Heaviside Enrichment

I have been looking at XFEM literature for a while and wondering why do we need 2 different heaviside enrichment functions in XFEM? In paper titled "A finite element method for crack growth without remeshing" by Nicolas Moes et al.(1999), they use the following function for heaviside enrichment:

 H(x,y)={1, for y>0

             -1,for y<0}