Cms silicon Strip Tracker Operation and Performance Gaëlle Boudoul

The Silicon Strip Tracker (SST) of the CMS experiment is, with 9.6 million readout channels, the largest strip tracker ever built. In order to correctly interpret and reconstruct the events recorded it needs to be precisely calibrated, thus ensuring that it fully contributes to the physics research program of the CMS experiment. In 2009 and 2010, the performance of the SST has been carefully studied using cosmic muons and tracks from proton-proton collisions at centre-of-mass energies of 900 GeV, 2.36 TeV and 7 TeV. In this paper, we present some results of the detector performance.

1. 
   Overview of the CMS experiment at LHC and specifications of the Silicon Strip Tracker
   

The LHC design luminosity of 10³⁴ cms^-2 s^-1 and the collision rate of 40MHz imposed major constraints on the design of the tracker. With an average of 20 minimum bias events per bunch crossing, more than 1000 reconstructible tracks are expected in the tracker. In order to keep the hit occupancy in the Silicon Strip Tracker (SST) of the order of a few percents, a fine granularity is mandatory. The tracking should be precise and robust in order to have a high efficiency and an excellent parameter resolution for both low and high transverse momenta. The primary and secondary vertices should be precisely reconstructed and the mass of the detector should be the lowest possible to minimize undesired particle interactions. The tracker has to be able to sustain high radiation doses and present a low risk of failure.

2. Overview of the Silicon Strip Tracker

After more than 15 years of development and construction, the SST was installed within CMS in 2008. It consists of 4 subdetectors: the Tracker Inner Barrel (TIB), with 4 cylindrical layers and 2724 modules, the Tracker Inner Disks (TID), with 3 disks on either side of the interaction point and 816 modules, the Tracker Outer Barrel (TOB), with 6 cylindrical layers and 5208 modules, and the Tracker Endcaps (TEC), with nine disks on each side and 6400 modules. The silicon detectors used in the SST are made of one (thin -320m) in the inner region or two (thick – 500m) sensors in the outer region. The Tracker covers the pseudo rapidity region |η|<2.5 [2].

Each detection module is composed of a silicon sensor made of p-type strips on n-type bulk. The interstrip distance (pitch) varies from 80 to 205 μm depending on the type of module and its position within the tracker. The readout of the signal is performed by the APV readout chip, which samples, amplifies, buffers and processes signal from 128 detector channels at a frequency of 40 MHz. In an operating mode known as peak, it produces a CR-RC pulse shape with a relatively slow rise time of 50 ns. In deconvolution mode, which is the standard mode of operation, it performs additional signal processing to constrain the signal to a single bunch crossing [3]. Each module is read out by 4 or 6 APV chips, depending on the number of strips, 512 or 768. On receipt of a Level1 Trigger, pulse height and bunch-crossing from the APV chips are converted to optical analogue signals through lasers. The data is then sent via optical links to the Front End Drivers (FED) [4], where the digitization of the data is performed. The FEDs perform pedestal subtraction, noise suppression, before carrying out cluster finding.

3. Commissioning and Operation

The SST commissioning is performed in several steps, which consist of tuning the laser gain for the optical readout links, optimizing the readout chip (pulse shape, analogue baseline), measuring the noise and the pedestals, and finally performing a fine delay scan to allow a synchronisation at the module-level down to 1 ns.

The operational fraction of the SST in 2010 is 98.1%, and the breakdown by subdetector is shown in Table 1.

Table 1- Operational fraction of the SST in 2010:

The signal-to-noise ratio is a benchmark for the tracker performance. It is measured using clusters associated to tracks and depends on the strip length and thickness. The signal-to-noise ratio in each subdetector is shown in Table 2. It varies from 19.2 and 23.7 in deconvolution mode, and is large enough to allow an efficient identification of particles with negligible background.

Table 2 - Signal to Noise ratio in peak mode (deconvolution) mode during 900GeV (7TeV) collisions:

Signal/noise	TIB	TID	TOB	TECthin	TECthick
Peak	27.4	26.7	34.1	28.8	35.7
Deconvolution	19.4	18.5	22.5	19.2	23.7

4.2 Hit Efficiency

To measure the hit efficiency of a given layer, the track reconstruction is performed without the layer under study. The efficiency is then computed considering the interpolation of these tracks onto the layer being studied, seeking hits on the modules through which the tracks should pass. The measured efficiencies are shown in Figure 2. Once the modules identified during commissioning as defective are removed from the measurement, the efficiency is above 99%. This method is thus also useful to monitor the SST and identify failures.

Figure 2 _ Hit efficiency as a function of the SST layer

The hit resolution is measured using overlapping modules within a layer, comparing the measured hit position and the position predicted by the interpolation. The advantage of this method is that it reduces the amount of material between two layers and reduces the effects of track extrapolation. The sensor resolution is known to depend strongly on the angle of the track and the pitch of the sensor. The results are therefore determined separately for different sensor pitches and in 10 degree intervals for the track incidence angle. The minimal value is achieved for an angle that corresponds to the optimal charge sharing. The results are shown in figure 3.

Figure 3_ Hit resolution as a function of the pitch for 3 different cluster sizes and for a local track angle between 0 and 10.

The 3.8T magnetic field causes the drift direction to tilt by the Lorentz Angle (LA), and modifies the distribution of the cluster charge on the strips. A correction must therefore be applied to the reconstructed cluster positions since any error in the assumed LA would result in an apparent misalignment of the silicon sensors. The correction is maximal in the barrel where the magnetic field is perpendicular to the electrical field.

The LA is measured with cosmic tracks by comparing the cluster width with the track angle on the sensors. The measured value of the tangent of the Lorentz Angle is 0.07 ±0.01 in the Inner Barrel and 0.09±0.01 in the Outer Barrel.

In deconvolution mode, an additional correction is needed, as a fraction of the charge from the sensor backplane does not reach the strips in time for readout, which biases the cluster position. The correction to the cluster position is measured using collision data and taken into account.

The mean energy loss through ionization (dE/dx) of a particle traversing a thin layer of material is given by the Bethe- Bloch formula [4]. For a fixed momentum and charge, the energy loss depends on the particle mass and it can be used for particle identification. A nice agreement is observed between data and simulation for the proton and kaon peaks (fig 4). The third peak observed in data is due to deuterons, which are not reproduced correctly in the simulation.

Figure 4 - Left: Reconstructed mass of low momenta hadrons (protons, kaons, deuterons). Right: Distribution of dE/dx versus momentum from 7 TeV collision events.

The performance of the CMS tracker after calibration with cosmic muon and collision data is excellent. A signal to noise ratio between 19.2 and 23.7 is measured, and high hit efficiencies and resolutions are obtained. Furthermore, the dE/dx measurements allow unambiguous observations of kaon, proton and deuterons in collision data.

[1] CMS Collaboration, The CMS Experiment at the CERN LHC, JINST

0803 (2008) S08004. doi:10.1088/1748-0221/3/08/S08004

[2] CMS collaboration, The Tracker Project, Technical Design Report,

CERN/LHC 1998-006, CMS TDR 5 (1998);

[3] M. Raymond et al, “The APV25 0.25 μm CMOS readout chip for the CMS tracker“, Nuclear Science Symposium Conference Record, 2000 IEEE, 9/113-9/118 (vol.2)