Carbon Flux with DAMPE Using Machine Learning Methods

DAMPE space-borne cosmic ray experiment has been collecting data since December 2015. Many high-impact results on the ion, electron and photon fluxes were obtained. This submission presents the carbon flux analysis with DAMPE using machine learning techniques. The readout electronics would saturate at energy deposits above several TeV in a single BGO bar of the DAMPE calorimeter. The total energy loss per event due to saturation can sometimes reach over a hundred TeV. We present a convolutional neural network model which can accurately recover the energy lost due to saturation and thus significantly increase the dynamic range of DAMPE. Another machine learning model combines the resolution of the hodoscopic BGO calorimeter and the high-resolution tracker of DAMPE to provide the best possible prediction of the direction of the incoming particle. This allows measuring charges at energies up to several hundred TeV. In this work, we present the application of these methods to carbon flux analysis.

暗物质粒子探测卫星（DAMPE）的空间宇宙射线实验自2015年12月起便一直在收集数据。在离子、电子和光子通量方面取得了许多具有重大影响力的成果。本论文利用机器学习技术，介绍了使用DAMPE对碳通量的分析。当能量沉积在DAMPE量能器的单个BGO晶体棒中超过数TeV时，读出电子学系统会发生饱和。由于饱和，每个事例的总能量损失有时会超过100 TeV。我们提出了一种卷积神经网络模型，它能够准确恢复因饱和而损失的能量，从而显著扩大DAMPE的动态范围。另一种机器学习模型结合了DAMPE的分层BGO量能器的分辨率和高分辨率径迹探测器，以尽可能精确地预测入射粒子的方向。这使得在高达数百TeV的能量下测量电荷成为可能。在这项工作中，我们介绍了这些方法在碳通量分析中的应用。