Following up from this post concerning the modeling of the invariant mass distribution using different monte carlo samples for the signal and three sources of background, I respun through the modeling algorithm with an added smearing to the masses.  The procedure is outlined below.

 Goal:

Properly model the signal and background to the invariant mass plots using four single particle MC sets normalized to fit the data.  Further, subtract out background contributions to arrive at a raw yield for each pt bin.

Details:

Data Vs. Single Particle MC (see here)

Cuts:

• Data events pass L2gamma trigger and L2gamma software trigger
• Cand Pt > 5.5
• Charged Track Veto
• At least one good strip in each SMD plane
• Z Vertex found and |vtx| < 60.
• Zgg < .7
• | particle eta | < .7

Bins:

9 pt bins, with boundries {5.5, 6., 6.5, 7., 7.75., 9., 11.5, 13.5, 16., 21.}

Smear:

The mass and of the single pions are smeared by sampling from a pt dependent normal distribution of the form N(1.005*Pt,.04).  Mass = smear*smear*old_mass.  This is supposed to mimic not only the detector resolution, but also the artificial increase in photon energy resultant from excess jet background at higher Pt.  

Obviously this is not the correct way to do this sort of smearing; it should be done at the BEMC level in the simulation, but this is a rapid way to test out assumption that smearing will better recreate the mass peak.  

Plots:

1)

Above is a plot of the invariant mass distributions for the 9 pt bins used in this analysis.  The black crosses represent the data (with errors.)  The four colored histograms show the invariant mass distributions of pion candidates found in single pion MC (purple), Single photon MC (red), single eta MC (blue) and mixed-event MC (green).  The four distributions are simultaneously fit to the data.  

2)

The above plot shows a data/MC comparison, where the red MC curve is the sum of the four single particle curves shown in plot 1.  As you can see (especially in bins 3-7) the single particle MC much better recreates the mass peak with the smearing.  It is not perfect, but compared to the original test, at least by eye, it looks better.  

Conclusions:

I think this test supports the conclusions that the BEMC response for single pion simulation was too precise originally.  Extra Pt dependent smearing should be added into the BEMC tower response.