No beam? That never stopped us before…

3 November 2008

Smiling collaborators in the Control Room on September 10th.

After the huge effort everybody put into getting ready for the first beam, and the very successful start-up we had on September 10th – 11th, beam operations stopped following the incident in Sector 3-4 of the LHC. This presented a new set of options for the experiment.

While the situation in the machine was being assessed in the days after the incident, we decided to keep ATLAS in full operation, taking valuable commissioning data while waiting for more concrete information. As it became clear that there would be no more beam operation in 2008, a new plan was put in place that accounted for both the need for more commissioning data with a combined system and the work to be done on the experiment over the shutdown.

There were several objectives to be fulfilled before the end of combined data taking in 2008. The SCT and Pixel detectors especially had only be in full operation for a few weeks, and it was important to collect enough data for a first characterisation of the detector elements, alignment and DAQ stability before the shutdown. Alignment was also a priority for the muon system, requiring many millions of muon tracks. In addition, following an earlier gas overpressure incident on the small wheels, it was crucial to collect enough data to assess the efficiency of chambers that might have been affected, to determine if repairs have to be done during the shutdown.

Data taken with the combined system is of course also the only way to study correlations between sub-detectors, like combined muon tracking, relations between tracks and calorimeter clusters, and the use of all systems in the high level trigger. And last, but by no means least, running the full system is the only way to assess our operational efficiency, from the basic scaling of detector and service systems to communication patterns in the control room, data quality assessment and much more.

Given the programme outlined above we decided to keep the experiment in full 24/7 operation until the opening of the experiment for shutdown work had to start towards the end of October. After concentrating on getting everything ready for collisions, the new programme required two rapid changes to our setup – on the timing and the trigger, in particular the high level trigger.

Already on the 10th of September we had successfully moved through the first part of timing-in our beam pickups (the timing reference for beam operations) with the trigger system, and had taken data with the first circulating beam and later the first RF-captured beam. Now we quickly had to change the setup back to be as efficient as possible for cosmic-ray muons, which have no phase relation to our clock, different time-of-flight for the top and bottom parts of the detector w.r.t. the IP, and are generally much less fun than collisions.

On the other hand many sub-detectors require the trigger timing to be stable within a few (ideally 1) clock cycle, to find the data in their read-out windows. The initial setup was respectable, delivering good events from the beginning, but over time the main triggers moved close to containing all events within one cycle! Moving from this setup to collision timing next year promises to be a much easier task than we faced this time around, thanks to the hard work put in by everyone, and the lessons learned in deriving our current setup.

Another system that needed to react very quickly was the high level trigger. Again all effort had been put into being ready for collisions, and suddenly a quite different problem had to be solved. Our L1 trigger rate for cosmic-ray muons is around 500 Hz, however, with an event size of several MB for the full system in this mode, only about 10% can be recorded.

On the other hand the rate of cosmic rays crossing the Inner Detector is much smaller, so small in fact that we can not afford to miss any. But the L1 trigger does not include the Inner Detector, and therefore the high level trigger moved from the intended initial observer role for collisions into the central role of saving all good events crossing the inner tracking volume. Dedicated algorithms were put in place and continuously improved based on feedback from daily data quality meetings, and, using both the L2 trigger and the Event Filter, the efficiency compared to reconstructed tracks in control samples is now very close to 100%.

To single out these aspects should of course not distract from the huge effort made by all systems, detectors as well as DAQ, infrastructure, cooling, magnets, computing, offline reconstruction, Tier-0, safety systems … to provide a stable system that is more than just the sum of its parts – a working ATLAS experiment! They all deserve articles of their own, and no doubt will get them before long.

Did we achieve our objectives? I say yes. Many tens of millions of events were recorded for muon alignment studies, and more Inner Detector tracks were recorded than we had hoped for. The programme of qualifying the small wheel TGC chambers was completed, with a first analysis showing no loss of efficiency. Millions of events with tracking and calorimeter hits await combined analysis. The recent ATLAS week was full of interesting results from both beam and cosmic ray data, and I am sure many more are on the way.

So nothing more to do? Far from it. The last weeks have been just as useful in discovering problems in the system as they are for the data we took. The combined run may have finished, but an intense programme of calibration and development is under way already, that will fill all the time we have from now until we have our next rendez-vous with Mr. Beam next year.


Thorsten Wengler

University of Manchester