| Jan 11, 2007 | Jan 18, 2007 | Feb 2, 2007 | March 1, 2007 | March 22, 2007 |
| April 5, 2007 | April 12, 2007 | May 3, 2007 | May 9, 2007 |
Draft of PrimexNote (ps and pdf).
Here are a first study of single-arm compton analysis. For these plots I analyzed 1 file of run 5003.
I applied the following cuts for event selection:
To clean up the spectra, I also add a timing cut.
Here is the beam photon energy spectrum.
Here is a plot of the ratio of the measured cluster energy by the calculated energy of the scattered photon from compton.
Here is the timing plot with the different cut regions.
The time difference distribution is not centered at zero. The offest is ~1ns.
Here is a fit to the sideband subtracted distribution.
Here are plots of x vs y in various bins of Emeas/Ecalc.
Here are many plots of Emeas/Ecalc for a given HYCAL id.
Here are the plots from the 4 corners of the figure above.
 |
 |
 |
 |
Here is a HYCAL element directly next to the beampipe.
I fixed a bug in my code in the calculation of z given the cluster energy, beam energy, x, and y. Here is a plot of z.
Here is a plot of Emeas/Ecalc vs z after the fix.
Here is a new plot of Emeas/Ecalc. The left plot is the same as above. The right plot contains the distribution after the timing cut (red) and after a cut along the recalculated z. The cut is 625cm < z < 825cm.
I analyzed 1486 of the 1500 files in Eric's pi0 analysis. Below are some plots of the stability of the Emeas/Ecalc over the run range.
For each file, I produced a plot of the energy ratio. I integrated the counts in 3 regions in the ratio: 0.5-0.9, 0.9-1.1,and 1.1-1.5. Then, I divided the counts in regions 1 and 3 by the counts in region 2. Here is an example from run 5003.
Here are histograms of these ratios of counts for all files. In the righthand plot, the black and red histograms are for runs less than and greater than 5100.
Here are plots of these ratios of counts by run number.
Here are individual histograms for different runs.
| File 5003-00 | File 5067-00 |
|
 |
| File 5068-00 | File 5159-00 |
 |
 |
I am tuning up my analysis code to fit the quantity Emeas/Ecalc for various angle theta bins. The cuts are same listed above for veto, hycal total sum trigger, and timing. I also subtract out-of-time background with the sideband technique. The results on this page are for 1 run which contains 9 files.
Here are the timing distributions for each theta bin.
Here are the means and the sigmas from the fits to the timing distributions as a function of theta bin.
| Mean | sigma |
 |
 |
Here are the plots of Emeas/Ecalc for each theta bin. The black and blue histograms are in-time and out-of-time sideband events, respectively.
Here are the fits to the background-subtracted Emeas/Ecalc distributions. For the fits, I used a gaussian for the peak and a 2nd-order polynomial for the background.
This week I am studying the hits in the PS in order to match the e- in the PS with the Compton photon in HYCAL. The plot below is the distribution of PS hits with a front/back coincidence per event.
These plots show the reconstructed PS time for hits on the left and right sides.
These plots show the number of PS hits per event for those hits in the timing peak (t=0-20ns).
These plots show the time difference between the PS hits with good timing and the time of the photon from the tagger.
This week I am making cuts with respect to which side of the PS was hit. I analyzed a single run, 5003. The cuts I applied are photon timing, veto, and PS side hits. In this run, the e- predominantly was detected in the right side.
This week I extracted the yield for the single-arm compton and plotted the normalized ratio.
I determined the yield by fitting the peak of the ratio of Ecluster to ECompton. The angular range was from 0.5o to 1.3o. I fit the peak with a gaussian and the background with a 2nd order polynomial. I subtracted the background shape and integrated the counts under the peak.
To get the number of beam photons, I used the pflux code on each file.
The plots below show the yield divided by number of photons for each run. In the right plot, I zoomed on the good runs.
 |
 |
I am wondering if the slope in the right plot above is coming from the number of beam photons. Below is a plot of the tagging ratio from Aram's talk at the June Collaboration meeting. I plan to discuss this with him.
The next step is to apply an acceptance correction and number of target centers. Also, I need to calculate the cross section from the Klein-Nishina formula.
It was suggested at the Friday Primex meeting to redo the normalized yield analysis with out the veto requirement. These next two plots still have the timing sideband subtraction.
Here is a plot of the energy ratio for run 5003 with(red) and without(black) the veto cut.
Here is the normalized yields with(blue) and without(red) the veto cut.
The 2nd suggestion was to use sidebands fromt he timing difference distribution which are not next to the main timing peak.
These plots show the energy ratio for each sideband. The blue is for nearest sidebands, the red is for next-nearest sidebands, and the black in the right plot is the unsubtracted yield from the timing peak.
 |
 |
Here is the normalized yields with nearest sidebands(blue) and next-nearest sidebands(red).
Next I tried to fit the spectrum before the timing sideband subtraction. A better background function is needed.
Today, I analyzed the single-arm compton allowing for any number of clusters per eveny. Below is the comparison of the energy ratio for various cluster numbers per event.
 |
 |
Here are the distributions for the timing sidebands.
 |
 |
I analyzed all of the carbon target runs and accepted events with 1 or 2 clusters. I repeated the procedure of cutting on the timing peak, and subtracting the events from the timing sidebands. Then I fit the energy ratio with a gaussian and second order polynomial.
Here is an example of the fit to the energy ratio with events with 1 or 2 clusters per event.
Here are the normalized yields by run number. The blue(red) data are with(without) a veto cut. For the plot on the right, I zoomed in on the runs after 5100. In the analysis of only 1 cluster per event, there was a jump at around run 5220. Ilia said there was a change in beam current there. Now the distribtuion is flat. Aram's plot of tagging efficiency per run can be found on July 6, 2006
 |
 |
These distribtuions look like Aram's plots of the relative tagging ratio. I need to update my primex libraries to get the latest photon flux numbers and rerun the analysis.
I also coded the Klein-Nishina formula for the Compton differential cross section into ROOT. Here it is as a function of energy and angle theta. With ROOT, I think I can reliably numerically integrate it.
I checked out the latest CVS release, rebuilt my libraries, and reanalyzed the carbon data. Here is the latest normalized yields.
I analyzed two dedicated compton runs(5080 and 5150) to compare with the results of the pi0 production run 5003. I also analyzed the differential cross section into incident photon energy bins based on the T-counters. I ran my usual machinary except that the fits for each ratio of Emeasured over Ecompton
was done per T-counter bin. The energy for each bin was take from the center of the range of energies of each T-counter.Here are the fits per T-counter where the first plot(left,top) is the total events over t he entire energy range and the others are the individual T-counter statistics. The run was the pi0 production run 5003.
Same as above but for the dedicated compton run 5080.
Same as above but for the dedicated compton run 5150.
Here is a comparison of the differential cross section for runs 5003(red), 5080(black), and 5150(magenta). The line is my calculation of the Klein-Nishina cross section with respect to incident photon energy.
Since the cross sections between runs 5150(dedicated compton) and 5003(pi0 production) do not agree, I thought that the problem might be in the incident photon selection. So I analyzed run 5003 again, but this time I accepted all photons with a trigger bit 2 and calculated the energy ratio. The distributions per T-counter is below. The first histogram(upper left corner is the sum of all T-counters. I appear to get good fits.
Now I am comparing to my previous results. In the plot below, the line is my calculation of Klein-Nishina cross section, the blue points are from run 5150, and the red and black are from run 5003. The black are the old analysis, and the red are with accepting all incident photons.
This week I wanted to check my cuts for the analysis of the two different runs (5003-pi0 prod. and 5150-compton). I fit the energy ratio distributions before timing and sideband subtraction (cut 1), after the timing cut and before the sideband subtraction (cut 2), and after both the timing cut and the sideband subtraction. The fits are shown below where the left column is cut 1, the middle column is cut 2, and the right column is cut 3. The top row is without the veto cut, and the bottom row is with the veto cut.
Here are the fits for run 5003.
Here are the fits for run 5150.
Here are some ratios of various yields.
In the left plot, I am normalizing the yields from each cut by cut 1 from each analysis type. The black line is run 5150 w/o the veto, the red dashed line is run 5150 w/ the veto cut, the blue line is run 5003 w/o the veto cut, and the magenta dashed line is from run 5003 w/ the veto cut.
For the right plot, I made the ratio of the yields with the veto cut to the yields without the veto. The black line is run 5150 and the blue line is run 5003.
Comparing cross sections with the 2*pi factor included.
I found a silly bug in my calculation of the experimental cross sections. To calculate the number of target centers, I used 12(for A) instead of 6(for Z). I recalculated the cross sections and converted the units to mbarns.
I also used a different theoretical cross section. Yelena gave me the total cross section, and I multiplied it by the acceptance of 0.08. For the experimental cross sections, I removed the acceptance factor from it.
Compton cross sections from pi0 production run 5003(black and red) and compton run 5150(blue and magenta). The line is the theoretical total cross section multiplied by the acceptance.
The difference in the experimental cross sections between the compton and pi0 production runs comes from a change in the HYACL total sum trigger threshold which was changed in hardware. For the pi0 prod run the threshold was 2.0 GeV. For the compton run, it was lowered. I am showing plots of the cluster energy vs the energy ratio to illustrate how the threshold affected our data sample.
Plots of Ecluster vs. Ecluster/Ecompton for run 5003 (pi0 prod). The plot of the left is for in-time events, and the plot on the right is for the timing sideband events.
Plots of Ecluster vs. Ecluster/Ecompton for run 5150 (compton run). The plot of the left is for in-time events, and the plot on the right is for the timing sideband events.
This week's try is in calculating the differential cross sections. I calculated the differential Klein-Nishina cross section for a single angular bin at theta=0.8deg (the bin limits are 0.5deg to 1.3deg.).
Compton differential cross sections from pi0 production run 5003(red) and compton run 5150(green). The line is the Klein-Nishina differential cross section.
I am more confident that the dedicated compton run result is correct and am going to move on to refinements like better background subtraction, latest photon flux numbers, and a more detailed acceptance calculation.
Also, I am going to determine the trigger inefficiency in the pi0 prod. run.
I recalculated the photon flux for each run with the latest changes that Aram made to the pflux code. I believe he modified the relative tagging ratios for each T-counter. Below is a plot of the realtive single-arm compton cross section with the new flux.
Here is the same quantity but divided up by T-counter.
Next, I calculated the weighted average for each run range.
I am testing the effect of the angular range on the single-arm compton differential cross section. I performed my analysis on the pi0 production data run 5003 with various angular thresholds. The plot below shows the extracted yields normalized by the photon flux and acceptance for each angular range. The lower limit is always 0.5deg..
Here are plots of the individual T-counter fits for each angular range.
Here is a plot of compton kinematics for a 5.5GeV incident photon.
Next, I looked at the effect of the angular range on the normalized yields with respect to run number.
Angular range : 0.5 - 1.3deg.
Angular range : 0.5 - 1.1deg.
Angular range : 0.5 - 0.9deg.
| Full run range. | Zoom of second run range. | Zoom of third run range. |
 |
 |
 |
Angular range : 0.5 - 0.7deg.
Since the background is greatest around the beamline and decreases as the polar angle increases, I made some new fits after raising the lower angle limit to 0.6deg and 0.7deg. The upper limit on theta was set to 0.9deg.
Here are plots of the individual T-counter fits for the previous and new angular ranges.
At the last Friday meeting Yelena suggested running the analysis with a different threshold cut on the cluster energy. I placed a cut to accept only those events with a cluster above the software threshold. The fits to the energy ratio distributions are below. I left the angular range at 0.5 to 1.3deg.
Here are plots of the individual T-counter fits for each energy threshold.
I re-analyzed my pi0 production data and dedicated compton data for the angular range of 0.5-0.9 degrees. I also recalculated the geometric acceptance for both cases and got similar results of 0.063. Below is a comparsion of the pi0 production run and the compton run.
Something is still wrong. Another mistake is that the theoretical differential cross section from Klein-Nishina is about 25 mb/sr.
I tried the analysis but this time I extracted the yield with a gaussian shape for the peak and an exponential function for the background. I also tightenned the fit range to get better fits in both cases (polynomial background and exponential background shapes). Here are the fits for run 5003 (pi0 production). I get similar chi-squared values from each type of background fit.
2nd Order Polynomial background
This plot is a comparison of the reduced chi-squared for the compton run (red=polynomial, green=exponential).
Here is a comparison of the differential cross sections for run 5150 (compton) with the different background shapes (red=polynomial, green=exponential). The right plot is a ratio of the green point over the red points.
 |
 |
This plot is a comparison of the reduced chi-squared for the pi0 run (red=polynomial, green=exponential).
Here is a comparison of the differential cross sections for run 5003 (pi0) with the different background shapes (red=polynomial, green=exponential). The right plot is a ratio of the green point over the red points.
 |
 |
Below is a plot of the compton differential cross section for each pi0 prodcution run. The geometrical acceptance have been applied.
In the plots below, I broke up the data into three run ranges and plotted the projection of the data points for each run range. I also fit the distributions with a gaussian.
Next, I fit each run range with a flat line and calculated the chi-squared per degree of freedom. The fits are shown below.
The table below gives the chi-squared per degree of freedom for each run range.
| Run Range | chi2 | ndf | chi2/ndf |
| 1 | 140.3 | 17 | 8.3 |
| 2 | 285.5 | 70 | 4.1 |
| 3 | 271.3 | 41 | 6.6 |
I am improving the fit to the energy ratio in order to get better agreement between the dedicated compton run cross section and the pi0 production run cross section. I extracted the yield in a different way by integrating the gaussian peak shape used to fit the energy ratio. The previuos method was to subtract the expontential background shape from the total histogram.
Below are the fits by T-counter for the energy ratio. This run is a pi0 production run. The red-dashed line is the total fit, the blue is the histogram after the background shape was subtracted, and the magenta is the gaussian shape after the fit.
These are fits to the energy ratio for the compton run. The legend is the same as the plots above.
Here is the comparison of the differential cross sections by method of yield extraction.
|
 |
Here I am studying the photon multiplicity for the pi0 production run 5003.
The plot below shows the number of tagged photons per event with a valid trigger bit 2.
These plots are the difference in the timing differences (HYCAL to tagger) between the first tagged photon and the subsequent tagged photons.
These are plots of the energy difference between the first tagged photon and the subsequent tagged photons. The red histograms are events where the 2 photons have the same time difference.
From the study of the spread in the single-arm compton cross sections per run, I had fit the spread over the three run ranges. In this study, I summed over all T-counters. In the plots below, I show the mean and RMS for each T-counter(id=1-11) and all T-counters(id=0).
I also calculated the chi-squared for the distribution for each T-counter.
Here are some plots from my tuning up of the acceptance calculation.
Like Eric's analysis, I am showing a comparison of the energy deposited in HyCal (for reconstruction) and the energy written in the generated banks.
After reconstruction, I am showing the familiar relationship between the E-channels and T-channels.
Here is the reconstructed energy distribution of incident photons.
Here are plots of the scattered photons energy vs polar angle with generated (left) and reconstructed (right).
Here are plots of the energy ratio with generated (left) and reconstructed (right).
I found that the generated incident photon spectrum should look like the plot below.
When I use the incident photon energy from the generated bank instead of from reconstruction, the get the following energy ratio.
Here is a very preliminary acceptance per T-counter.
I changed my analysis scheme to select the cluster and tagger photon accoeding to a likelihood method. In this way, I choose only one cluster/tagger photon per event and avoid a combinatorial analysis. I calculated 2 probabilities - photon timing difference and energy ratio.
Here is a plot of the likelihood for the pi0 production run.
Here are the fits for each T-counter from the single-arm compton analysis of a Compton run.
Here are the fits for each T-counter from the single-arm compton analysis of a pi0 production run.
Here is the cross section comparison after the likelihood analysis.
I changed my angular range in an attempt to reduce the background.
Here are the fits per T-counter for the range 0.85 - 0.9deg.
I made 2 changes in my analysis since the collaboration meeting
Here is the cross section in the angualr range 0.5 to 0.9deg.
Here are fits per T-counter for pi0 run in the angular range 0.5 to 0.9 deg.
Here are fits per T-counter for compton runs in the angular range 0.5 to 0.9 deg.
I analyzed all of the dedicated Compton runs on the carbon target and found some bad runs. Here is a plot of the cross section per run. The runs that deviate from the average value are labeled "good" in the book_keeping database. However, from Yelena's analysis note, they were not included.
I tried to improve my signal to background by increasing the lower limit on the angular range. Here is a plot of the energy ratio with different angualr ranges.
Next, I focused on the angular range of 0.6-0.9deg to improve the signal to background.
Here are example fits using a skewed gaussian and a 3rd order polynomial.
Here is a comparison of the reduced chi-squared per T-counter for different lower limits on the background.
Here are example fits using a skewed gaussian and a 2nd order polynomial.
Here is a comparison of the reduced chi-squared per T-counter for different lower limits on the background.
Here are example fits using a skewed gaussian and a gaussian for the background.
Here is a comparison of the reduced chi-squared per T-counter for different lower limits on the background.
Here is the cross section for the pi0 run (5003) with different analyses.
I analyzed run 5115 with the same single-arm compton method. This run is the first empty target run with the pi0 production conditions.
Here are plots by T-counter comparing the pi0 production run against the empty target run. I scaled the empty target distributions to the other run in the range of 0.15 to 0.3 in hte energy ratio. I do not get good agreement is I use a different area of the distribution to set the scale factor.
The good news is that the distributions match well below 0.5. The bad news is that the background distribution does not match above 0.5. By subtracting the two data runs, I get a background which is not easy to model.
