Diagram removal does not remove crossed diagrams

[Danilo Enoque Ferreira de Lima]

Dear MadGraph authors,

I am trying to generate the following process:

    import model HHVBF_UFO
    generate p p > h h j j QCD=0 --diagram_filter

I use the diagram filter to keep only diagrams where a Z/W decays into quarks in the final state (the code used is below). That is, I want to select the non-VBF diagrams to study them separately.

When doing this, I always get VBF diagrams anyway (for some quark configurations), although those diagrams disappear, when specifying the exact initial and final state. For example, I see some of the VBF diagrams in the u c~ > h h u c~ process with this:

generate u p > h h j j QCD=0 --diagram_filter

But not if I type directly:

generate u c~ > h h u c~ QCD=0 --diagram_filter

(in the latter case, all diagrams are removed by the filter)

I have identified that such diagrams are always produced by crossing symmetry in the MadGraph output. That is, I see messages such as the following one for the offending diagrams:

INFO: Crossed process found for u c~ > h h u c~, reuse diagrams.

Looking in the code, I have manually changed madgraph/code/diagram_generation.py in line 1785 (of MG5 2.6.5) leading to the following diff in the generate_multi_amplitudes function (where it checks for crossing diagrams):

daniloefl@daniloefl-work:~/workspace/MG5_aMC_v2_6_5$ diff madgraph/core/diagram_generation.py.orig madgraph/core/diagram_generation.py
1785c1785
< not process.get('is_decay_chain'):
---
> not process.get('is_decay_chain') and not diagram_filter:

After this, the diagrams that I wish were removed no longer appear, reinforcing my suspicion that this is a bug.

For completeness, I add my code in PLUGIN/user_filter.py below, if it is useful (as I could have made a mistake in it too).

Best regards,
Danilo

def remove_diag(diag):
    """keep all diagrams with a V->jj decay, where j is a final-state quark"""
    found_vjj = 0
    found_tchan = 0
    for vertex in diag['vertices']:
        if vertex.get('id') == 0: # special final vertex
             continue
        if len(vertex['legs']) != 3: # ignore 4-point vertices
            continue

        leg = vertex['legs'][2]
        if (abs(leg['id']) == 24 or abs(leg['id']) == 23):
            if leg.get('number') > 2:
                leg = vertex['legs'][0]
                if not(abs(leg['id']) < 6 and leg['state']):
                    continue
                leg = vertex['legs'][1]
                if not(abs(leg['id']) < 6 and leg['state']):
                    continue
                found_vjj += 1
        else:
            leg = vertex['legs'][-1]
            if leg.get('number') < 3:
                continue
            leg = vertex['legs'][-1]
            if not(abs(leg['id']) < 6 and leg['state']):
                continue
            i = 1
            if abs(vertex['legs'][i]['id']) == 24 or abs(vertex['legs'][i]['id']) == 23:
                i = 0
            leg = vertex['legs'][i]
            if not(abs(leg['id']) < 6 and leg['state']):
                continue
            found_vjj += 1
    order_ok = (diag['orders'].get('C3') == 0 and diag['orders'].get('C2V') == 0 and diag['orders'].get('CV') >= 2)
    accept = (found_vjj >= 1 and order_ok)
    return not accept

[Olivier Mattelaer (olivier-mattelaer)]

Hi Danilo,

Can you attach your model?
This model is not our online database of model.
I have also tried with the SM model, but there, your script removes 100% of the diagram.

Cheers,

Olivier

[Danilo Enoque Ferreira de Lima (daniloefl)]

Hi Olivier,

as I am not entirely sure I can share this publicly, I have sent you the model in a private email.

Cheers,
Danilo

[Olivier Mattelaer (olivier-mattelaer)]

Thanks,

Your fix is correct and I pushed it in the development version.

Thanks,

Olivier

[Danilo Enoque Ferreira de Lima (daniloefl)]

Thank you, Olivier.