Privacy and Security Notice

elastic

Final H(e,e'p) results (?) 05/31/06


The HMS acceptance cuts are:
abs(hsyptar)<0.03.and.abs(hsxptar)<0.055.and.abs(hsdelta)<6.0,  highest Q2
abs(hsyptar)<0.04.and.abs(hsxptar)<0.07.and.abs(hsdelta)<8.0, other two data points

The SOS acceptance cuts are:
abs(ssyptar)<0.07.and.abs(ssxptar)<0.04.and.abs(ssdelta-2.5)<12.5


Q2 (GeV2)
 Y data (counts/mC)
 Y simc (counts/mC)
 Ydata/Ysimc
 Stat. error
4.511
6.1  
7.47 
53.6289
11.5775
3.35994
 51.323 
10.9023
3.21941
 1.04493
1.06193
1.04365
0.0101589
0.0167812
0.0136594

ssxptar 05/31/06

I tried increasing the ssxptar cut from ssxptar<0.04 to  ssxptar<0.045.  John A. pointed out that we should let the collimator define the acceptance.  However, the normalized yields increased by less than 1% with this change and the yield ratios did not change to 5 decimal places.

I then tried reducing the ssxptar acceptance to ssxptar<0.020 to check the tails in the hsxptar distributions.
Nominal cuts, abs(ssxptar)<0.040

 New cuts, abs(ssxptar)<0.020

It appears that the tails in the hsxptar distribution move with the new cut.  Therefore, these tails are most likely due to resolution, and should not be cut.

BCMs, Form factor parameterization, etc. 05/31/06

I tried a few more things.  I removed the hsxptar cuts described below (05/31/06), and then used the XEM BCM calibrations (the summer global fit for our July data and their late November fit for our December data).
Q2 (GeV2)
 Ydata/Ysimc before
 Stat. error
 Ydata/Ysimc after
 Stat. error
4.511
6.1  
7.47 
1.04506
1.06413
1.0766 
0.0101609
0.0168231
0.0140541
 1.04493
1.0652 
1.06693
0.0101589
0.0168329
0.0139641

I then changed the form factor parameterization (John A.'s Rosenbluth form factors), because I noticed there was a mistake.  Above Q2=6 GeV2, the G_E elastic form factor needs to be set to G_dipole (or what appears to be very close to it).  This form factor does not contribute significantly at large Q2, however, the parameterization may become very large in this region.

Results with the corrected form factor parameterization
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
4.511
6.1  
7.47 
1.04493
1.06193
1.04365
0.0101589
0.0167812
0.0136594

I also tried Peter Bosted's parameterization again.  However, the yield ratios all moved up by <1%.

Here is the plot of the ratio after changing back to John A.'s parameterization.  Note- the nominal hsxptar acceptance was used here, and in the previous two tables.

Hsxptar cuts 05/31/06

I tried using hsxptar cuts to remove the long tails in the experimental data hsxptar distributions.  Previously, I had hsed hsxptar cuts to remove the edges of both the SIMC and data distributions.

This is how hsxptar looks with the new cuts :
abs(hsxptar+0.0025)<0.0225, at the middle and low Q2
abs(hsxptar+0.00125)<0.0175, at the highest Q2


The yield ratios with the new cuts are


Hsxptar cuts 05/30/06

The hsxptar cut abs(hsxptar)<0.15 seems to make the data yield agree better with SIMC.

I didn't really get anywhere with this, but this is what I have so far.
1) The coplanarity distribution is broader in the data compared to simc.   I am not sure why -see the work below (05/29/06).
2) Both cuts hsxptar<0.15 and hsxptar>-0.15 increase Ydata/Ysimc, however, the former is worse.
Nominal hsxptar cuts
Q2    Ydata/Ysimc   Stat error
4.511  1.04506        0.0101609
6.1    1.06413        0.0168231
7.47   1.0766         0.0152574

hsxptar<0.015  ... This has the worst problems
Q2    Ydata/Ysimc   Stat error
4.511  1.02616        0.0101572
6.1    1.04258        0.0166829
7.47   1.06588        0.01399

hsxptar>-0.015
Q2    Ydata/Ysimc   Stat error
4.511  1.01099        0.0106755
6.1    1.05829        0.0173991
7.47   1.02881        0.0138151

Tight cut on hsxptar (abs(hsxptar)<0.15)
Q2    Ydata/Ysimc   Stat error
4.511  0.988393       0.0106613
6.1    1.03519        0.0172414
7.47   1.01777        0.013746
3) The events with hsxptar<-0.15 seem to be reconstructed correctly
Q2
W
xbj
Em
Many plots for Q2 = 4.511   6.1    7.47   GeV2

Systematic error 05/30/06

The significant contributions are:
Proton absorption:  0.5% random and 2.0% scale
HMS fid effic:  1% random and 1.0% scale
SOS fid effic:  0.5% random and 0.5% scale
Radiative corrections: 0.5% random and 1.0% scale
Acceptance: 2.0% scale
Total systematic uncertainty: 1.3% random and 3.2% scale

Model dependence: 2-3% at the lowest Q2, and 3-4% at the other two points.

In the plot below, the solid line represents the random and scale uncertainty added in quadrature, while the dashed line represents the random systematic uncertainty.  The model uncertainty has not been included.  The error bars on the data points are the statistical uncertainty.

When an extra cut is included to remove the edges of the hsxptar distributions in both the data and SIMC, abs(hsxptar)<0.015, the ratios become

Coplanarity 05/29/06

In an attempt to find out why the hsxptar distributions have longer tails in the data compared to SIMC (see "Rebinned hsxptar plots 05/29/06"  below), I plotted
lab phi of the particle in the SOS - lab phi of the particle in the HMS - 180 deg
which should be zero.
The Lab phi angle was calculated from
      myhsphi=atan2(hsxptar, (sin(hs_spec_angle)-hsyptar*cos(hs_spec_angle)) )
      myssphi=atan2(ssxptar, (-sin(ss_spec_angle)-ssyptar*cos(ss_spec_angle)) )
      if (myssphi<-1.5) then
         myssphi=myssphi+2.0*3.14159265358
      endif
where, hsxptar has, in addition, the appropriate oop offsets.

In any case, these plots of the coplanarity show that the width of the SIMC distribution is smaller than the data by about 0.5 degrees, which corresponds to 9 mrad.  This discrepancy could be due to (1) resolution (2) multiple scattering (3) peaking approx.

Proton absorption 05/29/06

Ok, after correcting for the proton absorption once only ...
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.781453
0.988393
1.03519 
1.0184  
0.00396284
0.0106613
0.0172414
0.0137499

However, the results above use the nominal acceptance cuts, with a tight cut on hsxptar
abs(hsxptar)<0.015

When I use the nominal acceptance cuts (without the tight hsxptar cut),
abs(hsyptar)<0.04.and.abs(hsxptar)<0.07.and.abs(hsdelta)<8.0
abs(ssyptar)<0.07.and.abs(ssxptar)<0.04.and.abs(ssdelta-2.5)<12.5
I obtain the following yield ratios
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.817759
1.04506 
1.06413 
1.07826
0.00364188
0.0101609
0.0168231
0.0140651


HMS Cerenkov cut 05/29/06

I tried using a hcer cut to remove pions, without any corrections for the good protons removed due to knock-on electrons and Cerenkov blocking.  The cut was hcer_npe<0.3.
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
4.511
6.1  
7.47 
1.0177 
1.06676
1.04622
0.0111092
0.0179923
0.0143205

However, e-pi coincidences should not be present due to the cointime cut, and the W cut, abs(W-0.94)<.1.

Furthermore, knock-on electrons have been ignored.  Based on the rejection ratio measured for pions in the HMS Cerenkov for the cut hcer_npe>0.5, which was around 50:1, the expected fraction of protons that produce knock-on electrons is about 2%.  This is about the size of the effect seen above. 

Jason's write-up on knock-on electrons is here
http://hallcweb.jlab.org/xmptlog/0409_archive/040902160325.html

Hee'p offsets 05/29/06

The plots for the previously fitted heep offset results (see 05_28_06 below) are shown below.  SIMC with radiation, eloss, etc turned on is shown in blue.  The data are the green and red histograms, which should overlap.

Q2=0.633 GeV2
Q2=4.51 GeV2
Q2=6.1  GeV2
Q2=7.47 GeV2

The offsets were:

Ebeam(GeV)
 P hms(GeV)
 TH hms(deg)
 P sos(GeV)
 TH sos(deg)
 dE
 dth_ss
 dp_ss
 dth_hs
 dp_hs
 dssxptar
 dhsxptar
2.038
 -1.692
 25.005
 0.874
 54.90
 0.59
 0.0
 0.48
 0.6
 0.9
 -0.0032 -0.0011
5.012
 4.116
 18.70
 -1.7227
 50.00
 0.79
 -1.0
 11.0
 0.6
 0.9
 -0.0032 -0.0011
5.766
 4.867
 16.19
 -1.6868
 51.61
 0.2
 0.0
 8.11
 0.0
 0.9
 -0.0032
-0.0011
4.021
 3.232
 22.15
 -1.588
 50.00
 -0.75
 -1.0 3.85
 1.0
 0.9
 -0.0032
-0.0011
The format of the offsets "d[variablename]" are the same as those in "heepcheck.f".  For example, for the electron in the sos
         p_e = ssp*(1-dp_ss/1000.0)
         th_e = mysstheta-dth_ss/1000.0  [rad]
         p_p = hsp*(1-dp_hs/1000.0) 
         th_p = myhstheta-dth_hs/1000.0   [rad]
         newE=e_beam*(1+dE/1000.0)
         newssxptar= ssxptar- dssxptar  [rad]
         newhsxptar=hsxptar-dhsxptar  [rad]

Acceptance edges 05/29/06

Previously (see 05_24_06 below), I tried using cuts to select a narrow region in the center of the acceptance.  Now, the effect of removing events near the edges of the acceptance was measured.
Q2 (GeV2)
 Ydata/Ysimc before
 Stat. error
 Ydata/Ysimc after
 Stat. error
0.633
4.511
6.1  
7.47 
0.826934
1.04592 
1.09544 
1.07767
0.00419348
0.0112818
0.0182449
0.0145502
0.793018
1.05639 
1.07809 
1.08236
 0.00424169
0.0118871
0.0193476
0.0158556

The cuts used were:
abs(hsyptar)<0.03.and.abs(hsxptar)<0.015.and.abs(hsdelta)<6.0
abs(ssyptar)<0.055.and.abs(ssxptar)<0.03.and.abs(ssdelta-2.5)<10.


Rebinned hsxptar plots 05/29/06

The hsxptar distributions in the last three Q2 settings appear rounded, but the shape is not too different from the SIMC (blue) distributions.

Calculation of the data norm. yields 05/29/06

The links below show the values of the the efficiencies, etc. used in the calculation of the normalized yield.  An example is provided below.

list1
list2
list3
list4

setting names 
el1lh2, Q2 = 0.633 GeV2
el2lh2, Q2 = 4.511 GeV2
el3lh2, Q2 = 6.1   GeV2
el4lh2, Q2 = 7.47 GeV2

Example, run 49537 (Q2=6.1 GeV2, electron in the SOS):
yield = synccorr*blockcorr/scereff*cpre/ctrg*hslowfid*scerfid
      /(1-helecdt/100)/(1-selecdt/100)/Tproton*(pcounts-bcounts)
= 1.0*1.0/0.997471*1356/1356*/0.9300/0.9929
     /(1-0.0001)/(1.0)*1.058*(627-0)
= 720.291
Which can be seen to be equal to the value determined by the analysis code (720.298).

Definitions:
scereff = efficiency of the sos cerenkov
hslowfid = fid efficiency of the HMS for slow particles (protons).  The cuts used were (1) small or zero hcer signal (2) calorimeter signal above a low threshold (3) event type was a coincidence.
scerfid = fid efficiency of the SOS with a cut to select large scer and calorimeter signals
helecdt = electronic dead time (%)
1/Tproton = 1/(proton transmission),  the correction for proton absorption

The normalized yield for this setting is determined from the corrected bcm2 (ie. cbcm2).  Charge cuts were used for run 49537,because there was a trip in this run.  Most runs do not have trips, and so cbcm2 = bcm2.
norm yield = 720.291/ (72000/1000) = 10.004 counts/mC

Plots showing the normalized yield for each run.

The setting averaged normalized yields (with dummy subtraction) are:
Q2 (GeV2)
 xnyield
0.633
4.511
6.1  
7.47 
73578. 
43.1746
10.3873
3.14522


More checks 05/28/06

The results did not change significantly when the sync filter was used
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.828136
1.04829 
1.10506 
1.05668 
0.00419797
0.0112964
0.0183294
0.0143972

I then played with the heep offsets a little more to get the peaks centered even closer to SIMC.
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.826607
1.04597 
1.09549 
1.07767 
0.00419182
0.0112823
0.0182457
0.0145502

Here are some plots of ssshtrk (electrons in the SOS) and hcer_npe (protons in the HMS), and this is why these cuts on these variables have not been applied in the analysis.  The background from misidentified ep elastic events appears to be on the % level, and it seems that using cuts on these variables will unecessarily complicate the analysis (one needs to coincider Cerenkov blocking and knock-on electrons, for example).
Q2= 4.511 GeV2
Q2= 6.1     GeV2
Q2= 7.47   GeV2

More checks 05/27/06

The analysis was redone using Tanja's offsets (Table 3.2, page 107 of her thesis).  dphms=-0.13% and dtheta = 0 in both the hms and sos.
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.884794
1.1     
1.12457 
1.10806 
0.00392917
0.0107005
0.017709 
0.0146358

Then, I refit the offsets beginning with Dave's offsets  (dphms=-0.09% and dthetahms=-0.6mrad), and made some small adjustments to get the peaks to line up with simc.
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.87798
1.10472
1.13634
1.10915
0.00389827
0.0107451
0.0178939
0.0146505

Based on the entry below on 05/26/05, a tight hsxptar cut was tried to remove the long tails seen in the data.  The highest Q2 has the thinnest hsxptar distribution so the cut was:
abs(hsxptar)<0.15
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.839753
1.0484  
1.10683 
1.056
0.00424456
0.0112971
0.0183449
0.0148575


More checks 05/26/06

Latest hsxptar distributions (first three are the delta scan, last three are the smallest Q2 to largest Q2 settings).  Blue = simc.


Check of the random coincidences.  There are no random coincidences.
Large Q2

Small Q2


Cuts used in the hee'p analysis 05/26/05


Data cuts
PID and missing mass cuts
      (scer_npe>1.35.and.Em<0.025.and.abs(W-0.94)<0.1)
Acceptance
      abs(hsyptar)<0.04.and.abs(hsxptar)<0.07.and.abs(hsdelta)<8.0
      (abs(ssyptar)<0.07.and.abs(ssxptar)<0.04.and.abs(ssdelta-2.5)<12.5)
Jochen's cut (Pg 80 of his thesis)
       (ssyptar>-125.0+4.25*ssdelta+64.0*ssytar-1.7*ssdelta*ssytar)
       (ssyptar<125.0-4.25*ssdelta+64.0*ssytar-1.7*ssdelta*ssytar)
Coincidence blocking and sync cuts (values from the database)
       hsrawct>[myblockcut].and.(hsrawct+ssrawct)>[mysynccut]
 Angle cointime cuts (protons in the SOS)
       abs((cointime)-[a])<[b].and.(ssbeta)>[b1min].and.(ssbeta)<[b1max]
       (ssbeta)>[b2min].and.(ssbeta)<[b2max]
       (cointime)-[a]>-[b]+[m]*(-0.5)*((ssbeta)-[c]-abs((ssbeta)-[c]))
       (cointime)-[a]<[b]+[m]*(-0.5)*((ssbeta)-[c]-abs((ssbeta)-[c]))

SIMC cuts
      Em<0.025.and.abs(W-0.94)<0.1
      abs(hsyptar)<0.04.and.abs(hsxptar)<0.07.and.abs(hsdelta)<8.0
      abs(ssyptar)<0.07.and.abs(ssxptar)<0.04.and.abs(ssdelta-2.5)<12.5
     (ssyptar>-125.0+4.25*ssdelta+64.0*ssytar-1.7*ssdelta*ssytar)
     (ssyptar<125.0-4.25*ssdelta+64.0*ssytar-1.7*ssdelta*ssytar)
 


The latest Y_data/Y_SIMC 05/25/06

Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.87454
1.11799
1.15437
1.17445
0.00388384
0.0107887 
0.0179889 
0.0152029


Checks of elastic analysis 05/24/06


I checked the data normalized yields using the paw interface (independent of my existing kumacs).  The data yields I have been using are in reasonable agreement with the  simplified analysis.  One place to start looking may be in the HMS fid effic, which was 0.93 at some settings.  If this is artificially small, it may partially explain the higher data normalized yields.  Otherwise, I can not see any 10% discrepancies in the analysis of the experimental data.  I used the hadron fid (87-95%) for the proton arm and the electron fid (99.3-99.5%) for the electron arm.

I replayed the elastic runs with slightly different fid track effic calculations.  I required that there is a small (non-zero) signal in the hadron-arm calorimeter.  The fid effic was essentially unchanged.  As before, I still use the coincidence hadron fid efficiency (Cerenkov < 0.2 npe) for the hadron arm.

The W and xbj plots between data and SIMC have reasonable agreement.  This suggests that the PID cuts are reasonable.
W plots for Q2=0.633-0.64 and for Q2=4.511-7.47
xbj plots for Q2=0.633-0.64 and for Q2=4.511-7.47

Next, I checked that the infiles kinematics were set correctly.  This could be done by plotting hsp (data vs SIMC), ssp, hstheta and sstheta.  If the central kinematics were set incorrectly, these histograms will not match.  The results below show that the central kinematics were, in fact, set correctly.  The plots below are normalized so that SIMC has the same area as the data.
de4lh2, Q2 = 0.633
de5lh2, Q2 = 0.637
de6lh2, Q2 = 0.64
el2lh2, Q2 = 4.511
el3lh2, Q2 = 6.1
el4lh2, Q2 = 7.47

Increasing the generation volume in SIMC (more events outside of the acceptance) did not seem to change the results:
Q2 (GeV2)
 Ydata/Ysimc before
 Stat. error
 Ydata/Ysimc after
 Stat. error
0.633
0.637
0.64 
4.511
6.1  
7.47 
0.835908
0.959927
0.87903 
1.06541 
1.09743 
1.15238
0.00294076
0.00404626
0.00800635
0.00910868
0.0148896
0.012746 
0.838134
0.962656
0.870977
1.06562 
1.1014  
1.14734 
0.0029529
0.00406916
0.00794429
0.00911016
0.0149438
0.0126902

Using an Em cut of EM<25MeV (applied to both the data and simc) helped a little, expecially with the highest Q2 point, but there are still problems.  The delta scan runs are now worse.
Q2 (GeV2)
 Ydata/Ysimc before
 Stat. error
 Ydata/Ysimc after
 Stat. error
0.633
0.637
0.64 
4.511
6.1  
7.47 
0.835908
0.959927
0.87903 
1.06541 
1.09743 
1.15238
0.00294076
0.00404626
0.00800635
0.00910868
0.0148896
0.012746 
0.809157
0.903222
0.784015
1.05697 
1.09139 
1.11027 
0.00306525
0.00406864
0.00750922
0.0101257
0.0168878
0.0142297

I finally got somewhere!  I cut on only the central part of the acceptance, and the yield ratio came out very close to one.  The only runs with problems are the delta scan runs, that are outside of the central part of the acceptance.
Q2 (GeV2)
 Ydata/Ysimc before
 Stat. error Ydata/Ysimc before Stat. error
4.511
6.1  
7.47 
1.05697 
1.09139 
1.11027 
0.0101257
0.0168878
0.0142297
1.00957
1.02392
1.03172
 0.0225653
0.0354379
0.0281909
However, it seems that this was just a statistical fluctuation.

I found a problem with the way I was implementing the proton absorption.  Previously, I applied to absorption correction (1/0.945) to the data whenever histograms were produced, and not when the yield ratios were calculated.  As the SIMC histograms were normalized to match the data, the absorption correction was having no effect.  After implementing the correction, the yield ratios are now worse.
Q2 (GeV2)
 Ydata/Ysimc
 Stat. error
0.633
4.511
6.1  
7.47 
0.87454
1.11799
1.15437
1.17445
0.00388384
0.0107887 
0.0179889 
0.0152029



 
Elastic H(e,e'p) yield check 05/23/06

SIMC was run again, with John A's parameterization, and plots of spectrometer quantities were produced.  All SIMC plots are normalized to match the data.  There was almost no change in the ratio of the yields with the new form factor parameterization.

de4lh2, Q2 = 0.633
de5lh2, Q2 = 0.637
de6lh2, Q2 = 0.64
el2lh2, Q2 = 4.511
el3lh2, Q2 = 6.1
el4lh2, Q2 = 7.47

More plots vs. spectrometer variables are here
de4lh2, Q2 = 0.633
de5lh2, Q2 = 0.637
de6lh2, Q2 = 0.64
el2lh2, Q2 = 4.511
el3lh2, Q2 = 6.1
el4lh2, Q2 = 7.47

Elastic H(e,e'p) yield check 05/23/06


Elastic runs were analyzed and simulated with SIMC.

The data normalized yields were divided by the proton transmission (0.945 in the HMS and 0.95 in the SOS).

Peter Bosted's parameterization of the form factors was used in these results (and changed to John's parameterization in subsequent analysis).

The settings are  (table of settings)
* Delta scan for elastic e,e'p, Q2 = 0.646 GeV2, beam=2038, Electron in HMS
 de4lh2 48865 48866
de5lh2 48870
de6lh2 48872
 
* electron in SOS, Q2 = 4.56 GeV2, beam=4021
 el2lh2 49874 49875 49876 49877
* electron in SOS, Q2 = 6.17 GeV2, beam=5012
 el3lh2 49532 49533 49534 49535 49536 49537
* electron in SOS, Q2 = 7.37 GeV2, beam=5766
 el4lh2 52394 52395 52396 52397 52398 52399 52401 52402 52403 52411 52412 52413 52414 52415 52416



The first results for the yields are shown below
Q2     Ydata/Ysimc  Stat.Error
0.633  0.842488        0.0029639
0.637  0.967865        0.00407972
0.64    0.886777        0.00807693
4.511  1.07136          0.00915955
6.1      1.1012            0.0149407
7.47    1.1361            0.0125656


Plots of Q2 for the settings above, in this order, are shown below.  Red=data, Blue=simc.  The agreement between data and SIMC is poor, and the error bars for the data in the first three plots appear to be too small compared to the point-to-point variation.  The SIMC plots have been normalized to match the data.