This page features public results from the ATLAS Tracking Performance Group.

Filter Documents

Select the desired keywords to filter the results.
Selections within a section row are combined with a logical OR, while selections among different section rows are combined with a logical AND.

Quick links: Papers Confnotes Pubnotes

Figures

HL-LHC

Public Plots	Approval Date
NEW  Track finding performance plots for a Graph Neural Network pipeline on ATLAS ITk simulated data	October 2023
NEW  Expected tracking performance of the ATLAS Inner Tracker Upgrade for Phase-II	October 2023
NEW  Status of the ACTS Integration for Phase-II ATLAS Track Reconstruction	October 2023
Clustering and tracking in dense hadronic environments with the ATLAS Inner Tracker for the High-Luminosity LHC	November 2022
Parameter optimization algorithms within ACTS	June 2022
Latest Performance Studies of the ATLAS Inner Tracker (ITk) in the HL-LHC	February 2020
Hashing for track reconstruction : Hashing and similarity learning for track reconstruction	October 2019
A Fast Track Reconstruction for HL-LHC	September 2019
Expected Performance of the ATLAS Inner Tracker at the High-Luminosity LHC	March 2017
Vertex reconstruction performance in high pile-up conditions	October 2016
Study of HL-LHC luminous region scenarios	October 2016
Properties of clusters and seeds for the ITk	October 2016

Run 2 and Run 3

Public Plots	Approval Date
NEW  Software and computing for Run 3 of the ATLAS experiment at the LHC	April 2024
NEW  Software Performance of the ATLAS Track Reconstruction for LHC Run 3	August 2023
NEW  The ATLAS Experiment at the CERN Large Hadron Collider: A Description of the Detector Configuration for Run 3	May 2023
Reconstruction of V0 in photo-nuclear collisions using 5.02 TeV Pb+Pb data with ATLAS	September 2023
Proof of concept: Track finding with GPU-Based Annealer	May 2023
Performance plots for the ATLAS Track and Vertex Reconstruction in early LHC Run 3	November 2022
Initial Run 3 pixel and tracking performance in data and simulation	July 2022
Tracking performance in dense environments	July 2022
Tracking efficiency studies in dense environments	May 2022
Run 3 Large-radius tracking performance in data and simulation	May 2022
Expected Large-Radius Tracking Performance in Run 3	December 2021
ATLAS Primary Vertex Reconstruction with ACTS	November 2021
Performance of the improved ATLAS pixel number and position networks for LHC Run 3	July 2021
Novel Vertex Selection for VBF Higgs Invisible Events	April 2020
Low pT tracking in nominal pileup	April 2020
IBL sensor shape corrections	February 2020
ATLAS Pixel cluster splitting with Mixture Density Network	May 2019
Increasing track reconstruction efficiency for high momentum B hadrons at ATLAS	April 2019
Primary Vertex Selection in VBF Higgs to Invisibles at mu 200 with the ATLAS Experiment	April 2019
Identifying merged tracks with machine learning	April 2019
Shape of the ATLAS Insertable B-Layer Sensors from Track-to-Hit Residuals	April 2019
The ϕ-dependent radial distortion of the ATLAS Inner Detector	April 2019
Impact Parameter resolutions in di-jet events	September 2018
Jet core tracking pileup dependence	September 2018
Vertex performance in 2018 data	September 2018
SCT cluster splitting in high pT tau leptons	March 2018
Effects of reduced charge information on CTIDE neural network performances	March 2018
Pixel vertical position evolution	March 2018
Updates on long-term alignment monitoring and diagnostics for ATLAS ID misalignments	February 2018
Vertex-quality fraction of secondary vertices created with large-impact-parameter tracks	February 2018
Cluster size at B-Layer in 2016. Time dependence of dE/dx and cluster size in the IBL and B-layer in 2016.	December 2017
Vertexing performance in high mu 2017 data	November 2017
Average Number of Tracks against mu for Data 2017	November 2017
Comparison of vertex reconstruction quantities in ATLAS software releases 21 and 20.7 for 2016 proton-proton data	August 2017
Weak mode biases in final ID alignment for the 2016 pp dataset	July 2017
Inner Detector residuals from first 2017 alignment	July 2017
Time Dependent Average B-Layer dE/dX (superseded by plot from December 2017)	July 2017
Comparison Between Lorentz Angle at Beginning of Run1 and Run 2 Pixel Approved Plots	May 2017
pT-weighted track reconstruction efficiency inside jets with 2016 data	December 2016
Impact Parameter Resolution Using 2016 MB Data	December 2016
Preliminary ID systematic misalignment for the 2016 pp dataset	December 2016
ID alignment trend plots in 2016	December 2016
Number of tracks vs. mu with full 2016 data	November 2016
Number of reconstructed vertices versus number of interactions in 2016 data and simulation	October 2016
Pixel Clustering Neural Networks Performance Plots	September 2016
Performance plots for dynamic alignment in the pixel detector	August 2016
Heavy Ion Track Reconstruction Efficiency	July 2016
2016 Primary Vertex Performance Plots	July 2016
Signal track reconstruction efficiency in Large radius tracking	July 2016
IBL Alignment Performance in 2016	July 2016
Number of tracks vs. mu in 2015 and 2016	May 2016
Vertex Reco in 2016 and Release 20.7	May 2016
Alignment of the ATLAS Inner Detector with early 2016 LHC data	May 2016
2D Views of the ID with hadronic interactions	February 2016
Time dependent alignment corrections to IBL distortions	December 2015
Impact of IBL distortion on impact parameter resolution in MC	December 2015
Impact of IBL distortion on impact parameter resolution	December 2015
IBL Bowing	November 2015
KS resonance	July 2015
Impact parameter resolution Run-1 vs. Run-2	July 2015
Basic Inner Detector Tracking Performance at 13 TeV	June 2015
ATLAS Inner Detector Alignment Performance with 2015 Cosmic Ray Data	June 2015
Hadronic Interaction Candidates in 13 TeV data	June 2015
Photon Conversion Candidates in 900 GeV data	June 2015
Vertexing Plots from Initial 13 TeV Runs	June 2015

Run-1 (2012 Collision data)

Public Plot	Approval Date
Alignment framework updates with IBL	June 2014
ID material studies: SCT Extension efficiency	March 2014
Inner Detector performance in 25 ns and 50 ns runs	December 2013
Hadronic Interaction Vertices	April 2013
TRT performance in HI collisions	Feb 2013
Track Reconstruction Performance and Efficiency	October 2012
Improved impact parameter systematics	October 2012
Alignment systematics	May 2012
Neural network clustering	May 2012
Vertex reconstruction	November 2013
Impact parameter resolution	May 2012
TRT occupancy in 2012	May 2012
Minimum bias performance plots at 8 TeV	August 2012

Run-1 (2011 Collision data)

Vertex performance

The following plots are commented in document ATL-COM-PHYS-2011-1312

They refer to 2011 data-taking, notice that plots for 2012 data are already available in the new vertex reconstruction plots page

The data are physics_MinBIas stream from period G4 (runs 183038, 183045, 183054, 183078, 183079, 183081, 183127, 183129, 183130), random triggers applied The MC is minimum bias. The resolution is computed with he split method described in ATLAS-CONF-2010-069 The most prominent difference to what is described in the CONF note above is that in the track reconstruction the minimum pT of tracks was raised from 100 MeV to 400 MeV.

Estimated vertex resolution σX in 7 TeV data of 2011 as a function ofthe number of tracks per vertex. The data are from period G4 from 2011 data taking which comprises of runs183038, 183045, 183054, 183078, 183079, 183081, 183127, 183129, 183130.The stream is physics_MinBias which has 1.396.597 events. The random trigger applied is EF_rd0_filled_noAlg.	eps version of the figure
Estimated vertex resolution σY in 7 TeV data of 2011 as a function ofthe number of tracks per vertex. The data are from period G4 from 2011 data taking which comprises of runs183038, 183045, 183054, 183078, 183079, 183081, 183127, 183129, 183130.The stream is physics_MinBias which has 1.396.597 events. The random trigger applied is EF_rd0_filled_noAlg.	eps version of the figure
Estimated vertex resolution σZ in 7 TeV data of 2011 as a function ofthe number of tracks per vertex. The data are from period G4 from 2011 data taking which comprises of runs183038, 183045, 183054, 183078, 183079, 183081, 183127, 183129, 183130.The stream is physics_MinBias which has 1.396.597 events. The random trigger applied is EF_rd0_filled_noAlg.	eps version of the figure
Estimated vertex resolution σX in 7 TeV data of 2011 as a function of √ΣpT2) where pT is the transverse momentum of each track used in the vertex fit. The data are from period G4 from 2011 data taking which comprises of runs183038, 183045, 183054, 183078, 183079, 183081, 183127, 183129, 183130.The stream is physics_MinBias which has 1.396.597 events. The random trigger applied is EF_rd0_filled_noAlg.	eps version of the figure
Estimated vertex resolution σY in 7 TeV data of 2011 as a function of √ΣpT2) where pT is the transverse momentum of each track used in the vertex fit. The data are from period G4 from 2011 data taking which comprises of runs183038, 183045, 183054, 183078, 183079, 183081, 183127, 183129, 183130.The stream is physics_MinBias which has 1.396.597 events. The random trigger applied is EF_rd0_filled_noAlg.	eps version of the figure
Estimated vertex resolution σZ in 7 TeV data of 2011 as a function of √ΣpT2) where pT is the transverse momentum of each track used in the vertex fit. The data are from period G4 from 2011 data taking which comprises of runs183038, 183045, 183054, 183078, 183079, 183081, 183127, 183129, 183130.The stream is physics_MinBias which has 1.396.597 events. The random trigger applied is EF_rd0_filled_noAlg.	eps version of the figure

The following plots are commented in ATL-COM-PHYS-2011-571. They are updates of the document ATLAS-CONF-2010-069* The most prominent difference to what is described in the CONF note above is that in the track reconstruction the minimum pT of tracks was raised from 100 MeV to 400 MeV. The run is 180164 which had an integrated luminosity of ~24 pb-1. The trigger applied is EF_rd0_filled_noAlg. The stream is physics_MinBias which has ~494k events of which ~233k pass the trigger.

Two dimensional distribution of reconstructed primary vertices with NTrk > 3 in the transverse (x-y) plane in 7 TeV data of 2011. This plot updates Fig 2 of ATLAS-CONF-2010-069. The run is 180164 which had an integrated luminosity of ~24 pb-1. The trigger applied is EF_rd0_filled_noAlg. The stream is physics_MinBias which has ~494k events of which ~233k pass the trigger.	eps version of the figure
Two dimensional distribution of reconstructed primary vertices with NTrk > 3 in the (x-z) plane in 7 TeV data of 2011. This plot updates Fig 3 (left) of ATLAS-CONF-2010-069. In 2011 the ATLAS coordinate system has been rotated by 0.5 mrad about the y-axis to make it closer to its nominal definition with the z-axis parallel to the beam. The run is 180164 which had an integrated luminosity of ~24 pb-1. The trigger applied is EF_rd0_filled_noAlg. The stream is physics_MinBias which has ~494k events of which ~233k pass the trigger.	eps version of the figure
Two dimensional distribution of reconstructed primary vertices with NTrk > 3 in the (y-z) plane in 7 TeV data of 2011. This plot updates Fig 3 (right) of ATLAS- CONF-2010-069. Differently from the x coordinate the much smaller 0.08 mrad tilt about the x-axis has not been applied in the redefinition of the coordinate system. The run is 180164 which had an integrated luminosity of ~24 pb-1. The trigger applied is EF_rd0_filled_noAlg. The stream is physics_MinBias which has ~494k events of which ~233k pass the trigger.	eps version of the figure

Pile up

The previous set of plots from document ATL-COM-PHYS-2011-1426 has been removed since superseded by the public note ATLAS-CONF-2012-042

The following plots are commented in document ATL-COM-PHYS-2011-1312

Data are random trigger events from 2011 data up to mid-September. Simulation is a mixture of Non-Diffractive, Single Diffractive and Double Diffractive events based on Pythia cross sections (approximately 68% Non-Diffractive events).

Distribution of the number of tracks associated to the reconstructed vertices in data and MC. The MC sample used is a Pythia 6 minimum bias sample containing a mixture of Non-Diffractive, Single Diffractive and Double Diffractive events based on Pythia cross sections. Systematics uncertainties are not shown in the plot. Data are collected using a 5Hz random trigger	eps version of the figure
Distribution of √ΣpT2 of the tracks associated to the reconstructed vertices in data and MC. The MC sample used is a Pythia 6 minimum bias sample containing a mixture of Non-Diffractive, Single Diffractive and Double Diffractive events based on Pythia cross sections. Systematics uncertainties are not shown in the plot. Data are collected using a 5Hz random trigger	eps version of the figure
Distribution of the average number of reconstructed vertices as function of the average number of interactions per bunch crossing . Data are collected using a 5Hz random trigger (EF_rd0_filled_noAlg). The measurement is provided by the LUCID detector. The red curve is taken from a fit on minimum bias MC and superimposed on data fitting only the vertical offset (to account for a small component due to non-collision backgrounds in the data sample that is not simulated. The fitted intercept value is compatible with zero).The azure boxes show the mean number of reconstructed vertices per minute of data taking. The result agree with the MC expectation for a vertex reconstruction efficiency on Non-Diffractive vertices of about 95% and of about 10% on Diffractive ones	eps version of the figure

Run-1 (2010 Collision data)

Tracking in jets

The following plots are commented in document ATL-COM-PHYS-2011-1247

Mean number of shared pixel hits as a function of the distance ∆R between the track and the jet for bins in jet pT for data and Monte Carlo. Jets are reconstructed with the anti-kt(R = 0.6) algorithm and calibrated using the EM+JES calibration scheme.	eps version of the figure
Fraction of tracks with a TRT extension as a function of the distance ∆R 0 between the track and the jet for bins in jet pT for data and Monte Carlo. Jets are reconstructed with the anti-kt(R = 0.6) algorithm and calibrated using the EM+JES calibration scheme.	eps version of the figure

Interaction rate in SCT

The following plots are documented in ATL-COM-SOFT-2011-026, please refer to it for more details.

Fraction of charged particles with their last hits in the second and third layers of SCT (red for L2, green for L3) as a function of radius at which charged particles stops. The double structures in red and green come from the fact that the last hit in a layer occurs when the particle stopped in that layer or in the supporting material of the next layer. A small fraction of the tracks with their last hit in L4 are loopers and actually stop at small radii.	eps version of the figure
Comparisons of fractions of early/late interacting and non-interacting tracks from identified protons and anti-protons from Λ-decay to two G4 physics lists. The data and MC are plotted as a function of pT. The purity of the proton/anti-proton tracks is already before background subtraction (based on side-bands around the Λ mass).	eps version of the figure eps version of the figure eps version of the figure	eps version of the figure eps version of the figure eps version of the figure
Comparisons of fractions of early/late interacting and non-interacting tracks from identified pions from K0s-decay with all charged particle tracks. The data and MC are plotted as a function of pT. The purity of the pion tracks is already high before background subtraction (based on side-bands around the K0s mass).	eps version of the figure eps version of the figure eps version of the figure	eps version of the figure eps version of the figure eps version of the figure

Heavy ions tracking

The following plots are commented in document ATL-COM-PHYS-2010-1028

Track selection follows the special reconstruction cuts used in heavy ion tracking:

• pT > 500 MeV
• |η| < 2.5
• nSCT hits + nPixel hits ≥ 9
• no Pixel holes, at most 1 SCT hole
• at most one shared silicon hit

with the additional tighter selections

• nBLayer hits ≥ 1
• |d₀| < 1.5 mm (not for impact parameter distributions)
• |z₀sinθ| < 1.5 mm (not for impact parameter distributions)

Transverse impact parameter distribution for all reconstructed tracks.	eps version of the figure eps version of the figure
Longitudinal impact parameter distribution for all reconstructed tracks.	eps version of the figure eps version of the figure
Number of pixel hits on track (integrated and averages as function of η) for heavy ion collision and comparison with Monte Carlo expectations.	eps version of the figure eps version of the figure
Number of SCT hits on track (integrated and averages as function of η) for heavy ion collision and comparison with Monte Carlo expectations.	eps version of the figure eps version of the figure
Distribution of average number SCT holes for heavy ion collision and comparison with Monte Carlo expectations. A hole is defined as a missing measurement when it is expected i.e. a track crosses an operational silicon modules and have hits both in silicon modules preceeding and following it. Since tracking runs with a requirement of no holes in the Pixel Detector, only SCT holes can be counted.	eps version of the figure
Number of TRT hits on track (integrated and averages as function of η) for heavy ion collision and comparison with Monte Carlo expectations.	eps version of the figure eps version of the figure
Transverse impact parameter distribution for all tracks in centrality bins 0-10 and 70-100.	pdf version of the figure pdf version of the figure
Longitudinal impact parameter distribution for all tracks in centrality bins 0-10 and 70-100.	pdf version of the figure pdf version of the figure
Average number of pixel hits (total and versus pseudorapidity) on track in centrality bins 0-10 and 70-100.	pdf version of the figure pdf version of the figure
Average number of SCT hits (total and versus pseudorapidity) on track in centrality bins 0-10 and 70-100.	pdf version of the figure pdf version of the figure

Heavy ion vertexing

The following plots are commented in document ATL-COM-PHYS-2011-1312

The transverse error of the primary vertex vs. the number of tracks usedin the vertex fit. As expected, the error decreases as a square root with the number of tracks contributing to the vertex. The fact that the dependence is a power law isevident from linearity in log-log scale .The red line is Hijing MC, the black dots are datafrom run 169927 of 2010.	eps version of the figure
The longitudinal error of the primary vertex vs. the number of tracks usedin the vertex fit. As expected, the error decreases with the number of trackscontributing to the vertex. The red line is Hijing MC, the black dots are datafrom run 169927 of 2010.	eps version of the figure
The number of tracks contributing to the vertex vs. all reconstructed tracks in theevent. The top plot is data from run 169927 of 2010, the bottom plot is HIJING MC. Most of the discrepancy can be attributed to the different momentum spectrum between data and HIJING.	eps version of the figure eps version of the figure

Hadronic interactions

Older plots from ATL-COM-PHYS-2011-296 and ATL-COM-PHYS-2011-584 have been removed from this page.

Please use the ones from the published paper:
A measurement of the material in the ATLAS inner detector using secondary hadronic interactions

Public Plot	Approval Date
NEW  Two Dimensional Views of the Inner Detector Produced by the Use of Hadronic Interaction Vertices	February 2016

Run-1 (2009 Collision data)

Armenteros-Podolanski plot of all secondary vertices selected as follows. The selection of secondary vertices is identical to the one used for the K shorts, and Lambda(bar)s: The standard ATLAS V0 vertex finding code is used. No mass constraint is applied during the vertex fit. The two tracks are required to have opposite charge, more than six silicon hits each and a transverse momentum greater than 100 MeV. In addition, the distance of the reconstructed secondary vertex to the primary vertex has to be larger than 0.2 mm in the transverse x-y (=R phi) plane. Furthermore, a cut is applied to require that the flight path of the V0 candidate points back to the interaction point. The selection is applied on the angle in the transverse plane between the flight direction, defined from the line from the primary vertex to the secondary vertex, and the momentum sum of the two tracks. The transverse momentum of the positive tracks is plotted versus the asymmetry of longitudinal momenta of both tracks. Transverse and longitudinal momenta are defined with respect to the momentum sum of the two tracks (i.e. the momentum direction of the V0 candidate), and are given in the laboratory frame.	eps version of the figure
Inner Detector Material Studies
Several different geometries are used to compare simulated samples with data. The nominal sample is the default comparison, as the material model used in simulation is also used in reconstruction. The additional samples scale structures in the Inner Detector to produce roughly 10% and 20% more material in the simulation in terms of radiation length.
Fitted Kshort mean as a function of the decay radius for data and MC simulation.	eps version of the figure
Fitted Kshort mean as a function of the decay radius for various MC simulated material descriptions.	eps version of the figure

Simulation studies

Track reconstruction efficiency in jets

Internal documentation about these plots is available in document ATL-COM-PHYS-2012-139

Track reconstruction efficiency in simulated ttbar events, as a function of distance ΔR to the closest jet (ΔR=√(Δη2+Δφ2)). Cuts applied on the tracks: pT>500 MeV, at least seven silicon hits, at least two pixel hits, at least one hit in the innermost pixel layer, |d0|<1 mm.	eps version of the figure
Rate of reconstructed tracks with a poor truth matching score (<80%) in simulated ttbar events, as a function of distance ΔR to the closest jet (ΔR=√(Δη2+Δφ2)). Cuts applied on the tracks: pT>500 MeV, at least seven silicon hits, at least two pixel hits, at least one hit in the innermost pixel layer, |d0|<1 mm.	eps version of the figure

Electron energy loss recovery

Internal documentation about this plot is available in document ATL-COM-PHYS-2012-139

Some other plots using GSF instead of Global-χ² fitter are available in ATLAS-CONF-2012-047.

Reconstructed transverse momentum compared to true transverse momentum for simulated electrons with pT=5 GeV, using three different track fitting algorithms. A truth seeded pattern recognition was used, therefore the plot does not show the effect of pattern recognition failures.

eps version of the figure

Older Studies

* InDetTrackingPerformanceCSCFigures * ATLAS Detector Paper * CSC Book: ID Tracking Perfomance Note

" target="_top">https://twiki.cern.ch/twiki/bin/view/AtlasPublic/InDetTrackingPerformanceApprovedPlots-->

Major updates:
-- ChristianSchmitt - 19-Nov-2009

Responsible: ChristianSchmitt
Last reviewed by: Never reviewed