DIFFERENCES BETWEEN ATHENA-ATLFAST AND FORTRAN-ATLFAST


Differences

1) Athena-Atlfast has its default value for the eta transition of the barrel to forward detector (GlobalEventDataMaker.BarrelForwardEta) set at 3.2. In the Fortran the default value (YPAR(17)) is 3.0. This differences is due to the fact that 3.2 is the correct value for the Atlas detector.

2) The default values in Athena-Atlfast for the radius of the cone used by the ClusterCone Algorithm in the barrel and forward regions (ClusterMaker.RConeBarrel and ClusterMaker.RConeForward) are both 0.401. In the Fortran the defaults for these values (YPAR(11) and YPAR(16)) are 0.400. The values are set to 0.401 in Athena-Atlfast to avoid grid edge problems which are seen when the (radius of cone/calorimeter cell size) = integer value. To do a correct comparison with the Fortran, the Fortrans values also need to be set to 0.401 because the Fortran can also suffer from the grid edge problems.

3) Differences in the number of Isolated muons has been seen between the two Atlfasts. Both Athena-Atlfast and Fortran-Atlfast do Isolation of individual particles in the same way, but it is found that the order of the Isolation from Clusters is found to effect the result. Isolation from Clusters is done as follows:

  • Electrons : The nearest cluster to the electron is tested to see if it is the electrons associated Cluster (within RClusterMatch), then the electron is tested for isolation against all the Clusters and also the nearest if it was not its associated one. If the electron is found to be isolated its associated cluster is removed from the list of clusters.

  • Photons : The nearest cluster to the photon is tested to see if it is the photons associated Cluster (within RClusterMatch), then the photon is tested for isolation against all the Clusters and also the nearest if it was not its associated one. If the photon is found to be isolated its associated cluster is removed from the list of clusters.

  • Muons : The muon is tested for isolation against all the Clusters.

    In Athena-Atlfast the job is set to run in the AtlfastDoNotTouch.txt file with electron, then photon, then muon isolation, so muons are never tested for Isolation against isolated-electrons or isolated-photons associated clusters. In Fortran-Atlfast, the order is electron, muon then photon isolation, so the muon is tested for isolation against isolated-photons associated clusters. The Athena-Atlfast job can be configured to run the same as the Fortran-Atlfast by changing the order of the algorithms in the AtlfastDoNotTouch.txt.

    4) The Fortran-Atlfast has the value for Barrel-Forewad eta transition hardwired into the JetSmearer as 3.0. It is set to the GlobalEventDataMaker.BarrelForwadrEta value in Athena-Atlfast (3.2 as default).


    Extensions

    1) By default both Athena-Atlfast and Fortran-Atlfast calculate all the clusters and thus Jets as massless. Athena-Atlfast has an option (ClusterMaker.MasslessJets = false/true) which if set to false means that the Clusterss will be calculated as purely the sum of cells four-momenta, and thus they and the Jets will not be massless. (see here for more information.)


    A Note on the Grid Edge Problems

    It was found that when comparing Athena-Atlfast and Fortran-Atlfast on an event by event basis subtle differences were seen. This was due to probelms in the ClusterMaking. The Calorimeter grid was made up by 0.1 by 0.1 cells and the clustering algorithm had a cone of 0.400. It looks for any cells lying even slightly within a cone of this radius. It was common that the cone edge would fall on the boundry between cells. As a computer cannot have an infinately accurate number, the cone edge never fell exactly on a cell boundry, but just slightly over to one side. It was found that sometimes the cone edge would be just inside a cell in the Athena-Atlfast and just outside in the Fortran-Atlfast or vice-versa. This problem was solved by making the cone radius 0.401 and then the cone edge should never fall on a cell boundary.