bwcutoff in WWjj processes

Dear Jasper,

I think that you approach the problem by the wrong direction.

The decay-chain syntax is suppose to be used to select a sub-sample of Feynman diagram which does not interfere (or in a negligible way) with the other set of Feynman Diagram. The typical reason why the interference is negligible is because you have the breit-wigner enhancement. Obviously this enhancement is only present if you are (close to be) on shell. This is the reason of the bwcutoff. Using large/very large bwcutoff shows indicates that you are selecting diagrams beyond the validity range of asking the W to be on shell.

Using bwcutoff=1500 is therefore intrinsically inconsistent. Since you ask the “W” to be on shell and at the same time allow them to be very far from their on shell mass.

If you care about the tail, then those interference are important and you need to include them:
generate p p > l+ vl l+ vl j j QCD=0
If you know for sure that those interferences are negligible on the FULL tail, then yes you can use the decay-chain syntax with very large bwcutoff.
But using such large value might be problematic for the Phase-Space integration.
So it might be better to not ask the W+ to be on shell but just in s-channel with the following syntax:
generate p p > w+ > l+ vl l+ vl j j QCD=0
The current syntax do not allow to ask two W on s-channel so far.

The exact same comment apply to the second processes.
> We found that using bwcutoff = 35 to include the resonant Higgs contribution gives a good agreement with other generators.

This is already too large (I guess). Indeed putting this to 35 allow to have the H > WW* decay. But this is not two on shell decay.
If you want to include it correctly, you also need to have
generate p p > l+ vl l- vl~ j j QCD=0
or
generate p p > l+ vl l- vl~ j j $ t t~ QCD=0

Note that bwcutoff also apply to the “$” syntax so setting it very large kill effectively the top interference on a large range.
To have this consistent you have to use the same bwcutoff for this sample than the one that you use for your tt~ sample.
Such that you do not have any double counting/hole in your generation.

Cheers,

Olivier

> On Jan 18, 2016, at 11:02, Jasper Lauwers <email address hidden> wrote:
>
> New question #281121 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/281121
>
> Dear authors,
>
> We are trying to understand the correct setting for the bwcutoff in the following two processes:
> 1)
> generate p p > w+ w+ j j QED=4 QCD=0, w+ > l+ vl @ 1
> add process p p > w- w- j j QED=4 QCD=0, w- > l- vl~ @ 2
>
> 2)
> generate p p > w+ w- j j $ t t~ QED=4 QCD=0, w+ > l+ vl, w- > l- vl~
>
>
> For (1) we observe that increasing the bwcutoff to a very large value is needed to include the full contribution of the off-shell W's. We ended up using bwcutoff = 1500, after noticing this gives a cross section that is a few percent higher than using bwcutoff = 150.
>
> For (2) however, when increasing the bwcutoff there appear many extra events because of a tail in the W mass distribution. Is this a consequence of the narrow width limit not being respected? And what would be the correct set-up to generate this process?
>
> We found that using bwcutoff = 35 to include the resonant Higgs contribution gives a good agreement with other generators. But it is not clear to me if we are missing the part equivalent to what gets added for (1) when increasing the bwcutoff to higher values.
>
> Thanks,
> Jasper
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Dear Olivier,

Thank you very much. Your reply has been very helpful!

Your recommended syntax (p p > l+ vl l+ vl j j) is indeed what we needed. For (1) we found that this gives results in agreement within the statistical uncertainty with using the decay-chain syntax with a bwcutoff = 15. There are some extra diagrams, but they are negligible after our generator level cuts (mjj > 100 GeV).
Increasing the bwcutoff for (1) also increases the cross section. For my understanding, is this because of problems in the phase space integration you mentioned, since the interference seems to be negligible?

For the second process, there are extra diagrams with only one intermediate W boson that do contribute significantly in our signal phase space. And an additional difficulty here is that generating events with "p p > l+ vl l- vl~ j j" is quite computationally intensive.

Is it possible to separate the diagrams from (2) with the diagrams that give the same final state but only have one intermediate W boson? Assuming their interference is negligible.

Thanks!
Jasper

Hi Jasper,

We can do "at least one W boson in s-channel"
by the following syntax
p p > w+ | w- > l+ vl l- vl~ j j

Is that what you need?

Cheers,

Olivier

Hi Olivier,

What I meant was separating the diagrams that have exactly one intermediate W boson with the vector boson scattering diagrams that have two intermediate W bosons.

Do you also have an idea what exactly goes wrong when increasing the bwcutoff for (1)?

Thank you,
Jasper