Add saving of true flight distance information

[particleist]

We have all the information needed to save the true flight distance of the decaying particle from its parent vertex. In the first instance this should be truth only information, later we need to consider how to add smearing of PVs and SVs but that's a bit less trivial.

[gcowan]

We were just discussing this yesterday. Should be simple to add. We would also like to add the FD information for user-defined two-body combinations of the decay products.

[particleist]

I'm not sure I quite see the point of the user-defined combinations? How do you see that working?

[gcowan]

Donal asked us if RapidSim would be able to help him understand what fraction of B -> [K π π π]_D K events end up being reconstructed as B -> [Ks0 K π]_D K, where two of the pions are reconstructed as a flying Ks0. So if we can compute the FD for combination of two particles, we can see what efficiency a particle FD cut would have.

[particleist]

But as long as there is no smearing the pipi vertex will coincide exactly with the D vertex, no? So this I think only comes in once we start smearing things.

[dcraik]

Our idea was to define the "decay" vertex as the point of closest approach between the lines defined by the smeared momenta and true origin vertices, and use the decay of the most recent common ancestor as the "origin" vertex.

I don't have a feeling for the relative sizes of the effects on the FD due to the momentum and SV smearing so I'm not sure how useful it is to do this before implementing SV smearing.

[particleist]

I think there are a couple of things to take care of for the FD and vertex smearing to make sense

1/ Smearing PVs can be done based on measured resolutions in X/Y and Z as a function of the number of tracks in the PV. This can be trivially added to the beam parameters, generated along with the pileup, and propagated.

2/ For SV however, the momentum resolution needs to be complemented with some simulation of the multiple scattering in the VELO which is to first order decoupled from the P resolution. That could be naturally done once we implement the VELO material model for the brems, in the meantime I don't see a super easy way to do it?

Thoughts?

[gcowan]

For 2), let's wait for the VELO material model to go in.

For 1), is there any public information about the X/Y and Z resolutions? There must be in some VELO performance paper. What about as a function of nPVTracks? Obviously we don't have nPVTracks in RapidSim, so I we should just randomly sample from a TH3 of (nPVTracks, XYres, Zres).

[particleist]

For nPVTracks we can just use Pythia to generate a distribution to sample from, no?

For the resolutions you want this

https://cds.cern.ch/record/1756296?ln=en

figure 3

[gcowan]

Perfect for the resolutions. We can just parameterise these with an exponential. e.g.,

[gcowan]

I have added some initial code in 27fbd37 for a RapidVertex class. Doesn't do anything fancy, just holds a std::pair with he vertex and it's smeared version. Using this we can modify the existing computation of IP, FD to use the true or smeared vertex.