review impact parameter b tagging

[cbernet]

Hi Nicolò (@nfoppiani),

I would like to use your btagging parametrization for the new detector models we're setting up in papas (namely CLIC and IDEA), as the parametrization of the resolution on the impact parameter is commonly given as a target for future detectors.

So I started to review your code and to work on a better integration.

First, I am going to separate the impact parameter smearing and the jet tagging, to put them in different analyzers. This will make it possible to run several jet tagging algorithms on the same tracks after smearing.

Second, there are a few points that are puzzling me, and you can find my questions and comments here: clic...for_nic
They are marked with a COLIN->NIC tag

Could you please have a look? Then we can discuss these questions on skype if you want, or you can provide answers here if you prefer.

Based on my current understanding, the main items are the following:

• i suspect that your IP calculation might be wrong, as it does not give the right results in simple cases, see https://github.com/cbernet/heppy/blob/master/papas/test_path.py#L61. On the contrary, Lucas' calculation is correct. I could make yours work by changing the arguments of the minimize_scalar method.
• i'm not sure that the smearing should be done separately on the x and y components of the IP vector. How do you support this choice?
• i think that smearing should have an effect on the sign of the IP

Cheers & thanks in advance for your help!

Colin

[nfoppiani]

Hi Colin,

sorry for my late answer but I missed this mail.

Well, it is probably easier to discuss everything by skype, in particular the comments marked by COLIN->NIC.

For what concerns the three items:

• I remember that I tested the IP calculation with simple cases with typical numbers of the IP (like particles with high velocity and IP of the roder of tens of microns( and it seemed to be working.
  I also remember that I had troubles with that scipy function and I am not surprised that my tuning of
  the arguments of minimize_scalar was not perfect.
• This was a non trivial point. Basically the IP resolution reflects (at least in the core) a gaussian resolution of the components of a 2d vector, and you can see it in some plots of the IP resolution. For this reason we decided to smear separately the two x and y components.
• Yes I agree. I remember I thought about this point but I have not found any easy way to do it.

Anyway I have a few ideas to improve the algorithm, we can also discuss them by Skype!

Cheers
Nick