SMEFT negative distributions using MC weights

Hi,

This looks like the same pattern as the Z/photon interference term. Isn't it?

Cheers,

Olivier

> On 27 Jun 2023, at 18:45, James <email address hidden> wrote:
>
> New question #707128 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/707128
>
> Hi,
>
> I am generating charged current Drell-Yan events using Madgraph. I want to generate the SM process as well as three additional processes which receive a contribution from a single Wilson coefficient of the SMEFT. The three Wilson coefficients I am looking at are cll1, cHl1, and clq3. I use the model SMEFTsim_U35_MwScheme_UFO-CCDY and generate the process as follows:
>
> import model SMEFTsim_U35_MwScheme_UFO-CCDY
> define l+- = e+ e- mu+ mu-
> define vl+- = ve ve~ vm vm~
> generate p p > l+- vl+- NP<=1 SMHLOOP=0 NP^2==1
>
> I am then using AnalysisTop to plot distributions associated to the final state kinematics. I am able to identify the truth-level final-state neutrino and lepton; and plot distributions accordingly. I also include a weight "weight_mc" for each of the events in my histograms, from the MC generator weights. I plot two distributions for each variable, one corresponding to the electron decay channel and the muon decay channel. The SM distributions look as expected, as do the cll1 distributions. The cHl3 and clq3 distributions exhibit negative distributions, however. More specifically, the invariant mass of the W boson starts negative and close to zero at m_W = 0. It becomes increasingly more negative as m_W increases and suddenly swaps sign at the W-pole (~80 GeV) to be come very large and positive, gradually decreasing to zero as m_W moves away from the W-pole.
>
> Does anyone know why including the weights causes this strange effect to occur?
>
>
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

Thank you for your response, but I don't quite understand what you mean. Are you saying that the neutral current Drell-Yan generation with cHl3 and clq3 causes the same effect? The unweighted distribution looks as I expected, with a peak at the W-pole and no negative distributions. For some reason, adding in the MC weights causes the distribution of events with lepton-neutrino invariant mass below 80 GeV to shift below zero. In other words, the majority of events with invariant mass below 80 GeV have a negative weight. Is there a reason for this?

Thanks,
James

I just came across your post about SMEFT negative distributions using MC weights. It's fascinating to see the advancements in this field! By the way, have you considered the importance of regular maintenance, such as AC cleaning, to ensure optimal performance and longevity of your equipment? It's an essential aspect to keep in mind. Great work on your research!
https://www.ipurity.com/our-services/ac-and-ac-duct-cleaning-in-dubai/

Hi,

Interference are not positive definite, so you do expect to have positive and negative weighted event when you compute such quantity. So you always have to keep the weight into account.
When you pass the pole mass, you indeed change the phase of the amplitude and therefore you can have a change of sign in the interference term. Since the interference is large where the amplitude is large, it does make sense that when you do not consider the sign, you do see a peak, this is however not a normal peak in term of shape since it technically has a dip in the middle.

I do not know your SMEFT model, so I can not really comment on the model in details, but from what you present everything seems as expected to me.

Cheers,

Olivier

Hi Olivier,

Thank you for your answer, I did not realise that this was normal behaviour!

Cheers,
James

Thanks Olivier Mattelaer, that solved my question.

Hi Olivier,

I have spotted another difference between the SM generation and the linear interference generation which I think is unusual. I am able to construct the invariant mass distribution of the neutrino and charged lepton and compare it to the four-momentum of the intermediate boson, which is represented by a stock variable pdfinfo_Q included in the output of AnalysisTop. In the SM generation, to invariant mass and the pdfinfo_Q agree, as expected. In the linear interference generation, however, the invariant mass looks as expected, but the pdfinfo_Q looks different. The pdfinfo_Q distribution exhibits a peak at 40 GeV (half the W-mass) rather than 80 GeV and also features an enlarged tail in the lower mass regions.

I was wondering if you knew the reason for this difference? Perhaps AnalysisTop doesn't register the interference term for some reason and only sees a single SM contribution at the amplitude-squared level, and hence gives a peak which is half that of the W-mass?

Thanks,
James

Hi,

Not all events have the "W" information written inside the lhef. Only onshell "W" are written (and should) inside that file.
I would bet that the difference between your plot is coming from that specific reason.

Cheers,

Olivier

Hi Olivier,

I believe that I take this into account in my AnalysisTop code. The events which do not contain an intermediate boson seem to have a four-fermion vertex qq->ll rather than two vertices for the production and decay of the Z-boson, which I include. I am able to identify the final state leptons from each event yet I still observe a problem with the momentum distribution. It also seems strange that this effect would result in the four-momentum peak being half of the invariant mass peak.

Thanks,
James

Then I do not know, the issue is likely in the analysistop code but I have no experience with such code in order to comment.

Cheers,

Olivier

Hi Olivier,

Thank you for your help. I have asked some AnalysisTop experts and I have been told that the pdfinfo_Q variable is extracted by looking up TruthEventsAuxDyn.Q from the CollectionTree inside my TRUTH1 sample. The AnalysisTop experts seem to believe that the issue arises from how MadGraph writes the truth record for that process with the SMEFT model SMEFTsim_U35_MwScheme_UFO I am using. Does this information help you understand the issue?

Thanks,
James

Hi,

Not that much since I do not know what "TRUTH1" is, neither "pdfinfo_Q" , or "TruthEventsAuxDyn.Q".

> The AnalysisTop experts seem to believe that the issue arises from how MadGraph writes the truth record for that process with the SMEFT model SMEFTsim_U35_MwScheme_UFO I am using.

This make me things, that the issue should be related to the fact that we do write or not the "W" particles within the lhef file.
This information should not be used for analysis but only by the parton-shower to constraint the radiation pattern/recoil.

But I guess long story short ... do not use that variable within AnalysisTop.

Cheers,

Olivier