Negative mass squared in LowEnergy

[Avelino]

Hi Florian,

I found a problem with a SPheno module that I just obtained with SARAH. It works perfectly fine as long as I don't compute low energy observables. However, when I do that I get an error due to negative squared masses. I could trace it back to the CalculateLowEnergyConstraints routine, and in particular to the part labeled

Code: Select all

## SM only ##

Here some (but not all) BSM parameters are set to zero and one BSM VEV (but not all) is set to a small value. Then, when TreeMasses is called again the resulting hh squared mass matrix contains a negative eigenvalue. Is this OK or more parameters should be set to zero here as well?

Thanks in advance.

[FStaub]

Hi Avelino,

I guess you have a model where SM fermions mix with new physics? In these cases, all parameters which are responsible for that mixing should be set to zero.Otherwise, it's not possible to get a SM prediction for the observables.
I'm only using for VEVs not exactly zero but a small value to prevent numerical problems. This might screw up the BSM sector, but shouldn't affect the SM observables because both sectors should be decoupled at this point.

Cheers,
Florian

[Avelino]

I see, thanks for explaining. The problem appears in the scalar sector and I believe it's due to the fact that only some of the BSM parameters are set to zero, and not all.

More precisely, I have three CP-even scalars, the Higgs + 2 extra BSM scalars. It is true that the parameter choice made by SPheno decouples the Higgs and then, if it did not stop running, it would serve to compute the SM observables correctly. However, negative squared masses appear in the 2x2 BSM block, which has the structure

{{vBSM^2, vBSM mu} , {vBSM mu, vBSM_2^2}}

SPheno only sets vBSM -> 0.001, but leaves vBSM_2 and mu unaffected. Therefore, if I have {vBSM -> 0.001, mu -> -100, vBSM_2 -> 100} I get a negative squared mass and SPheno stops.

[FStaub]

Hmmm... ok, then it would be the easiest if you put also these VEVs to zero. Unfortunately, it's highly model dependent what the optimal way to decouple both sectors is. If you have any proposal for a general procedure, let me know.

Cheers,
Florian

[Avelino]

Thanks. That solution of course works, but I admit I do not have any clear idea about how to do it in general.