Is it true that the first step must be elastic definition in VUMAT?
And how to choose a initial young's modulus and poison's ratio?
I tried several initial young's modulus, some works some don' t. And it seems when it is not defined elasticly at the first step, there always errors.
Could I know why?
No, the first step does not have to be purely elastic. You can use the same equations and solution approach for the first step as in any other step.
If you don't use a purely elastic first step then ABAQUS will print a warning, but the simulation will continue without any problems and the results will be just fine.