The advantage of the finite element method (FEM) lies in its flexibility in addressing rugged interfaces in complex geological models. However, the efficiency of the FEM is relatively low for large-scale seismic wave modelling. Here, we introduce an order-corrected symplectic FEM (OCSFEM) with structure-preserving properties and parsimonious memory requirements for the elastic wave equation. In this method, the storage of the large sparse stiffness matrix is changed to the storage of the element Jacobian matrix. An efficient order-corrected symplectic method with third-order temporal accuracy is combined with a triangle-based FEM to construct the OCSFEM. The structure-preserving characteristics and high efficiency of the OCSFEM facilitate the high-fidelity modelling of large-scale and long-term wave phenomena. Complex and large-scale numerical examples show that the OCSFEM exhibits low numerical dispersion and high stability compared with conventional methods, such as the second-order symplectic FEM.

有限元方法（FEM）的优势在于它在解决复杂地质模型中粗糙界面时的灵活性。但是，对于大规模地震波建模，FEM的效率相对较低。在这里，我们介绍一种具有结构保留特性和弹性波方程的简约记忆要求的阶校正辛有限元（OCSFEM）。在这种方法中，将大的稀疏刚度矩阵的存储更改为元素雅可比矩阵的存储。一种有效的具有三阶时间精度的阶数校正辛方法与基于三角形的有限元法相结合，构造了OCSFEM。OCSFEM的结构保留特性和高效率有助于对大规模和长期波动现象进行高保真建模。