MadGraph5_aMC@NLO

Taking Interference Into Account

Asked by Ben Tannenwald

Hello,

I'm looking into higgs decays with a final state of l+ l- and gamma. There are a few ways this can happen, but simply calculating each process independently and adding up the results seems too naive. If I add each process to the same generation, does MadGraph calculate the total amplitude (interference terms included) or does it calculate each separately and then weight and sum the results? I've been reading some documentation on the Single-Diagram-Enhanced multi-channel integration that MadGraph uses, but I'm finding it a little hard to wrap my head around.

Specifically, I'm asking if I run the commands:

import model h_Zgamma-full
generate p p > h > z a, z > mu+ mu-
add process p p > h > mu+ mu-, mu+ > mu+ a
etc.

output
launch

will I get a result that takes into account the interference between each channel? Any help would be much appreciated. Thanks!

Ben

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
Last query:
Last reply:
Revision history for this message
Best Olivier Mattelaer (olivier-mattelaer) said : 		#1

Hi Ben,

I think the syntax that you are looking for is:

> import model h_Zgamma-full
> generate p p > h > mu+ mu- a

This will include both h -> z a and h -> mu+ mu- with a photon radiation, as well as the photon emitted from the initial parton.
(so all possibilities)
the "> h >" syntax forces that all diagrams present have at least one higgs in s-channel.
Note that such syntax are in general NOT gauge invariant, and the computation breaks down as soon as the higgs is off-shell.
But this shouldn't be a problem for your particular process.

> add process p p > h > mu+ mu-, mu+ > mu+ a

Note that this syntax is not valid
the presence of comma is valid only if the particles is on-shell which is clearly not possible for
mu+ > mu+ a
Also the decaying particle should have a non vanishing width otherwise the computation will crash (and is not valid anyway).

Please take a look at the following tutorial for information about the meaning of the various syntax, and which one
are gauge invariant or not.
http://indico.cern.ch/getFile.py/access?contribId=4&resId=2&materialId=slides&confId=234296

> will I get a result that takes into account the interference between each channel?

No you will not.

In addition, if you have more than one process defined with the same final states.
I would suggest to add at the end of the line a @X
so for example
generate p p > z a, z > mu+ mu- @0
add process p p > z, z > mu+ mu- a @1
This ensures at 100% that none of your contribution might remove one of your previous contributions.
We had this problems in the past with MG4, in principle this shouldn't be a problem anymore with MG5,
but it's better to be safe.

Cheers,

Olivier

On May 9, 2013, at 11:11 PM, Ben Tannenwald <email address hidden> wrote:

> New question #228557 on MadGraph5:
> https://answers.launchpad.net/madgraph5/+question/228557
>
> Hello,
>
> I'm looking into higgs decays with a final state of l+ l- and gamma. There are a few ways this can happen, but simply calculating each process independently and adding up the results seems too naive. If I add each process to the same generation, does MadGraph calculate the total amplitude (interference terms included) or does it calculate each separately and then weight and sum the results? I've been reading some documentation on the Single-Diagram-Enhanced multi-channel integration that MadGraph uses, but I'm finding it a little hard to wrap my head around.
>
> Specifically, I'm asking if I run the commands:
>
> import model h_Zgamma-full
> generate p p > h > z a, z > mu+ mu-
> add process p p > h > mu+ mu-, mu+ > mu+ a
> etc.
>
> output
> launch
>
> will I get a result that takes into account the interference between each channel? Any help would be much appreciated. Thanks!
>
> Ben
>
>
> --
> You received this question notification because you are a member of
> MadTeam, which is an answer contact for MadGraph5.

Revision history for this message
Ben Tannenwald (btannenwald) said : 		#2

Olivier,

Thank you so much for the prompt and incredibly pertinent response. After looking through the tutorial you pointed me to (also very informative!), I just want to clarify what I think I learned to make sure I've got it straight.

Using the process

generate p p > h > mu+ mu- a

will give me all the contributions, but won't, in general, be gauge invariant. The syntax

generate p p > h, h > mu+ mu- a

should be gauge invariant though, yes? A final question, since you said that the interference wasn't taken into account the way I was generating before, is there a way to implement the effects from interference in MadGraph?

Ben

Ben

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#3

Hi Ben,

> generate p p > h, h > mu+ mu- a
>
> should be gauge invariant though, yes?

Yes but you are going to loose the photon emission coming from the initial state particles.

> Is there a way to implement the effects from interference in
> MadGraph?

No, this is not possible.
The only way to have interferences is to generate them with a single command.
If this create problems for a specific process, I can consider to implement this
and/or give you a patch allowing you to do that for that special case.

Cheers,

Olivier

On May 10, 2013, at 12:36 AM, Ben Tannenwald <email address hidden> wrote:

> Question #228557 on MadGraph5 changed:
> https://answers.launchpad.net/madgraph5/+question/228557
>
> Status: Answered => Open
>
> Ben Tannenwald is still having a problem:
> Olivier,
>
> Thank you so much for the prompt and incredibly pertinent response.
> After looking through the tutorial you pointed me to (also very
> informative!), I just want to clarify what I think I learned to make
> sure I've got it straight.
>
> Using the process
>
> generate p p > h > mu+ mu- a
>
> will give me all the contributions, but won't, in general, be gauge
> invariant. The syntax
>
> generate p p > h, h > mu+ mu- a
>
> should be gauge invariant though, yes? A final question, since you said
> that the interference wasn't taken into account the way I was generating
> before, is there a way to implement the effects from interference in
> MadGraph?
>
> Ben
>
> Ben
>
> --
> You received this question notification because you are a member of
> MadTeam, which is an answer contact for MadGraph5.

Revision history for this message
Ben Tannenwald (btannenwald) said : 		#4

Olivier,

Thanks again for the informative respoonse. Last question (I promise!): when you say the only way to generate interferences is to generate them with a single command, what exactly do you mean? Is there a way for me to implement interference directly from the command line?

Ben

Revision history for this message
Ben Tannenwald (btannenwald) said : 		#5

Thanks Olivier Mattelaer, that solved my question.

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#6

What I mean is that when you do
generate p p > e+ e-
You have two diagrams. The code takes into account the interference between those two diagrams.
But as soon as you try to do this with two commands like
generate p p > a > e+ e-
add process p p > z > e+ e-
Then you will loose the interference between the two diagrams.

Cheers,

Olivier

Revision history for this message
Ben Tannenwald (btannenwald) said : 		#7

Olivier,

I think I finally understand. I realize I keep saying thanks, but, again, your help has been much, much appreciated.

Ben

Revision history for this message
Ben Tannenwald (btannenwald) said : 		#8

Olivier,

This probably shouldn't need asking, but I figured it couldn't hurt to check. As MadGraph calculates processes at tree level (leading order), the interference is only calculated to leading order as well, correct?

Ben

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#9

Sure.

If you need NLO, I would obviously advised to use aMC@NLO.

Cheers,

Olivier

On May 14, 2013, at 3:51 PM, Ben Tannenwald <email address hidden> wrote:

> Question #228557 on MadGraph5 changed:
> https://answers.launchpad.net/madgraph5/+question/228557
>
> Ben Tannenwald posted a new comment:
> Olivier,
>
> This probably shouldn't need asking, but I figured it couldn't hurt to
> check. As MadGraph calculates processes at tree level (leading order),
> the interference is only calculated to leading order as well, correct?
>
> Ben
>
> --
> You received this question notification because you are a member of
> MadTeam, which is an answer contact for MadGraph5.

Revision history for this message
Ben Tannenwald (btannenwald) said : 		#10

Olivier,

Thanks for the advice! I didn't know MG had a utility for going to NLO, and this is good news! When I try to use the specially written h_Zgamma model (http://cp3.irmp.ucl.ac.be/~omatt/h_za/h_za.html), I get the response

Command "generate p p > h, h > mu+ mu- a [QCD]" interrupted with error:
InvalidCmd : The model h_Zgamma-full cannot handle loop processes

Are there plans to implement loop processes in the h_Zgamma model? I imagine since it's a non-standard model (pun not intended) that any changes are pretty low on the totem pole, but I figured I'd ask. And again, thanks for all the help.

Ben

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#11

Hi Ben,

Since this vertex is a pure QED one, this is in principle straightforward.
Indeed this vertex will not enter any loop/real corrections and therefore you don't any counter term/R2/...
So you just have to "merge" the two model, and therefore can in principle do it yourself if you need it now.
In fact the manual merge will be needed at a point since the hZgamma was not implemented via FR anyway.
However the HZGamma vertex requires modification from MG5/aMC@NLO in order to be supported.
This is not a big problem, but prevent you to be able to run it anyway.

Just to be clear, with my previous answer.
All diagrams that are drawned have the interference taken into account with all the other one.
The fact that you have here (resolved) loop and standard tree diagram didn't change anything.

Cheers,

Olivier

On May 14, 2013, at 6:06 PM, Ben Tannenwald <email address hidden> wrote:

> Question #228557 on MadGraph5 changed:
> https://answers.launchpad.net/madgraph5/+question/228557
>
> Ben Tannenwald posted a new comment:
> Olivier,
>
> Thanks for the advice! I didn't know MG had a utility for going to NLO,
> and this is good news! When I try to use the specially written h_Zgamma
> model (http://cp3.irmp.ucl.ac.be/~omatt/h_za/h_za.html), I get the
> response
>
> Command "generate p p > h, h > mu+ mu- a [QCD]" interrupted with error:
> InvalidCmd : The model h_Zgamma-full cannot handle loop processes
>
> Are there plans to implement loop processes in the h_Zgamma model? I
> imagine since it's a non-standard model (pun not intended) that any
> changes are pretty low on the totem pole, but I figured I'd ask. And
> again, thanks for all the help.
>
> Ben
>
> --
> You received this question notification because you are a member of
> MadTeam, which is an answer contact for MadGraph5.

Revision history for this message
Ben Tannenwald (btannenwald) said : 		#12

Olivier,

Sounds good. When you say I can "merge" the two models myself, what exactly would I have to do? And you say that the HZgamma vertex requires modification... is this something I can also do on my own, or would this require some heavy additions to the HZgamma model? Again, thank you very much for all your help.

Ben

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#13

You have to take the information related to the hza vertex (interactions/coupling/lorentz structure)
and include them in a copy of the loop_sm model.
Please read the UFO paper http://arxiv.org/abs/arXiv:1108.2040 to have all information on the format of the model.
This part you can do yourself in principle.

As you know for using the hza model, you need a special version of MG5. That is accessible with bazaar.
That version doesn't contain the modification related to aMC@NLO. Merging those modifications to the code
is not something trivial. So the best is to wait that those two versions are merged at some point.

Cheers,

Olivier

On May 16, 2013, at 3:21 PM, Ben Tannenwald <email address hidden> wrote:

> Question #228557 on MadGraph5 changed:
> https://answers.launchpad.net/madgraph5/+question/228557
>
> Ben Tannenwald posted a new comment:
> Olivier,
>
> Sounds good. When you say I can "merge" the two models myself, what
> exactly would I have to do? And you say that the HZgamma vertex requires
> modification... is this something I can also do on my own, or would this
> require some heavy additions to the HZgamma model? Again, thank you very
> much for all your help.
>
> Ben
>
> --
> You received this question notification because you are a member of
> MadTeam, which is an answer contact for MadGraph5.