If the tensile load is normal to the edge crack, you already have pure Mode I. There is nothing to create share stresses along the crack front line. Perhaps you want to get rid of the T-stress, but this is unavoidable in your geometry.

Dear Andrey,

Thanks for your response. The loading is normal to the crack edge, but due to the loading condition we can also have mode II SIF. One solution could be applying analytical solution of mode I over the problem boundaries which needs to add some features to the our existing code to be applicable. I need another solution that deals only with loading or nodal boundary condition types to have pure mode I. It seems it's not possible to have the second solution in a numerical code.

Dear Mohamad,

i applied the analytical displacements for mode I in my xfem simulations. It was easy in Matlab. 

Bests,

p

Dear Vinh Phu,

I know the general scheme. But, since our code is not well prepared for this, I need some help on the implementation. Is it possible to access your XFEM codes?

You can probably utilize the idea of superposition and Green's functions, where the SIF of a cracked body uses stress from the uncracked body as the input to a problem with tractions applied directly to the cracked surfaces-see Bueckner and Rice.  In your case, you could use the uncracked shear stresses and apply them in the opposite direction to the crack on your actual model.  This should effectively cancel the mode II SIF.  Your boundary conditions would remain the same, and only new surface traction would be applied to your FEM model, assuming linear elastic material.

Well, I'll try to follow your solution Brian. Thanks.