Performance of quark/gluon discrimination in 8 TeV pp data at CMS (JME-13-002)

Abstract

A likelihood discriminator capable of distinguishing between jets originating from quarks and from gluons is presented. The separation is made available for all jets with transverse momentum greater than 30 GeV, up to pseudorapidities of η=4.7 The discrimination is based on observables sensitive to fundamental differences in the fragmentation properties of gluons and quarks. The performance of the tagger is evaluated using Z+jets and dijet events produced in pp collisions at √s=8 TeV.

Simulation studies

Figures	Caption
	Fig. 1 (left): shape comparison, in simulated QCD di-jet events, of the likelihood discriminator for jets with 40 < pT < 50 GeV in the central region of the detector. Expected distributions for light quark jets (blue) and gluon jets (red) are shown separately.
	Fig. 1 (right): discriminator performance curves of the quark-gluon tagger: green squares for jets with eta < 2 and 40 < pT < 50 GeV, open brown markers for jets with eta < 2 and 80 < pT < 100 GeV, yellow solid markers for jets with 3 < eta < 4.7 and 40 < pT < 50 GeV.

Validation on data

Figures	Caption
	Fig. 2 (left): Data-MC comparison, for jets with 80 < pT < 100 GeV and eta<2 in Z+jet events, of the jet candidate multiplicity. The data (black markers) are compared to the MADGRAPH/PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 2 (center): Data-MC comparison, for jets with 80 < pT < 100 GeV and eta<2 in Z+jet events, of the jet ptD. The data (black markers) are compared to the MADGRAPH/PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 2 (right): Data-MC comparison, for jets with 80 < pT < 100 GeV and eta<2 in Z+jet events, of the jet minor axis (sigma2). The data (black markers) are compared to the MADGRAPH/PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 3 (top left): Data-MC comparison for the quark-gluon discriminant in Z+jet events for jets in the central region with 40 < pT < 50 GeV. The data (black markers) are compared to the MADGRAPH/PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 3 (top center): Data-MC comparison for the quark-gluon discriminant in Z+jet events for jets in the central region with 80 < pT < 100 GeV. The data (black markers) are compared to the MADGRAPH/PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 3 (top right): Data-MC comparison for the quark-gluon discriminant in Z+jet events for jets in the forward region with 40 < pT < 50 GeV. The data (black markers) are compared to the MADGRAPH/PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 3 (bottom left): Data-MC comparison for the quark-gluon discriminant in dijet events for jets in the central region with 40 < pT < 50 GeV. The data (black markers) are compared to the PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Data-MC comparison for the quark-gluon discriminant in dijet events for jets in the central region with 40 < pT < 50 GeV. The data (black markers) are compared to the HERWIG simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 3 (bottom center): Data-MC comparison for the quark-gluon discriminant in dijet events for jets in the central region with 80 < pT < 100 GeV. The data (black markers) are compared to the PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).
	Fig. 3 (bottom right): Data-MC comparison for the quark-gluon discriminant in dijet events for jets in the forward region with 40 < pT < 50 GeV. The data (black markers) are compared to the PYTHIA simulation, on which the different components are shown: quarks (blue), gluon (red) and unmatched/pile up (grey).

Systematic uncertainties

Figures	Caption
	Fig. 4 (left): Validation of the smearing function method for jets with 50 < pT < 65 GeV and eta<2 in dijet events. The data (black markers) is compared to the simulation before (blue dotted line) and after (red solid line) the application of the smearing.
	Fig. 4 (right): Validation of the smearing function method for jets with 50 < pT < 65 GeV and eta<2 in Z+jet events. The data (black markers) is compared to the simulation before (blue dotted line) and after (red solid line) the application of the smearing.
	Effects of the dijet data-driven smearing on the Z+jet simulated events, separately for quark (blue) and gluon (red) jets with 50 < pT < 65 GeV and eta<2. The smeared distribution (markers) is compared to the simulation before the application of the smearing (filled histograms).
	Fig. 5 (left): Change in discriminating performance after the smearing for the quark-gluon discriminator in the center of the detector. The figure shows the efficiency as a function of jet pT for quarks (blue) and gluons (red) before (solid markers) and after (hollow markers) the smearing, for a fixed cut (>0.5) on the discriminator.
	Herwig++ efficiencies. Change in discriminating performance after the smearing for the quark-gluon discriminator in the center of the detector. The figure shows the efficiency as a function of jet pT for quarks (blue) and gluons (red) before (solid markers) and after (hollow markers) the smearing, for a fixed cut (>0.5) on the discriminator. This plot was updated following a re-evaluation of the smearing parameters.
	Direct comparison of tagging efficiency between smeared Pythia 6 (solid markers) and smeared Herwig++ (hollow markers) Monte Carlo generators in di-jet events.
	Fig. 5 (right): Change in discriminating performance after the smearing for the quark-gluon discriminator in the forward region of the detector. The figure shows the efficiency as a function of jet pT for quarks (blue) and gluons (red) before (solid markers) and after (hollow markers) the smearing, for a fixed cut (>0.5) on the discriminator.
	Pileup robustness in the central region. Change in discriminating performance after the smearing for the quark-gluon discriminator in the central region of the detector. The figure shows the efficiency as a function of number of reconstructed primary vertices in the event for quarks (blue) and gluons (red) before (solid markers) and after (hollow markers) the smearing, for a fixed cut (>0.5) on the discriminator.
	Pileup robustness in the forward region. Change in discriminating performance after the smearing for the quark-gluon discriminator in the forward region of the detector. The figure shows the efficiency as a function of number of reconstructed primary vertices in the event for quarks (blue) and gluons (red) before (solid markers) and after (hollow markers) the smearing, for a fixed cut (>0.5) on the discriminator.

Additional variables

Figures	Caption
	Fig. 6 (left): Single-variable perfomance comparison, in terms of quark efficiency and gluon rejection after applying a cut on each variable independently, of the studied discriminating variables for jets with eta<2 and 40 < pT < 50 GeV.
	Fig. 6 (center): Single-variable perfomance comparison, in terms of quark efficiency and gluon rejection after applying a cut on each variable independently, of the studied discriminating variables for jets with eta<2 and 80 < pT < 100 GeV.
	Fig. 6 (right): Single-variable perfomance comparison, in terms of quark efficiency and gluon rejection after applying a cut on each variable independently, of the studied discriminating variables for jets with 3<eta<5 and 50 < pT < 65 GeV.