sPHENIX MAPS tracker discussion

[Anthony Frawley <afrawley AT fsu.edu>]

Hello Ross,

That all sounds good.

Single tracks are preferable for this exercise, I think.

I agree with Christof that comparing the extrapolated reconstructed track position with the truth hit position at the first layer of the TPC would be better. Using the truth momentum and extrapolating to get the truth position, then taking the difference to the reconstructed track extrapolation, would be fine.

The truth hit position in the first layer is available, since that is an active volume. I am not sure if the truth tracks will be associated with hits in all active volumes, but if so you should be able to extract the g4hit position matched to the truth track ID. Otherwise, in a single track environment, the truth hit position in a TPC layer can be extracted from ntp_g4hit by asking for hits with "layer == xxx && gprimary ==1". To get it from inside the code, you would have to loop over the g4hit container and find the layer.

Cheers

Tony

---

From: Ross Corliss <rcorliss AT mit.edu>
Sent: Friday, October 5, 2018 11:14 AM
To: Anthony Frawley; christof.roland AT cern.ch
Cc: jhuang AT bnl.gov; Todoroki, Takahito; Carlos Perez; Chiu, Mickey; Sanghoon Lim; Molly Taylor; sphenix-intt-l AT lists.bnl.gov; sphenix-maps-l
Subject: Re: Meeting of the Inner Tracker Task Force VIRTUAL

 

Tony and Christof,

Thank you both for the comments.  I reply in-line (I believe these responses cover all of Christrof's bullets as well):

On Oct 4, 2018, at 10:56 PM, Anthony Frawley <afrawley AT fsu.edu> wrote:

Hello Ross and Molly,

Interesting! Some comments/questions:

1) Why electrons? There will be significant tails on the extrapolation for electrons after 8% of a radiation length in 4 layers of INTT. 

Electrons were a somewhat arbitrary, low-mass choice.  The next version will use pions, per other suggestions.

2) You extrapolated to the inner field cage at 20 cm. The inner tracking radius is 30 cm, where space charge distortions will be worst.  Can you push it to 30 cm?

The extrapolation past the IFC can be done, but is a little harder because it needs to know about the material layer at 20cm.  I will see if this is already handled, or easily added...

3) I am surprised  by the Z resolution for "11" and "111" on slide 4 and 5. I would have expected these to be much worse than those for any configuration having a "0" layer. The implication of this is that the Z resolution is dominated by the MVTX (which is actually not insane, it has 5 micron resolution per layer vs ~40 micron resolution for the INTT "0" layers). This raises the question of whether the INTT matters at all to the Z extrapolation accuracy. Can you try it with no INTT at all?

We will add this to the next iteration.  I, too, was surprised that the Z resolution was not so strongly impacted by the absence of 0-type layers.

4) You are using a fixed momentum of the particles, and the extrapolation trajectory in r-phi is always the same at fixed momentum. I assume that the extrapolation uses the momentum measured by the MVTX and INTT to predict the hit location, and you show the distribution around the mean.  The momenta we would actually be using would be a few hundred MeV/c to maybe 1 GeV/c, and we would not have the momentum from a precise TPC measurement until we have already corrected for space charge effects. It may be that we can take the TPC space points at the space charge crossover point to get a better estimate of the track momentum, but otherwise we rely on the momentum measurement by the MVTX + INTT for the extrapolation. Can you try it with a continuous range of particle momenta?

To be clear:  We use the kalman filter to fit the MVTX+INTT hits, which gives us a TrackRep.  From this, we can extract the predicted positions, as well as the covariances, at any desired radius (assuming the track reaches that radius).  The variables shown in yesterday's slides are not the differences between actual hit and predicted hit, but instead the errors, as extracted from the covariance matrix, when the track is extrapolated to R=20cm.  A good check (that is tricky to implement) would be to show that the actual hit locations on the IFC, or the layer of choice, are distributed similarly to the predicted covariance.

The seeding of this process may cheat and take in Truth momenta somehow, though in my read through the starting prediction at the origin seemed to be hard-coded rather than pulled from other data.

I think it is important to note that all of this is being done in a single-particle environment.  The matching to MVTX and inner TPC hits may get messier when pile-up is considered.

In the next iteration, we will:

- switch from monoenergetic electrons to charged pions with momenta from ~500MeV to 1GeV, if that is acceptable

- look at adding the vertex constraint

- extrapolate to 30cm so the values can be directly compared to the innermost TPC layer (though that won't directly tell us whether the covariance matrix is properly handled in genfit)

- look at performance with zero INTT layers

- look at adding one or two outer TPC hits (we can at least assemble the framework for this)

Thanks for the comments and suggestions.

-Ross

Best regards

Tony

---

From: Ross Corliss <rcorliss AT mit.edu>
Sent: Thursday, October 4, 2018 7:16 PM
To: christof.roland AT cern.ch
Cc: jhuang AT bnl.gov; Todoroki, Takahito; Carlos Perez; Anthony Frawley; Chiu, Mickey; Sanghoon Lim; Molly Taylor; sphenix-intt-l AT lists.bnl.gov; sphenix-maps-l
Subject: Re: Meeting of the Inner Tracker Task Force VIRTUAL

 

I've uploaded some slides to the indico page ( https://indico.bnl.gov/event/5107/contributions/24213/attachments/20273/27051/TPC_IFC_resolutions_from_INTT.pdf ) showing the resolutions we're getting by extrapolating the Kalman track stubs in the MVTX+INTT out to the inner field cage.  

-Ross

 

On Oct 2, 2018, at 11:10 AM, Christof Roland <christof.roland AT cern.ch> wrote:

Hi Everybody,

since some of us are at the hard probes conference we will make the 

task force meeting scheduled tomorrow virtual. There will not be an actual 

meeting, but please post new material you would like to share on the indico 

page or add it to the google doc so we can look at it offline and discuss by email

https://indico.bnl.gov/event/5107/

https://docs.google.com/document/d/17PJdu94Ydg25V-cTHIYfSbkfojH4ZUEfsfoZPm0RDb0/edit

Cheers

   Christof