Flavour-physics benchmarks for tracker-based particle identification at the FCC-ee

TL;DR

The study evaluates whether FCC-ee flavour physics can rely on tracker-based particle identification from CLD and IDEA detectors instead of dedicated PID subsystems. By training boosted decision trees on simulated particle-gun samples and using inputs such as $dE/dx$, time-of-flight, and drift-chamber $dN/dx$, it benchmarks PID performance across key flavour observables, including $b$-flavour tagging, rare decays, and $s$-jet tagging, over a range of timing resolutions. The results show substantial background suppression in many cases, especially at low momenta with moderate ToF ($\sim$30–$50\, ext{ps}$) and with IDEA benefiting from drift-chamber cluster-counting; however, in high-momentum regimes $dN/dx$ remains the dominant discriminator and $dE/dx$ can be less helpful. Overall, tracker-based PID can reach meaningful sensitivity improvements for several observables, while dedicated PID detectors may offer additional gains in specific scenarios, guiding detector-design decisions and highlighting directions for future work.

Abstract

The correct identification of charged hadrons plays a crucial role in flavour-physics measurements. The final detector configurations at the proposed FCC-ee are yet to be determined and this study aims to contribute to this discussion by benchmarking the particle identification (PID) performance of the proposed CLD and IDEA detectors using fully simulated events. At present, neither detector proposal includes dedicated PID systems, relying instead on information from the tracking subsystems. We estimate the expected level of contamination due to misidentified charged hadrons for $b$-flavour tagging, rare $b\to s$ transitions, and $s$-jet tagging. The PID information provided by silicon trackers, namely time-of-flight and energy-deposit measurements, leads to significant background suppression with high signal efficiency for the low-momentum hadrons considered for same-side $b$-flavour tagging. In order to improve the contamination in rare decays where momenta are in the medium range, only good timing resolution of 30ps and below can yield an improvement of one order of magnitude below the level achieved by kinematic criteria alone. Light-quark jet-flavour tagging requires identification of particles with very large momentum, which is not possible using only time-of-flight or energy-deposit information in silicon. Access to the number of clusters in a drift-chamber setup, as proposed for the IDEA detector however, results in strong background suppression in every case. This suppression can be further improved in some scenarios by time-of-flight resolution of 30-50ps or better. The PID quality generally exhibits only a small dependence on the cluster-counting efficiency. Whether dedicated PID detectors could further enhance flavour-physics sensitivity should be the subject of future study.

Figures (26)

• Figure 1: Input quantities for particle identification using CLD tracking detectors. Top: $\mathrm{d}\!E\!/\!\mathrm{d} x$ in the silicon sensors, calculated as the harmonic mean of all hits in silicon trackers. Bottom: speed measured using the time-of-flight and flight-distance between the first and the last hit in silicon trackers.
• Figure 2: Input quantities for particle identification using IDEA tracking detectors. Top: $\mathrm{d}\!E\!/\!\mathrm{d} x$ in the silicon sensors, calculated as the harmonic mean of all hits in silicon trackers. Middle: speed measured using the time-of-flight and flight-distance between the first and the last hit in silicon trackers. Bottom: $\mathrm{d}\!N\!/\!\mathrm{d} x$ interpolated from Garfield simulation (black line) with Poissonian variation at the generation stage (dots).
• Figure 3: Generator-level distributions of the charged-hadron momenta in the samples used to study specific flavour measurements and the particle-gun samples employed for training the classifiers.
• Figure 4: (Left/right) ROC for the (top) pion, (middle) kaon, and (bottom) proton identification at CLD/IDEA. The colours indicate the time-of-flight (ToF) resolution and the linestyle indicates whether the silicon-based $\mathrm{d}\!E\!/\!\mathrm{d} x$ information was used. A cluster-counting efficiency $\varepsilon_\text{cc}$ of 80% is assumed for the drift-chamber reconstruction at IDEA.
• Figure 5: (Left/right) AUC in different momentum ranges for the (top) pion, (middle) kaon, and (bottom) proton identification at CLD/IDEA. The colours indicate the time-of-flight (ToF) resolution and the linestyle indicates whether the silicon-based $\mathrm{d}\!E\!/\!\mathrm{d} x$ information was used. A cluster-counting efficiency $\varepsilon_\text{cc}$ of 80% is assumed for the drift-chamber reconstruction at IDEA.