Reconstruction, November 11, 2008

Hi Blake,

These are all good but tricky questions. Because the pi0 width depends on the energies, angles, etc. of two photons it is a bit more tricky to try to parameterize.

One issue is that making a chi^2 / dof cut on the fit to the pi^0 will certainly sculpt the invariant mass distribution since the chi^2 is derived from this distribution and the relevant error matrix.

Regarding the width of the pi^0, the key thing here is signal purity. We would like to select events by cutting on the pi^0 mass or the chi^2 / d.o.f. (which is slightly more sophisticated). If the peak is wide then we pull in a lot of background underneath the peak. The "figure of merit" for any analysis is really the ratio of true signal pi^0 to background in some window around the pi^0 mass.

We haven't done studies like this in detail so I'm writing a bit "off the cuff" but maybe you can venture of and blaze a new trail here. What you might want to do is:

- select a few benchmark reactions that include a pi^0, this would get the rough kinematics right for the signal pi^0's - generate signal MC for these reactions - generate some significant amount of PYTHIA background as well. Here you might want to veto all generated events that have a pi^0 in them. This is then no longer a physically realistic background from an analysis standpoint, but it might offer you some insight into how you can reject pi^0 backgrounds that don't peak in two-photon invariant mass. - come up with a selection cut in two-gamma invariant mass and look at the relative efficiencies for signal and background events to pass this selection cut - change something -- energy resolution, segmentation, etc. and repeat the previous step -- keep everything else unchanged

When you do this it will be important that you just look at the raw two-photon mass. Cutting on chi^2 will give you troubles because to calculate chi^2 one needs to know the error matrix. But the error matrix comes from MC for one particular set of detector configuration. Once you change something about the detector or its resolution you effectively change the error matrix and you would have to (by hand) propagate this change into the code to get the correct chi^2 for the new configuration.

Hope this helps to get you started, and sorry for the delayed reply. Your note caught me last week as I was in a rush to get prepared to go to Israel for the PANIC conference.

Cheers,

Matt

On Nov 5, 2008, at 12:53 PM, Blake Leverington wrote:

Hi Matt,

So before this error and a bunch of personal issues came up I was trying to figure out how to show how well the calorimeters perform using pi0 mass reconstruction. Is this something that can be shown as a function of length for the BCAL (except there are 2 photons, so...?) ? Should there be distributions for BCAL only, BCAL/FCAL and FCAL only? What probablility/chi^2 cut do I make? This looks like it affects the width of the pi0 mass distribution. It seems like a single pi0 distribution plot doesn't suffice to show how well something works. Also, I understand that a narrow distribution is better but what is considered a good pi0 width??

I'm also using a pi0-eta reaction. Would it be better to just use a pi0 cannon?

Sorry for all the questions at once.

-Blake