sPHENIX HCal discussion

["Sickles, Anne M" <sickles AT illinois.edu>]

Hi Edward,

I’m not sure I understand your comment.  I wasn’t suggesting that calorimeters were the only thing to use in jet measurements, but suggesting using the sampling information jet by jet to rescale the energy due to how we expect the response to differ for different jet fragmentation scenarios.  Obviously tracks can be of use here as well, but I think the calorimeter information provides a useful starting point.  This can only improve the jet energy resolution compared to measurements which don’t use the sampling information though the amount of improvement needs to be quantified and depends on the details of the detectors.

Best,

Anne

On Sep 13, 2017, at 12:07 PM, Edward Kistenev <kistenev AT bnl.gov> wrote:

Hi, Ann, 

ATLAS spent tons of money to make calorimeters uniformly responding. They respond differently to e/h but both e and h are seen more or less the same everywhere. This is not the case in sPHENIX and if it is treated on average - sPHENIX will be forced either to grossly overestimate errors (hiding ignorance) or may end with huge systematics and, what will really be bad, produce tails to distributions (and our physics is largely about those tails). sPHENX is designed to generate multiple measurements for every object produced in collision so why limit yourself to calorimeters only in jet measurements. 

Edward

On Sep 13, 2017, at 12:14 PM, Sickles, Anne M <sickles AT illinois.edu> wrote:

Hi Jamie & Edward,

ATLAS does something similar to what Jamie has proposed.  Basically the overall jet energy scale is set for all jets and then global properties of the the individual jets, rather than individual showers, are used to adjust the energy scale jet by jet in a way that does not change the average energy scale.  The variation in jet response as a function of each property is calculated and used to adjust the jet energy. For example, the energy response may vary by 5% as a function of what fraction of the jet energy is deposited in a specific calorimeter layer and then each jet energy is rescaled according to how much its energy is expected to be off from the nominal response because of its energy fraction in that layer.

I think such a scheme could work in sPHENIX, with for example the fraction of energy in the EMCal or IHCal being the variables.  I think with the energy cut off on showers than Jamie proposed there might be some sensitivity leftover to the precise placement of the cut and I think something like what ATLAS does which doesn’t change the overall energy scale could avoid that.

An ATLAS description of the procedure is in here:

https://arxiv.org/abs/1112.6426 in section 11.

Best,

Anne

On Sep 13, 2017, at 10:52 AM, Edward Kistenev <kistenev AT bnl.gov> wrote:

Dear Jamie,

in  my comments to Aaron’s mail on related subject (enclosed below) I tried to formulate the energy-flow strategy to energy (and jets) reconstruction which should be used to handle differences in calorimeter response to EM and Hadronic showers gracefully. One would need to build a chain able to 

- seed jets from tracking;

- recover all reconstruction data from tracking;

- collect energies on towers closest to trajectories;

- create probability distributions for energies “expected” in towers due to reconstructed particles;

- find the realization best fitting observed energy deposition pattern (along trajectory). This is when you’ll also correct the data for out of HCal energy leakage;

- subtract it from the whole of the picture in calorimeter;

- do the neutral shower search in the leftover energy pattern in EMC with identification via IHC;

and assign what will be left in calorimeters afterwards to “unidentified hadronic object” which will need recalibration using sampling fractions corrected for LCG.

The 20% nature dependent variation in response does not affect the trigger much so the calorimeter will stay the best trigger detector available but may badly affect the systematics if all what is used in analysis are “electromagnetic” and “hadronic” calibration coefficients assigned to named detectors.

Edward

Dear Aaron, thanks for absolutely perfect comments. And unfortunately they have no simple answers even in simulation. Jet is not a confluence of identical objects and calorimeter (sPHENIX like the others) is never (except if it is all built as a bolometer) will respond in similar ways even to the same object everywhere within its mechanical structure. It should be considered first and most a trigger device, sPHENIX will have a near perfect tracking which will never be equalled by calorimeter. Most of the particles in jets will be measured with precision which can never be matched in calorimeter (software cleanup). sPHENIX also have 1.5T field with cutoff for particles reaching calorimeters comparable to average transverse momenta suppressing underlying event (hardware cleanup).  The problem is how to share the energies seen in calorimeters into components already known from tracking and that one “yet to share”. This is when simulation is critical to account for the fact that the calorimeter is deeply structured longitudinally. In sPHENIX neighbor towers overlap ( due to the tilt of absorber plate) resulting in a double sampling along the same trajectory in a single calorimeter section. Simulation will predict the set of numbers for particle (energies in “sections”) distributed according currently non specified probability densities, it will be a job for calorimeter to reduce the uncertainties by offering linked measurements at different depths. 

What’s left afterwards is the subject to a separate considerations - but at least it is a challenge. 

Note that I am not even talking about punchthroughs - they are just part of probability distribution and must be handled gracefully.

Edward

On Sep 13, 2017, at 10:55 AM, Jamie Nagle <jamie.nagle AT colorado.edu> wrote:

Hello All,

At the sPHENIX Simulations meeting yesterday (https://indico.bnl.gov/conferenceDisplay.py?confId=2710), there was a good discussion about the different energy response in the calorimeters for hadronic and electromagnetic showers.   Christof asked about jets that have a leading pizero and thus a larger electromagnetic component, which is the opposite to those jets selected for fragmentation function measurements (i.e. with a leading charged hadron and thus no leading pizero).    In the TTrees generated by Dennis, the pizero are already decayed into two photons before compiling the particle listing (particle_pid[i]).    Thus, at Dennis' suggestion, I have plotted the jet reconstructed energy for truth jets > 50 GeV as a function of the fraction of the truth particle energy in electromagnetic form (in this case in photons and electrons incident on the front of the calorimeter).    I have plotted this for the Default sPHENIX Calorimeters, though it looks similar (maybe a bit worse) for the Aluminum with and without readout.

The figure attached shows a ~20% variation in the reconstructed energy response as a function of truth electromagnetic fraction.    It aligns with Jin's comment that the default in the reconstruction code is the EMCal energy scale set for pure EM showers.    If the jet consists almost entirely of a very high z pizero, one gets a reconstructed energy of 50 GeV, i.e. very close to the truth.    As the electromagnetic fraction is smaller, the reconstructed energy total decreases.  This is part of my concern expressed today that when we investigate changes in the inner HCal and as a function of leading z hadron -  we are likely walking up and down this curve.

After I am back from the Initial Stages meeting, I would be interested to see how this energy response varies with reconstructed energy fraction in the EMCal (which is something we will measure).  Also, as a small-step start in the direction Christof mentioned for particle flow jets, I would like to implement something that finds EMCal showers that look like pure EM showers (with a probability variable cut like we use in PHENIX).   For such showers above some energy (1-2 GeV), one could use the current energy scale, and then for all the remaining energy in the EMCal use a different hadronic response energy scale.   One can see if that already will improve the resolution by removing the effect shown in the plot.    All of this should lead to some eventual documented scheme for calibrating the jet energy scale in reality.

Sincerely,

Jamie

--

||--------------------------------------------------------------------------------
|| James L. Nagle   
|| Professor of Physics, University of Colorado, Boulder
|| EMAIL:   jamie.nagle AT colorado.edu
|| SKYPE:  jamie-nagle        
|| WEB:      http://spot.colorado.edu/~naglej 
||---------------------------------------------------------------------------------

<figure.pdf>_______________________________________________
sPHENIX-HCal-l mailing list
sPHENIX-HCal-l AT lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/sphenix-hcal-l

_______________________________________________
sPHENIX-HCal-l mailing list
sPHENIX-HCal-l AT lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/sphenix-hcal-l

Anne Sickles

Assistant Professor

University of Illinois at Urbana Champaign

sickles AT illinois.edu

Anne Sickles

Assistant Professor

University of Illinois at Urbana Champaign

sickles AT illinois.edu