Mismatch between precomputed opacity table and dsharp_opac.get_opacities() output

[doikiyoaki]

Hi Til,

Merry Christmas!

I’m writing because I have trouble reproducing the dust opacity shown in Fig. 4 of Birnstiel et al. (2018).

As I understand it, the package provides precomputed opacity tables as default_opacities(_smoothed).npz, and these tables were calculated in smoothed_opacities.ipynb using dsharp_opac.get_opacities().

When I load default_opacities(_smoothed).npz and size-average it, I reproduce the plot in Fig. 4. However, when I recompute the opacities locally using dsharp_opac.get_opacities(), I do not recover the same result. In other words, the provided opacity table and the output of dsharp_opac.get_opacities() are not consistent on my side.

For cross-checking, I also computed the DSHARP opacity using optool, and that result is consistent with the opacity recomputed by dsharp_opac.get_opacities().

I am attaching my script and the results (opacity_1mm_no_size_averaging.pdf, opacity_1mm_size_averaged.pdf, beta_size_averaged.pdf). The blue curve shows the opacity from the precomputed table, the orange curve shows the newly computed opacity, and the green curve shows the result from optool. Also, the size averaging is performed with q = 3.5.
The precomputed tables and recomputed ones match at small grain sizes, while they deviate at large grain sizes.

I would like to hear your ideas about the reason for this discrepancy, or if I am making any mistakes in using this package. One possibility I considered is that the refractive index (n,k) used by the code may have been updated since the precomputed tables were generated, but I could not verify this.

Best,
Kiyoaki

beta_size_averaged.pdf
opacity_1mm_no_size_averaging.pdf
opacity_1mm_size_averaged.pdf

compare_opacity.ipynb

[birnstiel]

Dear Kiyoaki,

Thanks for checking. Yes, I got aware of this earlier last month. I have also rerun the smoothed opacities before the Christmas break, but didn't find the time to check the results yet or find where the differences come from. There have been a few changes to the code over the years including the extrapolation and size averaging bug you found. But if I understand you correctly, the differences exist already before any size averaging?

Which commit are you using?

I compared current to old downloads of the (n,k) values, but they are identical.

[doikiyoaki]

Dear Til,

Thanks for your reply, and sorry for the late reply. I’ve just returned from vacation in Japan.

But if I understand you correctly, the differences exist already before any size averaging?

Yes, the non–size-averaged dust opacities already differ, so the issue does not seem to come from the size-averaging step. That’s why I initially suspected that the input (n,k) data might have changed, but since you confirmed the (n,k) values are identical, I’m not sure what is causing the discrepancy.

Which commit are you using?

I’m not sure about the commit ID, but I’m using v1.1.11 installed from PyPI.

Best wishes,
Kiyoaki