sPHENIX HCal discussion

[Lee Songkyo <songkyolee AT gmail.com>]

Hello all,

I am also jumping into this thread. The attached shows the first study on the separation between quark and gluon jets with different IHCAL configurations.

Two variables, multiplicity and p_{T} D, are investigated as shown on S4. Then, the ROC curves are constructed as shown on S5. Lastly, focusing on Tower jets only, results with 5 different IHCAL scenarios are compared in S6. It seems like there is no big difference in performance between different IHCAL configurations.

Any comments and suggestions would be very welcome!

Regards,

Songkyo

Attachment: 20171004_songkyo_QvsGjet.pdf
Description: Adobe PDF document

On 4 Oct 2017, at 10:28 AM, Rosi Reed <rjr215 AT lehigh.edu> wrote:

Hi John and Raghav,

I definitely agree that it is important to set things up in such a fashion that we can answer the next charge, even if that means taking a little longer at this point, as there will be a next time and a time-after-that.  

In terms of focus, it would be good to make clear plots of what we already have (even if these are only used internally, it's nice to be able to grab something to make a point).  

I do think the jet efficiency would be important to have on the time scale of the descoping decision, as one way to see what's happening in the "unmeasured" sector, for lack of a better term.  But, I don't anticipate that this will be much different.

Thanks to Raghav for all his hard work!  

Cheers,

Rosi

On Wed, Oct 4, 2017 at 11:08 AM, John Lajoie <lajoie AT iastate.edu> wrote:

Hi Raghav, Rosi: 

    That sounds great. Before you do the unfolding I think it would be beneficial to incorporate the tower scaling into the jet finder, since most of the JES and JER difference you are seeing is coming from the variation in the JES as a function of the EM fraction of the jet. 

    Jin has suggested that an easy way to do this would be to allow the user to specify scaling constants for the jet finder inputs.  You would then specify scaling constants for the CEMC and HCAL, and with properly chosen constants that would get you most of the benefit I have been able to show in my slides. Songkyo and I could look at implementing this. 

    Doing this might have a longer timescale than the IHCAL descoping decisions, but that's OK. I think it's worth building on the momentum we have gained to put together a set of tools that will allow a more thorough analysis of jet performance.  The questions we are called on to answer are only going to get more detailed as time passes. 

    Let me know what you think. 

Regards,
John

On 10/4/2017 9:52 AM, Raghav Kunnawalkam Elayavalli wrote:

Hi John and Rosi, 

My jobs are still running from yesterday so as soon as they are done I will make the comparison plots and yeah, I would also hope that there is not that large of a different with the new geometry. I can also updated the color on the new plots and plot the ratios of JES and JER with the regular steel HCALIN. 

Once my new jobs finish i will produce the jet reconstruction efficiency and do the toy unfolding plots which would show the effect of the worsened resolution. 

Cheers

Raghav 

**************************************

Raghav Kunnawalkam Elayavalli

PhD, Rutgers University

SRN-206W 

136 Frelinghyusen Rd.

Piscataway NJ 

office: +1 848 445 8769

cell : +1 732 532 9232  

**************************************

On Oct 3, 2017, at 4:26 PM, John Lajoie <lajoie AT iastate.edu> wrote:

Hi Raghav, 

    Very nice study!  It will be good to see how this changes with the latest HCAL geometry updates; I don't expect much of a difference. 

    I thought I would hijack your thread to send out some revised results that I have been working on. These have been run with the latest HCAL geometry updates, and in addition I applied a procedure to correct the HCAL and CEMC energy scales at the calorimeter tower level.  With these corrections, the JES is flat as a function of the jet electromagnetic fraction, and the differences between the various IHCAL scenarios are much more apparent since they are not washed out by variations in the JES. 

Regards,
John

On 10/3/2017 2:20 PM, Raghav Kunnawalkam Elayavalli wrote:

Hi All, 

I attached a set of slides to this email that looks at the different HCALIN descoping scenarios and their impact on the jet energy scale and resolution for a variety of jet radii (R = 0.2, 0.3 and 0.4). I included the case of having a 5in steel frame as well. 

Building of my previous set of plots, I looked at the impact of the resolution for quark vs gluons jets in a pT range 15-60 GeV. We see that the degradation in the resolution is quite similar to the inclusive sample, with quark-jets similar to inclusive at high pT and gluon-jets similar to inclusive at low pT. 

I also added some rudimentary detector level corrections on the jet pT (by just fitting the response as shown in slide 5) and the effect on the resolution is quite minimal for both inclusive and for quark vs gluon jets. 

Im also hoping to have some additional plots that study the fragmentation (and the high z part; statistics pertaining) and the jet shapes. 

Note: These plots are not with the latest hcal updates from Chris and Im currently running some jobs now that will compare the two.  

 

Cheers

Raghav 

**************************************

Raghav Kunnawalkam Elayavalli

PhD, Rutgers University

SRN-206W 

136 Frelinghyusen Rd.

Piscataway NJ 

office: +1 848 445 8769

cell : +1 732 532 9232  

**************************************

On Oct 2, 2017, at 12:33 PM, John Lajoie <lajoie AT iastate.edu> wrote:

Hi Dennis, 

    I'd like to have some additional statistics of your ">50 GeV, R=0.4 jet in |eta|<0.6" jet sample to have a look at high-z fragmentation.  Would it be possible for you to generate some additional hepmc files?  A factor of 2-4 would be good, 10 would be great.   I don't need the G4 simulations as I will run them through my analysis code, just the hepmc files. 

John

On 9/11/2017 6:07 PM, Perepelitsa, Dennis wrote:

Dear sPHENIX Jet Structure and HCal teams (and in particular John L., Aaron, Songkyo, Raghav),

As we agreed at the HCal meeting last week, we have begun coordinating the production of simulation jobs to investigate the sensitivity of the jet response (especially as a function of fragmentation pattern) to different Inner HCal scenarios.

I have run a set of 10k Pythia QCD hard scattering events with the requirement of a >50 GeV, R=0.4 jet in |eta|<0.6. A full G4 simulation and then standard (pp) jet reconstruction were performed under three scenarios: (1) the default IHCal description, (2) an Al-material IHCal, and (3) an Al-material IHCal which is not read out (this is implemented as tower reco jets not being given the HCALIN_TOWER container as an input to jet reconstruction in G4_Jets.C ).

The input Pythia8 HepMC files are here: /sphenix/user/dvp/gen/QCD35/

My G4 setup is here: /phenix/upgrades/decadal/dvp/production-IHCal/macros/macros/g4simulations/ 

At this point, I have not kept the resulting G4 hits files, but have rather “tree-ified” the results using a simple TreeMaker class you can find here: /direct/phenix+upgrades/decadal/dvp/production-IHCal/src/ . The trees store the (pt/eta/phi) for truth and tower-reco jets for R=0.2,0.3,0.4,0.5 cone sizes, as well as the pt/eta/phi/particle-ID of all final-state truth particles. A few comments here:

1. R=0.5 is a somewhat larger cone size than we normally look at, but I included it for completeness so analyzers can study the jet R-dependence of these different scenarios.
2. The final-state truth particles are included so that analyzers can also study the fragmentation pattern (for example, max-z) dependence. 
3. I think we have been wanting a “standard” tree-maker AKA ntuple-dumper like this for a while — we should flesh this out and formally add it to git soon.

Anyway, I have compiled the resulting output trees, which should have 10k events each, here: 

/sphenix/user/dvp/sims/sPHENIX-Sep17-QCD35-10k-DefaultIHCal.root
/sphenix/user/dvp/sims/sPHENIX-Sep17-QCD35-10k-AlIHCal.root           
/sphenix/user/dvp/sims/sPHENIX-Sep17-QCD35-10k-AlIHCalNoReadout.root

I would strongly and warmly encourage Aaron, Songkyo, Raghav and any other interested parties to please take a look at the setup and the resulting output trees. (I know that the Colorado group is planning to have a look at these outputs and try to produce some initial plots for the simulations meeting tomorrow, but it is highly preferable that this is not the only group doing so.)

Actually, I know that Raghav has taken the lead and run some initial simulations of his own — I attach some slides from him here which briefly summarize those studies. In the near future, it would be great to check that we are getting consistent answers in the region where our simulations overlap.

Depending on how the first studies turn out, we are happy to coordinate an additional production (including “tree-ification”) with HIJING embedding and a more extensive pT range, also in the near future.

All are welcome and encouraged to take a look,

Dennis

Dennis V. Perepelitsa
Assistant Professor, Physics Department
University of Colorado Boulder

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

-- 

John Lajoie

Professor of Physics

Iowa State University

 

(515) 294-6952

lajoie AT iastate.edu

 
Contact me:  john.lajoie

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

-- 

John Lajoie

Professor of Physics

Iowa State University

 

(515) 294-6952

lajoie AT iastate.edu

 
Contact me:  john.lajoie

<Tower Energy Scale Corrections and JES.pdf>

-- 

John Lajoie

Professor of Physics

Iowa State University

 

(515) 294-6952

lajoie AT iastate.edu

 
Contact me:  john.lajoie

-- 

Rosi Reed

Assistant Professor, Physics Department

Lehigh University
(610)758-3907

16 Memorial Drive East Office 406

Bethlehem, PA 18015

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