CMS Tracking POG Performance Plots for Vertex 2014

Track Reconstruction Performance

The reconstructed tracks are selected with the "high purity" flag (see arXiv:1405.6569).
The samples used are ttbar events simulated at sqrt(s)=13 TeV with different pile-up conditions (=25,40,70,140; bunch crossing=25 or 50 ns).
The releases used are either the one used during Run1 or the current development candidate for Run2 data-taking.
The matching of reconstructed track with simulated once requires that >75% of the reconstructed track hits come from the simulated track.

New Developments

Previous studies showed that timing of iterations seeded from pairs of strip hits diverges at high pile-up.
Two dedicated developments target time reduction of these iterations (pixelLess=iter5 and tobTec=iter6).
The first one is a new seeding algorithm, based on hit triplets whose main feature is a chi2 cut from straight line fit of 3 points in the RZ plane.
The second one is a cut on the strip cluster charge, rejecting clusters produced by out-of-time particles. It can be applied upfront tracking, during seeding or during pattern recognition and accounts for sensor thickness and trajectory crossing angle.
The sample used is ttbar simulated at sqrt(s)=13 TeV, 25ns bunch crossing and average pile-up 40. The release used is the current development candidate for Run2 data-taking.

Plot	Description
	Number of seeds vs eta for the strip-seeded iterations, comparing the pair-based algorithm with the new triplet-based one. Plots are stacked. The new algorithm reduces the number of seeds by a factor 2x.
	Number of reconstructed high purity tracks vs pT for the strip-seeded iterations, comparing the pair-based algorithm with the new triplet-based one. Plots are stacked. The new algorithm reconstructed approximately the same number of tracks.
	Tracking fake rate vs eta with and without the strip cluster charge cut. Fake rate is defined as number of non-associated reconstructed tracks divided by the number of reconstruced tracks. The cluster charge cut effectively reduces the fake rate by about 2x.
	Timing vs tracking iteration, adding sequentially the new developments described above. Both developments reduce the timing of the two strip-seeded iterations by 2x, achieving a 3x reduction on the total tracking timing.

Timing

Evaluated on Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz, comparing Run1 release and the current development candidate for Run2 data-taking. Timing is normalised to the time of Run1 tracking with PU=25 and BX=50ns.

Plot	Description
	Timing vs luminosity for ttbar+PU samples with 25 ns bunch crossing. Samples with different average pile-up are used as reported on the plot. Shown is timing of full CMS reconstruction and of the iterative tracking sequence. It was not possible to run on the sample with PU~140 with the Run1 release.
	Timing of iterative tracking sequence vs pile-up for ttbar samples with 25 ns bunch crossing. Large time reduction with respect to Run1 (~2x@PU25, ~3x@PU40, ~4x@PU70).

Track Reconstruction Performance

Track Selections for efficiency and fake rate

• consider only HighPurity tracks, track associator by hits (>75%)
• fake rate: all tracks
• efficiency:
  • vs : pT>0.9 GeV, VZ<30 cm, VR<3.5 cm
  • vs R: pT>0.5 GeV, VZ<150 cm, |η|<2.5 cm

General description

The plots that follow come always in three:

1. Track reconstruction efficiency vs eta
2. Fake and Duplicate rate vs eta
3. Track reconstruction efficiency vs track production radius

The definition of efficiency, fake rate and duplicate rate follow.

efficiency

it is defined as the number of matched reconstructed tracks divided by number of simulated tracks. The association between reconstructed and simulated tracks has been introduced and explained in the previous paragraph

fake rate

it is defined as the number of non-matched reconstructed tracks divided by number of reconstructed tracks.

duplicate rate

it is defined as the number of reconstructed tracks associated to the very same simulated track divided by number of reconstructed tracks.

Run1 vs Run2

Plot	Description
	Comparison of tracking efficiency for Run 1 and Run 2 typical PU conditions. Despite the harsher conditions for Run2 we were able to keep the same performance. We do expect the same physics performance from the tracking point of view in Run2 as we had in Run1.
	Comparison of the fake+duplicate rate for typical Run 1 and Run2 PU conditions. The fake+duplicate rate is at an acceptable level over the full eta range, with a slight increase in the central region for the Run2 release-conditions.
	Comparison of the tracking efficiency for Run 1 and Run 2 typical PU conditions as a function of the production radius. A slightly lower efficiency for displaced tracks in the Run 2 scenario is visible.

More specific comparisons

To better illustrate the effects of all the improvements that have been developed and integrated so far in the current development release, we also do a comparison between results obtained using the Run1 tracking vs the same obtained using the Run 2 candidate release.

PU25 BX50ns

In this specific comparison, which is tailored to mimic the harsher conditions met during Run 1, we can appreciate:

• Using Run2 tracking the same or higher efficiency for prompt tracks is observed.
• Run 2 tracking reached up to a factor of 2 reduction in fake rate
• The only downside is a slight efficiency loss for displaced tracks

Efficiency vs eta	Fake+duplicate rate	Efficiency vs production radius

PU25 BX25ns

In this specific comparison we can appreciate:

• Using Run2 tracking the same or higher efficiency for prompt tracks is observed.
• Run 2 tracking reached up to a factor of 4 reduction in fake rate
• The only downside is a slight efficiency loss for displaced tracks

Efficiency vs eta	Fake+duplicate rate	Efficiency vs production radius

PU40 BX25ns

In this specific comparison, tailored to mimic the harsher conditions that we will reasonably face during Run 2, we can appreciate:

• Using Run2 tracking the same or higher efficiency for prompt tracks is observed.
• Run 2 tracking reached up to a factor of 6 reduction in fake rate
• The only downside is a slight efficiency loss for displaced tracks

Efficiency vs eta	Fake+duplicate rate	Efficiency vs production radius

Primary Vertex Reconstruction Performance

Reconstructed vertices are selected requiring: ndof>4, |z|< 24 cm, R<2 cm.
Samples used are ttbar simulated at sqrt(s)=13 TeV, 25ns bunch crossing and different average pile-up conditions (=25,40,70,140).
Release used is the current development candidate for Run2 data-taking.
Matching with simulated vertices requiring: |deltaZ|<1 mm AND |deltaZ|<3sigmaZ (where sigmaZ is the reconstructed vertex Z uncertainty).

Plot	Description
	Number of reconstructed vertices vs number of generated pile-up interactions. Observe a linear scaling up to PU~80 and a strong deviation for higher PU.
	Number of reconstructed vertices matching a simulated vertex vs number of generated pile-up interactions. Observe an approximately linear scaling on the full range.
	Vertex reconstruction efficiency vs number of generated pile-up interactions. Linear decreasing trend is observed in the full range with an average of 70%. Efficiency defined as number of matched divided by number of simulated vertices.
	Vertex reconstruction fake rate vs number of generated pile-up interactions. Below 10% for PU<80, it shows a large increase (20-40%) for PU around 140. Increase in fake rate reflects poor tracking performance in the same PU range. Fake rate defined as number of non-matched reconstructed vertices divided by number of reconstructed vertices.
	Vertex merge rate vs longitudinal distance to the closest simulated vertex. Resulting curve is nearly PU-independent and thus reflects detector resolution and reconstruction performance. Above ~1 mm the merge rate is negligible, it reaches ~50% at 100um and 100% at 10 um. Merged rate defined as number of reconstructed vertices matched to more than one simulated vertex divided by number of matched reconstructed vertices.
	The signal vertex is identified as the reconstructed vertex with the highest sum of the squared pT of the tracks associated to it. The identification performance for ttbar events is similar for PU~25 to PU~70, with a large deterioration at PU~140 mostly due to increased fake and merged rate.

-- MarcoRovere - 09 Sep 2014