Enable EFT decomposition in decay syntax and Madspin?

Hi,

> I would like to understand exactly what the underlying issue is and whether it can be solved

This is mainly an issue of man-power. Combining those two features is a quite complex task.
We mainly need on volunteer to implement this.
If you have a PhD student available for that I can try to give him a MCnet short-term fellowship (4 month) to work on that.
https://www.montecarlonet.org/short-studentship

Cheers,

Olivier

> On 26 May 2020, at 13:41, Baptiste Ravina <email address hidden> wrote:
>
> New question #690976 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/690976
>
> Dear experts,
>
> I know there have been several tickets open in the past about the "squared amplitude constraints are not supported" error when using EFT models in MadGraph and MadSpin. I would like to understand exactly what the underlying issue is and whether it can be solved - indeed this would be tremendously helpful to a number of our colleagues! The baseline argument for allowing EFT decomposition would bring up the computing costs associated with generating the same (well-known) SM terms multiple times and the need for an efficient sampling of the Wilson coefficient phase-space in the context of LHC precision measurements and the efforts towards global EFT fits.
>
> To give some concrete examples, the following syntax is not allowed for the generation of the linear (SM/BSM interference) term only in a dileptonic ttbar process (with the insertion done in the ttbar production):
>
> import dim6top_LO_UFO
> generate p p > t t~ QCD<=2 QED=0 FCNC=0 DIM6=1 DIM6^2==1, (t > w+ b FCNC=0 DIM6=0, w+ > l+ vl FCNC=0 DIM6=0), (t~ > w- b~ FCNC=0 DIM6=0, w- > l- vl~ FCNC=0 DIM6=0)
>
> The following is however possible:
>
> generate p p > t t~ QCD<=2 QED=0 FCNC=0 DIM6=1 DIM6^2==1
>
> but if I try to run MadSpin to get the decay products of the tops (pure SM, no EFT insertion), it implicitly builds up the previous generate command and complains again.
>
> In an old ticket (https://answers.launchpad.net/mg5amcnlo/+question/255413), the following trick was suggested:
> 1) run "generate p p > t t~ QCD<=2 QED=0 FCNC=0 DIM6=1 DIM6^2==1"
> 2) edit the LHE file and proc card to remove the problematic dim6 syntax (and thus trick MadSpin into believing it's dealing with a SM ttbar process)
> 3) run "./bin/madevent" and "decay_events" to use MadSpin as usual
>
> This trick seems to work and produce a reasonable (within uncertainties) result, but clearly doesn't apply to EFT insertions in the decay of the tops, where one goes back to the complaint about constraints on squared amplitudes.
>
> Another alternative would be:
>
> generate p p > t t~ > b b~ l+ l- vl vl~ QCD<=2 QED=4 FCNC=0 DIM6=1 DIM6^2==1
>
> but as I understand it, this syntax makes the tops resonant and the top width independent of EFT contributions, and therefore might lead to unphysical results...
>
> Could you therefore comment on the plans for squared amplitude constraints in MadSpin and recommendations for achieving a consistent EFT decomposition without them in the meantime?
>
> Thanks for your help,
> Baptiste
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

> If you have a PhD student available for that I can try to give him a MCnet short-term fellowship (4 month) to work on that.

Thanks, I will look into that - would that be 4 months at 100% FTE?

In the meantime, could you comment on the validity of the resonant syntax to obtain the decomposition, e.g.

SM: generate p p > t t~ > b b~ l+ l- vl vl~ QCD<=2 QED=4 FCNC=0 DIM6=0
linear: generate p p > t t~ > b b~ l+ l- vl vl~ QCD<=2 QED=4 FCNC=0 DIM6=1 DIM6^2==1
quadratic: generate p p > t t~ > b b~ l+ l- vl vl~ QCD<=2 QED=4 FCNC=0 DIM6=1 DIM6^2==2

are there any side effects I should be aware of? are the spin correlations correctly handled? do I need to worry about the top width?

Cheers,
Baptiste

Hi,

Yes that would be 4 months pay at 100% and where the PhD need to work in Belgium (at least in principle since the covid might relax that constraint in theory). If know someone interested please contact me quickly by email since the deadline starts to be short (and remaining place quite limited). In the above link you will have full details on the scheme.

> In the meantime, could you comment on the validity of the resonant syntax to obtain the decomposition, e.g.

I do not know your model, so I can only guess if this is correct or not.
Be careful that
 QED=4 is the same as QED<=4

Otherwise, it sounds surprising to me that QCD (or QED) is not reduced by one for the quadratic, but this depend of the model author choice.

> are there any side effects I should be aware of?
- A big one, interference term can have huge issue with the phase-space integration, they are not guarantee that such computation converges (and if it does that it converges trough the physical number)
- A second one is that you need to be carefull about the dynamical scale choicde that you use for those three processes. The default dynamical scale will not be the same for the three processes (the interference run_card can be very different of the two others actually). For such computation using a dynamical scale like HT/2 is a better choice (and default for the linear term) than using the CKKW algortithm to pick a scale (which would be default for SM and for quadratic)

Cheers,

Olivier