Total inclusive jet cross section

[Generic User (don't modify these fields)]

Dear Sherpa Team,

I run the Sherpa for inclusive jet production @7TeV using the following Run.dat

(fragmentation){

  FRAGMENTATION = Ahadic
  DECAYMODEL    = Hadrons

}(fragmentation)

(me){
  ME_SIGNAL_GENERATOR   = Comix
  EVENT_GENERATION_MODE = Weighted
}(me)

(beam){
  BEAM_1          = 2212
  BEAM_ENERGY_1   = 3500.
  BEAM_2          = 2212
  BEAM_ENERGY_2   = 3500.
}(beam)
(mi){
  MI_HANDLER    = Amisic
  SCALE_MIN     = 2.0745
}(mi)

(processes){
Process 93 93 -> 93 93 93{2};
   Order_EW 0; CKKW sqr(20/E_CMS);
   Max_N_Quarks 4;

   Integration_Error 0.02 {2};
   Enhance_Function pow(PPerp2(p[2]),2) {2};

   Integration_Error 0.03 {3};
   Enhance_Function pow(max(PPerp2(p[2]),PPerp2(p[3]),PPerp2(p[4])),2) {3};
   Enhance_Factor 4 {3};

   Integration_Error 0.04 {4};
   Enhance_Function pow(max(PPerp2(p[2]),PPerp2(p[3]),PPerp2(p[4]),PPerp2(p[5])),2) {4};
   Enhance_Factor 16 {4};

End process;
}(processes);

(selector){
  NJetFinder 2 25.0 0 1
}(selector)

I get slightly different results for total cross section, if I run it with MI_HANDLER=Amisic and MI_HANDLER=None options. The difference is ~2.2%, but the accuracy of calculation is ~0.7%. So, it look like a systematic effect, rather than statistical fluctuation.

So, my question is whether such a difference is expected or there is a bug in my steering?

I have attached the logs from the two runs.

Thank you in advance, Pavel

[Generic User (don't modify these fields)]

log files

[Frank Siegert]

I'm trying whether I can reproduce this with Sherpa trunk over night.

[Frank Siegert]

I can reproduce it indeed with the following additional findings:

• Hadronisation doesn't play a role, the UE alone makes the difference
• This happens only with the 2->4 included, not if I use only 93 93 -> 93 93 93{1}
• I've tried with a pure 2->4 run (without merging) and don't see any problem, but here the fluctuations are also bigger after 1M weighted events:

  2.32744e+06 pb +- ( 99792.7 pb = 4.28 % )    (no UE) vs.
  2.48211e+06 pb +- ( 124733 pb = 5.02 % )     (with UE)
  

• The integrated cross sections are not affected, and even if I use the same Results directory for both runs the generated cross section shows the strange behaviour.

So it seems to come from the jet veto in the shower, but I don't quite understand yet why the MPI should have an influence on that. The effect is very small, so it's quite hard to debug.

[Frank Siegert]

We think this is understood now: When MPI are on, the beam remnant phase happens to request event retries in a few percent of the events (due to colour). At this stage the shower had already passed a jet veto (which would reduce the cross section) but now it has the possibility to fail the jet veto once more in the same event. In this way a small but noticable reduction of the cross section is introduced.

The solution is to not re-generate the shower evolution in such cases but keep it and only retry the MPI and beam remnant phases. Will try to implement that tomorrow such that it makes it into 1.2.2.

Thanks for reporting the issue.

[Frank Siegert]

Implemented on trunk. This will be part of Sherpa 1.2.2 which is in validation right now and will be released tomorrow if all goes well.

[Stefan Hoeche]

Milestone old deleted