Global Chi2 Alignment Using Cosmics

This page summarises studies into using simulated cosmics samples to align the Pixel and SCT components of the ATLAS Inner Detector using the Global Chi2 algorithm.

These studies were done in release 14.1.0 using InDetAlignExample-00-00-44, following the instructions for setting up this package here InDetAlignExampleHowTo. Only alignment of the silicon was attempted, there is no TRT alignment.

Cosmics Samples Used

Two different simulated cosmics samples have been used in this study:

• BField off cosmics sample simulated with ATLAS-CommNF-02-00-00 and described here (produced by Lauren Thompkins). This sample has been produced such that the hits on a track are only digitised if the track has both a TRT barrel hit AND a hit in the Pixel - thus it is optimized for tracks in the Pixel detector volume. Currently only using the High energy cosmics files, 296 files with ~450k events total - results correspond to ~80 hours of cosmics data assuming 100% trigger efficiency.

• B Field on cosmics sample simulated with ATLAS-Comm-01-00-00 adding alignment tag CMCCOND-CSC-00-01-00 (resulting in CSC misaligned geometry) and digitized with ATLAS-CSC-02-01-00 (again the misaligned geometry). This is the FDR2 cosmics sample described here CosmicSimulationSamples. It is optimized for tracks in the TRT barrel volume - a track is required to have a hit in the TRT barrel volume before it is digitized. The sample contains ~50k events, which correspond to only 40 minutes of data taking (cosmic rate in TRT barrel 20Hz and assuming 100% trigger efficiency).

The statistics in the BField off sample are thus considerably greater than the BField On.

Alignment from Nominal - Level 1 and Level 2

Cosmics samples simulated with the CSC misaligned geometry are initially reconstructed using the conditions DB tag which places the ID modules in their nominal positions OFLCOND-CSC-00-00-00. The Global Chi2 alignment is then run for 6 iterations (three at level 1 and three at level 2, NO level 3 iterations) in an attempt to realign the detector. The alignment is run out-of-the-box using the default settings for cosmics alignment. All events available are used (see above). No collisions events are used in the alignment.

The summary plots from the final (6th) iteration are shown here:

• B Field off ps.gz
• B Field on ps.gz

The following results show the level 1 and level 2 alignment constants that were arrived at after the full six iterations. The results are compared to the constants in OFLCOND-CSC-00-01-05 (previously determined CSC alignment used in recent detector paper) and the ideal alignment case (i.e. the actual known deformations from nominal that were put into the CSC geometry, described here SiliconMisaCSC).

Units are micrometers for x,y,z and mrads for alpha, beta, gamma.

These are the alignment constants text files for Bfield off and Bfield on from which the results below are derived.

Note that currently the CoG transformation is not applied to the resulting alignment constants

Conclusions

• ~80 hours of BField off cosmics data alone achieves a CSC realignment of the silicon at Level 2 that is at least as good as was achieved previously with the multimuons + cosmics (OFLCOND-CSC-00-01-05).
• This statement cannot be made for the Level 1 alignment currenty....need to find out the effective L1 constants for OFLCOND-CSC-00-01-05.

Level 1 Alignment Results

Shown is the relative alignment of the Pixel and SCT Barrel (Pixel Barrel - SCT Barrel) for each of the global alignment parameters. No L1 alignment constants were produced in the CSC exercise (so none in OFLCOND-00-01-05).

The cosmics result in very large L1 translations in the global Z direction. In order to investigate have added results for running alignment from nominal with 100k multimuons, and also running the alignment with multimuons starting from the ideal alignment. Can see that L1 Z shifts are still introduced but not to the same extent as in the cosmics.

Pixel Level 2 Alignment Results

Shown is the relative alignment of the Pixel B-Layer wrt the other two layers in the Pixel barrel for each of the global alignment parameters.

SCT Level 2 Alignment Results

Shown is the relative alignment of the first SCT layer wrt the other three layers in the SCT barrel for each of the global alignment parameters.

Alignment from Curl Deformation

The cosmics samples are initially reconstructed using the CSCMisaligned_RDeltaPhi_02 conditions database tag (the "Curl" deformation on top of the CSC misalignment produced by Tobias Goettfert, see InDetResidualMisalignment), and six iterations of the GX2 algorithm are run (again three at level 1 and three at level 2) in an attempt to realign the silicon.

So far I only have results with the BField Off sample.

The summary plots from the final (6th) iteration are shown here:

• B Field off ps.gz

Unfortunately it is not so straightforward to interpret the alignment constants results here, and do a comparison with the L1 and L2 ideal constants to see how well the weak mode has been removed. This is because, although the curl deformation is L2 deformation the systematic deformations are actually implemented as distortions at L3, and so although the ideal alignment may well have been fully recovered, this will not be apparent from looking at the resulting L2 constants since we need to look at L2*L3 to know the true relative positions of the layers. (This can be done with IDAlignVisualization but I haven't done this yet)

Here you can see how the InDetAlignmentMonitoring plots look before and after the alignment for a 20k Zmumu collisions data sample (pdf). It appears that the clocking effect in the barrel has been completely removed by the cosmics! (see slides 6, 9 & 10). The alignment has also resulted in a global shift of the detector relative to the beamline (slide 7), and a significant bias in z0 (slide 8). Slide 4 possibly indicates that the cosmics alignment has resulted in a shifting of the Pixel B-layer in the local Y direction, and maybe this explains the z0 bias?

Future Plans

• Get hold of a larger statistics B field on cosmics sample to make the comparison between constants achieved with B field on and off more meaningful.
• Run level 3 alignment with the cosmics (need large stats.).
• Run cosmics alignment on systematic deformations sets - can we remove weak modes with cosmics?

-- BenCooper - 05 Sep 2008