Single photons with EM37 - comparing converted and non-converted photons
<(Et-E)/Et> histograms versus eta.

---

Comment: I wrote an athena algorithm that extracts E and Et for unconverted photons. This is somewhat different from Mike's algorithm, as described below. Note that the root file used here is different from Mike's root files, because I had to modify the actual athena algorithm and add some Ntuple branches. The new root files, however, should be backward compatible with old code.
Here is the new athena algorithm code: Conversion.cxx and Conversion.h (search for "Anton" inside to find edited places)
Here is the new ROOT macro code and the root file on which it operates.

Converted photons only	Non-converted photons (uses AOD Photon objects)	Non-converted photons (uses CaloCluster energy directly)
Event basis: Only converted photons are used. Their energy is reconstructed in the following way: We use the energy of a CaloCluster whose eta is matched to the MC conversion point. If the match is successful, we use this photon; otherwise we discard it. Note: the brem-loss of energy by the electron/positron pair is not compensated!	Event basis: Only non-converted photons are used. Their energy is taken to be the AOD's reconstructed Photon energy. Moreover, we preselect only those events that satisfy the following set of criteria: There is only one MC photon There are no other MC particles besides the photon There is only one reconstructed photon (which, this, has to correspond to the above MC photon) We put in a consistency cut: the absolute difference of the truth and reconstructed energy must be less than 30 GeV, and their difference in eta must be less than 1.0. If these cuts are not satisfied, we drop the event In other words, these are the events where only one photon is generated at the interaction point, and it travels all the way to the LAr calorimeter unconverted. Note: This set of criteria is perhaps overly restrictive. But we actually get enough statistics (see the plots). In the future, it will be possible to lift most of these criteria by doing a statistical matching of each of several MC photons to reconstructed Photon objects (there is an AnalysisTools algorithm for this in Athena).	Event basis: Only non-converted photons are used. Unlike in the second column, their energy is taken to be the energy of a CaloCluster (and not the energy of Reconstructed AOD Photons). This time, we preselect only those events that satisfy the following set of criteria: There is only one MC photon There are no other MC particles besides the photon There is only one CaloCluster that fired up for this event In other words, these criteria are 99% compatible with column 2, i.e. they pretty much contain the same events (as can be seen from their "# of points in the average" plots). Important: As we can see, the energy response in this case is extremely similar to that in the middle column. It appears that the energy of reconstructed AOD photons is just the energy of a cluster. If needed, I can plot the difference between the two energy responses (using Reco photons vs directly using CaloClusters).

Some comments:

• This study is totally work in progress, and I would like to hear some suggestions/hints.
• It is important to realize that for unconverted photons, I am considering only very "nice" events - with only one particle coming from interaction point that leaves only one calo cluster fired.
• In the future, I may repeat these plots by considering not only singe non-converted photon events, but events with an arbitrary number of unconverted photons. This will allow to collect more statistics (for example, by looking at higgs->gamma,gamma files).
• The comparison between the converted and non-converted cases cannot be easily made becase I use different statistics. Thus, the second and third plot look much smoother at least in part due to much higher statistics, especially at lower etas.