Bad line profile in rt_line_one

[rjrw]

The default line profile used in rt_line_one doesn't look anything like a Voigt profile.

A more accurate implementation can now be switched on by changing the branch used in function getVoigt -- it needs to be verified that this version is correct, and the impact on run time assessed.

[rjrw]

I introduced variable names to the code at r4358 which explain the issues.

The alternative branch will slow the code; Peter says he can do something better.

[gary]

Robin's plot of the profile

[gary]

code in question is rt_line_one.cpp:395

[peter]

In r4619, r4620, and r4621 several (wrapper) routines were added that should help in calculating Voigt profiles.

VoigtU() calculates a Voigt profile U(a,v) using the normalization in Eq. 9-45 of Mihalas.

VoigtH() calculates a Voigt profile H(a,v) using the normalization in Eq. 9-44 of Mihalas.

VoigtU0() calculates U(a,0), i.e. the central depth of the profile.

VoigtH0() calculates H(a,0).

The last two routines are faster and more accurate than the generic Voigt routine, but obviously only work at the line center.

[peter]

In r4624 a fast version of VoigtH() was added, called FastVoigtH().

[peter]

In r4627 FastVoigtH() was tuned further and some comments were added regarding its accuracy.

[peter]

In r4630 a higher order term was added to FastVoigtH() to increase its accuracy and extend the region where it can be applied to higher a values.

[peter]

A bug in r4624 was fixed in r4695.

[peter]

In r4703 the unit tests were fixed and int_ex2() was renamed to dawson().