[Bonuslist] BoNuS12

[Sebastian Kuhn]

Dear Bonusites,

I did a quick calculation regarding the issue of accidental coincidences. Please check whether I made a mistake!

1) During BoNuS6, we ran with an average luminosity of about 0.5*10^33 /cm^2/s on an effective target of 16 cm, which translates to a luminosity of 3*10^31 /cm^2/s per cm of target length. This is a LOT less than we had originally planned, but unless I made a math mistake, that's what I get from the typical beam currents (about 20 nA at 5.3 GeV) that we ran with. NOTE THAT WE WANT TO INCREASE THIS BY A FACTOR OF 40 for BoNuS12!

2) For our coincidence events, we required that the difference in vertex position between the RTPC proton and the CLAS electron be less than 15 mm. This means that any "fake VIP" produced accidentally within a 3 cm stretch of the target would have been counted as a coincidence. This corresponds to about 1*10^32 /cm^2/s "effective luminosity" within vertex cuts to produce accidentals.

3) We also required a (rather loose) timing cut through our sdist and edist cuts. It's a bit hard to translate those into time windows, but roughly speaking, a cut from -3 to +8 mm on either sdist or edist translate to about 1/3 of the typical length of a well-contained track, and since the total drift time for the full track is 6 µs, we can conclude that this cut corresponds to an effective 2 µs coincidence window.

4) Both Slava and Nate find that, within these cuts, we have on average 23% accidental coincidences. Since our VIP "efficiency" is around 2% for BoNuS6, this translates to a 0.5% likelihood that a proton that LOOKS like a good VIP coincidence but is really accidental appears each time we "open" one of these 2 µs windows (electron trigger in CLAS). Keeping round numbers, we can conclude that the random "fake VIP" rate is about 2.5 kHz for an effective luminosity of 1*10^32 /cm^2/s within vertex cuts.

5) One obviously unknown ingredient is how this rate will change by going from 5 to 11 GeV beam energy. My gut feeling would be that it won't change all that much - after all, it takes only a few MeV to produce a "fake VIP", so the beam energy will be relatively unimportant. (Most "fake VIP's" will come from - nearly - real photons given off by beam electrons, and my understanding is that the number of photons per energy bin delta-k is roughly given by s/k where s is the target thickness in radiation lengths and k the photon energy, largely independent of the incident electron energy). But I'm eager to hear more expert opinions on this. (One COULD try to run a Monte Carlo that simulated inclusive hadron rates to see how they depend on beam energy, but I'm not sure that I would trust this for low-momentum protons).

6) Assuming we want to run with a luminosity of 2*10^34 /cm^2/s with BoNuS12, and assuming 4)-5) above, we can estimate the likely accidental background rate under various scenarios. I am proposing one of these scenarios below - others might want to change my numbers:
- Assume we use 40 cm of effective target length (meaning that the non-shielded part of the BoNuS12 target extends 40 cm) but the same target density as before: In that case, the luminosity is 5*10^32 /cm^2/s per cm target length.
- Further assume that we can improve the vertex resolution such that we can apply a vertex cut of +/- 5 mm , so the effective luminosity within cuts is 5*10^32 /cm^2/s . This makes some sort of forward Vertex detector for CLAS12 crucial - we must find out whether we can use the µMEGAs in the forward direction. We probably also need to improve the RTPC vertex resolution. Here, it might help if we can increase the drift length AND the amount of charge produced per cm (using Ne/DME instead of He/DME) so more pads are involved in each track fit. BTW, this argues that we really need to double the number of electronics channels (or multiplex somehow) so we don't have to make the pads longer than the 5 mm they have now.
- Adding all these assumptions, I now come up with a prediction of 12.5 kHz "fake VIPs" within vertex cuts.
- We MUST reduce our timing window to at most 1 µs or, preferably, 0.5 µs. This should be quite possible, if we use a somewhat "faster" chamber (higher HV and/or faster gas), 4 cm instead of 3 cm track length, and tighter sdist/edist cuts. Regarding the latter: If you look at Nate's figures (4.40 and 4.41 in the analysis note), you can see that already we could cut a bit tighter without losing too many good tracks. Hopefully, this will improve further if we can manage to produce absolutely "wrinkle-free" cathode foils and SEMs, and spend (even) more effort on an accurate time-distance calibration. In any case, we can always CLAIM we will in the proposal... ;-)
- Assuming our VIP efficiency remains 2%, we get an accidental fraction of 0.6% / 2% = 0.3, which is quite manageable. In reality, we will of course (claim that we will) increase our acceptance for VIPs significantly (by eliminating the "ridges" and getting more backward angle acceptance) *), but this will affect accidentals the same way as true coincidences, so the accidental fraction 0.3 is not going to change.

So I am cautiously optimistic we can convince the PAC that accidentals are manageable. I am still worried about the latency of charge inside the drift gas - hopefully someone (else) can look at the experience from EG6 and come up with some estimate at which point this may become a problem. (Again, it will help to have a longer detector in this case IF what matters is the latent charge/unit volume, not the overall amount inside the drift region).

The last issue is the trigger rate. As far as I remember, we ran at 500 Hz for BoNuS6. Yesterday I heard that we can run as high as 3 kHz with the EG6 setup. That's a factor of 6, but still not enough for the factor 40 increase in luminosity we are proposing (!!!). Therefore, we MUST pursue either a hardware level 2 trigger (cutting down on junk triggers so we collect only good electrons - at 3 kHz) and/or a new RTPC DAQ system (preferably one that is much faster, simpler and more tailored to our RTPC than the ALICE system). Again, I'm not an expert, but I've heard some good suggestions (FPAs custom-made at JLab? Getting something from our collaborators at Argonne or Los Alamos National Labs?...)

'nuff said - Sebastian

*) I think it WOULD be helpful to have a standalone simple simulation of the RTPC acceptance to see how much it increases if we have NO dead regions in phi AND a longer detector - at least 40 cm, slightly set back relative to the target.