Chapter 5 Simulations and Acceptance Corrections
5.1 Detector Simulation
The cornerstone of our simulations is GSIM,[34] the standard GEANT-based[35] sim- ulation of the CLAS detector. GSIM handles detector geometry, tracking through materials and electromagneticfields, the physics that govern interactions between par- ticles and the detector, and the detector response. Figure 5.1 illustrates the GSIM simulation of CLAS along with the charged particle tracks of a single simulated event in a cross sectional view. Appendix B includes the contents of the ffread files, the
Figure 5.1 Cross sectional view of the CLAS detector as visualized using
GEANT-based GSIM. The charged particle tracks of a single event are illustrated. Track curvatures reflect the influence of the magnetic field.
configuration files that drive GSIM, for E1F and E16.
Beyond the idealized GSIM simulation, the GSIM Post-Processor (GPP, a com- mon CLAS software package) smears the simulated detector signals and suppresses detector elements to match the experimental conditions. GPP smears TOF times
with consideration to scintillator bar lengths and drift chamber times according to the distance of closest approach to the wires. The degree of GPP smearing is sep- arately tunable for TOF and each of the three regions of the DC. In the current analysis, the three DC regions were required to share a single smearing factor. The parameters were determined by scanning a grid of TOF and DC parameter values to find the pair of parameters that led to the best match between simulation and experiment in terms of (a) electron time resolution and (b) elastic scattering peak width. The GPP parameters used for E1F and E16 appear in Table 5.1. Note that Table 5.1 GPP parameters for E1F and E16. A single DC smearing parameter value is used for all three regions – gpp parametersa,b, andc for regions 1, 2, and 3, respectively – to model spatial resolution. The TOF smearing parameter – gpp parameter f – was used to model the SC time resolution
.
Parameter E1F E16 a/b/c 1.357 1.5
f 1.05 1.35
the GPP parameters determined with this method translate reasonably well to the ω mass peak that is reflected in the missing mass of X in ep→epX, as seen in Fig- ure 5.2. The systematic error introduced by thefigure’s illustrated difference between background-subtracted experiment and simulation amounts to less than 2% by way of our sideband background subtraction method, as discussed in Section 4.3.
For each experiment, about 250 million ep → eωp → eπ+π−π0p events, par- titioned into Q2 ranges of 1 GeV2 width to ensure sufficient statistics at high Q2, were generated over the full range of W and Q2 using GENEV, an event generator commonly used in CLAS analyses. GENEV is based on CLAS photoproduction dif- ferential cross sections with a dipole function to describe the Q2 evolution. However, it does not incorporate experimental differential cross sections for ω production. In- stead, theωevent model was adapted from more general three-pion data by imposing
0.65 0.70 0.75 0.80 0.85 0.90 0.95 M MX[epX] (GeV)
E16
exp with bgexp without bg
sim, recon
0.65 0.70 0.75 0.80 0.85 0.90 0.95 M MX[epX] (GeV)
E1F
exp with bgexp without bg
sim, recon
Figure 5.2 Simulated versus experimental distributions of M MX(epX) for E16
(left) and E1F (right). The highest, green distribution represents experimental data with background from the epπ+π− detection configuration; the red distribution is
the background-subtracted experimental data; the smooth blue distribution is the simulated data.
the invariant mass ofω onto the three-pion kinematics. The angular distributions of ωin the hadron CMS (φ∗ and cosθ∗ distributions) were forced to be phase-space-like. Additionally, a radiative tail in the invariant mass was simulated according to Mo and Tsai [36] as already implemented in GENEV. The radiative tail contributes to theM MX(epX) distribution of Figure 5.2, and its effect is treated by the acceptance
corrections. The additional impact of the radiative effects in terms of W and Q2 bin migration is separately treated in Section 6.1.
The level of agreement between simulated and experimental lab-frame particle kinematics is illustrated in Figure 5.3 for E16 and Figure 5.4 for E1F. The exper- imental distributions include non-ω background, so some differences are expected. There are several notable differences, including between (1) simulated and exper- imental electron momentum distributions, (2) simulated and experimental proton momentum distributions, and (3) thrown and reconstructed electron distributions in E16. Differences 1 and 2 reflect shortcomings of GENEV’s ω event model and represent sources of systematic error that would be improved by a second round of
p
e exp thrown reconcosθ
eφ
ep
pcosθ
pφ
p 0 1 2 3 4 5 p(GeV /c)p
π+ 0.5 0.6 0.7 0.8 0.9 1.0 cosθcosθ
π+ 0 π3 23π π 43π 53π φ(radians)φ
π+Figure 5.3 Simulated and experimental lab-frame particle kinematics for E16. Lab-frame distributions of momentum magnitude (left), cosine of the polar angle (middle column), and azimuthal angle (right) of particles in the epπ+
(top/middle/bottom) detection configuration. Experiment with background appears in blue; and thrown and reconstructed simulated events appear in green and red, respectively.
simulations that are informed by the results of this analysis. These systematic errors have not yet been studied. Difference 3 is an expected feature of E16 and reflects the electron acceptance limitation that motivated the modified target position and magneticfield strength of E1F. The improved electron acceptance is evident by com- paring the thrown and reconstructed electron momentum distributions of E1F in the top-left panel of Figure 5.4 (contrasting to the same for E16 in Figure 5.3).
p
e exp thrown reconcosθ
eφ
ep
pcosθ
pφ
p 0 1 2 3 4 5 p(GeV /c)p
π+ 0.5 0.6 0.7 0.8 0.9 1.0 cosθcosθ
π+ 0 π3 23π π 43π 53π φ(radians)φ
π+Figure 5.4 Simulated and experimental lab-frame particle kinematics for E1F. Lab-frame distributions of momentum magnitude (left), cosine of the polar angle (middle column), and azimuthal angle (right) of particles in the epπ+
(top/middle/bottom) detection configuration. Experiment with background appears in blue; and thrown and reconstructed simulated events appear in green and red, respectively.
so it is important to check that its effect is propagated into the kinematic variables of interest in the simulation as it is in the experiment. The effect manifests as the dominant feature of the hadron CMSφ� distribution. Agreement between simulation and experiment is illustrated in Figure 5.5 for representative W values in E16 (top panels) and E1F (bottom panels).
Ideally, thefinal cross sections of the current analysis would inform a second round of thrown events, with improved W, Q2, and angular distributions from which new
0 E16 W=1.77 GeV exp with bg sim, recon W=1.97 GeV exp with bg sim, recon W=2.17 GeV exp with bg sim, recon −3−2−1 0 1 2 3 φ� (radians) 0 E1F W=1.77 GeV exp with bg sim, recon −3−2−1 0 1 2 3 φ� (radians) W=1.97 GeV exp with bg sim, recon −3−2−1 0 1 2 3 φ� (radians) W=2.17 GeV exp with bg sim, recon
Figure 5.5 Reconstructed simulated (blue) and experimental (green) azimuthal angle distributions ofω in hadronic CMS for E16 (top) and E1F (bottom). Three different W ranges are presented: [1.76,1.78) GeV, [1.96,1.98) GeV, and
[2.16,2.18) GeV (from left to right).
acceptance corrections would be obtained. Fortunately, as indicated in the introduc- tion to this chapter, the thrown event distribution is not as influential as the quality of the detector simulation in determining acceptance corrections in our case. Previous studies [7, 37] with coarser binning determined that systematic uncertainties resulting from thrown event model differences were 5-10%. Morand et al. [7] quoted, addition- ally, much higher uncertainties (up to 20%) in the special cases where a model was used tofill acceptance holes; however, in the current analysis, this hole-filling method was not used.