耗散各向异性介质中改进的哈密顿驱动最短路径射线追踪
Modified Hamiltonian-driven shortest-path ray tracing in dissipative anisotropic media
AI总结：
本研究将g*-哈密顿公式融入最短路径射线追踪算法，对比三种核函数的表现，发现MEV驱动的算法在耗散各向异性介质的qSV建模中计算稳定性最佳，可有效处理传统方法的缺陷。
AI中文摘要：
增强耗散各向异性介质中的射线追踪颇具挑战性，因为地下介质既存在地质与流变学不连续性，又具有方向相关的衰减特性。传统实射线追踪及复能量-速度公式难以处理qSV尖点、波前三重影和突变不连续性。近期已提出共轭法与最短路径扩展方法来缓解这些问题，但它们依赖近似哈密顿公式，且通常仅在平滑背景模型中测试，此外很少涉及初至折射或波传播与能量耗散的耦合行为。本研究将基于g*-哈密顿的公式融入适用于层状耗散各向异性介质的最短路径射线追踪算法。在局部尺度上，我们首先利用三个核函数计算复能量速度及对应的射线量，即射线速度、射线衰减和射线品质因子。这些核函数以不同方式为所有体波生成复能量速度，分别为原始复能量速度（OEV）、共轭实射线追踪（C-RRT）和g*-哈密顿（MEV）。随后我们分析它们的差异与数值行为，并检验若干优化策略。结果表明，所有核函数对qSH波和qP波的计算结果一致，而OEV在qSV建模中会沿射线路径产生伪反射或数值不稳定性及非物理衰减；C-RRT会产生大误差，包括在部分qP和qSV走时值中出现非物理负衰减，且在我们的参数 regime 下会导致射线品质因子严重偏离真实值。在所有测试的核函数中，MEV驱动的最短路径射线追踪产生的初至行为计算稳定性最佳。
英文摘要：
Enhancing ray tracing in dissipative anisotropic media is challenging because the subsurface exhibits both geologic and rheological discontinuities and directionally dependent attenuation. Conventional real-ray tracing and complex energy-velocity formulations struggle with qSV cusps, wavefront triplications, and abrupt discontinuities. Recent conjugate and shortest-path extensions have been proposed to mitigate these issues, yet they rely on approximate Hamiltonian formulations that are typically tested in smooth background models. In addition, they rarely address first-arrival refraction or the coupled behavior of wave propagation and energy dissipation. Here, the g*-Hamiltonian-based formulation is incorporated into a shortest-path-based ray-tracing algorithm for layered dissipative anisotropic media. At the local scale, we first compute the complex energy velocities and corresponding ray quantities, which are ray velocity, ray attenuation, and ray quality factor, using three kernels. These kernels produce complex energy velocities for all body waves in different fashions, namely, the original complex energy velocity (OEV), conjugate real ray tracing (C-RRT), and g*-Hamiltonian (MEV). We then analyze their discrepancies and numerical behavior and examine several optimization strategies. The results demonstrate that all kernels agree for qSH and qP waves, whereas OEV generates pseudo-reflection or numerical instability and unphysical attenuation along ray paths in qSV modeling. C-RRT produces large errors, including non-physical negative attenuation in some qP and qSV traveltime values, and strongly biased ray quality factors under our parameter regime. Among the tested kernels, MEV-driven shortest-path ray tracing produces the most computationally stable first-arrival behavior.
