Detailed analysis of multi-line molecular distributions in the Seyfert galaxy NGC 1068: Possible effect of the AGN outflow to the starburst ring

Hiroma Okubo, Toshiki Saito, Shuro Takano, Nario Kuno, Akio Taniguchi, Taku Nakajima, Nanase Harada, Ken Mawatari

Published 2025-06-19Version 1

We apply principal component analysis (PCA) to the integrated intensity maps of 13 molecular lines of the nearby type-2 Seyfert galaxy NGC 1068 obtained by Atacama Large Millimeter/sub-millimeter Array (ALMA) to objectively visualize the features of its center, (1) within a radius of about 2 kpc ($\sim$ 27".5; hereafter the "overall region") and (2) the ring shaped starburst region between 750 pc ($\sim$ 10") and 2 kpc ($\sim$ 27".5) of the galaxy (hereafter the "SB ring region"). PCA is a powerful unsupervised machine learning technique that extracts key information through dimensionality reduction. The PCA results for the overall region have a possibility to reconstruct a map representing the approximate H$_2$ column density and difference of volume density and/or chemical composition between the circumnuclear disk (CND) and the starburst ring (SB ring). Additionally, the PCA results for the SB ring region have a possibility to reconstruct a map representing the approximate H$_2$ column density and distinction between starburst dominated region and shock dominated region. Furthermore, the PCA results for the SB ring region indicate a possible interaction between the Active Galactic Nucleus (AGN) outflow and gas in the SB ring. Although further investigation is required, we suggest that the AGN outflow interacts with gas in the SB ring, as this feature is consistent with the direction of the AGN outflow and is contributed by CN, C$_2$H and HCN, which are known to be enhanced by the AGN outflow. These results demonstrate that PCA can effectively extract features even for galaxies with complex structures, such as AGN + SB ring. This study also implies that PCA has the potential to uncover previously unrecognized phenomena by visualizing latent structures in multi-line data.