sPHENIX discussion of software

[Edouard Kistenev <kistenev AT bnl.gov>]

I bet it will not be much different. Want it or not but this calorimeter is coming (in simulation) much better (in simulation)  and much cheaper (in design) then anything actually built under the same name. It all boils down to a very simple fact - by sheer chance it is well optimized for measuring low pT and high pT components in essentially different devices with different sampling fractions. And it generates enough of multiple scattering to make it almost insensitive to tilt angle. Unfortunately it makes its energy resolution misbehaving when we are trying to push it into framework of stochastic+constant terms, it is plain wrong.

Separate (and very important) question - will it live to its promise. Definitely not. Geant maybe good for any of the components taken separately but fails miserably when applied to multicomponent system. Even subtle difference in e/h and e/m between components will …… immediately be reflected in spoiled low/high energy tails (the fact that one disappears and another one appears with momentum is just an e/h impersonation). 

In that respect (how to make components members of the same family) - we will probably never be able to do it in the central calorimeter, they will fight creating artificial breaks at 10GeV/c or somewhere else. But I am sure the problem can be solved in the forward direction where we are still free with our design choices both in structure and in readout. 

Edward

On Oct 21, 2015, at 12:33 PM, John Lajoie <lajoie AT iastate.edu> wrote:

Dear sPHENIX Folks:

    After the spirited discussions in both the simulations and HCAL meeting yesterday regarding the energy resolution we are seeing in the MC, John Haggerty pointed me to a very nice paper from the CALICE collaboration:

http://arxiv.org/abs/1510.04063

I have been pretty vocal about voicing skepticism about the low-energy neutron response in GEANT4, but they show time structure data that seem to show the late time (neutron) hits are well-reproduced with the QGSP_BERT_HP physics list in a tungsten calorimeter.  They also claim (but don't show) that "for steel the high precision modelling has much less influence", which is consistent with what we have seen when comparing HCAL simulations with QGSP_BERT and QGSP_BERT_HP physics lists (QGSP_BERT is our default).  OK, I'll shut up about the neutrons.

Something else that was discussed in the HCAL meeting in particular was whether or not what I am showing is inconsistent with what has been shown in the proposal. It certainly is (see the second of the attached slides).  The low energy response is significantly better in the new simulations, which is reflected in the improved stochastic term when fitting the full distribution. However, the fit over all momentum does have a very poor Chi2.  If I fit only to points >16GeV (last slide) the Chi2 is much better and the projection to lower energies overshoots the data.

Edward suggested that some of this might be due to the fact that the EMCal is quite different in the new simulations (current simulations use the 2D spacal) and low-energy hadronic showers deposit a fair fraction of their energy in the EMCal.  The next-to-last slide shows the EMCal energy fraction distributions as a function of energy for pi- at eta=0 and a selection of energies.  There is certainly a transition below 10GeV.

Over the weekend I'll try to get resolution results with towering and clustering for comparison.

Regards,
John

--

John Lajoie

PHENIX Deputy Spokesperson

Professor of Physics

Iowa State University

 

(515) 294-6952

lajoie AT iastate.edu

Contact me: john.lajoie

<compare2.pdf>_______________________________________________
Sphenix-software-l mailing list
Sphenix-software-l AT lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/sphenix-software-l