基于卷积神经网络预测南海海底地形

针对南海区域,使用3种重力信号(垂线偏差、重力异常、垂直重力梯度异常)训练卷积神经网络模型,并将预测结果与船测数据和国外模型结果进行对比分析。将3种重力信号分成4组数据：重力异常,重力异常与垂直重力梯度异常,重力异常与垂线偏差,以及重力异常、垂线偏差和垂直重力梯度异常。4种组合方式的反演结果与船测水深之间的标准差分别为104.780 m、102.778 m、93.788 m、88.289 m,表明随着不同类型重力数据的加入,水深预测精度明显提高,并且在深度大于2 000 m时,反演结果精度提升效果更为显著。将训练集占总数据集的比例分别设置为80%、70%、60%和50%,反演结果与船测水深之间的标准差分别为88.289 m、91.256 m、92.833 m、96.022 m,表明数据量的增多可以有效提高模型学习结果的精度。     

Strain effects on optical polarisation properties in (11-22) plane GaN films

We present the theoretical results of the electronic band structure of wurtzite GaN films under biaxial strains in the (11^-22)-plane. The calculations are performed by the κ• p perturbation theory approach through using the effective-mass Hamiltonian for an arbitrary direction. The results show that the transition energies decrease with the biaxial strains changing from -0.5% to 0.5%. For films of (11^-22)-plane, the strains are expected to be anisotropic in the growth plane. Such anisotropic strains give rise to valence band mixing which results in dramatic change in optical polarisation property. The strain can also result in optical polarisation switching phenomena. Finally, we discuss the applications of these properties to the (11^-22) plane GaN-based light-emitting diode and lase diode.