Difference between revisions of "Solid eventgenerator"
|Line 28:||Line 28:|
Revision as of 01:56, 12 July 2020
- 1 general note
- 2 inclusive event generator
- 3 solid_bggen
- 4 SIDIS event generator (outdated)
- 5 JPsi
- 6 genTCS
Lund file format
GEMC reads input files in the Lund format. GEMC 2.x writes entries from the header (first) line to output banks:
- For versions through GEMC 2.5, all header entries except the first and fifth are written to variables var1, var2, var3... in the "header" bank.
- For versions from GEMC 2.6 on, all header entries are written to variables userVar001, userVar002, userVar003... in the "userHeader" bank.
- GEMC 1.x doesn't have header bank so no 1st row entry will pass into GEMC output. The workaround is to use the information from the LUND file or from eicRate root tree output, pay special attention the make sure the event index is right.
General rule is to use (crossection*luminosity*phasespace/Ntrials) as weight for events, then sum events with weight in any kinematic region to get rate in Hz in that region.
Ntrials is equal to Nevent if no trial is rejected within the phasespace.
events in one file should use Ntrials/Nevent for that file, then when sum events from many files with equal number of events, it needs to divide number of files to get correct normalization for example, there are 100 files, each file has 10000 events, the rate in each file has normalized by 10000, then add 100 files, the final result needs normalization by 10000*100=1e6, so the weight needs to divide by number of file 100
It's a good practice to put weight as the tenth number in the Lund file header line, it will be transferred to GEMC output as var8 in the header bank (for GEMC versions through 2.5) or userVar010 in the userHeader bank (GEMC versions 2.6 and up).
inclusive event generator
What it can do
It can generate:
- <math>e^- </math> DIS, elastic, Moller, or P. Bosted events, uniformly distributed with weight
- <math>\pi^+ , \pi^-, \pi^0, K^+, K^-, K_s, p</math> with cross section given by the WISER fits, either uniformly distributed with weight or distributed according to cross section. The <math>\pi^0</math> then decay in two photons. The two photons output are in the tree as pi0_g1 and pi0_g2.
The output is then written also in LUND and SOLLUND format (the decay products will be part of the same event)
The vertex position is decided in a uniform way randomly inside the raster. For the two decayed <math>\gamma</math>s from the <math>\pi^0</math> the vertex is shifted by a displacement determined selecting randomly the decay time. If the radius is set to "0.0" in the input file, the vertex will not be generated.
see "readme" for details
eicRate -i <input file> -o <root output file> -n <# events> -m <model #>
-i (or -input) input file name -o (or -output) output file name -n (or -nevent) number of events will be produced -m (or -model) which physics model will be used, see below
The input file
the input file it commands the event generator with information in common to every process. an example can be found at https://jlabsvn.jlab.org/svnroot/solid/evgen/eicRate_20101102/output/input.dat The flags are (they need to be in the same order as in the example)
- lumin : Luminosity for nuclei (not for nucleon) (for example 1.0e37 Hz/cm^2)
- runtime : Runtime (for example 48.0 hours)
- e-_mom: Beam momentum (for example 11.0 GeV)
- e-_the: Beam direction polar angle (for example 0 rad)
- e-_phi: Beam direction azimuthal angle (for example 0 rad)
- ion_mom : Momentum of the ions of the target (for example 0 GeV)
- ion_the: Ion direction polar angle (for example 0 rad)
- ion_phi: Ion direction azimuthal angle (for example 0 rad)
- ion_m : Mass of the ions of the target (for example 0.938 GeV)
- ion_Z : Number of protons in the target (for example 1)
- ion_N : Number of neutrons in the target (for example 1)
- dens : Density of the target (g/cm^3): if 0.0, will try to search one in his own database. If you rather than want to use your own (or it is not in the database), specify it here (for example 0.162 g/cm^3)
- length : Length of the target (cm): if 0.0, will try to search one in his own database. If you rather than want to use your own (or it is not in the database), specify it here (for example 40 cm)
- ras_x : size in X of the raster (cm , it is assumed to be of cylindrical shape): if 0.0 The vertex will not be generated (for example 0.2 cm)
- ras_y : size in Y of the raster (cm , it is assumed to be of cylindrical shape): if 0.0 The vertex will not be generated (for example 0.2 cm)
- offset: center of the target (for example 0.0 0.0 10.0 cm)
- output : Output format: 0=root ; 1=root+LUND; 2=root+SOLLUND (for example 3)
The output file name is specified at the time of execution of eicRate. If not specified the default 'output.root' will be used. It is also possible to have text output in lund format in parallel of the root format
- root format
- The output root file contain a tree "T" with variables below with details explained here https://jlabsvn.jlab.org/svnroot/solid/evgen/eicRate_20101102/output/README
- LUND format
- used for input into GEMC. flag 'output 2' in the inputfile (a script will create another file, named changing the string "root" in the output name with "lund", so by default 'output.lund'
- the lund format is at , the entries defined here are
OUT << "1" << " \t " << Wprate << " \t " << Wmrate << " \t " << "0" << " \t " << "0" << " \t " << x << " \t " << y << " \t " << W << " \t " << Q2 << " \t " << rate << endl; OUT << " \t " << "1" << " \t " << charge << " \t " << "1" << " \t " << particle_id << " \t " << "0" << " \t " << "0" << " \t " << px << " \t " << py << " \t " << pz << " \t " << Ef << " \t " << mass << " \t " << p_vertex.X() << " \t " << p_vertex.Y() << " \t " << p_vertex.Z() << endl;
- SOLLUND format
- this is an alternative workaround to overcome the shortcoming of GEMC 1.x
- flag 'output 3' in the inputfile (a script will create another file, named changing the string "root" in the output name with "sollund", so by default 'output.sollund'
- based on LUND format, except one more entry "weight" is added between "Number of particles" and "Number of target nucleons"
- SOLLUND file can be as input for solgemc, the "weight" will be taken into output.
2.5M events for 4.4GeV, 5.5GeV, 6.6GeV, 11.0GeV at http://hallaweb.jlab.org/12GeV/SoLID/download/event_gen/test_generator.tgz
Conversion from root to lund format
Following is shown a tool to convert and play with the output of the event generator https://jlabsvn.jlab.org/svnroot/solid/evgen/root_to_lund/
Arguments for constructor are as follows name mass width charge 2*spin parity C-conjugation 2*Isospin 2*Isospin3 G-parity type lepton number baryon number PDG encoding stable lifetime decay table shortlived subType anti_encoding
name, 0.0*MeV, 0.0*MeV, 0.0, 0, 0, 0, 0, 0, 0, "geantino", 0, 0, 0, true, 0.0, NULL, false, "geantino", 0
DEFINITION CHARGED GEANTINO
name, 0.0*MeV, 0.0*MeV, +1.*eplus, 0, 0, 0, 0, 0, 0, "geantino", 0, 0, 0, true, 0.0, NULL, false, "geantino", 0
ROOT TO LUND CONVERSION
I modified my root to lund conversion so that one can filter the data based on the angle of scattering of the electron. I have also put the option to use a "lund format" file as input for filtering based on the angle (you have these files right now). There is also an option to have in the output a geantino or a charged geantino, rather than the electron. If you running without flags you get this message (how to use it): root2lund : This Program translate the output from eicRate in LUND format and Filter the scattered electron angles Usage: root2lund -o outputfile -i inputfile [-th_min theta_min] [-th_max theta_max] [-ph_min phi_min] [-ph_max phi_max] [-l 1-0] [-g 2-1-0] -o outputfile : output file name (example pluto.lund) -i inputfile : input file name (example pluto.root) -h help, print this message [optional] -th_min theta_min : specify theta minimum for scattered electron -th_max theta_max : specify theta maximum for scattered electron -ph_min phi_min : specify phi minimum for scattered electron -ph_max phi_max : specify phi maximum for scattered electron -l 1-0 : specify format input file ( lund-> 1 (just angle filtering); root->0 (default) ) -g 2-1-0 : specify if particle in output is a geantino ( geantino-> 1 ; charged geantino-> 2 ; normal->0 (default) )
Based on bbgen from halld to generate dominant exclusive events of photonproduction on proton, electron production on any nuclear target is added by assuming EPA for virtual photons and bremsstrahlung for real photons. scattered electron is not included in the output.
Rakitha's note https://solid.jlab.org/cgi-bin/private/ShowDocument?docid=3
some lund file generated /work/halla/solid/evgen/solid_bggen
Some information and instructions from Rakitha:
The last entry in the lund file contains total rate divided by the total events in the file. So this number is small if the file contain large no.of events. I think the two files you were comparing may have been 10k (old file) and 100K (new file). The variable fEvQ2 which is the one before the last entry stores the total absolute rate for the particle. This is per Zhiwen's request. So this number can be compared between lund files with different event sizes.
I wanted to give you the steps I followed to generate these files. Here they are, modify the fort.15 macro file to set no.of events and random number value then run bggen using "./bggen" 1. First run the hall D code and this will generate a PAW file with default file name, "bggen.nt" 2. Convert the PAW file to root format using h2root command for example "h2root bggen.nt /home/rakithab/Simulation_Analysis/Generators/HallD/ElectroProduction/FixRate_10k_rootfiles/bggen_LH40cm_output_300k_1.root" 3. Then I have a script called HallD_gen_lund.cc available in remoll repository, https://github.com/JeffersonLab/remoll_solid/tree/master/generators/halld/scripts to generate lund files. 4. For example "./HallD_gen_lund /home/rakithab/Simulation_Analysis/Generators/HallD/ElectroProduction/FixRate_10k_rootfiles/bggen_LH40cm_output_300k_8.root 1 2 1 " you can see what these options are in the code.
The final step will create 6 different lund files.
Compile and run solid_bggen on the current ifarm environment (centos7)
Previously, compiled version with instruction /group/solid/apps/evgen/solid_bggen (it only works on ifarm1101 with centos6.5 for now because of dependence on /site/cernlib/i386_rhel6)
here is new instruction by Ye Tian on centos7
- 1) cp -r /group/solid/apps/evgen/solid_bggen target directory
- 2) cd target directory/solid_bggen/bggen/
- 3) cp -r /w/halla-scifs17exp/solid/tianye/group_bggen/bggen/build_ifarm ./
Here, within the /build_ifarm/ directory, the copied "/code/" directory contains all the adjusted files, which are needed to compile the "bggen" on current ifarm machine. Basically, Jixie commanded out all the HallD tracking software "HDDM", which are not needed for our purpose.
- 4) source ifarm.bashrc or ifarm.cshrc, depends on the shell you preferred.
- 5) cd ../run/, here you can modify file "fort.15" to adjust RNDMSEQ, TRING, and ... parameters to satisfy your purpose.
- 6) ../build_ifarm/hdgen.Linux.x86_64.exe ---------run the code
- 7) h2root bggen.nt--------convert to root file
- 8) cd ../../scripts/ -------- to the directory to generate corresponding "lund" files
- 9) make --------- use HallD_gen_lund to create "lund" files to feed to GEMC simulation.
- 10) ./HallD_gen_lund ../bggen/run/bggen.root 1 1 1 -------example to generate "lund" files. The explanation of "1 1 1 " is shown in HallD_gen_lund.cc file.
If you want to submit the jobs on ifarm, this is an example job submission scrip: /w/halla-scifs17exp/solid/tianye/group_bggen/bggen/build_ifarm/submitjob_bggenpi0
For current version "bggen", you have to change target length, radiation length, and beam current in the file "/build_ifarm/code/bremsbeam_ini.F" and recompile the code. The more user friendly version is under working by Jixie.
If you any question on compiling the "bggen" on ifarm, just send email to Ye Tian (firstname.lastname@example.org)
the rate on He3 cell glass window seems underestimated
window has Z=17 A=35 lumi_nuclei=0.054e36Hz/cm^2 https://jlabsvn.jlab.org/svnroot/solid/evgen/eicRate_20101102/output/input_solid_SIDIS_He3_window_upstream.dat
For files Rakitha made at /work/halla/solid/evgen/solid_bggen/lund_format/10k_lundfiles/SIDIS_He3 pi+ for He3 of "hallD_pion_p_3He_10k_1.lund", the rate factor at the end of header line is 7704 pi+ for one window of "hallD_pion_p_DST_Winu_10k_1.lund", the same factor is 230 the ratio is 7704/230=33 Assuming pi+ crosssection is same for neutron and proton, this ratio is only proportional to nucleon luminosity ratio
nucleon luminosity ratio between 3He and one window should 1.6 refer to https://hallaweb.jlab.org/wiki/index.php/SoLID_full_simulation_and_file_sharing#luminosity_and_radiation_thickness
So there is a discrepancy of factor 21=33/1.6
Even adding Z/A is 2/3 for He3 gas and 1/2 for window and pi+ crosssection is actually different for neutron and proton, it should not be so different
Have a complete generator which can output both scattered electron and all hadrons
it would be very useful to study fake coincidence among detectors in single particle trigger and more complicated situation in multiple particles coincidence trigger
1. add scattered electron into output into solid_bbgen, maybe by EPA (Yuxiang and Rakitha's conclusion is it's not possible)
2. Follow bbgen approach to combine electron resonance model and electron pythia (haven't tried)
SIDIS event generator (outdated)