Study of the Pi+ Pi- P events from g6c run of CLAS

The purpose of this study is :
a) to setup and test the software framework to be used for future analysis of more complex topologies;
b) to come up with some reasonable procedures and cuts for data selection;
c) to verify that PWA results for this well-known topology do make sense.

Data Sample

The initial data sample consisted of events in which incoming beam photon and outgoing proton and pi+ were positively identified. A loose missing mass cut for (gamma p -> pi+ p) was applied. There are 34567 events in this sample. It can be divided further into events with pi- and without pi- as well as into events with some extra particles (denoted X) and without them. Here is the breakdown:

(A) Pi+ Pi- P 4175 events (Mass plots)
(B) Pi+ P 15457 events (Mass plots)
(C) Pi+ Pi- P X 1747 events (Mass plots)
(D) Pi+ P X 13188 events (Mass plots)

Therefore, "clean" events of type (A) constitute only 12% of the initial sample. It would be nice to recover some events from (B), (C) and (D) to increase the final statistics. In case (B), this is a question of how well assigning the missing momentum to pi- works. In case (C), it is a question of how to distinguish "real" extra particles X from those which do not belong to a particular event. Case (D) is a combination of (C) and (B).

Study of extra particles

Bottom right plot on Mass plots in (C) and (D) shows the invariant mass of all extra particles. Those are mostly photons, with a significant fraction of neutrons as well. Kaons are also seen. Also, there were events with antiprotons, De, extra p, pi+ or pi- as well as "unknown" particles (their contribution is not that big and they were discarded for now). A few conclusions:

1. Neutrons. Almost all extra neutrons are "fake". When they are disregarded, all (C) or (D) distributions look similar to (A) or (B). Their inclusion, to the contrary, shifts missing mass significantly away from the expected values.

2. Kaons. There are almost no kaons in (C). However, there are a lot of K- in (D) when pi- was not detected. This points to a case of mistaken identity. Simply reassigning pi- ids to those K- works well and produces distributions similar to (B).

3. Photons. Initially, it was hoped for most of the photons to be "fake" (out of time, noise, ADC pedestals, etc.) in order to "recover" those events by simply disregarding photons in them. However, there is a clear pi0 peak from the 2-photon events (bottom right on plot (C)). This rules out non-physical origin of many "photons" (aka noise, pedestals). Also, a strong pi0 peak in (C) is almost gone in (D). I don't know of any reason why there should be a correlation between pi- and "background" pi0 (but I can imagine a correlation between pi- and pi0 from the same event). The problem here is that missing mass is not very sensitive to those pi0's, and even more so to photons in single-photon events (which are by far the largest fraction of those "extra X" events - see the leftmost bin on the above-mentioned plots). I'm looking now at the possibility of applying non-kinematic cuts (i.e., TDC time window) in order to distinguish real and fake photons and to recover events with the fake ones.

Acceptance corrections

This plot shows M(pi+pi-) mass before and after acceptance correction for events (A) and (B). The good news is that initially non-similar distributions look more similar after acceptance correction. The bad news is that what looked like a strong f2(1270) peak is much less prominent after acceptance correction. Of course, nothing wrong with that (I don't know rho and f2 photoproduction cross sections) but... We shall see what comes out of PWA.

Missing mass studies

To my surprise, I was puzzled for quite a while with such a simple distribution as a missing mass spectrum. Missing mass for events gamma p -> pi+pi-p (A) (top left) is unusually narrow while missing mass for events gamma p -> pi+p (B) (bottom left) is much-much-much wider (even so it peaks at the pion mass as expected). Naively, I was expecting to see about the same missing mass resolution for events (B) as for events (A) assuming that the only difference between two samples is in the detection of pi-. Dissimilarity in the widths of two distributions was attributed to a large fraction of junk in sample (B), and I spent a lot of time trying unsuccessfully to clean it out.

Finally, I took sample (A) and artificially turned it into type (B) by simply discarding a pi-. Surprisingly, the resulting "missing mass" (top right plot) was anything but narrow. I was so amused that I even did a primitive "kinematic fit" to a zero missing momentum for events (A) (see bottom row of this plot). Logically enough, a delta-function for MM2(gamma p -> pi+ pi- p)=0 resulted in a delta-function for MM2(gamma p -> pi+ p)=M2(pi-). However, a barely noticeable widening of the former resulted in a significant widening of the later.

Two implications of this result:
1. It looks like some kind of a cut on MM2(gamma p -> pi+ p) was applied to the initial sample. We may want to relax this cut: it looks like a lot of type (A) events were killed by this cut (see top right plot) even if the overall missing mass for those events is almost perfect (top left).
2. Because a narrow appearance of a total missing mass distribution does not necessarily lead to narrow distributions in other variables we may want to consider implementing some kind of a kinematic fitter.