IF runs classifiers inconsistent

[maxbriel]

I'm currently comparing Initial-Final runs with Nearest Neighbour runs for interactions on the HMS-HMS grid. Specifically, I'm looking at the stable "reverse" mass transfer interaction. I've selected systems that produce a high spin + interp_class_HMS_HMS=stable_reverse_MT. In the efficiency plot below you can see there's a significant difference between the interpolation and the nearest neighbour runs.

When looking at the grid slices at $Z=0.1Z_\odot$, there's no stable reverse mass transfer where the systems are (black triangles in bottom plots). At other $q$ slices, there is stable reverse mass transfer where the models are, but I'm not sure it is correct, especially since there are also systems that go through this interaction, while they shouldn't even have a stable reverse mass transfer interaction (black triangles in the "unstable contact phase" area).

The culprit seems to be that the interp_class_HMS_HMS and mt_history_HMS_HMS are not consistent with one another. interp_class_HMS_HMS=stable_reverse_MT contains other mt_history_HMS_HMS, such as 'stable mass transfer postMS' and 'no RLOF'.

q=0.85 q=0.95

[mkruckow]

Please, show all neighboring slices. Otherwise we can't estimate the 3D effects. E.g. you black triangles, which don't match might be close to the next q slice, which has the stable reverse MT at a different area in M1 and P. Maybe the best way would be to only plot the stable MT symbols, but for two neighboring q slices and over plot the pop-syn systems in-between those two. I'm a bit confused when your first plot shows there should be no stable reverse MT in the slice of 0.1Zsun (NN) but your bottom plots show. I guess, this is related to the fact, that you additionally selected by high spin, which comes from an independent interpolator then the reverse MT classification. I think, we need a deeper investigation on the data, to get a solid ground on what is actually wrong and what are plotting/selection effects.

[maxbriel]

Let me try to clarify it a bit. I am performing an additional selection on effectively the rotation rate (m_disk_radiated actually).

There are 2 things at play:

1. There are systems with different mt_history_HMS_HMS and interp_class_HMS_HMS that should not be there. stable_reverse_MT should not give no_RLOF as a mt_history. I've attached the other slices below for Z=0.1Zsun. The values are not in regions near stable_reverse_MT in nearby grid slices. I've also checked if the right metallicity populations are plotted.
2. Now the 2nd component: IF interpolator is returning values where there might be high spins in a region of stable_reverse_MT. This is something I'm less worried about. Although the models do not contain these high spins, this might be a resolution thing and the IFinterpolator finds them correctly. That is a different investigation and more of an interpolator verification, but that's not the issue I wanted to address here.

[mkruckow]

OK, thanks.

1. Yes, your examples are in areas, where the neighboring slices don't show reverse MT either. (FYI: there are some cases, e.g. in the q=0.8 slice which fall under the category of having stable MT in the q=0.85 slice.) So, as there are low numbers of data points for this class, we should check how the convective hull is determined in those cases, @philipp-rajah .
2. I made this point, because most black triangles are at places with reverse MT, hence removing the ones in the contact area won't change the efficiency plot much, because the numbers of systems will change by a factor <<2, which is tiny on a logarithmic scale.