Problems of spatially varying DMI

[yuanshh]

Dear fidimag team:
How can I add a spatially varying DMI in atomistic simulation ? I have tried to do this with a function like this :

but it returns :

[davidcortesortuno]

Hi, I think you found a bug in the code. Thanks for reporting it. In the fidimag/atomistic/dmi.py function, in the section of the code (line 133) we have

        elif self.dmi_type == 'interfacial':

            clib.compute_dmi_field_interfacial(m,
                                               self.field,
                                               self.mu_s_inv,
                                               self.energy,
                                               self.D,
                                               self.neighbours,
                                               self.n,
                                               self.n_ngbs,
                                               self.n_ngbs_dmi,
                                               self.DMI_vector
                                               )

where self.D must be self._D. I'll commit the changes

[davidcortesortuno]

Hey, I realized DMI doesn't accept spatially varying DMI for interfacial systems, we can implement it maybe next week. Spatial only works for bulk DMI

[yuanshh]

Hi, @davidcortesortuno
Thanks for your reply! I have updated the 'dmi.py' file, but unfortunately it returns another error:

And when I use the bulk type, it also has an error:

[davidcortesortuno]

Hi, I just updated the documentation and DMI classes to allow spatial dependence of the interfacial DMI. Can you build Fidimag again and test if it now works for you?

[yuanshh]

Hi, @davidcortesortuno
I have rebuilt Fidimag and use the function I used before. And it returns:

It is the same as the error in bulk type.

[davidcortesortuno]

Ah, the error is because your DMI function should return 6 values, which are the DMI values for every neighbour (the neighbour list depends on the mesh you use, check the documentation for the DMI function), something like

def D_pos(r):
    x, y, z = r
    if x < L/2:
        return np.zeros(6)
    else:
        return np.ones(6.) * D

However you might also set these values at the boundary. At x = L/2 - dx where dx is a lattice unit, you might need something something like [0, D, 0, 0, 0, 0] since the neighbour at +x should have a D strength DMI. Similarly at x=L/2 + dx you might need something like [D, 0, D, D, D, D].
This is currently how we have implemented spatial varying DMI but we will think if there might be a more optimal way

[yuanshh]

OK，it works.
Thanks a lot for your patiently reply!