Recoil PID

Page maintainer: Caro

This is a working page which will continuously change. The content is not reviewed and is intended for people working on this specific topic.

Back to Recoil Analysis Page

Intro

Charged particle identification can be established with the individual layers of the Recoil detector: Silicon Strip Detector (SSD), Scintillating Fiber Tracker (SFT) and Photon Detector (PD). The main purpose is to separate protons from positively charged pions.

Essentials

PID plane, hard cuts and parent distributions

Energy deposition versus reconstructed momentum for one (out of nine) layers of the Recoil Detector. This example is for the inner SSD layer from 07c data. The regions in the PID plane populated by different particle species are indicated.

For a given track in the Recoil Detector, up to 9 layers (2 SSD, 4 SFT and 3 PD) can deliver an independent piece of information on the particle type, looking at the momentum reconstructed by SSD and SFT and the energy deposition in the respective layer (see PID plane on the right hand side).

Hard PID cuts can be performed separately on each of the layers by defining an appropriate separation function between positive pions and protons. However, in order to minimize contamination and maximize statistics, it is desirable to combine all available subdetector responses for a given track.

A clean data sample in one of the Recoil sublayers can be defined by performing hard PID cuts on the detector layers other than the one considered. By doing so, parent distributions can be extracted for the detector layer under consideration, i.e. the detector response for a given particle type with certain momentum.

The official Recoil parent distributions were extracted under the following conditions:

From real data
Only M7 tracks, pion hypothesis
Only single tracks with 4 spacepoints (1111)
Separately for each quadrant
Correction for incident angle
Momentum binning: The SFT parent distributions have 10 bins in the momentum range from 0 to 1 GeV/c, the width of each bin is 100 MeV/c. As there is very few statistics in the first two bins (0 ~ 200 MeV/c), the parent distribution from the third bin (200 ~ 300 MeV/c) is simply used for the first two bins. For SSD, the momentum range 0~1GeV/c is divided into 17 bins: the first bin is 0~200MeV/c, the left range is equally divided with bin-width 50 MeV/c. For SSD, the parent distribution in each momentum bin is extracted separately.

Probabilistic (Bayesian) PID algorithm

In analogy to what has been carried out for the front spectrometer in the past, the advanced identification of pions and protons in the Recoil detector is based on a Bayesian algorithm that uses conditional probabilities.

For an explanation of the algorithm, see

rdPID values

The Bayesian algorithm yields a PID value for each detector layer, labeled:

rdPID1	inner SSD
rdPID2	outer SSD
rdPID3	inner SFT, parallel layer
rdPID4	inner SFT, stereo layer
rdPID5	outer SFT, parallel layer
rdPID6	outer SFT, stereo layer
rdPID7	PD layer A
rdPID8	PD layer B
rdPID9	PD layer C

which means that the clusters are the basic rdPID units.

The rdPID values in the uDST tables

For the 07c0 and later uDST productions, the rdPID values are stored in g1RDPID.
For each track reconstructed with method 7 and having positive charge, the pion hypothesis is fed into the recoil PID library to calculate the individual PID values.
For tracks with only 2 spacepoints in the SSD (none in SFT), there is no pion hypothesis provided and thus no PID value available; the track is assumed to be a proton.
The entry g1RDTrack.g1RDPID links to the table g1RDPID holding the PID values. Be careful: In C, the arrays start counting at 0.

Example: rdPID values in 07c1

Combined rdPID distributions Fehler beim Parsen (Konvertierungsfehler. Der Server („https://wikimedia.org/api/rest_“) hat berichtet: „Cannot get mml. Server problem.“): {\displaystyle \left.\mathrm {rdPID} =\sum _{i}{\mathrm {rdPID} _{i}}\right.} for Recoil quadrants 1, 2, 3, 4 (from left to right) and different track topologies (only tracks with p < 700 MeV/c)
rdPID>0 selects protons, rdPID<0 positively charged pions

The following plots show the momentum dependence of the rdPID distributions in bins of each 100 MeV/c wide from p = 100 MeV/c to 1 GeV/c. (Quadrant 3)

The rdPIDlib and Parent Distributions

As for hrc front spectrometer data, there are no rdPID values available in hrc/xtc type data. One can get access to rdPID values by linking ones user analysis code to an external library, the rdPIDlib. This library contains the latest parent distributions of the SSD and SFT (and PD).

/group02/rcoilgrp/Software/rdPIDLIB

Parent distributions (in subdirectory of corresponding production):

/group02/rcoilgrp/Software/rdPIDLIB/PDs

Copy of source code:

/group02/rcoilgrp/Software/rdPIDLIB/src

How to use rdPIDlib in a C-code:
A detailed example on how to setup rdPIDlib and gather required track and spacepoint information in your code can be found here:

/group02/rcoilgrp/Software/rdPIDLIB/hanna_example

To use this library link your program against libRDPID.a and set the environment variable RDPIDLIB to the directory in which the data files are located. In case you want to use the rdPIDLIB library within Hanna++, please the also set the environment variable RDPIDLIBBASE to the base directory of the library. In case of the public installation this is /group02/rcoilgrp/Software/rdPIDLIB.

Note that the rdPIDlib can not be used with uDST data because not all pieces of information necessary to feed into the rdPIDlib are available. Spacepoint and cluster information are not stored in the uDSTs in order to limit written output.

More Info

Documentation

Documentation by X. Lu on Recoil PID
See also rdPIDlib

Ongoing Studies

Last update: 7.5.2009

Check of PID algorithm (method of hard cuts to produce parent distributions) on MC data. see [RDPID-006]
PID contamination and efficiency in dependence of PIDcut obtained from "PEPSI-challenge" (input: one true MC type and check output of hard-cut method). see [RDPID-008]
Basic MC-data comparison (delta(E) in bins of momentum: widths compatible?), is also useful for the general understanding of the MC. see [RDPID-007]
How to use the rdPID values in an analysis?
- Track selection based on PID? What if 5 layers say proton and 1 layer says pion?
- How to add up $\left.\sum _{i}\omega _{i}\mathrm {PID} _{i}\right.$ $\left.\sum _{i}\omega _{i}\mathrm {PID} _{i}\right.$ ?
  - Give different weights Fehler beim Parsen (MathML mit SVG- oder PNG-Rückgriff (empfohlen für moderne Browser und Barrierefreiheitswerkzeuge): Ungültige Antwort („Math extension cannot connect to Restbase.“) von Server „https://wikimedia.org/api/rest_v1/“:): {\displaystyle \left.\omega_i\right.} to different layers?
  - What happens if there is no information in a layer?
Fluxes:
- How important are they? (have not yet been calculated)
- Depend on MC generator / physics process (compare DISng, Pythia, gmcDVCS, ...)

Talks and Reports on Recoil PID

[RDPID-010] X. Lu, The HERMES Recoil Detector: Particle Identification and Determination of Detector Efficiency of the Scintallating Fiber Tracker (Diploma Thesis September 2009)
[RDPID-009] RDPID for 06e, X. Lu, TRM 5.5.2009
[RDPID-008] Study on Recoil PID based on 06e and 07c, including contamination, efficiency, 3SPs, quadrant variation and suggestion for PID cut, X. Lu, Talk given at Recoil Meeting 22.4.2009
[RDPID-007] Data-MC (DISng) comparison, X. Lu, RDPID meeting 14.4.2009
[RDPID-006] Study of Recoil PID on MC data, X. Lu, Talk given at Recoil Meeting 13.2.2009, including contamination and efficiency
[RDPID-005] SFT PID, W. Yu, Talk given at Recoil Meeting 1.8.2008
[RDPID-004] Particle Identification with SSD for 07c data, X. Lu, Talk given at Recoil Meeting 1.8.2008
[RDPID-003] Correction for incident angle, X. Lu, Report 14.5.2008
[RDPID-002] Recoil Detector Studies on Particle Identification, C. Riedl, Nov. 2007 (Internal Note 07-019)
[RDPID-001] Data Analysis with the Recoil Detector -- Separation of Pions and Protons, S. Sorger, Summer Student report 2007

PID with the Front Spectrometer