Bryan – herwig author (UK), did postdoc at Lab
torbjorn – pythia author
bell anderson – original pythia
Frank Krauss – Sherpa (Germany origins)
Rich – Feynman field generators
peter richardson – herwig
peter skands – pythia
stefan hosche – SLAC, MC generators
Jonathon Butterworth – Celebrity outreach in UK (check youtube)
Three sections today: Two main pieces of MC generators
1. Equation 1: This let’s us separate the problem into different steps
Momentum phase space (phi_n), parton-level cross section and the phase space
Factorization and renormalization scale:
mu_F: factorization scale: QCD is highly coupled at small energy scale. If you use all orders of perturbation, then it is independent of mu_F. In practice, it is dependent on mu_F. mu_F is scale at which you evaluate PDFs.
For example, for Z process, mu_F=m_Z and mu_R=m_Z and vary these scales to estimate uncertainty on this. Use mu, mu/2 and 2*mu to evaluate systematics.
Sometimes functional form of mu might change the scales non-trivially. For W, you use mW and if you have W+jets, then you have pT which is a variable
In general, use same value for mu_F and mu_R and then change them. It depends on which theorist you talk to!
2. Slide 11 and 12 from Jennet’s presentation:
Agreement with pythia is same as agreeing with data here
If ATLAS data didn’t exist,you should use dark plot with grey uncertainty is what Madgraph author would say
3. How do you pick mu_F and mu_R?
Generally pick something from control region in data, some processes only you have an obvious scale. You can use pT or something.
If you use same PDFs in different generators, then calculation should be similar
4. NLO:
Extra pieces: Extra radiation can be calculated separately, you can calculate loop pieces separately and combine it with LO
LO+emission-absorption of gluon (virtual) interference gives same order as Radiated diagram (real)
Radiated diagram: IR and collinear divergences and these cancel with loop level diagram
5. k factor:
You use all the time
MC generator generator up to order alpha_s^(something). You use angular distribution and differential cross-section from here but scale the cross-section from MC to the NLO one that you trust
There is no reason why k factor should be correct but if you move into corners of phase space where the shape depends on order of calculation.
To do diff cross section, you want the highest order calculation
Examples of where we get shape completely wrong at LO? Bbar production. In general any gluon splitting processes as this is NLO process
6. Large Nc limit?
Not relevant
7. When do parton showers describe jets well?
Soft radiation and collinear.
In collinear limit:
Page 23: Split angular part into two parts involving angle between radiated gluon and quark/anti-quark
MC generators generally don’t get njets in large angle region right. Splitting function is Leading Log and it works best in collinear region not at large angles! So for ee→3 jets, you shouldn’t rely on calculation from ee→2jets + parton shower, use ee→3jets Matrix Element
Another example is when both tops are hadronically decaying, you don’t trust calculation where 5th and 6th jets come from Parton Shower
8. Scale changes as you shower:
Scale decreases until you reach 1 GeV or so when you stop showering and hadronize
9. Backwards radiation:
ISR computation.
In MC, you start with partons and hard scattering ME rather than protons.You could select the partons and use PDFs and evolve them forward with the showering and hope answer is reasonable, but it is not. You mirror image forward evolution back, you start with negative scale and evolve backwards so that it becomes less negative.
You start with them after ISR which is hard scattering scale and then evolve backwards until you hit 1 Gev ~ proton mass