Invariant mass spectrum for a new resonance is too symmetric?

Hi,

Difficult to answer that without the model.
My guess is that your operator ggy0 or y0aa enhance the high energy which make sense since those are typically dimension 5/6 operator which breaks unitarity as some point.

Cheers,

Olivier

> On Mar 9, 2016, at 04:32, Martin <email address hidden> wrote:
>
> New question #288317 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/288317
>
> Hi,
>
> Using Feynrules, I have added an effective vertex for a new scalar produced via gluon fusion,which decays to two photons via another effective vertex.
>
> When I plot the invariant mass of the photons, I get a very symmetric distribution, even when a large mass and width are used (e.g. 1500 GeV mass, 300 GeV width). However, I expect to see an asymmetric distribution, since the gluon parton density is decreasing up to high mass, which means that the high tail of the Breit Wigner gets suppressed relative to the low tail.
>
> I've generated the resonance using:
>
> generate g g > a a
>
> which only gives a single diagram in my EFT (as expected). Using g g > y0, y0 > a a makes no difference, which is presumably expected from the UFO model (there is no off-shell calculation?).
>
> Have I misunderstood something about the symmetry of the distribution? It looks to me like it simply uses a Breit-Wigner with no additional correction.
>
> Cheers,
>
> Martin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi,

The model indeed has dimension 5 operators (products of field stength tensors plus the new scalar, i.e. G_munu G^munu Y0, for the SU(3), SU(2) and U(1) field strength tensors).

It seems odd that these give an exact symmetry of the invariant mass distribution however, since this implies an exact cancellation of the effect of falling gluon density which would be a crazy accident. I will try looking at the distribution for an SM Higgs with a changed mass and width to see if I see the same thing there.

Do you have a case that gives an asymmetric invariant mass distribution?

Cheers,

Martin

Hi,

> Do you have a case that gives an asymmetric invariant mass distribution?

This sounds a quite small effect but yes it is present.
I have made multiple plot for g g > a a in heft model and I can see a systematic asymmetry,
I look in particular to mh=1500, width=10 and I forbid the Higgs to be between 1400 and 1600 GeV.

Cheers,

Olivier

> On Mar 9, 2016, at 22:17, Martin <email address hidden> wrote:
>
> Question #288317 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/288317
>
> Status: Answered => Open
>
> Martin is still having a problem:
> Hi,
>
> The model indeed has dimension 5 operators (products of field stength
> tensors plus the new scalar, i.e. G_munu G^munu Y0, for the SU(3), SU(2)
> and U(1) field strength tensors).
>
> It seems odd that these give an exact symmetry of the invariant mass
> distribution however, since this implies an exact cancellation of the
> effect of falling gluon density which would be a crazy accident. I will
> try looking at the distribution for an SM Higgs with a changed mass and
> width to see if I see the same thing there.
>
> Do you have a case that gives an asymmetric invariant mass distribution?
>
> Cheers,
>
> Martin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

HI Olivier,

Many thanks for your help so far.

I can reproduce your observed asymmetry in the heft model, but it is much smaller than I expect. This might be fine, but I want to check understanding.

I'm working with Lydia Roos and Yee Yap who have previously looked at BSM Higgs studies using PowHeg. They see a much larger asymmetry in the invariant mass of the photons, as you can see in the two slides I have put here:

https://dl.dropboxusercontent.com/u/52286446/madgraph_tests.pdf

In the PowHeg case, the distribution is well matched by a simple model that convolves the Breit-Wigner distribution with the squared matrix element for the production and the parton luminosity. You can see on the second slide that the effect of the falling gluon luminosity is to cause a pronounced skew of the distribution even on a linear scale.

With Madgraph, I see a small asymmetry on a log scale, but it is hard to resolve on a linear scale,and it is clearly a much smaller effect.

Now, this could result from differences in the underlying physics model but is there any way we can check in more detail? Your suggestion that it is due to dimension 5 operators favouring the high mass tail (which then leads to less of an asymmetry after the PDF is considered) should give a greater skew at lower invariant mass presumably, since here the invariant masses probed by the interaction are well under the cut off scale in the theory? i.e. the high mass tail is less hyped, which means the skew to the low tail should be enhanced? Or do I misunderstand? In any case, that is not what I see.

Cheers,

Martin

Hi,

> Now, this could result from differences in the underlying physics model
> but is there any way we can check in more detail? Your suggestion that
> it is due to dimension 5 operators favouring the high mass tail (which
> then leads to less of an asymmetry after the PDF is considered) should
> give a greater skew at lower invariant mass presumably, since here the
> invariant masses probed by the interaction are well under the cut off
> scale in the theory? i.e. the high mass tail is less hyped, which means
> the skew to the low tail should be enhanced? Or do I misunderstand? In
> any case, that is not what I see.

The HEFT model breaks for mh>2*mt so you are clearly out of his validation range.
So it is clearly possible to have huge effect.

> I can reproduce your observed asymmetry in the heft model, but it is
> much smaller than I expect. This might be fine, but I want to check
> understanding.

Thanks for checking this, keep me in touch.

Cheers,

Olivier

> On Mar 15, 2016, at 05:22, Martin <email address hidden> wrote:
>
> Question #288317 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/288317
>
> Status: Answered => Open
>
> Martin is still having a problem:
> HI Olivier,
>
> Many thanks for your help so far.
>
> I can reproduce your observed asymmetry in the heft model, but it is
> much smaller than I expect. This might be fine, but I want to check
> understanding.
>
> I'm working with Lydia Roos and Yee Yap who have previously looked at
> BSM Higgs studies using PowHeg. They see a much larger asymmetry in the
> invariant mass of the photons, as you can see in the two slides I have
> put here:
>
> https://dl.dropboxusercontent.com/u/52286446/madgraph_tests.pdf
>
> In the PowHeg case, the distribution is well matched by a simple model
> that convolves the Breit-Wigner distribution with the squared matrix
> element for the production and the parton luminosity. You can see on the
> second slide that the effect of the falling gluon luminosity is to cause
> a pronounced skew of the distribution even on a linear scale.
>
> With Madgraph, I see a small asymmetry on a log scale, but it is hard to
> resolve on a linear scale,and it is clearly a much smaller effect.
>
> Now, this could result from differences in the underlying physics model
> but is there any way we can check in more detail? Your suggestion that
> it is due to dimension 5 operators favouring the high mass tail (which
> then leads to less of an asymmetry after the PDF is considered) should
> give a greater skew at lower invariant mass presumably, since here the
> invariant masses probed by the interaction are well under the cut off
> scale in the theory? i.e. the high mass tail is less hyped, which means
> the skew to the low tail should be enhanced? Or do I misunderstand? In
> any case, that is not what I see.
>
> Cheers,
>
> Martin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier,

I have had some more time to check details on this (in discussion with the colleagues mentioned above, plus Stefano Frixione). I have put some slides here:

https://dl.dropboxusercontent.com/u/52286446/madgraph_diphoton_tests.pdf

The executive summary is that the parton density is indeed having a significant effect (proved by setting the PDF by hand to 1 in every event, giving rise to a clear enhancement of the high mass tail). I have some remaining questions and misunderstandings, however:

1) In the case of my EFT, I can explicitly put the cutoff scale to high values which should suppress the effect. This presumably occurs when the width of the scalar is set to "auto" and the width is actually calculated correctly.

2) What does Madgraph do in the case of a fixed width (this is ultimately what we want to use)? There is some set of Lagrangian parameters that maps to the width in this case, but the mapping is many to one. My impression is that the Lagrangian parameters are ignored in this case, and Madgraph simply uses a Breit-Wigner of the required width. But then how does Madgraph know about the high mass tail enhancement, and is it necessarily wrong (i.e. the cutoff scale that maps to that width might be quite high)? How are the operators of the EFT actually used in this case?

3) I see a similar enhancement of the high mass tail in SM Z production and decay. Is this expected?

I don't think Madgraph is doing anything wrong here, I just need to understand what is going on under the hood.

Cheers,

Martin

Hi Martin,

> 1) In the case of my EFT, I can explicitly put the cutoff scale to high
> values which should suppress the effect. This presumably occurs when the
> width of the scalar is set to "auto" and the width is actually
> calculated correctly.

I do not understand this comment.

> 2) What does Madgraph do in the case of a fixed width (this is
> ultimately what we want to use)? There is some set of Lagrangian
> parameters that maps to the width in this case, but the mapping is many
> to one. My impression is that the Lagrangian parameters are ignored in
> this case, and Madgraph simply uses a Breit-Wigner of the required
> width.

The total width is only used in multiple place in the code:
1) in the associate propagator of the matrix element. This is creating the Breit-Wigner shape.
2) as a cut-off if you use g g > y0 , y0 > a a then the y0 is associate to an on-shell cut which is proportional to the width.
if you use g g > a a this cut should not be used at all.
3) the width enter in some of the change of variable used in the phase-space integration. Those one should not have any impact on the final result.

The width is a free parameter that you set (Auto is a special value which set the width for you at a fixed value)

> But then how does Madgraph know about the high mass tail
> enhancement, and is it necessarily wrong (i.e. the cutoff scale that
> maps to that width might be quite high)?

Madgraph integrates over the full allowed phase-space and the integration will automatically adapt to your integrand.
At first iteration only a few phase-space point will be tested in the tail since they are suppose to be suppress but this is enough to detect that they are larger than expected and during the following iteration the code will adapt to probe more in details that part of the phase-space.

> How are the operators of the
> EFT actually used in this case?

The EFT operator are treated as any other Feynman Rules. Nothing special is done (but if you ask MadGraph to only compute the interference term of the EFT with the SM)

> 3) I see a similar enhancement of the high mass tail in SM Z production
> and decay. Is this expected?

Setting the PDF to one will certainly enhance the high energy region for any processes.

Cheers,

Olivier

> On Mar 29, 2016, at 06:52, Martin <email address hidden> wrote:
>
> Question #288317 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/288317
>
> Status: Answered => Open
>
> Martin is still having a problem:
> Hi Olivier,
>
> I have had some more time to check details on this (in discussion with
> the colleagues mentioned above, plus Stefano Frixione). I have put some
> slides here:
>
> https://dl.dropboxusercontent.com/u/52286446/madgraph_diphoton_tests.pdf
>
> The executive summary is that the parton density is indeed having a
> significant effect (proved by setting the PDF by hand to 1 in every
> event, giving rise to a clear enhancement of the high mass tail). I have
> some remaining questions and misunderstandings, however:
>
> 1) In the case of my EFT, I can explicitly put the cutoff scale to high
> values which should suppress the effect. This presumably occurs when the
> width of the scalar is set to "auto" and the width is actually
> calculated correctly.
>
> 2) What does Madgraph do in the case of a fixed width (this is
> ultimately what we want to use)? There is some set of Lagrangian
> parameters that maps to the width in this case, but the mapping is many
> to one. My impression is that the Lagrangian parameters are ignored in
> this case, and Madgraph simply uses a Breit-Wigner of the required
> width. But then how does Madgraph know about the high mass tail
> enhancement, and is it necessarily wrong (i.e. the cutoff scale that
> maps to that width might be quite high)? How are the operators of the
> EFT actually used in this case?
>
> 3) I see a similar enhancement of the high mass tail in SM Z production
> and decay. Is this expected?
>
> I don't think Madgraph is doing anything wrong here, I just need to
> understand what is going on under the hood.
>
> Cheers,
>
> Martin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.