Some questions concerning integration channels

Hi,

> What does the second number after the dot mean?

This correspond to different choice of particle assumed to be on shell/off-shell (below/above the threshold) for the phase-space integration.
This is a ternary number.

> 2) For each subprocess, the results of the integration channels add up to the total result for the subprocess. Aren't there any interference terms between the graphs? Are they somehow included in the results of the channels?

The interference is indeed include in the computation.
You can see the definition of the computation in the following lecture:
https://cp3.irmp.ucl.ac.be/projects/madgraph/attachment/wiki/Shanghai2015/MG5_Shangai.pdf

> 3) Some of the diagrams MadGraph finds seem to be not included in any integration channel (they do not appear in the config_subproc_map.inc). For example, diagrams that include vertices with four lines are in most cases discarded. How does MadGraph decide which diagrams to include?

All four point interaction (or more) are discarded since we expect those phase-space to be cover by the three point interaction.
For the three points, we keep only the first diagram of a given topology

> 4) Is it possible to manually switch off some of the integration channels (and even some of the subprocesses), for example those that contribute less than the uncertainty of the combined result? I can imagine that this could speed up the calculation/event generation.

Depends of what you mean switch off.
1) Removing completely a channel of integration.
2) Assuming the channel to return 0 (or very small number with large but insignificant error)

The first method will modify all the other channel of integration (since the result is independent of the number of channel of integration) and can lead to a slower code in some case.

The second method is actually done automatically by the code, so they are no need to do that manually.

Note that we have two way to perform the integration:
1) The standard mode of integration

a) the standard one runs first a quick evaluation of each channel (we call this the survey)
b) then we refine the channel that are usefull (i.e. where the number of event generated during the survey are not enough for the number of requested event).
During the second step the precision of the cross-section will automatically increase (it is quite poor after the survey)

2)The gridpack mode (intended to be used for generating large number of events with the same benchmark/cut)
a) we run the same survey but this time with quite high precision to have a very reliable value of the cross-section for each channel.
b) we create a new code where the result of the survey is the starting point. All events generated during the survey are discarded.

c) the new code can start directly the refine step. For channel with low number of events needed (those where in the previous run the survey was enough to cover those in case). we launch a random number to see if the channel is going to produce at least one event. If it is not the case, then the channel is discarded. Otherwise it is run normally.

More details on the gridpack mode:
https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/GridDevelopment

Cheers,

Olivier

> On Aug 4, 2016, at 09:27, Eric <email address hidden> wrote:
>
> New question #321938 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/321938
>
> Dear MadGraph Team,
>
> I have some questions concerning the contributions of the individual integration channels listed in the html results pages.
>
> 1) In the list including the contributions of the channels/graphs to the subprocesses, some graphs are denoted just by their number (e.g. G9) while others have a second number separated by a dot (e.g. G12.02 or G5.25). What does the second number after the dot mean?
>
> 2) For each subprocess, the results of the integration channels add up to the total result for the subprocess. Aren't there any interference terms between the graphs? Are they somehow included in the results of the channels?
>
> 3) Some of the diagrams MadGraph finds seem to be not included in any integration channel (they do not appear in the config_subproc_map.inc). For example, diagrams that include vertices with four lines are in most cases discarded. How does MadGraph decide which diagrams to include?
>
> 4) Is it possible to manually switch off some of the integration channels (and even some of the subprocesses), for example those that contribute less than the uncertainty of the combined result? I can imagine that this could speed up the calculation/event generation.
>
> Cheers,
>
> Eric
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.