Elliptic crack instead of line crack in Phase-field method

What should be the crack initial condition in phase field fracture simulation?

By using phi=0 (broken phase) for one line segment of nodes  crack can
be specified. but with the loading applied we expect some crack opening
displacement. now as there is only one node specifying both upper and
lower crack faces displacements +u and -u at that node will be zero. 

My opinion which could be wrong on this point is that, one should take
elliptical crack to distinguish crack upper and bottom surfaces. Image showing phi=0 by red coloured nodes.
I am not sure whether crack surface which is diffuse is given by phi=0 or phi=1.