Open Dataset Variables#

The following variables are defined:

Event variables#

Variable Name	Description
`nJets`	Number of jets in the event.
`primaryVertexToBeamDisplacementX`	X component of the displacement from the primary vertex to the center of the beam spot.
`primaryVertexToBeamDisplacementY`	Y component of the displacement from the primary vertex to the center of the beam spot.
`primaryVertexToBeamDisplacementZ`	Z component of the displacement from the primary vertex to the center of the beam spot.
`primaryVertexToTruthVertexDisplacementX`	X component of the displacement from the primary vertex to the truth origin.
`primaryVertexToTruthVertexDisplacementY`	Y component of the displacement from the primary vertex to the truth origin.
`primaryVertexToTruthVertexDisplacementZ`	Z component of the displacement from the primary vertex to the truth origin.

Jet variables#

Variable Name	Description
`eta_btagJes`	Jet $\eta$ in input (D)AOD.¹
`pt_btagJes`	Jet $p_{\rm T}$ in input (D)AOD.¹
`eta`	Calibrated $\eta$ of jet.
`phi`	Jet $\phi$ .
`pt`	Calibrated $p_{\rm T}$ of the jet. Flavor tagging is generally quantified in terms of this variable.
`isJvtHS`	Truth hard-scatter label. Truth information! Do not use for training!
`isJvtPU`	Truth pile up label. Truth information! Do not use for training!
`eventNumber`	Event number of the event the jet is associated with.
`mcEventWeight`	Event weight of MC simulated event. Truth information! Do not use for training!
`HadronConeExclExtendedTruthLabelID`	more detailed version of the HadronCone algorithm, with labels for double b-jets: 55 (double b-jets), 54 (bc jet). Truth information! Do not use for training!
`HadronConeExclTruthLabelID`	jet label, using a geometric cone around the jet rather than the jet clustering algorithm. If a parton with a transverse momentum of more than 5 GeV is found within dR(q, jet) < 0.3 of the jet direction, the jet is labelled as a jet with the parton's flavour. The label should be one of: 0 (light jet), 4 (charm jet), 5 (bottom jet), or 15 (tau jet). Results should be similar to the ghost labeling. Truth information! Do not use for training!
`matchedToTruthJet`	Jet is matched to an $R = 0.4$ anti-Kt truth jet.
`ptFromTruthJet`	Matched $R = 0.4$ anti-Kt truth jet $p_{\rm T}$ .
`DL1dv01_pb`	Output from DL1d, $b$ -jet probability.³ ²
`DL1dv01_pc`	Output from the DL1d, $c$ -jet probability.³ ²
`DL1dv01_pu`	Output from the DL1d, light-jet probability.³ ²
`GN2v01_pb`	Output from the GN2, $b$ -jet probability.³
`GN2v01_pc`	Output from the GN2, $c$ -jet probability.³
`GN2v01_ptau`	Output from the GN2, $\tau$ -jet probability.³
`GN2v01_pu`	Output from the GN2, light-jet probability.³

Track variables#

Variable Name	Description
`d0RelativeToBeamspot`
`d0RelativeToBeamspotUncertainty`
`valid`	valid flag, for more robust selection, true for any track that is defined
`d0`	transverse impact parameter relative to PV. (aka distance of closest approach of the track to the primary vertex point in the r-phi projection). Defined here⁶
`d0Uncertainty`	uncertainty on track d0
`phiUncertainty`	uncertainty on track phi
`pt`	track transverse momentum
`qOverP`	track charge divided by momentum magnitude
`qOverPUncertainty`	uncertainty on track qOverP
`theta`	track momentum polar angle
`thetaUncertainty`	uncertainty on track theta
`z0RelativeToBeamspot`	longitudinal impact parameter projected onto the direction perpendicular to the track, relative to beamspot
`z0RelativeToBeamspotUncertainty`	uncertainty on track z0RelativeToBeamspot
`z0SinTheta`	longitudinal impact parameter projected onto the direction perpendicular to the track, relative to the PV. Defined here⁶
`z0SinThetaUncertainty`	uncertainty on track z0SinTheta
`numberOfInnermostPixelLayerHits`	number of hits in the IBL: could be 0, 1, or 2
`numberOfInnermostPixelLayerSharedHits`	number of shared hits (contributing to the track fit and to another track) in the IBL
`numberOfInnermostPixelLayerSplitHits`	number of split hits in the IBL
`numberOfNextToInnermostPixelLayerHits`	number of hits in the next-to-innermost pixel layer: could be 0, 1, or 2
`numberOfPixelHits`	combined number of hits in the pixel layers (including the IBL)
`numberOfPixelSharedHits`	number of shared hits (contributing to the track fit and to another track + not marked as split hit) in the pixel layers (including the IBL)
`numberOfPixelSplitHits`	number of split hits in the pixel layers (including the IBL; split hit = hit is identified as being created by multiple charged particles during ambiguity solver stage at pattern recognition level)
`numberOfSCTHits`	combined number of hits in the SCT layers (since 2 strip hits are required for a full SCT spacepoint, this number is divided by two in the track selection)
`numberOfSCTSharedHits`	Number of shared hits (contributing to the track fit and to another track) in the SCT layers (since 2 strip hits are required for a full SCT spacepoint, this number is divided by two in the track selection)
`deta`	pseudorapidity distance between track and jet
`dphi`	azimuthal angle distance between track and jet
`ptfrac`	fraction of the jet pt carried by the track
`lifetimeSignedD0`	The lifetime-signed transverse impact parameter.⁴
`lifetimeSignedD0Significance`	The lifetime-signed track $d_0$ significance ( $d_{0} / \sigma_{d_{0}}$ ).⁴
`lifetimeSignedZ0SinTheta`	The lifetime-signed longitudinal impact parameter, multiplied by $\sin \theta$ .⁴
`lifetimeSignedZ0SinThetaSignificance`	The lifetime-signed track $z_0$ significance ( $z_{0} \sin \theta / \sigma_{z_{0} \sin \theta}$ ).⁴
`ftagTruthOriginLabel`	Truth origin of the track.⁵
`ftagTruthParentBarcode`	barcode of the parent B/C hadron (if one exists) of the truth particle linked to this track (if one exists). If no truth particles are linked to this track a value of -1 is returned. A value of -2 indicates that there is a truth particle linked to the track but it is not a B/C hadron. Used for linking to particles in the `truth_hadrons` dataset.
`ftagTruthVertexIndex`	truth vertex index of the track. 0 is reserved for the truth PV, any SVs are indexed arbitrarily with a int >0. Truth vertices within 0.1mm are merged.
`GN2v01_trackOrigin`	GN2 prediction for the track origin (ftagTruthOriginLabel).⁵
`GN2v01_vertexIndex`	GN2 prediction for the track vertex index (ftagTruthVertexIndex).

Truth variables#

Variable Name	Description
`mass`
`valid`	validity flag, for more robust selection, true for any truth particle that is defined
`Lxy`	truth particle decay displacement⁷
`decayVertexDPhi`	truth particle decay $\phi$ offset, relative to the jet axis⁷
`decayVertexZ`	Z component of the truth particle decay displacement⁷
`deta`	pseudorapidity distance between truth particle and jet
`dphi`	azimuthal angle distance between truth particle and jet
`pt`	truth particle transverse momentum
`barcode`	generator level barcode of the truth particle, used to link to objects in other datasets
`ftagTruthParentBarcode`	barcode of the parent B hadron (if one exists) of the truth particle.
`pdgId`	truth particle pdgId

Please reach out to the flavour tagging group if you have any questions:

ATLAS FTAG mattermost channel

In production, flavor tagging algoirhtms run before the final jet calibration is applied.The jet energy scale at this stage is set by the jet calibration which is applied when jets are built.Jets saved in DAODs are calibrated to this scale as well.For consistency, we use this jet energy scale in the training as well. ↩↩
The DL1 series includes DL1, DL1r, and DL1d. These are legacy taggers which are not recommended for physics analysis. They are included here only for comparison. ↩↩↩
This is a high-level output from a neural network. It should not be considered a stable input to design new taggers. ↩↩↩↩↩↩↩
The lifetime signed impact parameters are given a positive sign if $|\phi_0 - \phi_{\mathrm{jet}}| < \pi / 2$ and a negative sign otherwise, where $\phi_0$ and $\phi_{\mathrm{jet}}$ are the $\phi$ components of the track displacement and the jet momentum, respectively. ↩↩↩↩
Tracks labels: 0 = pileup, 1 = fake, 2 = prompt, 3 = from $b$ -hadron, 4 = from $c$ child of a $b$ hadron, 5 = from $c$ -hadron, 6 = from $\tau$ lepton, 7 = from other secondary decay. See InDetTrackTruthOriginDefs for ore information. ↩↩
The conventional $d_0$ takes the sign of $(\vec{p}_0 \times \vec{d}_0) \cdot \hat{z}$ , where $\vec{d}_0$ is the transverse track displacement at the closest approach to the primary vertex and $\vec{p}_0$ is the momentum at that point. ↩↩
For truth variables the displacement is defined as the decay vertex position, relative to the production vertex for the simulated event. ↩↩↩