Particle Identification in Belle II Silicon Vertex Detector

We report a particle identification (PID) method developed for charged pions, kaons, and protons using specific ionization information in the silicon-strip vertex detector (SVD) of Belle II with D∗+ → D[→ K−π+]π+ and Λ → pπ− decay samples. The study is based on e+e− collision data recorded at the Υ (4S) resonance by the Belle II detector. The introduction of additional information from the SVD is found to improve the overall PID performance in the low-momentum region.


Introduction
Identification of charged particles such as pions, kaons, and protons is important to the physics program of the Belle II experiment [1]. Belle II has an excellent particle identification (PID) system comprising three main subdetectors, the central drift chamber (CDC), time-of-propagation counter, and aerogel ring-imaging Cherenkov counter. Low-momentum charged particles having a transverse momentum p T < 65 MeV/c are unable to reach the CDC, owing to their highly curved trajectories. Our goal is to exploit specific ionization (dE/dx) [2] by these low-momentum particles in the silicon-strip vertex detector (SVD) towards identifying them. Even if the particles have a p T greater than 65 MeV/c and are thus able to reach the CDC, the dE/dx values measured in the SVD can provide complementary information to that obtained from the main PID subdetectors [3].
The study is based on e + e − collision data recorded at the Υ (4S) resonance by the Belle II detector. We use relatively clean samples of D * + → D 0 (K − π + )π + and Λ → pπ decays to first obtain the SVD dE/dx calibration for pions, kaons, and protons. Later, we check the impact of dE/dx information on overall PID performance using the same decay channels. To assess the impact of SVD dE/dx information on the overall PID performance, we plot the identification efficiency and fake rate as a function of momentum applying a requirement on the binary PID likelihood L(i/j) > 0.5. The efficiency i is defined as the ratio of the number of charged particle tracks identified with PID requirement under the particle hypothesis i and the number of charged particle tracks identified kinematically under the hypothesis i. The fake rate (f j→i ) is the ratio of the number of charged particle tracks identified with PID requirement under the hypothesis i and the number of charged particle tracks identified kinematically under the hypothesis j.

SVD dE/dx calibration
The D * + → D 0 (→ K − π + )π + decay is used to calibrate the pion and kaon PIDs based on dE/dx information in SVD. We require the charged particle tracks to have a transverse (longitudinal) impact parameter less than 0.5 cm (2.0 cm). These tracks must have at least one SVD hit and a track-fit χ 2 probability value greater than 10 −5 . To further purify the sample, we require the reconstructed D 0 mass to lie between 1.85 and 1.88 GeV/c 2 , corresponding to a ±3σ window around the nominal D 0 mass. The reconstructed D * mass must be within 1.95 and 2.05 GeV/c 2 . We apply a loose criterion on kaon and pion PID likelihoods, calculated without SVD information, to remove low-momentum secondary pions and kaons produced due to hadronic interaction in the detector material. We model the signal and background shape in the D * -D 0 mass difference (∆m) by a sum of two Gaussian functions with a common mean and a threshold function, respectively. The s Plot [4] technique is used to subtract the residual background contributions. The Λ → pπ decay is used to calibrate the proton PID based on dE/dx information in SVD. We require the reconstructed pπ invariant mass of Λ candidates to be in the range [1.10, 1.13] GeV/c 2 , and they are further subjected to a vertex fit. To remove the random combination of two tracks, the distance between the interaction point and the vertex of the Λ candidates is required to be greater than 1.0 cm and the vertex-fit χ 2 probability must be greater than 10 −3 . We also require at least one SVD hit for both daughter tracks. We suppress the contamination of charged pions coming from the K 0 S decay by rejecting events that have the M π + π − value in the range [488, 508] MeV/c 2 , corresponding to a ±3σ window around the nominal K 0 S mass. Similarly, events with electrons from converted photons are suppressed by excluding M e + e − < 50 MeV/c 2 . We impose an additional requirement of at least one CDC hit and a loose criterion on the proton PID calculated without SVD information to remove low-momentum secondary pions produced due to hadronic interaction with the detector material. We model the signal shape in M pπ with the sum of a Gaussian and two asymmetric Gaussian functions of a common mean and the background shape with a second-order Chebyshev polynomial. Again the s Plot [4] technique is used to subtract the residual background contributions. The fitted distributions of ∆m from the D * sample and M pπ from the Λ sample are shown in Fig. 1. As shown in Fig. 2, the two-dimensional distributions of dE/dx vs. momentum shows a clear separation between different particles in the low momentum region. These background subtracted two-dimensional histograms are used as probability density functions for various particle hypotheses and uploaded to the calibration database.

PID performance
To assess the impact of the SVD dE/dx information to the overall PID, we use a separate set of data sample processed including the PID information from SVD. We study the efficiency and fake rate as a function of momentum by varying the PID likelihood L(i/j). The PID likelihood criterion is varied from 0 to 1 in order to produce these plots. The efficiency vs. fake rate distributions shown in Fig. 3 confirm the improvement in PID performance by adding the SVD dE/dx information. The data-MC difference in performance arises due to imperfect simulation of the cluster energy distribution for which the work is underway.
Nonetheless, our study confirms that for a given fake rate the addition of dE/dx information improves the efficiency in the low-momentum region.

Conclusion
We have developed a PID method for charged pions, kaons, and protons using energy loss information in Belle II SVD with D * + → D 0 (→ K − π + )π + and Λ → pπ − decay samples. The study tells that adding the SVD information improves the overall PID performance in the low-momentum region.