Nima's Marmoset

Here is one more splinter of Nima Arkani-Hamed's CERN visit. Apart from a disappointing seminar for theorists, Nima gave another talk advertising his MARMOSET to a mostly experimental audience. OK, I know it was more than two weeks ago, but firstly it's summertime, and secondly, i'm still doing better with the schedule than the LHC.

MARMOSET is a new tool for reconstructing the fundamental theory from the LHC data. When you ask phenomenologists their opinion about MARMOSET, officially they just burst out laughing. Off the record, you could hear something like "...little smartass trying to teach us how to analyze data..." often followed by *!%&?#/ ^@+`@¦$. I cannot judge to what extent this kind of attitude is justified. I guess, it is partly a reaction to overselling the product. To my hopelessly theoretical mind, the talk and the whole idea appeared quite interesting.

In the standard approach, the starting point to interpreting the data is a lagrangian describing the particles and interactions. From the lagrangian, all the necessary parton level amplitudes can be calculated. The result is fed to Monte Carlo simulations that convolute the amplitudes with the parton distribution functions, calculate the phase space distributions and so on. At the end of this chain you get the signal+the SM background that you can compare with the observations.

Nima pointed out several drawbacks of such an approach. The connection between the lagrangian and the predicted signal is very obscure. The lagrangians have typically a large number of free parameters, of which only a few combinations affect the physical observables. Typically, the signal, e.g. a pT distribution, has a small dependence on the precise form of the amplitude. Moreover, at the dawn of the LHC era we have little idea which underlying theory and which lagrangian will turn out relevant. This is in strong contrast with the situation that has reigned in the last 30 years, when the discovered particles (the W and Z bosons, the top quark) were expected and the underlying lagrangian was known. Nima says that this new situation requires new strategies.

Motivated by that, Nima&co came up earlier this year with a paper proposing an intermediate step between the lagrangian and the data. The new framework is called an On-Shell Effective Theory (OSET). The idea is to study physical processes using only kinematic properties of the particles involved. Instead of the lagrangian, one specifies the masses, production cross sections and decay modes of the new particles. The amplitudes are parameterized by one or two shape variables. This simple parameterization is claimed to reproduce the essential phenomenology that could equally well be obtained from more complicated and more time-consuming simulations in the standard approach.

MARMOSET is a package allowing OSET-based Monte Carlo simulations of physical processes. As the input it requires just the new particles + their production and decay modes. Based on this, it generates all possible event topologies and scans the OSET parameters, like production and decay rates, in order to fit the data. The failure implies necessity to add new particles or new decay channels. In this recursive fashion one can extract the essential features of the underlying fundamental theory.

This sounds very simple. So far, the method has been applied under greenhouse conditions to analyze the "black boxes" prepared for the LHC olympics. Can it be useful when it comes to real data? Proffesionals say that MARMOSET does not offer anything they could not, if necessary, implement within half an hour. On the other hand, it looks like a useful tool for laymen. If a clear signal is discovered at the LHC, the package can provide a quick check if your favourite theory is able to reproduce the broad features of the signal. Convince me if I'm wrong... Anyway, we'll see in two years.

The video recording available here.