Thank you for you prompt reply.
In both models, I didn’t specify the special_bonds command.
In fix shake model, bonds and angles are constrained by fix shake command. In the fix rigid model, bonds and angles are not specified in data file, since I think they are unnecessary for rigid bodies.
The thermo output for ebond and eanlge are all zero during the simulations. evdwl on step 0 are almost the same (9503.941 kcal/mol vs. 9503.9436 kcal/mol).