Benjamin Hooberman, with Slawek Tkaczyk, Andrea Venturi and Kevin Burkett

Introduction

As discussed in http://venturia.home.cern.ch/venturia/GlobalRuns/PeakVsDeco/PeakVsDeco.html, there has been an observation of a difference in the alignment geometries obtained using CRAFT09 peak and deconvolution data. This is explained in a model by Andrea Venturi (Venturi model) which attributes the effect to a loss of charge from the sensor backplane, which takes the most time to traverse the detector and result in charge accumulation on the strips. This effect is greatest in deconvolution mode when the time window of charge integration is significantly shorter than in peak mode. The result is a bias in the position of the reconstructed hit in the detector module local u direction. When performing the alignment procedure, this bias translates to a shift in the extracted position of the detector in the local w direction (normal to the detector surface, pointing from the backplane to the strips). In TOB where the sensors are thickest and the effect is therefore maximized, an average deviation of 20 microns is observed between the aligned w positions of the detector modules between the peak and deconvolution alignment geometries.

In this study, we investigate the alignment bias in peak vs. deconvolution data using an alternative method, by looking at the track-cluster residuals in the two readout modes. This allows us to study the bias on a track-by-track basis, and in particular to look at the bias as a function of the track incidence angle and the arrival time of the cosmic muon with respect to the readout trigger.

Results

The plots shown below are based on a comparison of 1M events in Run 109046 (peak) and Run 110998 (deco) of the dataset /Cosmics/CRAFT09-TrackingPointing-CRAFT09_R_V4_CosmicsSeq_v1/RAW-RECO. Both runs are processed using the alignment geometry based on CRAFT09 peak data.

We first look at the track-cluster residual distributions in the u direction, and the corresponding delta w shift, as shown below.

A shift of 15 microns in deco data with respect to peak data is observed for TOB, while the corresponding shift for TIB is 2 microns. These values are consistent with those obtained by Andrea (20 microns for TOB and 2 microns for TIB) by comparing the peak and deco alignment geometries as shown below. The larger shift in TOB is due to the thicker sensors, for which the drift time from the sensor backplane to the strips is larger.

If the observed shift is due to the loss of charge from the backplane as posited in the Venturi model, we expect to see a correlation between the delta u shift and the difference between the track incidence angle and the Lorentz angle. This is checked by plotting delta u vs. tan(theta_trk)-tan(theta_LA) as shown below, and computing the correlation coefficient C.

As expected, positive correlations between delta u and tan(theta_trk)-tan(theta_LA) are observed, in particular for deco data. For TOB, the correlation is significantly larger than for TIB, as expected due to the thicker sensors.

The double-peak structure observed in these plots is due to two effects. First, the distribution of the Lorentz angle has two peaks at -0.1 and 0.1, corresponding to the two orientations of the detector v direction with respect to the global B-field. Also, the delta u distribution is broadened near tan(theta_trk)-tan(theta_LA)=0 since tracks which have such an incidence angle give smaller clusters, and in particular give a higher fraction of single-strip clusters for which the detector spatial resolution is degraded. This effect is greater in peak data, since in deco data additional cross-talk between the strips leads to fewer single-strip clusters.

We measure the muon arrival time using the time from the drift tubes (DT). Below are the distributions of DT time, the error in the DT time, and the number of DT hits used to calculate this time. To ensure that the DT time is well-measured, we require delta(DT time)<10 ns and number of DT hits>=25. The observed shift in DT time in peak vs. deco is under investigation.

Next, we look at the correlation between the delta w shift and the DT time as shown below for TOB and TIB.

For both TIB and TOB, the correlation between delta w and DT time is small in peak mode. In both TOB and TIB deco data a positive correlation is observed as expected, with the magnitude of the correlation about twice as large for TOB as for TIB. We divide the data into 8 bins of DT time specified by the boundaries DT time = -20, -12, -8, -4, 0, 4, 8, 12, 20 ns and take the average delta w and dt time for each bin. The results are summarized below.

A line is fitted to the delta w vs. DT time and the slope is extracted. A slope of 2.2 (1.0) microns/ns is determined for TOB (TIB).

To DO

• The comparison will be repeated using the full CRAFT09 peak and deco samples.
• We plan to look at collision data to determine if this effect is present. We will examine deco data in which the readout time wrt the trigger has been scanned in increments.