sPHENIX discussion of software

[Alexander Bazilevsky <shura AT bnl.gov>]

Hi Kurt,

More commenting on "purely electromagnetic showers" giving prob~0.

First of all, what you show looks reasonable to me. For the ideal case 
one should expect prob distribution to be uniform (just a const). What I 
see in your plot: in the fist bin of prob distribution there is an 
excess by a factor of 5-8, which mean that if you apply cut prob>0.01, 
for which you would expect to cut only 1% of electrons, you will 
actually cut 5-8%. Why?

Jin's comment is exactly right - there is a quite sizable chance for an 
electron to radiate photon, and that makes a response in the EMCal being 
not purely from a single shower. If you require really pure events (no 
photon(s) radiated) you will see that the peak at zero will decrease by 
a factor of two-three, leading to "only" 3-4% loss for prob>0.01.

Another effect comes from the tails of EMCal response. I don't know if 
you select electrons from the peak (of expected energy) or take all of 
them, including low energy tail. But if you plot prob for electrons from 
the tail, you will again see that they also contribute to prob~0.

... But even if you select really clean single electromagnetic shower, 
you'll still get some excess at prob~0. I'd say, the level of 3-4% is ok 
(the level of how I have tuned it by now). Probably some more tuning may 
improve it a bit. But that's for a future exercise.

Sasha.


On 11/6/17 8:46 PM, Huang, Jin wrote:
> Hi, Kurt
>
> Thanks for continuing the efforts on the particle flow algorithm.
>
> > Even at substantial energy, the purely electromagnetic showers from the 
> > single electrons have a significant chance to produce a prob value of near 0. 
> > Is this expected?
> Interesting. Sasha may better comment on it.
>
> Meanwhile, there is a non-negligible chance for the electrons to produce a 
> hard radiative photon in the tracking detectors prior to indenting the EMCal. 
> Then it can be pulled away from the photon by the magnetic field and produce 
> two cluster or a prolonged cluster, depending on how much is the magnetic 
> pull. The latter case could make the electron shower non-standard and low 
> prob.
>
>
> > ; I assume the default energy scale given to the EMC is calibrated to 
> > electromagnetic showers.
> Correct.
>   
>
> > If so, is the EMC hadronic energy scale known, or should I calculate it? Any 
> > advice here would be appreciated.
> User has to define and apply it.
>
> It depends on the shower development (MIP through or initiate a shower) and 
> energy (lower and higher energy has different scale). In the leading order, 
> one may choose a constant scale factor to convert the EM-shower energy scale 
> to hadronic energy scale, and then deal with the dependencies in the next 
> order corrections.
>
> Songkyo may talk about that tomorrow in the simulation meeting.
>
>
> > I plan to call into the simulations meeting tomorrow, so we could chat about 
> > it then if there is time.
> Sounds great. Would you like to be added to the agenda?
>
>
> Cheers
>
> Jin
>
>
> ______________________________
>
> Jin HUANG
>
> Associate Physicist
> Brookhaven National Laboratory
> Physics Department, Bldg 510 C
> Upton, NY 11973-5000
>
> Office: 631-344-5898
> Cell:   757-604-9946
> ______________________________
>
> From: sPHENIX-software-l [mailto:sphenix-software-l-bounces AT lists.bnl.gov] On 
> Behalf Of Kurt Keys Hill
> Sent: Monday, November 6, 2017 7:28 PM
> To: sphenix-software-l AT lists.bnl.gov
> Subject: [Sphenix-software-l] Looking into PHENIX clustering in sPHENIX
>
> Hi All,
>
> The Colorado group is interested in particle flow jet reconstruction in 
> sPHENIX. As a first step, we would like to separate electromagnetic from 
> hadronic EMCal clusters within the jet and apply the appropriate energy scale 
> to each type. We plan to eventually apply a track-cluster replacement scheme 
> as a next step.
>
> Moving toward EMCal cluster identification, I have merged Sasha's pull 
> request into a local copy of g4cemc. I ran single pions and electrons through 
> an otherwise default sPHENIX and plotted the chi2 and prob distributions in 
> four different EMC cluster energy selections (see attached slides). In the 
> lowest energy (0-1 GeV) the charged pion deposits MI energy which seem to 
> look a lot like low energy EM showers. At higher energy, there seems to be 
> fairly good separation between the EM and hadronic showers, and at the 
> highest energies the pi0 curve separates from the electrons - presumably as 
> the two photon showers merge.
>
> Even at substantial energy, the purely electromagnetic showers from the 
> single electrons have a significant chance to produce a prob value of near 0. 
> Is this expected?
>
> As an aside, I would like to understand the current applied calorimeter 
> calibrations better. I am using the calibrated calorimeter towers and 
> clusters; I assume the default energy scale given to the EMC is calibrated to 
> electromagnetic showers. If so, is the EMC hadronic energy scale known, or 
> should I calculate it? Any advice here would be appreciated.
>
> I plan to call into the simulations meeting tomorrow, so we could chat about 
> it then if there is time.
>
> Thanks,
> Kurt
>
> _______________________________________________
> sPHENIX-software-l mailing list
> sPHENIX-software-l AT lists.bnl.gov
> https://lists.bnl.gov/mailman/listinfo/sphenix-software-l

--
Alexander Bazilevsky
Brookhaven National Laboratory, Bldg. 510D, 2-232
Upton, NY 11973 Tel: 631-344-3712
Email: shura AT bnl.gov
-------------------------------------------------