Carl pointed out to me that the min-bias requirement actually only looks at the ADC of BBC-small tiles. While this may not change the message of the results, it is useful to examine the effect of tagging coincidence events on either small or small+large tiles.

Figure 1

BBC Large and Small	BBC Small

As Fig. 1 shows, there is a substantial increase in no-coincidence events relative to coincidence events when only looking to small tiles. I am actually surprised there is this much of a change, as now there are actually slightly more no-coincidence events (27888 vs. 29059). Moreover, the actual min-bias trigger places a constraint on the TAC and ΔT, which means I am actually tagging too many coincidence events. Given the event vertex cut, such requirements should have no effect on my di-photon samples.

Figure 2

BBC Large and Small	BBC Small

The rate of di-photons for each ID changes little for coincidence events (0.1692 to 0.1795) while it changes rather substantially for no-coincidence events (0.007898 to 0.04966).

Figure 3

BBC Large and Small	BBC Small

The resulting mass and eta-phi distributions seem relatively similar between the two samples. However, the final arithmetic is a bit odd. Given the final numbers (5005 coincidence di-photons and 1443 no-coincidence di-photons), I estimate a BBC efficiency of 77.6% which is quite a bit lower than the quoted 88%. If I assume the difference is due to additional single-beam background, I would expect some 671 non-collision di-photons in my no-coincidence sample, i.e. 46% of the no-coincidence sample. It seems likely that such a contribution would noticably change the shape of the no-coincidence distributions, however, no such difference appears to my eyes.

Figure 4

BBC Large and Small
BBC Small

One final check is to normalize the coincidence and no-coincidence samples and subtract the two, looking for variations signaling a background distribution. In Fig. 4, I show the results of such a procedure. In the "large and small" sample the constant fit seems perfectly reasonable, within 2σ of 0. For the "small only" sample the fits return perfectly reasonable χ² values, but the mass distributions are clearly different. The φ distribution is within 1σ of 0. The mass distribution is 2.6σ from 0, and there appears to be a dip within the signal mass window.