RS part of NLO calculation is spoiled if done with Comix or External and using some selectors
Hi all, I've noticed that when one uses Comix or External as RS_ME_Generator with, for example, a Variable Selector the result is clearly wrong. Here the tests I've made to let you understand what I'm talking about.
Run cards: Sherpa_amegic.yaml, Sherpa_comix.yaml, Sherpa_ext.yaml and the generator for the fsrchannels Sherpa_opt.yaml to have a faster integration with External.
The selectors used are:
SELECTORS:
- [Mass, 11, -11, 10., E_CMS]
- [Mass, 13, -13, 10., E_CMS]
- VariableSelector:
Variable: m
Flavs: [11, -11, 13, -13]
Ranges: [60.,360.]
Fixed_Order's total XS:
- Amegic:
BVI: 0.0820188 pb +- ( 0.000593193 pb = 0.723241 % )
RS: 0.00276642 pb +- ( 8.52978e-05 pb = 3.08333 % ) 150000 ( 300752 -> 54.3 % )
- Comix:
RS: 0.222791 pb +- ( 0.0119382 pb = 5.35848 % ) 905000 ( 1676994 -> 55.2 % )
- External:
BVI: 0.0828534 pb +- ( 0.000818447 pb = 0.987825 % )
RS: 0.0860101 pb +- ( 0.005062 pb = 5.88536 % ) 3010000 ( 5899236 -> 50.9 % )
As you can already see the only reasonable RS contribution is given by Amegic, the others are just too big. The Born is not reported but it's close to the BVI result. The run with Comix has no BVI because it throws this error:
Fatal error thrown in OpenLoops_Virtual::Calc:
Unknown coupling.
It seems that it doesn't define the functions AlphaQED and AlphaQCD but this is not the point here.
MC@NLO's total XS:
- Amegic:
BVI: 0.0811681 pb +- ( 0.000509591 pb = 0.627821 % )
RS: 0.00232608 pb +- ( 0.000180375 pb <-> 0.0803565 pb ) 150000 ( 310824 -> 62.7 % )
- Comix:
RS: 0.147752 pb +- ( 0.00695058 pb <-> 0.000814439 pb ) 1000000 ( 1678309 -> 62.4 % )
- External:
BVI: 0.0830993 pb +- ( 0.000592904 pb = 0.713489 % )
RS: 0.0557233 pb +- ( 0.00515385 pb <-> 0.000830993 pb ) 1725000 ( 3051768 -> 57 % )
Again a similar situation...
I then tried removing VariableSelector and now they all more or less agree:
Fixed_Order's total XS:
- Amegic:
BVI: 0.0952725 pb +- ( 0.000708745 pb = 0.743914 % ) 150000 ( 170318 -> 96.1 % )
RS: 0.00323475 pb +- ( 0.000233064 pb <-> 0.0943198 pb ) 150000 ( 265287 -> 61.7 % )
- Comix:
RS: 0.00328103 pb +- ( 0.000131699 pb = 4.01396 % ) 215000 ( 341021 -> 70.6 % )
- External:
BVI: 0.0937142 pb +- ( 0.000637275 pb = 0.680019 % )
RS: 0.00312722 pb +- ( 0.000111327 pb <-> 0.000937142 pb ) 150000 ( 235557 -> 64.8 % )
I also tried a last check. I used VariableSelector but I put cut values such that it is like it is disabled, that means:
- VariableSelector:
Variable: m
Flavs: [11, -11, 13, -13]
Ranges: [0.,10000.]
The result is now somewhat weird.
Fixed_Order's total XS:
- Amegic:
BVI: 0.0957957 pb +- ( 0.000670655 pb = 0.700088 % )
RS: 0.00258519 pb +- ( 0.000272319 pb <-> 0.0937273 pb ) 150000 ( 256513 -> 67.5 % )
- Comix:
RS: 0.00354693 pb +- ( 0.00011914 pb = 3.35894 % ) 300000 ( 478094 -> 67.5 % )
- External:
BVI: 0.0937238 pb +- ( 0.000621498 pb = 0.663116 % )
RS: 0.00457488 pb +- ( 0.00073498 pb <-> 0.000937238 pb ) 150000 ( 236269 -> 64.3 % )
In the past months I had seen a similar behavior also when using jet selectors in a NLO calculation.