Add interesting sources to high-dimensional visualizations

[sebastian-schindler]

Indicate interesting sources in high-dimensional visualization plots like corner plot or UMAP plot. It is already interesting to see where in the phase space they lie (maybe they are all clumped up together, or far away from the rest in a small cluster).

Definitely interesting: NGC 1068. Would expect it to be outside the main cluster, because it is very close by and thus absolutely brighter.

Interesting sources identified previously:

• TXS 0506+056
• NGC 1068
• NGC 1320
• NGC 3227
• NGC 3786
• NGC 4151
• NGC 4235
• NGC 5290
• NGC 5506

The NGC sources are Seyferts from the Neronov, Savchenko, Semikoz paper.

• [ ] Allow marking specific points in high-dimensional visualizations, possibly with a label.
• [ ] Mark these sources in the plots.
• [ ] Identify further interesting sources.

[sebastian-schindler]

The IDs in SOIs.pkl file written previously are 1RXS identifiers, but current catalog has only 2RXS identifiers (or AllWISE or others). So some identifier of these sources included in the current catalog need to be figured out.

• [x] Check which identifiers are available in the catalog, and find these identifiers for the interesting sources.

[sebastian-schindler]

The available identifiers are for the...

• intermediate catalog: AllWISE ID (all entries), 2RXS/XMMSL2 ID (depending on detection)
• final catalog: AllWISE ID (all entries), field name containing the 2RXS or XMMSL2 ID, depending on detection

Therefore, the AllWISE identifier would be the best option, also because most objects will have a detection by WISE.

[sebastian-schindler]

In Sarah's final catalog the AllWISE identifiers are formatted in a strange way. For example, the following ID is displayed in the notebook as follows:

• J235256.39-203209.4 displayed as b"b'J235256.39-203209.4 '"

The first b and set of quotation marks is expected, as the data is saved as byte sequences, which are like this marked in python. Therefore, the following would have been expected to be displayed:

• b"J235256.39-203209.4"

However, the inner set of quotation marks including the b are part of the string itself and thus superfluous, as is the trailing whitespace. The values can be corrected by this:

lambda x: x[2:-1].strip()

The values remain as a byte sequence.

[sebastian-schindler]

With the above described preparations, the interesting sources can be searched in the catalog. It turns out that neither in the final nor the intermediate catalog any of the sources are contained. Is this expected or plausible?

One of the constraints of the catalog is detection of a source in x-rays. Is it plausible that none of the interesting sources are x-ray bright (enough)?

• [ ] #6

[sebastian-schindler]

All interesting sources do have an x-ray detection in 2RXS, as checked by querying SIMBAD/VizieR. Thus, the requirement of an x-ray detection is no reason that these sources are excluded from the catalog.

[sebastian-schindler]

After checking the cuts applied in Sarah's catalog, the following was identified as the cause that removes all interesting sources:

This plot shows the position of the interesting sources in the same plane:

The cut line removes the entire lower population, which contains all interesting sources.

What is puzzling, is that all interesting sources are Seyfert galaxies, which are AGN; but the cuts that remove them are supposedly designed to select AGN. It appears that these cuts are too strict, because they remove a lot of interesting AGN.

• [ ] Can the catalog even be used, given its strict cuts that remove many interesting AGN?

[sebastian-schindler]

The cut was developed by Salvato et al. based on another survey at lower x-ray energies that separates AGN from stars and galaxies, and extended to higher x-ray energies to be applied to 2RXS data. They mention an efficiency of AGN selection of 99 % and impurities of < 0.1 % regarding stars.

• [ ] Is it just coincidence that the interesting sources are cut away, even though the AGN selection is supposed to be very efficient?

[sebastian-schindler]

After talking with Markus Ackermann from DESY, it became apparent that the AGN cut by Salvato et al. does indeed remove NGC 1068 and other interesting sources. According to him, the reason is the high obscuredness of NGC 1068, i.e. the high column density of dust along the line of sight to the black hole. This causes strong absorption of the x-ray flux, while the IR flux is affected less, putting NGC 1068 further left in the plot than it should be. Would it have a higher x-ray flux, it would be further to the right, and thus potentially above the cut line.

This means that the used cut is not good at distinguishing AGN when it comes to highly obscured AGN. This is a problem, as we do want to keep also obscured AGN, and especially NGC 1068 and similar sources.

[sebastian-schindler]

However, this does not explain some of the other Seyferts being cut: From the Seyferts shown here, only NGC 1320 also has a high column-density according to the paper, and some are actually only Seyfert 1 galaxies (NGC 3227, NGC 4151, NGC 4235). The argument of overly attenuated x-rays compared to IR doesn't work for these sources, and thus does not explain why they are cut away.