Minimal Jet Radius?

Dear Andrew,

I'm a bit confused on what you are trying to do. Are you generating MC@NLO events, and looking at the cross section? Without showering? That would be inconsistent.

Having said that, there are some small cut-offs in the MC@NLO code. For example, around lines 511-523 of SubProcesses/cuts.f there is the removal of events when two partons are very soft. Also, when very close to soft/collinear limits, the real-emission matrix elements are approximated by their values in the limits. But maybe most importantly, is that the phase-space integration as well as the "virt-tricks" (see section 2.4.3 of arXiv:1405.0301) are not very well optimised for these configurations. Hence, I'm not surprised that this does not work out-of-the-box.

best,
Rikkert

Hi Rikkert,

Thanks for the information! This is helpful.

However, let me say what exactly I am doing because I don't understand how this is "inconsistent". I generate pp > Zj [ all = QCD ] events. The "fixed-order" flag and "shower" flags are turned off. I demand that the Z boson has pT > 500 GeV and cluster jets with radius R. MadGraph generates then 10k weighted events, which I analyze with my own code.

Specifically, as long as I measure IRC safe observables on these weighted NLO events, I should produce the differential cross section at NLO for the observable that I study (at least down to some IR cutoff). Is this correct? I have long used MadGraph in this way and if this is inconsistent, then I am very confused as to what the weighted events represent in this framework. I had long assumed that the events that MadGraph produces at NLO are like those from, say EVENT2. Is my understanding correct?

Thanks!
Andrew

Hi Andrew,

I'm afraid your understanding is not correct.

You should have gotten the following warning when you run the code in this way:

"WARNING: You have chosen not to run a parton shower. NLO events without showering are NOT physical. Please, shower the LesHouches events before using them for physics analyses. You have to choose NOW which parton-shower you WILL use and specify it in the run_card. " (in blue!)

The events that are generated are unweighted (up to a sign) and they include the shower subtraction terms, which, without including the appropriate shower, make these events non-NLO correct. This is very different from EVENT2: there are no shower subtraction terms there (and therefore those events can also not be unweighted).

It is possible to generate weighted events (just like the ones in EVENT2), using fixed-order=ON and setting the FO_ANALYSIS_FORMAT to LHE in the FO_analyse_card.dat. Furthermore, to keep track of the separate flavours at fixed order, you'll have to set the separate_flavour_configurations (or something like that) to true in the FKS_params.dat file (can be found in the Cards directory). Note that these events will be weighted. Which means you can easily get 10s of GBs of data in your final event file, even with relatively moderate "required accuracy". Therefore, the use of this is typically not recommended, and it is better to perform your analysis on the fly with a dedicated analysis file directly linked to the MadGraph5_aMC@NLO code.

Best,
Rikkert

Hi Rikkert,

Ahh, okay, thank you! This clarifies a lot.

So, I think what I want to do may just not be possible in MadGraph, then. I want to specifically calculate the NLO cross section for flavor kT jets, say. In question #695046, Marco stated that parton flavor is not passed to the internal FastJet implementation. Is this still true?

Best,
Andrew

Hello Andrew,

That is still true at the level of the generation cuts. Of course, at the level of analysis cuts, you can do whatever you want for the cuts -- as long as they are tighter than the generation cuts.

Best,
Rikkert