Wiki
Clone wikinjet / NJetSherpa / NJetSherpa
Home | About | Installation | Using with Sherpa | Physics Results | Issues |
Welcome to NJet project wiki
These instructions are relevant for NJet version 2.
Fixed order NLO event generation with NJet+Sherpa
The following method has been used with Sherpa version 2.0.beta and version 2.0.beta2.
NOTE in Sherpa 2.0.0 the parameter INTEGRATOR 6
inside the (integration)
section of the Sherpa run cards listed below MUST be removed.
Basic strategy
Use the basic template below for parts RS (real+subtraction), I (integrated subtraction) and B (born). For V (virtual) we will use the
USE_DUMMY_VIRTUALS
option for the grid.
Sherpa should be compiled with --enable-root
, --enable-mpi
, --enable-fastjet
and --enable-lhole
. For example:
../configure --enable-gzip --enable-lhapdf=<path-to-lhapdf> --enable-lhole --enable-mpi --enable-fastjet=<path-to-fastjet> --enable-rivet=<path-to-rivet> --enable-root=<path-to-root>
The procedure goes in 3 stages
1. Generate Comix process maps. Do this on a single core per NLO part. In (integration)
section
set the option Finish_Optmization=Off
and set ERROR=5.0
. This step
can take a while, especially for RS.
Sherpa -f Run<NLOpart>.dat
2. Delete the result directory created in step 1. Use mpi to generate integration grid.
Turn Finish_Optmization=On
, keep the ERROR
large around 0.3
.
mpiexec -np <cores> Sherpa -f Run<NLOpart>.dat
This step can also be run on a single core if mpi is not available.
3. Event generation specify output file using EVENT_OUTPUT=Root[<file>]
which will store events
in EVT_FILE_PATH/<file>. e.g.
Sherpa -f Run<NLOpart>.dat EVENT_OUTPUT=Root[event1] -R 1
Event sets maybe generated in parallel using different random seeds.
This can be set either by -R <seed>
or -R "<seed1> <seed2>"
from the command line.
NB: If the directory EVT_FILE_PATH
is set in the Run card then
ensure the directory exists.
Extra parameters for Virtuals
To link NJet to Sherpa add:
SHERPA_LDADD SherpaLHOLE njet
to the (run) parameters.
For steps 1 and 2 add USE_DUMMY_VIRTUALS=1
to the (run)
parameters and Loop_Generator Internal
to (process)
. For event generation switch to Loop_Generator LHOLE
and make sure OLE_contract.lh
exists. This means running Sherpa -f RunV.dat
once after adding the LHOLE generator, this will issue
some errors in the cross checks and create the OLE_order.lh file.
IR regularization scheme in OLE_order.lh MUST be changed to CDR
when running V separately from B, I and RS.
For > 2j we recommend adding the NJetType
option to the order files to make use of the faster de-symmetrized colour sums.
IRregularisation CDR NJetType loopds
To create the OLE_contract.lh file run
njet.py -o OLE_contract.lh OLE_order.lh
Run Cards for Multi-jet Production
RunB.dat
RunV.dat
RunI.dat
RunRS.dat
Benchmark results for di-jets
To help validate the set-up we provide the following breakdown of NLO components for di-jet production using:
p_T^{1st jet} > 80 GeV, p_T^{all jets} > 60 GeV |eta| < 2.8, anti-kt with R=0.4, MSTW2008 PDFs, Mu_R = Mu_F = H_T/2, IR scheme for V,I : CDR alpha_dipole 0.1
MuR H_T/4 H_T/2 H_T LO 1271.1482 ( 1.518 ) nb 956.29871 ( 1.133 ) nb 735.58093 ( 0.868 ) nb B 1030.0985 ( 1.218 ) nb 800.64112 ( 0.938 ) nb 633.34814 ( 0.738 ) nb V -102.76486 ( 0.883 ) nb 216.89281 ( 1.076 ) nb 153.42467 ( 1.186 ) nb I -251.45876 ( 0.361 ) nb -275.40691 ( 0.212 ) nb -52.369789 ( 0.131 ) nb RS 639.37296 ( 4.288 ) nb 448.29426 ( 2.868 ) nb 322.79007 ( 1.994 ) nb
where LO uses MSTW2008lo68cl PDFs rather than MSTW2008nlo68cl for B,V,I and RS.
Total XS (LO) = 956.29871 (1.132) ^(314.84949) _(220.71778) nb Total XS (NLO) = 1190.42128 (3.210) ^(124.82656) _(133.22819) nb
NB: The scale choice of H_T/2 is not ideal for the di-jet cross section compared to data.
Updated