Search for the decay

arXiv:2607.25607 · hep-ex · Submitted 2026-07-28 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Experiment. The day in high-energy experiment: detectors, searches and limits.

Kai: We're looking at the paper: "Search for the decay".

Mira: Here's the gist of the paper. The paper presents the first search for B0 → K0S τ+τ− decays using Belle and Belle II data and sets an upper limit on the branching fraction at 90% confidence level.

Kai: Search Methodology The search looks for signal decays in B0B¯0 events produced in asymmetric-energy electron-positron collisions using samples from the Belle and Belle II detectors, comprising 1.16 billion Υ(4S) events In Υ(4S) → B0B¯0 decays, the non-signal B¯0 meson is fully reconstructed in a hadronic channel For the signal B0 meson, τ-lepton decays into final states with a single charged particle are selected The analysis uses a tagging approach where one B meson is fully reconstructed in a hadronic decay mode, and the signal is searched for among the remaining particles The selection requires the beam-constrained mass to satisfy Mbc = q s/(4c 4) − ⃗p∗ B¯0 tag2 /c2 > 5.27 GeV/c2 5.27 GeV/c2.> and the energy difference to satisfy −0.15 < ∆E = E∗ B¯0 tag−√s/2 < 0.10 GeV Pairs of oppositely charged particles assumed to be pions are required to satisfy 450 < M(π+π−) < 550 MeV/c2 The background composition varies depending on the τ+τ− final state, with dominant backgrounds being B → D(→ XSL)lνl decays for the ll category To suppress these backgrounds above, a separate BDT classifier is trained for each of the five categories for both Belle and Belle II The input observables comprise quantities related to the signal side such as the invariant mass of the K0S and t−, M(K0S t−), to characterize backgrounds where the two particles have the same origin (mostly a D− meson) Furthermore, observables like q2rec tend to populate higher values in backgrounds Signal Extraction The signal extraction is performed through a binned maximum likelihood fit to the modified BDT output observable O′ using templates constructed from simulation In each τ+τ− category (denoted as C), three templates are considered: signal events, BB¯ background, and qq¯ background The likelihood function L is a product of Poisson probability density functions P and combines the information from all bins of signal and background samples The branching fraction B(B0 → K0S τ+τ−), free in the fit, sets the normalisation of all signal templates and is extracted from a simultaneous fit across all categories Measurement and Limits The measurement of the branching fraction in the entire allowed q2 region yields B(B0 → K0S τ+τ−) = (1.2 ± 3.3 ± 2.9) × 10−4 The observed

expected: exclusion limit is evaluated at the 90% confidence level yielding B(B0 → K0S τ+τ−) < 8.3

7.4: × 10−4 Systematic Uncertainties The total systematic uncertainty on B(B0 → K0S τ+τ−) is 2.9 This includes contributions from leading B-decay branching fractions, simulation sample size, and lepton ID uncertainties The largest systematic contribution stems from the limited knowledge of the signal-side B composition Combined Result The combination with the recent measurement of B+ → K+τ+τ− determines B(B → Kτ +τ−) < 5.4 × 10−4 at the 90% confidence level The resulting observed

expected: upper limit is B(B → Kτ+τ−) < 5.4

5.0: × 10−4 at the 90% confidence level Pruning redundant reasoning tokens The analysis uses five mutually exclusive final-state categories for τ+τ− decays with similar sensitivities to signal The background composition varies depending on the τ+τ− final state, with dominant backgrounds being B → D(→ XSL)lνl decays for the ll category How the training signal is built The signal extraction is performed through a simultaneous binned maximum likelihood fit to the classifier outputs in all reconstructed τ+τ− categories in the Belle and Belle II datasets The choice of binning for O′ is optimized per category using signal efficiency quantiles, with additional constraints to ensure sufficient event counts per bin Where the gains come from The precision of the measurement is limited by the statistical uncertainty The total systematic uncertainty on B(B0 → K0S τ+τ−) is 2.9 This includes contributions from leading B-decay branching fractions, simulation sample size, and lepton ID uncertainties The largest systematic contribution stems from the limited knowledge of the signal-side B composition Combined Result The combination with the recent measurement of B+ → K+τ+τ− determines B(B → Kτ +τ−) < 5.

Title and authors: Kai: So we've got the title "Search for the decay". It sounds like a straightforward search, but there's a lot of technical machinery behind it.

Mira: Exactly. The authors are trying to find evidence for this B zero to K zero S tau plus tau minus decay by looking at those specific events in their massive datasets >

Lev: And the paper describes the method they used, which involves tagging one B meson fully reconstructed in a hadronic channel and searching for the signal among the remaining particles >

Kai: It’s all about this beam constrained mass condition being greater than five point two seven gigaelectron volts squared over light speed squared, and an energy difference between zero point fifteen and zero point ten gigaelectron volts >

Mira: And they also require the invariant mass of the pion plus pion to be between four hundred fifty and five hundred fifty megaelectron volts per light speed square >

Lev: That’s just setting the boundaries for what they consider a candidate particle pair, so if you were building a detector for this, those constraints are the initial filters you'd apply >

Kai: Now, when we look at the background characterization in this paper, they say the background composition really changes depending on which tau plus tau minus final state is being looked at >

Mira: And they highlight that the dominant backgrounds are actually B zero decays into a D meson decaying into X S L nu L decays for those electron plus and electron minus categories >

Lev: That means their biggest challenge is separating this signal from those D meson contaminations, especially in the two lepton category where things get messy >

Kai: To handle that, they mention training a separate BDT classifier for each of the five tau plus tau minus categories for both Belle and Belle II data >

The paper's summary: Kai: So to summarize what this paper is actually saying, they performed the first search for this decay using their combined Belle and Belle II data sets >

Mira: And the main finding they report is that they observed no evidence for the signal itself in these searches >

Lev: They set an upper limit on the branching fraction for B zero to K zero S tau plus tau minus at a ninety percent confidence level, which is less than eight point three times ten to the minus four >

Kai: That eight point three times ten to the minus four figure is what they found when they looked only at this specific decay channel >

Mira: And because they combined this result with a recent measurement of the B plus zero decay into K plus tau plus tau minus, they can combine them to determine an upper limit for the isospin averaged branching fraction B zero to K tau plus tau minus >

Lev: That combination leads them to set an upper limit on that isospin averaged branching fraction at less than five point four times ten to the minus four at a ninety percent confidence level >

The paper's improvements: Kai: The authors discuss some ways they could improve the search methodology for this decay >

Mira: They suggest a few improvements, starting with how they perform the signal extraction itself. Instead of doing separate fits for each tau plus tau minus category, they propose a binned maximum likelihood fit to classifier outputs across all reconstructed categories simultaneously >

Lev: That would allow them to determine the branching fraction free in the fit rather than having to do separate fits per category >

Kai: And they also talk about how they can combine results from different data sets, specifically combining Belle and Belle II data, while accounting for those correlations between systematic uncertainties common to both analyses >

Mira: So, essentially they are pushing toward a joint likelihood fit that uses a single shared parameter corresponding to the isospin averaged branching fraction B zero to K tau plus tau minus >

Conclusion: Kai: Wrapping up, this paper, "Search for the decay", confirms that in this search space so far, they haven't seen any signal >

Mira: The implication is that the upper limit they set on the branching fraction B zero to K zero S tau plus tau minus is less than eight point three times ten to the minus four at a ninety percent confidence level >

Lev: From an engineering view, what this means for us in terms of real hardware is that we need a signal rate significantly higher than what they observed for this particular decay mode if we want to see any bumps >

Kai: Exactly. It shows where the sensitivity lies and where the next generation of experiments needs to focus their resources >

Mira: And combining this with the B plus zero measurement, they've narrowed down that combined upper limit for B zero to K tau plus tau minus down to less than five point four times ten to the minus four >

Lev: So, even with that combination, we are still talking about a very small branching fraction and the systematic uncertainty of two point nine is still significant >

Kai: It’s a good result for setting constraints on new physics, even if it doesn't find the decay itself yet >

Mira: Yeah, this paper is important because it sets a constraint that helps guide future experiments searching for these rare B decays and what those are really doing out there >

Belle Collaboration · Belle II Collaboration

hep-ex

Submitted: 2026-07-28

Updated: 2026-11-12

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 78/100

The gist: The paper presents the first search for B0 → K0S τ+τ− decays using Belle and Belle II data and sets an upper limit on the branching fraction at 90% confidence level.

Key concepts

Search Methodology
The search involved looking for signal decays in B0B¯0 events from asymmetric electron-positron collisions using data from Belle and Belle II. They selected events based on beam-constrained mass and energy difference criteria, requiring specific particle masses to be pions.
Background Characterisation and Suppression
The background varies depending on the tau+tau final state, with B → D(→ XSL)ℓνℓ decays being dominant in one category. A separate BDT classifier is trained for each of the five categories to suppress these backgrounds by using signal-side observables like M(K0S t−).
Signal Extraction
Signal extraction uses a binned maximum likelihood fit on modified BDT output observables (O') with templates constructed from simulations. This fit simultaneously extracts the branching fraction B(B0 → K0S τ+τ−) and normalizes signal templates across all tau+tau categories.
Systematic Uncertainties
The total systematic uncertainty is 2.9, including factors like leading B-decay branching fractions, simulation sample size, and lepton ID uncertainties. The largest systematic issue is the limited knowledge regarding the composition of the signal-side B meson.

Terminology

Summary

The paper presents the first search for B0 → K0S τ+τ− decays using Belle and Belle II data and sets an upper limit on the branching fraction at 90% confidence level. p. 2

Where the paper says it

We observe no evidence for the signal and set an upper limit on the branching fraction B(B0 → K0 S τ +τ −) < 8. 3 × 10−4 at the 90% confidence level.

Search Methodology

The search looks for signal decays in B0B¯0 events produced in asymmetric-energy electron-positron collisions using samples from the Belle and Belle II detectors, comprising 1.16 billion Υ(4S) events In Υ(4S) → B0B¯0 decays, the non-signal B¯0 meson is fully reconstructed in a hadronic channel For the signal B0 meson, τ-lepton decays into final states with a single charged particle are selected The analysis uses a tagging approach where one B meson is fully reconstructed in a hadronic decay mode, and the signal is searched for among the remaining particles The selection requires the beam-constrained mass to satisfy Mbc = q s/(4c 4) − ⃗p∗ B¯0 tag2 /c2 > 5. p. 2

Where the paper says it

In Υ(4S) → B0B¯0 decays, the non-signal B¯0 meson is fully reconstructed in a hadronic channel. For the signal B0 meson, τ -lepton decays into final states with a single charged particle are selected.

27 GeV/c2 5. p. 4
Where the paper says it

To avoid experimenter bias, all selections are optimised on simulation and finalised before examining the signal region. 2 Detectors and samples The Belle II experiment is located at the SuperKEKB accelerator, which collides 7 GeV electrons and 4 GeV positrons at centre-of-mass (CM) energies near the Υ(4S) resonance [16, 17].

27 GeV/c2.> and the energy difference to satisfy −0. p. 6
Where the paper says it

27 GeV/c2 (3. 1) and the energy difference to satisfy −0.

15 < ∆E = E∗ B¯0 tag−√s/2 < 0. p. 5
Where the paper says it

We select events with Υ(4S) candidates decaying into a pair of neutral B mesons. One of the neutral B mesons, referred to as the tag-side B¯0 meson (B¯0 tag), is formed using the full event interpretation (FEI) algorithm [37].

10 GeV Pairs of oppositely charged particles assumed to be pions are required to satisfy 450 < M(π+π−) < 550 MeV/c2

Background Characterisation and Suppression

The background composition varies depending on the τ+τ− final state, with dominant backgrounds being B → D(→ XSL)lνl decays for the ll category To suppress these backgrounds above, a separate BDT classifier is trained for each of the five categories for both Belle and Belle II The input observables comprise quantities related to the signal side such as the invariant mass of the K0S and t−, M(K0S t−), to characterize backgrounds where the two particles have the same origin (mostly a D− meson) Furthermore, observables like q2rec tend to populate higher values in backgrounds

Signal Extraction

The signal extraction is performed through a binned maximum likelihood fit to the modified BDT output observable O′ using templates constructed from simulation In each τ+τ− category (denoted as C), three templates are considered: signal events, BB¯ background, and qq¯ background The likelihood function L is a product of Poisson probability density functions P and combines the information from all bins of signal and background samples The branching fraction B(B0 → K0S τ+τ−), free in the fit, sets the normalisation of all signal templates and is extracted from a simultaneous fit across all categories

Measurement and Limits

The measurement of the branching fraction in the entire allowed q2 region yields B(B0 → K0S τ+τ−) = (1.2 ± 3.3 ± 2.9) × 10−4 The observed [expected] exclusion limit is evaluated at the 90% confidence level yielding B(B0 → K0S τ+τ−) < 8. p. 2

Where the paper says it

We observe no evidence for the signal and set an upper limit on the branching fraction B(B0 → K0 S τ +τ −) < 8. 3 × 10−4 at the 90% confidence level.

3 [7.4] × 10−4

Systematic Uncertainties

The total systematic uncertainty on B(B0 → K0S τ+τ−) is 2.9 This includes contributions from leading B-decay branching fractions, simulation sample size, and lepton ID uncertainties The largest systematic contribution stems from the limited knowledge of the signal-side B composition

Combined Result

The combination with the recent measurement of B+ → K+τ+τ− determines B(B → Kτ +τ−) < 5.4 × 10−4 at the 90% confidence level The resulting observed [expected] upper limit is B(B → Kτ+τ−) < 5. p. 2

Where the paper says it

3 × 10−4 at the 90% confidence level. Combining this with the recent measurement of the isospin-partner decay B+ → K+τ +τ −, we determine an upper limit B(B → Kτ +τ −) < 5.

4 [5.0] × 10−4 at the 90% confidence level

Pruning redundant reasoning tokens

The analysis uses five mutually exclusive final-state categories for τ+τ− decays with similar sensitivities to signal The background composition varies depending on the τ+τ− final state, with dominant backgrounds being B → D(→ XSL)lνl decays for the ll category

How the training signal is built

The signal extraction is performed through a simultaneous binned maximum likelihood fit to the classifier outputs in all reconstructed τ+τ− categories in the Belle and Belle II datasets The choice of binning for O′ is optimized per category using signal efficiency quantiles, with additional constraints to ensure sufficient event counts per bin

Where the gains come from

The precision of the measurement is limited by the statistical uncertainty The total systematic uncertainty on B(B0 → K0S τ+τ−) is 2.9 This includes contributions from leading B-decay branching fractions, simulation sample size, and lepton ID uncertainties The largest systematic contribution stems from the limited knowledge of the signal-side B composition

The combination with the recent measurement of B+ → K+τ+τ− determines B(B → Kτ +τ−) < 5.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved signal extraction can simultaneously perform a binned maximum likelihood fit to classifier outputs across all reconstructed tau plus tau final state categories in both Belle and Belle II datasets. This allows for a simultaneous determination of the branching fraction B(B0 -> K0 S tau+tau-) free in the fit, rather than performing separate fits per category.

  2. The improved system can perform a joint likelihood fit combining the binned signal-extraction models of two measurements, B0 -> K0 S tau+tau- and B+ -> K+ tau+tau-, using a single shared parameter of interest corresponding to the isospin averaged branching fraction B(B -> Ktau +tau-).

  3. The AI system can accurately measure the branching fraction by combining the results from separate fits to either Belle or Belle II data, accounting for correlations between systematic uncertainties common to both analyses.

Abstract

We present the first search for B 0 to K 0 S τ+τ- decays. We look for signal decays in B 0 B 0 events produced in asymmetric-energy electron-positron collisions. This work uses samples from the Belle and Belle II detectors, comprising 1.16 billion Υ(4S) events. In Υ(4S) to B 0 B bar 0 decays, the non-signal B bar 0 meson is fully reconstructed in a hadronic channel. For the signal B 0 meson, τ-lepton decays into final states with a single charged particle are selected. We observe no evidence for the signal and set an upper limit on the branching fraction B(B 0 to K 0 S τ+τ-) < 8.3 times 10-4 at the 90% confidence level. Combining this with the recent measurement of the isospin-partner decay B+ to K+τ+τ-, we determine an upper limit B(B to Kτ+τ-) < 5.4 times10-4 at the 90% confidence level.

Sources

Related papers