Stellarator divide by 0 warning

[timothy-nunn]

There is a divide by 0 warning where the tprofile has not been set in Stellarator runs, first reported as a possible cause of #2961.

[timothy-nunn]

/Users/timothynunn/process/process/impurity_radiation.py:445: RuntimeWarning: divide by zero encountered in log
  np.log(teprofile),

The warning arises from the last element of profile_y being 0. I suspect this is because the profile domain is normalised to [0, 1] and that the behaviour at x=1 is undefined, and it should be very slightly below 1. @ym1906 does this make sense? Could we just normalise the profile_x by saying profile_x := profile_x / max(profile_x) + 1 rather than just by the max(profile_x)? Or is the issue a bit more nuanced than that?

[grmtrkngtn]

The last point of the profile is 0 because of the parabolic analytic formula, it should be zero even though this isn't physically meaningful (undefined behaviour).

I'm not sure I understand how your suggestion would fix the issue? An alternative might be to add a small floor to the number.

[timothy-nunn]

Hi @ym1906 thanks for the reply. My suggestion is just to stop profile_x having a 1 entry which should stop profile_y from going to 0 (since it happens when profile_x = 1). When you say a small floor you mean modify elements of profile_y to be profile_y = np.max(profile_y, 1e-8)?

[timothy-nunn]

Sorry @ym1906 I will bump this because it probably got lost of Xmas

[grmtrkngtn]

Yes, I meant have a very small value for the last element. What do you think?