" correction to the

  uncertainties gave fits vs CoinRate with ChiSquare closer to 1

  - this "ad-hoc" correction was added in quadrature to the other uncertainties

- since this "ad-hoc" correction is one of the larger systematic uncertainties,

  it's worth looking at in a bit more detail

  - instead of adding it as a constant in the denominator, tried attributing

    this as a systematic in the EDTM instead

    - this gives a small rate-dependence to the uncertainty

    - resulting ChiSquare is slightly better, 1.01 vs 1.07

- Dave had earlier suggested that Nathan treat the EDTM uncertainty in this

  way, so that's done now

  - will call it the "EDTM systematic uncertainty" rather than "ad-hoc"

  - will update the Lumi report accordingly

Kathleen

--------

PionLT analysis setup

- still working hard on classes, going through Junaid's analysis instructions

  when has time

Chi Kin

-------

KaonLT Q2=3.0, W=3.14 LT-sep

- shifted MM Lambda peak to SIMC positions, instead of other way around

- looking in detail at match of SIMC Lambda peak to data for each t-phi bin

  - keeping the bin if passes criteria:

    - counts>100

    - yield+error>0

    - yield ratio of the Lambda(1.10-1.14) region to pi+ region>5%

- slightly adjusted t-bins, now has 4 t-bins and 12 phi-bins

  - still looking at adjustments to t-bin limits

- Data/MC ratios

  - generally looks good, but gets some bumps in phi-distribution for the

    highest 2 t-bins

  - using sigT=p4*exp(-|p5*t|), monotonically falling

    - has questions whether this is a good form to use, Garth responds that

      T is expected to be fairly flat, but does not have to fall monotonically

      for K+, it could have a shallow hump in it, we will discuss this more in

      Nacer's presentation

  - no Q2 or W dependence in any of the functions.  Nacer suggests to restore a

    simple W-factor

- Data and MC focal plane and physics variable plot comparisons

  - comparison is notably worse for the Left setting than the Center setting

- looking into further improvements

  - describing Sigma0 peak w/ CrystalBall function, as described earlier

    - sometimes this gave a long tail underneath the Lambda, leading to

      over-subtraction

  - looking at an alternate fit form to avoid this, Gaussian+Tail+Baseline

    gives a good description.  Then exclude the Baseline, subtracting just

    Gaussian+Tail

  - some t-phi bins have a weird background between the Lambda and Sigma peaks

    which doesn't follow either fit form, the statistics for the shown bin

    looks marginal

    - looking at Landau distribution with a polynomial

    - *NB* Garth: is it essential that we retain these t-phi bins, i.e. if this

      same bin is populated more cleanly by a different SHMS setting, then it

      would be better to just exclude this bin (for both data and MC), rather

      than have a large systematic uncertainty from the poorly constrained

      background fit

Nacer

-----

KaonLT Q2=0.5 LT-sep refinements

- trying different functional forms for sigT, while keeping L=LT=TT=0 in fit

- simple Wfac=1/(W^2-mp^2)^2 form used

- last week showed a fit that gave fairly good Data/MC ratios, the question is

  if this can be improved further, to give ratios more uniformly closer to 1

  (rather than a t-dependent ratio) and fewer phi-wiggles

- Try sigT=(p1*|t|+p2)*exp(-|p3*t|)

  - Data/MC Ratios fairly flat with phi, but ~1.6 for low -t, higher -t near 1

  - the issue is that the model has difficulty accommodating the hump in sigT

    vs t, cross section drops too quickly at low -t, becoming negative

- Try sigT=(p1*|t|+p2)^2 *exp(-|p3*t|)

  - this forces sigT to not go negative at low -t

  - ratios much closer to 1 at low -t, low epsilon, but high epsilon is worse

    ~1.3

- Try sigT=p1*exp(-|p2*t|)/(p3*|t|+p4)^2

  - ratios look promising, phi-wiggles but under control

- Also tried sigT=p1*exp(-|p2*t|)/(p3*|t|+p4)^2

  - gave good ratios after 3rd iteration, except at low -t

Next steps:

- will go back to sigT=p1/Q^2 + |t|^p2/(Q^2+p3^2)^2

  and introduce a simple function for sigL

Alicia

------

Investigating the feasibility of u-channel DVCS info from KaonLT data

- looking at plots of MM^2=EM^2-PM^2 to see how well DVCS and exclusive pi0 are

  separated

  - proton PID and random subtraction applied, but no dummy subtraction

  - default normalization in study: 90% pi0, 10% DVCS

Low epsilon settings, no MM^2 shift is needed in omega region, so MM^2

reconstruction in pi0 region is presumably reliable

- Q2=2.1, W=2.95

  - DVCS region has poor statistics

- Q2=3.0, W=2.32

  - DVCS outside coincidence MM^2 region for this setting

- Q2=3.0, W=3.14

  - pi0 peak is mostly consistent with SIMC MM^2 shape, except for a shoulder

    on left side that is presumably DVCS

    - the shoulder indicates DVCS is more than 10% of pi0, could be as much as

      20%

- Q2=4.4, W=2.14

  - also see a DVCS shoulder on pi0, but not as prominent as previous setting

- Q2=5.5, W=3.02

  - statistics are poor

High epsilon settings, no diamond cut is applied to equalize to low epsilon

acceptance

- Q2=3.0, W=2.32

  - MM^2 resolution is poor, but statistics are very good

  - DVCS region appears easier to interpret than omega region, due to fewer

    contributing processes

  - to describe DVCS region, would need both DVCS and pi0 SIMC plus a

    polynomial background

- Q2=3.0, W=3.14

  - applied a MM^2 shift, perhaps too far

- Q2=4.4, W=2.74

- Q2=5.5, W=3.02

  - bad statistics, will get worse if a diamond cut is applied

In general, it appears there is enough there to be worthwhile for a MSc or

undergraduate summer student to study in more detail.  Will go back to omega

analysis now

- Q2=3.0, W=2.32

  - investigate whether MM^2 rather than MM makes it easier to separate omega

    from pi+n leakthrough

  - conclusion: the separation between the peaks is larger with MM^2, but so is

    the width of each peak, so no difference in separation capability

Next Week Meetings

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

- Thurs: Dec 4 @ 16:00 Eastern/15:00 Regina

  - KaonLT will go first

- Fri: Dec 5 @ 11:00 Eastern/10:00 Regina

  - we will continue where we left off