[Halld-physics] PID & Vertex Reconstruction

[Paul Mattione]

Yes, I don't see a way around it.  

 - Paul

On Sep 16, 2011, at 5:52 PM, Elton Smith wrote:

> Hi Paul,
> 
> Your proposal sounds very reasonable to me. As I understand it, it also 
> assumes that the start counter time resolution is sufficient to 
> determine the RF beam bucket for a given track.
> 
> Cheers, Elton.
> 
> Elton Smith
> Jefferson Lab MS 12H5
> 12000 Jefferson Ave
> Suite #16
> Newport News, VA 23606
> (757) 269-7625
> (757) 269-6331 fax
> 
> 
> On 9/16/11 1:45 PM, Paul Mattione wrote:
>> To continue our discussion from the physics meeting on Monday, I'd like to propose a different method of doing PID&  vertex reconstruction of charged tracks (below).  Please let me know what you think, or if you have any questions.
>> 
>> I propose that PID&  vertex reconstruction be performed in several steps:
>>  - Calculate PID chi-squares from track reconstruction, drift chamber dE/dx, and TOF dE/dx.
>>  - Group the tracks into different physics events based on which RF beam bunches their SC hit-times match with.
>>  - For each physics event, reconstruct the positions of the vertices using the track hypotheses with the best FOM from the chi-squares determined so far.
>>  - Calculate the timing chi-squared for each hypothesis by projecting the RF time and BCAL/FCAL/TOF hit-time to the reconstructed vertices and comparing (although any point on the track would do).
>>  - Perform PID: calculate the PID FOM by first making the chi-squares from the different sources have equal weight.
>>  - Calculate the reconstructed vertex time from the average of the track times.
>> 
>> The tricky part is selecting the corresponding RF beam bunch for each track, since a track with two different PID hypotheses may match with two different beam bunches (especially if it is slow, small-angled, or doesn't have an SC hit).  Also, there are ambiguities in using the photon time to select the RF beam bunch during increased luminosity running, and/or if some of the photons were untagged (counter inefficiency, or photon energy outside of tagging range).  These problems are only for a small percentage of tracks/events, but its important to take them into account.
>> 
>> I think the best way to select the RF beam bunches for each track is to first project the SC time from each track hypothesis to its POCA to the beamline, and determine the closest RF bunch for each.  Then these matched RF bunches can be compared with the RF bunches that match the tagger photons.  Of the RF bunches that photons were matched to, the one with the most matching tracks would be selected.  Then every track with at least one hypothesis matching to it would be grouped together with that RF beam bunch.  This would be repeated until no tracks are remaining.
>> 
>> This should work unless the true photon wasn't tagged.  If there's a large number of tracks in the event (3+), most of them correspond to a single RF beam bunch, and none of the photons matched it, then the true photon probably wasn't tagged.  In this case, then the RF beam bunch with the majority of the tracks would be selected.  I don't want to go into every case in this email, but some algorithm like this can be implemented.
>> 
>> Since there are still going to be some backgrounds from mis-identification, I think that ultimately the best way of doing PID is by testing how well the tracks match the specific reaction that you are studying.  For example, for the Lambda->p, pi- decay you can force your proton&  pi- candidates have a common spacetime vertex, and then remove the false candidates with chi-squared and/or 3-sigma cuts (etc.).  For complicated reactions, I think it's probably best that the PID&  vertexing results from the track reconstruction code be regarded as a "first iteration" that in some cases can be improved upon during physics analysis.
>> 
>>  - Paul
>> 
>> 
>> _______________________________________________
>> Halld-physics mailing list
>> Halld-physics at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/halld-physics
> _______________________________________________
> Halld-physics mailing list
> Halld-physics at jlab.org
> https://mailman.jlab.org/mailman/listinfo/halld-physics