Improving the Direct Determination of $|V_{ts}|$ using Deep Learning

An $s$-jet tagging approach to determine the Cabibbo-Kobayashi-Maskawa matrix component $|V_{ts}|$ directly in the dileptonic final state events of the top pair production in proton-proton collisions has been previously studied by measuring the branching fraction of the decay of one of the top quarks by $t \to sW$. The main challenge is improving the discrimination performance between strange jets from top decays and other jets. This study proposes novel jet discriminators, called DISAJA, using a Transformer-based deep learning method. The first model, DISAJA-H, utilizes multi-domain inputs (jets, leptons, and missing transverse momentum). An additional model, DISAJA-L, further improves the setup by using lower-level jet constituent information, rather than the high-level clustered information. DISAJA-L is a novel model that combines low-level jet constituent analysis with event classification using multi-domain inputs. The model performance is evaluated via a CMS-like LHC Run 2 fast simulation by comparing various statistical test results to those from a Transformer-based jet classifier which considers only the individual jets. This study shows that the DISAJA models have significant performance gains over the individual jet classifier, and we show the potential of the measurement during Run 3 of the LHC and the HL-LHC.

💡 Research Summary

This paper presents a significant advancement in the precision measurement of the CKM matrix element $|V_{ts}|$ through the introduction of a novel deep learning framework called DISAJA. The fundamental challenge addressed in this research is the difficulty of “s-jet tagging”—the ability to distinguish strange jets, which originate from the $t \to sW$ decay process, from other light-flavor jets in the dileptonic final states of top-pair production during proton-proton collisions.

The researchers propose a Transformer-based architecture to overcome the limitations of traditional jet classifiers. The study introduces two distinct models: DISAJA-H and DISAJA-L. The DISAJA-H model utilizes a multi-domain approach, incorporating high-level features from various event components, including jets, leptons, and missing transverse momentum ($E_T^{miss}$). By leveraging the attention mechanism of the Transformer, DISAJA-H can capture the complex correlations and global topology of the entire event, rather than focusing solely on the jet’s properties.

Building upon this, the DISAJA-L model introduces an even more granular approach by utilizing low-level jet constituent information. Instead of relying on pre-clustered high-level features, DISAJA-L analyzes the individual particles that constitute a jet. This allows the model to extract subtle physical signatures from the particle-level distribution, effectively combining low-level constituent analysis with high-level event classification. This hybrid approach represents a significant leap in the ability to identify flavor-specific jets.

The performance of these models was evaluated using a CMS-like LHC Run 2 fast simulation. The results demonstrate that the DISAJA models achieve significant performance gains compared to standard Transformer-based classifiers that only consider individual jets. The study proves that integrating multi-domain event information and low-level constituent data is crucial for improving discrimination power.

The implications of this research are profound for the future of high-energy physics. As the LHC enters Run 3 and moves toward the High-Luminosity LHC (HL-LHC) era, the ability to perform high-precision measurements of CKM elements will be vital for searching for physics beyond the Standard Model. The DISAJA framework provides a robust and scalable methodology for enhancing the sensitivity of flavor-tagging algorithms, paving the way for more accurate fundamental physics discoveries in the coming decade.