Table of Content |
1.Introduction and Disclaimers |
2.Cluster Variables |
Shortcuts to external documentation: |
3.Global Jet Variables |
Shortcuts to external documentation: |
4.Jet Structure Variables |
Shortcuts to external documentation: |
This page contains some documentation for the root-tuple variables available for the SLAC Jet/MissingEt retreat
March 2007. The tuples have been produced from csc11
production, which in large parts used Athena release 11.0.42 and some
11.0.5†. The content of
the tuples is selective, with a focus on the the calorimeter signal and jet algorithms. Generally, there are variables
for vertices, tracks (TrackParticle objects), egamma object
variables (not Electron!), sliding window and topological
cluster variables for electron reconstruction, uncalibrated and (first order) calibrated topological clusters for
jets, missing Et blocks from various sources, and muons from the MuID package.
The data structure is flat, i.e. all variables independent of their type are leaves of only one tree in
root. The the Athena Aware Ntuple
(CBNTAA) technology has been used to book and fill the variables. The production code is based
on release 12.0.3, with several new tags for jet reconstruction and CBNTAA jet code now available in the main
development branch (toward 13.0.0). The main difference to 12.0.6 and the main branch may be a slightly different
local hadronic calibration package lacking the recent improvements concerning dead material corrections, cluster tagging,
and out-of-cluster corrections.
Attention! |
Some of the variable names used for clusters and jets in this version of the root tuple are not (yet) standard. We expect most of them to become the standard for 13.0.0, though. Some may of course disappear, too! |
†11.0.5 includes trigger information which is not included in the tuples available for this retreat.
The basis of the hadronic final state signal (jets, missing Et) in the calorimeter are topological
cell clusters, based on nearest neighbour clustering. Alternatively, calorimeter cell towers on a regular grid
of Δη × Δφ = 0.1 × 0.1 are still considered for jet finding, mainly as a benchmark.
Topological clustering exploits the local cell geometry best by forming three-dimensional, topologically connected cell
clusters (nearest neighbour strategy). Each cluster starts with a seed cell with |E| > 4σ, to which neighbouring cells
with signals |E| > 2σ are added. A final "ring" of cells with |E| > 0 is added to collect even the smallest signals.
If a cluster has more than one signal maxium, it is split. This leads to clusters sharing cell signals, as cells are typically
not completely assigned to one or the other cluster. Rather, their signal is split between clusters i and
j proportional to the
cluster energy, with wi = Ei / ( Ei + Ej ) and wj =
Ej / ( Ei + Ej ) being the corresponding cell signal weights.
Two index-parallel cluster collections are stored in the tuple. Cluster variables with a _rawtopo
belong to uncalibrated (electromagnetic) energy scale clusters, while variables tagged _caltopo
belong to calibrated (hadronic) energy scale clusters. Note that a cluster at index i in any of the uncalibrated cluster
variable arrays has the same index in the calibrated cluster variable array - the cell composition of the cluster
is independent of the calibration.
A full list of cluster variables, together with some brief description, is available for both collections in this
file.
Topological clusters actually have shape parameters (cluster moments), which are also available in the tuple. These are
useful for cluster classification, (hadronic) calibration, and local dead material corrections. This brief
note describes how some of these moments are actually calculated (starts at the
bottom of page 54). Also, this Wiki page,
maintained by Sven Menke (MPI, Munich), the principal author of the topological clustering, contains an up-to-date
description of the algorithms.
〈Algorithm〉 | 〈Parameter〉 | Algorithm |
C | 4 | Seeded Cone Algorithm with cone size R = 0.4 |
7 | Seeded Cone Algorithm with cone size R = 0.7 | |
Kt | 4 | Kt Algorithm with distance parameter D = 0.4 |
6 | Kt Algorithm with distance parameter D = 0.6 |
The algorithm indicator 〈Algorithm〉 and the 〈Parameter`〉 are summarized in the table to the right. The algorithms and their parameter space considered for the tuple corresponds to the standard jet collections available in default reconstruction.
The available signal sources (〈SignalSource〉 parameter) are
Truth | jets from (stable) particles in the truth event in Monte Carlo; |
Tower | jets from projective calorimeter cell towers (size Δη×Δφ = 0.1 × 0.1); |
CalClus | jets from calibrated topological clusters (hadronic energy scale, corrected for e/h > 1 and some dead material), reconstructed using 4/2/0 cell signal significance cuts (in σnoise, as seed/neighbour/perimeter cell cut); |
Clus | jets from uncalibrated topological clusters (electromagnetic energy scale, no further corrections), reconstructed using 4/2/0 cell signal significance cuts (in σnoise, as seed/neighbour/perimeter cell cut); |
The global jet variables available for all jets, independent of the signal source, are (also see this file for a full listing):
VarName | Type | Index Range | Range | Comment |
Num | UInt_t | 0..500 | number of jets in event | |
Eta | vector〈double〉 | 0..Num-1 | [-5,5] | pseudo-rapidity η of jets |
Phi | vector〈double〉 | 0..Num-1 | [-π,+π] | azimuth φ of jets |
E | vector〈double〉 | 0..Num-1 | jet energy in MeV | |
Et | vector〈double〉 | 0..Num-1 | jet transverse energy in MeV | |
M | vector〈double〉 | 0..Num-1 | jet mass in MeV | |
Px | vector〈double〉 | 0..Num-1 | jet momentum component px in MeV | |
Py | vector〈double〉 | 0..Num-1 | jet momentum component py in MeV | |
Pz | vector〈double〉 | 0..Num-1 | jet momentum component pz in MeV | |
Emf | vector〈double〉 | 0..Num-1 | 0..1 | electromagnetic energy fraction in jet (see below) |
Size | vector〈long〉 | 0..Num-1 | number of constituents in jet |
The electromagnetic energy fraction in a jet is calculated in different
ways, depending on the signal source. For a truth particle jet, the energy fraction carried by photons, electrons, and
positrons is used. For a calorimeter cluster jet or a calorimeter tower jet, the energy fraction deposited in
the electromagnetic calorimeter samplings is calculated. Note that for jets from topological clusters the energy in tagged
electromagnetic clusters can also be used as an indicator of electromagnetic energy fraction in jets. It is available through
the jet constituent links described below.
The constituents for each jet are also available in the root tuple. These are basically the kinematic
four-vector for all jets, indpendent of signal source or calibration. Note that for jets from uncalibrated towers and uncalibrated
clusters these four-vectors are also not fully calibrated. Each constituent also has a kinematic weight with which it contributes
to the jet (always 1 for the moment). Also, a reference to other constituent stores is provided by the index pointing into the
corresponding tuple array, which for the moment only works for jets from topological clusters. This index gives access to more
cluster variables for a given jet.
A detailed list of all jet variables is available. Jets from clusters have already
more information for each constituent attached. This allows a quick evaluation of the effect of local hadronic calibration, for
example. Also, the absolute cluster barycenter (x,y,z) is given for studies of the longitudinal and radial energy distribution in
jets. In general it is quite possible to study the calorimeter signals for jets in quite some detail. The variable name pattern for
constituent variables is∗:
Note that VarNames ending in f denote variables calculated using the final calibration of the constituents, which for uncalibrated cluster and tower jet input means electromagnetic energy scale, while for locally calibrated clusters the variables are calculated on hadronic energy scale. In any case the names of variables calculated using the electromagnetic scale end with i.
∗due to a bug in some cluster related jet variables, not all start with jetC - to be fixed for 13.0.0. The documentation here shows the "wrong" names for consistency with the data on disk, which follow a 〈Algorithm〉〈Parameter〉〈SignalSource〉jetC〈VarName〉 pattern.