= MC@NLO: Next-to-Leading-Order Monte Carlo Generator = == General == * Source: [[http://www.hep.phy.cam.ac.uk/theory/webber/MCatNLO/|MC@NLO Home Page]] * Latest official version: [[http://www.hep.phy.cam.ac.uk/theory/webber/MCatNLO/Package33.tar.gz|v3.3]] * ATLAS will be using a patched version that includes spin correlations and additional choices for top quark decays: v3.31. Patch can be obtained from [[https://twiki.cern.ch/twiki/bin/view/Atlas/MCatNLOForAtlas|ATLAS MC@NLO TWiki]] * Current installation directories: /afs/ifh.de/group/atlas/users/scratch/husemann/mcprod/MC@NLO/v3.''x'' * [[http://www.hep.phy.cam.ac.uk/theory/webber/MCatNLO/MCatNLO_man33.pdf|Manual for v3.3]] * Parton shower with [[http://hepwww.rl.ac.uk/theory/seymour/herwig/herwig65.html|HERWIG]] * MC production in two steps 1. NLO MC generation (first time: additional numerical integration step): output = text file with 4-vectors of hard process 1. Parton shower MC generation (HERWIG): output = HEPEVT format file. In ATLAS this step will be integrated in ATHENA, with ATLAS tunes for HERWIG. * Steering concept: * Call a (bash) shell script for all actions (compilation, running) * Actually two files 1. All inputs (''MCatNLO.inputs''): to be adapted by user 1. All actions (''MCatNLO.Script'') * Call ''bash MCatNLO.inputs'' to start === Installation === 1. Create installation directory and untar v3.3 archive into installation directory 1. Untar addendum for v3.31 on top of v3.3 1. Dependence on other packages: only PDFs (internal or PDFLIB or LHAPDF), LHAPDF installation described below 1. Compilation * Must be under SL4: MC@NLO default compiler is ''g77'' (from gcc3.4), not ''gfortran'' (from gcc4.x). * Compile LHAPDF first 1. The LHAPDF default compiler is ''gfortran'': ''setenv FC g77'' to switch to ''g77'' 1. Configure needs absolute target path (this will be hard-coded in the library to retrieve the PDF set files): ''configure --prefix <absolute path>'' 1. Compile: ''make; make install'' * Final compilation is done with the script ''bash MCatNLO.inputs'', see below * Problems with the shipped ''Makefile'': no proper linking to LHAPDF * Added full path to link target by hand * LHAPDF needs ''libmathlib.a'' from the CERN library * Result: see /afs/ifh.de/group/atlas/users/scratch/husemann/mcprod/MC@NLO/v3.31/Makefile === Running MC@NLO === * The procedure is described well in the [[http://www.hep.phy.cam.ac.uk/theory/webber/MCatNLO/MCatNLO_man33.pdf|MC@NLO Manual]] * Essential steps 1. Edit input file ''MCatNLO.inputs''with the right input parameters (see discussion below) 1. ''bash MCatNLO.inputs'' will trigger compilation and event generation * If the last command in the input file is ''runNLO'', only the NLO step is called. This is sufficient for the ATLAS MC production. Otherwise you need to install HERWIG as well, see below. * Running parton shower MC: * Download latest version of HERWIG source code from [http://hepwww.rl.ac.uk/theory/seymour/herwig/herwig65.html|HERWIG]]. You need ''herwig6510.f'', ''HERWIG65.INC'', and ''herwig6510.inc''. * Copy HERWIG files to work directory or create symbolic links (Not sure if this is necessary, but I ran into a problem when I tried to give the full HERWIG path in the input file: the path length is restricted to 80 characters... good old FORTRAN days...) * Patch ''herwig6510.f'' as described in MC@NLO manual: comment out the user subroutines ''UPEVNT'', ''UPINIT'', ''PDFSET'', and ''STRUCTM'', which are implemented in the MC@NLO code * With ''runMCatNLO'' both the NLO and the parton shower MC are compiled and executed. * Output: * The NLO step puts out an event file that one can easily parse and use for generator level plots * The parton shower step puts out the final HEPEVT file and performs some dummy analysis with ''topdrawer'' output (reminder: ''td <infile> POSTSCR,ORIENT=3'') * Looking under the hood of the master script ''MCatNLO.Script'', the script does not have too much (essential) functionality * Enforce correct naming of executables and output files * Make sure the right main program and utilities get compiled (e.g. ttbar production is a QQbar process and uses ''mcatnlo_qqmain.f'', vector boson + Higgs production is in ''mcatnlo_vhmain.f'') * Create input data card for main programm * Create links to PDF library === Choice of Parameters === * Using ATLAS conventions for [[https://twiki.cern.ch/twiki/bin/view/Atlas/TopMCProductionParameters|Basic MC parameters]] * Random seeds: * Internal random number generator: MC@NLO has their own implementation of what ''Numerical Recipes'' calls the "Park/Miller minimal standard generator" (see Chapter 7.1, under Portable Random Number Generators) * Periodicity is 2e9, not clear if this is enough * Possible solutions for seeds: * Gia: start from 780207, increase by 1000 for every job * Wouter: pick from library of 10000 random seeds * PDF choice: * Take PDFs from LHAPDF (Les Houches Accord PDF library) * LHA PDF #10000: CTEQ6M * Open question: which top quark width? * Currently using 1.42 GeV (sort of a random value) * Top width changes with mass, should we account for that in the MC production? * References for top width: * M. Jezabek and J.H. Kühn, ''Top quark width: Theoretical update'', [[http://prola.aps.org/abstract/PRD/v48/i5/pR1910_1|PRD 48 (1993) R1910]] * QCD and electroweak correction beyond narrow width approximation (in which production and decay factorize) * Table I: width as a function of top mass * W. Bernreuther, ''Top quark physics at the LHC'', [[http://www.iop.org/EJ/abstract/0954-3899/35/8/083001/|J. Phys. G: Nucl. Part. Phys. 35 (2008) 083001]] * Simple parameterization of tree level cross section and QCD correction, fits well with tabulated values by Jezabek/Kühn * Comparison of references (and simple power law fit): [[attachment:masswidth.eps]] == Useful Links == * [[https://twiki.cern.ch/twiki/bin/view/Atlas/TopMC|ATLAS Top MC Page]]