自旋轨道耦合（SOC）在磷光、系间窜跃等自旋禁阻的激发态衰减过程中起着关键作用。然而，高效、准确地计算磷光、热活化延迟荧光（TADF）等材料的旋轨耦合常数仍面临挑战。其主要困难是高精度的多参考态方法通常计算量很大，而含时密度泛函理论（TDDFT）虽然具有计算量小、速度快的优点，但其缺点是精度需要系统地标定。目前，不同程序中实现了基于TDDFT计算SOC矩阵元的几种不同方案，为了理解这些计算方案的准确性和差别，本文以2SeU分子光物理过程为例，通过以多参考态方法计算结果作为参考，比较了PySOC、ORCA、Dalton及BDF程序中基于TDDFT计算SOC常数的方案，系统地考察了如下不同之处带来的影响：（1）旋轨耦合算符（平均场和有效电荷近似）、（2）TDDFT和TDA方法、（3）TDDFT下不同计算方案（TDA公式、运动方程、二次响应方法）。得到的主要结论包括以下几点：（1）同样采用TDA方法时，各程序相对于BDF程序计算结果的平均绝对误差最大仅为2.1cm^-1，说明对该体系不同SOC算符的影响可以忽略。（2）TDDFT方法下，PySOC和ORCA计算SOC矩阵元时均利用TDA公式且归一化条件相同，最终可以得到一致的SOC常数。（3）TDDFT方法下，Dalton和BDF程序计算基态-激发态间SOC时，二者SOC矩阵元的计算均采用线性响应理论且归一化条件相同，最终的SOC结果一致。但与PySOC和ORCA的结果存在差别，主要由于后者使用了TDA的计算公式。（4）TDDFT方法下，对于激发态-激发态间SOC的计算，Dalton采用二次响应理论计算SOC矩阵元，BDF程序可以采用运动方程公式或二次响应理论来计算。BDF采用运动方程公式的结果与PySOC和ORCA的结果比较接近，而BDF和Dalton均采用二次响应理论的结果在有些结构下差别较大，其原因可能是对于发散项的处理不同，该问题还需后续更细致的研究。

Spin-orbit coupling (SOC) plays a key role in spin-forbidden processes of excited states such as phosphorescence and intersystem crossing. However, it is still a great challenge to efficiently and accurately compute spin-orbit coupling constants for phosphorescence and thermally activated delayed fluorescence (TADF) materials. The main difficulty is that high-level multi-reference methods usually require a large amount of computational cost, while the time-dependent density functional theory (TDDFT) method is more efficient, but less accurate. The accuracy of TDDFT needs to be calibrated systematically before actual applications. Currently, different schemes for calculating SOC matrix elements based on TDDFT have been implemented in different programs. In order to understand the accuracy and difference of different schemes, the photophysical process of the 2SeU molecule was taken as an example in this work and the results obtained from multi-reference methods were used as a reference for comparison. The schemes for calculating SOC constants based on TDDFT in PySOC, ORCA, Dalton and BDF programs were compared and the impacts of the following differences were investigated systematically: (1) spin-orbit coupling operators (mean field approximation and effective charge approximation), (2) TDDFT and Tamm-Dancoff approximation (TDA), (3) different calculation schemes within TDDFT (TDA formula, equation of motion, quadratic response methods). The main conclusions were as follows: (1) With TDA, the mean absolute error (MAE) of the SOC matrix elements obtained from different programs relative to the results of the BDF program was at most 2.1cm^-1, indicating that the influence of the different SOC operators on the system can be neglected. (2) At the TDDFT level, both PySOC and ORCA used the TDA formula to calculate the SOC matrix elements with the same normalization conditions, such that the results were very similar. (3) For SOC between the ground state and excited states, both the Dalton and BDF programs used the formula obtained from linear response theory, such that the obtained SOC results were consistent and different from the results obtained from PySOC and ORCA. (4) For the SOC matrix elements between excited states, Dalton used quadratic response theory and the BDF program used the equation of motion (EOM) method or quadratic response theory. The EOM results of BDF were close to those of PySOC and ORCA, while the results of BDF and Dalton using quadratic response theory were different at some structures. The reason could be that divergent terms were treated differently, and further studies on this issue need to be carried out in future.