Cosmic calibrations for the Inner Detector

20 October 2008

Offline tracking uses filters to counteract the noise seen in the detectors



The last month has been a busy one for those commissioning the Inner Detector; due to the previous problems with the cooling system, it has actually been the first time that all three sub-detectors have been able to run together. With both silicon detectors finally properly cooled, the ID teams have been pressing ahead, using cosmic data to fine-tune their detectors.

A major element of this tweaking is the alignment of the thousands of silicon sensors. Pinpointing exactly where each sensor is in the Pixel and SCT detectors – how it is orientated and how it is displaced from its nominal position – means that any misalignment can then be compensated for at the track-fitting and calculation stages, thereby increasing the resolution of particle reconstructions.

But finding the positions of the sensors, each carrying close to 50,000 individual pixels, requires a little detective work. By looking at where the sensors are relative to computed particle trajectories, it is possible to figure out whether a particular sensor is displaced or is in the correct position.

“It’s a bit of a chicken and egg,” explains Thijs Cornelissen, a CERN fellow in the ATLAS Data Processing Group (ADP). “For a given sensor, you assume that a trajectory is perfect and try to ‘move’ the sensor such that it lines up with the trajectory,” he begins, “but of course, this trajectory is also affected by the misalignments of other sensors, so what happens is you need to iterate this a few times until all the sensors really do line up with all the trajectories.”

Initially, misalignments were of the order of about one millimetre; not insignificant in a detector which ought to be able to measure a track with a precision of a few microns if everything is perfectly aligned and calibrated. “Those really big offsets have already been corrected and the pixel detector is now moved by about one millimetre in the reconstruction, so it reconstructs tracks with much better quality,” says Thijs. “But we still have to perform the more detailed alignments down to the level of each separate sensor, to get a precision of say 10 microns.” In the case of the Pixel Detector, there may not be enough data to do this until collisions happen.

Happening concurrently to the alignment effort is the calibration of the ID. “The idea for collisions is to have a 24-hour based calibration running, to provide the best possible calibration alignment before we start the bi-processing of the data,” explains Christian Schmitt, also a CERN fellow with the ATLAS ADP. “Obviously for cosmics, we don’t get as many data as for collisions, so we’re still at the phase of setting up all the procedures.”

Right now, there are regular shifters in place looking through the latest data for problems like noisy pixels in order to provide updated constants and noise masks for the off-line reconstruction. There are still a few sensors that need to be debugged, but, according to Thijs, 95 percent are now being read out successfully.

Another important part of the calibration procedure is looking at TRT electron drift times. This is the time taken for free charge, left in the particle’s wake as it shoots through the TRT straw ionising the gas, to reach the wire at the centre of the straw. This time measurement can be converted to a radius – the distance of the passing particle from the wire – and used in the reconstruction.

Since cosmic rays only pass through the small Pixel Detector once every few seconds, it will be a few more weeks yet before the commissioners have hit their target of gathering at least 100,000 tracks. “You need a lot of tracks to have crossed each sensor before you can get an idea about its correct position,” explains Thijs.

By this point, the larger SCT and TRT detectors should themselves have recorded up to a million cosmic tracks, providing us with a solid understanding of the ID ahead of first collisions in 2009.


Ceri Perkins

ATLAS e-News