== MadGraph5: a version 5 of MadGraph entirely in Python ==

=== First step: a topology generator for MadGraph2 ===

==== Why ? ====
The big limitation of current MG2 is the way topologies are generated, ie in a way completely independent of the interaction structure of the model, thus rather inefficient. As a first step, we propose an external Python script helping MG2 on this. This script can evolve to a full version when we have more info on how the HELAS routines are going to be organized.

==== What ? ====
Input: particles.dat, interactions.dat and proc_card.dat

Output: A set of text files (one per subprocess) containing list of valid topologies (empty diagrams) in a way which is readable by mg2. A set of postscript files with the Feynman diagrams.

==== How ? ====
Two steps process:
   * Identify which topologies are OK with a new algorithm
   * Fill them to produce feynman diagrams (filling for matrix element is done by old mg2)

New algorithm:
   1. From interactions.dat, create a (pseudo) dictionary (Python surclassing with new routines to take care of lists in a transparent way) which, giving two lists of particles (potentially only one particle in a list) output a list of possible particle to complete a 3-vertex (with order ?). All k-vertices are divided into k-2 3-vertices and a pseudo particle.
   1. Loop over all possible subprocesses
   1. Give a tag to each particle in the subproc (1,2 initial state, 3 to n for the other ones)
   1. Order these tags according to the number of possible interactions of the associated particle (a=lowest,...) (that is the great idea!!!)
   1. Look at a and b in the dictionary and stop if empty output. If no empty output, replace a and b by this list, increment order counter (stop if counter too high), and restart diagram analysis
   1. restart previous step until the end, if it does not stop before, it's a valid topology, write it down like ((a,b),c) (intermediate particles are not identified!
   1. Once it is done, fill topologies like in MG, and draw diagrams (with FeynMF ? with feynpy ?)





   



-- Main.MichelHerquet - 14 Aug 2008