Long Lived Particles Break HEP May 13, 2009Posted by gordonwatts in Conference, physics.
In my last post I mentioned that long lived particles break some basic assumptions that we make in the way we design our software and hardware in HEP. One fascinating example of this that was brought into clear relief for me at this workshop is the interaction between Monte Carlo generation and detector simulation. Look again at the picture I had up last time:
While what I’ve shown above is real data, lets imagine it was a simulation for sake of discussion. Simulation is crucial – it allows us to compare what we think we know against nature. You might imagine that the code that generates everything that happens at the very center of that picture on the left is different from the code that propagates the particles out through the detector (the green, yellow, and blue lines). In fact, this is exactly how we structure our code in HEP – as a two step process.
The first step is to generate the actual physics interaction. Say a top quark production, or Higgs production and decay, or Hidden Valley decay. As output the generator produces a list of particles and what directions they are heading. Most of them will then stream through the detector leaving tracks and data similar to the right side of that above picture. At this point we’ve got the starting point for all those “lines” or particle trajectories on the left.
Then the detector simulator program takes over. Its job is to simulate the detector. It takes each one of the particles and steps it, a millimeter at a time, through the detector. As it moves through the detector it decides if it should loose some energy interacting with the material, or leave a signal in a detector, etc. Once the simulation is done what we have is something that looks like the experiment was actually run – we can feed it through the same software that we use for real data to find electrons, tracks, etc.
But some of these long lived particle models have particles that interact as they move through the detector. The Quirk model is the poster-boy for this (odd, a model without a web page! At least that I could find). As pairs of these move through the detector they interact with each other and with the material they are traveling through. In short – the detector simulation has to act a bit like the generator – we are mixing these two things.
The main detector simulation program (GEANT4) – written in C++ and carefully planned out – does not look anything like an event generator – written in FORTRAN (common blocks!? ‘nuff said – wait, that was flame bait, wasn’t it?). My guess is it will take a year or so to get GEANT4 updated to accommodate models like Quirks. While it isn’t a complete rewrite of the package – it was quite generally designed – the GEANT4 folks probably didn’t think of a modification to allow interactions like this as a possibility.
Which makes me wonder if in the future generators will really just be subroutines (methods, sub-classed objects, etc.) in detector simulations?🙂 We all know that detectors are the most important things out there, after all!