Everything as a single pdf file (341 pages, 8.2Mb)
Pibero Djawotho
Indiana University
July 31, 2006
Simulation were done by Jason for the SVT review.
Figure 1: Fitted peak integral vs. fit residual sum (U+V) from st_jpsi input stream (J/psi trigger only).  Figure 2: Fitted peak integral vs. fit residual sum (U+V) from st_physics input stream (all triggers except express stream triggers). 
The separation between photons and pions was achieved by using Les cut in the above figures where photons reside above the curve and pions below. The data set used is the st_jpsi express stream.
The transverse profile of an electromagnetic shower in the SMD can be parametrized by the equation below in each SMD plane:
f(x) is the energy in MeV as a function of SMD strip x. The algorithm performs a simultaneous fit in both the U and V plane. The maximal residual (data  fit) is then calculated. A single photon in the SMD should be well descibed by the equation above and therefore will have a smaller maximal residual. A neutral pion, which decays into two photons, should exhibit a larger maximal residual. Typically, the response would be a double peak, possibly a larger peak and a smaller peak corresponding to a softer photon.
This directory contains images of single event SMD responses in both U and V plane. The file name convention is SMD_RUN_EVENT.png. The fit function for a single peak is the one described in the section above with 5 parameters:
Pibero Djawotho
Indiana University
August 4, 2006
The dataset used in this analysis is the 2005 p+p collision at √s=200 GeV with the endcap calorimeter hightower1 (eemcht1mb = 96251) and hightower2 (eemcht2mb = 96261) triggers.
The file catalog query used to locate the relevant files is:
get_file_list.pl keys 'path,filename' delim / cond 'production=P05if, trgsetupname=ppProduction,filetype=daq_reco_MuDst,filename~st_physics, tpc=1,eemc=1,sanity=1' delim 0
Pibero Djawotho
Indiana University
August 6, 2006
A detailed description of the EEMC slow simulator is presented at the STAR EEMC Web site.
The following settings were used in running the slow simulator:
// // Initialize slow simulator // StEEmcSlowMaker *slowSim = new StEEmcSlowMaker("slowSim"); slowSim>setDropBad(1); // 0=no action, 1=drop chn marked bad in db slowSim>setAddPed(1); // 0=no action, 1=ped offset from db slowSim>setSmearPed(1); // 0=no action, 1=gaussian ped, width from db slowSim>setOverwrite(1); // 0=no action, 1=overwrite muDst values slowSim>setSource("StEvent"); slowSim>setSinglePeResolution(0.1); slowSim>setNpePerMipSmd(2.0); slowSim>setNpePerMipPre(3.9); slowSim>setMipElossSmd(1.00/1000); slowSim>setMipElossPre(1.33/1000);
EEMC Fast Simulator 
EEMC Slow Simulator 
The plots that follow are sums of individual SMD responses in each plane centered around a common mean (here 0), over a +/40 strips range. The convention for the parameters in the fits below is:
The simulation is from single photons thrown at the EEMC with the following p_{T} distribution:
The highest tower above 4 GeV in total energy is selected and the corresponding SMD sector fitted for peaks in both planes, where the area of the peaks in the U plane is constrained to be identical to that of the peak in the V plane. The peak is shifted to be centered at 0 where peaks from other events are then summed. The summed SMD response in each plane is displayed below:
Ditto in log scale.
Ditto by sector.
Sector #  SMDu σ_{core}  SMDu σ_{tail}  SMDv σ_{core}  SMDv σ_{tail} 
Sector 1  0.869033 ± 0.0142868  3.42031 ± 0.10226  0.84379 ± 0.0185107  3.03287 ± 0.0775009 
Sector 2  0.814959 ± 0.0169271  2.99941 ± 0.0730426  0.889892 ± 0.0163065  3.35288 ± 0.0911979 
Sector 3  0.84862 ± 0.0148706  3.07648 ± 0.0909689  0.914377 ± 0.014706  3.72821 ± 0.0966915 
Sector 4  0.924398 ± 0.0144207  3.74458 ± 0.10611  0.888146 ± 0.0180771  3.06618 ± 0.0647075 
Sector 5  0.934218 ± 0.0163887  3.45149 ± 0.0944309  0.911209 ± 0.0175273  3.28633 ± 0.0890581 
Sector 6  0.797976 ± 0.0148133  3.20464 ± 0.0986085  0.822437 ± 0.018835  3.30595 ± 0.118813 
Sector 7  0.836936 ± 0.0150085  3.28589 ± 0.0853598  0.873338 ± 0.0173883  3.16654 ± 0.0838938 
Sector 8  0.828403 ± 0.0167005  3.05517 ± 0.075584  0.891045 ± 0.0152102  3.34806 ± 0.0836394 
Sector 9  0.832881 ± 0.0127855  3.3214 ± 0.0762928  0.8436 ± 0.0175466  3.0183 ± 0.079444 
Sector 10  0.804059 ± 0.0160906  3.0943 ± 0.0897946  0.874845 ± 0.015788  3.18113 ± 0.0748357 
Sector 11  0.930286 ± 0.0187086  3.40024 ± 0.0951671  0.854395 ± 0.0167265  3.21076 ± 0.0812402 
Sector 12  3.33227 ± 0.111911  0.848668 ± 0.0142344  0.895174 ± 0.0160939  3.48527 ± 0.12061 
In this sample, high tower triggers, eemcht1mb (96251) and eemcht2mb (96261), from the 2005 p+p at √s=200 GeV ppProduction are selected. The highest tower above 4 GeV is chosen and the corresponding SMD sector is searched for peaks in both planes. Peaks from several events are summed together taking care of shifting them around to have a common mean.
Ditto in log scale.
Here, I try to pick a representative sample of electrons from the 2005 pp200 dataset. The cuts used to pick out electrons are:
The selection for electrons is illustrated in the dE/dx plot below, where the pions should be on the left and the electrons on the right.
This study is motivated by Weihong's photon energy loss study where an etadependence of reconstructed photon energy to generated photon energy in EEMC simulation was observed.
In this study, the etadependence is investigated by running the EEMC slow simulator with the new readjusted weights for the preshower and postshower layers of the EEMC. Details on this are here.
The parameters from the fits are used to plot the fit functions for comparison between Weihong's and Pibero's results.
While the adjusted weights for the different EEMC layers contribute to bringing the ratio of reconstructed energy to generated energy closer to unity, they do not remove the etadependence.
In the plot below, I use the energy of the single tower (tower with max energy) presumably the tower the photon hit. The nonlinearity seems to disappear.
In the plot below, I use the energy of the 3x3 cluster of tower centered around the tower with the max energy. The nonlinearity is restored.
The plot below shows E_{tower}/E_{cluster} vs. eta where the cluster consists of 3x3 towers centered around the max energy tower.
Below is the profile of E_tower/E_cluster vs. eta.
The plot below shows the energy sampled by the entire calorimeter as a function of eta, i.e. sampling fraction as a function of eta.
Sampling fraction integrated over all eta's.
10k muons thrown by Will with:
E=1 GeV 
E=2 GeV 
Summary of Reconstructed/Monte Carlo Photon Energy 
Description 
The study presented here uses Monte Carlo data sets generated by Will Jacobs at different photon energies (5, 10, 20, 40, 80, 160 GeV):
For each photon energy, the ratio E_reco/E_MC vs. eta was plotted and fitted to the function p0+p1*(1eta), where E_reco is the reconstructed photon energy integrated over the
entire
EEMC. The range of the fit was fixed from 1.15 to 1.95 to avoid EEMC edge effects. The advantage of parametrizing the etadependence of the ratio in this way is that p0 is immediately interpretable as the ratio in the middle of the EEMC. The parameters p0 and p1 vs. photon energy were subsequently plotted for the EEMC fast and slow simulator.
EEMC Fast/Slow Simulator Results 
Conclusions 
The parameter p0, i.e. the ratio E_reco/E_MC in the midregion of the EEMC, increases monotonically from 0.74 at 5 GeV to 0.82 at 160 GeV, and the parameter p1, i.e. the slope of the etadependence, also increases monotonically from 0.035 at 5 GeV to 0.038 at 160 GeV. There appears to be a magic energy around 10 GeV where the response of the EEMC is nearly flat across its entire pseudorapidity range. The anomalous slope p1 at 160 GeV for the EEMC slow simulator is an EEMC hardware saturation effect. The EEMC uses 12bit ADC's for reading out tower transverse energies and is set for a 60 GeV range. Any particle which deposits more than 60 GeV in E_T will be registered as depositing only 60 GeV as the ADC will return the maximum value of 4095. This translates into a limit on the eta range of the EEMC for a particular energy. Let's say that energy is 160 GeV and the EEMC tops at 60 GeV in E_T, then the minimum eta is acosh(160/60)=1.6. This limitation is noticeable in a plot of E_reco/E_MC vs. eta. This anomaly is not observed in the result of the EEMC fast simulator because the saturation behavior was not implemented at the time of the simulation (it has since been corrected). The energydependence of the parameter p0 is fitted to p0(E)=a+b*log(E) and the parameter p1 to p1(E)=a+b/log(E). The results are summarized below:
EEMC fast simulator fit p0(E)=a+b*log(E) a = 0.709946 +/ 0.00157992 b = 0.022222 +/ 0.000501239 EEMC slow simulator fit p0(E)=a+b*log(E) a = 0.733895 +/ 0.00359237 b = 0.0177537 +/ 0.0011397 EEMC fast simulator fit p1(E)=a+b/log(E) a = 0.0849103 +/ 0.00556979 b = 0.175776 +/ 0.0138241 EEMC slow simulator fit p1(E)=a+b/log(E) a = 0.0841488 +/ 0.0052557 b = 0.187769 +/ 0.0130445
The following
from the IUCF STAR Web site gives a brief overview of the SMD gamma/pi0 discrimination algorithm using the method of maximal sided fit residual (data  fit). This technique comes to STAR EEMC from the Tevatron via Les Bland via Jason Webb. The specific fit function used in this analysis is:
f(x)=[0]*(0.69*exp(0.5*((x[1])/0.87)**2)/(sqrt(2*pi)*0.87)+0.31*exp(0.5*((x[1])/3.3)**2)/(sqrt(2*pi)*3.3))
x is the strip id in the SMDu or SMDv plane. The widths of the narrow and wide Gaussians are determined from empirical fits of shower shape response in the EEMC from simulation.
In the rest of this analysis, only those photons which have reconstructed pt > 5 GeV are kept. There is no requirement that the photon doesn't convert. The dividing curve between photons and pions is:
f(x)=4*x+1e7*x**5
The yaxis is integrated yield over the SMDu and SMDv plane, and the xaxis is the sum of the maximal sided residual of the SMDu and SMDv plane.
Following exchanges with Scott Wissink, the idea is to move from a quintic to a quadratic to reduce the number of parameters. In addition, the perpendicular distance between the curve and a point in the plane is used to estimate the likelihood of a particle being a photon or pion. Distances above the curve are positive and those below are negative. The more positive the distance, the more likely the particle is a photon. The more negative the distance, the more likely the particle is a pion.
Hi Pibero, With your new "linear plus quintic" curve (!) ... how did you choose the coefficients for each term? Or even the form of the curve? I'm not being picky, but how to optimize such curves will be an important issue as we (hopefully soon) move on to quantitative comparisons of efficiency vs purity. As a teaser, please see attached  small loss of efficiency, larger gain in purity. Scott
Hi Pibero, I just worked out the distance of closest approach to a curve of the form y(x) = a + bx^2 and it involves solving a cubic equation  so maybe not so trivial after all. But if you want to pursue this (not sure it is your highest priority right now!), the cubic could be solved numerically and "alpha" could be easily calculated. More fun and games. Scott
Hi Pibero, I played around with the equations a bit more, and I worked out an analytic solution. But a numerical solution may still be better, since it allows more flexibility in the algebraic form of the 'boundary' line between photons and pions. Here's the basic idea: suppose the curved line that cuts between photons and pions can be expressed as y = f(x). If we are now given a point (x0,y0) in the plane, our goal is to find the shortest distance to this line. We can call this distance d (I think on your blackboard we called it alpha). To find the shortest distance, we need a straight line that passes through (x0,y0) and is also perpendicular to the curve f(x). Let's define the point where this straight line intersects the curve as (x1,y1). This means (comparing slopes) (y1  y0) / (x1  x0) = 1 / f'(x1) where f'(x1) is the derivative of f(x) evaluated at the point (x1,y1). Rearranging this, and using y1 = f(x1), yields the general result f(x1) f'(x1)  y0 f'(x1) + x1  x0 = 0 So, given f(x) and the point (x0,y0), the above is an equation in only x1. Solve for x1, use y1 = f(x1), and then the distance d of interest is given by d = sqrt[ (x1  x0)^2 + (y1  y0)^2 ] Example: suppose we got a reasonable separation of photons and pions using a curve of the form y = f(x) = a + bx^2 Using this in the above general equation yields the cubic equation (2b^2) x1^3 + (2ab + 1  2by0) x1  x0 = 0 Dividing through by 2b^2, we have an equation of the form x^3 + px + q = 0 This can actually be solved analytically  but as I mentioned, a numerical approach gives us more flexibility to try other forms for the curve, so this may be the way to go. I think (haven't proved rigorously) that for positive values of the constants a, b, x0, and y0, the cubic will yield three real solutions for x1, but only one will have x1 > 0, which is the solution of interest. Anyway, it has been an interesting intellectual exercise! Scott
I made use of the ROOT function TMath::RootsCubic to solve the cubic equation numerically for computing distances of each point to the curve. With the new quadratic curve f(x)=100+0.1*x^2 the efficiency is 63% and the rejection is 82%.
The plot on the left below shows the efficiency of identifying photons over the pt range of 1030 GeV and the one on the right shows the rejection rate of single neutral pions. Both average about 75% over the pt range of interest.
The plot below shows background rejection vs. signal efficiency for different energy ranges of the thrown gamma/pi0.
Below on the left is a plot of the ratio of the sum of preshower 1 and 2 to tower energy for both photons (red) and pions (blue). On the right is the rejection of pions vs. efficiency of photons as I cut on the ratio of preshower to tower. It is clear from these plots that the preshower layer is not a good gamma/pi0 discriminator, although can be used to add marginal improvement to the separation preovided by the shower max.
ALL ENERGIES 

E=2040 GeV 

E=4060 GeV 

E=6080 GeV 

E=8090 GeV 

The plots below show the distribution of clusters in the endcap calorimeter for different partonic pT ranges. 2000 events were generated for each pT range. A cluster is made up of a central high tower above 3 GeV in pT and its surounding 8 neighbors. The total cluster pT must exceed 4.5 GeV.
Below is the pT of direct and decay photons from the Pythia record. Note how the two subsets are well separated at a given partonic pT. Any contamination to the direct photon signal would have to come from higher partonic pT.
Jet 
Gamma 
The gamma fitter runs out of the box. The code consists of the classes StGammaFitter
and StGammaFitterResult
in CVS. After checking out a copy of offline/StGammaMaker, cd into the offline directory and run:
root4star StRoot/StGammaMaker/macros/RunGammaFitterDemo.C
The following plots will be generated on the ROOT canvas and dumped into PNG files.
The class
computes the maximal sided residual of the SMD response in the u and vplane for gamma candidates. It is based on C++ code developed by Jason Webb from the original code by Les Bland who got the idea from CDF (?) The algorithm follows the steps below:
[0]*(0.69*exp(0.5*((x[1])/0.87)**2)/(sqrt(2*pi)*0.87)+0.31*exp(0.5*((x[1])/3.3)**2)/(sqrt(2*pi)*3.3))
/star/institutions/iucf/balewski/prodOfficial06_muDst/
The parameters of the isolation cut were suggested by Steve Vigor:
Hi Pibero, In general, I believe people have used smaller cone radii for isolation cuts than for jet reconstruction (where the emphasis is on trying to recover full jet energy). So you might try something like requiring that no more than 10 or 20% of the candidate cluster E_T appears in scalar sum p_T for tracks and towers within a cone radius of 0.3 surrounding the gamma candidate centroid, excluding the considered cluster energy. The cluster may already contain energy from other jet fragments, but that should be within the purview of the gamma/pi0 discrimination algo to sort out. For comparison, Les used a cone radius of 0.26 for isolation cuts in his original simulations of gamma/pi0 discrimination with the endcap. Using much larger cone radii may lead to accidental removal of too many valid gammas. Steve
This analysis is based on the work of Jan and Justin on SMD Profile Analysis for different TPC momenta. See here for a list of cuts. The original code used by Jan and Justin is here.
/star/institutions/iucf/balewski/prodOfficial06_muDst/
Figure 1: Number of tracks surviving each successive cut
Figure 2a: Number of tracks per trigger id for all electron candidates. Most common trigger ids are:
127652  eemcjp0etotmbL2jet  EEMC JP > th0 (32, 4 GeV) and ETOT > TH (109, 14 GeV), minbias condition, L2 Jet algorithm, reading out slow detectors, transverse running 
127271  eemcjp1mb  EEMC JP > th1 (49, 8 GeV) && mb, reading out slow detectors, transverse running 
127641  eemchttpmbl2gamma  EEMC HT > th1 (12, 2.6 GeV, run < 7100052;13, 2.8 GeV, run >=7100052) and TP > TH1 (17, 3.8 GeV, run < 710052; 21, 4.7 GeV, run>=7100052 ), minbias condition, L2 Gamma algorithm, reading out slow detectors, L2 thresholds at 3.4, 5.4, transverse running 
127622  bemcjp0etotmbL2jet  BEMC JP > th0 (42, 4 GeV) and ETOT > TH (109, 14 GeV), minbias condition, L2 Jet algorithm, reading out slow detectors, transverse running; L2jet thresholds at 8.0,3.6,3.3 
Figure 2b: Number of tracks per trigger id for all electron candidates for p_{T} > 4 GeV. The dominant trigger ids become:
127641  eemchttpmbl2gamma  EEMC HT > th1 (12, 2.6 GeV, run < 7100052;13, 2.8 GeV, run >=7100052) and TP > TH1 (17, 3.8 GeV, run < 710052; 21, 4.7 GeV, run>=7100052 ), minbias condition, L2 Gamma algorithm, reading out slow detectors, L2 thresholds at 3.4, 5.4, transverse running 
127262  eemcht2mbemul  EEMC HT > th2 (22, 5.0 GeV) && mb, reading out slow detectors, emulated in L2, transverse running, different threshold from 117262 
127271  eemcjp1mb  EEMC JP > th1 (49, 8 GeV) && mb, reading out slow detectors, transverse running 
127652  eemcjp0etotmbL2jet  EEMC JP > th0 (32, 4 GeV) and ETOT > TH (109, 14 GeV), minbias condition, L2 Jet algorithm, reading out slow detectors, transverse running 
Figure 3: p_{T} distribution of tracks before E/p, dE/dx and pT cut
Figure 4: p_{T} distribution of electron candidates with p_{T} > 4 GeV
Figure 5: η distribution of electron candidates (all p_{T})
Figure 6: φ distribution of electron candidates (all p_{T})
Figure 7: dE/dx of tracks before E/p and dE/dx cuts (all p_{T})
Figure 8: dE/dx of tracks before E/p and dE/dx cuts (p_{T} > 4 GeV)
Figure 9: dE/dx of tracks before E/p and dE/dx cuts (all p_{T} and 0.8 < η < 1.0)
Figure 10: dE/dx of tracks before E/p and dE/dx cuts (all p_{T} and 1.0 < η < 1.2)
Figure 11: dE/dx of tracks before E/p and dE/dx cuts (all p_{T} and 1.2 < η < 1.4)
Figure 12: dE/dx of tracks before E/p and dE/dx cuts (all p_{T} and 1.4 < η < 1.6)
Figure 13: dE/dx of tracks before E/p and dE/dx cuts (all p_{T} and 1.6 < η < 1.8)
Figure 14: dE/dx of tracks before E/p and dE/dx cuts (all p_{T} and 1.8 < η < 2.0)
/star/institutions/iucf/hew/2006ppLongRuns/
Figure 2.1
Figure 2.2: The dominant trigger ids are:
137273  eemcjp1mb  EEMC JP > th1 (52, 8.7 GeV) && mb, reading out slow detectors, longitudinal running 2 
137641  eemchttpmbl2gamma  EEMC HT > th1 (16, 3.5 GeV) and TP > th1 (20, 4.5 GeV), minbias condition, L2 Gamma algorithm, reading out slow detectors, L2 thresholds at 3.7, 5.2, longitudinal running 2 
137262  eemcht2mbemul  EEMC HT > th2 (22, 5.0 GeV) && mb, reading out slow detectors, emulated in L2, longitudinal running 2 
137222  bemcjp1mb  BEMC JP > th1 (60, 8.3 GeV) && mb, reading out slow detectors, longitudinal running 2 
Figure 2.3a
Figure 2.3
Figure 2.4
Figure 2.5
Figure 2.6
Figure 2.7
Figure 2.8
Figure 2.9
Figure 2.10
Click here for a tarball of the code used in this analysis.
f(x)=p0*(0.69*exp(0.5*((xp1)/0.87)**2)/(sqrt(2*pi)*0.87)+0.31*exp(0.5*((xp1)/3.3)**2)/(sqrt(2*pi)*3.3))
p0 = yield
p1 = centroid
21 electrons
99 electrons
This analysis to look for etas at higher energy is in part motivated by this study. The interest in etas, of course, is that their decay photons are well separated at moderate energies (certainly more separated than the photons from pi0 decay). I ran Weihong's pi0 finder with tower seed threshold of 0.8 GeV and SMD seed threshold of 5 MeV (I believe his default SMD seed setting is 2 MeV). I then look in the 2photon invariant mass region between 0.45 and 0.65 GeV (the PDG nominal mass for the eta is 0.54745 +/ 0.00019 GeV). I observe what looks like a faint eta peak. The dataset processed is the longitudinal 2 run of 2006 from the 20 runs sitting on the IUCF disk in Weihong's directory (/star/institutions/iucf/hew/2006ppLongRuns/).
Within the reconstructed mass window 0.45 to 0.65 GeV, I take a look at the decay photon shower profiles in the SMD. The samples are saved in the file etas.pdf. For the most part, these shower shapes are cleaner than the original sample. Although the statistics are not great.
Hal did a comparison of the widths of the shower shapes between Monte Carlo and data. Below is a description of what was done.
I took the nominal central value, either from the maxHit or the nominal central value, and added the energy in the +/ 12 strips. Then I computed the mean strip (which may have been different from the nominal central value!!). I normalized the shape to give unit area for each smd cluster, and added to the histograms separately for U and V and for MC and data (= Will's events). I did NOT handle Will's events correctly, just using whatever event was chosen randomly, rather than going through his list sequentially. Note I ran 1000 events, and got 94 events in my shower shape histos. So, there are several minor problems. 1) I didn't go through Will's events sequentially. 2) I normalized, but perhaps not to the correct 25 strips, because the mean strip and the nominal strip may have differed. 3) there may have been a cutoff on some events due to being close to one end of the smd plane (near strip 0 or 287). My sense from looking at the plots is that these don't matter much. The conclusion is that the MC shape is significantly narrower than the shape from Will's events, which is obviously narrower than the random clusters we were using at first with no selection for the etas. Hence, we are not wasting our time with this project.
A few histograms were added to the code:
Figure 1:
Data vs. MC mean ustrip
Figure 2:
Data vs. MC mean vstrip
Figure 3:
Data vs. MC ustrip sigma
Figure 4:
Data vs. MC vstrip sigma
Figure 5:
MC E
_{v}
vs. E
_{u}
Figure 6:
Data E
_{v}
vs. E
_{u}
Figure 7:
MC energy asymmetry in SMD planes
Figure 8:
Data energy asymmetry in SMD planes
Figure 9:
Shower shape library index used (picked at random)
Single events shower shapes are displayed in
or
.
In case you missed it, the first look is
. I processed
from the 2006 pp longitudinal 2 runs and picked events tagged with the L2gamma trigger id (137641). I ran the StGammaMaker on the MuDst files from these runs and produced gamma trees. These gamma trees are available at
/star/institutions/iucf/pibero/2007/etaLong/
. Within the StGammaMaker framework, I developed code to seek candidate etas with emphasis on high purity. The macros and source files are:
Note, the workhorse function is
StEtaFinder::findTowerPoints()
.
I fit the diphoton invariant mass with two Gaussians, one for the pi0 peak (p0p2) and another one for the eta peak (p3p5) plus a quadratic for the background (p6p8). The Gaussian is of the form p0*exp(0.5*((xp1)/p2)**2)
and the quadratic is of the form p6*+p7*x+p8*x**2
. A slightly better chi2/ndf in the fit is achieved by using BreitWigner functions instead of Gaussians for the signal here. I calculate the raw yield of etas from the fit as p3*sqrt(2*pi)*p5/bin_width = 85
where each bin is 0.010 GeV wide. I select candidate etas in the mass range 0.45 to 0.55 GeV and plot their photon response in the shower maximum detector here. Since we are interested in collecting photons of pT > 7 GeV, only those candidate photons with pT > 5 GeV will be used in the shower shape library. I also calculate the background under the signal region by integrating the background fit from 0.45 to 0.55 GeV and get 82 counts.
Pibero Djawotho Last updated at Tue Mar 4 16:21:13 EST 2008
The following is a revisited study of E_reco/E_MC for photons with the addition of the SMD energy to E_reco.
SectorVsRunNumber.png 10Feb2010 12:22 14K ShowerShapes.png 10Feb2010 12:22 17K chiSquareMC.png 10Feb2010 12:22 13K chiSquarePibero.png 10Feb2010 12:22 16K chiSquareWill.png 10Feb2010 12:22 16K chiSquareWillAndMC.png 10Feb2010 12:22 17K
The plots below show the conversion process before the Endcap. I look at prompt photons heading towards the Endcap from a MC gammajet sample with a partonic pT of 911 GeV. I identify those photons that convert using the GEANT record. The top left plot shows the total number of direct photons and those that convert. I register a 16% conversion rate. This is consistent with Jason's 2006 SVT review. The top right plot shows the source of conversion, where most of the conversions emanate from the SVT support cone, also consistent with Jason's study. The bottom left plot shows the separation in the SMD between the projected location of the photon and the location of the electron/positron from conversion.
This
file shows several shower shapes in a single plot for comparison:
These Shower Shapes are binned by:
(pre1==0&&pre2==0)
and (pre1>0pre2>0)
They are then fitted with a tripleGaussian of the form:
[0]*([2]*exp(0.5*((x[1])/[3])**2)/(sqrt(2*pi)*[3])+[4]*exp(0.5*((x[1])/[5])**2)/(sqrt(2*pi)*[5])+(1[2][4])*exp(0.5*((x[1])/[6])**2)/(sqrt(2*pi)*[6]))
All fits to MC are with reference to the old Monte carlo fit function:
[0]*(0.69*exp(0.5*((x[1])/0.87)**2)/(sqrt(2*pi)*0.87)+0.31*exp(0.5*((x[1])/3.3)**2)/(sqrt(2*pi)*3.3))
All fits to the data are with reference to a single
. The fit function is:
[0]*([2]*exp(0.5*((x[1])/[3])**2)/(sqrt(2*pi)*[3])+[4]*exp(0.5*((x[1])/[5])**2)/(sqrt(2*pi)*[5])+(1[2][4])*exp(0.5*((x[1])/[6])**2)/(sqrt(2*pi)*[6]))
Partonic p_{T}=911 GeV  Partonic p_{T}=911 GeV 
During Run 6, the L2gamma trigger (trigger id 137641) sampled 4717.10 nb^{1} of integrated luminosity. By restricting the jet to the Barrel, η_{jet}<1, and the gamma to the Endcap, 1<η_{gamma}<2, the yield of gammajets is estimated as the product of the luminosity, the cross section, and the fraction of events in the phasespace above. The total cross section reported by Pythia for gammajet processes at different partonic p_{T} thresholds is listed in the table below. No efficiencies are included.
p_{T} threshold [GeV]  Total cross section [mb]  Fraction  N_{gammajets} 
5  6.551E05  0.0992  30654 
6  3.075E05  0.1161  16840 
7  1.567E05  0.1150  8500 
8  8.654E06  0.1131  4617 
9  4.971E06  0.1223  2868 
10  2.953E06  0.1151  1603 
The p_{T} slope is exp(0.69*pT)=2^(pT)
, so the statistics are halved with each 1 GeV increase in p_{T}.
The purpose of this study is to estimate the photon yield per trigger in the Endcap Electromagnetic Calorimeter during Run 6. The trigger of interest is the L2gamma trigger. Details of the STAR triggers during Run 6 were compiled in the 2006 p+p run (run 6) Trigger FAQ by Jamie Dunlop. The triggers relevant to this study are reproduced in the table below for convenience.
Trigger id  Trigger name  Description 
117641  eemchttpmbl2gamma  EEMC HT > th1 (12, 2.6 GeV) and TP > TH1 (17, 3.8 GeV), minbias condition, L2 Gamma algorithm, reading out slow detectors, L2 thresholds at 2.9, 4.5 
127641  eemchttpmbl2gamma  EEMC HT > th1 (12, 2.6 GeV, run < 7100052;13, 2.8 GeV, run >=7100052) and TP > TH1 (17, 3.8 GeV, run < 710052; 21, 4.7 GeV, run>=7100052 ), minbias condition, L2 Gamma algorithm, reading out slow detectors, L2 thresholds at 3.4, 5.4, transverse running 
137641  eemchttpmbl2gamma  EEMC HT > th1 (16, 3.5 GeV) and TP > th1 (20, 4.5 GeV), minbias condition, L2 Gamma algorithm, reading out slow detectors, L2 thresholds at 3.7, 5.2, longitudinal running 2 
The luminosity sampled by each trigger was also caclulated here by Jamie Dunlop. The luminosity for the relevant triggers is reproduced in the table below for convenience. The figureofmerit (FOM) is calculated as FOM=Luminosity*P_{B}*P_{Y} for transverse runs and FOM=Luminosity*P_{B}^{2}*P_{Y}^{2} for longitudinal runs where P_{B} is the polarization of the blue beam and P_{Y} is the polarization of the yellow beam. Naturally, in spin physics, the FOM is the better indicator of statistical precisison.
Trigger  First run  Last run  Luminosity [nb^{1}]  Figureofmerit [nb^{1}] 
117641  7093102  7096017  118.88  11.89 
127641  7097009  7129065  3219.04  1099.43 
137641  7135050  7156028  4717.10  687.65 
For this study, only the trigger of longitudinal running 2 (137641) is used. As mentioned above, at level0, an EEMC high tower above 3.5 GeV and its associated trigger patch above 4.5 GeV in transverse energy coupled with a minimum bias condition, which is simply a BBC coincidence to ensure a valid collision, is required for the trigger to fire. The EEMC has trigger patches of variable sizes depending on their location in pseudorapidity. (The BEMC has trigger patches of fixed sizes, 4x4 towers.) At level2, a high tower above 3.7 GeV and a 3x3 patch above 5.2 GeV in transverse energy is required to accept the event.
In addition to selecting events that were tagged online by the L2gamma trigger, the offline
looks for tower clusters with minimum transverse energy of 5 GeV. These clusters along with their associated TPC tracks, preshower and postshower tiles, and SMD strips form gamma candidates. Gamma trees for the 2006 trigger ID 137641 with primary vertex are located at
/star/institutions/iucf/pibero/2006/gammaTrees/
.
The gamma candidate is required to have no track pointing to any of its towers.
The gamma candidate is required to have 85% of the total transverse energy in a cone of radius 0.3 in etaphi space around the position of the gamma candidate. That is E
_{T}^{gamma}
/E
_{T}^{cone}
> 0.85 and R=√Δη
^{2}
+Δφ
^{2}
=0.3 is the cone radius.
The gamma candidate is matched to the best awayside jet with neutral fraction < 0.9 and cos(φ
_{gamma}
φ
_{jet}
) < 0.8. The 2006 jet trees are produced by Murad Sarsour at PDSF in
/eliza13/starprod/jetTrees/2006/trees/
. A local mirror exists at RCF under the directory
/star/institutions/iucf/pibero/2006/jetTrees/
.
Jan Balewski has an excellent writeup, Offline spin DB at STAR, on how to get spin states. I obtain the spin states from the skim trees in the jet trees directory. In brief, the useful spin states are:
Blue Beam Polarization  Yellow Beam Polarization  Spin4 
P  P  5 
P  N  6 
N  P  9 
N  N  10 
L2gamma triggers  730128 
Endcap gamma candidates  723848 
Track isolation  246670 
EMC isolation  225400 
Awayside jet  99652 
SMD max sided residual  19281 
Barrelonly jet  15638 
Note the number of L2gamma triggers include only those events with a primary vertex and at least one gamma candidate (BEMC or EEMC).
After selecting Endcap gamma candidates out of L2gamma triggers, applying track and EMC isolation cuts, and matching the Endcap gamma candidate to an awayside jet, I record the number of jets below per event. Surprisingly, 8% of the events only have 1 jet. Those are events where the Endcap gamma candidate was not reconstructed as a neutral jet by the jet finder. The question is why.
I display both Barrel and Endcap calorimeter towers (the zaxis represents tower energy) and draw a circle of radius 0.3 around the gamma candidate and a circle of radius 0.7 around the awayside jet for 2006 pp200 run 7136022. Even though many of the gamma candidates not reconstructed by the jet finder are at the forward edge of the Endcap, it is not at all clear why those that are well within the detector are not being reconstructed.
Note:
No cuts on residuals applied.
jet A_LL systematic possibility
While Pythia does a pretty good job of simulating prompt photon production in p+p collisions, it does not include polarization for the colliding protons nor partons. A statistical method for assigning polarization states for each event based on A_{LL} [1] is demonstrated in this section. For an average number of interactions for each unpolarized bunch crossing, N_{eff}, the occurence of an event with a particular polarization state obeys a Poisson distribution with average yield of events per bunch crossing:
For simplicity, the polarizations of the blue and yellow beams are assumed to be P
_{B}
=P
_{Y}
=0.7 and N
_{eff}
=0.01. The "+" spin state defines the case where both beams have the same helicities and the "" spin state for the case of opposite helicities. The asymmetry A
_{LL}
is calculated from the initial states polarized and unpolarized parton distribution functions and partonlevel asymmetry:
The algorithm then consists in alternatively drawing a random value N
_{int}
from the Poisson distributions with mean μ
_{+}
and μ
_{}
until N
_{int}
>0 at which point an interaction has occured and the event is assigned the current spin state. The functioning of the algorithm is illustrated in Figure 1a where an input A
_{LL}
=0.2 was fixed and N
_{trials}
=500 different asymmetries were calculated. Each trial integrated N
_{total}
=300 events. It is then expected that the mean A
_{LL}
~0.2 and the statistical precision~0.1:
Indeed, both the A
_{LL}
and its error are reproduced. In addition, variations on the number of events per trial were investigated (N
_{total}
) in Figure 1b. The extracted width of the Gaussian distribution for A
_{LL}
is consistent with the prediction for the error (red curve).
Figure 1a  Figure 1b 
For this study, the gammajets Monte Carlo sample for all partonic p_{T} were used. As an example, the prompt photon processes for the partonic p_{T} bin 911 GeV and their total cross sections are listed in the table below. Each partonic p_{T} bin was divided into 15 files each of 2000 events.
============================================================================== I I I I I Subprocess I Number of points I Sigma I I I I I III (mb) I I I I I I N:o Type I Generated Tried I I I I I I ============================================================================== I I I I I 0 All included subprocesses I 2000 9365 I 3.074E06 I I 14 f + fbar > g + gamma I 331 1337 I 4.930E07 I I 18 f + fbar > gamma + gamma I 2 8 I 1.941E09 I I 29 f + g > f + gamma I 1667 8019 I 2.579E06 I I 114 g + g > gamma + gamma I 0 1 I 1.191E10 I I 115 g + g > g + gamma I 0 0 I 0.000E+00 I I I I I ==============================================================================
The cross sections for the different partonic p_{T} bins has been tabulated by Michael Betancourt and is reproduced here for convenience.
Partonic p_{T} [GeV]  Cross Section [mb] 
34  0.0002962 
45  0.0000891 
57  0.0000494 
79  0.0000110 
911  0.00000314 
1115  0.00000149 
1525  0.000000317 
2535  0.00000000990 
3545  0.000000000449 
These events were processed through the 2006 pp200 analysis chain, albeit without any cuts on the SMD. The simulated quantities were taken from the Pythia record and the reconstructed ones from the analysis.
Figure 2 
Figure 3 
Figure 4 
Figure 5 
Figure 6 
Figure 7 
Figure 8 
Figure 9 
PartonicKinematics.C 
DeltaG.C 
pythia6.tar.gz
from the ROOT site ftp://root.cern.ch/root/pythia6.tar.gz and unpack.
tar zxvf pythia6.tar.gzA directory
pythia6/
will be created and some files unpacked into it. Cd into it and compile the Pythia 6 interface to ROOT.
cd pythia6/ ./makePythia6.linuxFor more information, consult Installing ROOT from Source and skip to the section Pythia Event Generators.
tar zxvf pythia8108.tgzA directory
pythia8108/
will be created. Cd into it and follow the instructions in the README
file to build Pythia 8. Set the environment variables PYTHIA8
and PYTHIA8DATA
(preferably in /etc/profile.d/pythia8.sh
):
export PYTHIA8=$HOME/pythia8108 export PYTHIA8DATA=$PYTHIA8/xmldocRun configure with the option for sharedlibrary creation turned on.
./configure enableshared make
tar zxvf root_v5.20.00.source.tar.gz cd root/ ./configure linux withpythia6libdir=$HOME/pythia6 \ enablepythia8 \ withpythia8incdir=$PYTHIA8/include \ withpythia8libdir=$PYTHIA8/lib make make installSet the following environment variables (preferably in
/etc/profile.d/root.sh
):
export ROOTSYS=/usr/local/root export PATH=$PATH:$ROOTSYS/bin export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$ROOTSYS/lib:/usr/local/pythia6 export MANPATH=$MANPATH:$ROOTSYS/manYou should be good to go. Try running the following Pythia 6 and 8 sample macros:
root $ROOTSYS/tutorial/pythia/pythiaExample.C root $ROOTSYS/tutorial/pythia/pythia8.C
7136022.pdf 7136033.pdf 7136034.pdf 7137036.pdf 7138001.pdf 7138010.pdf 7138032.pdf 7140046.pdf 7143012.pdf 7144014.pdf 7145018.pdf 7145024.pdf 7146020.pdf 7146077.pdf 7147052.pdf 7148027.pdf 7149005.pdf 7152062.pdf 7153008.pdf 7155052.pdf