Environment of Quiescent Low-mass Galaxies Hosting AGNs in MaNGA

The discovery of active galactic nuclei (AGNs) in low-mass (M * ≤ 5 × 109 M ⊙) galaxies has pushed forward the idea that AGN feedback may play a role in quenching star formation in the low-mass regime. In order to test whether AGNs can be a dominant quenching mechanism, we must first disentangle the effects of internal and external processes caused by a galaxy’s environment. We have used the Sloan Digital Sky Survey IV Mapping Nearby Galaxies at Apache Point Observatory survey to produce resolved Baldwin, Phillips, & Terlevich diagrams, and we find 41 AGNs (∼1.3%) in low-mass galaxies. We have studied the group richness (the number of group members) of our AGN and non-AGN samples as a proxy for determining the possible effect of the environment on the gas reservoir in these galaxies. We find that low-mass galaxies hosting AGNs are more likely to be found in isolation or in low-mass groups than galaxies in the non-AGN samples. This preference is even more clear when we split our samples into star-forming and quiescent subsamples. This suggests that environment is not the main cause of quenching in these galaxies, though we cannot rule out the possibility of past mergers.


Introduction
Since the discovery that there is a bimodal distribution of galaxy morphology and color, studies of galaxy evolution have focused on determining the underlying cause of these two populations.Strateva et al. (2001) linked the separation between the blue cloud and red sequence to galaxy morphology by showing that the blue cloud is generally populated by spirals while the red sequence comprises mainly ellipticals.A similar study by Schawinski et al. (2014) showed that a galaxy's star formation rate (SFR) is correlated to its color such that the blue cloud is star-forming and the red sequence is passive.
Combining these discoveries suggested an evolutionary trend in the galaxy color and morphology as the star formation rate decreased.An initial picture was proposed in which starforming galaxies would move from the blue cloud to the red sequence as the gas supply required for star formation was exhausted.A morphological transformation would also happen at some time during this transition.The question then follows: what mechanisms affect a galaxy's gas supply such that it quenches its star formation?
The galaxy could simply run out of gas via secular star formation.As long as the galaxy does not accrete more gas to replenish its reservoir, then it will not form new stars (e.g., Kereš et al. 2005;Kennicutt & Evans 2012).The galaxy's cold interstellar medium could be stripped via a dense intragroup medium or intracluster medium in a process known as ram pressure stripping (e.g., Gunn & Gott 1972;Jaffé et al. 2015).The winds of massive stars or supernovae can heat the local gas or eject it from the galaxy.Therefore, star formation can be quenched on local scales by stellar feedback (e.g., Larson 1974;Dekel & Birnboim 2006).Feedback from active galactic nuclei (AGNs) in the form of radiation, winds, and jets can also provide a mechanism to heat or expel the gas needed for star formation (e.g., Bundy et al. 2008;Cano-Díaz et al. 2012).The galaxy's halo gas could be removed in a process known as starvation or strangulation, which would prevent the galaxy from replenishing its gas supply in the future (e.g., Larson et al. 1980;Peng et al. 2015).
We can split the quenching mechanisms into internal (stellar and AGN feedback) and external processes.These external processes are dependent on the environment of the galaxy (i.e., size of the galaxy group, density of the intragroup medium, distance to the nearest galaxy neighbor, etc.).There is a separate trend with stellar mass in which environment can play a more dominant role in quenching low-mass galaxies than more massive galaxies.Bell et al. (2005) showed that low-mass galaxies are currently in the process of quenching from z ∼ 1 to z ∼ 0 while the majority of high-mass galaxies were already quenched.Therefore, we aim to disentangle the effects of quenching through environmental processes or via internal mechanisms.
Over the past decade, an increasing number of AGNs have been discovered in low-mass galaxies (M * 5 × 10 9 M e ).Serendipitous discoveries of AGNs in low-mass galaxies followed by large sky surveys of nearby galaxies (e.g., Reines et al. 2011Reines et al. , 2013Reines et al. , 2014) ) have allowed for larger samples of these objects.It is only within the last several years that we have had large enough samples of low-mass AGNs to study the properties of the host galaxies in a statistically significant manner.
In the first dedicated survey of AGNs in low-mass galaxies, Reines et al. (2013) used Sloan Digital Sky Survey (SDSS; York et al. 2000;Abazajian et al. 2009;Eisenstein et al. 2011) DR7 spectroscopy to discover 151 AGN candidates out of 25,974 dwarf galaxies.They produced [O III]/Hβ versus [N II]/Hα Baldwin, Phillips, & Terlevich (BPT; Baldwin et al. 1981) diagrams from which they found 136 AGN candidates.They also included an additional 15 galaxies that had broad Hα emission lines.
In a separate survey, Moran et al. (2014) discovered 28 AGNs out of 9526 nearby (D < 80 Mpc) galaxies with stellar masses M * < 10 10 M e .In order to select only galaxies that had clear AGN signatures, they removed galaxies with lowionization nuclear emission regions (LINERs; Heckman 1980) because the ionization source is unclear (Belfiore et al. 2016).They excluded the LINERs from their sample using the [N II], [S II], and [O I] BPT diagrams unless their spectra contained other AGN indicators (e.g., He II λ4686, [Fe VII] λ6087, or [Fe X] λ6375).
Similarly, Sartori et al. (2015) found 336 AGNs out of 48,000 dwarf galaxies (M * < 10 9.5 M e , z < 0.1) using a combination of AGN selection methods: BPT diagram, Shirazi He II diagram (Shirazi & Brinchmann 2012), and Widefield Infrared Survey Explorer (WISE) mid-IR color-color diagram.With only three galaxies out of the 336 AGN candidates being selected by all three selection techniques, there was almost no overlap between the optical and mid-IR selected AGNs.This suggests that optical and mid-IR selection techniques are probing different regions of the AGN (i.e., narrow-line and broad-line regions versus the dusty torus).If the inner regions of the AGN are obscured by the dusty torus, then the mid-IR colors can detect the AGNs that would otherwise be missed by optical selection methods.Any future studies of AGNs in lowmass galaxies should consider multiwavelength selection criteria in order to determine the presence of an AGN.
Based on the results of these dwarf galaxy surveys, the AGN fraction in low-mass galaxies remains around 0.3%-0.7%.This is lower than the approximately constant 10% AGN fraction found by Bongiorno et al. (2016) in more massive galaxies, though the AGN fraction at higher masses (M * ∼ 10 11 M e ) had a redshift dependence, making it vary between 3% and 20%.
With the larger samples of AGNs in low-mass galaxies comes the possibility of understanding how the presence of an AGN can affect the host galaxy and its evolution in the lowmass regime.Environmental mechanisms (e.g., galaxy interactions, density of the intragroup medium, etc.) and stellar feedback in the form of supernovae and stellar winds from massive stars have been shown to quench low-mass galaxies.Can an AGN also influence the galaxy's star formation and ultimately play a dominant role in quenching the galaxy?
Detecting an AGN in a low-mass galaxy can be extremely difficult, though, because we expect these active black holes to have lower mass than their counterparts in high-mass galaxies.Single-fiber spectroscopy may not detect low-luminosity AGNs if there is enough activity from star formation.Single-fiber spectroscopy also has the problem that the fiber itself can cover a large portion of the centers of low-mass galaxies, especially as we look to galaxies at higher redshift.Therefore, integral field unit (IFU) spectroscopy is needed to better detect large amounts of photoionization of high energy-level atoms in the centers of these low-mass galaxies.In some cases, the AGN can have a radio component, which can instead be detected by radio telescopes; see reviews by Ho (2008) and Padovani et al. (2017) and references therein.Previous radio observations of low-mass galaxies with AGNs have been limited to individual galaxies or small samples (e.g., Moran et al. 1999;Greene et al. 2006;Reines et al. 2011).A recent survey by Reines et al. (2020) studied 111 AGNs in low-mass galaxies with the Very Large Array at X-band frequencies (∼8-12 GHz).However, these observations generally require long integration times (unless the source is very bright) so it is difficult to obtain large samples of AGNs in low-mass galaxies using this method.
The Sloan Digital Sky Survey IV (SDSS-IV; Blanton et al. 2017) Mapping Nearby Galaxies at Apache Point Observatory (MaNGA; Bundy et al. 2015) survey has performed IFU spectroscopy for 3251 galaxies with stellar masses 8.5 < log(M * /M e ) < 9.7, which provides us with a new way to view the centers of low-mass galaxies without too much contamination from nearby star formation.
In order to distinguish between the effects of the environment and AGNs in quenching low-mass galaxies, we need to study the environments of galaxies with and without AGNs.Kristensen et al. (2020) used single-fiber spectroscopic data of 62,258 low-mass galaxies in SDSS to determine the presence of an actively accreting massive black hole.Their analysis of the density of the local environment showed that there was no difference between the environments of low-mass galaxies with AGNs and those without AGNs.With MaNGA's capability to study low-mass galaxies in detail, more low-luminosity AGNs or star formation-contaminated AGNs may be found.This makes it important to continue studying the environments of AGN and non-AGN galaxies so that we can better separate the effects of environment and the AGN on the host galaxy.
We define a sample of low-mass MaNGA galaxies with AGNs classified using resolved BPT diagrams of the galaxies' central spaxels.It is still uncertain whether an AGN in a lowmass galaxy can be a dominant quenching mechanism over the environment.Generally, simulations of low-mass galaxies have focused on whether AGN feedback could play a greater role than supernovae in quenching low-mass galaxies as opposed to disentangling the effects of AGN feedback and environmental processes (e.g., Silk 2017; Dashyan et al. 2018;Koudmani et al. 2019Koudmani et al. , 2021Koudmani et al. , 2022)).Geha et al. (2012) showed that the observed quenched fraction of isolated galaxies decreases to negligible values at low stellar masses, suggesting that the quenching of low-mass galaxies and environmental processes are intertwined.Similarly, Peng et al. (2010) found that lowmass galaxies only tend to be quenched if they are living in a dense environment.Dickey et al. (2019) studied quiescent isolated dwarf galaxies and found that 16 out of 20 of their sample had emission line ratios similar those of AGNs.Given that these studies have not addressed the combination of AGN feedback and environment as a quenching mechanism, we will focus on the relationship between the presence of an AGN and the host galaxyʼs star formation rate and environment.In this paper, we define environment to be the group richness or the number of galaxies within a group, and we characterize the environments of the galaxies using two methods to measure the group richness: (1) the Tempel et al. (2017) group catalog, and (2) a simple SDSS position-velocity search for nearby galaxies.
In Section 2, we will discuss our sample selection of AGNs, our definition of quenching, and our environment classification methods.Section 3 presents the results of our environmental analysis of AGNs and our quiescent and starforming subsamples of low-mass galaxies.Finally, in Section 4, we will discuss potential AGN selection effects in low-mass galaxies and a comparison of our environmental analysis with an optically variable AGN sample from Baldassare et al. (2020).

Sample Selection
We use the MaNGA MPL-11 data release (Bundy et al. 2015) to select low-mass galaxies based on the selection criteria from the Penny et al. (2018) study of low-mass galaxies in an earlier MaNGA data release (MPL-5): galaxies with M * 10 9.7 M e , and we also restrict the low-mass galaxies to have a lower bound on stellar mass of M * = 10 8.5 M e in order to better match the distribution of stellar masses in MaNGA.There are 3251 galaxies that meet these criteria.To investigate the role of AGNs in quenching, we separate our low-mass sample into two subsamples: galaxies on the star-forming main sequence and quiescent galaxies.
We have chosen to use star formation rates calculated from the WISE (Wright et al. 2010) mid-infrared 12 μm band in order to account for dust-obscured star formation.We obtain the 12 μm band magnitudes from the AllWISE catalog (Wright et al. 2019;Cutri et al. 2021).In Figure 1, we show the WISE W1 − W2 versus W2 − W3 color-color diagram for the lowmass galaxies in MaNGA.The majority of the low-mass galaxies have fairly blue W1 − W2 colors with a wider range of W2 − W3 colors.Regardless of whether we consider all galaxies with W1 − W2 0.8 as AGNs (Stern et al. 2012) or whether we only classify galaxies within the Jarrett et al. (2011) box as AGNs, the number of low-mass galaxies in MaNGA identified as AGNs remains <15 (0.3% of all low-mass galaxies).
We convert the WISE magnitudes to fluxes and luminosities using the methods in Wright et al. (2010).Cluver et al. (2017) have calibrated the WISE W3 and W4 band luminosities to estimate the star formation rate (see their Equations (4) and ( 6)), and we use the W3 band to calculate the star formation rates of our MaNGA galaxies.Based on the SFR-M * plot of all MaNGA galaxies in Figure 2, we find that the passive sequence is well separated from the star-forming main sequence (SFMS) by a linear regression that is 0.75 dex below the SFMS.Therefore, we consider galaxies to be quiescent if their total star formation rate falls at least 0.75 dex below the SFMS.This selection can be seen in Figure 2 as the blue dashed line.In this way, 416 low-mass galaxies were classified as quiescent while 2828 have some ongoing star formation, and seven galaxies did not have WISE detections.However, the majority of the lowmass galaxies fall below the SFMS, with star formation rates in between the SFMS and the quiescent sequence.

MaNGA Data
The MaNGA survey (Bundy et al. 2015) uses a combination of the 2.5 m optical telescope at Apache Point Observatory (Gunn et al. 2006) and the SDSS-BOSS spectrograph (Dawson et al. 2013;Smee et al. 2013) to perform IFU spectroscopy of ∼10,000 nearby galaxies covering about 2700 deg 2 in the sky.The IFU sizes range from 19 to 127 fibers with each fiber covering 2″ on the sky (Drory et al. 2015).At an average redshift of z ∼ 0.03, this corresponds to a physical resolution of 1-2 kpc.Each galaxy observation had three sets of exposures in order to reach the threshold for signal-to-noise ratio (S/N) (Yan et al. 2016a).There were also IFU fiber bundles set aside to observe standard stars for flux calibration in addition to fibers for science observations (Yan et al. 2016b).
The survey targets galaxies in the NASA Sloan Atlas v1_0_1 (NSA; Blanton et al. 2011), and the targets were chosen so that the observed galaxies have a roughly flat stellar mass distribution at M * > 10 9 M e .The Primary sample of galaxies had fiber coverage out to 1.5 effective radii (R e ) while the Secondary sample (∼33% of observed galaxies) had fiber coverage out to 2.5 R e .The color-enhanced supplement (Wake et al. 2017), accounting for ∼17% of the MaNGA targets, added galaxies to the Primary sample that were located in underdense regions of the NUV − i versus M i color-magnitude plane in order to allow for further observations of green valley galaxies.
The MaNGA Data Reduction Pipeline

AGN Selection
In a recent study of the environment of AGNs in low-mass galaxies, Kristensen et al. (2020) found no correlation between the presence of an AGN and the density of the environment.However, these studies have included the socalled low-ionization nuclear emissions regions (LINERs) in their sample.Heckman (1980) first coined the term LINER as a way to describe this class of objects that has a lower ratio of [O III]/Hβ and higher ratios of [N II]/Hα, [S II]/Hα, and [O I]/Hα than AGNs.In the past, LINERs were thought to be a low-luminosity class of AGNs (e.g., Kauffmann et al. 2003;Kewley et al. 2006;Ho 2008), but Belfiore et al. (2016) showed that LINER emission is not necessarily restricted to galaxy centers.Other theories suggest that LINER emission can be produced by heating from hot, evolved stars (e.g., postasymptotic giant branch stars; Binette et al. 1994;Stasińska et al. 2008;Cid Fernandes et al. 2011) or strong shocks (Heckman et al. 1989).This does not rule out that some galaxies with LINER emission in their centers are associated with low-luminosity AGNs, but the ambiguity is enough to study the environments of AGNs and LINERs separately.Therefore, we will distinguish between the environments of AGNs and LINERs to determine whether environment and AGN activity are really uncorrelated.
We split our sample into subsamples of AGN and LINER galaxies.We distinguish between these different sources of photoionization using BPT diagrams and the demarcation lines developed by Kauffmann et al. (2003) and Kewley et al. (2006).Additionally, because MaNGA is an IFU survey, we can create resolved BPT diagrams to better characterize photoionization at the center of each galaxy, which allows us to potentially detect low-luminosity AGNs in spectra that might otherwise be dominated by star formation.In Figure 3, we show examples of BPT diagrams we created for two galaxies in our sample-one low-mass AGN and one low-mass LINER.We classify galaxies as AGNs or LINERs if their central nine spaxels fall within the AGN or LINER region on both the [N II]/Hα and [S II]/Hα BPT diagrams.We find that 41 low-mass galaxies have emission line ratios suggestive of nuclear activity, and none of the AGNs appear to have broad-line components.Based on our definition of quenching, 11 of our low-mass AGN galaxies are quiescent and 30 have ongoing star formation.We note that several of the BPTselected AGNs and LINERs in our sample have red mid-IR colors that are indicative of a potential AGN as seen in Figure 1.However, these red AGN colors will also contaminate the magnitude of the W3 band so we caution against using the exact values of the W3-measured SFR.For the galaxies where this is a concern, we find that the W3 magnitude is several magnitudes lower than it is for the rest of the low-mass galaxies, which pushes them well above the SFMS.Because we are only interested in a binary definition of quenching (i.e., star-forming or quenched), we place these galaxies in our starforming sample.
Our final sample includes 3251 low-mass galaxies in MaNGA of which 3244 had WISE detections.Of these 3244 low-mass galaxies, 416 are quiescent.There are 41 AGNs and 66 LINERs, with 12 AGNs and 24 LINERs being quenched.Therefore, we have an AGN fraction of ∼1.3% in low-mass galaxies, and ∼29% of the AGNs are quenched.However, the color-enhanced sample from MaNGA was upweighted to include more green valley galaxies.The green valley tends to host a greater fraction of BPT-selected AGNs than the blue cloud or red sequence of galaxies (Schawinski et al. 2007).Therefore, our calculated AGN fractions may be too high for low-mass galaxies in general.

Classification of Environment
We use the Tempel et al. (2017) group catalog (hereafter called the Tempel catalog) to characterize the environment of our low-mass galaxy sample.The Tempel catalog was created using a friends-of-friends group finder algorithm performed on the SDSS DR12 (see Tempel et al. 2014 for a more complete description of this algorithm).Out of 584,449 galaxies, they found 88,662 galaxy groups consisting of at least two members.
However, because the Tempel catalog was based on SDSS DR12 and our sample is based on SDSS DR16, 10% of our low-mass AGN sample and 14% of our low-mass LINER sample do not have group classifications in the Tempel catalog.Therefore, we performed a simple environment search in SDSS DR16 for nearby galaxies to our sample galaxies.We looked for galaxies within a 1 h −1 Mpc radius and with velocity dispersions of 100 km s −1 , 200 km s −1 , and 250 km s −1 .From the galaxies that met these initial criteria, we found those galaxies that lay within 500 kpc of another selected galaxy, and these became our final SDSS environment samples.Based on the results of many studies that have used different groupfinding techniques (such as friends-of-friends or nth nearest neighbor algorithms or counting the number of galaxies within a comoving volume; see Muldrew et al. 2012 for more details), galaxy groups tend to have radii of 0.5-1 h −1 Mpc and velocity dispersions of 250-500 km s −1 (e.g., Carlberg et al. 2001;Eke et al. 2004;Knobel et al. 2012).Generally, group radius and velocity dispersion increase with the number of group members (Brough et al. 2006).Our selection cuts on radius and velocity dispersion are consistent with the expected measurements for small groups (N < 10), and we expect to miss some group members in larger groups with velocity dispersions outside our selection criteria.
Figure 4 shows a comparison of the group richnesses of our 250 km s −1 environment search in SDSS and those of the Tempel catalog search.This comparison is limited to the galaxies that had successful cross-matches with the Tempel catalog.In groups of fewer than about 10 members, our simple environment search in SDSS tends to find more group members than were found by the Tempel catalog (positive N SDSS250 -N Tempel values).This is most likely due to a mismatch in the velocity dispersions of the galaxies with respect to the group.We have restricted our velocity range of the group to a maximum of 500 km s −1 , which is typical for intermediatemass galaxy groups but too small for clusters.Therefore, in isolated galaxies, pairs, or small groups, our velocity range may be too high, so we could be selecting galaxies that are moving too quickly to be bound to a group.Similarly, in larger groups bordering on clusters, we will miss a large fraction of the galaxies that fall outside this velocity range, so we will underestimate the number of group members (negative values of N SDSS250 -N Tempel ).
Another caveat in our SDSS search arises in larger groups where group members may fall outside a 1 Mpc search radius.It is especially important to note that our target low-mass galaxy is unlikely to be the central brightest galaxy in a group, so the center of our radius search could potentially fall on the outskirts of the group.Therefore, a simple 1 Mpc search around the low-mass galaxy may select nearby galaxies that are not actually a part of the group.

Results
As shown in the left panels of Figures 5 and 6, we compare the environments of low-mass galaxies in the Tempel catalog to the group richness of AGNs and LINERs in our sample.We find that ∼54% of the low-mass AGNs and ∼44% of the lowmass LINERs are isolated, while this same statistic drops to ∼39% for all of the inactive (non-AGN and non-LINER) lowmass galaxies.If we consider the groups of five or fewer members, ∼81% of AGNs and ∼84% of LINERs are found in small groups, compared with ∼73% of all low-mass galaxies.An Anderson-Darling two-sample test between the distribution of all inactive low-mass galaxies and the distribution of all AGN galaxies results in a significance level of less than 0.05.Therefore, we can reject the null hypothesis that they come from the same sample, and we can conclude that the environments of AGNs are not the same as the environments of inactive low-mass galaxies.However, performing the same Anderson-Darling two-sample test for the distributions of the inactive low-mass galaxies and the LINERs gives a significance level of ∼0.10 so we cannot reject the null hypothesis.
After performing this same analysis for each velocity range of our SDSS environment search, we see a slightly different trend.Unsurprisingly, as we extend our search to greater velocity ranges, the number of nearby galaxies increases, so the number of isolated galaxies in our samples decreases.In the 250 km s −1 search, we find ∼44% of the AGNs, ∼41% of the LINERs, and ∼36% of all low-mass galaxies are isolated.∼71% of the AGNs and LINERs and ∼66% of all low-mass galaxies are found in low-mass groups (N 5).Performing Anderson-Darling two-sample tests for the 250 km s −1 search between the low-mass galaxies, AGNs, and LINERs results in significance levels of >0.25 for both the AGNs and the LINERs.

Comparison of Quiescent and Star-forming Galaxy Environments
As discussed in Section 2, we split our samples of low-mass galaxies, AGNs, and LINERs into quiescent and star-forming samples using the star formation rates calculated from the WISE 12 μm band.In the center and right panels of Figures 5  and 6, we show the environments of each of these samples in the Tempel catalog.In the Tempel catalog, the profiles of the star-forming galaxies in each bin of the density of the environment are very similar, with most of the star-forming galaxies located in isolated or low-density environments regardless of whether the galaxy is classified as an AGN or LINER.Anderson-Darling two-sample tests of the starforming samples result in significance levels >0.25 for the AGNs and =0.193 for the LINERs in comparison with the lowmass galaxies.Therefore, we cannot reject the null hypothesis that the environments of the star-forming AGNs and LINERs are different than those of the star-forming low-mass galaxies.
In contrast, the density profiles for the quenched AGNs and LINERs show a greater probability that these galaxies are isolated or in low-mass groups (N 5) than quenched lowmass galaxies.∼64% of the AGNs and ∼35% of the LINERs in our quiescent sample are located in isolation while only ∼18% of quiescent low-mass galaxies are found in isolated galaxies.This is reflected in the Anderson-Darling two-sample tests, which gave significance levels <0.001 for both the quenched AGN and LINER samples.This means that we reject the null hypothesis that the environments of the quenched AGNs and LINERs are the same as the group richness of quenched low-mass galaxies in general.
These trends generally hold when we look at the results of our SDSS environment search, though there are fewer isolated galaxies.The significance levels of the quenched AGNs and LINERs are <0.001 and =0.003, respectively, while the Interestingly, there is a small, but significant sample of AGNs residing in quiescent low-mass galaxies that are isolated.This is true for both the Tempel catalog (seven AGNs; ∼35% of Tempel isolated galaxies) and the SDSS 250 km s −1 search (five AGNs; ∼28% of SDSS isolated galaxies).This raises a couple of questions: (1) What quenched these galaxies?and (2) What is feeding the AGNs?These are open questions, which we will only touch upon in this paper.

Discussion
Our results indicate that quiescent low-mass galaxies hosting AGNs are more likely than their non-AGN counterparts to be isolated or live in small groups.We see a similar trend for the quiescent LINERs.

AGN Selection Effects
The use of the BPT diagram to select AGNs has worked well for galaxies in which either there is little star formation to contaminate the emission line ratios or the spectroscopic fibers cover only a small portion of the galaxy's angular size.In the second scenario, the fiber will only cover the central region of the galaxy, thus capturing light from only the nucleus, where the ionizing photons from the active black hole dominate the spectrum.
Out of 3244 low-mass galaxies in the MaNGA survey with WISE detections, we classify 2828 (87%) as star-forming in comparison with ∼71% of intermediate-mass (10 9.7 < M * /M e < 10 10.5 ) and ∼42% of high-mass (M * /M e > 10 10.5 ) galaxies.Thus, low-mass galaxies tend to be star-forming, which makes the detection of AGNs significantly harder.
Another problem arises from the fact that low-mass galaxies typically have lower metallicities, which have been shown by Groves et al. (2006) to push optical emission line ratios to the left on a BPT diagram.This can be enough to move a galaxy from the AGN regime to solidly within the star-forming regime.Therefore, it is likely that the majority of optical surveys will locate AGNs in quiescent, low-mass galaxies and they will miss the population of AGNs found in star-forming or low-metallicity, low-mass galaxies.However, in the case of this project where we want to have a large sample of quiescent AGNs, selecting AGNs via BPT diagrams is not a major concern.

Optical Variability
Instead of using a BPT diagram to classify AGNs, Baldassare et al. (2020) used optical variability as an indicator of the presence of an AGN in low-mass galaxies.They selected a sample of galaxies with stellar masses (M * 2 × 10 10 M e ) from the NASA Sloan Atlas version 0 (NSAv0; Blanton & Roweis 2007;Aihara et al. 2011;Blanton et al. 2011), though they also include any galaxies that have high mass but fell within the field of view of one of their target low-mass galaxies.In addition, they include two samples of broad-line AGNs from Greene & Ho (2007) and Reines & Volonteri (2015).For each galaxy, they constructed light curves using data from the Palomar Transient Factory (Rau et al. 2009;Law et al. 2009), and these light curves had baselines of 3 days to 2156 days.In order to determine whether these galaxies were AGN candidates, they compared their variability with a damped random walk model, which has been shown to mimic AGN variability (e.g., Kelly et al. 2009;MacLeod et al. 2010).This produced 417 AGN candidates out of 47,125 galaxies, and 131 have stellar masses comparable to our low-mass galaxies in MaNGA (10 8.5 M * /M e 10 9.7 ).Baldassare et al. (2020) found that optically variable AGNs in low-mass galaxies tend to inhabit the star-forming regime of the BPT diagram.This suggests that we cannot use solely the BPT diagram for selecting AGNs in low-mass galaxies.Our sample from MaNGA will be skewed toward those galaxies that have either little star formation to contaminate their spectra or more luminous AGNs.Given this possible bias in our MaNGA AGN sample, we perform the same environment analysis (using the Tempel catalog and the simple SDSS search methods) for the optically variable AGNs as described earlier in this paper in order to understand the environments of all AGNs in low-mass galaxies.Out of the 131 optically variable AGN candidates in their sample that have low stellar masses, 12 met our criteria for being quenched.When we cross-match the Baldassare et al. (2020) sample with the Tempel catalog, there are 57 low-mass galaxies with successful matches, seven of which are quenched.
In Figure 7, we compare the environments of the Baldassare et al. (2020) AGN candidates measured by the Tempel catalog to those of the MaNGA AGN candidates.The optically variable AGNs, regardless of star formation rate, tend to be in larger groups than the BPT-selected AGNs.There is no statistical difference between the environments of the quiescent AGN and non-AGN galaxies.However, there is a statistical difference in the environments of the star-forming optically variable AGN candidates and the non-AGN galaxies.The starforming non-AGN galaxies are more likely than the starforming AGN ones to reside in small groups.We only see this trend in the galaxies cross-matched with the Tempel catalog.When we perform our SDSS search for nearby galaxies, the Anderson-Darling significance level for the star-forming samples increases to 0.027, and the significance level for the quenched samples decreases to <0.001, with the quenched AGN galaxies residing in larger groups than quenched non-AGN ones.
Given the disparity between our results for the optically variable AGNs and for those selected via the BPT diagram, we consider several possible explanations.

Scenario 1
If one of the AGN selection methods is not selecting galaxies with accretion onto a massive black hole in a low-mass galaxy, then we will not be comparing AGNs with AGNs.One concern is that diagrams of emission line ratio were created using massive galaxies, so BPT diagrams may not be applicable to low-mass systems, especially in very low-metallicity galaxies (Groves et al. 2006).However, this leads to missing AGNs in low-mass galaxies as opposed to misidentifying AGNs.Another concern is that the origin of optical variability has not been confirmed to be solely the accretion of matter onto a massive black hole, though AGNs are known to be variable at all wavelengths, and searches for quasars using optical variability have been highly successful (e.g., Sesar et al. 2007;Schmidt et al. 2010;MacLeod et al. 2011).Kelly et al. (2009) suggest that the variability could be the result of thermal instabilities in the accretion disk.Similarly, Burke et al. (2021) conclude that the variability is dominated by rapid accretion causing fluctuations in the UV-emitting region of the accretion disk.
Some possibilities for non-AGN mechanisms include phenomena produced by stellar processes common in starbursting galaxies, such as supernovae and luminous blue variables (LBVs, Smith et al. 2011).Baldassare et al. (2020) do take into account possible contamination by these non-AGN mechanisms.They note that LBVs are most likely too faint to be included in their optically variable sample, and they have removed any objects that have burst-like variability that might be indicative of a supernova.

Scenario 2
Scenario 2 is that optical variability is selecting AGNs, but they are of a different breed than the BPT-selected AGNs.
The BPT diagram focuses on how the ionization from the AGNs affects the narrow-line region, so BPT-selected AGNs tend to be of type 2. Optical variability is thought to arise from variations in the accretion disk's thermal emission.Therefore, selection via optical variability is biased toward type-1 AGNs where the line of sight is not obscured by a dusty torus.In this case, there is a possibility that we are seeing an intrinsic difference between the environments of type-1 and type-2 AGNs.However, if the only difference between type-1 and type-2 AGNs is the viewing angle, then we would not expect significant differences in their environments.
Another reason to study the differences between these samples is evident in Figure 1.The Baldassare et al. (2020) optically variable AGNs are well separated from the BPTselected AGNs and LINERs in mid-IR color-color space.The BPT-selected AGNs and LINERs have colors that are similar to normal spiral and elliptical galaxies, while the optically variable AGNs have much redder W2 − W3 colors, and they tend to fall in the region designated as ultraluminous infrared galaxies.Therefore, these samples may indeed be different types of AGNs.

Scenario 3
Scenario 3 is that both the BPT diagram and optical variability are selecting AGNs.We are seeing different primary quenching mechanisms in the Baldassare et al. (2020) sample and the BPT-selected AGNs.The low-mass galaxies for which the AGN is the dominant quenching mechanism will be more likely to be isolated so that there is no environmental quenching.In the case of the BPT-selected AGNs, these galaxies are also more likely to be quenched because our selection method biased our AGN sample toward galaxies that are not dominated by star formation.Because we find that the optically variable AGNs with star formation are more likely than non-AGN galaxies to be found in galaxy groups with more members, this suggests that the environment will play a more dominant role in quenching these galaxies than the presence of an AGN.

Summary
We have produced resolved BPT diagrams of 3251 lowmass galaxies in the MaNGA survey, and we find 41 AGNs and 66 LINERs.We split each sample into quiescent and starforming subsamples using the star formation rate calculated from the WISE 12 μm band.We cross-matched our AGN and LINER samples with the Tempel et al. (2017) group catalog, and we performed a simple environment search in SDSS DR16 for nearby galaxies within velocity ranges of 100, 200, and 250 km s −1 .Out main results are: 1. Quiescent, low-mass galaxies hosting AGNs show a preference for isolation or low-mass groups.Similarly, quiescent, low-mass LINERs are more likely to be found in low-mass groups than the quiescent non-AGN/ LINERs.However, the LINERs are just as likely to be found in isolation as the non-AGN/LINERs.2. The environments of optically variable AGNs from the Baldassare et al. (2020) sample do not show the same results as the BPT-selected AGNs.Based on our analysis using the Tempel catalog, there is no difference between the environments of the quenched optically variable AGN galaxies and the non-AGN ones.However, the optically variable AGNs with star formation are more likely than the star-forming non-AGN galaxies to be found in larger groups.3.In isolated, quiescent, low-mass galaxies, we might be seeing the possibility of AGNs as a dominant quenching mechanism.
(DRP; Law et al. 2016) reduces observations into row-stacked spectra and data cubes.The DRPALL catalog file stores the pointing information for the target galaxy, its photometric observations, and the stellar mass provided by the NSA catalog.The MaNGA Data Analysis Pipeline (DAP; Westfall et al. 2019; Belfiore et al. 2019) further analyzes the data cubes to produce resolved measurements (MAPS) of stellar and gas line-of-sight velocity,

Figure 1 .
Figure 1.WISE W1 − W2 vs. W2 − W3 color-color diagram of the low-mass galaxies in MaNGA (gray points).The BPT-selected AGNs and LINERs in low-mass galaxies are represented as red crosses and blue triangles, respectively.We also include the mid-IR colors of the low-mass galaxies from the Baldassare et al. (2020) sample of optically variable AGNs.Two classifications are used for identifying mid-IR AGNs: W1 − W2 0.8 (solid line) and the Jarrett et al. (2011) demarcation box (dashed lines).

Figure 2 .
Figure 2. Star formation rates calculated using the WISE W3 (12 μm) band as a function of stellar mass for all the galaxies in MaNGA MPL-11 (gray contours).We include the linear regression to the SFMS (red solid line) and 0.75 dex below the SFMS (blue dashed line).We show a scatterplot of the lowmass AGNs in our sample (crosses), low-mass LINERs (circles), and a color bar representing the NUV − r color (NUV is near-ultraviolet).

Figure 3 .
Figure 3. MaNGA footprints for two sample galaxies with their associated resolved BPT diagrams.We include an example of an AGN (top, MaNGA 1-351790) and a LINER (bottom, MaNGA 1-131278) in our sample.Each point is a different spaxel, with larger points representing spaxels that are closer to the center of the galaxy.We show the Kewley et al. (2001) extreme starburst line (solid), the Kauffmann et al. (2003) demarcation line (dashed), and the Kewley et al. (2006) Seyfert/LINER separation line (dotted-dashed).The color bar shows the D n 4000 value for each spaxel.

Figure 4 .
Figure 4. Comparison of the group richness (number of group members) estimated by the SDSS environment search within 250 km s −1 and the Tempel catalog for all the low-mass galaxies in MaNGA MPL-11 that successfully cross-matched with the Tempel catalog.N SDSS250 -N Tempel measures the difference between the group richnesses of the SDSS search and the Tempel catalog such that a value of zero represents when the SDSS search and the Tempel catalog agreed on the group richness.Negative values of N SDSS250 -N Tempel represent groups where the SDSS search did not find as many group members as the Tempel catalog.The color bar shows the number of target galaxies in each 2D bin of N SDSS250 -N Tempel vs. N Tempel .

Figure 5 .
Figure 5. Cumulative distribution functions of the group richness (measured by the number of group members in the Tempel catalog) for low-mass galaxies (blue) and low-mass AGNs (red).Left: environments of all low-mass galaxies and all low-mass AGNs.The significance level of the Anderson-Darling two-sample test is equal to 0.029.Center: environments of quenched low-mass galaxies and quenched low-mass AGNs.The significance level of the Anderson-Darling two-sample test is less than 0.001.Right: environments of star-forming low-mass galaxies and star-forming low-mass AGNs.The significance level of the Anderson-Darling two-sample test is greater than 0.25.

Figure 6 .
Figure 6.Cumulative distribution functions of the group richness (measured by the number of group members in the Tempel catalog) for low-mass galaxies (blue) and low-mass LINERs (red).Left: environments of all low-mass galaxies and all low-mass LINERs.The significance level of the Anderson-Darling two-sample test is equal to 0.092.Center: environments of quenched low-mass galaxies and quenched low-mass LINERs.The significance level of the Anderson-Darling two-sample test is less than 0.001.Right: environments of star-forming low-mass galaxies and star-forming low-mass LINERs.The significance level of the Anderson-Darling two-sample test is equal to 0.193.

Figure 7 .
Figure 7. Cumulative distribution functions of the group richness (measured by the number of group members in the Tempel catalog) for low-mass galaxies (blue) and Baldassare AGN candidates (red).Left: environments of all low-mass galaxies and all low-mass AGNs in the Baldassare et al. (2020) optically variable sample.The significance level of the Anderson-Darling two-sample test is equal to 0.185.Center: environments of quenched low-mass galaxies and quenched low-mass AGNs.The significance level of the Anderson-Darling two-sample test is greater than 0.25.Right: environments of star-forming low-mass galaxies and star-forming lowmass AGNs.The significance level of the Anderson-Darling two-sample test is less than 0.001.