Photometry and kinematics of extragalactic star-forming complexes

1 Introduction

Our current understanding of the star formation process is mostly based on information obtained within our Galaxy. A detailed description of all the major galactic star-forming regions and complexes can be found in Reipurth (2008). However, it is difficult to get a complete understanding of star formation using only data from our Galaxy since they are relevant for a limited range of physical parameters. However, now we are able to explore star formation in detail in other galaxies. Studying extragalactic active star-forming regions (star-forming complexes [SFCs]) is helpful to fill the gaps in observable parameters of stars and to understand the process of star formation in other physical environments.

This includes comparison of SFC observational data in different lines and photometric bands. There were many studies devoted to CO and HI (21 cm) emission from SFCs (see Heyer and Dame 2015, Fukui and Kawamura 2010 for reviews). In the work of Kennicutt and Evans (2012), dust emission and H α emission were also considered.

In the present work, we focus on “dust” bands: 24, 160, and 8 μ m . It is widely assumed that emission in 24 μ m band is hot dust emission, 160 μ m is cold dust emission, and 8 μ m is polycyclic aromatic hydrocarbon emission (Draine and Li 2007). We also examine correlations between fluxes in dust bands and CO emission.

CO emission and dust emission are tracers of interstellar matter, or, more precisely, of dense clouds. These clouds are birthplaces of stars, so one can associate CO and dust emission with star formation process (Calzetti et al. 2007).

We developed a procedure to relate emission from different spectrum bands in a selected area and to decide if this area can be considered as a SFC candidate. This procedure also allows us to detect cases where an SFC candidate is distinguishable in dust bands, but not visible in CO and HI emission lines. This information might help clarify some details of star formation models.

This work is organized as follows. In Section 2, we describe our data selection procedure along with a method of candidate SFC identification. In Section 3, photometry of SFCs is performed and correlations between emission in different bands are revealed. In Section 4, we discuss the kinematics of SFCs. The conclusions are given in Section 5.

2 Galaxy selection and method of SFC identification

For our study, we need galaxies, where separate SFCs can be easily identified. Given the angular resolution of available surveys, we require 1 kpc regions to have an angular size of several arcseconds. We also want to consider galaxies with active ongoing star formation. This implies the following selection criteria.

1. Distance is less than 15 Mpc, which is required to examine 1 kpc SFCs with angular resolution of ∼ 1 arcsecond.

2. The galaxy is of a spiral or irregular morphological type.

3. Inclination of the galaxy symmetry plane is less than 45°.

4. The galaxy has been observed at 8 and 160 μ m (data from “Spitzer,” Werner et al. 2004, and “Herschel,” Pilbratt et al. 2010) and in the near UV band (data from GALEX telescope, Conti et al. 2011).

In Section 4, we will use CO-spectral surveys PHANGS (Leroy et al. 2021) and HERACLES (Leroy et al. 2009). These surveys provide information about CO emission (2–1 transition) with high angular resolution (PHANGS) and low angular resolution (HERACLES).

Using these criteria, we selected 17 galaxies from the HyperLeda database (Makarov et al. 2014). For details see Table 1. We use the most recent available data.

We detected SFCs in these galaxies, matching “Spitzer” and “Herschel” data. When maps in different spectral bands are matched, it is important to ensure they have the same resolution. For this purpose, all used images were convolved to the worst of angular resolution of all used images. Comparison of images in 8 μ m (Figure 1) shows that some information is lost during convolution, and a complex identified on a low-resolution map may actually encompass several star formation complexes. The Spitzer images were convolved to the Herschel point spread function, using kernels from Aniano et al. (2011).

Figure 1

Comparison between unconvolved (on the left) and convolved (on the right) images of NGC628 in 8 μ m band.

After that we pick out from 20 to 200 complex candidates in each galaxy, using circular apertures, and overlap different emission maps.

For a given pair of regions, let D 12 be the distance between centers of these regions and r ¯ be the average radius of regions. Overlapping regions were classified as belonging to one of the four types (Figure 2).

1. A close type corresponds to the case when a single region on the 160 μ m map overlaps with a single region on the 8 μ m map, and the distance between region centers, D 12 , is less than the average radius of regions, r ¯ .

2. A cross type corresponds to the case when two regions (at 160 μ m and at 8 μ m ) overlap, but D 12 > r ¯ .

3. An empty type implies that there is no region on the 8 μ m map that overlaps with a region on the 160 μ m map.

4. An irregular type encompasses all other cases.

Figure 2

Red color corresponds to 160 μ m regions and apertures, green color corresponds to 8 μ m ones. A number in the bottom-right corner of each picture indicates the type of the region according to our notation (see text).

We assume that an SFC is reliably detected if it belongs to the close type. A stricter criterion may reduce a sample to statistically insignificant one, but a softer criterion may lead to selection of numerous false SFCs. Summary of SFC distribution over types is presented in Table 2.

In general, the number of close-type SFCs is significantly greater in spiral galaxies than in irregular galaxies. Notable exceptions from this rule are NGC 4736 (which contains significantly less SFCs than other spiral galaxies) and IC 2574 (which hosts much more SFCs than other irregular galaxies). It is interesting to note that Donovan Meyer et al. (2013) also examined NGC 4736 and identified only a few giant molecular clouds in this galaxy. Hence, both galaxies are exceptions from a rule and deserve further investigation.

Initially, we included emission in the GALEX near-UV (NUV) band in consideration, in order to compare NUV emission with dust emission. However, it turned out that NUV emission is highly correlated with 8 μ m emission so there is no need to include the NUV band in a future analysis.

3 Photometry and comparison of various star formation tracers

As mentioned in Section 1, our aim is to investigate whether SFC emission fluxes in different bands (24, 160 and 8 μ m ) are correlated with each other. For this purpose, we performed photometry analysis for each SFC. Flux in each aperture was calculated and then the local background was subtracted from the total flux as described by Khramtsova et al. (2013). Flux in each aperture was calculated as follows. We sum fluxes in each aperture, taking into account partially occupied pixels. We also moved the aperture by ± 2 pixels along both axes and repeated flux computation. Then the mean flux and flux error were computed. The local background was calculated using the 6-pixel annulus around each aperture. Background then was subtracted from the flux. Total flux error was calculated as σ stat 2 + σ dist 2 , where σ dist is an error related to distance uncertainty.

We present the results in Figure 3. All fluxes are reduced to the same distance of 10 Mpc. Effects of extinction were not taken into account since they are not significant at these spatial scales and wavelengths. It is easy to see that relation F 160 ∼ F 8 is valid for spiral galaxies and the slope is the same for all the galaxies. This is not surprising since all these galaxies are nearby and of similar metallicity (Table 1).

Figure 3

Relations between aperture fluxes in 160 μ m band and 8 μ m band for spiral and irregular galaxies. All fluxes are reduced to a distance D = 10  Mpc. The line on the left panel shows a least-square fit with a slope of 0.98 on a logarithmic scale.

One can also see that there is no simple relation for irregular galaxies, but there are many regions with the absence of sufficient emission in one band. These SFCs are promising targets for future investigation.

One can get similar results for spiral galaxies in the other “dust” bands: F 160 ∼ F 24 , F 24 ∼ F 8 . We do not give an illustration for these relations since they are nearly identical to the F 8 – F 160 relation. Fluxes from SFCs in irregular galaxies are much less correlated.

The PHANGS data are available for galaxies NGC 628, NGC 4254, and NGC 5236. This allows us to consider relations between flux in CO(2–1) spectral line band and fluxes in “dust” bands. It turns out to be linear, F CO ∼ F dust , similar to the ones described above (Figure 4).

Figure 4

F CO – F dust relation for NGC 5236. This relation seems to be close to a linear one since the line shows a least-square fit with a slope of 0.91 on a logarithmic scale. Yet correlation coefficient r = 0.5 may indicate that more complicated relation takes place.

4 Kinematics of star formation complexes

We initially selected galaxies seen almost face-on, so that it is unlikely that two unrelated SFCs appear randomly on the same line of sight. Nevertheless, we decided to take a closer look at spectra of SFCs. PHANGS data contain images in the CO(2–1) line with a resolution of 2 arcseconds for galaxies NGC 628, NGC 4254, and NGC 5236, which are included in our list. Therefore, we can obtain spectra for some of our SFCs to check whether there are SFC candidates with several kinematic components, which would imply that these are not true SFCs, but rather a combination of two (several) SFCs along the same line of sight (although see, e.g., Tenorio-Tagle et al. 1996 for the discussion of possible several velocity components in the spectrum of a single star-forming region).

We found that there are indeed two peaks in some spectra. Figure 5 shows three examples of two-peak spectrum for three regions in NGC 4254. Blue (orange) colored lines are spectra from PHANGS (HERACLES) data. This may indicate the existence of two spectrally (and therefore kinematically) separated regions disguised as one photometric region. The difference between these peaks is about 20 – 30 km s − 1 . However, we should keep in mind that this can be an effect of degraded resolution because of convolution. We used this information to check if there are apertures in convolved images with more than one SFC candidate inside. However, this procedure could also be used for unconvolved data if one wants to find out if there are several kinematically distinct regions within an aperture.

Figure 5

Spectra of three regions taken from PHANGS (blue) and HERACLES (orange) data. All three regions are two-peak regions in NGC 4254. Horizontal line represents SNR ≈ 3 in HERACLES data.

Our examination showed that in most “two-peak” cases the two regions can be spatially resolved using maps with higher resolution (without convolution). We also noticed that lower resolution maps in PHANGS lead to more “two-peak” regions. For example, most of two-peak spectra came from regions in galaxy NGC 4254 that happened to have the worst resolution among the three galaxies mentioned above.

We also compared PHANGS spectra with HERACLES spectra. We conclude that at least as we speak about the number of peaks, there is about 100% coincidence between these two datasets even though the spectral resolution is sufficiently higher in PHANGS data.

In these three galaxies, we have no need to examine such two-peak regions since they all can be resolved on unconvolved 8 μ m “Spitzer” images. By the way, this means that our method of identifying SFCs works satisfactorily. Our method of finding kinematically distinct regions can be applied if there are no high-resolution images, but a spectrum of the region is available.

5 Conclusion

We developed an approach to identify SFCs in external galaxies matching emission in 160 μ m band and 8 μ m band and using well-defined classification of overlapping regions. One can use this approach to identify SFCs in galaxies that were not included in our dataset or to check automatic algorithms of SFC selection.

For all the selected spiral galaxies a strong correlation was found between emission in dust and CO bands: F 8 ∼ F 24 ∼ F 160 ∼ F CO . However, we should keep in mind that SFCs were selected comparing fluxes in 160 and 8 μ m bands so correlation between dust fluxes may be weaker than it is in this article. Our selection procedure is based on the presence of emission at 160 and 8 μ m , so we can miss some regions that are bright in one spectral band and faint in another. This is the reason why correlations between dust bands may be influenced by selection effects.

We did not find such a correlation for irregular galaxies. These galaxies are hosts for several unusual regions (see, e.g., two leftmost points on graph for irregular galaxies in Figure 3), which show the absence of emission in one band. As we mentioned in Section 1, all the considered bands are related to the star formation process so we expect to see some emission in all the bands, if these regions are indeed SFCs. This peculiar behavior goes beyond the scope of this work and has to be explained in the future.

An approach of finding kinematically distinct regions was proposed. We used it to check our procedure of SFC identification but one can use it if there is a bad resolution image and a spectrum of an SFC. We used this approach only to check whether the aperture contains more than one region. But one can apply this technique in other cases (e.g., without possibility return back to good resolution).