Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Galaxy Activity, Torus, and Outflow Survey (GATOS): N. Unveiling physical processes in local active galaxies. Unsupervised hierarchical clustering of JWST MIRI/MRS observations

L. Hermosa Muñoz, J. R. González Fernández, A. Alonso-Herrero, I. García-Bernete, O. González-Martín, M. Pereira-Santaella, E. López-Rodríguez, C. Ramos Almeida, S. García-Burillo, L. Zhang, A. Audibert, E. Bellochi, F. Combes, T. Díaz-Santos, D. Esparza-Arredondo, B. García-Lorenzo, M. García-Marín, E. K. S. Hicks, Á. Labiano, N. A. Levenson, M. Martínez-Paredes, C. Packham, R. A. Riffel, D. Rigopoulou, J. Schneider, M. Villar-Martín

Abstract

With the rise of the integral field spectroscopy, we are currently dealing with large amounts of spatially resolved data, whose analysis has become challenging, especially when observing complex objects such as nearby galaxies. We aim to develop a method to automatically separate different physical regions within the central parts (1"~160 pc, on average) of galaxies. This can allow us to better understand the systems, and provide an initial characterisation of the main ionisation sources affecting its evolution. We have developed an unsupervised hierarchical clustering algorithm to analyse data cubes based on spectral similarity. It clusters together spaxels with similar spectra, which is useful to disentangle between different physical processes. We have applied this method to a sample of 15 nearby (distances <100 Mpc) galaxies, 7 from the Galaxy Activity, Torus, and Outflow Survey (GATOS) and 8 archival sources, all observed with the medium resolution spectrometer (MRS) of the Mid-Infrared Instrument (MIRI) on board of the JWST. From the clusters, we computed their median spectrum and measured the line and continuum properties. We used these measurements to train random forest models and create several empirical mid-IR diagnostic diagrams for the MRS channel 3 wavelength range, including among others the bright [Ne II], [Ne III], and [Ne V] lines, several H2 transitions, and PAH features. The clustering technique allows to differentiate emission coming from an AGN, the disc, and star forming regions in galaxies, and other composite regions, potentially ionised by several sources simultaneously. This is supported by the results from the empirical diagnostic diagrams, that are indeed able to separate physically distinct regions. This innovative method serves as a tool to identify regions of interest in any data cube prior to an in-depth analysis of the sources. [abridged]

The Galaxy Activity, Torus, and Outflow Survey (GATOS): N. Unveiling physical processes in local active galaxies. Unsupervised hierarchical clustering of JWST MIRI/MRS observations

Abstract

With the rise of the integral field spectroscopy, we are currently dealing with large amounts of spatially resolved data, whose analysis has become challenging, especially when observing complex objects such as nearby galaxies. We aim to develop a method to automatically separate different physical regions within the central parts (1"~160 pc, on average) of galaxies. This can allow us to better understand the systems, and provide an initial characterisation of the main ionisation sources affecting its evolution. We have developed an unsupervised hierarchical clustering algorithm to analyse data cubes based on spectral similarity. It clusters together spaxels with similar spectra, which is useful to disentangle between different physical processes. We have applied this method to a sample of 15 nearby (distances <100 Mpc) galaxies, 7 from the Galaxy Activity, Torus, and Outflow Survey (GATOS) and 8 archival sources, all observed with the medium resolution spectrometer (MRS) of the Mid-Infrared Instrument (MIRI) on board of the JWST. From the clusters, we computed their median spectrum and measured the line and continuum properties. We used these measurements to train random forest models and create several empirical mid-IR diagnostic diagrams for the MRS channel 3 wavelength range, including among others the bright [Ne II], [Ne III], and [Ne V] lines, several H2 transitions, and PAH features. The clustering technique allows to differentiate emission coming from an AGN, the disc, and star forming regions in galaxies, and other composite regions, potentially ionised by several sources simultaneously. This is supported by the results from the empirical diagnostic diagrams, that are indeed able to separate physically distinct regions. This innovative method serves as a tool to identify regions of interest in any data cube prior to an in-depth analysis of the sources. [abridged]

Paper Structure

This paper contains 20 sections, 23 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Data and methodology
  3. Data sample
  4. Unsupervised hierarchical clustering technique
  5. Random forest classifier
  6. Results
  7. Main properties of the clustering maps
  8. Median spectrum per cluster
  9. Line ratios per cluster
  10. Discussion
  11. The importance of the wavelength range: ch3-short vs ch3-all
  12. Caveats of the methodology
  13. Clustering technique
  14. Automatic classification of the ionising source of the clusters
  15. Identifying the ionising source of the clusters
  16. ...and 5 more sections

Figures (23)

  • Figure 1: Flowchart of the methodology discussed in Sect. \ref{['Sect2:Data']}.
  • Figure 2: Clustering results of the ch3-short cube (top) and the complete ch3 channel cube (bottom) for NGC 7172. We show in the left panels the median flux in logarithm scale of the ch3-short and ch3-all cube (top and bottom, respectively; see Appendix \ref{['Appendix_MedianMaps']}). Middle panels show the cluster maps, while right panels show the median spectra per cluster in logarithm scale, normalised to the total integrated flux (see Sect. \ref{['SubSect2:Clustering']}). The maps are centred in the original observed position (north is up and east to the left). The white star indicates the photometric centre. We assigned the same colours to the clusters and their respective spectrum. Colours are calculated automatically by dividing the 'jet' palette in matplotlib. We note that both the cluster colours and numbering are arbitrary, have no physical meaning, and are assigned independently in the top and bottom panels. We mark with dashed, vertical, gray lines the main emission lines, and with grey bands the PAH features in the spectrum. The wavelength is in rest frame.
  • Figure 3: Same as Fig. \ref{['Fig:Cluster_NGC7172']} but for NGC 5728. We note that, for the top, right panel, we do not show the spectrum for cluster 8, as it is a low S/N cluster.
  • Figure 4: Histogram of the average, relative importance of the features measured in the ch3-all cubes obtained from the automatic classification of the clusters (see Sect. \ref{['SubSect3:Results_Ionisation']}). The errorbars are estimated as the standard deviation of all the importances for each feature calculated using Monte Carlo simulations (n$=$1000, see Sect. \ref{['SubSect2:RFtecnhique']}).
  • Figure 5: Diagnostic diagrams based on the best preferred line ratios using the ch3-all cubes, in a logarithmic scale (see Fig. \ref{['Fig:HistogramRatiosImportance']} and details in Sect. \ref{['SubSect3:Results_Ionisation']}). Each point is a cluster from the galaxies used as the training sample, colour-coded by their assigned class probability (AGN in orange, SF in green, and Other in blue; see details in Sect. \ref{['SubSect3:Results_Ionisation']}), with darker colours indicating a higher probability. The initial training labels of the clusters (see Table \ref{['Table:2']}) were obtained from previous detailed JWST MIRI/MRS analysis of the sources (see references in Table \ref{['Table:1']}, and details in Sect. \ref{['SubSect2:RFtecnhique']}). Contours show the kernel density estimate (KDE) of the distribution for each class at four probability levels: 0.5, 0.6, 0.75, and 0.9.
  • ...and 18 more figures