amc nlo
play

aMC@NLO Olivier Mattelaer University of Illinois at - PowerPoint PPT Presentation

Oct 27, 2022 •356 likes •943 views

UIUC aMC@NLO Olivier Mattelaer University of Illinois at Urbana-Champaign for the MadGraph/aMC@NLO team Full list of contributors: http://amcatnlo.web.cern.ch/amcatnlo/people.htm LoopFest 2013 1 UIUC Plan of the Talk aMC@NLO

  1. UIUC aMC@NLO Olivier Mattelaer University of Illinois at Urbana-Champaign for the MadGraph/aMC@NLO team Full list of contributors: http://amcatnlo.web.cern.ch/amcatnlo/people.htm LoopFest 2013 1

  2. UIUC Plan of the Talk • aMC@NLO ➡ MadLoop ➡ MadFKS ➡ NLO+PS • DEMO • MadSpin (decay of unstable particles) • Work in progress • Conclusion 2 O. Mattelaer, LoopFest 2013 aMC@NLO

  3. UIUC aMC@NLO: A Joint Venture MadGraph FKS FKS MC@NLO CutTools 3 O. Mattelaer, LoopFest 2013 aMC@NLO

  4. UIUC aMC@NLO • Why automation? ➡ Time: Less tools, means more time for physics ➡ Robust: Easier to test, to trust ➡ Easy: One framework/tool to learn 4 O. Mattelaer, LoopFest 2013 aMC@NLO

  5. UIUC aMC@NLO • Why automation? ➡ Time: Less tools, means more time for physics ➡ Robust: Easier to test, to trust ➡ Easy: One framework/tool to learn • Why NLO? ➡ Reliable prediction of the total rate ➡ Reduction of the theoretical uncertainty 4 O. Mattelaer, LoopFest 2013 aMC@NLO

  6. UIUC aMC@NLO • Why automation? ➡ Time: Less tools, means more time for physics ➡ Robust: Easier to test, to trust ➡ Easy: One framework/tool to learn • Why NLO? ➡ Reliable prediction of the total rate ➡ Reduction of the theoretical uncertainty • Why matched to the PS? ➡ Parton are not an detector observables ➡ Matching cure some fix-order ill behaved observables 4 O. Mattelaer, LoopFest 2013 aMC@NLO

  7. UIUC NLO Basics Real NLO Virtual Born Z Z Z σ NLO = d ( d ) σ V + d ( d ) σ R + d (4) σ B m +1 m m 5 O. Mattelaer, LoopFest 2013 aMC@NLO

  8. UIUC NLO Basics Real NLO Virtual Born Z Z Z σ NLO = d ( d ) σ V + d ( d ) σ R + d (4) σ B m +1 m m Need to deal with singularities Z Z Z Z σ NLO = d ( d ) ( σ V + d ( d ) ( σ R − C ) + d (4) σ B d φ 1 C ) + 1 m +1 m m 5 O. Mattelaer, LoopFest 2013 aMC@NLO

  9. UIUC NLO Basics Real NLO Virtual Born Z Z Z σ NLO = d ( d ) σ V + d ( d ) σ R + d (4) σ B m +1 m m Need to deal with singularities Z Z Z Z σ NLO = d ( d ) ( σ V + d ( d ) ( σ R − C ) + d (4) σ B d φ 1 C ) + 1 m +1 m m MadLoop MadFKS MadGraph 5 O. Mattelaer, LoopFest 2013 aMC@NLO

  10. UIUC MADLOOP The virtual 6 O. Mattelaer, LoopFest 2013 aMC@NLO

  11. UIUC The OPP Method • Reduce the Amplitudes at the Integrand level. m � 1 m � 1 ⇤ ⌅ � d i 0 i 1 i 2 i 3 + ˜ ⇥ N ( l ) = d i 0 i 1 i 2 i 3 ( l ) D i i 0 <i 1 <i 2 <i 3 i ⇥ = i 0 ,i 1 ,i 2 ,i 3 m � 1 m � 1 ⇤ ⌅ � ⇥ + c i 0 i 1 i 2 + ˜ c i 0 i 1 i 2 ( l ) D i i 0 <i 1 <i 2 i ⇥ = i 0 ,i 1 ,i 2 m � 1 ⌅ m � 1 ⇤ b i 0 i 1 + ˜ � ⇥ + b i 0 i 1 ( l ) D i i 0 <i 1 i ⇥ = i 0 ,i 1 m � 1 ⌅ m � 1 ⇤ � ⇥ + a i 0 + ˜ a i 0 ( l ) D i i 0 i ⇥ = i 0 m � 1 + ˜ ⇥ P ( l ) D i i • Feed CutTools with loop numerator and obtain the coefficients (including R 1 Term ) • Add R2 counter-terms. [Ossola, Papadopoulos, Pittau 2006] 7 O. Mattelaer, LoopFest 2013 aMC@NLO

  12. UIUC OPP in a nutshell • In OPP reduction we reduce the system at the integrand level. • We can solve the system numerically: we only need a numerical function of the (numerator of) integrand. We can set-up a system of linear equations by choosing specific values for the loop momentum l, depending on the kinematics of the event • OPP reduction is implemented in CutTools (publicly available). Given the integrand, CutTools provides all the coefficients in front of the scalar integrals and the R1 term • The OPP reduction leads to numerical unstabilities whose origins are not well under control. Require quadruple precision. • Analytic information is needed for the R2 term, but can be compute once and for all for a given model [See C. Degrande Talk] 8 O. Mattelaer, LoopFest 2013 aMC@NLO

  13. UIUC MADLOOP 3 2 g g • Diagram Generation d~ ➡ Generate diagrams d~ d~ with 2 extra particles d ➡ Need to filter result g g 2>2 • Evaluation of the Numerator: 1 4 ➡ OpenLoops techniques 9 O. Mattelaer, LoopFest 2013 aMC@NLO

  14. UIUC MADLOOP 3 2 g g • Diagram Generation d~ ➡ Generate diagrams d~ d~ with 2 extra particles d ➡ Need to filter result g g 2>2 • Evaluation of the Numerator: 1 4 ➡ OpenLoops techniques 9 O. Mattelaer, LoopFest 2013 aMC@NLO

  15. UIUC MADLOOP 3 2 g g • Diagram Generation d~ d~ ➡ Generate diagrams d~ d~ with 2 extra particles d d ➡ Need to filter result g g 2>4 1 4 9 O. Mattelaer, LoopFest 2013 aMC@NLO

  16. UIUC MADLOOP 3 2 g g • Diagram Generation d~ d~ ➡ Generate diagrams d~ d~ with 2 extra particles d d ➡ Need to filter result g g 2>4 • Evaluation of the Numerator: 1 4 9 O. Mattelaer, LoopFest 2013 aMC@NLO

  17. UIUC MADLOOP 3 2 g g • Diagram Generation d~ d~ ➡ Generate diagrams d~ d~ with 2 extra particles d d ➡ Need to filter result g g 2>4 • Evaluation of the Numerator: 1 4 ➡ OpenLoops techniques [S. Pozzorini & al.(2011)] ... W 0 1 r max µ 0 µ 1 ··· µ r l µ 0 l µ 1 · · · l µ r W 3 C ( r ) X N ( l µ ) = V 0 5 V 1 4 1 r =0 V 1 V 0 3 W 1 W 2 2 2 4 W 1 3 [See F. Cascioli Talk] 9 O. Mattelaer, LoopFest 2013 aMC@NLO

  18. UIUC MADFKS The real 10 O. Mattelaer, LoopFest 2013 aMC@NLO

  19. UIUC FKS substraction • Find parton pairs i , j that can give collinear singularities • Split the phase space into regions with one collinear singularities • Integrate them independently ➡ with an adhoc PS parameterization ➡ can be run in parallel • # of contributions ~ n^2 More details in F. Caola Talk [S. Frixione, Z Kunst, A Signer (1995)] 11 O. Mattelaer, LoopFest 2013 aMC@NLO

  20. UIUC MC@NLO Matching to the shower 12 O. Mattelaer, LoopFest 2013 aMC@NLO

  21. UIUC Sources of double counting Parton shower ... Born+Virtual: ... Real emission: • There is double counting between the real emission matrix elements and the parton shower: the extra radiation can come from the matrix elements or the parton shower • There is also an overlap between the virtual 13 O. Mattelaer, LoopFest 2013 aMC@NLO

  22. UIUC MC@NLO procedure Parton shower ... Born+Virtual: ... Real emission:  � d σ NLOwPS Z Z I ( m ) = d Φ m ( B + V + d Φ 1 MC ) MC ( O ) dO loop  � I ( m +1) + d Φ m +1 ( R − MC ) ( O ) MC • Double counting is explicitly removed by including the “shower subtraction terms” 14 O. Mattelaer, LoopFest 2013 aMC@NLO

  23. UIUC MC@NLO properties • Good features of including the subtraction counter terms 1. Double counting avoided : The rate expanded at NLO coincides with the total NLO cross section 2. Smooth matching : MC@NLO coincides (in shape) with the parton shower in the soft/collinear region, while it agrees with the NLO in the hard region 3. Stability : weights associated to different multiplicities are separately finite. The MC term has the same infrared behavior as the real emission (there is a subtlety for the soft divergence) • Not so nice feature (for the developer): 1. Parton shower dependence : the form of the MC terms depends on what the parton shower does exactly. Need special subtraction terms for each parton shower to which we want to match 15 O. Mattelaer, LoopFest 2013 aMC@NLO

  24. UIUC Four-lepton production • 4-lepton invariant mass is almost insensitive to parton shower effects. 4-lepton transverse momenta is extremely sensitive [Frederix, Frixione, Hirschi, maltoni, Pittau & Torrielli (2011)] 16 O. Mattelaer, LoopFest 2013 aMC@NLO

Recommend

More recommend