A Comparison of Outflow Properties in AGN Dwarfs versus Star-forming Dwarfs

Our goal was to build a suitable control sample to use for comparison to the study of L20. Thus, we targeted SF dwarf galaxies with similar redshift and stellar mass that have prominent outflows but no signs of AGN activity.

To obtain this sample, we selected all galaxies from the NASA–Sloan Atlas with M_* < 10¹⁰ M_⊙ and z < 0.05 that were optically classified by the Sloan Digital Sky Survey (SDSS) as an SF or starburst galaxy. We then used the open-source Python code Bayesian AGN Decomposition Analysis for SDSS Spectra (BADASS; Sexton et al. 2020) to fit the SDSS spectra of these galaxies in order to search for broad components to the [O iii] λ5007 emission lines, which are indicative of outflow activity. Only 72 out of 26,560 galaxies (0.27%) showed significant broad components, a fraction consistent with that found in samples of more massive galaxies (Matzko et al. 2022). From this, we selected a subset with the most visible broad [O iii] λ5007 components and observed 10 of these targets (shown in Figure 1 and listed in Table 1) with the Keck Cosmic Web Imager (KCWI). These included three targets from the Manzano-King et al. (2019) SF dwarf sample. The targets were matched in redshift and stellar mass to those of L20 (their distributions are shown in Figure 2) and were classified as SF in the Baldwin, Phillips & Terlevich (BPT) diagram (Baldwin et al. 1981; Veilleux & Osterbrock 1987; see Figure 3), which is the diagnostic diagram used to distinguish between AGNs and SF galaxies based on the ratios of optical emission lines. The SF dwarfs are also similar in size to the AGN dwarfs if we compare them based on their half-light radius (average radius of 1.1 kpc for the AGN dwarfs versus 1.0 kpc for the SF dwarfs).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

All the targets were observed with KCWI (Morrissey et al. 2018) through the Keck program 2021A-U080 (PI: G. Canalizo) on 2021 March 19 and 2021 March 20. All targets were observed using the blue grating (BL), which covers a wavelength range of 3500 to 5500 Å, and the small slicer setup (spectral resolution ∼80 km s⁻¹ FWHM at 4550 Å). The observations were obtained under clear sky conditions and a typical seeing of ∼08.

The PSF of these IFU observations was measured using observations of the spectrophotometric standard stars that were taken before and after the on-target observations throughout the night. We stacked narrowband images in the range of 5000–5100 Å and fit 2D Moffat profiles (Moffat 1969) to them. We used a Moffat profile as opposed to the Gaussian profile used by L20, as a Moffat profile fits the radial profile of the star better, as referenced by previous studies (Li et al. 2016). Moreover, there was no significant difference in the FWHMs obtained from both profiles, so we adopted the median FWHM obtained from the best-fit Moffat profiles. There would be some discrepancy in the PSF as we could not take observations of the standard stars at the same time as the science target, but the seeing (as measured by the Maunakea Weather Center) ranged from 075 to 10 across both nights, and the FWHM we measured was well within this range.

All the targets were reduced with the KCWI Data Extraction and Reduction Pipeline (KDERP), ¹⁰ following the standard procedures listed in the manual. The pipeline also included both wavelength and flux calibrations. We then used routines from the IFSRED (Rupke 2014a) library to resample the cubes with square spaxels as individual exposures had rectangular spaxels. The data cubes were resampled to 015 ×015 using IFSR_KCWIRESAMPLE, and then the data cubes of the same target were median-combined into a single data cube using IFSR_MOSAIC, which also corrected for shifts caused by dithering.

We spatially binned the data cubes using the Voronoi binning method (Cappellari & Copin 2003). Binning is a useful technique that can be used to preserve the spatial resolution of the data while obtaining a higher signal-to-noise ratio (S/N). This would help us to accurately characterize the broad components in the emission lines to determine the outflows, which might otherwise be missed due to a lower S/N. The Voronoi binning method, in particular, bins or divides the data into hexagonal bins, so data from neighboring pixels are grouped as one to achieve the target S/N. Thus, the size of the bins increases with the distance from the photocenter of the target to match the required S/N. We set an S/N of 5 for our targets as we found that was the highest S/N we could obtain before we started losing spatial information from the binning. All subsequent analysis was done on the Voronoi-binned spectra.

3.2.1. Fitting the [O iii] Emission Lines

3.2.2.  F-tests to Determine the Optimum Number of Components to Fit the [O iii] Emission Lines

The F-test determines the significance of the fit between a complex (higher-order) and a simple (lower-order) model in order to justify the number of Gaussian components we include in our fit. To do this, we calculated the standard deviation of the residuals of the fit (given by σ) for each spaxel. We did the F-test, given as , for one versus two components as well as for two versus three components for the Gaussian fits to the [O iii] λλ4959, 5007 lines. The F-test was calculated over a wavelength range of 4800–5500 Å, which includes the [O iii] λλ4959, 5007 lines. An F value greater than 3 indicates that the higher-order fit (i.e., the fit using more components) is justifiable. We performed the F-test for the fits to the [O iii] λλ4959, 5007 lines in every spaxel of our targets and mapped the results of the F-test. All targets failed the F-test for one component, so more than one component was required to fit the profile. For some of the targets (J1325, J1622, J1047, and J1014), it was evident that we needed only two components. The rest of the targets were more ambiguous since they contained a fraction of spaxels with F values close to, but below, 3. For those targets, we inspected the fits visually to assess the effect that including an additional component had on the residuals and to determine whether the additional component was fitting a real feature in the emission lines. From Figure 4(a), we see that fitting a single component is not sufficient, and it misses a crucial part of the emission line. In Figure 4(c), we see that the third component does not fit any real feature. We, therefore, chose to use a maximum of two components (Figure 4(b)) for all the targets to keep the fits as simple as possible, with the caveat that our flux measurements may miss a small fraction in the wings of the emission lines. This is also different from the analysis of L20, in which three of the eight targets needed three components to get the best fit, while two needed just one component.

Zoom In Zoom Out Reset image size

3.2.3. Fitting the Full Spectrum

The SFRs of the SF dwarfs were calculated from the [O ii] lines using the following equation (Kewley et al. 2003):

where [log(O/H) + 12] was taken to be 8.9 for solar metallicities (as adopted by L20). If we consider a 0.5× solar metallicity, the SFR would be ∼30% lower.

These values are listed in column (10) of Table 1. We find that for most of our targets, the SFR is higher than the values listed for the AGN dwarfs from L20, which have an average upper limit of 0.20 M_⊙ yr⁻¹. In comparison, the average for the SF dwarfs is 0.25 M_⊙ yr⁻¹. We note that the SFR values for the SF dwarfs are likely to be higher than those for the AGN dwarfs, as there could be a significant contribution from the AGN to the [O ii] emission lines, which in turn could lead to an overestimation of the value of the SFR.

The SFR for 9 of the 10 objects in our sample could also be obtained from the MPA-JHU catalog, in which the SFRs were calculated based on the techniques used in Brinchmann et al. (2004). These values are much higher than our calculated values, with an average of 1.38 M_⊙ yr⁻¹. This could be because the MPA-JHU estimates used multiple lines as SF indicators, while we only used the [O ii] lines for our estimates. The biggest difference was for J0838, which had an MPA-JHU estimate for the SFR of 3.94 M_⊙ yr⁻¹ as compared to the SFR of 0.45 M_⊙ yr⁻¹ we measured from [O ii].

We calculated the median, minimum, and maximum values of v₅₀ and W₈₀ for the emission lines in our targets, similar to the methodology adopted by L20 for AGN dwarfs. The parameter v₅₀ is defined as the median velocity at the 50th percentile of the total flux, while W₈₀ is the line width that includes 80% of the total flux. W₈₀ was calculated as v₁₀ – v₉₀, where v₁₀ and v₉₀ are the velocities at the 10th and 90th percentiles of the flux, calculated from the red side of the line. For a schematic diagram depicting the different quantities mentioned, refer to Figure 3 in L20.

These values were determined relative to the systemic velocity of the galaxy, which we determined by fitting stellar absorption features in the integrated spectrum of the data cube. Uncertainties in the velocity measurements were estimated by Monte Carlo methods over 100 iterations.

Measurement uncertainties in v₅₀ estimated in this way are typically below 5 km s⁻¹. However, velocity maps (see Figure 5) show bin-to-bin variations in the order of 20 km s⁻¹ in some cases. These fluctuations are more likely to be due to error measurements than to true changes in velocity. Therefore, we adopted 20 km s⁻¹ as a conservative error in our velocity measurements. Our measurements for the whole sample are summarized in Table 2.

Zoom In Zoom Out Reset image size

A maximum of two components were fit to the emission lines for each of the targets. The C1 component refers to the narrow component, and C2 refers to the broad component. L20 assumed that C2, or in some cases C3, refers to the outflow component since nearly all of their C2/C3 v₅₀ values were blueshifted, which is a clear indicator of outflows. In our case, the broad components were approximately at the systemic velocity or even redshifted, with only certain regions showing significantly blueshifted velocity (Figure 5). We only considered spaxels with a blueshifted velocity relative to the systemic to constitute an outflow. Redshifted values of the C2 v₅₀ could indicate disturbed or stirred gas in the galaxy. If we had chosen redshifted spaxels to constitute the outflows as well, then that would have indicated a larger fraction of the spaxels contributed to outflowing material, which could change the values of the energetics carried out by the flow. This could lead to either overestimation or underestimation of the effect of the outflow. Since we did not have a good indicator to determine whether the redshifted spaxels truly constituted an outflow, we tended to agree with the generally used assumption that only blueshifted spaxels are likely to depict an outflow. However, for completeness, we considered both cases (using blueshifted spaxels versus using both blue- and redshifted spaxels) in our analysis of the energetics of the outflows (see Section 4.8).

On average, the median values of v₅₀ for the C2 component were close to 0 km s⁻¹. This could be due to the fact that most of the spaxels did not have a significant velocity offset. However, since we considered only blueshifted velocity values to constitute the outflows, the average of the minimum (maximally blueshifted) values of v₅₀ was found to be ∼−60 km s⁻¹. From the velocity maps in Figure 5, we see that blueshifted outflows constitute a small fraction of the kinematics in the galaxy. The C2 W₈₀ velocity maps are plotted in Figure 6 and show a significant spread in the dispersion values. The average of the median values of W₈₀ was ∼480 km s⁻¹.

Zoom In Zoom Out Reset image size

As mentioned earlier, the shape of the line profile of the broad component in our targets is not preferentially blueshifted and is symmetric, in contrast to the clear blueshifted line profile in L20's AGN dwarfs. This is consistent with previous observations of line profiles associated with star-formation-driven outflows (Concas et al. 2017; Davies et al. 2019; Matzko et al. 2022). The symmetric shape of the line profiles in the SF dwarfs could indicate lower extinction along the line of sight in at least some of the outflows. AGN-powered outflows originate from the dusty centers of galaxies, and so they are affected by extinction, which would lead to their profiles being more blueshifted (Heckman et al. 1981). However, stellar-driven outflows are not restricted to originating from the centers of galaxies, as energetic stellar processes can occur anywhere in the galaxy. Any outflows near the edge of galaxies, i.e., those that are least affected by extinction, are more likely to dominate the flux of the final spectrum, leading to a profile that appears to be more symmetric, with velocity values centered around 0 km s⁻¹ (S. Kadir et al. 2023, in preparation).

We plot the v₅₀ maps for the C1 component and the stellar fields in Figures 7 and 8, respectively. The C1 component represents the gas that follows the stellar component in the galaxy. Comparing Figures 5 and 7, we see that for nearly all the targets, the C1 and C2 components have distinct kinematics, implying that the two components are likely tracing different physical processes. For instance, the velocity fields for the C1 and C2 components for J1622 in Figures 5 and 7 are perpendicular to each other, possibly indicating that the outflow is oriented perpendicularly to the rotation of the galaxy. As mentioned earlier, the gas in the C1 component follows the stellar component in the galaxy as can be readily seen by comparing Figures 8 and 7. We see signs of rotation in the stellar field of some targets such as J1047, J1307, J1622, and J1717. For the rest of the targets, there does not seem to be a clear rotation in the stellar velocity, indicating that either we are seeing the galaxies face-on in some cases, or the galaxies may not be supported by rotation in other cases.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The W₈₀ velocity maps of the C1 component of J1325 in Figure 9(a) show narrower widths of the line profile compared to those of the C2 component. This trend was prevalent among all objects in the sample, possibly indicating that the C2 component likely traces processes like outflows that can cause the broadening of the line profiles, while the C1 component is relatively quiescent. We refer the reader to the appendix for a more detailed analysis of the v₅₀ and W₈₀ maps of the C1 and C2 components of the individual targets. The stellar velocity dispersion of all the targets also has low values as indicated by Figure 9(b).

Zoom In Zoom Out Reset image size

Figure 5 also shows the spatial distribution of the C2 component for each galaxy, with the axes in units of kiloparsecs. The outflows appear to extend to significantly larger radii than the half-light radius of these targets as given by SDSS (see Table 1). We also mapped the radial profile of the outflows and compared it to the PSF (see Appendix B). We fit a Moffat profile to a standard star that was observed close to the target to reduce uncertainties caused by changes in atmospheric turbulence and to determine the physical extent of the outflows in our targets. The deviation from the profile of the star for all the targets shows that the outflows are spatially extended.

The [O iii] λ5007 flux maps of our targets show that multiple H ii regions are present in some of the targets (see Figures 19 and 23). These are most apparent in J1307 and J1717, but they are also likely present in J0838, J0850, J1014, and J1302. This makes it challenging to set a single radius for the outflow while determining the energetics (see Section 4.8), since different regions may have outflows of different sizes. L20 used a single spherical shell model for their sample of AGN dwarfs as AGN-powered outflows originate from a central source and the morphology of the outflows is spherical on the 2D sky plane. For comparison purposes, we considered a similar spherical shell model for the SF dwarfs. The radius of the outflow was taken to be the extent to which the outflow component could be measured from the radial profiles and velocity maps. Thus, the results need to be analyzed with caution since the radius of the outflows could be much smaller.

Since our main goal was to compare outflows powered by star formation to those powered by AGNs, it was important to make sure the galaxies in our sample do not host any AGN activity. The targets in our sample were classified as SF based on integrated SDSS spectra (Figure 3). However, a weak AGN signal could be diluted by the strong emission lines associated with SF regions.

Using MaNGA IFU data, Mezcua & Domínguez Sánchez (2020) found AGN activity in individual spaxels based on the ionized line ratios of lines used in the BPT diagram that was not apparent in the integrated spectra of the same targets. We took a similar approach for our data cubes to ensure that all of the ionization is likely caused by SF activity and to rule out any possibility of AGN activity that can manifest itself in the various comparison ratios.

To do this, we obtained the [O iii]/Hβ ratio for each spaxel and compared it to the value of the Kauffmann demarcation line (KO3; Kauffmann et al. 2003) between AGN and SF galaxies. We chose the Kauffmann demarcation line because it has lower limits than the Kewley line (Kewley et al. 2001). Since the KCWI wavelength coverage does not extend beyond 5500 Å, we were unable to measure the [N ii]/Hα ratio directly. So for each target, we obtained the value of [N ii]/Hα from SDSS and determined the corresponding [O iii]/Hβ ratio on the Kauffmann line. Note that this value of [N ii]/Hα was obtained using an integrated aperture and thus may not be indicative of the actual spaxel-to-spaxel variation of the line ratio. This could lead to the biasing of the ratio toward the center of the galaxy and to significant errors while determining values in outer regions. Being mindful of this caveat, we subtracted the obtained value from our calculated value and plotted those spaxels where the difference was greater than 0 (see Figure 10). If the difference is greater than 0, the ratio is higher than the accepted value for possible SF origins.

Zoom In Zoom Out Reset image size

For 8 of the 10 targets, no part of the galaxy indicated [O iii]/Hβ ratios above the SF line. In two targets, J0838 and J0850, the ratios in some spaxels were 0.2 dex greater than the expected value; Figure 10 shows the specific location of these spaxels. While the higher [O iii]/Hβ ratios in these spaxels are likely due to shocks, we cannot rule out the possibility that there may be some AGN contamination in these two targets. However, all the outflow properties in these two targets appear to be consistent with the rest of the sample, so their outflows are most likely powered by star formation.

The electron densities of the targets can be obtained from the [S ii] line ratios or the [O ii] line ratios from the relation given by Sanders et al. (2016). We took the [S ii] line ratios from SDSS for all our targets (similar to L20). The values of the electron densities (average value of ∼165 cm⁻³) were lower compared to the values for the AGN dwarfs (average value of ∼400 cm⁻³). This difference between the two samples could have been caused by the fact that AGNs are capable of producing strong ionization and thus have higher values of ionized electron densities than SF dwarf galaxies.

Dust extinction in the targets was measured using the Hβ/Hγ emission line ratios of their integrated spectrum. The extinction was calculated using the extinction curve from Cardelli et al. (1989) with an R_v = 3.1 and an intrinsic Hβ/Hγ ratio of 2.13 using case B from Osterbrock & Ferland (2006) at a temperature of 10⁴ K. We also compared these values to the extinction calculated from the Hα/Hβ line ratios obtained from SDSS, and they were comparable within the error limits. So we used the values of extinction A_v determined from the Hβ/Hγ ratios for further calculations. The values of the electron densities and extinction for the SF dwarfs are given in columns (3) and (4) of Table 3.

Table 3. Energetics of the Outflows

Name	Comp	ne (cm−3)	Extinction (E(B – V))	log(M/M⊙)	Rout (kpc)	log[(dM/dt)/(M⊙ yr−1)]		log[(dE/dt)/(erg s−1)]		log[(cdp/dt)/(L⊙)]
						Blue	Red+Blue	Blue	Red+Blue	Blue	Red+Blue
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
J0838	C1	121 ± 0	0.2		2.9	−4.1	−1.9		37.0		7.4
	C2				2.9	−2.3	−2.0		34.0		7.4
J0850	C1	150 ± 0	0.2		1.2	−2.8	−2.0		37.03		7.6
	C2				1.2	−3.5	−2.5	35.5	36.7	5.8	6.9
J1014	C1	63 ± 24	0.1		2.7	−1.9	−2.3		36.2	7.6	6.8
	C2				2.7	−1.9	−2.5		36.2	7.6	6.7
J1047	C1	265 ± 2	0.2		2.0	−5.1	−2.4		35.8		7.0
	C2				2.0	−1.8	−2.7		36.2		6.7
J1302	C1	146 ± 2	0.2		2.1	−3.3	−2.4		37.1		7.0
	C2				2.5	−4.2	−2.6		37.2		7.0
J1307	C1	113 ± 15	0.2	6.0	1.9	−1.8	−2.1		36.8		7.2
	C2				1.4	−2.7	−2.5		37.3		7.2
J1325	C1	233 ± 40	0.1	5.4	1.6	−3.0	−3.2		35.0		5.7
	C2				1.5	−2.2	−2.8		36.3		6.6
J1415	C1	271 ± 18	0.2		1.5	−3.7	−3.0		35.9		6.1
	C2				1.4	−2.5	−3.1		35.3		6.0
J1622	C1	193 ± 5	0.1		3.2	−2.0	−2.2		36.5		7.0
	C2				2.9	−2.1	−2.3		37.0		7.2
J1717	C1	85 ± 11	0.2		3.4	−2.4	−2.0		37.4		7.4
	C2				3.0	−2.4	−2.4		37.5		7.2

Notes. Column (1): short name of the target. Column (2): component number. Column (3): electron densities measured from the [S ii] line ratios from SDSS. Column (4): the extinction values calculated from the Hβ/Hγ measured from the integrated spectrum of the KCWI data. Column (5): ionized gas mass of the outflow. Column (6): outflow radius used in the calculation of the energetics. Columns (7) and (8): the mass-loss rate calculated using only the blueshifted spaxels and both the blueshifted and redshifted spaxels, respectively, to constitute the outflow. Columns (9) and (10): the energy rate calculated using only the blueshifted spaxels and both the blueshifted and redshifted spaxels, respectively. Columns (11) and (12): the momentum rate calculated using only the blueshifted spaxels and both the blueshifted and redshifted spaxels, respectively. The errors in the energetics are dominated by extinction.

Download table as:  ASCIITypeset image

Following L20, the ionized gas masses were calculated using the Hβ luminosity of the outflowing gas. We converted the Hβ luminosity to the Hα luminosity by a correction factor L_Hα,corr = 2.86 L_Hβ,corr. The correction factor is the intrinsic Hα/Hβ ratio for H ii regions, which is appropriate for the case B recombination assumed from Osterbrock & Ferland (2006) with T = 10⁴ K. We also corrected for extinction based on an intrinsic Hβ/Hγ ratio of 2.13 (for more details, refer to Section 4.7). The equation for the ionized mass was adapted from Equation (29) in Veilleux et al. (2020) by assuming solar metallicity and is given as

We also calculated the mass-loss rates (dM/dt), the momentum rates in units of energy (cdp/dt), and the kinetic energy rates (dE/dt). We assumed a time-averaged, thin-shell free wind model, which has been previously used by Shih & Rupke (2010), Rupke & Veilleux (2013b), and L20, with a spherically symmetric outflow.

The energetics were calculated by summing up quantities over individual spaxels:

where m_out is the outflowing mass in each spaxel, v_50,out is the value of v₅₀ in each spaxel, and σ_out is the velocity dispersion calculated as W₈₀/2.73. R_out is the radial extent of the spherical outflow and was determined from the maximum radial extent indicated in the C2 velocity maps of the individual targets (Figures 5 and 6). The angle between the velocity vector of the outflow in 3D space and the line of sight is defined as θ = sin⁻¹(r_spaxel/R_out).

The values of the energetics are shown in Table 3. The ionized gas mass in column (5) was calculated using Equation (2) and it gives a measure of the gas moved by the outflows for C2. The values of the energetics in columns (7), (9), and (11) indicate the mass-loss rates, energy rates, and momentum rates of the ionized gas, respectively.

As indicated earlier, we only considered the blueshifted values of v_50,out as being part of the outflow. This was added over the entire radial extent to get the total values of the energetics. If we also include the redshifted spaxels to constitute the outflow, we find that the value of dE/dt is on average 10% larger than the values calculated for only the blueshifted spaxels, implying that there would be a small increase in the values of the energetics. The values of the energetics calculated by considering both redshifted and blueshifted spaxels to constitute the outflow are given in columns (8), (10), and (12) of Table 3 for comparison.

Comparison of the values given in Tables 2 and 3 to the SFRs in Table 1 shows that there is no correlation between the SFR (or specific SFR) and outflow properties for SF dwarfs. However, this could be due to the poor constraints we have on SFRs, as noted in Section 3.3.

As in the analysis done by L20, we compared the outflow velocities of our sample of SF dwarfs to the escape velocities of the galaxies in order to determine whether a significant fraction of the outflowing gas escapes the galaxy.

While there are several methods to calculate the escape velocities of galaxies, we used abundance matching in order to obtain results that can be compared directly to those of L20. We calculated the escape velocities by first obtaining the halo masses of the galaxies using abundance matching (Moster et al. 2013). We assumed a Navarro–Frenk–White dark matter density profile (Navarro et al. 1996). Using the integrated spectrum of the broad component of [O iii] λ5007 for each galaxy, we found the ratio of the flux that has velocities higher than the escape velocity to the total flux of the line, and we defined this as the escape fraction. The escape fraction in percentage for the SF dwarfs is given in Table 4. Note that we defined the escape fraction based on the flux ratios, not on the mass ratios, so the caveats listed in L20 still apply here. Our escape velocities were also measured from the center of the galaxy and should be considered conservative upper limits, as escape velocities are highest at the center.

The escape fractions listed in Table 4 demonstrate that only a small fraction of the gas (if any) is able to escape the galaxies. Since escape velocities derived via abundance matching are susceptible to significant systematic biases, we also estimated them using stellar velocities and dispersions. We determined the circular velocity, v_circ = , where v_* and σ_* are the maximum stellar velocity and stellar dispersion, respectively. Note that this is a lower limit, as many of our rotation curves do not extend to the flat part of the curve. We then used the relation v_esc ≈ 3v_circ (Veilleux et al. 2020) to determine the escape velocity. We found that the escape velocities determined in this way are on average ∼30% higher than the escape velocities given in Table 4, indicating that an even lower fraction of the gas would be likely to escape.

There is a difference in kinematics, extent, and energetics between the SF dwarfs and AGN dwarfs. The average of the median values of v₅₀ of the SF dwarfs, found in Section 4.1, is lower than the average of the median values of the AGN dwarfs (−64 km s⁻¹). Comparison of the maximally blueshifted velocities also shows that star-formation-driven outflows are slower than AGN-driven outflows (average of ∼−65 km s⁻¹ for SF dwarfs versus −145 km s⁻¹ for AGN dwarfs). The median values are redshifted or centered around 0, compared to the highly blueshifted values for AGN dwarfs. If we consider just the blueshifted v₅₀ to contribute significantly to the outflow, we can see that there are very few blueshifted spaxels in the broad component of the SF dwarfs, whereas nearly all the spaxels in the broad component of the L20 AGN dwarfs are blueshifted. This could possibly indicate that strongly blueshifted outflows are not as prominent in SF dwarfs as they are in AGN dwarfs, and most of the effect of stellar activity may be to stir the gas rather than to expel it from the hosts.

The overall extent of stellar-driven outflows is greater than that of AGN-driven outflows as can be clearly seen in Figure 11. Note that the detectability limits were comparable in the measurement of outflow extent in both the SF and AGN samples. Therefore the difference in sizes is real and not a result of different sensitivities in the two samples. The AGN-driven outflows extend to distances between 0.3 and 3.1 kpc, with an average of 1.2 kpc. On the other hand, the stellar-powered outflows extend to further distances, between 1.2 and 3.2 kpc, with an average of 2.2 kpc. This could be because stellar-powered outflows may originate at multiple places within the galaxy (see Section 4.4), and are not necessarily confined to the center as AGN-powered outflows are. They can thus reach larger scales, even though they might not be as fast as AGN-powered outflows.

Zoom In Zoom Out Reset image size

The mass, momentum, and kinetic energy rates are all much lower for the SF dwarfs than for the AGN dwarfs, as indicated in Figure 12. As mentioned earlier, assuming both the redshifted and blueshifted spaxels contribute to the outflows, we find that the values of the energetics increase by a small fraction, but overall they are still lower by 2 orders of magnitude compared to those of the AGN dwarfs.

Zoom In Zoom Out Reset image size

The differences between the quantities in the two samples increase from the mass-loss rates to the momentum rates and are maximal for the energy rates. These calculations depend on both line luminosities and velocity terms. While the [O iii] λ5007 luminosities of the AGN dwarfs are comparable to those of the SF dwarfs, the median v₅₀ of the AGN-driven outflows are much higher than those of the SF-driven outflows.

The radial extent of the outflows could also play a role in lowering the energetics for the SF dwarfs, as the outflows are more radially extended than those of the AGN dwarfs.

The escape fractions from Table 4 for the SF galaxies are comparable to the escape fractions for the AGN galaxies. However, the AGN dwarfs have a maximum of 6% for two galaxies. Although these fractions are not significant enough to translate to an effective portion of the gas being ejected from the host galaxies by the outflows, they could still show that AGN dwarfs could play a more effective role in stirring up gas, which can considerably affect their environment. The escape velocities are also upper limits as they are calculated from the center of the galaxy, and we would also need to factor in the frequencies of these outflows.

There is a significant difference between the outflows powered by AGNs and those powered by stellar processes. The lower speeds of the star-formation-driven outflows as compared to the AGN-driven outflows indicate that the gas is moving more slowly, and the lower energetic rates indicate that they are possibly not transporting large amounts of mass or energy from the galaxy. The greater values of kinematics and energetics from the AGN-driven outflows suggest that AGNs could play a more significant role in stripping the galaxy of gas and halting or slowing down star formation. Thus it is important to consider AGN contribution to feedback in dwarf galaxies in the study of galaxy evolution.

From the color–magnitude diagram for the AGN dwarfs and SF dwarfs presented in Figure 13, we see that the AGN dwarfs are redder than the SF dwarfs. The correlation between galaxy colors and the nature of the galaxy has been extensively studied (Martin et al. 2007; Schawinski et al. 2007, 2010). The bluer colors of the SF dwarfs could indicate that their outflows have not played a significant role in effectively quenching star formation. On the other hand, the more energetic AGN-powered outflows may have hampered star formation. We note, however, that the positions of the galaxies in the color–magnitude diagram could be a selection effect, as it is generally easier to detect AGNs in galaxies with lower SFRs.

Zoom In Zoom Out Reset image size

The extent of the SF outflows is found to be greater than that of the AGN outflows. This could indicate that the SF outflows can reach greater scales, and thus their overall time-averaged kinetic energy outflow rates appear lower. SF outflows could be comparable to AGN outflows in stirring up the gas at smaller scales in the host galaxy. The escape velocities could also be lower at these scales, so it would be easier for the material in the SF outflows to escape the host galaxy.

While star formation could be instantaneously less energetic compared to AGN-driven outflows, it is important to consider the long-term effects of the two processes. Multiple bursts from stellar activity could lead to stellar processes dominating over AGN-driven outflows in regulating the environment of the galaxy. For AGN-driven outflows to be significant over a long time, it would be important to consider the duty cycles of AGN activity or how frequently AGN-powered outflows occur as compared to stellar-driven outflows.

A number of studies have measured the kinematics of ionized gas in dwarf galaxies, and we compare our results with some of the work found in the literature. Schwartz & Martin (2004) used high-resolution echelle spectrographs to study the Na D absorption lines in six nearby dwarf starburst galaxies. They discovered that three of them have outflows, with an average outflow velocity of −27 km s⁻¹. This is lower (more blueshifted) than our median v₅₀ values. However, our maximum blueshifted v₅₀ values for the broad component are comparable.

van Eymeren et al. (2009) looked at emission lines in H i and Hα to estimate outflow velocities in the nearby SF irregular dwarf galaxy NGC 2366. They characterized the galaxy as having two major outflows, one with blueshifted velocities of 30 km s⁻¹ and the other with redshifted velocities up to 50 km s⁻¹ with a radial extent of 1.4 kpc. It is interesting that they considered redshifted velocities to indicate an outflowing component, in contrast to our analysis. We find that if we consider the redshifted velocities in our calculations of the energetics, the values do not change significantly. However, it would imply that it is possible that the redshifted velocities in our C2 maps are indicative of actual outflowing gas.

Our results are also in line with Marasco et al. (2023), who studied ionized gas kinematics in a sample of 19 nearby starburst galaxies. The line profile they observed as depicted in their Figure 6 also appears to be symmetric, and there is no apparent blueshift. They found ionized gas outflow rates in the range of 10⁻⁴–10⁻¹ M_⊙ yr⁻¹, which is an order of magnitude higher than the values calculated in this work for SF dwarfs. However, they still concluded that stellar feedback only stimulates a gentle gas cycle and is less likely to cause major blowouts that affect the evolution of the dwarf galaxy.

Manzano-King et al. (2019) observed 3 of the 10 SF dwarfs in this work and found wind speeds that are much higher than what we measured. This could be due to the fact that they defined outflow velocity differently from this work. It could also be attributed to the slit spectroscopy used as compared with the higher resolution power of KCWI and the larger field of view, which helped us make more accurate measurements. Some differences in profiles were also observed by L20 in their comparison of KCWI data with LRIS data for the AGN dwarfs. While the results generally agreed, they found that for two of the targets, the KCWI data showed narrower profiles with smaller blueshifts as compared to the LRIS data (for a more in-depth analysis, refer to Section 4.3 in L20).

Our analysis shows that stellar feedback might be less significant than previously expected. With the spatial information provided by IFS, we are able to see the effects of stellar-powered outflows in galaxies. There appear to be some regions of strong outflow activity, but even then, it is not as powerful as AGN activity. It is possible that, at least in many cases, stellar feedback might only disturb the gas locally and not have major effects on the host galaxy as a whole.

Figure 14 shows maps of the [O iii] λ5007 flux, velocity, and stellar kinematics for J0838. The first row represents the broad component, the second row represents the narrow component, and the third row of panels represents the total flux map. The flux is normalized to the value indicated by the black cross to mark the approximate center of the galaxy. This is slightly different from the case of L20, who marked the black cross with the highest flux value, which also coincided with the center of the galaxy due to the nature of AGN dwarfs. In our sample, the maximum value of the flux need not be close to the center as there is no preferential location for strong stellar activity. To maintain uniformity with the L20 analysis, we consider our normalizing flux value to come from a spaxel in the apparent center of the galaxy. The maximum values of the flux are indicated above the panels. The plot of the C2 flux shows some extended emissions distinct from those traced by C1.

Zoom In Zoom Out Reset image size

The C2 v₅₀ values have a significant gradient compared to the C1 values. There does not seem to be a strongly blueshifted component, but there is evidence of a redshifted component. In contrast, the C1 velocities are relatively quiescent with some being very close to 0 km s⁻¹. The strongly redshifted blobs on the right are likely due to noise and do not have any physical meaning. The v₅₀ maps of the broad component are clearly distinct from those of the narrow component, indicating that the broad component does not trace the gas in the galaxy. There are also no apparent correlations between the C2 flux and the strength of the velocity maps.

The relation between W₈₀ and σ indicating dispersion in the line profile is given as W₈₀ = 2.63σ. The W₈₀ maps indicate that the narrow component does not have a wide dispersion, which is expected as the narrow component traces the quiescent gas in the galaxy. In contrast, the broad component clearly has greater values of W₈₀, showing that the component it traces is significantly disturbed. These values range between 300 and 700 km s⁻¹, with a median of 397 km s⁻¹.

In the [O iii] λ5007 flux maps for J0850 indicated in Figure 15, the C2 flux component shows an extended flux making it distinct from the C1 flux component. The velocity maps of C1 show that the systemic gas in the galaxy is likely rotating. However, we see that the C2 component once again does not trace the gas in the galaxy. There does not seem to be a pattern in the C2 velocity maps, but that could just indicate that the broad component is tracing stirred and disturbed gas in the galaxy, which could be caused by stellar processes. The average value of the C2 v₅₀ is roughly 0 km s⁻¹ when computed over the entire galaxy. The W₈₀ maps for C2 seem to trace certain regions of low dispersion, and other regions show higher dispersion, which could possibly indicate turbulence caused by the C2 component.

Zoom In Zoom Out Reset image size

Figure 16 traces the [O iii] λ5007 flux and velocity maps. The C2 flux component seems diminished, indicating that there is not much contribution from the broad component to the flux. There seems to be a faint blob near the top, but it is more prominent in the C1 flux map as compared to C2. It also does not translate over to the v₅₀ maps, as we do not see any corresponding distinct feature near the top of these maps. The C1 component v₅₀ appears quiescent, and the bottom blue component is out of place and is likely caused by noise. The C2 v₅₀ does not have any significant features, although it does appear to be more strongly redshifted as compared to the narrow gas component. The velocity dispersion indicated by the W₈₀ maps is also higher in the outer regions for C2, and there seems to be a region of lower dispersion being traced out as well.

Zoom In Zoom Out Reset image size

From Figure 17, we see that the v₅₀ maps for C2 are clearly distinct from those for C1. They appear to be more blueshifted, indicating that this target likely has some prominent stellar outflows, also indicated by the high values of dispersion from the W₈₀ velocity maps for C2. C1 is once again quiescent with little dispersion.

Zoom In Zoom Out Reset image size

Figure 18 shows the [O iii] λ5007 flux maps and velocity maps for J1302. C1 traces the gas in the galaxy and shows that it is likely to be rotating. The C2 velocity maps do not follow the rotation but they do not have any distinct features. They mostly seem to have low velocity values, although they do show significant dispersion in their W₈₀ values as compared to C1.

Zoom In Zoom Out Reset image size

From Figure 19, it is evident that J1307 has multiple sources (visible in the flux maps). They appear to be fainter in the C2 maps as compared to the C1 maps, suggesting that these sources likely do not contribute to the broad component. The v₅₀ maps of C1 show some rotation in the galaxy, but there does not seem to be a strong correlation between the bright-flux region in the flux map and the velocity maps. There are slightly blueshifted velocities observable in the C2 velocity map in the regions around the bright sources, but they are not significantly high. It is possible that the bright-flux region corresponds to another stellar source producing high ionization activity, but it still does not produce fast outflows. This is supported by the absence of significant dispersion in the W₈₀ velocity maps in the same region.

Zoom In Zoom Out Reset image size

The C2 v₅₀ map (Figure 20) shows that the outflow is either rotating or oriented biconically. In either case, it is clearly distinct from the C1 gas component, which is nearly quiescent. This indicates that the gas in the galaxy is nearly stationary, but there is a distinct outflow present. The C2 W₈₀ velocity map also shows a significant dispersion, unlike the C1 W₈₀ velocity map.

Zoom In Zoom Out Reset image size

The C2 v₅₀ map (Figure 21) shows some regions of blueshift, which could indicate an outflow. It is distinct once again from the C1 v₅₀ map, which is mostly quiescent but has some hints of a rotating gas. The C2 W₈₀ velocity map also shows significant dispersion indicating that C2 is distinct from the C1 quiescent gas.

Zoom In Zoom Out Reset image size

The v₅₀ flux maps of J1622 (Figure 22) show extremely interesting structures going from C1 to C2. The C2 component can constitute an outflow because of the significant blueshift in the velocities but they appear to be oriented perpendicularly to the rotating gas in the galaxy. The significant velocity dispersion in the C2 W₈₀ velocity map is indicative of an outflow. The narrow component does not have a significant dispersion once again.

Zoom In Zoom Out Reset image size

J1717 is at the lowest redshift compared to the rest of the galaxies and thus occupies a larger fraction in the sky. The galaxy appears extended and also has interesting flux maps as seen in Figure 23. The C1 flux map indicates the presence of several sources of SF regions near the center and one slightly near the bottom, which is only faintly seen in the C2 flux map. The galaxy also appears to have a rotating gas component as seen in the C1 v₅₀ map. The C2 component seems to trace this gas slightly based on the rotating pattern observed, but it has lower values of v₅₀. There also does not seem to be a prominent dispersion, although the map is significantly noisy. We could not go to higher S/N as we started losing spatial information. There seems to be a slight visual correlation between the flux maps and the velocity maps near the regions of the additional sources, although it is not too apparent from the W₈₀ velocity maps. This is different from the case of J1307, for which we also have obvious multiple sources but no apparent correlations between the flux and velocity maps (see Figure 19).

Zoom In Zoom Out Reset image size