学术文献库

The form of the Lagrangian proposed in Part I of this study has been previously used for obtaining stationary ray paths between two endpoints in isotropic media. We extended it to general anisotropy by replacing the isotropic medium velocity with the ray (group) velocity magnitude which depends on both, the elastic properties at the ray location and the ray direction. This generalization for general anisotropy is not trivial and in this part we further elaborate on the correctness, physical interpretation, and advantages of this original arclength-related Lagrangian. We also study alternative known Lagrangian forms and their relation to the proposed one. We then show that our proposed first-degree homogeneous Lagrangian (with respect to the ray direction vector) leads to the same kinematic ray equations as the alternative Lagrangians representing first- and second-degree homogeneous functions. Using different anisotropic examples, we further validate/demonstrate the correctness of the proposed Lagrangian, analytically (for a canonical case of an ellipsoidal orthorhombic medium) and numerically (including the most general medium scenario: spatially varying triclinic continua). Finally, we analyze the commonly accepted statement that the Hamiltonian and the Lagrangian can be related via a resolvable Legendre transform only if the Lagrangian is a time-related homogeneous function of the second-degree with respect to the vector tangent to the ray. We show that this condition can be bypassed, and a first-degree homogeneous Lagrangian, with a singular Hessian matrix, can be used as well, when adding a fundamental physical constraint which turns to be the Legendre transform itself. In particular, the momentum equation can be solved, establishing, for example, the ray direction, given the slowness vector.