AI 中文总结
针对scNMSSM中h1h1 -> 4a1 -> 8τ信号，利用事件级分类器（AUC 0.941）优于单喷注分类（AUC 0.759），有效区分于标准模型本底。
AI 中文摘要
在半约束NMSSM中，当λ较大且tan(beta)较小时，如果最轻的赝标量介于b-bbar阈值之下，双希格斯产生可通过h1h1 -> 4a1 -> 8τ过程进行。每个a1衰变产生的τ对发生准直并在单个喷注内合并。以先前工作中确定的基准点m_a1 = 8.2 GeV为例，我们考察哪些观测量能够将这些事件与产生类似末态的SM过程区分开来。信号和本底样本使用MadGraph5_aMC@NLO、Pythia 8和Delphes 3模拟，大半径喷注重建参数R = 0.8。通常用于双叉标记的子结构观测量在此失效：喷注质量和N-subjettiness比值tau_21相对于增强的Z -> tau tau本底，其区分能力分别仅为0.02和0.09。两者均为双τ喷注，且τ衰变产生的中微子将重建质量移动到接近18 GeV的共同值，尽管母粒子质量相差一个数量级。区分能力转而来自整个事件。基于单个胖喷注训练的增强决策树AUC达到0.759 ± 0.002，而事件级分类器达到0.941 ± 0.001，其性能由增强物体的多重性和次主导物体的质量驱动。真值匹配确认这些喷注在77%的事件中对应真实的a1衰变。顶夸克对产生重现了多物体结构，相对于ttbar -> bbbar τ τ本底，AUC降至0.855 ± 0.008。对于两个模拟本底的截面加权组合，分类器达到0.908 ± 0.002，在10%本底效率下信号效率为71%。这些是快速模拟下的分类指标，而非发现显著性：可约本底、触发接受度和堆积效应未包含在内。
英文摘要
In the semi-constrained NMSSM at large lambda and low tan(beta), di-Higgs production can proceed through h1h1 -> 4a1 -> 8tau when the lightest pseudoscalar lies below the b-bbar threshold. The tau pairs from each a1 decay are collimated and merge within a single jet. Using the benchmark point m_a1 = 8.2 GeV identified in earlier work as a representative signal topology, we examine which observables separate such events from Standard Model processes producing similar final states. The exotic branching fraction of this point exceeds present experimental limits, and all results concern the shape of the distributions, not the signal rate. The substructure observables normally used for two-prong tagging prove ineffective here: the jet mass and the N-subjettiness ratio tau_21 have separation powers of 0.02 and 0.09 against boosted Z -> tau tau. Both are ditau jets, and the neutrinos from tau decays displace the reconstructed masses to a common value near 18 GeV despite parent masses differing by an order of magnitude. Discrimination comes instead from the event as a whole. A boosted decision tree trained on single fat jets reaches an AUC of 0.759 +- 0.002, while an event-level classifier reaches 0.941 +- 0.001, driven by the multiplicity of boosted objects and the masses of the subleading ones. Top-quark pair production reproduces the multi-object structure, and against ttbar -> bbbar tau tau the AUC falls to 0.855 +- 0.008. For a cross-section-weighted combination of the two simulated backgrounds the classifier reaches 0.908 +- 0.002, with 71% signal efficiency at 10% background efficiency. These are classification metrics from fast simulation, not discovery significances.
Comments19 pages, 3 figures, 7 tables; v2: added discussion of experimental constraints on the benchmark point
