J integral estimation in ABAQUS

Dear all,

I am a Ph.D student. I am trying to calculate the value of the j-integral for a 3D penny shaped crack embeddee in a continuum linear elastic solid using ABAQUS.

the first difficulty that I have met is that j-integral values are negative forr some nodes along the crack front.

I am pretty sure that the value of the normal to the crack front that I have used is correct.

 Do you know some other reason for that?

The second problem is that the J-integral and the SIFs  are strongly path-dependent.......why?Do I have to create an ad-hoc partition of my model?

Thanks very much

Francesca