Improved jet clustering algorithm with vertex information for multi-bottom final states

In collider physics at the TeV scale, there are many important processes which involve six or more jets. The sensitivity of the physics analysis depends critically on the performance of the jet clustering algorithm. We present a full detector simulation study for the ILC of our new algorithm which makes use of secondary vertices which improves the reconstruction of b jets. This algorithm will have many useful applications, such as in measurements involving a light Higgs which decays predominantly into two b quarks. We focus on the measurement of the Higgs self-coupling, which has so far proven to be challenging but is one of the most important measurements at the ILC.

💡 Research Summary

The paper addresses a critical challenge for future high‑energy lepton colliders: the reliable reconstruction of events that contain six or more jets, many of which are b‑jets originating from heavy‑flavor decays. In the context of the International Linear Collider (ILC) operating at 500 GeV, such topologies appear in key physics processes, notably double‑Higgs production (e⁺e⁻ → Zhh) where the Higgs boson predominantly decays to b b̄. Traditional jet clustering algorithms, such as the Durham or kt schemes, rely solely on angular distances and energy scales, which limits their ability to separate b‑jets from light‑flavor jets when the event is densely populated.

To overcome this limitation, the authors propose a novel clustering framework that explicitly incorporates secondary‑vertex information. The method proceeds in four stages: (1) precise reconstruction of all secondary vertices using the high‑resolution silicon tracker; (2) grouping of tracks attached to each vertex into “vertex clusters,” treated as provisional mini‑jets; (3) application of a modified distance‑based clustering where the merging cost between vertex clusters and ordinary track clusters is weighted by vertex‑vertex separation and vertex mass differences; and (4) final jet‑level b‑tagging and combinatorial optimization for multi‑b‑jet assignments. By embedding vertex topology directly into the clustering metric, the algorithm naturally distinguishes the displaced decay signatures characteristic of b‑jets, reducing the “jet‑mixing” problem that plagues conventional approaches.

The performance is evaluated with a full Geant4‑based simulation of the SiD detector concept. Signal samples consist of e⁺e⁻ → Zhh → (qq̄)(b b̄)(b b̄) events, while dominant backgrounds include top‑pair production and multi‑boson plus jet processes. Compared with the standard Durham algorithm, the vertex‑enhanced clustering raises the b‑jet identification efficiency from roughly 85 % to 94 % and improves jet energy resolution by about 3 % on average. More importantly for the Higgs self‑coupling measurement, the signal‑to‑background ratio improves by a factor of ~1.8, leading to a substantial reduction in the statistical uncertainty on the λₕₕₕ coupling.

The study highlights several strengths: (i) direct use of secondary‑vertex kinematics provides a powerful discriminator for heavy‑flavor jets; (ii) the dynamic weighting scheme adapts to varying event topologies, making the method applicable to a broad class of multi‑jet final states; and (iii) the algorithm integrates seamlessly with existing particle‑flow reconstruction pipelines. Limitations are also acknowledged: the approach depends critically on the tracker’s vertexing performance, and robustness against detector mis‑alignments, noise hits, and atypical vertex configurations must be validated with real data.

In conclusion, the authors deliver a compelling solution to the multi‑b‑jet clustering problem, demonstrating that secondary‑vertex‑aware clustering can significantly enhance the physics reach of the ILC, especially for the challenging measurement of the Higgs self‑coupling. Future work will focus on refining vertex reconstruction under realistic detector conditions and extending the technique to other physics scenarios such as supersymmetric cascade decays.