β衰变寿命是原子核的基本物理性质之一，在原子核物理与核天体物理中都起着重要作用。本文利用贝叶斯神经网络（Bayesian Neural Network，BNN）方法预测原子核的β衰变寿命及其不确定性。研究发现，在神经网络输入层引入β衰变能以及与原子核对效应相关的物理量，并采用β衰变寿命的对数作为网络输出，可以显著提高学习精度。对于寿命小于1 s的原子核，其预言精度约为0.2个数量级，取得了与BNN方法学习β衰变寿命的实验值和理论值的对数之差相当的精度。当外推至未知核区，其预言的β衰变寿命在误差范围内与其他理论模型的结果符合较好，尤其是对于Z?50的原子核。

Mechanical exfoliation (ME) and chemical vapor deposition (CVD) MoS₂ monolayers have been extensively studied, but the large differences of nonlinear optical performance between them have never been clarified. Here, we prepared MoS₂ monolayers using ME and CVD methods and investigated the two-photon absorption (TPA) response and its saturation. We found that the TPA coefficient of the ME monolayer was about (1.88 ± 0.21) × 10³ cm/GW, nearly two times that of the CVD one at (1.04 ± 0.15) × 10³ cm/GW. Furthermore, we simulated and compared the TPA-induced optical pulse modulation in multilayer cascaded structures, which is instructive and meaningful for the design of optical devices such as a beam shaper and optical limiter.

Questions hovering over the modulation of bandgap size and excitonic effect on nonlinear absorption in two-dimensional transition metal dichalcogenides (TMDCs) have restricted their application in micro/nano optical modulator, optical switching, and beam shaping devices. Here, degenerate two-photon absorption (TPA) in the near-infrared region was studied experimentally in mechanically exfoliated MoS2 from single layer to multilayer. The layer-dependent TPA coefficients were significantly modulated by the detuning of the excitonic dark state (2p). The shift of the quasiparticle bandgap and the decreasing of exciton binding energy with layers were deduced, combined with the non-hydrogen model of excitons in TMDCs and the scaling rule of semiconductors. Our work clearly demonstrates the layer modulation of nonlinear absorption in TMDCs and provides support for layer-dependent nonlinear optical devices, such as optical limiters and optical switches.