Kaon_LT Experiment

                Mar 19-20/26 PionLT/KaonLT Analysis Meeting Notes

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

                                (Notes by GH)

                    Today: PionLT will be discussed first

Thursday: Present

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

Regina - Garth Huber, Alicia Postuma, Nermin Sadoun, Muhammad Junaid,

   Nacer Hamdi

CUA - Chi Kin tam, Sameer Jain

Virginia - Richard Trotta

JLab - Dave Gaskell

Junaid

------

PionLT Q2=3.85 W=2.62 LT-sep

- added t-shifts calculated by GH

  - needed to adjust first 4 t-bins to keep ~2500 counts/bin

      old 3 bins:  0.16-0.208   0.208-0.226   0.226-0.264

      new 3 bins:  0.16-0.206   0.206-0.226   0.226-0.262

  - had MM offsets with both signs so not all data shifted in same way

    - average effect is data shifted to slightly smaller -t

  - Data/MC ratios very similar to before

    - some oscillations in 5th t-bin at low epsilon, was there before too

    - Vijay also had some low epsilon high -t oscillations

  - Rosenbluth "money plots"

    - before: d2sig/dtdphi for bins 2-4 had big dip at phi=180 where

      sig_hieps<sig_loeps, even if sig_hieps>sig_loeps overall

    - now: the sig_hieps has shifted up by enough that at dip

      sig_hieps=sig_loeps

  - separated cross-sect: sigL increased by ~1sigma at low -t, changes smaller

    at high -t

- met with Nathan yesterday to discuss comparison analysis, meeting again

  tomorrow

Nacer

-----

KaonLT Q2=0.5 LT-sep

- added t-shifts calculated by GH

  - MM shift is always to the left (i.e. needed to shift data to lower MM), so

    data need to shift to lower -t as well

  - biggest t-shift=0.00375, which is comparable to the width of a t-bin

  - no changes made to binning, 1 iteration only

  - sigL changes shape a lot, goes strongy negative for first bin

  - lowest t-bin Data/MC ratio is very bad

  - yields unchanged to 7-sig-figs, something is claerly wrong

  - will look carefully at scripts

- Garth: *NB* probably the t-binning had to be adjusted as well, take a look at

  yields/bin, as Junaid did

Alicia

------

KaonLT u-channel LT-sep

  - more u-binning studies for Q2=3.0 W=3.14

- u-bin limits

  - previously, showed bkd-subtracted omega yields vs u-bin for

    Left,Center,Right

  - to get a cleaner sample, tightened cut 0.75<MM<0.85

  - shifted bins slightly lower in -u

- compare t-binning vs u-binning

  - with u-binning the MM plots seem to have better defined omega peaks

  - confirmed from Bill's thesis that he used u-binning

- u-phi scatter plot

  - add an offset (-u+0.3) because u passes through zero

  - left SHMS has some u>0 events

  - investigated at what value of u we lose total phi coverage

- investigated u reconstruction resolution

  - compared u vs u_true in SIMC

  - sigma of (u-u_true) is 0.008 GeV2, which is much smaller than the

    u-binning

- checking which u-phi bins are dominated by omega radiative tail events

  - will eventually do a Rad-Off vs Rad-On study in SIMC

  - for a quick check, looked at peak to tail ratio

      tail: 0.85<MM<1.0   peak: 0.75<MM<0.85

  - tail/(tail+peak) gets >50% for -u>0.6, which is higher than where full

    phi-coverage ends

- try 2 u-bins vs 3 u-bins, due to statistics

   0<-u<0.12 - many fewer backgrounds contribute than higher -u

   0.12<-u<0.40 - center SHMS: harder to pick out omega by eye

                  left SHMS: RF time is not available

		  right SHMS: has RF time

- binned shape study

   0<-u<0.12 center SHMS: rho fit fails for some cases due to narrower

     0.65<MM<0.95 range of data

   0.12<-u<0.40 center SHMS: K data leakthrough has a lot of piDelta events

  - instead of using both K and pi data leakthroughs, will use pi data

    leakthrough and add to it the Klambda MC

  - low statistics and systematic uncertainties in bkd subtraction will need

    careful work

    - will revisit u-phi binning after more bkd studies

Next steps

- need to run more Pythia MC, since most events are at higher -u and not needed

  for the high MM shape study

- will meet with Bill next Tuesday to get his opinion on binning and shape

  studies done so far

Friday: present

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

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

   Muhammad Junaid, Nacer Hamdi

CUA - Chi Kin Tam, Sameer Jain, Tanja Horn

Ohio - Julie Roche

Virginia - Richard Trotta

Glasgow - Kathleen Ramage

JMU - Cabriel Niculescu

FIU - Pete Markowitz

Kathleen

--------

PionLT LD+ cut study

- acceptance cuts

  - using cuts from Vijay's thesis (p. 167)

    - these are probably okay, will need to check what Matrix Elements were

      used

  - now applying acceptance cuts first, before the others

- then apply CoinTime cut, since it removes most events, before looking at

  other cuts

- RF cut vs MM

  - Junaid: RF time looks good, no evidence of a double peak, like Nermin found

    (and corrected)

- HGC

  - made a plot of X,Y vs NPE, sees clearly where the "hole" is

  - if a HGC cut is used, a cut needs to be drawn around this region to exclude

    it in both data and MC

  - Alicia: suggests to first see if you can get away with just Aerogel and RF

    cuts, and no HGC, in which case no HGC hole cut would be needed

- NGC

  - don't apply a cut for pi+ data, but use the pi+ data to determine the

    correction to apply to pi- data (where a cut will be necessary)

- Aerogel

  - made a plot f X,Y vs NPE

    - some discussion about small region at x=-45cm that is slightly

      inefficient

    - there is also evidence of a small gap between 2 rows of tiles at x=-30cm  

  - Junaid: aerogel tray cut is applied in replay, cuts were set by Petr

    Stepanov, so the tray cut is probably okay

  - will try placing cut at 1 NPE, to remove protons

- MM plot:

  - want to see if events to left of pi+n peak goes to zero

    - Nathan: need to subtract both Randoms and Dummy Target

    - Gabriel: need to normalize using Qeff from Beam Charge, TrackEff,

      LiveTime, etc

    - Garth: plot with y-axis going slightly negative, to check for

      over-subtraction

Nermin

------

PionLT LD+ detector efficiencies

- analyzing 9.117GeV Q2=1.6 W=3.08 SHMS right (6deg)

- applied: Aerogel>0.2NPE  -2.0<CoinTime<+2.0  No RFTIme cut

- HMS Cer Eff 99.1%

  - Gabriel: please confirm that binomial errors are used for detector

    efficiencies

  - Alicia: yes it should be set up this way in the Report Files

- SHMS Aero Eff 97%

  - Garth: this seems low, pions should have a very high aerogel efficiency

    - wonders if there is proton contamination in the event sample used to

      calculate the efficiency

  - Nathan: *NB* need to apply the HGC to clean up the particle sample when

    calculating pion aerogel efficiency

    - suggests to apply NPE>5 cut on HGC

  - Junaid: don't worry about the exact NPE value to use in the cut, as long as

    it's sufficiently high to exclude protons, NPE>3 is probably okay too

Next steps:

- will look at LD- detector efficiencies, same Q2-W setting

- after that, will start LD2 Lumi study

- *NB* Junaid will send Kathleen and Julie his Efficiencies Report

Nathan

------

PionLT Q2=3.85 W=2.62 analysis cross-checks

- made a spreadsheet of Yield Comparisons between NH and MJ

  - Low Epsilon Center yields are different ~5%, other settings are exactly the

    same

  - cuts could be different, still checking

  - Garth: a possible explanation is that a run replay failed and you're

    comparing a different number of runs

  - Nacer: an explanation alternate to run fail could be symlinks pointing to

    the wrong place, remove symlinks to be sure you're using the same file

- using ROC2, found an error in Junaid's code that ROC1 was used by mistake

- different MM offsets could lead to small yield variations

  - NH and MJ are both fitting SIMC MM, expecting the same result, but

    obtaining small differences

  - Garth: you need to be sure both SIMC files have exactly the same

    statistical sample, if even 1 event is different, all subsequent events

    will have a different random number sample

Sameer

------

KaonLT CoinTime blocking

- instead of a dynaminc timing window, went to a fixed window for each run

  period

  - all windows are 140ns wide:

    Oct/18 - multiple timing windows, each time the DAQ configuration changed

    Dec/18 - 1 window: -11 to +129ns

    Mar/19 - 1 window: -25 to +115ns

  - showed plots of blocking correction vs CoinRate

    - plots looked good, studies are nearly completed

- Richard will send his blocking corr for Q2=3.0 W=3.14 for cross checks

- Questions for Nathan: did you calculate the cable length delays?

  - No, looked at photos of oscilloscope traces and added a bit extra to

    account for jitter

  - the way Sammer did it looks good

- Q: how did you calculate the systematic uncertainty w/5ns window variation?

  - Nathan: don't want the cut too tight, want to remain in the shoulders of

  - good randoms

- will try to make a plot of corrected yields after CoinCorr vs Rate, like

  Nathan did, to confirm everything is correct

- Nacer: will the CoinTime blocking correction be calculated in the Report

  Files?

  - Yes, still waiting on last checks, will update the files soon

Chi Kin

-------

KaonLT systematic studies

- SIMC systematic Rad On/Off

  - get 2x difference in yield after applying Diamond Cut

  - Garth: this indicates the MM cut is too narrow, it excludes all of the

    radiative tail

    *NB* the MM cut needs to be broadened

  - Gabriel: the data looks a lot more like the RadOn dist than RadOff

    - maybe tweaking one of the RadCorr params in SIMC is better than

      RadOn/RadOff, which is an extreme change

  - *NB* CKT will contact Dave for a meeting on how to proceed

- acceptance systematics

  - will vary target cuts

    - ssxptar cut looks a bit wide

  - delta cut: can narrow down the delta cut, but not make wider

    - Tanja: our knowledge of the magnetic optics is not perfect, there are

      discrepancies between data and MC even within the well understood region.

      We want to quantify that

Richard

-------

KaonLT LT-sep

- made a write-up on statistical uncertainty calculations that will be posted

- will consult GH offline on the exact input values used in his t-shift calc

- comparison of thetaCM dists for data and MC

  - sees a spike at thetaCM=0 in data but not SIMC

  - Gabriel: what happens in the code when cos(thetaCM)>1 due to resolution

    effects?  This can lead to a spike at thetaCM=0 when taking the ArcCos when

    calculating thetaCM

    - *NB* to avoid this effect, don't convert to thetaCM

  - Tanja: we need to understand this issue, does it lead to the spike in sigL

    at low -t?

Next Meetings

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

- Thurs: Mar 26 @ 16:00 Eastern/14:00 Regina

  - KaonLT will go first

- Fri: Mar 27 @ 11:00 Eastern/9:00 Regina

  - we will continue where we left off