Difference between revisions of "G2p Analysis Minutes"

Minutes of the weekly analysis meetings

9/16/2015

Present: Min, JP, Jixie, Melissa
By Phone: Vince, Jie, Chao, Pengjia, Alex, Ryan, Toby, Ellie

Feature Presentations:

• Jie
• Gave an update on his simulation study using MUDIFI. Last time he showed that he uses a
least square fit to go from endplane to endplane. For the g2p simulation, it's used to go from
(x,y,θ,φ,dp) in the target plane to (x,y,θ,φ,dp) in the focal plane. SNAKE produces (x,y,θ,φ,dp)
data in both planes and a polynomial function can be used to determine the relationship
between the two planes. First the maximum power for each variable is determined, symmetries
are included (if applicable), then the priority of each function must be determined. A Gram-
Schmidt transformation is used to do the minimization, and the final step is to decide whether
to accept the function fi. JP/Vince asked how the priority for each function is chosen, since the
choice could introduce some ordering dependence. Since we don't have enough data to
constrain, JP suggested introducing some physics. He also suggested getting rid of dp in the
focal plane (since it is really just dp in the target plane) so that there is one less dimension to
fit. More details can be seen in Jie's slides here.
• Ryan
• Gave an update on evaluating physics quantities using models. His goal was to do a
comparison of the hyperfine splitting calculation in C. Carlson's 2006/2008 paper using MAID
and Hall B models. Previously he showed a comparison using the Hall B model, where he saw
good agreement between his calculation and Carlson's result. For this time he also did a
comparison using the MAID model. There is a large difference between the two models for the
value of the integrand of Δ2, over the full range of Q2. JP suggested comparing the value of g2
between the two models, instead of Δ2. More details can be seen in his slides here.

General Discussion:

• Toby
• Last time he showed a comparison between data and cross sections calculated using the
Bosted model (seen here). The model was not continuous for one momentum setting. It
turns out the BPM was not recording for that setting, so he re-calculated the XS using
values for θ and φ from surrounding runs and found that the result was now consistent with
the neighboring momentum settings. He also played around with the acceptance cuts he
showed, and found a region using the "extreme cut" where the sawtooth pattern go away
and the yields become flat.

9/9/2015

Present: Min, JP, Jixie, Melissa
By Phone: Vince, Jie, Chao, Ryan, Toby, Karl

Feature Presentations:

• Jie
• Gave an overview of his simulation study using SNAKE and MUDIFI. Using SNAKE, the
idea is to trace the particle trajectory in a series of free boxes, then an endplane can be
defined within the free boxes to determine the trajectory from endplane to endplane. In
g2psim this will be used to trace the trajectory between the sieve plane and the focal
plane. There are two important issues to consider, the magnetic field in the free boxes,
and the transportation method between the two endplanes. A Runge-Kutta method is
used for the transportation, and an adaptive method (which changes with field size and
gradient) is used to determine the step size. For the magnetic field, a 3D Tosca field
map is used for the septum, and an analytic model is used for the spectrometer magnets.
He also described the method for choosing the transportation function to determine the
25 matrix terms for each kinematic variable (up to 4th order). JP asked how to determine
which order is more important, and cautioned to be careful not to over fit the data. He and
Jie will discuss this more offline. More details can be seen in his slides here.
• Chao
• Gave an optics status update. The current problem is that changes in the horizontal beam
position changes the dp reconstruction. This was fixed by applying a linear correction in
BPM-x and BPM-y to θ, φ and dp. Applying this correction made the data/simulation
agree, but it is still necessary to understand the correction. Using dp scan data with beam
position scan data still leaves an offset in θ and φ. He also tried two other methods to do
the calibration. Method 1 uses the central hole to determine a constant offset, uses the
horizontal beam position scan data to determine all the matrix elements related to y, and
uses dp scan runs to fit the additional matrix elements. Method 2 uses beam position
scan data to determine the matrix elements that are not related to xfp. Neither of these
methods works very well, even when the fitting order of the matrix elements are adjusted.
He will continue to check more fitting procedures. More details can be found in his slides
here.
• Pengjia
• Gave an update on comparing asymmetries from data/models for the 2.2 GeV, 5T
longitudinal setting. He first radiated the Bosted model XS; he showed the break down of
the different radiative corrections including the elastic tail (internal and external) and the
radiative effects from inelastic scattering. He showed a comparison with Ryan's results
which had good agreement, though there is some deviation when he used different proton
form factors (provided by Moshe). In order to compare the asymmetry result with data, he
needs to do the radiative corrections for the polarized MAID model, which he will continue
to work on. More details can be seen in his slides here.

9/2/2015

Present: JP, Min, Kalyan, Melissa
By Phone: Vince, Alex, Jie, Chao, Ryan, Toby, Ellie, Karl

Feature Presentations:

• Melissa
• Gave an update on determining the correction to asymmetries for pion contamination.
Previously, she used the 2.2 GeV, 5T longitudinal setting as an example, for this time
she used the transverse configuration, as it has the largest pion contamination. She
modified her method for differentiating between electrons and pions; previously she
identified pions as events that do not trigger the Cherenkov, this time she used a cut
below the single photo electron peak to select pions. JP suggested using stricter cuts
for selecting pions when determining the pion asymmetry, to make sure it is a clean
sample. Although this setting has a large pion contamination, the number of pion
events that remain after lead glass cuts are applied is low, so the overall correction
is very small. More details can be seen in her slides here.
• Min
• Gave an update on acceptance tuning. She looked at results from the 1.7 GeV setting
and found that the height of the peaks is different between data and simulation for
different holes in the sieve slit. Looking at the θ vs φ distribution, she noticed on the
positive θ side there were fewer holes in simulation than in data, and on the negative
θ side there were more holes in simulation compared to data. She tried adjusting the
aperture cuts by changing the physical location of the Q1, Q2 and septum planes to
uncover events that were previously blocked. After the adjustment, the data/simulation
results agreed much better. She showed a similar procedure for a dilution run and was
able to see better agreement between data/simulation after the aperture adjustments
were made. JP/Vince suggested being cautious when adjusting the aperture cuts, as
the apertures should be well defined already. They suggested adjusting things that are
less well known first, such as the septum field. More details can be seen in her slides
here.
• Ryan
• Working on generating physics quantities using models to eventually compare with data.
So far, he has incorporated the MAID2007 and Hall B models. He calculated the
hyperfine splitting quantities Δ1 and Δ2 using the Hall B EG1 model and compared them
to the results in C. Carlson's 2006 paper. His results agreed very well, he only saw a
small deviation for Δ2 in the lowest Q2 range. To check that nothing strange was going
on in the low Q2 range, he calculated Δ2 for various small ranges around zero, to ensure
that the integration method is well behaved. The Carlson paper also includes calculations
using the Simula model, he would like to do a check of this method as well, but is unsure
where to find the Simula model. Since he was able to reproduce the hyperfine splitting
result, he is confident in his calculation method; it should be straightforward to calculate
other spin structure function integrals. More details on his method can be found in his
slides here.
• Toby
• Showed an update on his acceptance/scattering angle study. He applied cuts of varying
strictness on the reconstructed θ vs φ distribution at the target. He then calculated the
real scattering angle before/after scattering for each cut. Using these scattering angles
as inputs for the Bosted model he was able to calculate the XS and compare it to data.
He showed the comparison between data/Bosted model for each acceptance cut. He
hoped that the more extreme acceptance cuts on the data would flatten the yield
distribution, but it actually just enhanced the sawtooth pattern. JP questioned why the
Bosted prediction for one momentum bin (the green one) was so much lower than the
neighboring ones. Toby said the only input that goes into the model is the scattering angle,
so it must be considerably different for this momentum setting. He also commented that
the model prediction is scaled to match the data, so the scaling enhances the jump in
the model. More details on his acceptance cuts and method can be found in his slides
here.

8/26/2015

Present: Jixie, Alex, Min, Kalyan, JP, Chao, Melissa
By Phone: Vince, Jie, Ryan, Toby, Karl

Feature Presentations:

• Chao
• Gave an update on optics studies. Previously, he showed a comparison between optics data
and simulation where he cut on individual sieve holes. For this week, he included a cut on y
and dp that he didn't include previously. In the plot of θ vs y, JP pointed out that there were
additional tails on some sieve holes, Chao says these tails go away when the φ cut is
included. He showed a 2D plot of θ vs φ, along with the 1D projection of these variables. JP
commented that the data is more smeared out than the simulation, so it is difficult to compare
data with simulation results. It's possible that some of the smearing comes from events that
punch through the sieve slit and are not being cut out (the simulation currently assumes these
events are stopped by the sieve). Looking at the center hole, the data/simulation agree
reasonably well, but the agreement gets worse for holes on the edge of the sieve slit. Chao
will continue to look into this. More details can be seen in his slides here.
• He also mentioned that ~1 month ago he reported a problem with the BPM simulation.
He found it was actually a problem with the simulation itself, which has now been updated
to correct the problem.

General Discussion:

• Pengjia has submitted his BPM NIM paper.

8/19/2015

Present: Chao, Jie, Min, Jixie, Kalyan, Melissa
By Phone: Vince, Ryan, Toby, Karl, Pengjia

Feature Presentations:

• Melissa
• Showed a method for determining the correction to asymmetries for pion contamination. The
measured asymmetry can be written in terms of the electron and pion asymmetries, multiplied
by the fraction of events of that particle. To differentiate between electrons and pions, a cut is
made on events that trigger the Cherenkov detector (electrons) and events that do not trigger
the Cherenkov (pions). Kalyan commented that this method of selecting pions may be too blind,
and suggested making a 2D cut on E/p vs Cherenkov instead. To determine the fraction of
electrons/pions, "good event" cuts are applied (loose acceptance cuts, single track events and
pion rejector cuts) and the number of electrons/pions that survive the cuts are counted. Although
the pion asymmetry is much larger than the electron asymmetry, the number of residual pions
is very small, so the overall correction is small. For this example the 2.2 GeV, 5T longitudinal
setting was used, but the correction may be larger for the 5T transverse settings, which she
will do next. More details can be seen in her slides here.
• Min
• Gave an update on acceptance studies. She used the 1.7 GeV, 2.5T setting (with 3rd septum
configuration) to test the acceptance cuts using the yield ratio of data to simulation. She
showed a plot of target θ vs φ for both data and simulation, and applied a graphical cut to get
rid of events on the edges. Looking at the resulting 1D plots shows reasonable agreement
between θ and φ. Vince pointed out that there seems to be some additional structure in the
dp distribution, and suggested expanding the δ cut out to &pm; 4% to see if there is more
hidden structure. Min said she has seen this in other settings as well and will look into it. She
also tried adjusting the cuts on θ, φ and δ and looked at how the ratio of yields from data/
simulation changed; the ratio starts to drop off as the cuts get wider. Next she will work on
calculating the yields using W bins. More details can be seen in her slides here.

8/12/2015

Present: Jie, Chao, Kalyan, Min, Jixie, Melissa
By Phone: Pengjia, Ryan, Toby, Karl, Ellie, Alex

Feature Presentations:

• Chao
• Gave an update on optics studies. Previously, Min has shown comparisons of data/simulation
where she looked at different columns of sieve holes. Chao continued this study by cutting on
individual sieve holes. Looking at the center sieve hole, the φ distribution looks reasonable, but
the θ distribution has some issues. Specifically, the data is wider on the left side (negative θ)
than the simulation. He showed results for several sieve holes, including one that was in the
row as the center hole, but in the farthest column on the left side of the sieve slit. In this case
the θ distribution had similar width for both data and simulation, but the height of the simulation
distribution was larger. This may suggest that something is wrong with the acceptance in the
simulation. The problem with θ seems to be systematic; for most of the sieve holes there are
more events on the left side of the data than in the simulation. Kalyan asked whether the
thickness of the sieve slit was taken into account, Chao says it was included in the simulation
and optimizer. Jixie suggested comparing the dp distributions for each sieve hole. More details
can be seen in Chao's slides here.

7/29/2015

Present: Chao, Min, JP, Melissa
By Phone: Alex, Pengjia

Feature Presentations:

• Min
• Gave an update on acceptance studies. Previously, Chao showed a correction for the beam
position in g2psim. Min checked this correction by looking at optics runs. On the data she
applied a cut on dp (to select elastic events) and on the focal plane (to get rid of junk
events). On the simulation, only an aperture cut was applied. She compared the data and
simulation results for θ and φ both before and after drifting from the sieve slit to the target.
The effect of the target field on θ is not very large, though there is a shift down in angle.
However, the shape of the θ distribution is different between data and simulation. The
effect of the target field on φ is significant; it's not clear why this is the case. To try to
determine the reason for the discrepancies, she applied a cut along each column of sieve
slit holes in φ and compared the data and simulation results for θ,φ and dp. Column 2
seems to have the best agreement (for dp), columns 1 and 2 have a shift while columns
3-6 have a difference in width. Chao suggested also cutting along each row of sieve slit
holes to better study the effect of the drift on φ. Min will work on identifying the differences
between data and simulation, specifically whether the boundaries are different between
the two. More details can be seen in her slides here.

7/22/2015

Present: JP, Min, Jixie, Melissa
By Phone: Alex, Ryan, Toby, Ellie, Karl, Pengjia

Feature Presentations:

• Melissa
• Showed various checks for packing fraction runs that have large discrepancies, including
beam position, beam energy, spectrometer central momentum setting, septum current, and
the location of anneals. For both the 2.2 and 1.1 GeV settings, there are significant shifts in
the x and y beam position (2mm or more). She showed the average and standard deviation
of epics readings for the LHRS septum current for each run. For several runs, there are small
number of epics readings that drift from the average current. It's not clear if this is the
current is actually changing or if it is a readout problem. She will look into the yield vs time
to try to determine this. JP suggested making a raster cut on the center of the beam to get
rid of any potential scraping. Pengjia says there is a way to cut on raster size, but there will
be an additional BPM uncertainty. Melissa will discuss this more offline with Pengjia. More
details can be seen in her slides here.
• Pengjia
• Gave an update on calculating asymmetries/cross sections using the MAID model. This time,
he used the Bosted model to get the unpolarized cross sections for the proton. Using unradiated
cross section models, he still saw a factor of 6 difference when compared with data. Ryan also
calculated asymmetries using unradiated models, and the results were consistent with Pengjia's,
so it seems the radiative effects contribute significantly to the factor of 6 difference. More
details can be seen in his slides here.

7/15/2015

Present: JP, Min, Chao, Jixie, Melissa
By Phone: Ryan, Toby, Ellie, Karl, Vince, Pengjia

Feature Presentations:

• Chao
• Doing a check of simulation/data results for the 2.2 GeV, longitudinal setting in order
to check that the optics calibration matrix will work for all momentum settings (not
just elastic). He showed a comparison for several different momentum settings.
Vince suggested simulating more events, since the statistics start to drop off for
lower momentum settings. In general, the data is broader then the result from
simulation. In the θ distribution, the "tails" seen in the data are not symmetric. JP
suggested carefully selecting cuts to remove any background from the data to
ensure a good comparison with simulation. More details can be seen in his
slides here.
• Pengjia
• More discussion on Pengjia's slides from last week. Pengjia has seen a factor of
six difference between his asymmetry calculated from data compared to the
asymmetry determined using the MAID model. Ryan did not see this factor of six
in his study, but in addition to using radiated XS models, he is using a combination
of the Bosted and MAID models (Pengjia is using just the MAID model). Karl
commented that the XS from MAID is not very good for our kinematics. Pengjia
also showed an estimated NH3 XS. JP commented that it was surprising that the
delta peak wasn't visible. Pengjia will look into this for next time.

7/8/2015

Present: JP, Min, Chao, Melissa
By Phone: Toby, Vince

Feature Presentations:

• Pengjia
• Working on comparing asymmetries determined from MAID to those calculated from
data; last time he showed a factor of 6 difference between the two results. This
time, he showed two different methods for calculating the differential XS from the
virtual photon XS. The first method calculates the differential XS directly from
the virtual photon XS, and the second method calculates the differential XS using
F1, F2, g1 and g2. He also showed two different methods for calculating the
asymmetry from the virtual photon XS; one method uses A1 and A2, while the other
is calculated from the virtual photon XS. There was a discrepancy between the two
methods, but he found that there was actually a mistake in one of the equations.
While the results from the two methods agree with each other, they are still a factor
of 6 larger than the results from data. More details can be seen in his slides here.
• Toby
• Showed an update of his scattering angle study for model reconstruction. He is
using BPM and optics data to determine the scattering angle. He showed an example
of the calculated scattering angle; JP questioned whether the range was really from
0-16 deg. Toby pointed out that there really aren't many events at the boundaries,
the distribution is peaked between 4.5-9 deg. He then uses this scattering angle as
input for the Bosted model. At low ν the simulation results match reasonably well
with the data, but the discrepancy starts to become larger as ν increases. He took
a closer look at the reconstructed scattering angle for each central momentum, and
found that a second peak is visible for smaller values of p0. If this second peak was
real, the simulation should recreate the data, so this could suggest a problem with
the reconstruction. However, Chao pointed out that, for this energy setting, the optics
calibration is not complete (currently using the longitudinal optics matrix), so the
reconstructed θ and φ may not be correct yet. For next time, Toby will test this
analysis using the 1.7 GeV setting, which has calibrated optics. More details can
be seen in his slides here.
• Melissa
• Gave a summary of packing fraction analysis. For some settings, the variation in the
yields (and Pf) is due to a fluctuation in the beam position. This will hopefully be
resolved by Jie's beam position/acceptance study. For the 2.2 GeV, 2.5T transverse setting
there is also an issue of a shift in ν between runs. For the 1.1 GeV, 2.5T transverse
setting, the yields vary depending on the beam current used for each run. JP suggested
a few things to check as the possible cause for these discrepancies including drift in the
BCM calibration, anneals of the target material, drift in field (HRS or septum), etc. She
has posted a technote draft here, feedback would be appreciated. More details can be seen in her slides here.
• Min
• Gave an update on her acceptance study. Last time, she showed that the simulation results
are more narrow than the data. For this time, she tried cutting on just the center hole,
and found that the dp distribution from data better matched the simulation. JP suggested
cutting on each of the different holes individually to determine which hole causes the dp
distribution to be wider. More details can be seen in her slides here.
• Chao
• Gave an update on optics analysis, currently checking whether simulation results match
with data, starting with optics data. First he checked whether the geometry was correct,
some deviation was found in the recent optics meeting. The position of the BPMs is hard-
coded into the simulation, and the drifting algorithm is used to drift the electrons
backward to the BPMs to simulate the readout of BPM A and B. He compared the θ
distribution before drifting in the target field using a fixed and non-fixed BPM location.
The results using a fixed BPM match well with the data. He is currently working on
checking this for all momentum settings, and will follow up with Pengjia about the BPM
problem. More details can be seen in his slides here.

7/1/2015

Present: Chao, Vince, Jie, JP, Min, Melissa
By Phone: Pengjia

Feature Presentations:

• Min
• Gave an update on acceptance studies. She showed a summary table of each energy
setting and septum configuration. There hasn't been any problems in the 2.2 GeV, 5T
longitudinal and 1.7 GeV, 2.5T transverse settings, but some settings (2.2 GeV, 2.5T
transverse, 1.2 GeV, 2.5T transverse and 2.2 GeV, 5T transverse) have a discrepancy
in the focal plane between data and simulation. She compared the data divided by the
Mott XS to the simulation results without XS, and found that the simulation results
of θ and φ are narrower then the data. JP commented that, while including the Mott
XS might change the shape of the distribution, it shouldn't change the boundary. She
work on figuring out the cause for this discrepancy before calculating the acceptance.
More details can be seen in her slides here.
• Pengjia
• Comparing asymmetries determined from the MAID model to those calculated from data.
Last time he showed a factor of 6 difference between the two results. He determined the
scattering angle and Q2 by fitting the data and compared them to the quantities calculated
from MAID. He also showed the results of calculating the quantities g1, g2, F1, F2, AL, AT,
dXSL, dXST, XStot andXSmott using MAID. He hasn't included radiative corrections in his
calculations, but this probably won't account for the factor of 6. More details can be seen
in his slides here.