Data rate of Atlas and CMS, LHC's detectors too slow?

Question

There is a technical question I always curious about to ask a CERN expert? I have read, http://nordberg.web.cern.ch/PAPERS/JINST08.pdf, page 5,

that the data sampling rate, number of stills taken from the collisions on the LHC’s Detectors Atlas and CMS are about 40KHz and probably this number to be upgraded up to 100KHz at the end of this year when the LHC is restarted with the higher luminosity beams (i.e. more collisions per second, currently at 1GHz). From these 40K/s (one event still, sampled every 25μs) to 100K/s (one event still every 10μs, upgraded later this year) sampled detector events from the collisions via the L1-trigger system only 200/s events are selected and recorded as being statistical important. All the other sampled events are ignored and deleted. Thus, in the best case one sampled event (still) every 10μs! (also, recorded events 200Hz thus about one event recorded as possible significant every 5ms in average).

Don’t you believe this rate to be too SLOW and that a very fast decay event could be missed out and therefore a potentially statistical significant result being never recorded for further analysis especially now where we are looking for new very high energy physics at dimensions of the order of 10E-17 cm?

I mean I find this sampling rate used, too slow and a bottleneck and low probability in order to catch these new hypothetical high energy particles?

Of course this could be statistically fixed by recording data for many years like the latest LHCb possible related to leptoquarks 3.1σ potential discovery. They collected these data for the last 10 years!! to reach only at 3.1σ:

https://physicsworld.com/a/has-a-new-particle-called-a-leptoquark-been-spotted-at-cern/

But then again what if a decay event duration is much less than 10μs, a fraction of the sampling period wouldn't that imply that more and more significant collision events would be more and more totally missed out and more and more years of recording events will be needed as we probe higher and higher energies to come to a statistical significant result? Maybe in the worst case scenario an exponential function?

What is the time-of-flight resolution of the detector sensors? I expect this to be a millionth fraction of a picosecond ? Right?

I believe the CERN members are aware of this bottleneck of the detectors and are blaming it on budgetary limitations on CERN.

Answer 1

There seems to be some confusion about how collider experiments works, particularly with the question, "Don’t you believe this rate to be too SLOW and that a very fast decay event could be missed".

There is no relation between the event rate and the decay rate of an event.

I am familiar with DESY, not CERN, so I don't have any numbers (other than 27km and 7 TeV) available, but it goes like this: the protons (counter)circulate in bunches, and the bunches collide at known times. During collisions, the detector is looking for events.

When an event occurs, whether is interesting, exotic, fast or slow, everything traverses the detectors ultra-relativistically, effectively at $c$, so the duration of the event is the size of the detector divided by $c$. (See note 1 at the end).

The problem starts with the beams: they are protons. A 7 TeV proton, in the lab frame, is basically a 2 dimensional object with no time evolution. It is as flat, frozen, pancake with 3 valence quarks and pretty much an unbound number of low energy QCD vacuum fluctuations called sea quarks and gluons. These all carry a fraction of the 7 TeV of momentum, and lower energy things have high cross sections (because of unitarity).

That means most events are garbage collisions with not enough energy to probe beyond the standard model. Recording them would render doing any physics impossible, as all detector systems have a dead time after triggering (I don't not know what the numbers are at LHC, but even a few nanosecond means the detectors and the read-out, write systems would be paralyzed with garbage).

To avoid this, there are various levels of hardware and software triggers. Hardware triggers just look at AND, OR, NOT, XORs, etc of electronic logic gates coming directly from detectors (so they are fast). They may be multi-level, and multi channel (for different processes).

If the hardware triggers, the data goes to fast software (or maybe FPGA firmware, first), for more filtering. Eventually, an interesting event tells the system to save the event.

After reading and recording an event, the whole system needs time to recover, so there is a minimum time until the next event can be recorded, which may be less than the average time between event because of downstream bottlenecks. Since many experiments measure absolute cross sections, correcting for dead time and trigger efficiencies is hugely complex, requiring sophisticated Monte Carlo analysis and various system measurements on the lab bench.

Note 1: So you maybe ask about producing a massive boson at rest in the lab. It will just sit there until it decays. The lifetime of massive particles is on the order of the time it take light to cross a proton, so it's irrelevant.

Moreover, even though the COM of the protons is the lab frame, the progenitor quarks (or gluons) do not have equal an opposite momenta, as they are randomly selected from a blob of QCD vacuum fluctuations, and occasional valence quarks.

That greatly complicates the experiment, and is why $e^+e^-$ colliders are so attractive. With them, the progenitors of event are exactly known, they just can't be stored in a 27 km ring at 7 TeV thanks to synchrotron radiation.