Question on excluding s-channel particles with $$

Hi Benjamin, Andrew

I have simply no idea of what those parameters are.
So Benj are you able to comment?

Cheers,

Olivier

Hi all,

I do not see why the cross section should be independent of the sign of the elements of the neutralino mixing matrix. Naively, I would say that the squared matrix element is a sum of terms in N[i,j] U[k,l] and N[i,j] V[k,l] (assuming everything real) where N is the neutralino mixing matrix and U/V the chargino mixing matrices.

Cheers,

Benjamin

Hi Benjamin,

Is the squared matrix element summed in terms of N[i,j]^2 or just N[i,j]U[k,l]? I'm wondering because perhaps I must change the sign of terms in U and V if I change the sign of a column in N.

Also, I thought that maybe this effect could be coming from the fact that I used "$$" to exclude some diagrams, so I ran on the full process instead. While the difference is now smaller, it is still statistically significant.

I ran one sample with default values of the second column of the NMIX matrix, and one with the values flipped. I got a difference in cross section of about 2.5% which was outside of the statistical uncertainty! The difference is 15 times larger than the statistical uncertainty.

Default: sigma = 0.2858+-0.0005 pb
Rotated RN2*: sigma = 0.2933+-0.0005 pb

Here's the details on my param_card.dat.
One sample I ran with the settings:

Masses
  1000022 4.49837e+02 # Mneu1
  1000023 -4.79916e+02 # Mneu2
  1000025 5.06467e+02 # Mneu3
  1000035 4.00161e+03 # Mneu4
  1000024 4.78022e+02 # Mch1
  1000037 4.00161e+03 # Mch2

NMIX
....
    2 1 3.38520e-02 # RNN21
    2 2 -1.35155e-02 # RNN22
    2 3 7.05399e-01 # RNN23
    2 4 -7.07872e-01 # RNN24
....

And for the other I flipped the sign of the RNN2? parameters only.
other sample...
    2 1 -3.38520e-02 # RNN21
    2 2 1.35155e-02 # RNN22
    2 3 -7.05399e-01 # RNN23
    2 4 7.07872e-01 # RNN24
.....

Any ideas or things for me to check?
Thanks,
Andy

Hi Andy,

> Is the squared matrix element summed in terms of N[i,j]^2 or just
> N[i,j]U[k,l]? I'm wondering because perhaps I must change the sign of
> terms in U and V if I change the sign of a column in N.
If you are producing two neutralinos, then you have products like N[imj] N[k,l] in the squared matrix element (let’s again forget about complex conjugation). In contrast, if you produce an associated pair with one chargino and one neutralino, then you will get products like N[i,j] U[k,l[] and N[i,j] V[k,l]. Of course, an extra dependence on N/U/V could arise from specific diagrams containing extra vertices involving neutralinos and/or charginos, but some diagrams will exactly involve two of such vertices, which will lead to the above-mentionned structure.

But the physics should not be sensible to such a sign, whatever you modify or not U/V. So I am a bit puzzled.

> Also, I thought that maybe this effect could be coming from the fact
> that I used "$$" to exclude some diagrams, so I ran on the full process
> instead. While the difference is now smaller, it is still statistically
> significant.
> I ran one sample with default values of the second column of the NMIX
> matrix, and one with the values flipped. I got a difference in cross
> section of about 2.5% which was outside of the statistical uncertainty!
> The difference is 15 times larger than the statistical uncertainty.
I do not know what to tell. First, the $$ might break gauge invariance, which could explain what you initially observe. Second, maybe could you try without the parton densities? To see if the problem persists. Equivalently, you could also check what is going on at the matrix element level (using madgraph-standalone). I hope this helps.

Cheers,

Benjamin