## Wednesday, June 06, 2018

### Monte Carlo fails at leptonic top pair production

The LHC collaborations have produced hundreds of papers that mostly say "everything agrees with the Standard Model, nothing new to be seen here".

Well, a new CMS preprint
Measurements of differential cross sections for $$t \bar t$$ production in proton-proton collisions at $$\sqrt{s} = 13\TeV$$ using events containing two leptons
says something completely different.

Go to the page 58/64 and you will see something like:
Measurements were made of the cross section for the appearance of the top quark and its antiparticle in the proton-proton collisions. Some overall cross sections could agree with the theory. However, the next-to-leading Monte Carlo simulations in any combinations suck at predicting the jet multiplicity. They suck at predicting the transverse momentum of the leptons, quarks, invariant masses, and all other kinematic observables you could think of. Higher-order models seem to suck, too.
Nice.

What should you think about it? I am not sure. This is a surprising disagreement, I think, because the top quarks should be relatively easy. They're very heavy so they should be created like new elementary particles that don't care much about the weak glue that attaches them to other quarks and gluons. The theoretical simulations shouldn't be this bad.

On the other hand, the very fact that the Monte Carlo simulators have to be used indicates that the QCD phenomena must matter. This is not an experiment whose results could be easily predicted just by considering the elementary particles such as top quarks – the gluons and their interactions matter. So I would say that this is a failure of the theory in a regime where both the simple Feynman diagrams describing high-mass top quarks; and messy calculations where Monte Carlo calculations are useful have to be combined.

When the combination is needed, theoretical predictions seem to be off.

MG5 aMC@NLO, MG5 aMC@NLO, PYTHIA8, MADSPIN and other calculations have to be made in Monte Carlo because the casino produces the random numbers that are needed in the calculations. Donald Trump is planning the American answer to the Monte Carlo simulations in several casinos he owns. The first generator will be MLVGA1 (Make Las Vegas Great Again One).

Of course, this could be a hint of new physics. Perhaps the top quark is composite. Or something else. But this is a complicated enough experiment – lots of processes are combined to produce the final results – that it may be very hard to reverse engineer the discrepancy. In other words, the discrepancy is no straightforward guide towards the right way to fix our theory if our theory is incomplete at all.