You might think it odd that work has already started to upgrade parts of the ATLAS detector, even before the LHC has started running!
As I wrote in a previous post, one of the key operating parameters for the LHC is luminosity, i.e., the beam intensity. The design calls for a luminosity of 1034/cm2/sec; this means that if you look at the beam head-on, it will contain 1034 protons/sec spread over an area of 1 sq. cm. In reality, each beam consists of 2800 bunches each containing ~1011 protons and about 0.03 mm in radius. Two such bunches collide every 25 nanoseconds, i.e., 40 million times/second; this is known as a bunch crossing.
The number of events that we collect is directly proportional to the luminosity. I had also written that the most common type of events that occur have a very large rate. What this implies is that when two proton bunches collide, most of the time we produce these “ordinary” events. For instance, when the luminosity is 1034, at each bunch crossing we get an average of 23 such events; they are easy to recognize, in the sense that they produce mainly low energy particles. So, when we produce an interesting event, say, a pair of top quarks or a Higgs boson, that event will be sitting atop this low-level “fuzz”. This background produces detectable signals in ATLAS, causing higher load on the trigger system and readout electronics, and pattern recognition problems in the reconstruction software. For instance, the (TRT) sub-detector, could have an occupancy rate as high as 10-20%, i.e., a non-negligible fraction of all wires will register hits; pattern recognition will be tough, but manageable. Other tracking sub-detectors are not a problem since they have much finer granulation (see Seth’s post on tracking). In addition, as particles produced in these “ordinary” events travel through the various sub-systems that are made out of silicon, they cause a slow degradation in them; detectors and electronics are made to resist this deterioration, but they do eventually fail.
When the luminosity starts to increase, as the LHC program calls for, these problems become much more serious. For instance, at a luminosity of 1035, each interesting event will be sitting atop a background of about 400 of these “junk” events; there is no way around it. As for the TRT, you can see that it will be useless.
The upgrade is planned in two phases. In the first phase, corresponding to a luminosity of 3*1034, currently scheduled for 2015 (but will depend on well the LHC performs), we plan to insert an extra layer of silicon sensors between the beam-pipe and the first silicon layer; by then, the first layer’s performance will have deteriorated enough to be noticeable. We will also upgrade some other detectors and electronics that sit close to the beam.
In the second phase, corresponding to a luminosity of 1035, all of the sub-systems that are used for tracking charged particles, i.e., the silicon layers and the TRT, will be replaced. All electronic readout for the detector will be replaced with more sophisticated stuff; the trigger logic and associated electronics will undergo a big overhaul, etc. This is scheduled for 2020.
These upgrades take time and effort, hence the early start. Most groups in the US and UK are already steering resources, in the form of people and money, towards the Phase I upgrade, e.g., my colleague, Hal Evans at Indiana University, is looking at some trigger improvements, and other countries are also joining in. We are used to this mode of operation; at one of my previous experiments (at the accelerator at Cornell University), we were simultaneously, analyzing data collected with version II of the experiment, starting to commission version II.5, and doing R&D on version III!
–Vivek Jain, Indiana University