D6EFTStudies

EFT studies with Dim6 Warsaw basis

madgraph_model

Modification to the existing SMEFTSim () Madgraph model used for the generation of events. To produce a sample:

• to prepare the environment on lxplus is to use LCG environments, for example (this alredy has a version of lhapdf as well):

  source /cvmfs/sft.cern.ch/lcg/views/LCG_96/x86_64-centos7-gcc8-opt/setup.sh 
  

• change the location of the pdf data cards

  export LHAPDF_DATA_PATH=/cvmfs/sft.cern.ch/lcg/external/lhapdfsets/current/
  export LHAPDF_DATA_PATH=/cvmfs/sft.cern.ch/lcg/views/LCG_96/x86_64-centos7-gcc8-opt/share/LHAPDF/:$LHAPDF_DATA_PATH
  

• download a Madgraph release and the existing SMEFTSim model:

  wget https://cms-project-generators.web.cern.ch/cms-project-generators/MG5_aMC_v2.6.5.tar.gz
  cd MG5_aMC_v2_6_5/models
  wget http://feynrules.irmp.ucl.ac.be/raw-attachment/wiki/SMEFT/SMEFTsim_A_U35_MwScheme_UFO_v2.1.tar.gz
  tar xzf SMEFTsim_A_U35_MwScheme_UFO_v2.1.tar.gz
  

• copy in the model folder the additional files present in the D6EFTStudies project, from the right folder depending on the version of the UFO:

  cp ../../../D6EFTStudies/madgraph_model/[folder]/* .
  

• trivial examples of Madgraph syntax can be found here
• get back to the Madgraph folder and prepare the process folder:

  ./bin/mg5_aMC
  import model SMEFTsim_A_U35_MwScheme_UFO_v2_1-cW_massless
  generate p p > e+ ve mu+ vm j j QCD=0 NP=1
  output SSeu_RcW_test
  quit
  

  Example of generation of VBF Higgs > WW > fully leptonic. The following syntax allows for having EFT entering both in the production and decay vertices of the Higgs boson, while it remains not present in the W decays.
  • NB1 with the W decay, no restrictions on NP^2 are accepted by Madgraph apparently
  • NB2 SMHLOOP not zero turns on loops coupling with gluons and bosons, hence one should most probably turn it off with no expected big impact, as its impact is O(alpha^8)

  ./bin/mg5_aMC
  import model SMEFTsim_A_U35_MwScheme_UFO_v2_1-cHW_massless
  define q = u c d s u~ c~ d~ s~
  generate p p > q q W+ W- QCD=0 SMHLOOP=0 NP=1, W+ > l+ vl, W- > l- vl~ 
  output VBFHWW_RcHW_WWdecay_fs
  quit
  

  One operator of interest which enters in the W decay is Hl3, therefore for productions involving that operator one needs to generate the full VBS process:

  working on it
  

The comments issued in the Madgraph shell correspond to the model used, the process generated, the type of generation:

• the model used for the generation is issued with the command import, which takes as an argument the model name ad possible restrictions, as coded in restriction files. In the example, the model is the one contained in the folder SMEFTsim_A_U35_MwScheme_UFO_v2_1, while the restriction is the one in the file restrict_cW_massless.dat. The restriction files present in the D6EFTStudies project produce events with massless light leptons and quarks, diagonal CKM matrix and rigorously zero value for all the Wilson coefficients, but the one present in the file name.
• the restriction restrict_VBS_massless.dat allows to modify cW, cHW, cHDD, cHbox, cHB, cHWB, cll, cHl3, which are the ones marked as relevant for TGC/QGC and hVV in I. Brivio sildes at 19/01/2019 VBSCan WG1 meeting.
• the process is generated according to the Madgraph syntax. In particular:
  • NP=1 means that no more than a single BSM operator is introduced in each diagram
  • NP^2==1, when present, means that only the interference is calculated
  • NP^2==2, when present, means that only the squared bsm part is calculated
• the default run_card.dat produced by Madgraph may be changed by modifying the file Template/LO/Cards/run_card.dat in the Madgraph folder
• besides the interactive operation, Madgraph may be used with instruction scripts, containing for example:

  import model SMEFTsim_A_U35_MwScheme_UFO_v3_1-cW_massless
  generate p p > e+ ve mu- vm~ NP=1 NP^2==1
  add process p p > e- ve~ mu+ vm NP=1 NP^2==1
  output WW_LI
  

• the script create1Dfolders.py generates the Madgpraph code for the event generation with one Wilson coefficient different from zero, for the linear and quadratic EFT terms, according to the list of operators switched on in the script.
• the script create2Dfolders.py generates the Madgpraph code for the event generation with two Wilson coefficients interfering between each other, according to the list of operators switched on in the script.

generation

Scripts to submit jobs to condor, for the generation of events in Madgraph. To submit a job:

• copy the folder present in the Madgraph directory in this folder:

cp -r ../../MG5_aMC_v2_6_0/SSeu_RcW_test .

• run the submission script

./submit.py SSeu_RcW_test

Config file examples can be found in the subfolders generation/SSeu_RcW/Cards/, generation/SSeu_RcW_int/Cards/, generation/SSeu_RcW_bsm/Cards/, where the param_card.dat, run_card.dat and proc_card.dat examples for same-sign WW and cW non zero are reported. The three folders refer to the SM case, the interference between SM and BSM and the BSM case respectively. Operators of interest are listed here

• The script postProcess.py takes as input a *_results folder, controls some basic parameters for the success of the generation, unpacks the LHE files, creates a summary of the run, and cleans the folder from unnecessary log files when called with the clean option (keeping those of failed jobs). It also creates the input cfg file for read_03.cpp, adding the cfg file to the folder itself

analysis

• read_02.cpp reads LHE files and produces sets of histograms, NOT MAINTAINED
• read_03.cpp reads LHE files and produces sets of ntuples
  • takes as input argument a config file: ./read_03 cfg/ntuple_SSeu_RcW_0p3.cfg
  • an additional argument sets the total number of events read per sample: ./read_03 cfg/ntuple_SSeu_RcW_0p3.cfg 100
  • all variables used later on are created here, for the time being
• checkEntries.cpp checks the number of entries of all ntuples stored in a root file

DatacarCreator

datacard_creator_2.cpp

The code has been mostly taken from datacard_creator.cpp. Fragmented in subfunctions implemented in dcutils.*, reads a config file containing the list of variables to be studied, the selections to be applied, the files to be read and some detailed configuration parameters

• files have to be produced with read_03.cpp, which is where variables are built. If a new variable is needed, for the time being a new run of read_03.cpp is required.
• for each variable as an input in the cfg file,
  datacard_creator_2.cpp produces a pair of files: a root file containing histograms and a datacard for Combine.
• In the folder where datacards are stored, a shell script gets created, that contains the commands to be run to create the roofit workspaces needed by Combine, as well as one to issue the combine fit. The script should be sourced from an environment built according to the indications present in EFT/Fit. Single .sub and .sh files to be launched with condor get created as well.

datacard_TwoOp_creator_2.cpp

• Prepares the combine datacards for 2D combine fits.
• In the folder where datacards are stored, a shell script gets created, that contains the commands to be run to create the roofit workspaces needed by Combine, as well as one to issue the combine fit. The script should be sourced from an environment built according to the indications present in EFT/Fit. Single .sub and .sh files to be launched with condor get created as well.

Condor submission

To submit to condor the creation of the workspaces and the combine fitting procedure, launch condor_submit with as argument the submit*.sub file of interest from the created datacard folders. Once a set of folders have been created by datacard_creator_2 or datacard_twoOp_creator_2, once can submit all jobs at once like this: for fil in `find . -maxdepth 2 -name "submit*ub" | grep SSWW` ; do condor_submit $fil ; done In this case, all files that have the name like submit*ub and are in a folder which has SSWW in its name will be sent to condor.

read_results.cpp

Runs over the rootfiles resulting from the combine fits and produces 1D likelihood scans

Fits with Combine

Instructions on how to install and use combine for likelihood scan.

• The reference for the building of anomalous coupling analytic fitting are here.
• A Combine tutorial can be found here

You need to be either on MiB cluster/virgilio (read the message after logging in on what to source to be able to use cmssw) or on lxplus.

• The Combine repository and some instruction are here
• The main Combine documentation is here

The installation is needed also when using the scripts created by datacard*_creator_2, which takes care of the steps C and D.

A) install Combine

Prepare a CMSSW release where to install combine, which sets the proper compiling environment with a consistent set of libraries to be linked in the compilation process (the command cmsenv sets all the shell environment variables needed for compiling and running once and for all). Since the CMSSW and Combine code is taken from a Git repository, it's suggested to install it outside of the EFT folder.

cmsrel CMSSW_10_2_13
cd CMSSW_10_2_13/src
cmsenv

Download the Combine tool inside the CMSSW release environment and enter the subfolder:

git clone https://github.com/cms-analysis/HiggsAnalysis-CombinedLimit.git HiggsAnalysis/CombinedLimit

Go to the Combine subfolder, and get the right stable version of Combine compatible with our setup:

cd $CMSSW_BASE/src/HiggsAnalysis/CombinedLimit
git fetch origin
git checkout v8.0.1

Compile combine with the scramv1 compiling tool of the CMS software

scramv1 b clean; scramv1 b

B) install the fitting model

Download our fitting model and compile it

cd $CMSSW_BASE/src/HiggsAnalysis
git clone [email protected]:amassiro/AnalyticAnomalousCoupling.git
scramv1 b -j 20

C) generate a RooFit workspace starting from the datacards

The workspace is the RooFit container that holds all the information needed by Combine di perform fits. THe following script creates the workspace starting from the txt and root files produced by datacard_creator and datacard_creator_2 of EFT/DatacardCreator.

text2workspace.py datacard.txt \
                  -P HiggsAnalysis.AnalyticAnomalousCoupling.AnomalousCoupling:analiticAnomalousCoupling \
                  --PO=k_my_1,r \
                  -o  model_test.root   

where:

option	meaning
-P PHYSMODEL	Physics model to use. It should be in the form (module name):(object name)
HiggsAnalysis.AnalyticAnomalousCoupling	module name of the physics model, it's a folder
analiticAnomalousCoupling	object name of the physics model
--PO = k_my_1,r	define the physics observables to be k_my_1 and r
-o model_test.root	filename of the workspace created

The module HiggsAnalysis.AnalyticAnomalousCoupling is actually implemented in "

• HiggsAnalysis/AnalyticAnomalousCoupling/python/AnomalousCoupling.py: old version

• HiggsAnalysis/AnalyticAnomalousCoupling/python/AnomalousCouplingTwoOp.py: version with one single Wilson coefficient free to float

• HiggsAnalysis/AnalyticAnomalousCoupling/python/AnomalousCouplingOneOp.py: version with two Wilson coefficients free to float

• HiggsAnalysis/AnalyticAnomalousCoupling/python/AnomalousCouplingEFT.py: version with all SMEFTSim Wilson coefficients free to float. NB the root file is searched for in the subfolder indicated in the txt file, by the text2workspace.py script. The workspace created can be printed for inspection:

  root -b model_test.root w->Print ()

When using datacard_creator_2, two shell script are also created, that contain the commands to be issued to actually create the workspaces, and that run Combine to perform the fits. Those scripts use the AnomalousCouplingEFT of the EFT model.

D) fit

Simulate with k_my_1 set to 1 and r=1

combine -M MultiDimFit model_test.root             \
        --algo=grid --points 120  -m 125           \
        -t -1 --expectSignal=1                     \
        --redefineSignalPOIs k_my_1                  \
        --freezeParameters r --setParameters r=1   \ 
        --setParameterRanges k_my_1=-20,20           \
        --verbose -1

Where:

option	meaning
-M MultiDimFit	use the method MultiDimFit to extract upper limits
--algo=grid	choose the algorithm grid to compute uncertainties
--points 120	set to 120 the number of points in DOING SOMETHING
-m 125	set the Higgs boson mass to 125 GeV
-t -1	number of Toy MC extractions
--expectSignal=1	generate signal toys instead of background ones, with the specified signal strength.
--redefineSignalPOIs k_my_1	set k_my_1 as the only physics observable in the fit
--freezeParameters r	the parameter r has a value which will not change in the fit
--setParameters r=1	the value to which r is frozen
--setParameterRanges k_my_1=-20,20	range of variability of the free parameter considered in the fit
--verbose -1	verbosity set to minimum

To simulate "sm" part, k_my_1 == 0 (this is how we simulate the expected result with "sm") :

combine -M MultiDimFit model_test.root                  \           
        --algo=grid --points 120  -m 125                \              
        -t -1 --expectSignal=1                          \      
        --redefineSignalPOIs k_my_1                       \     
        --freezeParameters r --setParameters r=1,k_my_1=0 \                             
        --setParameterRanges k_my_1=-20,20

E) plot the profile likelihood obtained

cd $CMSSW_BASE/src/HiggsAnalysis/AnalyticAnomalousCoupling/test/
root -l higgsCombineTest.MultiDimFit.mH125.root  \
        higgsCombineTest.MultiDimFit.mH125.root draw.cxx

PostPlots

read_results.cpp

Performes the detailed inspection of each single operator for 1D scans, produces the likelihood scans, calculates the crossing points with the horisontal thresholds for 1sigma and 2sigma, plots all the results and produces a summary txt CSV (comma-separated values) file containing the scan of all operators and variables considered in the generation campaign. Takes as input the cfg file fed to DatacardCreator and optionally the folder where the *result folder are saved.

read_results_2D.cpp

Performes the detailed inspection of each single operator for 2D scans, calculates the crossing intersections with the horisontal thresholds for 1sigma and 2sigma, plots the obtained contours ranked in area for the 1sigma CL. Takes as input the cfg file fed to DatacardCreator and optionally the folder where the *result folder are saved.

summaryPlots.cpp

Compares two different set of results (for example SSWW VBS and inclusive prodcution), taking as input the output CSV files of read_results.cpp.

useful commands and links

• unzip madgraph results within a folder

for fil in  `find  . -name "*gz"` ; do gunzip $fil ; done

• prepare list of lhe files for cfg files (rem unalias ls)

find . -name "*lhe" | sed -e s%\.%`pwd`% | tr "\n" ","

• calculate the cross-section of the sample from the generated events, in the generation folder:

./postProcess.py [folder containing the job outputs and lhe files]

• Madgraph totorial here