mg5 amc nlo
play

MG5_aMC@NLO looping up to be mad! Olivier Mattelaer IPPP/Durham - PowerPoint PPT Presentation

MG5_aMC@NLO looping up to be mad! Olivier Mattelaer IPPP/Durham work in progress with V. Hirschi Type of generation NLO NLO NLO NLO Loop Loop Tree Tree (QCD) (QCD) (EW) (EW) Induced Induced (SM)


  1. MG5_aMC@NLO looping up to be mad! Olivier Mattelaer IPPP/Durham � work in progress with V. Hirschi

  2. Type of generation NLO � NLO � NLO � NLO � Loop � Loop Tree Tree (QCD) (QCD) � (EW) � (EW) � Induced Induced (SM) (BSM) (SM) (BSM) (SM) (BSM) (SM) (BSM) Fix Order +Parton Shower Merged � Sample Mattelaer Olivier Looping up to be MAD 2

  3. Type of generation NLO � NLO � NLO � NLO � Loop � Loop Tree Tree (QCD) (QCD) � (EW) � (EW) � Induced Induced (SM) (BSM) (SM) (BSM) (SM) (BSM) (SM) (BSM) Fix Order +Parton Shower Merged � Sample Mattelaer Olivier Looping up to be MAD 2

  4. Loop Induced Why? • Main production mechanism for Higgs & Higgs associated processes � • Contribution for NNLO computation � • Correction to shape of observables � • We have the tool available Mattelaer Olivier Looping up to be MAD 3

  5. Loop Induced Why? Difficulties? • Main production • The phase-space mechanism for Higgs & integration is based Higgs associated on the born diagram � processes � • Loop evaluation are • Contribution for NNLO extremely slow � computation � • Need Leading Color • Correction to shape information for of observables � writing Events associated to the • We have the tool loop � available � Mattelaer Olivier Looping up to be MAD 3

  6. Loop Induced Difficulties? OLD Solution • Use Effective Field • The phase-space Theory (=> Tree) � integration is based • And correct the mass on the born diagram � effect � • Loop evaluation are W new = | M new | 2 | M old | 2 ∗ W old extremely slow � � • Need Leading Color � information for • Difficult control on � writing Events numerical uncertainty � associated to the • Wrong Leading Color loop � information/helicity � � • Not generic � � Mattelaer Olivier Looping up to be MAD 4

  7. Loop Induced Difficulties? OLD Solution • Use Effective Field • The phase-space Theory (=> Tree) � integration is based • And correct the mass on the born diagram � effect � • Loop evaluation are W new = | M new | 2 Not the method of choice for BSM � | M old | 2 ∗ W old extremely slow � � Not the method that we will choose • Need Leading Color � information for • Difficult control on � writing Events numerical uncertainty � associated to the • Wrong Leading Color loop � information/helicity � � • Not generic � � Mattelaer Olivier Looping up to be MAD 4

  8. Exact Integration Difficulties? New Solution • Contract the loop to • The phase-space have tree-level integration is based diagrams which drive on the born diagram � the integration multi- channel � • Loop evaluation are extremely slow � • Use Monte-Carlo over helicity � • Need Leading Color information for • Compute the loop with writing Events the color flow algebra � associated to the • more parallel code � loop � � � Mattelaer Olivier Looping up to be MAD 5

  9. parallelization MadEvent | M 1 | 2 | M 2 | 2 | M | 2 = | M 1 | 2 + | M 2 | 2 | M | 2 + | M 1 | 2 + | M 2 | 2 | M | 2 Mattelaer Olivier Looping up to be MAD 6

  10. parallelization MadEvent | M 1 | 2 | M 2 | 2 Z Z Z | M | 2 = | M 1 | 2 + | M 2 | 2 | M | 2 + | M 1 | 2 + | M 2 | 2 | M | 2 Mattelaer Olivier Looping up to be MAD 6

  11. parallelization MadEvent | M 1 | 2 | M 2 | 2 Z Z Z | M | 2 = | M 1 | 2 + | M 2 | 2 | M | 2 + | M 1 | 2 + | M 2 | 2 | M | 2 • Iteration 1 
 • Iteration 1 
 � � • Grid Refinement 
 • Grid Refinement 
 � � • Iteration 2 
 • Iteration 2 
 � � • Grid Refinement � • Grid Refinement � � � � � Mattelaer Olivier Looping up to be MAD 6

  12. parallelization MadEvent | M 1 | 2 | M 2 | 2 Z Z Z | M | 2 = | M 1 | 2 + | M 2 | 2 | M | 2 + | M 1 | 2 + | M 2 | 2 | M | 2 • Iteration 1 
 • Iteration 1 
 � � • Grid Refinement 
 • Grid Refinement 
 � � • Iteration 2 
 • Iteration 2 
 � � • Grid Refinement � • Grid Refinement � � � � � Mattelaer Olivier Looping up to be MAD 6

  13. parallelization New MadEvent | M 1 | 2 | M 2 | 2 Z Z Z | M | 2 = | M 1 | 2 + | M 2 | 2 | M | 2 + | M 1 | 2 + | M 2 | 2 | M | 2 • Iteration 1 
 • Iteration 1 
 • Grid Refinement 
 • Grid Refinement 
 • Iteration 2 
 • Iteration 2 
 • Grid Refinement � • Grid Refinement � � � � � Mattelaer Olivier Looping up to be MAD 7

  14. First Example: g g> h User Input • generate g g > h [QCD] � • output � • launch Loop Induced 2 2 g g b t σ loop = 15 . 74(2) pb h h 3 b 3 t b~ t~ g g 1 1 2 HEFT g h 3 g 1 σ heft = 17 . 63(2) pb Mattelaer Olivier Looping up to be MAD 8

  15. First Example: g g> h User Input • generate g g > h [QCD] � • output � • launch Loop Induced 2 2 g g b t σ loop = 15 . 74(2) pb h h 3 b 3 t b~ t~ g g 1 1 2 HEFT No bottom loop 2 g g t h h 3 3 t t~ g g 1 1 σ heft = 17 . 63(2) pb σ toploop = 17 . 65(2) pb Mattelaer Olivier Looping up to be MAD 8

  16. Validation p p > h j Mattelaer Olivier Looping up to be MAD 9

  17. Validation p p > h j Mattelaer Olivier Looping up to be MAD 9

  18. Validation p p > h j • b effect only important at low pt � • at large pt, this is just a re-scaling Mattelaer Olivier Looping up to be MAD 10

  19. Matching/Merging KT MLM Q match = 50 GeV Mattelaer Olivier Looping up to be MAD 11

  20. BSM Example: 2HDM BSM technicalities • Our code is fully ready for (all) BSM � • We (only) need NLO-UFO model � ➡ Except if you provide the loop matrix-element. Mattelaer Olivier Looping up to be MAD 12

  21. BSM Example: 2HDM BSM technicalities • Our code is fully ready for (all) BSM � • We (only) need NLO-UFO model � ➡ Except if you provide the loop matrix-element. Benchmark Point Mattelaer Olivier Looping up to be MAD 12

  22. Z+Scalar Processes Exact Phase-Space integration Reweighting (1503.01656) gg → Zh 0 gg → ZH 0 gg → ZA 0 113 +30% 686 +30% 0.622 +32% B1 − 21% − 22% − 23% 85.8 +30 . 1% 1544 +30% 0.869 +34% B2 − 21% − 22% − 23% 167 +31% 0.891 +33% 1325 +28% B3 − 19% − 21% − 21% Mattelaer Olivier Looping up to be MAD 13

  23. Z+Scalar Processes Exact Phase-Space integration Reweighting (1503.01656) gg → Zh 0 gg → ZH 0 gg → ZA 0 113 +30% 686 +30% 0.622 +32% B1 − 21% − 22% − 23% 85.8 +30 . 1% 1544 +30% 0.869 +34% B2 − 21% − 22% − 23% 167 +31% 0.891 +33% 1325 +28% B3 − 19% − 21% − 21% Mattelaer Olivier Looping up to be MAD 13

  24. Charged Higgs Exact Phase-Space integration Mattelaer Olivier Looping up to be MAD 14

  25. Charged Higgs Mattelaer Olivier Looping up to be MAD 15

  26. Type of generation NLO � NLO � NLO � NLO � Loop � Loop Tree Tree (QCD) (QCD) � (EW) � (EW) � Induced Induced (SM) (BSM) (SM) (BSM) (SM) (BSM) (SM) (BSM) Fix Order +Parton Shower Merged � Sample Mattelaer Olivier Looping up to be MAD 16

  27. Type of generation NLO � NLO � NLO � NLO � Loop � Loop Tree Tree (QCD) (QCD) � (EW) � (EW) � Induced Induced (SM) (BSM) (SM) (BSM) (SM) (BSM) (SM) (BSM) Fix Order +Parton Shower Merged � Sample • 2 to 2 processes: OK on a laptop � • 2 to 3 processes: OK on a small size cluster � • 2 to 4 processes: Specific case Mattelaer Olivier Looping up to be MAD 16

Recommend


More recommend